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 PL_expect = XOPERATOR;
1049 yylval.opval = (OP*)newSVOP(OP_CONST, 0, newSVpvn("",0));
1053 if (PL_lex_casemods) { /* oops, we've got some unbalanced parens */
1054 PL_lex_state = LEX_INTERPCASEMOD;
1058 /* Is there a right-hand side to take care of? (s//RHS/ or tr//RHS/) */
1059 if (PL_lex_repl && (PL_lex_inwhat == OP_SUBST || PL_lex_inwhat == OP_TRANS)) {
1060 PL_linestr = PL_lex_repl;
1062 PL_bufend = PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart = SvPVX(PL_linestr);
1063 PL_bufend += SvCUR(PL_linestr);
1064 SAVEFREESV(PL_linestr);
1065 PL_lex_dojoin = FALSE;
1066 PL_lex_brackets = 0;
1067 PL_lex_casemods = 0;
1068 *PL_lex_casestack = '\0';
1070 if (SvEVALED(PL_lex_repl)) {
1071 PL_lex_state = LEX_INTERPNORMAL;
1073 /* we don't clear PL_lex_repl here, so that we can check later
1074 whether this is an evalled subst; that means we rely on the
1075 logic to ensure sublex_done() is called again only via the
1076 branch (in yylex()) that clears PL_lex_repl, else we'll loop */
1079 PL_lex_state = LEX_INTERPCONCAT;
1080 PL_lex_repl = Nullsv;
1086 PL_bufend = SvPVX(PL_linestr);
1087 PL_bufend += SvCUR(PL_linestr);
1088 PL_expect = XOPERATOR;
1089 PL_sublex_info.sub_inwhat = 0;
1097 Extracts a pattern, double-quoted string, or transliteration. This
1100 It looks at lex_inwhat and PL_lex_inpat to find out whether it's
1101 processing a pattern (PL_lex_inpat is true), a transliteration
1102 (lex_inwhat & OP_TRANS is true), or a double-quoted string.
1104 Returns a pointer to the character scanned up to. Iff this is
1105 advanced from the start pointer supplied (ie if anything was
1106 successfully parsed), will leave an OP for the substring scanned
1107 in yylval. Caller must intuit reason for not parsing further
1108 by looking at the next characters herself.
1112 double-quoted style: \r and \n
1113 regexp special ones: \D \s
1115 backrefs: \1 (deprecated in substitution replacements)
1116 case and quoting: \U \Q \E
1117 stops on @ and $, but not for $ as tail anchor
1119 In transliterations:
1120 characters are VERY literal, except for - not at the start or end
1121 of the string, which indicates a range. scan_const expands the
1122 range to the full set of intermediate characters.
1124 In double-quoted strings:
1126 double-quoted style: \r and \n
1128 backrefs: \1 (deprecated)
1129 case and quoting: \U \Q \E
1132 scan_const does *not* construct ops to handle interpolated strings.
1133 It stops processing as soon as it finds an embedded $ or @ variable
1134 and leaves it to the caller to work out what's going on.
1136 @ in pattern could be: @foo, @{foo}, @$foo, @'foo, @:foo.
1138 $ in pattern could be $foo or could be tail anchor. Assumption:
1139 it's a tail anchor if $ is the last thing in the string, or if it's
1140 followed by one of ")| \n\t"
1142 \1 (backreferences) are turned into $1
1144 The structure of the code is
1145 while (there's a character to process) {
1146 handle transliteration ranges
1147 skip regexp comments
1148 skip # initiated comments in //x patterns
1149 check for embedded @foo
1150 check for embedded scalars
1152 leave intact backslashes from leave (below)
1153 deprecate \1 in strings and sub replacements
1154 handle string-changing backslashes \l \U \Q \E, etc.
1155 switch (what was escaped) {
1156 handle - in a transliteration (becomes a literal -)
1157 handle \132 octal characters
1158 handle 0x15 hex characters
1159 handle \cV (control V)
1160 handle printf backslashes (\f, \r, \n, etc)
1162 } (end if backslash)
1163 } (end while character to read)
1168 S_scan_const(pTHX_ char *start)
1170 register char *send = PL_bufend; /* end of the constant */
1171 SV *sv = NEWSV(93, send - start); /* sv for the constant */
1172 register char *s = start; /* start of the constant */
1173 register char *d = SvPVX(sv); /* destination for copies */
1174 bool dorange = FALSE; /* are we in a translit range? */
1175 bool didrange = FALSE; /* did we just finish a range? */
1176 bool has_utf8 = FALSE; /* embedded \x{} */
1179 I32 utf = (PL_lex_inwhat == OP_TRANS && PL_sublex_info.sub_op)
1180 ? (PL_sublex_info.sub_op->op_private & (OPpTRANS_FROM_UTF|OPpTRANS_TO_UTF))
1182 I32 this_utf8 = (PL_lex_inwhat == OP_TRANS && PL_sublex_info.sub_op)
1183 ? (PL_sublex_info.sub_op->op_private & (PL_lex_repl ?
1184 OPpTRANS_FROM_UTF : OPpTRANS_TO_UTF))
1186 const char *leaveit = /* set of acceptably-backslashed characters */
1188 ? "\\.^$@AGZdDwWsSbBpPXC+*?|()-nrtfeaxcz0123456789[{]} \t\n\r\f\v#"
1191 while (s < send || dorange) {
1192 /* get transliterations out of the way (they're most literal) */
1193 if (PL_lex_inwhat == OP_TRANS) {
1194 /* expand a range A-Z to the full set of characters. AIE! */
1196 I32 i; /* current expanded character */
1197 I32 min; /* first character in range */
1198 I32 max; /* last character in range */
1200 i = d - SvPVX(sv); /* remember current offset */
1201 SvGROW(sv, SvLEN(sv) + 256); /* never more than 256 chars in a range */
1202 d = SvPVX(sv) + i; /* refresh d after realloc */
1203 d -= 2; /* eat the first char and the - */
1205 min = (U8)*d; /* first char in range */
1206 max = (U8)d[1]; /* last char in range */
1210 "Invalid [] range \"%c-%c\" in transliteration operator",
1211 (char)min, (char)max);
1215 if ((isLOWER(min) && isLOWER(max)) ||
1216 (isUPPER(min) && isUPPER(max))) {
1218 for (i = min; i <= max; i++)
1222 for (i = min; i <= max; i++)
1229 for (i = min; i <= max; i++)
1232 /* mark the range as done, and continue */
1238 /* range begins (ignore - as first or last char) */
1239 else if (*s == '-' && s+1 < send && s != start) {
1241 Perl_croak(aTHX_ "Ambiguous range in transliteration operator");
1244 *d++ = (char)0xff; /* use illegal utf8 byte--see pmtrans */
1256 /* if we get here, we're not doing a transliteration */
1258 /* skip for regexp comments /(?#comment)/ and code /(?{code})/,
1259 except for the last char, which will be done separately. */
1260 else if (*s == '(' && PL_lex_inpat && s[1] == '?') {
1262 while (s < send && *s != ')')
1265 else if (s[2] == '{' /* This should match regcomp.c */
1266 || ((s[2] == 'p' || s[2] == '?') && s[3] == '{'))
1269 char *regparse = s + (s[2] == '{' ? 3 : 4);
1272 while (count && (c = *regparse)) {
1273 if (c == '\\' && regparse[1])
1281 if (*regparse != ')') {
1282 regparse--; /* Leave one char for continuation. */
1283 yyerror("Sequence (?{...}) not terminated or not {}-balanced");
1285 while (s < regparse)
1290 /* likewise skip #-initiated comments in //x patterns */
1291 else if (*s == '#' && PL_lex_inpat &&
1292 ((PMOP*)PL_lex_inpat)->op_pmflags & PMf_EXTENDED) {
1293 while (s+1 < send && *s != '\n')
1297 /* check for embedded arrays
1298 (@foo, @:foo, @'foo, @{foo}, @$foo, @+, @-)
1300 else if (*s == '@' && s[1]
1301 && (isALNUM_lazy_if(s+1,UTF) || strchr(":'{$+-", s[1])))
1304 /* check for embedded scalars. only stop if we're sure it's a
1307 else if (*s == '$') {
1308 if (!PL_lex_inpat) /* not a regexp, so $ must be var */
1310 if (s + 1 < send && !strchr("()| \n\t", s[1]))
1311 break; /* in regexp, $ might be tail anchor */
1315 if (*s == '\\' && s+1 < send) {
1316 bool to_be_utf8 = FALSE;
1320 /* some backslashes we leave behind */
1321 if (*leaveit && *s && strchr(leaveit, *s)) {
1327 /* deprecate \1 in strings and substitution replacements */
1328 if (PL_lex_inwhat == OP_SUBST && !PL_lex_inpat &&
1329 isDIGIT(*s) && *s != '0' && !isDIGIT(s[1]))
1331 if (ckWARN(WARN_SYNTAX))
1332 Perl_warner(aTHX_ WARN_SYNTAX, "\\%c better written as $%c", *s, *s);
1337 /* string-change backslash escapes */
1338 if (PL_lex_inwhat != OP_TRANS && *s && strchr("lLuUEQ", *s)) {
1343 /* if we get here, it's either a quoted -, or a digit */
1346 /* quoted - in transliterations */
1348 if (PL_lex_inwhat == OP_TRANS) {
1355 if (ckWARN(WARN_MISC) && isALNUM(*s))
1356 Perl_warner(aTHX_ WARN_MISC,
1357 "Unrecognized escape \\%c passed through",
1359 /* default action is to copy the quoted character */
1364 /* \132 indicates an octal constant */
1365 case '0': case '1': case '2': case '3':
1366 case '4': case '5': case '6': case '7':
1368 STRLEN len = 0; /* disallow underscores */
1369 uv = (UV)scan_oct(s, 3, &len);
1372 goto NUM_ESCAPE_INSERT;
1374 /* \x24 indicates a hex constant */
1378 char* e = strchr(s, '}');
1380 yyerror("Missing right brace on \\x{}");
1384 STRLEN len = 1; /* allow underscores */
1385 uv = (UV)scan_hex(s + 1, e - s - 1, &len);
1386 if (PL_hints & HINT_UTF8)
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 && (to_be_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 (to_be_utf8 || 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", 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", ftst);
2880 if (PL_expect == XOPERATOR)
2885 else if (*s == '>') {
2888 if (isIDFIRST_lazy_if(s,UTF)) {
2889 s = force_word(s,METHOD,FALSE,TRUE,FALSE);
2897 if (PL_expect == XOPERATOR)
2900 if (isSPACE(*s) || !isSPACE(*PL_bufptr))
2902 OPERATOR('-'); /* unary minus */
2909 if (PL_expect == XOPERATOR)
2914 if (PL_expect == XOPERATOR)
2917 if (isSPACE(*s) || !isSPACE(*PL_bufptr))
2923 if (PL_expect != XOPERATOR) {
2924 s = scan_ident(s, PL_bufend, PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
2925 PL_expect = XOPERATOR;
2926 force_ident(PL_tokenbuf, '*');
2939 if (PL_expect == XOPERATOR) {
2943 PL_tokenbuf[0] = '%';
2944 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1, TRUE);
2945 if (!PL_tokenbuf[1]) {
2947 yyerror("Final % should be \\% or %name");
2950 PL_pending_ident = '%';
2969 switch (PL_expect) {
2972 if (!PL_in_my || PL_lex_state != LEX_NORMAL)
2974 PL_bufptr = s; /* update in case we back off */
2980 PL_expect = XTERMBLOCK;
2984 while (isIDFIRST_lazy_if(s,UTF)) {
2985 d = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
2986 if (isLOWER(*s) && (tmp = keyword(PL_tokenbuf, len))) {
2987 if (tmp < 0) tmp = -tmp;
3002 d = scan_str(d,TRUE,TRUE);
3005 SvREFCNT_dec(PL_lex_stuff);
3006 PL_lex_stuff = Nullsv;
3008 /* MUST advance bufptr here to avoid bogus
3009 "at end of line" context messages from yyerror().
3011 PL_bufptr = s + len;
3012 yyerror("Unterminated attribute parameter in attribute list");
3015 return 0; /* EOF indicator */
3019 SV *sv = newSVpvn(s, len);
3020 sv_catsv(sv, PL_lex_stuff);
3021 attrs = append_elem(OP_LIST, attrs,
3022 newSVOP(OP_CONST, 0, sv));
3023 SvREFCNT_dec(PL_lex_stuff);
3024 PL_lex_stuff = Nullsv;
3027 if (!PL_in_my && len == 6 && strnEQ(s, "lvalue", len))
3028 CvLVALUE_on(PL_compcv);
3029 else if (!PL_in_my && len == 6 && strnEQ(s, "locked", len))
3030 CvLOCKED_on(PL_compcv);
3031 else if (!PL_in_my && len == 6 && strnEQ(s, "method", len))
3032 CvMETHOD_on(PL_compcv);
3033 /* After we've set the flags, it could be argued that
3034 we don't need to do the attributes.pm-based setting
3035 process, and shouldn't bother appending recognized
3036 flags. To experiment with that, uncomment the
3037 following "else": */
3039 attrs = append_elem(OP_LIST, attrs,
3040 newSVOP(OP_CONST, 0,
3044 if (*s == ':' && s[1] != ':')
3047 break; /* require real whitespace or :'s */
3049 tmp = (PL_expect == XOPERATOR ? '=' : '{'); /*'}(' for vi */
3050 if (*s != ';' && *s != tmp && (tmp != '=' || *s != ')')) {
3051 char q = ((*s == '\'') ? '"' : '\'');
3052 /* If here for an expression, and parsed no attrs, back off. */
3053 if (tmp == '=' && !attrs) {
3057 /* MUST advance bufptr here to avoid bogus "at end of line"
3058 context messages from yyerror().
3062 yyerror("Unterminated attribute list");
3064 yyerror(Perl_form(aTHX_ "Invalid separator character %c%c%c in attribute list",
3072 PL_nextval[PL_nexttoke].opval = attrs;
3080 if (PL_last_lop == PL_oldoldbufptr || PL_last_uni == PL_oldoldbufptr)
3081 PL_oldbufptr = PL_oldoldbufptr; /* allow print(STDOUT 123) */
3097 if (PL_lex_brackets <= 0)
3098 yyerror("Unmatched right square bracket");
3101 if (PL_lex_state == LEX_INTERPNORMAL) {
3102 if (PL_lex_brackets == 0) {
3103 if (*s != '[' && *s != '{' && (*s != '-' || s[1] != '>'))
3104 PL_lex_state = LEX_INTERPEND;
3111 if (PL_lex_brackets > 100) {
3112 char* newlb = Renew(PL_lex_brackstack, PL_lex_brackets + 1, char);
3113 if (newlb != PL_lex_brackstack) {
3115 PL_lex_brackstack = newlb;
3118 switch (PL_expect) {
3120 if (PL_lex_formbrack) {
3124 if (PL_oldoldbufptr == PL_last_lop)
3125 PL_lex_brackstack[PL_lex_brackets++] = XTERM;
3127 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
3128 OPERATOR(HASHBRACK);
3130 while (s < PL_bufend && SPACE_OR_TAB(*s))
3133 PL_tokenbuf[0] = '\0';
3134 if (d < PL_bufend && *d == '-') {
3135 PL_tokenbuf[0] = '-';
3137 while (d < PL_bufend && SPACE_OR_TAB(*d))
3140 if (d < PL_bufend && isIDFIRST_lazy_if(d,UTF)) {
3141 d = scan_word(d, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1,
3143 while (d < PL_bufend && SPACE_OR_TAB(*d))
3146 char minus = (PL_tokenbuf[0] == '-');
3147 s = force_word(s + minus, WORD, FALSE, TRUE, FALSE);
3148 if (UTF && !IN_BYTE && is_utf8_string((U8*)PL_tokenbuf, 0) &&
3149 PL_nextval[PL_nexttoke-1].opval)
3150 SvUTF8_on(((SVOP*)PL_nextval[PL_nexttoke-1].opval)->op_sv);
3158 PL_lex_brackstack[PL_lex_brackets++] = XSTATE;
3163 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
3168 if (PL_oldoldbufptr == PL_last_lop)
3169 PL_lex_brackstack[PL_lex_brackets++] = XTERM;
3171 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
3174 OPERATOR(HASHBRACK);
3175 /* This hack serves to disambiguate a pair of curlies
3176 * as being a block or an anon hash. Normally, expectation
3177 * determines that, but in cases where we're not in a
3178 * position to expect anything in particular (like inside
3179 * eval"") we have to resolve the ambiguity. This code
3180 * covers the case where the first term in the curlies is a
3181 * quoted string. Most other cases need to be explicitly
3182 * disambiguated by prepending a `+' before the opening
3183 * curly in order to force resolution as an anon hash.
3185 * XXX should probably propagate the outer expectation
3186 * into eval"" to rely less on this hack, but that could
3187 * potentially break current behavior of eval"".
3191 if (*s == '\'' || *s == '"' || *s == '`') {
3192 /* common case: get past first string, handling escapes */
3193 for (t++; t < PL_bufend && *t != *s;)
3194 if (*t++ == '\\' && (*t == '\\' || *t == *s))
3198 else if (*s == 'q') {
3201 || ((*t == 'q' || *t == 'x') && ++t < PL_bufend
3205 char open, close, term;
3208 while (t < PL_bufend && isSPACE(*t))
3212 if (term && (tmps = strchr("([{< )]}> )]}>",term)))
3216 for (t++; t < PL_bufend; t++) {
3217 if (*t == '\\' && t+1 < PL_bufend && open != '\\')
3219 else if (*t == open)
3223 for (t++; t < PL_bufend; t++) {
3224 if (*t == '\\' && t+1 < PL_bufend)
3226 else if (*t == close && --brackets <= 0)
3228 else if (*t == open)
3234 else if (isALNUM_lazy_if(t,UTF)) {
3236 while (t < PL_bufend && isALNUM_lazy_if(t,UTF))
3239 while (t < PL_bufend && isSPACE(*t))
3241 /* if comma follows first term, call it an anon hash */
3242 /* XXX it could be a comma expression with loop modifiers */
3243 if (t < PL_bufend && ((*t == ',' && (*s == 'q' || !isLOWER(*s)))
3244 || (*t == '=' && t[1] == '>')))
3245 OPERATOR(HASHBRACK);
3246 if (PL_expect == XREF)
3249 PL_lex_brackstack[PL_lex_brackets-1] = XSTATE;
3255 yylval.ival = CopLINE(PL_curcop);
3256 if (isSPACE(*s) || *s == '#')
3257 PL_copline = NOLINE; /* invalidate current command line number */
3262 if (PL_lex_brackets <= 0)
3263 yyerror("Unmatched right curly bracket");
3265 PL_expect = (expectation)PL_lex_brackstack[--PL_lex_brackets];
3266 if (PL_lex_brackets < PL_lex_formbrack && PL_lex_state != LEX_INTERPNORMAL)
3267 PL_lex_formbrack = 0;
3268 if (PL_lex_state == LEX_INTERPNORMAL) {
3269 if (PL_lex_brackets == 0) {
3270 if (PL_expect & XFAKEBRACK) {
3271 PL_expect &= XENUMMASK;
3272 PL_lex_state = LEX_INTERPEND;
3274 return yylex(); /* ignore fake brackets */
3276 if (*s == '-' && s[1] == '>')
3277 PL_lex_state = LEX_INTERPENDMAYBE;
3278 else if (*s != '[' && *s != '{')
3279 PL_lex_state = LEX_INTERPEND;
3282 if (PL_expect & XFAKEBRACK) {
3283 PL_expect &= XENUMMASK;
3285 return yylex(); /* ignore fake brackets */
3295 if (PL_expect == XOPERATOR) {
3296 if (ckWARN(WARN_SEMICOLON)
3297 && isIDFIRST_lazy_if(s,UTF) && PL_bufptr == PL_linestart)
3299 CopLINE_dec(PL_curcop);
3300 Perl_warner(aTHX_ WARN_SEMICOLON, PL_warn_nosemi);
3301 CopLINE_inc(PL_curcop);
3306 s = scan_ident(s - 1, PL_bufend, PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
3308 PL_expect = XOPERATOR;
3309 force_ident(PL_tokenbuf, '&');
3313 yylval.ival = (OPpENTERSUB_AMPER<<8);
3332 if (ckWARN(WARN_SYNTAX) && tmp && isSPACE(*s) && strchr("+-*/%.^&|<",tmp))
3333 Perl_warner(aTHX_ WARN_SYNTAX, "Reversed %c= operator",(int)tmp);
3335 if (PL_expect == XSTATE && isALPHA(tmp) &&
3336 (s == PL_linestart+1 || s[-2] == '\n') )
3338 if (PL_in_eval && !PL_rsfp) {
3343 if (strnEQ(s,"=cut",4)) {
3357 PL_doextract = TRUE;
3360 if (PL_lex_brackets < PL_lex_formbrack) {
3362 #ifdef PERL_STRICT_CR
3363 for (t = s; SPACE_OR_TAB(*t); t++) ;
3365 for (t = s; SPACE_OR_TAB(*t) || *t == '\r'; t++) ;
3367 if (*t == '\n' || *t == '#') {
3385 if (PL_expect != XOPERATOR) {
3386 if (s[1] != '<' && !strchr(s,'>'))
3389 s = scan_heredoc(s);
3391 s = scan_inputsymbol(s);
3392 TERM(sublex_start());
3397 SHop(OP_LEFT_SHIFT);
3411 SHop(OP_RIGHT_SHIFT);
3420 if (PL_expect == XOPERATOR) {
3421 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
3424 return ','; /* grandfather non-comma-format format */
3428 if (s[1] == '#' && (isIDFIRST_lazy_if(s+2,UTF) || strchr("{$:+-", s[2]))) {
3429 PL_tokenbuf[0] = '@';
3430 s = scan_ident(s + 1, PL_bufend, PL_tokenbuf + 1,
3431 sizeof PL_tokenbuf - 1, FALSE);
3432 if (PL_expect == XOPERATOR)
3433 no_op("Array length", s);
3434 if (!PL_tokenbuf[1])
3436 PL_expect = XOPERATOR;
3437 PL_pending_ident = '#';
3441 PL_tokenbuf[0] = '$';
3442 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1,
3443 sizeof PL_tokenbuf - 1, FALSE);
3444 if (PL_expect == XOPERATOR)
3446 if (!PL_tokenbuf[1]) {
3448 yyerror("Final $ should be \\$ or $name");
3452 /* This kludge not intended to be bulletproof. */
3453 if (PL_tokenbuf[1] == '[' && !PL_tokenbuf[2]) {
3454 yylval.opval = newSVOP(OP_CONST, 0,
3455 newSViv(PL_compiling.cop_arybase));
3456 yylval.opval->op_private = OPpCONST_ARYBASE;
3462 if (PL_lex_state == LEX_NORMAL)
3465 if ((PL_expect != XREF || PL_oldoldbufptr == PL_last_lop) && intuit_more(s)) {
3468 PL_tokenbuf[0] = '@';
3469 if (ckWARN(WARN_SYNTAX)) {
3471 isSPACE(*t) || isALNUM_lazy_if(t,UTF) || *t == '$';
3474 PL_bufptr = skipspace(PL_bufptr);
3475 while (t < PL_bufend && *t != ']')
3477 Perl_warner(aTHX_ WARN_SYNTAX,
3478 "Multidimensional syntax %.*s not supported",
3479 (t - PL_bufptr) + 1, PL_bufptr);
3483 else if (*s == '{') {
3484 PL_tokenbuf[0] = '%';
3485 if (ckWARN(WARN_SYNTAX) && strEQ(PL_tokenbuf+1, "SIG") &&
3486 (t = strchr(s, '}')) && (t = strchr(t, '=')))
3488 char tmpbuf[sizeof PL_tokenbuf];
3490 for (t++; isSPACE(*t); t++) ;
3491 if (isIDFIRST_lazy_if(t,UTF)) {
3492 t = scan_word(t, tmpbuf, sizeof tmpbuf, TRUE, &len);
3493 for (; isSPACE(*t); t++) ;
3494 if (*t == ';' && get_cv(tmpbuf, FALSE))
3495 Perl_warner(aTHX_ WARN_SYNTAX,
3496 "You need to quote \"%s\"", tmpbuf);
3502 PL_expect = XOPERATOR;
3503 if (PL_lex_state == LEX_NORMAL && isSPACE((char)tmp)) {
3504 bool islop = (PL_last_lop == PL_oldoldbufptr);
3505 if (!islop || PL_last_lop_op == OP_GREPSTART)
3506 PL_expect = XOPERATOR;
3507 else if (strchr("$@\"'`q", *s))
3508 PL_expect = XTERM; /* e.g. print $fh "foo" */
3509 else if (strchr("&*<%", *s) && isIDFIRST_lazy_if(s+1,UTF))
3510 PL_expect = XTERM; /* e.g. print $fh &sub */
3511 else if (isIDFIRST_lazy_if(s,UTF)) {
3512 char tmpbuf[sizeof PL_tokenbuf];
3513 scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
3514 if ((tmp = keyword(tmpbuf, len))) {
3515 /* binary operators exclude handle interpretations */
3527 PL_expect = XTERM; /* e.g. print $fh length() */
3532 GV *gv = gv_fetchpv(tmpbuf, FALSE, SVt_PVCV);
3533 if (gv && GvCVu(gv))
3534 PL_expect = XTERM; /* e.g. print $fh subr() */
3537 else if (isDIGIT(*s))
3538 PL_expect = XTERM; /* e.g. print $fh 3 */
3539 else if (*s == '.' && isDIGIT(s[1]))
3540 PL_expect = XTERM; /* e.g. print $fh .3 */
3541 else if (strchr("/?-+", *s) && !isSPACE(s[1]) && s[1] != '=')
3542 PL_expect = XTERM; /* e.g. print $fh -1 */
3543 else if (*s == '<' && s[1] == '<' && !isSPACE(s[2]) && s[2] != '=')
3544 PL_expect = XTERM; /* print $fh <<"EOF" */
3546 PL_pending_ident = '$';
3550 if (PL_expect == XOPERATOR)
3552 PL_tokenbuf[0] = '@';
3553 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1, FALSE);
3554 if (!PL_tokenbuf[1]) {
3556 yyerror("Final @ should be \\@ or @name");
3559 if (PL_lex_state == LEX_NORMAL)
3561 if ((PL_expect != XREF || PL_oldoldbufptr == PL_last_lop) && intuit_more(s)) {
3563 PL_tokenbuf[0] = '%';
3565 /* Warn about @ where they meant $. */
3566 if (ckWARN(WARN_SYNTAX)) {
3567 if (*s == '[' || *s == '{') {
3569 while (*t && (isALNUM_lazy_if(t,UTF) || strchr(" \t$#+-'\"", *t)))
3571 if (*t == '}' || *t == ']') {
3573 PL_bufptr = skipspace(PL_bufptr);
3574 Perl_warner(aTHX_ WARN_SYNTAX,
3575 "Scalar value %.*s better written as $%.*s",
3576 t-PL_bufptr, PL_bufptr, t-PL_bufptr-1, PL_bufptr+1);
3581 PL_pending_ident = '@';
3584 case '/': /* may either be division or pattern */
3585 case '?': /* may either be conditional or pattern */
3586 if (PL_expect != XOPERATOR) {
3587 /* Disable warning on "study /blah/" */
3588 if (PL_oldoldbufptr == PL_last_uni
3589 && (*PL_last_uni != 's' || s - PL_last_uni < 5
3590 || memNE(PL_last_uni, "study", 5)
3591 || isALNUM_lazy_if(PL_last_uni+5,UTF)))
3593 s = scan_pat(s,OP_MATCH);
3594 TERM(sublex_start());
3602 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack
3603 #ifdef PERL_STRICT_CR
3606 && (s[1] == '\n' || (s[1] == '\r' && s[2] == '\n'))
3608 && (s == PL_linestart || s[-1] == '\n') )
3610 PL_lex_formbrack = 0;
3614 if (PL_expect == XOPERATOR || !isDIGIT(s[1])) {
3620 yylval.ival = OPf_SPECIAL;
3626 if (PL_expect != XOPERATOR)
3631 case '0': case '1': case '2': case '3': case '4':
3632 case '5': case '6': case '7': case '8': case '9':
3633 s = scan_num(s, &yylval);
3634 DEBUG_T( { PerlIO_printf(Perl_debug_log,
3635 "### Saw number in '%s'\n", s);
3637 if (PL_expect == XOPERATOR)
3642 s = scan_str(s,FALSE,FALSE);
3643 DEBUG_T( { PerlIO_printf(Perl_debug_log,
3644 "### Saw string in '%s'\n", s);
3646 if (PL_expect == XOPERATOR) {
3647 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
3650 return ','; /* grandfather non-comma-format format */
3656 missingterm((char*)0);
3657 yylval.ival = OP_CONST;
3658 TERM(sublex_start());
3661 s = scan_str(s,FALSE,FALSE);
3662 DEBUG_T( { PerlIO_printf(Perl_debug_log,
3663 "### Saw string in '%s'\n", s);
3665 if (PL_expect == XOPERATOR) {
3666 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
3669 return ','; /* grandfather non-comma-format format */
3675 missingterm((char*)0);
3676 yylval.ival = OP_CONST;
3677 for (d = SvPV(PL_lex_stuff, len); len; len--, d++) {
3678 if (*d == '$' || *d == '@' || *d == '\\' || UTF8_IS_CONTINUED(*d)) {
3679 yylval.ival = OP_STRINGIFY;
3683 TERM(sublex_start());
3686 s = scan_str(s,FALSE,FALSE);
3687 DEBUG_T( { PerlIO_printf(Perl_debug_log,
3688 "### Saw backtick string in '%s'\n", s);
3690 if (PL_expect == XOPERATOR)
3691 no_op("Backticks",s);
3693 missingterm((char*)0);
3694 yylval.ival = OP_BACKTICK;
3696 TERM(sublex_start());
3700 if (ckWARN(WARN_SYNTAX) && PL_lex_inwhat && isDIGIT(*s))
3701 Perl_warner(aTHX_ WARN_SYNTAX,"Can't use \\%c to mean $%c in expression",
3703 if (PL_expect == XOPERATOR)
3704 no_op("Backslash",s);
3708 if (isDIGIT(s[1]) && PL_expect != XOPERATOR) {
3712 while (isDIGIT(*start) || *start == '_')
3714 if (*start == '.' && isDIGIT(start[1])) {
3715 s = scan_num(s, &yylval);
3718 /* avoid v123abc() or $h{v1}, allow C<print v10;> */
3719 else if (!isALPHA(*start) && (PL_expect == XTERM || PL_expect == XREF)) {
3723 gv = gv_fetchpv(s, FALSE, SVt_PVCV);
3726 s = scan_num(s, &yylval);
3733 if (isDIGIT(s[1]) && PL_expect == XOPERATOR) {
3772 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
3774 /* Some keywords can be followed by any delimiter, including ':' */
3775 tmp = ((len == 1 && strchr("msyq", PL_tokenbuf[0])) ||
3776 (len == 2 && ((PL_tokenbuf[0] == 't' && PL_tokenbuf[1] == 'r') ||
3777 (PL_tokenbuf[0] == 'q' &&
3778 strchr("qwxr", PL_tokenbuf[1])))));
3780 /* x::* is just a word, unless x is "CORE" */
3781 if (!tmp && *s == ':' && s[1] == ':' && strNE(PL_tokenbuf, "CORE"))
3785 while (d < PL_bufend && isSPACE(*d))
3786 d++; /* no comments skipped here, or s### is misparsed */
3788 /* Is this a label? */
3789 if (!tmp && PL_expect == XSTATE
3790 && d < PL_bufend && *d == ':' && *(d + 1) != ':') {
3792 yylval.pval = savepv(PL_tokenbuf);
3797 /* Check for keywords */
3798 tmp = keyword(PL_tokenbuf, len);
3800 /* Is this a word before a => operator? */
3801 if (*d == '=' && d[1] == '>') {
3803 yylval.opval = (OP*)newSVOP(OP_CONST, 0, newSVpv(PL_tokenbuf,0));
3804 yylval.opval->op_private = OPpCONST_BARE;
3805 if (UTF && !IN_BYTE && is_utf8_string((U8*)PL_tokenbuf, len))
3806 SvUTF8_on(((SVOP*)yylval.opval)->op_sv);
3810 if (tmp < 0) { /* second-class keyword? */
3811 GV *ogv = Nullgv; /* override (winner) */
3812 GV *hgv = Nullgv; /* hidden (loser) */
3813 if (PL_expect != XOPERATOR && (*s != ':' || s[1] != ':')) {
3815 if ((gv = gv_fetchpv(PL_tokenbuf, FALSE, SVt_PVCV)) &&
3818 if (GvIMPORTED_CV(gv))
3820 else if (! CvMETHOD(cv))
3824 (gvp = (GV**)hv_fetch(PL_globalstash,PL_tokenbuf,len,FALSE)) &&
3825 (gv = *gvp) != (GV*)&PL_sv_undef &&
3826 GvCVu(gv) && GvIMPORTED_CV(gv))
3832 tmp = 0; /* overridden by import or by GLOBAL */
3835 && -tmp==KEY_lock /* XXX generalizable kludge */
3837 && !hv_fetch(GvHVn(PL_incgv), "Thread.pm", 9, FALSE))
3839 tmp = 0; /* any sub overrides "weak" keyword */
3841 else { /* no override */
3845 if (ckWARN(WARN_AMBIGUOUS) && hgv
3846 && tmp != KEY_x && tmp != KEY_CORE) /* never ambiguous */
3847 Perl_warner(aTHX_ WARN_AMBIGUOUS,
3848 "Ambiguous call resolved as CORE::%s(), %s",
3849 GvENAME(hgv), "qualify as such or use &");
3856 default: /* not a keyword */
3859 char lastchar = (PL_bufptr == PL_oldoldbufptr ? 0 : PL_bufptr[-1]);
3861 /* Get the rest if it looks like a package qualifier */
3863 if (*s == '\'' || (*s == ':' && s[1] == ':')) {
3865 s = scan_word(s, PL_tokenbuf + len, sizeof PL_tokenbuf - len,
3868 Perl_croak(aTHX_ "Bad name after %s%s", PL_tokenbuf,
3869 *s == '\'' ? "'" : "::");
3873 if (PL_expect == XOPERATOR) {
3874 if (PL_bufptr == PL_linestart) {
3875 CopLINE_dec(PL_curcop);
3876 Perl_warner(aTHX_ WARN_SEMICOLON, PL_warn_nosemi);
3877 CopLINE_inc(PL_curcop);
3880 no_op("Bareword",s);
3883 /* Look for a subroutine with this name in current package,
3884 unless name is "Foo::", in which case Foo is a bearword
3885 (and a package name). */
3888 PL_tokenbuf[len - 2] == ':' && PL_tokenbuf[len - 1] == ':')
3890 if (ckWARN(WARN_BAREWORD) && ! gv_fetchpv(PL_tokenbuf, FALSE, SVt_PVHV))
3891 Perl_warner(aTHX_ WARN_BAREWORD,
3892 "Bareword \"%s\" refers to nonexistent package",
3895 PL_tokenbuf[len] = '\0';
3902 gv = gv_fetchpv(PL_tokenbuf, FALSE, SVt_PVCV);
3905 /* if we saw a global override before, get the right name */
3908 sv = newSVpvn("CORE::GLOBAL::",14);
3909 sv_catpv(sv,PL_tokenbuf);
3912 sv = newSVpv(PL_tokenbuf,0);
3914 /* Presume this is going to be a bareword of some sort. */
3917 yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
3918 yylval.opval->op_private = OPpCONST_BARE;
3920 /* And if "Foo::", then that's what it certainly is. */
3925 /* See if it's the indirect object for a list operator. */
3927 if (PL_oldoldbufptr &&
3928 PL_oldoldbufptr < PL_bufptr &&
3929 (PL_oldoldbufptr == PL_last_lop
3930 || PL_oldoldbufptr == PL_last_uni) &&
3931 /* NO SKIPSPACE BEFORE HERE! */
3932 (PL_expect == XREF ||
3933 ((PL_opargs[PL_last_lop_op] >> OASHIFT)& 7) == OA_FILEREF))
3935 bool immediate_paren = *s == '(';
3937 /* (Now we can afford to cross potential line boundary.) */
3940 /* Two barewords in a row may indicate method call. */
3942 if ((isIDFIRST_lazy_if(s,UTF) || *s == '$') && (tmp=intuit_method(s,gv)))
3945 /* If not a declared subroutine, it's an indirect object. */
3946 /* (But it's an indir obj regardless for sort.) */
3948 if ( !immediate_paren && (PL_last_lop_op == OP_SORT ||
3949 ((!gv || !GvCVu(gv)) &&
3950 (PL_last_lop_op != OP_MAPSTART &&
3951 PL_last_lop_op != OP_GREPSTART))))
3953 PL_expect = (PL_last_lop == PL_oldoldbufptr) ? XTERM : XOPERATOR;
3959 PL_expect = XOPERATOR;
3962 /* Is this a word before a => operator? */
3963 if (*s == '=' && s[1] == '>') {
3965 sv_setpv(((SVOP*)yylval.opval)->op_sv, PL_tokenbuf);
3966 if (UTF && !IN_BYTE && is_utf8_string((U8*)PL_tokenbuf, len))
3967 SvUTF8_on(((SVOP*)yylval.opval)->op_sv);
3971 /* If followed by a paren, it's certainly a subroutine. */
3974 if (gv && GvCVu(gv)) {
3975 for (d = s + 1; SPACE_OR_TAB(*d); d++) ;
3976 if (*d == ')' && (sv = cv_const_sv(GvCV(gv)))) {
3981 PL_nextval[PL_nexttoke].opval = yylval.opval;
3982 PL_expect = XOPERATOR;
3988 /* If followed by var or block, call it a method (unless sub) */
3990 if ((*s == '$' || *s == '{') && (!gv || !GvCVu(gv))) {
3991 PL_last_lop = PL_oldbufptr;
3992 PL_last_lop_op = OP_METHOD;
3996 /* If followed by a bareword, see if it looks like indir obj. */
3998 if ((isIDFIRST_lazy_if(s,UTF) || *s == '$') && (tmp = intuit_method(s,gv)))
4001 /* Not a method, so call it a subroutine (if defined) */
4003 if (gv && GvCVu(gv)) {
4005 if (lastchar == '-' && ckWARN_d(WARN_AMBIGUOUS))
4006 Perl_warner(aTHX_ WARN_AMBIGUOUS,
4007 "Ambiguous use of -%s resolved as -&%s()",
4008 PL_tokenbuf, PL_tokenbuf);
4009 /* Check for a constant sub */
4011 if ((sv = cv_const_sv(cv))) {
4013 SvREFCNT_dec(((SVOP*)yylval.opval)->op_sv);
4014 ((SVOP*)yylval.opval)->op_sv = SvREFCNT_inc(sv);
4015 yylval.opval->op_private = 0;
4019 /* Resolve to GV now. */
4020 op_free(yylval.opval);
4021 yylval.opval = newCVREF(0, newGVOP(OP_GV, 0, gv));
4022 yylval.opval->op_private |= OPpENTERSUB_NOPAREN;
4023 PL_last_lop = PL_oldbufptr;
4024 PL_last_lop_op = OP_ENTERSUB;
4025 /* Is there a prototype? */
4028 char *proto = SvPV((SV*)cv, len);
4031 if (strEQ(proto, "$"))
4033 if (*proto == '&' && *s == '{') {
4034 sv_setpv(PL_subname,"__ANON__");
4038 PL_nextval[PL_nexttoke].opval = yylval.opval;
4044 /* Call it a bare word */
4046 if (PL_hints & HINT_STRICT_SUBS)
4047 yylval.opval->op_private |= OPpCONST_STRICT;
4050 if (ckWARN(WARN_RESERVED)) {
4051 if (lastchar != '-') {
4052 for (d = PL_tokenbuf; *d && isLOWER(*d); d++) ;
4054 Perl_warner(aTHX_ WARN_RESERVED, PL_warn_reserved,
4061 if (lastchar && strchr("*%&", lastchar) && ckWARN_d(WARN_AMBIGUOUS)) {
4062 Perl_warner(aTHX_ WARN_AMBIGUOUS,
4063 "Operator or semicolon missing before %c%s",
4064 lastchar, PL_tokenbuf);
4065 Perl_warner(aTHX_ WARN_AMBIGUOUS,
4066 "Ambiguous use of %c resolved as operator %c",
4067 lastchar, lastchar);
4073 yylval.opval = (OP*)newSVOP(OP_CONST, 0,
4074 newSVpv(CopFILE(PL_curcop),0));
4078 yylval.opval = (OP*)newSVOP(OP_CONST, 0,
4079 Perl_newSVpvf(aTHX_ "%"IVdf, (IV)CopLINE(PL_curcop)));
4082 case KEY___PACKAGE__:
4083 yylval.opval = (OP*)newSVOP(OP_CONST, 0,
4085 ? newSVsv(PL_curstname)
4094 if (PL_rsfp && (!PL_in_eval || PL_tokenbuf[2] == 'D')) {
4095 char *pname = "main";
4096 if (PL_tokenbuf[2] == 'D')
4097 pname = HvNAME(PL_curstash ? PL_curstash : PL_defstash);
4098 gv = gv_fetchpv(Perl_form(aTHX_ "%s::DATA", pname), TRUE, SVt_PVIO);
4101 GvIOp(gv) = newIO();
4102 IoIFP(GvIOp(gv)) = PL_rsfp;
4103 #if defined(HAS_FCNTL) && defined(F_SETFD)
4105 int fd = PerlIO_fileno(PL_rsfp);
4106 fcntl(fd,F_SETFD,fd >= 3);
4109 /* Mark this internal pseudo-handle as clean */
4110 IoFLAGS(GvIOp(gv)) |= IOf_UNTAINT;
4112 IoTYPE(GvIOp(gv)) = IoTYPE_PIPE;
4113 else if ((PerlIO*)PL_rsfp == PerlIO_stdin())
4114 IoTYPE(GvIOp(gv)) = IoTYPE_STD;
4116 IoTYPE(GvIOp(gv)) = IoTYPE_RDONLY;
4117 #if defined(WIN32) && !defined(PERL_TEXTMODE_SCRIPTS)
4118 /* if the script was opened in binmode, we need to revert
4119 * it to text mode for compatibility; but only iff it has CRs
4120 * XXX this is a questionable hack at best. */
4121 if (PL_bufend-PL_bufptr > 2
4122 && PL_bufend[-1] == '\n' && PL_bufend[-2] == '\r')
4125 if (IoTYPE(GvIOp(gv)) == IoTYPE_RDONLY) {
4126 loc = PerlIO_tell(PL_rsfp);
4127 (void)PerlIO_seek(PL_rsfp, 0L, 0);
4129 if (PerlLIO_setmode(PerlIO_fileno(PL_rsfp), O_TEXT) != -1) {
4130 #if defined(__BORLANDC__)
4131 /* XXX see note in do_binmode() */
4132 ((FILE*)PL_rsfp)->flags |= _F_BIN;
4135 PerlIO_seek(PL_rsfp, loc, 0);
4139 #ifdef PERLIO_LAYERS
4140 if (UTF && !IN_BYTE)
4141 PerlIO_apply_layers(aTHX_ PL_rsfp, NULL, ":utf8");
4154 if (PL_expect == XSTATE) {
4161 if (*s == ':' && s[1] == ':') {
4164 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
4165 if (!(tmp = keyword(PL_tokenbuf, len)))
4166 Perl_croak(aTHX_ "CORE::%s is not a keyword", PL_tokenbuf);
4180 LOP(OP_ACCEPT,XTERM);
4186 LOP(OP_ATAN2,XTERM);
4192 LOP(OP_BINMODE,XTERM);
4195 LOP(OP_BLESS,XTERM);
4204 (void)gv_fetchpv("ENV",TRUE, SVt_PVHV); /* may use HOME */
4221 if (!PL_cryptseen) {
4222 PL_cryptseen = TRUE;
4226 LOP(OP_CRYPT,XTERM);
4229 if (ckWARN(WARN_CHMOD)) {
4230 for (d = s; d < PL_bufend && (isSPACE(*d) || *d == '('); d++) ;
4231 if (*d != '0' && isDIGIT(*d))
4232 Perl_warner(aTHX_ WARN_CHMOD,
4233 "chmod() mode argument is missing initial 0");
4235 LOP(OP_CHMOD,XTERM);
4238 LOP(OP_CHOWN,XTERM);
4241 LOP(OP_CONNECT,XTERM);
4257 s = force_word(s,WORD,FALSE,TRUE,FALSE);
4261 PL_hints |= HINT_BLOCK_SCOPE;
4271 gv_fetchpv("AnyDBM_File::ISA", GV_ADDMULTI, SVt_PVAV);
4272 LOP(OP_DBMOPEN,XTERM);
4278 s = force_word(s,WORD,TRUE,FALSE,FALSE);
4285 yylval.ival = CopLINE(PL_curcop);
4299 PL_expect = (*s == '{') ? XTERMBLOCK : XTERM;
4300 UNIBRACK(OP_ENTEREVAL);
4315 case KEY_endhostent:
4321 case KEY_endservent:
4324 case KEY_endprotoent:
4335 yylval.ival = CopLINE(PL_curcop);
4337 if (PL_expect == XSTATE && isIDFIRST_lazy_if(s,UTF)) {
4339 if ((PL_bufend - p) >= 3 &&
4340 strnEQ(p, "my", 2) && isSPACE(*(p + 2)))
4342 else if ((PL_bufend - p) >= 4 &&
4343 strnEQ(p, "our", 3) && isSPACE(*(p + 3)))
4346 if (isIDFIRST_lazy_if(p,UTF)) {
4347 p = scan_ident(p, PL_bufend,
4348 PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
4352 Perl_croak(aTHX_ "Missing $ on loop variable");
4357 LOP(OP_FORMLINE,XTERM);
4363 LOP(OP_FCNTL,XTERM);
4369 LOP(OP_FLOCK,XTERM);
4378 LOP(OP_GREPSTART, XREF);
4381 s = force_word(s,WORD,TRUE,FALSE,FALSE);
4396 case KEY_getpriority:
4397 LOP(OP_GETPRIORITY,XTERM);
4399 case KEY_getprotobyname:
4402 case KEY_getprotobynumber:
4403 LOP(OP_GPBYNUMBER,XTERM);
4405 case KEY_getprotoent:
4417 case KEY_getpeername:
4418 UNI(OP_GETPEERNAME);
4420 case KEY_gethostbyname:
4423 case KEY_gethostbyaddr:
4424 LOP(OP_GHBYADDR,XTERM);
4426 case KEY_gethostent:
4429 case KEY_getnetbyname:
4432 case KEY_getnetbyaddr:
4433 LOP(OP_GNBYADDR,XTERM);
4438 case KEY_getservbyname:
4439 LOP(OP_GSBYNAME,XTERM);
4441 case KEY_getservbyport:
4442 LOP(OP_GSBYPORT,XTERM);
4444 case KEY_getservent:
4447 case KEY_getsockname:
4448 UNI(OP_GETSOCKNAME);
4450 case KEY_getsockopt:
4451 LOP(OP_GSOCKOPT,XTERM);
4473 yylval.ival = CopLINE(PL_curcop);
4477 LOP(OP_INDEX,XTERM);
4483 LOP(OP_IOCTL,XTERM);
4495 s = force_word(s,WORD,TRUE,FALSE,FALSE);
4527 LOP(OP_LISTEN,XTERM);
4536 s = scan_pat(s,OP_MATCH);
4537 TERM(sublex_start());
4540 LOP(OP_MAPSTART, XREF);
4543 LOP(OP_MKDIR,XTERM);
4546 LOP(OP_MSGCTL,XTERM);
4549 LOP(OP_MSGGET,XTERM);
4552 LOP(OP_MSGRCV,XTERM);
4555 LOP(OP_MSGSND,XTERM);
4561 if (isIDFIRST_lazy_if(s,UTF)) {
4562 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, TRUE, &len);
4563 if (len == 3 && strnEQ(PL_tokenbuf, "sub", 3))
4565 PL_in_my_stash = find_in_my_stash(PL_tokenbuf, len);
4566 if (!PL_in_my_stash) {
4569 sprintf(tmpbuf, "No such class %.1000s", PL_tokenbuf);
4577 s = force_word(s,WORD,TRUE,FALSE,FALSE);
4584 if (PL_expect != XSTATE)
4585 yyerror("\"no\" not allowed in expression");
4586 s = force_word(s,WORD,FALSE,TRUE,FALSE);
4587 s = force_version(s);
4592 if (*s == '(' || (s = skipspace(s), *s == '('))
4599 if (isIDFIRST_lazy_if(s,UTF)) {
4601 for (d = s; isALNUM_lazy_if(d,UTF); d++) ;
4603 if (strchr("|&*+-=!?:.", *t) && ckWARN_d(WARN_PRECEDENCE))
4604 Perl_warner(aTHX_ WARN_PRECEDENCE,
4605 "Precedence problem: open %.*s should be open(%.*s)",
4611 yylval.ival = OP_OR;
4621 LOP(OP_OPEN_DIR,XTERM);
4624 checkcomma(s,PL_tokenbuf,"filehandle");
4628 checkcomma(s,PL_tokenbuf,"filehandle");
4647 s = force_word(s,WORD,FALSE,TRUE,FALSE);
4651 LOP(OP_PIPE_OP,XTERM);
4654 s = scan_str(s,FALSE,FALSE);
4656 missingterm((char*)0);
4657 yylval.ival = OP_CONST;
4658 TERM(sublex_start());
4664 s = scan_str(s,FALSE,FALSE);
4666 missingterm((char*)0);
4668 if (SvCUR(PL_lex_stuff)) {
4671 d = SvPV_force(PL_lex_stuff, len);
4674 for (; isSPACE(*d) && len; --len, ++d) ;
4677 if (!warned && ckWARN(WARN_QW)) {
4678 for (; !isSPACE(*d) && len; --len, ++d) {
4680 Perl_warner(aTHX_ WARN_QW,
4681 "Possible attempt to separate words with commas");
4684 else if (*d == '#') {
4685 Perl_warner(aTHX_ WARN_QW,
4686 "Possible attempt to put comments in qw() list");
4692 for (; !isSPACE(*d) && len; --len, ++d) ;
4694 sv = newSVpvn(b, d-b);
4695 if (DO_UTF8(PL_lex_stuff))
4697 words = append_elem(OP_LIST, words,
4698 newSVOP(OP_CONST, 0, tokeq(sv)));
4702 PL_nextval[PL_nexttoke].opval = words;
4707 SvREFCNT_dec(PL_lex_stuff);
4708 PL_lex_stuff = Nullsv;
4713 s = scan_str(s,FALSE,FALSE);
4715 missingterm((char*)0);
4716 yylval.ival = OP_STRINGIFY;
4717 if (SvIVX(PL_lex_stuff) == '\'')
4718 SvIVX(PL_lex_stuff) = 0; /* qq'$foo' should intepolate */
4719 TERM(sublex_start());
4722 s = scan_pat(s,OP_QR);
4723 TERM(sublex_start());
4726 s = scan_str(s,FALSE,FALSE);
4728 missingterm((char*)0);
4729 yylval.ival = OP_BACKTICK;
4731 TERM(sublex_start());
4738 if (isDIGIT(*s) || (*s == 'v' && isDIGIT(s[1]))) {
4739 s = force_version(s);
4742 *PL_tokenbuf = '\0';
4743 s = force_word(s,WORD,TRUE,TRUE,FALSE);
4744 if (isIDFIRST_lazy_if(PL_tokenbuf,UTF))
4745 gv_stashpvn(PL_tokenbuf, strlen(PL_tokenbuf), TRUE);
4747 yyerror("<> should be quotes");
4755 s = force_word(s,WORD,TRUE,FALSE,FALSE);
4759 LOP(OP_RENAME,XTERM);
4768 LOP(OP_RINDEX,XTERM);
4791 LOP(OP_REVERSE,XTERM);
4802 TERM(sublex_start());
4804 TOKEN(1); /* force error */
4813 LOP(OP_SELECT,XTERM);
4819 LOP(OP_SEMCTL,XTERM);
4822 LOP(OP_SEMGET,XTERM);
4825 LOP(OP_SEMOP,XTERM);
4831 LOP(OP_SETPGRP,XTERM);
4833 case KEY_setpriority:
4834 LOP(OP_SETPRIORITY,XTERM);
4836 case KEY_sethostent:
4842 case KEY_setservent:
4845 case KEY_setprotoent:
4855 LOP(OP_SEEKDIR,XTERM);
4857 case KEY_setsockopt:
4858 LOP(OP_SSOCKOPT,XTERM);
4864 LOP(OP_SHMCTL,XTERM);
4867 LOP(OP_SHMGET,XTERM);
4870 LOP(OP_SHMREAD,XTERM);
4873 LOP(OP_SHMWRITE,XTERM);
4876 LOP(OP_SHUTDOWN,XTERM);
4885 LOP(OP_SOCKET,XTERM);
4887 case KEY_socketpair:
4888 LOP(OP_SOCKPAIR,XTERM);
4891 checkcomma(s,PL_tokenbuf,"subroutine name");
4893 if (*s == ';' || *s == ')') /* probably a close */
4894 Perl_croak(aTHX_ "sort is now a reserved word");
4896 s = force_word(s,WORD,TRUE,TRUE,FALSE);
4900 LOP(OP_SPLIT,XTERM);
4903 LOP(OP_SPRINTF,XTERM);
4906 LOP(OP_SPLICE,XTERM);
4921 LOP(OP_SUBSTR,XTERM);
4927 char tmpbuf[sizeof PL_tokenbuf];
4929 expectation attrful;
4930 bool have_name, have_proto;
4935 if (isIDFIRST_lazy_if(s,UTF) || *s == '\'' ||
4936 (*s == ':' && s[1] == ':'))
4939 attrful = XATTRBLOCK;
4940 /* remember buffer pos'n for later force_word */
4941 tboffset = s - PL_oldbufptr;
4942 d = scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
4943 if (strchr(tmpbuf, ':'))
4944 sv_setpv(PL_subname, tmpbuf);
4946 sv_setsv(PL_subname,PL_curstname);
4947 sv_catpvn(PL_subname,"::",2);
4948 sv_catpvn(PL_subname,tmpbuf,len);
4955 Perl_croak(aTHX_ "Missing name in \"my sub\"");
4956 PL_expect = XTERMBLOCK;
4957 attrful = XATTRTERM;
4958 sv_setpv(PL_subname,"?");
4962 if (key == KEY_format) {
4964 PL_lex_formbrack = PL_lex_brackets + 1;
4966 (void) force_word(PL_oldbufptr + tboffset, WORD,
4971 /* Look for a prototype */
4975 s = scan_str(s,FALSE,FALSE);
4978 SvREFCNT_dec(PL_lex_stuff);
4979 PL_lex_stuff = Nullsv;
4980 Perl_croak(aTHX_ "Prototype not terminated");
4983 d = SvPVX(PL_lex_stuff);
4985 for (p = d; *p; ++p) {
4990 SvCUR(PL_lex_stuff) = tmp;
4998 if (*s == ':' && s[1] != ':')
4999 PL_expect = attrful;
5002 PL_nextval[PL_nexttoke].opval =
5003 (OP*)newSVOP(OP_CONST, 0, PL_lex_stuff);
5004 PL_lex_stuff = Nullsv;
5008 sv_setpv(PL_subname,"__ANON__");
5011 (void) force_word(PL_oldbufptr + tboffset, WORD,
5020 LOP(OP_SYSTEM,XREF);
5023 LOP(OP_SYMLINK,XTERM);
5026 LOP(OP_SYSCALL,XTERM);
5029 LOP(OP_SYSOPEN,XTERM);
5032 LOP(OP_SYSSEEK,XTERM);
5035 LOP(OP_SYSREAD,XTERM);
5038 LOP(OP_SYSWRITE,XTERM);
5042 TERM(sublex_start());
5063 LOP(OP_TRUNCATE,XTERM);
5075 yylval.ival = CopLINE(PL_curcop);
5079 yylval.ival = CopLINE(PL_curcop);
5083 LOP(OP_UNLINK,XTERM);
5089 LOP(OP_UNPACK,XTERM);
5092 LOP(OP_UTIME,XTERM);
5095 if (ckWARN(WARN_UMASK)) {
5096 for (d = s; d < PL_bufend && (isSPACE(*d) || *d == '('); d++) ;
5097 if (*d != '0' && isDIGIT(*d))
5098 Perl_warner(aTHX_ WARN_UMASK,
5099 "umask: argument is missing initial 0");
5104 LOP(OP_UNSHIFT,XTERM);
5107 if (PL_expect != XSTATE)
5108 yyerror("\"use\" not allowed in expression");
5110 if (isDIGIT(*s) || (*s == 'v' && isDIGIT(s[1]))) {
5111 s = force_version(s);
5112 if (*s == ';' || (s = skipspace(s), *s == ';')) {
5113 PL_nextval[PL_nexttoke].opval = Nullop;
5118 s = force_word(s,WORD,FALSE,TRUE,FALSE);
5119 s = force_version(s);
5131 yylval.ival = CopLINE(PL_curcop);
5135 PL_hints |= HINT_BLOCK_SCOPE;
5142 LOP(OP_WAITPID,XTERM);
5150 static char ctl_l[2];
5152 if (ctl_l[0] == '\0')
5153 ctl_l[0] = toCTRL('L');
5154 gv_fetchpv(ctl_l,TRUE, SVt_PV);
5157 gv_fetchpv("\f",TRUE, SVt_PV); /* Make sure $^L is defined */
5162 if (PL_expect == XOPERATOR)
5168 yylval.ival = OP_XOR;
5173 TERM(sublex_start());
5178 #pragma segment Main
5182 Perl_keyword(pTHX_ register char *d, I32 len)
5187 if (strEQ(d,"__FILE__")) return -KEY___FILE__;
5188 if (strEQ(d,"__LINE__")) return -KEY___LINE__;
5189 if (strEQ(d,"__PACKAGE__")) return -KEY___PACKAGE__;
5190 if (strEQ(d,"__DATA__")) return KEY___DATA__;
5191 if (strEQ(d,"__END__")) return KEY___END__;
5195 if (strEQ(d,"AUTOLOAD")) return KEY_AUTOLOAD;
5200 if (strEQ(d,"and")) return -KEY_and;
5201 if (strEQ(d,"abs")) return -KEY_abs;
5204 if (strEQ(d,"alarm")) return -KEY_alarm;
5205 if (strEQ(d,"atan2")) return -KEY_atan2;
5208 if (strEQ(d,"accept")) return -KEY_accept;
5213 if (strEQ(d,"BEGIN")) return KEY_BEGIN;
5216 if (strEQ(d,"bless")) return -KEY_bless;
5217 if (strEQ(d,"bind")) return -KEY_bind;
5218 if (strEQ(d,"binmode")) return -KEY_binmode;
5221 if (strEQ(d,"CORE")) return -KEY_CORE;
5222 if (strEQ(d,"CHECK")) return KEY_CHECK;
5227 if (strEQ(d,"cmp")) return -KEY_cmp;
5228 if (strEQ(d,"chr")) return -KEY_chr;
5229 if (strEQ(d,"cos")) return -KEY_cos;
5232 if (strEQ(d,"chop")) return -KEY_chop;
5235 if (strEQ(d,"close")) return -KEY_close;
5236 if (strEQ(d,"chdir")) return -KEY_chdir;
5237 if (strEQ(d,"chomp")) return -KEY_chomp;
5238 if (strEQ(d,"chmod")) return -KEY_chmod;
5239 if (strEQ(d,"chown")) return -KEY_chown;
5240 if (strEQ(d,"crypt")) return -KEY_crypt;
5243 if (strEQ(d,"chroot")) return -KEY_chroot;
5244 if (strEQ(d,"caller")) return -KEY_caller;
5247 if (strEQ(d,"connect")) return -KEY_connect;
5250 if (strEQ(d,"closedir")) return -KEY_closedir;
5251 if (strEQ(d,"continue")) return -KEY_continue;
5256 if (strEQ(d,"DESTROY")) return KEY_DESTROY;
5261 if (strEQ(d,"do")) return KEY_do;
5264 if (strEQ(d,"die")) return -KEY_die;
5267 if (strEQ(d,"dump")) return -KEY_dump;
5270 if (strEQ(d,"delete")) return KEY_delete;
5273 if (strEQ(d,"defined")) return KEY_defined;
5274 if (strEQ(d,"dbmopen")) return -KEY_dbmopen;
5277 if (strEQ(d,"dbmclose")) return -KEY_dbmclose;
5282 if (strEQ(d,"END")) return KEY_END;
5287 if (strEQ(d,"eq")) return -KEY_eq;
5290 if (strEQ(d,"eof")) return -KEY_eof;
5291 if (strEQ(d,"exp")) return -KEY_exp;
5294 if (strEQ(d,"else")) return KEY_else;
5295 if (strEQ(d,"exit")) return -KEY_exit;
5296 if (strEQ(d,"eval")) return KEY_eval;
5297 if (strEQ(d,"exec")) return -KEY_exec;
5298 if (strEQ(d,"each")) return -KEY_each;
5301 if (strEQ(d,"elsif")) return KEY_elsif;
5304 if (strEQ(d,"exists")) return KEY_exists;
5305 if (strEQ(d,"elseif")) Perl_warn(aTHX_ "elseif should be elsif");
5308 if (strEQ(d,"endgrent")) return -KEY_endgrent;
5309 if (strEQ(d,"endpwent")) return -KEY_endpwent;
5312 if (strEQ(d,"endnetent")) return -KEY_endnetent;
5315 if (strEQ(d,"endhostent")) return -KEY_endhostent;
5316 if (strEQ(d,"endservent")) return -KEY_endservent;
5319 if (strEQ(d,"endprotoent")) return -KEY_endprotoent;
5326 if (strEQ(d,"for")) return KEY_for;
5329 if (strEQ(d,"fork")) return -KEY_fork;
5332 if (strEQ(d,"fcntl")) return -KEY_fcntl;
5333 if (strEQ(d,"flock")) return -KEY_flock;
5336 if (strEQ(d,"format")) return KEY_format;
5337 if (strEQ(d,"fileno")) return -KEY_fileno;
5340 if (strEQ(d,"foreach")) return KEY_foreach;
5343 if (strEQ(d,"formline")) return -KEY_formline;
5348 if (strnEQ(d,"get",3)) {
5353 if (strEQ(d,"ppid")) return -KEY_getppid;
5354 if (strEQ(d,"pgrp")) return -KEY_getpgrp;
5357 if (strEQ(d,"pwent")) return -KEY_getpwent;
5358 if (strEQ(d,"pwnam")) return -KEY_getpwnam;
5359 if (strEQ(d,"pwuid")) return -KEY_getpwuid;
5362 if (strEQ(d,"peername")) return -KEY_getpeername;
5363 if (strEQ(d,"protoent")) return -KEY_getprotoent;
5364 if (strEQ(d,"priority")) return -KEY_getpriority;
5367 if (strEQ(d,"protobyname")) return -KEY_getprotobyname;
5370 if (strEQ(d,"protobynumber"))return -KEY_getprotobynumber;
5374 else if (*d == 'h') {
5375 if (strEQ(d,"hostbyname")) return -KEY_gethostbyname;
5376 if (strEQ(d,"hostbyaddr")) return -KEY_gethostbyaddr;
5377 if (strEQ(d,"hostent")) return -KEY_gethostent;
5379 else if (*d == 'n') {
5380 if (strEQ(d,"netbyname")) return -KEY_getnetbyname;
5381 if (strEQ(d,"netbyaddr")) return -KEY_getnetbyaddr;
5382 if (strEQ(d,"netent")) return -KEY_getnetent;
5384 else if (*d == 's') {
5385 if (strEQ(d,"servbyname")) return -KEY_getservbyname;
5386 if (strEQ(d,"servbyport")) return -KEY_getservbyport;
5387 if (strEQ(d,"servent")) return -KEY_getservent;
5388 if (strEQ(d,"sockname")) return -KEY_getsockname;
5389 if (strEQ(d,"sockopt")) return -KEY_getsockopt;
5391 else if (*d == 'g') {
5392 if (strEQ(d,"grent")) return -KEY_getgrent;
5393 if (strEQ(d,"grnam")) return -KEY_getgrnam;
5394 if (strEQ(d,"grgid")) return -KEY_getgrgid;
5396 else if (*d == 'l') {
5397 if (strEQ(d,"login")) return -KEY_getlogin;
5399 else if (strEQ(d,"c")) return -KEY_getc;
5404 if (strEQ(d,"gt")) return -KEY_gt;
5405 if (strEQ(d,"ge")) return -KEY_ge;
5408 if (strEQ(d,"grep")) return KEY_grep;
5409 if (strEQ(d,"goto")) return KEY_goto;
5410 if (strEQ(d,"glob")) return KEY_glob;
5413 if (strEQ(d,"gmtime")) return -KEY_gmtime;
5418 if (strEQ(d,"hex")) return -KEY_hex;
5421 if (strEQ(d,"INIT")) return KEY_INIT;
5426 if (strEQ(d,"if")) return KEY_if;
5429 if (strEQ(d,"int")) return -KEY_int;
5432 if (strEQ(d,"index")) return -KEY_index;
5433 if (strEQ(d,"ioctl")) return -KEY_ioctl;
5438 if (strEQ(d,"join")) return -KEY_join;
5442 if (strEQ(d,"keys")) return -KEY_keys;
5443 if (strEQ(d,"kill")) return -KEY_kill;
5449 if (strEQ(d,"lt")) return -KEY_lt;
5450 if (strEQ(d,"le")) return -KEY_le;
5451 if (strEQ(d,"lc")) return -KEY_lc;
5454 if (strEQ(d,"log")) return -KEY_log;
5457 if (strEQ(d,"last")) return KEY_last;
5458 if (strEQ(d,"link")) return -KEY_link;
5459 if (strEQ(d,"lock")) return -KEY_lock;
5462 if (strEQ(d,"local")) return KEY_local;
5463 if (strEQ(d,"lstat")) return -KEY_lstat;
5466 if (strEQ(d,"length")) return -KEY_length;
5467 if (strEQ(d,"listen")) return -KEY_listen;
5470 if (strEQ(d,"lcfirst")) return -KEY_lcfirst;
5473 if (strEQ(d,"localtime")) return -KEY_localtime;
5479 case 1: return KEY_m;
5481 if (strEQ(d,"my")) return KEY_my;
5484 if (strEQ(d,"map")) return KEY_map;
5487 if (strEQ(d,"mkdir")) return -KEY_mkdir;
5490 if (strEQ(d,"msgctl")) return -KEY_msgctl;
5491 if (strEQ(d,"msgget")) return -KEY_msgget;
5492 if (strEQ(d,"msgrcv")) return -KEY_msgrcv;
5493 if (strEQ(d,"msgsnd")) return -KEY_msgsnd;
5498 if (strEQ(d,"next")) return KEY_next;
5499 if (strEQ(d,"ne")) return -KEY_ne;
5500 if (strEQ(d,"not")) return -KEY_not;
5501 if (strEQ(d,"no")) return KEY_no;
5506 if (strEQ(d,"or")) return -KEY_or;
5509 if (strEQ(d,"ord")) return -KEY_ord;
5510 if (strEQ(d,"oct")) return -KEY_oct;
5511 if (strEQ(d,"our")) return KEY_our;
5514 if (strEQ(d,"open")) return -KEY_open;
5517 if (strEQ(d,"opendir")) return -KEY_opendir;
5524 if (strEQ(d,"pop")) return -KEY_pop;
5525 if (strEQ(d,"pos")) return KEY_pos;
5528 if (strEQ(d,"push")) return -KEY_push;
5529 if (strEQ(d,"pack")) return -KEY_pack;
5530 if (strEQ(d,"pipe")) return -KEY_pipe;
5533 if (strEQ(d,"print")) return KEY_print;
5536 if (strEQ(d,"printf")) return KEY_printf;
5539 if (strEQ(d,"package")) return KEY_package;
5542 if (strEQ(d,"prototype")) return KEY_prototype;
5547 if (strEQ(d,"q")) return KEY_q;
5548 if (strEQ(d,"qr")) return KEY_qr;
5549 if (strEQ(d,"qq")) return KEY_qq;
5550 if (strEQ(d,"qw")) return KEY_qw;
5551 if (strEQ(d,"qx")) return KEY_qx;
5553 else if (strEQ(d,"quotemeta")) return -KEY_quotemeta;
5558 if (strEQ(d,"ref")) return -KEY_ref;
5561 if (strEQ(d,"read")) return -KEY_read;
5562 if (strEQ(d,"rand")) return -KEY_rand;
5563 if (strEQ(d,"recv")) return -KEY_recv;
5564 if (strEQ(d,"redo")) return KEY_redo;
5567 if (strEQ(d,"rmdir")) return -KEY_rmdir;
5568 if (strEQ(d,"reset")) return -KEY_reset;
5571 if (strEQ(d,"return")) return KEY_return;
5572 if (strEQ(d,"rename")) return -KEY_rename;
5573 if (strEQ(d,"rindex")) return -KEY_rindex;
5576 if (strEQ(d,"require")) return -KEY_require;
5577 if (strEQ(d,"reverse")) return -KEY_reverse;
5578 if (strEQ(d,"readdir")) return -KEY_readdir;
5581 if (strEQ(d,"readlink")) return -KEY_readlink;
5582 if (strEQ(d,"readline")) return -KEY_readline;
5583 if (strEQ(d,"readpipe")) return -KEY_readpipe;
5586 if (strEQ(d,"rewinddir")) return -KEY_rewinddir;
5592 case 0: return KEY_s;
5594 if (strEQ(d,"scalar")) return KEY_scalar;
5599 if (strEQ(d,"seek")) return -KEY_seek;
5600 if (strEQ(d,"send")) return -KEY_send;
5603 if (strEQ(d,"semop")) return -KEY_semop;
5606 if (strEQ(d,"select")) return -KEY_select;
5607 if (strEQ(d,"semctl")) return -KEY_semctl;
5608 if (strEQ(d,"semget")) return -KEY_semget;
5611 if (strEQ(d,"setpgrp")) return -KEY_setpgrp;
5612 if (strEQ(d,"seekdir")) return -KEY_seekdir;
5615 if (strEQ(d,"setpwent")) return -KEY_setpwent;
5616 if (strEQ(d,"setgrent")) return -KEY_setgrent;
5619 if (strEQ(d,"setnetent")) return -KEY_setnetent;
5622 if (strEQ(d,"setsockopt")) return -KEY_setsockopt;
5623 if (strEQ(d,"sethostent")) return -KEY_sethostent;
5624 if (strEQ(d,"setservent")) return -KEY_setservent;
5627 if (strEQ(d,"setpriority")) return -KEY_setpriority;
5628 if (strEQ(d,"setprotoent")) return -KEY_setprotoent;
5635 if (strEQ(d,"shift")) return -KEY_shift;
5638 if (strEQ(d,"shmctl")) return -KEY_shmctl;
5639 if (strEQ(d,"shmget")) return -KEY_shmget;
5642 if (strEQ(d,"shmread")) return -KEY_shmread;
5645 if (strEQ(d,"shmwrite")) return -KEY_shmwrite;
5646 if (strEQ(d,"shutdown")) return -KEY_shutdown;
5651 if (strEQ(d,"sin")) return -KEY_sin;
5654 if (strEQ(d,"sleep")) return -KEY_sleep;
5657 if (strEQ(d,"sort")) return KEY_sort;
5658 if (strEQ(d,"socket")) return -KEY_socket;
5659 if (strEQ(d,"socketpair")) return -KEY_socketpair;
5662 if (strEQ(d,"split")) return KEY_split;
5663 if (strEQ(d,"sprintf")) return -KEY_sprintf;
5664 if (strEQ(d,"splice")) return -KEY_splice;
5667 if (strEQ(d,"sqrt")) return -KEY_sqrt;
5670 if (strEQ(d,"srand")) return -KEY_srand;
5673 if (strEQ(d,"stat")) return -KEY_stat;
5674 if (strEQ(d,"study")) return KEY_study;
5677 if (strEQ(d,"substr")) return -KEY_substr;
5678 if (strEQ(d,"sub")) return KEY_sub;
5683 if (strEQ(d,"system")) return -KEY_system;
5686 if (strEQ(d,"symlink")) return -KEY_symlink;
5687 if (strEQ(d,"syscall")) return -KEY_syscall;
5688 if (strEQ(d,"sysopen")) return -KEY_sysopen;
5689 if (strEQ(d,"sysread")) return -KEY_sysread;
5690 if (strEQ(d,"sysseek")) return -KEY_sysseek;
5693 if (strEQ(d,"syswrite")) return -KEY_syswrite;
5702 if (strEQ(d,"tr")) return KEY_tr;
5705 if (strEQ(d,"tie")) return KEY_tie;
5708 if (strEQ(d,"tell")) return -KEY_tell;
5709 if (strEQ(d,"tied")) return KEY_tied;
5710 if (strEQ(d,"time")) return -KEY_time;
5713 if (strEQ(d,"times")) return -KEY_times;
5716 if (strEQ(d,"telldir")) return -KEY_telldir;
5719 if (strEQ(d,"truncate")) return -KEY_truncate;
5726 if (strEQ(d,"uc")) return -KEY_uc;
5729 if (strEQ(d,"use")) return KEY_use;
5732 if (strEQ(d,"undef")) return KEY_undef;
5733 if (strEQ(d,"until")) return KEY_until;
5734 if (strEQ(d,"untie")) return KEY_untie;
5735 if (strEQ(d,"utime")) return -KEY_utime;
5736 if (strEQ(d,"umask")) return -KEY_umask;
5739 if (strEQ(d,"unless")) return KEY_unless;
5740 if (strEQ(d,"unpack")) return -KEY_unpack;
5741 if (strEQ(d,"unlink")) return -KEY_unlink;
5744 if (strEQ(d,"unshift")) return -KEY_unshift;
5745 if (strEQ(d,"ucfirst")) return -KEY_ucfirst;
5750 if (strEQ(d,"values")) return -KEY_values;
5751 if (strEQ(d,"vec")) return -KEY_vec;
5756 if (strEQ(d,"warn")) return -KEY_warn;
5757 if (strEQ(d,"wait")) return -KEY_wait;
5760 if (strEQ(d,"while")) return KEY_while;
5761 if (strEQ(d,"write")) return -KEY_write;
5764 if (strEQ(d,"waitpid")) return -KEY_waitpid;
5767 if (strEQ(d,"wantarray")) return -KEY_wantarray;
5772 if (len == 1) return -KEY_x;
5773 if (strEQ(d,"xor")) return -KEY_xor;
5776 if (len == 1) return KEY_y;
5785 S_checkcomma(pTHX_ register char *s, char *name, char *what)
5789 if (*s == ' ' && s[1] == '(') { /* XXX gotta be a better way */
5790 if (ckWARN(WARN_SYNTAX)) {
5792 for (w = s+2; *w && level; w++) {
5799 for (; *w && isSPACE(*w); w++) ;
5800 if (!*w || !strchr(";|})]oaiuw!=", *w)) /* an advisory hack only... */
5801 Perl_warner(aTHX_ WARN_SYNTAX,
5802 "%s (...) interpreted as function",name);
5805 while (s < PL_bufend && isSPACE(*s))
5809 while (s < PL_bufend && isSPACE(*s))
5811 if (isIDFIRST_lazy_if(s,UTF)) {
5813 while (isALNUM_lazy_if(s,UTF))
5815 while (s < PL_bufend && isSPACE(*s))
5820 kw = keyword(w, s - w) || get_cv(w, FALSE) != 0;
5824 Perl_croak(aTHX_ "No comma allowed after %s", what);
5829 /* Either returns sv, or mortalizes sv and returns a new SV*.
5830 Best used as sv=new_constant(..., sv, ...).
5831 If s, pv are NULL, calls subroutine with one argument,
5832 and type is used with error messages only. */
5835 S_new_constant(pTHX_ char *s, STRLEN len, const char *key, SV *sv, SV *pv,
5839 HV *table = GvHV(PL_hintgv); /* ^H */
5843 const char *why1, *why2, *why3;
5845 if (!table || !(PL_hints & HINT_LOCALIZE_HH)) {
5848 why2 = strEQ(key,"charnames")
5849 ? "(possibly a missing \"use charnames ...\")"
5851 msg = Perl_newSVpvf(aTHX_ "Constant(%s) unknown: %s",
5852 (type ? type: "undef"), why2);
5854 /* This is convoluted and evil ("goto considered harmful")
5855 * but I do not understand the intricacies of all the different
5856 * failure modes of %^H in here. The goal here is to make
5857 * the most probable error message user-friendly. --jhi */
5862 msg = Perl_newSVpvf(aTHX_ "Constant(%s): %s%s%s",
5863 (type ? type: "undef"), why1, why2, why3);
5865 yyerror(SvPVX(msg));
5869 cvp = hv_fetch(table, key, strlen(key), FALSE);
5870 if (!cvp || !SvOK(*cvp)) {
5873 why3 = "} is not defined";
5876 sv_2mortal(sv); /* Parent created it permanently */
5879 pv = sv_2mortal(newSVpvn(s, len));
5881 typesv = sv_2mortal(newSVpv(type, 0));
5883 typesv = &PL_sv_undef;
5885 PUSHSTACKi(PERLSI_OVERLOAD);
5897 call_sv(cv, G_SCALAR | ( PL_in_eval ? 0 : G_EVAL));
5901 /* Check the eval first */
5902 if (!PL_in_eval && SvTRUE(ERRSV)) {
5904 sv_catpv(ERRSV, "Propagated");
5905 yyerror(SvPV(ERRSV, n_a)); /* Duplicates the message inside eval */
5907 res = SvREFCNT_inc(sv);
5911 (void)SvREFCNT_inc(res);
5920 why1 = "Call to &{$^H{";
5922 why3 = "}} did not return a defined value";
5931 S_scan_word(pTHX_ register char *s, char *dest, STRLEN destlen, int allow_package, STRLEN *slp)
5933 register char *d = dest;
5934 register char *e = d + destlen - 3; /* two-character token, ending NUL */
5937 Perl_croak(aTHX_ ident_too_long);
5938 if (isALNUM(*s)) /* UTF handled below */
5940 else if (*s == '\'' && allow_package && isIDFIRST_lazy_if(s+1,UTF)) {
5945 else if (*s == ':' && s[1] == ':' && allow_package && s[2] != '$') {
5949 else if (UTF && UTF8_IS_START(*s) && isALNUM_utf8((U8*)s)) {
5950 char *t = s + UTF8SKIP(s);
5951 while (UTF8_IS_CONTINUED(*t) && is_utf8_mark((U8*)t))
5953 if (d + (t - s) > e)
5954 Perl_croak(aTHX_ ident_too_long);
5955 Copy(s, d, t - s, char);
5968 S_scan_ident(pTHX_ register char *s, register char *send, char *dest, STRLEN destlen, I32 ck_uni)
5978 e = d + destlen - 3; /* two-character token, ending NUL */
5980 while (isDIGIT(*s)) {
5982 Perl_croak(aTHX_ ident_too_long);
5989 Perl_croak(aTHX_ ident_too_long);
5990 if (isALNUM(*s)) /* UTF handled below */
5992 else if (*s == '\'' && isIDFIRST_lazy_if(s+1,UTF)) {
5997 else if (*s == ':' && s[1] == ':') {
6001 else if (UTF && UTF8_IS_START(*s) && isALNUM_utf8((U8*)s)) {
6002 char *t = s + UTF8SKIP(s);
6003 while (UTF8_IS_CONTINUED(*t) && is_utf8_mark((U8*)t))
6005 if (d + (t - s) > e)
6006 Perl_croak(aTHX_ ident_too_long);
6007 Copy(s, d, t - s, char);
6018 if (PL_lex_state != LEX_NORMAL)
6019 PL_lex_state = LEX_INTERPENDMAYBE;
6022 if (*s == '$' && s[1] &&
6023 (isALNUM_lazy_if(s+1,UTF) || strchr("${", s[1]) || strnEQ(s+1,"::",2)) )
6036 if (*d == '^' && *s && isCONTROLVAR(*s)) {
6041 if (isSPACE(s[-1])) {
6044 if (!SPACE_OR_TAB(ch)) {
6050 if (isIDFIRST_lazy_if(d,UTF)) {
6054 while ((e < send && isALNUM_lazy_if(e,UTF)) || *e == ':') {
6056 while (e < send && UTF8_IS_CONTINUED(*e) && is_utf8_mark((U8*)e))
6059 Copy(s, d, e - s, char);
6064 while ((isALNUM(*s) || *s == ':') && d < e)
6067 Perl_croak(aTHX_ ident_too_long);
6070 while (s < send && SPACE_OR_TAB(*s)) s++;
6071 if ((*s == '[' || (*s == '{' && strNE(dest, "sub")))) {
6072 if (ckWARN(WARN_AMBIGUOUS) && keyword(dest, d - dest)) {
6073 const char *brack = *s == '[' ? "[...]" : "{...}";
6074 Perl_warner(aTHX_ WARN_AMBIGUOUS,
6075 "Ambiguous use of %c{%s%s} resolved to %c%s%s",
6076 funny, dest, brack, funny, dest, brack);
6079 PL_lex_brackstack[PL_lex_brackets++] = (char)(XOPERATOR | XFAKEBRACK);
6083 /* Handle extended ${^Foo} variables
6084 * 1999-02-27 mjd-perl-patch@plover.com */
6085 else if (!isALNUM(*d) && !isPRINT(*d) /* isCTRL(d) */
6089 while (isALNUM(*s) && d < e) {
6093 Perl_croak(aTHX_ ident_too_long);
6098 if (PL_lex_state == LEX_INTERPNORMAL && !PL_lex_brackets)
6099 PL_lex_state = LEX_INTERPEND;
6102 if (PL_lex_state == LEX_NORMAL) {
6103 if (ckWARN(WARN_AMBIGUOUS) &&
6104 (keyword(dest, d - dest) || get_cv(dest, FALSE)))
6106 Perl_warner(aTHX_ WARN_AMBIGUOUS,
6107 "Ambiguous use of %c{%s} resolved to %c%s",
6108 funny, dest, funny, dest);
6113 s = bracket; /* let the parser handle it */
6117 else if (PL_lex_state == LEX_INTERPNORMAL && !PL_lex_brackets && !intuit_more(s))
6118 PL_lex_state = LEX_INTERPEND;
6123 Perl_pmflag(pTHX_ U16 *pmfl, int ch)
6128 *pmfl |= PMf_GLOBAL;
6130 *pmfl |= PMf_CONTINUE;
6134 *pmfl |= PMf_MULTILINE;
6136 *pmfl |= PMf_SINGLELINE;
6138 *pmfl |= PMf_EXTENDED;
6142 S_scan_pat(pTHX_ char *start, I32 type)
6147 s = scan_str(start,FALSE,FALSE);
6150 SvREFCNT_dec(PL_lex_stuff);
6151 PL_lex_stuff = Nullsv;
6152 Perl_croak(aTHX_ "Search pattern not terminated");
6155 pm = (PMOP*)newPMOP(type, 0);
6156 if (PL_multi_open == '?')
6157 pm->op_pmflags |= PMf_ONCE;
6159 while (*s && strchr("iomsx", *s))
6160 pmflag(&pm->op_pmflags,*s++);
6163 while (*s && strchr("iogcmsx", *s))
6164 pmflag(&pm->op_pmflags,*s++);
6166 pm->op_pmpermflags = pm->op_pmflags;
6168 PL_lex_op = (OP*)pm;
6169 yylval.ival = OP_MATCH;
6174 S_scan_subst(pTHX_ char *start)
6181 yylval.ival = OP_NULL;
6183 s = scan_str(start,FALSE,FALSE);
6187 SvREFCNT_dec(PL_lex_stuff);
6188 PL_lex_stuff = Nullsv;
6189 Perl_croak(aTHX_ "Substitution pattern not terminated");
6192 if (s[-1] == PL_multi_open)
6195 first_start = PL_multi_start;
6196 s = scan_str(s,FALSE,FALSE);
6199 SvREFCNT_dec(PL_lex_stuff);
6200 PL_lex_stuff = Nullsv;
6202 SvREFCNT_dec(PL_lex_repl);
6203 PL_lex_repl = Nullsv;
6204 Perl_croak(aTHX_ "Substitution replacement not terminated");
6206 PL_multi_start = first_start; /* so whole substitution is taken together */
6208 pm = (PMOP*)newPMOP(OP_SUBST, 0);
6214 else if (strchr("iogcmsx", *s))
6215 pmflag(&pm->op_pmflags,*s++);
6222 PL_sublex_info.super_bufptr = s;
6223 PL_sublex_info.super_bufend = PL_bufend;
6225 pm->op_pmflags |= PMf_EVAL;
6226 repl = newSVpvn("",0);
6228 sv_catpv(repl, es ? "eval " : "do ");
6229 sv_catpvn(repl, "{ ", 2);
6230 sv_catsv(repl, PL_lex_repl);
6231 sv_catpvn(repl, " };", 2);
6233 SvREFCNT_dec(PL_lex_repl);
6237 pm->op_pmpermflags = pm->op_pmflags;
6238 PL_lex_op = (OP*)pm;
6239 yylval.ival = OP_SUBST;
6244 S_scan_trans(pTHX_ char *start)
6255 yylval.ival = OP_NULL;
6257 s = scan_str(start,FALSE,FALSE);
6260 SvREFCNT_dec(PL_lex_stuff);
6261 PL_lex_stuff = Nullsv;
6262 Perl_croak(aTHX_ "Transliteration pattern not terminated");
6264 if (s[-1] == PL_multi_open)
6267 s = scan_str(s,FALSE,FALSE);
6270 SvREFCNT_dec(PL_lex_stuff);
6271 PL_lex_stuff = Nullsv;
6273 SvREFCNT_dec(PL_lex_repl);
6274 PL_lex_repl = Nullsv;
6275 Perl_croak(aTHX_ "Transliteration replacement not terminated");
6278 New(803,tbl,256,short);
6279 o = newPVOP(OP_TRANS, 0, (char*)tbl);
6281 complement = del = squash = 0;
6282 while (strchr("cds", *s)) {
6284 complement = OPpTRANS_COMPLEMENT;
6286 del = OPpTRANS_DELETE;
6288 squash = OPpTRANS_SQUASH;
6291 o->op_private = del|squash|complement|
6292 (DO_UTF8(PL_lex_stuff)? OPpTRANS_FROM_UTF : 0)|
6293 (DO_UTF8(PL_lex_repl) ? OPpTRANS_TO_UTF : 0);
6296 yylval.ival = OP_TRANS;
6301 S_scan_heredoc(pTHX_ register char *s)
6304 I32 op_type = OP_SCALAR;
6311 int outer = (PL_rsfp && !(PL_lex_inwhat == OP_SCALAR));
6315 e = PL_tokenbuf + sizeof PL_tokenbuf - 1;
6318 for (peek = s; SPACE_OR_TAB(*peek); peek++) ;
6319 if (*peek && strchr("`'\"",*peek)) {
6322 s = delimcpy(d, e, s, PL_bufend, term, &len);
6332 if (!isALNUM_lazy_if(s,UTF))
6333 deprecate("bare << to mean <<\"\"");
6334 for (; isALNUM_lazy_if(s,UTF); s++) {
6339 if (d >= PL_tokenbuf + sizeof PL_tokenbuf - 1)
6340 Perl_croak(aTHX_ "Delimiter for here document is too long");
6343 len = d - PL_tokenbuf;
6344 #ifndef PERL_STRICT_CR
6345 d = strchr(s, '\r');
6349 while (s < PL_bufend) {
6355 else if (*s == '\n' && s[1] == '\r') { /* \015\013 on a mac? */
6364 SvCUR_set(PL_linestr, PL_bufend - SvPVX(PL_linestr));
6369 if (outer || !(d=ninstr(s,PL_bufend,d,d+1)))
6370 herewas = newSVpvn(s,PL_bufend-s);
6372 s--, herewas = newSVpvn(s,d-s);
6373 s += SvCUR(herewas);
6375 tmpstr = NEWSV(87,79);
6376 sv_upgrade(tmpstr, SVt_PVIV);
6381 else if (term == '`') {
6382 op_type = OP_BACKTICK;
6383 SvIVX(tmpstr) = '\\';
6387 PL_multi_start = CopLINE(PL_curcop);
6388 PL_multi_open = PL_multi_close = '<';
6389 term = *PL_tokenbuf;
6390 if (PL_lex_inwhat == OP_SUBST && PL_in_eval && !PL_rsfp) {
6391 char *bufptr = PL_sublex_info.super_bufptr;
6392 char *bufend = PL_sublex_info.super_bufend;
6393 char *olds = s - SvCUR(herewas);
6394 s = strchr(bufptr, '\n');
6398 while (s < bufend &&
6399 (*s != term || memNE(s,PL_tokenbuf,len)) ) {
6401 CopLINE_inc(PL_curcop);
6404 CopLINE_set(PL_curcop, PL_multi_start);
6405 missingterm(PL_tokenbuf);
6407 sv_setpvn(herewas,bufptr,d-bufptr+1);
6408 sv_setpvn(tmpstr,d+1,s-d);
6410 sv_catpvn(herewas,s,bufend-s);
6411 (void)strcpy(bufptr,SvPVX(herewas));
6418 while (s < PL_bufend &&
6419 (*s != term || memNE(s,PL_tokenbuf,len)) ) {
6421 CopLINE_inc(PL_curcop);
6423 if (s >= PL_bufend) {
6424 CopLINE_set(PL_curcop, PL_multi_start);
6425 missingterm(PL_tokenbuf);
6427 sv_setpvn(tmpstr,d+1,s-d);
6429 CopLINE_inc(PL_curcop); /* the preceding stmt passes a newline */
6431 sv_catpvn(herewas,s,PL_bufend-s);
6432 sv_setsv(PL_linestr,herewas);
6433 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = s = PL_linestart = SvPVX(PL_linestr);
6434 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
6437 sv_setpvn(tmpstr,"",0); /* avoid "uninitialized" warning */
6438 while (s >= PL_bufend) { /* multiple line string? */
6440 !(PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = filter_gets(PL_linestr, PL_rsfp, 0))) {
6441 CopLINE_set(PL_curcop, PL_multi_start);
6442 missingterm(PL_tokenbuf);
6444 CopLINE_inc(PL_curcop);
6445 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
6446 #ifndef PERL_STRICT_CR
6447 if (PL_bufend - PL_linestart >= 2) {
6448 if ((PL_bufend[-2] == '\r' && PL_bufend[-1] == '\n') ||
6449 (PL_bufend[-2] == '\n' && PL_bufend[-1] == '\r'))
6451 PL_bufend[-2] = '\n';
6453 SvCUR_set(PL_linestr, PL_bufend - SvPVX(PL_linestr));
6455 else if (PL_bufend[-1] == '\r')
6456 PL_bufend[-1] = '\n';
6458 else if (PL_bufend - PL_linestart == 1 && PL_bufend[-1] == '\r')
6459 PL_bufend[-1] = '\n';
6461 if (PERLDB_LINE && PL_curstash != PL_debstash) {
6462 SV *sv = NEWSV(88,0);
6464 sv_upgrade(sv, SVt_PVMG);
6465 sv_setsv(sv,PL_linestr);
6466 av_store(CopFILEAV(PL_curcop), (I32)CopLINE(PL_curcop),sv);
6468 if (*s == term && memEQ(s,PL_tokenbuf,len)) {
6471 sv_catsv(PL_linestr,herewas);
6472 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
6476 sv_catsv(tmpstr,PL_linestr);
6481 PL_multi_end = CopLINE(PL_curcop);
6482 if (SvCUR(tmpstr) + 5 < SvLEN(tmpstr)) {
6483 SvLEN_set(tmpstr, SvCUR(tmpstr) + 1);
6484 Renew(SvPVX(tmpstr), SvLEN(tmpstr), char);
6486 SvREFCNT_dec(herewas);
6487 if (UTF && !IN_BYTE && is_utf8_string((U8*)SvPVX(tmpstr), SvCUR(tmpstr)))
6489 PL_lex_stuff = tmpstr;
6490 yylval.ival = op_type;
6495 takes: current position in input buffer
6496 returns: new position in input buffer
6497 side-effects: yylval and lex_op are set.
6502 <FH> read from filehandle
6503 <pkg::FH> read from package qualified filehandle
6504 <pkg'FH> read from package qualified filehandle
6505 <$fh> read from filehandle in $fh
6511 S_scan_inputsymbol(pTHX_ char *start)
6513 register char *s = start; /* current position in buffer */
6519 d = PL_tokenbuf; /* start of temp holding space */
6520 e = PL_tokenbuf + sizeof PL_tokenbuf; /* end of temp holding space */
6521 end = strchr(s, '\n');
6524 s = delimcpy(d, e, s + 1, end, '>', &len); /* extract until > */
6526 /* die if we didn't have space for the contents of the <>,
6527 or if it didn't end, or if we see a newline
6530 if (len >= sizeof PL_tokenbuf)
6531 Perl_croak(aTHX_ "Excessively long <> operator");
6533 Perl_croak(aTHX_ "Unterminated <> operator");
6538 Remember, only scalar variables are interpreted as filehandles by
6539 this code. Anything more complex (e.g., <$fh{$num}>) will be
6540 treated as a glob() call.
6541 This code makes use of the fact that except for the $ at the front,
6542 a scalar variable and a filehandle look the same.
6544 if (*d == '$' && d[1]) d++;
6546 /* allow <Pkg'VALUE> or <Pkg::VALUE> */
6547 while (*d && (isALNUM_lazy_if(d,UTF) || *d == '\'' || *d == ':'))
6550 /* If we've tried to read what we allow filehandles to look like, and
6551 there's still text left, then it must be a glob() and not a getline.
6552 Use scan_str to pull out the stuff between the <> and treat it
6553 as nothing more than a string.
6556 if (d - PL_tokenbuf != len) {
6557 yylval.ival = OP_GLOB;
6559 s = scan_str(start,FALSE,FALSE);
6561 Perl_croak(aTHX_ "Glob not terminated");
6565 /* we're in a filehandle read situation */
6568 /* turn <> into <ARGV> */
6570 (void)strcpy(d,"ARGV");
6572 /* if <$fh>, create the ops to turn the variable into a
6578 /* try to find it in the pad for this block, otherwise find
6579 add symbol table ops
6581 if ((tmp = pad_findmy(d)) != NOT_IN_PAD) {
6582 OP *o = newOP(OP_PADSV, 0);
6584 PL_lex_op = (OP*)newUNOP(OP_READLINE, 0, o);
6587 GV *gv = gv_fetchpv(d+1,TRUE, SVt_PV);
6588 PL_lex_op = (OP*)newUNOP(OP_READLINE, 0,
6589 newUNOP(OP_RV2SV, 0,
6590 newGVOP(OP_GV, 0, gv)));
6592 PL_lex_op->op_flags |= OPf_SPECIAL;
6593 /* we created the ops in PL_lex_op, so make yylval.ival a null op */
6594 yylval.ival = OP_NULL;
6597 /* If it's none of the above, it must be a literal filehandle
6598 (<Foo::BAR> or <FOO>) so build a simple readline OP */
6600 GV *gv = gv_fetchpv(d,TRUE, SVt_PVIO);
6601 PL_lex_op = (OP*)newUNOP(OP_READLINE, 0, newGVOP(OP_GV, 0, gv));
6602 yylval.ival = OP_NULL;
6611 takes: start position in buffer
6612 keep_quoted preserve \ on the embedded delimiter(s)
6613 keep_delims preserve the delimiters around the string
6614 returns: position to continue reading from buffer
6615 side-effects: multi_start, multi_close, lex_repl or lex_stuff, and
6616 updates the read buffer.
6618 This subroutine pulls a string out of the input. It is called for:
6619 q single quotes q(literal text)
6620 ' single quotes 'literal text'
6621 qq double quotes qq(interpolate $here please)
6622 " double quotes "interpolate $here please"
6623 qx backticks qx(/bin/ls -l)
6624 ` backticks `/bin/ls -l`
6625 qw quote words @EXPORT_OK = qw( func() $spam )
6626 m// regexp match m/this/
6627 s/// regexp substitute s/this/that/
6628 tr/// string transliterate tr/this/that/
6629 y/// string transliterate y/this/that/
6630 ($*@) sub prototypes sub foo ($)
6631 (stuff) sub attr parameters sub foo : attr(stuff)
6632 <> readline or globs <FOO>, <>, <$fh>, or <*.c>
6634 In most of these cases (all but <>, patterns and transliterate)
6635 yylex() calls scan_str(). m// makes yylex() call scan_pat() which
6636 calls scan_str(). s/// makes yylex() call scan_subst() which calls
6637 scan_str(). tr/// and y/// make yylex() call scan_trans() which
6640 It skips whitespace before the string starts, and treats the first
6641 character as the delimiter. If the delimiter is one of ([{< then
6642 the corresponding "close" character )]}> is used as the closing
6643 delimiter. It allows quoting of delimiters, and if the string has
6644 balanced delimiters ([{<>}]) it allows nesting.
6646 The lexer always reads these strings into lex_stuff, except in the
6647 case of the operators which take *two* arguments (s/// and tr///)
6648 when it checks to see if lex_stuff is full (presumably with the 1st
6649 arg to s or tr) and if so puts the string into lex_repl.
6654 S_scan_str(pTHX_ char *start, int keep_quoted, int keep_delims)
6656 SV *sv; /* scalar value: string */
6657 char *tmps; /* temp string, used for delimiter matching */
6658 register char *s = start; /* current position in the buffer */
6659 register char term; /* terminating character */
6660 register char *to; /* current position in the sv's data */
6661 I32 brackets = 1; /* bracket nesting level */
6662 bool has_utf8 = FALSE; /* is there any utf8 content? */
6664 /* skip space before the delimiter */
6668 /* mark where we are, in case we need to report errors */
6671 /* after skipping whitespace, the next character is the terminator */
6673 if (UTF8_IS_CONTINUED(term) && UTF)
6676 /* mark where we are */
6677 PL_multi_start = CopLINE(PL_curcop);
6678 PL_multi_open = term;
6680 /* find corresponding closing delimiter */
6681 if (term && (tmps = strchr("([{< )]}> )]}>",term)))
6683 PL_multi_close = term;
6685 /* create a new SV to hold the contents. 87 is leak category, I'm
6686 assuming. 79 is the SV's initial length. What a random number. */
6688 sv_upgrade(sv, SVt_PVIV);
6690 (void)SvPOK_only(sv); /* validate pointer */
6692 /* move past delimiter and try to read a complete string */
6694 sv_catpvn(sv, s, 1);
6697 /* extend sv if need be */
6698 SvGROW(sv, SvCUR(sv) + (PL_bufend - s) + 1);
6699 /* set 'to' to the next character in the sv's string */
6700 to = SvPVX(sv)+SvCUR(sv);
6702 /* if open delimiter is the close delimiter read unbridle */
6703 if (PL_multi_open == PL_multi_close) {
6704 for (; s < PL_bufend; s++,to++) {
6705 /* embedded newlines increment the current line number */
6706 if (*s == '\n' && !PL_rsfp)
6707 CopLINE_inc(PL_curcop);
6708 /* handle quoted delimiters */
6709 if (*s == '\\' && s+1 < PL_bufend && term != '\\') {
6710 if (!keep_quoted && s[1] == term)
6712 /* any other quotes are simply copied straight through */
6716 /* terminate when run out of buffer (the for() condition), or
6717 have found the terminator */
6718 else if (*s == term)
6720 else if (!has_utf8 && UTF8_IS_CONTINUED(*s) && UTF)
6726 /* if the terminator isn't the same as the start character (e.g.,
6727 matched brackets), we have to allow more in the quoting, and
6728 be prepared for nested brackets.
6731 /* read until we run out of string, or we find the terminator */
6732 for (; s < PL_bufend; s++,to++) {
6733 /* embedded newlines increment the line count */
6734 if (*s == '\n' && !PL_rsfp)
6735 CopLINE_inc(PL_curcop);
6736 /* backslashes can escape the open or closing characters */
6737 if (*s == '\\' && s+1 < PL_bufend) {
6739 ((s[1] == PL_multi_open) || (s[1] == PL_multi_close)))
6744 /* allow nested opens and closes */
6745 else if (*s == PL_multi_close && --brackets <= 0)
6747 else if (*s == PL_multi_open)
6749 else if (!has_utf8 && UTF8_IS_CONTINUED(*s) && UTF)
6754 /* terminate the copied string and update the sv's end-of-string */
6756 SvCUR_set(sv, to - SvPVX(sv));
6759 * this next chunk reads more into the buffer if we're not done yet
6763 break; /* handle case where we are done yet :-) */
6765 #ifndef PERL_STRICT_CR
6766 if (to - SvPVX(sv) >= 2) {
6767 if ((to[-2] == '\r' && to[-1] == '\n') ||
6768 (to[-2] == '\n' && to[-1] == '\r'))
6772 SvCUR_set(sv, to - SvPVX(sv));
6774 else if (to[-1] == '\r')
6777 else if (to - SvPVX(sv) == 1 && to[-1] == '\r')
6781 /* if we're out of file, or a read fails, bail and reset the current
6782 line marker so we can report where the unterminated string began
6785 !(PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = filter_gets(PL_linestr, PL_rsfp, 0))) {
6787 CopLINE_set(PL_curcop, PL_multi_start);
6790 /* we read a line, so increment our line counter */
6791 CopLINE_inc(PL_curcop);
6793 /* update debugger info */
6794 if (PERLDB_LINE && PL_curstash != PL_debstash) {
6795 SV *sv = NEWSV(88,0);
6797 sv_upgrade(sv, SVt_PVMG);
6798 sv_setsv(sv,PL_linestr);
6799 av_store(CopFILEAV(PL_curcop), (I32)CopLINE(PL_curcop), sv);
6802 /* having changed the buffer, we must update PL_bufend */
6803 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
6806 /* at this point, we have successfully read the delimited string */
6809 sv_catpvn(sv, s, 1);
6812 PL_multi_end = CopLINE(PL_curcop);
6815 /* if we allocated too much space, give some back */
6816 if (SvCUR(sv) + 5 < SvLEN(sv)) {
6817 SvLEN_set(sv, SvCUR(sv) + 1);
6818 Renew(SvPVX(sv), SvLEN(sv), char);
6821 /* decide whether this is the first or second quoted string we've read
6834 takes: pointer to position in buffer
6835 returns: pointer to new position in buffer
6836 side-effects: builds ops for the constant in yylval.op
6838 Read a number in any of the formats that Perl accepts:
6840 0(x[0-7A-F]+)|([0-7]+)|(b[01])
6841 [\d_]+(\.[\d_]*)?[Ee](\d+)
6843 Underbars (_) are allowed in decimal numbers. If -w is on,
6844 underbars before a decimal point must be at three digit intervals.
6846 Like most scan_ routines, it uses the PL_tokenbuf buffer to hold the
6849 If it reads a number without a decimal point or an exponent, it will
6850 try converting the number to an integer and see if it can do so
6851 without loss of precision.
6855 Perl_scan_num(pTHX_ char *start, YYSTYPE* lvalp)
6857 register char *s = start; /* current position in buffer */
6858 register char *d; /* destination in temp buffer */
6859 register char *e; /* end of temp buffer */
6860 NV nv; /* number read, as a double */
6861 SV *sv = Nullsv; /* place to put the converted number */
6862 bool floatit; /* boolean: int or float? */
6863 char *lastub = 0; /* position of last underbar */
6864 static char number_too_long[] = "Number too long";
6866 /* We use the first character to decide what type of number this is */
6870 Perl_croak(aTHX_ "panic: scan_num");
6872 /* if it starts with a 0, it could be an octal number, a decimal in
6873 0.13 disguise, or a hexadecimal number, or a binary number. */
6877 u holds the "number so far"
6878 shift the power of 2 of the base
6879 (hex == 4, octal == 3, binary == 1)
6880 overflowed was the number more than we can hold?
6882 Shift is used when we add a digit. It also serves as an "are
6883 we in octal/hex/binary?" indicator to disallow hex characters
6889 bool overflowed = FALSE;
6890 static NV nvshift[5] = { 1.0, 2.0, 4.0, 8.0, 16.0 };
6891 static char* bases[5] = { "", "binary", "", "octal",
6893 static char* Bases[5] = { "", "Binary", "", "Octal",
6895 static char *maxima[5] = { "",
6896 "0b11111111111111111111111111111111",
6900 char *base, *Base, *max;
6906 } else if (s[1] == 'b') {
6910 /* check for a decimal in disguise */
6911 else if (s[1] == '.' || s[1] == 'e' || s[1] == 'E')
6913 /* so it must be octal */
6917 base = bases[shift];
6918 Base = Bases[shift];
6919 max = maxima[shift];
6921 /* read the rest of the number */
6923 /* x is used in the overflow test,
6924 b is the digit we're adding on. */
6929 /* if we don't mention it, we're done */
6938 /* 8 and 9 are not octal */
6941 yyerror(Perl_form(aTHX_ "Illegal octal digit '%c'", *s));
6945 case '2': case '3': case '4':
6946 case '5': case '6': case '7':
6948 yyerror(Perl_form(aTHX_ "Illegal binary digit '%c'", *s));
6952 b = *s++ & 15; /* ASCII digit -> value of digit */
6956 case 'a': case 'b': case 'c': case 'd': case 'e': case 'f':
6957 case 'A': case 'B': case 'C': case 'D': case 'E': case 'F':
6958 /* make sure they said 0x */
6963 /* Prepare to put the digit we have onto the end
6964 of the number so far. We check for overflows.
6969 x = u << shift; /* make room for the digit */
6971 if ((x >> shift) != u
6972 && !(PL_hints & HINT_NEW_BINARY)) {
6975 if (ckWARN_d(WARN_OVERFLOW))
6976 Perl_warner(aTHX_ WARN_OVERFLOW,
6977 "Integer overflow in %s number",
6980 u = x | b; /* add the digit to the end */
6983 n *= nvshift[shift];
6984 /* If an NV has not enough bits in its
6985 * mantissa to represent an UV this summing of
6986 * small low-order numbers is a waste of time
6987 * (because the NV cannot preserve the
6988 * low-order bits anyway): we could just
6989 * remember when did we overflow and in the
6990 * end just multiply n by the right
6998 /* if we get here, we had success: make a scalar value from
7004 if (ckWARN(WARN_PORTABLE) && n > 4294967295.0)
7005 Perl_warner(aTHX_ WARN_PORTABLE,
7006 "%s number > %s non-portable",
7012 if (ckWARN(WARN_PORTABLE) && u > 0xffffffff)
7013 Perl_warner(aTHX_ WARN_PORTABLE,
7014 "%s number > %s non-portable",
7019 if (PL_hints & HINT_NEW_BINARY)
7020 sv = new_constant(start, s - start, "binary", sv, Nullsv, NULL);
7025 handle decimal numbers.
7026 we're also sent here when we read a 0 as the first digit
7028 case '1': case '2': case '3': case '4': case '5':
7029 case '6': case '7': case '8': case '9': case '.':
7032 e = PL_tokenbuf + sizeof PL_tokenbuf - 6; /* room for various punctuation */
7035 /* read next group of digits and _ and copy into d */
7036 while (isDIGIT(*s) || *s == '_') {
7037 /* skip underscores, checking for misplaced ones
7041 if (ckWARN(WARN_SYNTAX) && lastub && s - lastub != 3)
7042 Perl_warner(aTHX_ WARN_SYNTAX, "Misplaced _ in number");
7046 /* check for end of fixed-length buffer */
7048 Perl_croak(aTHX_ number_too_long);
7049 /* if we're ok, copy the character */
7054 /* final misplaced underbar check */
7055 if (lastub && s - lastub != 3) {
7056 if (ckWARN(WARN_SYNTAX))
7057 Perl_warner(aTHX_ WARN_SYNTAX, "Misplaced _ in number");
7060 /* read a decimal portion if there is one. avoid
7061 3..5 being interpreted as the number 3. followed
7064 if (*s == '.' && s[1] != '.') {
7068 /* copy, ignoring underbars, until we run out of
7069 digits. Note: no misplaced underbar checks!
7071 for (; isDIGIT(*s) || *s == '_'; s++) {
7072 /* fixed length buffer check */
7074 Perl_croak(aTHX_ number_too_long);
7078 if (*s == '.' && isDIGIT(s[1])) {
7079 /* oops, it's really a v-string, but without the "v" */
7085 /* read exponent part, if present */
7086 if (*s && strchr("eE",*s) && strchr("+-0123456789",s[1])) {
7090 /* regardless of whether user said 3E5 or 3e5, use lower 'e' */
7091 *d++ = 'e'; /* At least some Mach atof()s don't grok 'E' */
7093 /* allow positive or negative exponent */
7094 if (*s == '+' || *s == '-')
7097 /* read digits of exponent (no underbars :-) */
7098 while (isDIGIT(*s)) {
7100 Perl_croak(aTHX_ number_too_long);
7105 /* terminate the string */
7108 /* make an sv from the string */
7111 #if defined(Strtol) && defined(Strtoul)
7114 strtol/strtoll sets errno to ERANGE if the number is too big
7115 for an integer. We try to do an integer conversion first
7116 if no characters indicating "float" have been found.
7123 if (*PL_tokenbuf == '-')
7124 iv = Strtol(PL_tokenbuf, (char**)NULL, 10);
7126 uv = Strtoul(PL_tokenbuf, (char**)NULL, 10);
7128 floatit = TRUE; /* Probably just too large. */
7129 else if (*PL_tokenbuf == '-')
7131 else if (uv <= IV_MAX)
7132 sv_setiv(sv, uv); /* Prefer IVs over UVs. */
7137 nv = Atof(PL_tokenbuf);
7142 No working strtou?ll?.
7144 Unfortunately atol() doesn't do range checks (returning
7145 LONG_MIN/LONG_MAX, and setting errno to ERANGE on overflows)
7146 everywhere [1], so we cannot use use atol() (or atoll()).
7147 If we could, they would be used, as Atol(), very much like
7148 Strtol() and Strtoul() are used above.
7150 [1] XXX Configure test needed to check for atol()
7151 (and atoll()) overflow behaviour XXX
7155 We need to do this the hard way. */
7157 nv = Atof(PL_tokenbuf);
7159 /* See if we can make do with an integer value without loss of
7160 precision. We use U_V to cast to a UV, because some
7161 compilers have issues. Then we try casting it back and see
7162 if it was the same [1]. We only do this if we know we
7163 specifically read an integer. If floatit is true, then we
7164 don't need to do the conversion at all.
7166 [1] Note that this is lossy if our NVs cannot preserve our
7167 UVs. There are metaconfig defines NV_PRESERVES_UV (a boolean)
7168 and NV_PRESERVES_UV_BITS (a number), but in general we really
7169 do hope all such potentially lossy platforms have strtou?ll?
7170 to do a lossless IV/UV conversion.
7172 Maybe could do some tricks with DBL_DIG, LDBL_DIG and
7173 DBL_MANT_DIG and LDBL_MANT_DIG (these are already available
7174 as NV_DIG and NV_MANT_DIG)?
7180 if (!floatit && (NV)uv == nv) {
7182 sv_setiv(sv, uv); /* Prefer IVs over UVs. */
7190 if ( floatit ? (PL_hints & HINT_NEW_FLOAT) :
7191 (PL_hints & HINT_NEW_INTEGER) )
7192 sv = new_constant(PL_tokenbuf, d - PL_tokenbuf,
7193 (floatit ? "float" : "integer"),
7197 /* if it starts with a v, it could be a v-string */
7203 while (isDIGIT(*pos) || *pos == '_')
7205 if (!isALPHA(*pos)) {
7207 U8 tmpbuf[UTF8_MAXLEN+1];
7210 s++; /* get past 'v' */
7213 sv_setpvn(sv, "", 0);
7216 if (*s == '0' && isDIGIT(s[1]))
7217 yyerror("Octal number in vector unsupported");
7220 /* this is atoi() that tolerates underscores */
7223 while (--end >= s) {
7228 rev += (*end - '0') * mult;
7230 if (orev > rev && ckWARN_d(WARN_OVERFLOW))
7231 Perl_warner(aTHX_ WARN_OVERFLOW,
7232 "Integer overflow in decimal number");
7235 tmpend = uv_to_utf8(tmpbuf, rev);
7236 utf8 = utf8 || rev > 127;
7237 sv_catpvn(sv, (const char*)tmpbuf, tmpend - tmpbuf);
7238 if (*pos == '.' && isDIGIT(pos[1]))
7244 while (isDIGIT(*pos) || *pos == '_')
7253 sv_utf8_downgrade(sv, TRUE);
7260 /* make the op for the constant and return */
7263 lvalp->opval = newSVOP(OP_CONST, 0, sv);
7265 lvalp->opval = Nullop;
7271 S_scan_formline(pTHX_ register char *s)
7275 SV *stuff = newSVpvn("",0);
7276 bool needargs = FALSE;
7279 if (*s == '.' || *s == /*{*/'}') {
7281 #ifdef PERL_STRICT_CR
7282 for (t = s+1;SPACE_OR_TAB(*t); t++) ;
7284 for (t = s+1;SPACE_OR_TAB(*t) || *t == '\r'; t++) ;
7286 if (*t == '\n' || t == PL_bufend)
7289 if (PL_in_eval && !PL_rsfp) {
7290 eol = strchr(s,'\n');
7295 eol = PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
7297 for (t = s; t < eol; t++) {
7298 if (*t == '~' && t[1] == '~' && SvCUR(stuff)) {
7300 goto enough; /* ~~ must be first line in formline */
7302 if (*t == '@' || *t == '^')
7305 sv_catpvn(stuff, s, eol-s);
7306 #ifndef PERL_STRICT_CR
7307 if (eol-s > 1 && eol[-2] == '\r' && eol[-1] == '\n') {
7308 char *end = SvPVX(stuff) + SvCUR(stuff);
7317 s = filter_gets(PL_linestr, PL_rsfp, 0);
7318 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = SvPVX(PL_linestr);
7319 PL_bufend = PL_bufptr + SvCUR(PL_linestr);
7322 yyerror("Format not terminated");
7332 PL_lex_state = LEX_NORMAL;
7333 PL_nextval[PL_nexttoke].ival = 0;
7337 PL_lex_state = LEX_FORMLINE;
7338 PL_nextval[PL_nexttoke].opval = (OP*)newSVOP(OP_CONST, 0, stuff);
7340 PL_nextval[PL_nexttoke].ival = OP_FORMLINE;
7344 SvREFCNT_dec(stuff);
7345 PL_lex_formbrack = 0;
7356 PL_cshlen = strlen(PL_cshname);
7361 Perl_start_subparse(pTHX_ I32 is_format, U32 flags)
7363 I32 oldsavestack_ix = PL_savestack_ix;
7364 CV* outsidecv = PL_compcv;
7368 assert(SvTYPE(PL_compcv) == SVt_PVCV);
7370 SAVEI32(PL_subline);
7371 save_item(PL_subname);
7374 SAVESPTR(PL_comppad_name);
7375 SAVESPTR(PL_compcv);
7376 SAVEI32(PL_comppad_name_fill);
7377 SAVEI32(PL_min_intro_pending);
7378 SAVEI32(PL_max_intro_pending);
7379 SAVEI32(PL_pad_reset_pending);
7381 PL_compcv = (CV*)NEWSV(1104,0);
7382 sv_upgrade((SV *)PL_compcv, is_format ? SVt_PVFM : SVt_PVCV);
7383 CvFLAGS(PL_compcv) |= flags;
7385 PL_comppad = newAV();
7386 av_push(PL_comppad, Nullsv);
7387 PL_curpad = AvARRAY(PL_comppad);
7388 PL_comppad_name = newAV();
7389 PL_comppad_name_fill = 0;
7390 PL_min_intro_pending = 0;
7392 PL_subline = CopLINE(PL_curcop);
7394 av_store(PL_comppad_name, 0, newSVpvn("@_", 2));
7395 PL_curpad[0] = (SV*)newAV();
7396 SvPADMY_on(PL_curpad[0]); /* XXX Needed? */
7397 #endif /* USE_THREADS */
7399 comppadlist = newAV();
7400 AvREAL_off(comppadlist);
7401 av_store(comppadlist, 0, (SV*)PL_comppad_name);
7402 av_store(comppadlist, 1, (SV*)PL_comppad);
7404 CvPADLIST(PL_compcv) = comppadlist;
7405 CvOUTSIDE(PL_compcv) = (CV*)SvREFCNT_inc(outsidecv);
7407 CvOWNER(PL_compcv) = 0;
7408 New(666, CvMUTEXP(PL_compcv), 1, perl_mutex);
7409 MUTEX_INIT(CvMUTEXP(PL_compcv));
7410 #endif /* USE_THREADS */
7412 return oldsavestack_ix;
7416 Perl_yywarn(pTHX_ char *s)
7418 PL_in_eval |= EVAL_WARNONLY;
7420 PL_in_eval &= ~EVAL_WARNONLY;
7425 Perl_yyerror(pTHX_ char *s)
7428 char *context = NULL;
7432 if (!yychar || (yychar == ';' && !PL_rsfp))
7434 else if (PL_bufptr > PL_oldoldbufptr && PL_bufptr - PL_oldoldbufptr < 200 &&
7435 PL_oldoldbufptr != PL_oldbufptr && PL_oldbufptr != PL_bufptr) {
7436 while (isSPACE(*PL_oldoldbufptr))
7438 context = PL_oldoldbufptr;
7439 contlen = PL_bufptr - PL_oldoldbufptr;
7441 else if (PL_bufptr > PL_oldbufptr && PL_bufptr - PL_oldbufptr < 200 &&
7442 PL_oldbufptr != PL_bufptr) {
7443 while (isSPACE(*PL_oldbufptr))
7445 context = PL_oldbufptr;
7446 contlen = PL_bufptr - PL_oldbufptr;
7448 else if (yychar > 255)
7449 where = "next token ???";
7450 #ifdef USE_PURE_BISON
7451 /* GNU Bison sets the value -2 */
7452 else if (yychar == -2) {
7454 else if ((yychar & 127) == 127) {
7456 if (PL_lex_state == LEX_NORMAL ||
7457 (PL_lex_state == LEX_KNOWNEXT && PL_lex_defer == LEX_NORMAL))
7458 where = "at end of line";
7459 else if (PL_lex_inpat)
7460 where = "within pattern";
7462 where = "within string";
7465 SV *where_sv = sv_2mortal(newSVpvn("next char ", 10));
7467 Perl_sv_catpvf(aTHX_ where_sv, "^%c", toCTRL(yychar));
7468 else if (isPRINT_LC(yychar))
7469 Perl_sv_catpvf(aTHX_ where_sv, "%c", yychar);
7471 Perl_sv_catpvf(aTHX_ where_sv, "\\%03o", yychar & 255);
7472 where = SvPVX(where_sv);
7474 msg = sv_2mortal(newSVpv(s, 0));
7475 Perl_sv_catpvf(aTHX_ msg, " at %s line %"IVdf", ",
7476 CopFILE(PL_curcop), (IV)CopLINE(PL_curcop));
7478 Perl_sv_catpvf(aTHX_ msg, "near \"%.*s\"\n", contlen, context);
7480 Perl_sv_catpvf(aTHX_ msg, "%s\n", where);
7481 if (PL_multi_start < PL_multi_end && (U32)(CopLINE(PL_curcop) - PL_multi_end) <= 1) {
7482 Perl_sv_catpvf(aTHX_ msg,
7483 " (Might be a runaway multi-line %c%c string starting on line %"IVdf")\n",
7484 (int)PL_multi_open,(int)PL_multi_close,(IV)PL_multi_start);
7487 if (PL_in_eval & EVAL_WARNONLY)
7488 Perl_warn(aTHX_ "%"SVf, msg);
7491 if (PL_error_count >= 10) {
7492 if (PL_in_eval && SvCUR(ERRSV))
7493 Perl_croak(aTHX_ "%"SVf"%s has too many errors.\n",
7494 ERRSV, CopFILE(PL_curcop));
7496 Perl_croak(aTHX_ "%s has too many errors.\n",
7497 CopFILE(PL_curcop));
7500 PL_in_my_stash = Nullhv;
7505 S_swallow_bom(pTHX_ U8 *s)
7508 slen = SvCUR(PL_linestr);
7512 /* UTF-16 little-endian */
7513 if (s[2] == 0 && s[3] == 0) /* UTF-32 little-endian */
7514 Perl_croak(aTHX_ "Unsupported script encoding");
7515 #ifndef PERL_NO_UTF16_FILTER
7516 DEBUG_p(PerlIO_printf(Perl_debug_log, "UTF-LE script encoding\n"));
7518 if (PL_bufend > (char*)s) {
7522 filter_add(utf16rev_textfilter, NULL);
7523 New(898, news, (PL_bufend - (char*)s) * 3 / 2 + 1, U8);
7524 PL_bufend = (char*)utf16_to_utf8_reversed(s, news,
7525 PL_bufend - (char*)s - 1,
7527 Copy(news, s, newlen, U8);
7528 SvCUR_set(PL_linestr, newlen);
7529 PL_bufend = SvPVX(PL_linestr) + newlen;
7530 news[newlen++] = '\0';
7534 Perl_croak(aTHX_ "Unsupported script encoding");
7539 if (s[1] == 0xFF) { /* UTF-16 big-endian */
7540 #ifndef PERL_NO_UTF16_FILTER
7541 DEBUG_p(PerlIO_printf(Perl_debug_log, "UTF-16BE script encoding\n"));
7543 if (PL_bufend > (char *)s) {
7547 filter_add(utf16_textfilter, NULL);
7548 New(898, news, (PL_bufend - (char*)s) * 3 / 2 + 1, U8);
7549 PL_bufend = (char*)utf16_to_utf8(s, news,
7550 PL_bufend - (char*)s,
7552 Copy(news, s, newlen, U8);
7553 SvCUR_set(PL_linestr, newlen);
7554 PL_bufend = SvPVX(PL_linestr) + newlen;
7555 news[newlen++] = '\0';
7559 Perl_croak(aTHX_ "Unsupported script encoding");
7564 if (slen > 2 && s[1] == 0xBB && s[2] == 0xBF) {
7565 DEBUG_p(PerlIO_printf(Perl_debug_log, "UTF-8 script encoding\n"));
7570 if (slen > 3 && s[1] == 0 && /* UTF-32 big-endian */
7571 s[2] == 0xFE && s[3] == 0xFF)
7573 Perl_croak(aTHX_ "Unsupported script encoding");
7585 * Restore a source filter.
7589 restore_rsfp(pTHXo_ void *f)
7591 PerlIO *fp = (PerlIO*)f;
7593 if (PL_rsfp == PerlIO_stdin())
7594 PerlIO_clearerr(PL_rsfp);
7595 else if (PL_rsfp && (PL_rsfp != fp))
7596 PerlIO_close(PL_rsfp);
7600 #ifndef PERL_NO_UTF16_FILTER
7602 utf16_textfilter(pTHXo_ int idx, SV *sv, int maxlen)
7604 I32 count = FILTER_READ(idx+1, sv, maxlen);
7609 New(898, tmps, SvCUR(sv) * 3 / 2 + 1, U8);
7610 if (!*SvPV_nolen(sv))
7611 /* Game over, but don't feed an odd-length string to utf16_to_utf8 */
7614 tend = utf16_to_utf8((U8*)SvPVX(sv), tmps, SvCUR(sv), &newlen);
7615 sv_usepvn(sv, (char*)tmps, tend - tmps);
7621 utf16rev_textfilter(pTHXo_ int idx, SV *sv, int maxlen)
7623 I32 count = FILTER_READ(idx+1, sv, maxlen);
7628 if (!*SvPV_nolen(sv))
7629 /* Game over, but don't feed an odd-length string to utf16_to_utf8 */
7632 New(898, tmps, SvCUR(sv) * 3 / 2 + 1, U8);
7633 tend = utf16_to_utf8_reversed((U8*)SvPVX(sv), tmps, SvCUR(sv), &newlen);
7634 sv_usepvn(sv, (char*)tmps, tend - tmps);