3 * Copyright (C) 1991, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999, 2000,
4 * 2001, 2002, 2003, 2004, 2005, 2006, 2007, 2008 by Larry Wall and others
6 * You may distribute under the terms of either the GNU General Public
7 * License or the Artistic License, as specified in the README file.
12 * 'It all comes from here, the stench and the peril.' --Frodo
14 * [p.719 of _The Lord of the Rings_, IV/ix: "Shelob's Lair"]
18 * This file is the lexer for Perl. It's closely linked to the
21 * The main routine is yylex(), which returns the next token.
25 #define PERL_IN_TOKE_C
28 #define new_constant(a,b,c,d,e,f,g) \
29 S_new_constant(aTHX_ a,b,STR_WITH_LEN(c),d,e,f, g)
31 #define pl_yylval (PL_parser->yylval)
33 /* YYINITDEPTH -- initial size of the parser's stacks. */
34 #define YYINITDEPTH 200
36 /* XXX temporary backwards compatibility */
37 #define PL_lex_brackets (PL_parser->lex_brackets)
38 #define PL_lex_brackstack (PL_parser->lex_brackstack)
39 #define PL_lex_casemods (PL_parser->lex_casemods)
40 #define PL_lex_casestack (PL_parser->lex_casestack)
41 #define PL_lex_defer (PL_parser->lex_defer)
42 #define PL_lex_dojoin (PL_parser->lex_dojoin)
43 #define PL_lex_expect (PL_parser->lex_expect)
44 #define PL_lex_formbrack (PL_parser->lex_formbrack)
45 #define PL_lex_inpat (PL_parser->lex_inpat)
46 #define PL_lex_inwhat (PL_parser->lex_inwhat)
47 #define PL_lex_op (PL_parser->lex_op)
48 #define PL_lex_repl (PL_parser->lex_repl)
49 #define PL_lex_starts (PL_parser->lex_starts)
50 #define PL_lex_stuff (PL_parser->lex_stuff)
51 #define PL_multi_start (PL_parser->multi_start)
52 #define PL_multi_open (PL_parser->multi_open)
53 #define PL_multi_close (PL_parser->multi_close)
54 #define PL_pending_ident (PL_parser->pending_ident)
55 #define PL_preambled (PL_parser->preambled)
56 #define PL_sublex_info (PL_parser->sublex_info)
57 #define PL_linestr (PL_parser->linestr)
58 #define PL_expect (PL_parser->expect)
59 #define PL_copline (PL_parser->copline)
60 #define PL_bufptr (PL_parser->bufptr)
61 #define PL_oldbufptr (PL_parser->oldbufptr)
62 #define PL_oldoldbufptr (PL_parser->oldoldbufptr)
63 #define PL_linestart (PL_parser->linestart)
64 #define PL_bufend (PL_parser->bufend)
65 #define PL_last_uni (PL_parser->last_uni)
66 #define PL_last_lop (PL_parser->last_lop)
67 #define PL_last_lop_op (PL_parser->last_lop_op)
68 #define PL_lex_state (PL_parser->lex_state)
69 #define PL_rsfp (PL_parser->rsfp)
70 #define PL_rsfp_filters (PL_parser->rsfp_filters)
71 #define PL_in_my (PL_parser->in_my)
72 #define PL_in_my_stash (PL_parser->in_my_stash)
73 #define PL_tokenbuf (PL_parser->tokenbuf)
74 #define PL_multi_end (PL_parser->multi_end)
75 #define PL_error_count (PL_parser->error_count)
78 # define PL_endwhite (PL_parser->endwhite)
79 # define PL_faketokens (PL_parser->faketokens)
80 # define PL_lasttoke (PL_parser->lasttoke)
81 # define PL_nextwhite (PL_parser->nextwhite)
82 # define PL_realtokenstart (PL_parser->realtokenstart)
83 # define PL_skipwhite (PL_parser->skipwhite)
84 # define PL_thisclose (PL_parser->thisclose)
85 # define PL_thismad (PL_parser->thismad)
86 # define PL_thisopen (PL_parser->thisopen)
87 # define PL_thisstuff (PL_parser->thisstuff)
88 # define PL_thistoken (PL_parser->thistoken)
89 # define PL_thiswhite (PL_parser->thiswhite)
90 # define PL_thiswhite (PL_parser->thiswhite)
91 # define PL_nexttoke (PL_parser->nexttoke)
92 # define PL_curforce (PL_parser->curforce)
94 # define PL_nexttoke (PL_parser->nexttoke)
95 # define PL_nexttype (PL_parser->nexttype)
96 # define PL_nextval (PL_parser->nextval)
100 S_pending_ident(pTHX);
102 static const char ident_too_long[] = "Identifier too long";
103 static const char commaless_variable_list[] = "comma-less variable list";
105 #ifndef PERL_NO_UTF16_FILTER
106 static I32 utf16_textfilter(pTHX_ int idx, SV *sv, int maxlen);
107 static I32 utf16rev_textfilter(pTHX_ int idx, SV *sv, int maxlen);
111 # define CURMAD(slot,sv) if (PL_madskills) { curmad(slot,sv); sv = 0; }
112 # define NEXTVAL_NEXTTOKE PL_nexttoke[PL_curforce].next_val
114 # define CURMAD(slot,sv)
115 # define NEXTVAL_NEXTTOKE PL_nextval[PL_nexttoke]
118 #define XFAKEBRACK 128
119 #define XENUMMASK 127
121 #ifdef USE_UTF8_SCRIPTS
122 # define UTF (!IN_BYTES)
124 # define UTF ((PL_linestr && DO_UTF8(PL_linestr)) || (PL_hints & HINT_UTF8))
127 /* The maximum number of characters preceding the unrecognized one to display */
128 #define UNRECOGNIZED_PRECEDE_COUNT 10
130 /* In variables named $^X, these are the legal values for X.
131 * 1999-02-27 mjd-perl-patch@plover.com */
132 #define isCONTROLVAR(x) (isUPPER(x) || strchr("[\\]^_?", (x)))
134 #define SPACE_OR_TAB(c) ((c)==' '||(c)=='\t')
136 /* LEX_* are values for PL_lex_state, the state of the lexer.
137 * They are arranged oddly so that the guard on the switch statement
138 * can get by with a single comparison (if the compiler is smart enough).
141 /* #define LEX_NOTPARSING 11 is done in perl.h. */
143 #define LEX_NORMAL 10 /* normal code (ie not within "...") */
144 #define LEX_INTERPNORMAL 9 /* code within a string, eg "$foo[$x+1]" */
145 #define LEX_INTERPCASEMOD 8 /* expecting a \U, \Q or \E etc */
146 #define LEX_INTERPPUSH 7 /* starting a new sublex parse level */
147 #define LEX_INTERPSTART 6 /* expecting the start of a $var */
149 /* at end of code, eg "$x" followed by: */
150 #define LEX_INTERPEND 5 /* ... eg not one of [, { or -> */
151 #define LEX_INTERPENDMAYBE 4 /* ... eg one of [, { or -> */
153 #define LEX_INTERPCONCAT 3 /* expecting anything, eg at start of
154 string or after \E, $foo, etc */
155 #define LEX_INTERPCONST 2 /* NOT USED */
156 #define LEX_FORMLINE 1 /* expecting a format line */
157 #define LEX_KNOWNEXT 0 /* next token known; just return it */
161 static const char* const lex_state_names[] = {
180 #include "keywords.h"
182 /* CLINE is a macro that ensures PL_copline has a sane value */
187 #define CLINE (PL_copline = (CopLINE(PL_curcop) < PL_copline ? CopLINE(PL_curcop) : PL_copline))
190 # define SKIPSPACE0(s) skipspace0(s)
191 # define SKIPSPACE1(s) skipspace1(s)
192 # define SKIPSPACE2(s,tsv) skipspace2(s,&tsv)
193 # define PEEKSPACE(s) skipspace2(s,0)
195 # define SKIPSPACE0(s) skipspace(s)
196 # define SKIPSPACE1(s) skipspace(s)
197 # define SKIPSPACE2(s,tsv) skipspace(s)
198 # define PEEKSPACE(s) skipspace(s)
202 * Convenience functions to return different tokens and prime the
203 * lexer for the next token. They all take an argument.
205 * TOKEN : generic token (used for '(', DOLSHARP, etc)
206 * OPERATOR : generic operator
207 * AOPERATOR : assignment operator
208 * PREBLOCK : beginning the block after an if, while, foreach, ...
209 * PRETERMBLOCK : beginning a non-code-defining {} block (eg, hash ref)
210 * PREREF : *EXPR where EXPR is not a simple identifier
211 * TERM : expression term
212 * LOOPX : loop exiting command (goto, last, dump, etc)
213 * FTST : file test operator
214 * FUN0 : zero-argument function
215 * FUN1 : not used, except for not, which isn't a UNIOP
216 * BOop : bitwise or or xor
218 * SHop : shift operator
219 * PWop : power operator
220 * PMop : pattern-matching operator
221 * Aop : addition-level operator
222 * Mop : multiplication-level operator
223 * Eop : equality-testing operator
224 * Rop : relational operator <= != gt
226 * Also see LOP and lop() below.
229 #ifdef DEBUGGING /* Serve -DT. */
230 # define REPORT(retval) tokereport((I32)retval, &pl_yylval)
232 # define REPORT(retval) (retval)
235 #define TOKEN(retval) return ( PL_bufptr = s, REPORT(retval))
236 #define OPERATOR(retval) return (PL_expect = XTERM, PL_bufptr = s, REPORT(retval))
237 #define AOPERATOR(retval) return ao((PL_expect = XTERM, PL_bufptr = s, REPORT(retval)))
238 #define PREBLOCK(retval) return (PL_expect = XBLOCK,PL_bufptr = s, REPORT(retval))
239 #define PRETERMBLOCK(retval) return (PL_expect = XTERMBLOCK,PL_bufptr = s, REPORT(retval))
240 #define PREREF(retval) return (PL_expect = XREF,PL_bufptr = s, REPORT(retval))
241 #define TERM(retval) return (CLINE, PL_expect = XOPERATOR, PL_bufptr = s, REPORT(retval))
242 #define LOOPX(f) return (pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)LOOPEX))
243 #define FTST(f) return (pl_yylval.ival=f, PL_expect=XTERMORDORDOR, PL_bufptr=s, REPORT((int)UNIOP))
244 #define FUN0(f) return (pl_yylval.ival=f, PL_expect=XOPERATOR, PL_bufptr=s, REPORT((int)FUNC0))
245 #define FUN1(f) return (pl_yylval.ival=f, PL_expect=XOPERATOR, PL_bufptr=s, REPORT((int)FUNC1))
246 #define BOop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)BITOROP)))
247 #define BAop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)BITANDOP)))
248 #define SHop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)SHIFTOP)))
249 #define PWop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)POWOP)))
250 #define PMop(f) return(pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)MATCHOP))
251 #define Aop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)ADDOP)))
252 #define Mop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)MULOP)))
253 #define Eop(f) return (pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)EQOP))
254 #define Rop(f) return (pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)RELOP))
256 /* This bit of chicanery makes a unary function followed by
257 * a parenthesis into a function with one argument, highest precedence.
258 * The UNIDOR macro is for unary functions that can be followed by the //
259 * operator (such as C<shift // 0>).
261 #define UNI2(f,x) { \
262 pl_yylval.ival = f; \
265 PL_last_uni = PL_oldbufptr; \
266 PL_last_lop_op = f; \
268 return REPORT( (int)FUNC1 ); \
270 return REPORT( *s=='(' ? (int)FUNC1 : (int)UNIOP ); \
272 #define UNI(f) UNI2(f,XTERM)
273 #define UNIDOR(f) UNI2(f,XTERMORDORDOR)
275 #define UNIBRACK(f) { \
276 pl_yylval.ival = f; \
278 PL_last_uni = PL_oldbufptr; \
280 return REPORT( (int)FUNC1 ); \
282 return REPORT( (*s == '(') ? (int)FUNC1 : (int)UNIOP ); \
285 /* grandfather return to old style */
286 #define OLDLOP(f) return(pl_yylval.ival=f,PL_expect = XTERM,PL_bufptr = s,(int)LSTOP)
290 /* how to interpret the pl_yylval associated with the token */
294 TOKENTYPE_OPNUM, /* pl_yylval.ival contains an opcode number */
300 static struct debug_tokens {
302 enum token_type type;
304 } const debug_tokens[] =
306 { ADDOP, TOKENTYPE_OPNUM, "ADDOP" },
307 { ANDAND, TOKENTYPE_NONE, "ANDAND" },
308 { ANDOP, TOKENTYPE_NONE, "ANDOP" },
309 { ANONSUB, TOKENTYPE_IVAL, "ANONSUB" },
310 { ARROW, TOKENTYPE_NONE, "ARROW" },
311 { ASSIGNOP, TOKENTYPE_OPNUM, "ASSIGNOP" },
312 { BITANDOP, TOKENTYPE_OPNUM, "BITANDOP" },
313 { BITOROP, TOKENTYPE_OPNUM, "BITOROP" },
314 { COLONATTR, TOKENTYPE_NONE, "COLONATTR" },
315 { CONTINUE, TOKENTYPE_NONE, "CONTINUE" },
316 { DEFAULT, TOKENTYPE_NONE, "DEFAULT" },
317 { DO, TOKENTYPE_NONE, "DO" },
318 { DOLSHARP, TOKENTYPE_NONE, "DOLSHARP" },
319 { DORDOR, TOKENTYPE_NONE, "DORDOR" },
320 { DOROP, TOKENTYPE_OPNUM, "DOROP" },
321 { DOTDOT, TOKENTYPE_IVAL, "DOTDOT" },
322 { ELSE, TOKENTYPE_NONE, "ELSE" },
323 { ELSIF, TOKENTYPE_IVAL, "ELSIF" },
324 { EQOP, TOKENTYPE_OPNUM, "EQOP" },
325 { FOR, TOKENTYPE_IVAL, "FOR" },
326 { FORMAT, TOKENTYPE_NONE, "FORMAT" },
327 { FUNC, TOKENTYPE_OPNUM, "FUNC" },
328 { FUNC0, TOKENTYPE_OPNUM, "FUNC0" },
329 { FUNC0SUB, TOKENTYPE_OPVAL, "FUNC0SUB" },
330 { FUNC1, TOKENTYPE_OPNUM, "FUNC1" },
331 { FUNCMETH, TOKENTYPE_OPVAL, "FUNCMETH" },
332 { GIVEN, TOKENTYPE_IVAL, "GIVEN" },
333 { HASHBRACK, TOKENTYPE_NONE, "HASHBRACK" },
334 { IF, TOKENTYPE_IVAL, "IF" },
335 { LABEL, TOKENTYPE_PVAL, "LABEL" },
336 { LOCAL, TOKENTYPE_IVAL, "LOCAL" },
337 { LOOPEX, TOKENTYPE_OPNUM, "LOOPEX" },
338 { LSTOP, TOKENTYPE_OPNUM, "LSTOP" },
339 { LSTOPSUB, TOKENTYPE_OPVAL, "LSTOPSUB" },
340 { MATCHOP, TOKENTYPE_OPNUM, "MATCHOP" },
341 { METHOD, TOKENTYPE_OPVAL, "METHOD" },
342 { MULOP, TOKENTYPE_OPNUM, "MULOP" },
343 { MY, TOKENTYPE_IVAL, "MY" },
344 { MYSUB, TOKENTYPE_NONE, "MYSUB" },
345 { NOAMP, TOKENTYPE_NONE, "NOAMP" },
346 { NOTOP, TOKENTYPE_NONE, "NOTOP" },
347 { OROP, TOKENTYPE_IVAL, "OROP" },
348 { OROR, TOKENTYPE_NONE, "OROR" },
349 { PACKAGE, TOKENTYPE_NONE, "PACKAGE" },
350 { PMFUNC, TOKENTYPE_OPVAL, "PMFUNC" },
351 { POSTDEC, TOKENTYPE_NONE, "POSTDEC" },
352 { POSTINC, TOKENTYPE_NONE, "POSTINC" },
353 { POWOP, TOKENTYPE_OPNUM, "POWOP" },
354 { PREDEC, TOKENTYPE_NONE, "PREDEC" },
355 { PREINC, TOKENTYPE_NONE, "PREINC" },
356 { PRIVATEREF, TOKENTYPE_OPVAL, "PRIVATEREF" },
357 { REFGEN, TOKENTYPE_NONE, "REFGEN" },
358 { RELOP, TOKENTYPE_OPNUM, "RELOP" },
359 { SHIFTOP, TOKENTYPE_OPNUM, "SHIFTOP" },
360 { SUB, TOKENTYPE_NONE, "SUB" },
361 { THING, TOKENTYPE_OPVAL, "THING" },
362 { UMINUS, TOKENTYPE_NONE, "UMINUS" },
363 { UNIOP, TOKENTYPE_OPNUM, "UNIOP" },
364 { UNIOPSUB, TOKENTYPE_OPVAL, "UNIOPSUB" },
365 { UNLESS, TOKENTYPE_IVAL, "UNLESS" },
366 { UNTIL, TOKENTYPE_IVAL, "UNTIL" },
367 { USE, TOKENTYPE_IVAL, "USE" },
368 { WHEN, TOKENTYPE_IVAL, "WHEN" },
369 { WHILE, TOKENTYPE_IVAL, "WHILE" },
370 { WORD, TOKENTYPE_OPVAL, "WORD" },
371 { YADAYADA, TOKENTYPE_IVAL, "YADAYADA" },
372 { 0, TOKENTYPE_NONE, NULL }
375 /* dump the returned token in rv, plus any optional arg in pl_yylval */
378 S_tokereport(pTHX_ I32 rv, const YYSTYPE* lvalp)
382 PERL_ARGS_ASSERT_TOKEREPORT;
385 const char *name = NULL;
386 enum token_type type = TOKENTYPE_NONE;
387 const struct debug_tokens *p;
388 SV* const report = newSVpvs("<== ");
390 for (p = debug_tokens; p->token; p++) {
391 if (p->token == (int)rv) {
398 Perl_sv_catpv(aTHX_ report, name);
399 else if ((char)rv > ' ' && (char)rv < '~')
400 Perl_sv_catpvf(aTHX_ report, "'%c'", (char)rv);
402 sv_catpvs(report, "EOF");
404 Perl_sv_catpvf(aTHX_ report, "?? %"IVdf, (IV)rv);
407 case TOKENTYPE_GVVAL: /* doesn't appear to be used */
410 Perl_sv_catpvf(aTHX_ report, "(ival=%"IVdf")", (IV)lvalp->ival);
412 case TOKENTYPE_OPNUM:
413 Perl_sv_catpvf(aTHX_ report, "(ival=op_%s)",
414 PL_op_name[lvalp->ival]);
417 Perl_sv_catpvf(aTHX_ report, "(pval=\"%s\")", lvalp->pval);
419 case TOKENTYPE_OPVAL:
421 Perl_sv_catpvf(aTHX_ report, "(opval=op_%s)",
422 PL_op_name[lvalp->opval->op_type]);
423 if (lvalp->opval->op_type == OP_CONST) {
424 Perl_sv_catpvf(aTHX_ report, " %s",
425 SvPEEK(cSVOPx_sv(lvalp->opval)));
430 sv_catpvs(report, "(opval=null)");
433 PerlIO_printf(Perl_debug_log, "### %s\n\n", SvPV_nolen_const(report));
439 /* print the buffer with suitable escapes */
442 S_printbuf(pTHX_ const char *const fmt, const char *const s)
444 SV* const tmp = newSVpvs("");
446 PERL_ARGS_ASSERT_PRINTBUF;
448 PerlIO_printf(Perl_debug_log, fmt, pv_display(tmp, s, strlen(s), 0, 60));
457 * This subroutine detects &&=, ||=, and //= and turns an ANDAND, OROR or DORDOR
458 * into an OP_ANDASSIGN, OP_ORASSIGN, or OP_DORASSIGN
462 S_ao(pTHX_ int toketype)
465 if (*PL_bufptr == '=') {
467 if (toketype == ANDAND)
468 pl_yylval.ival = OP_ANDASSIGN;
469 else if (toketype == OROR)
470 pl_yylval.ival = OP_ORASSIGN;
471 else if (toketype == DORDOR)
472 pl_yylval.ival = OP_DORASSIGN;
480 * When Perl expects an operator and finds something else, no_op
481 * prints the warning. It always prints "<something> found where
482 * operator expected. It prints "Missing semicolon on previous line?"
483 * if the surprise occurs at the start of the line. "do you need to
484 * predeclare ..." is printed out for code like "sub bar; foo bar $x"
485 * where the compiler doesn't know if foo is a method call or a function.
486 * It prints "Missing operator before end of line" if there's nothing
487 * after the missing operator, or "... before <...>" if there is something
488 * after the missing operator.
492 S_no_op(pTHX_ const char *const what, char *s)
495 char * const oldbp = PL_bufptr;
496 const bool is_first = (PL_oldbufptr == PL_linestart);
498 PERL_ARGS_ASSERT_NO_OP;
504 yywarn(Perl_form(aTHX_ "%s found where operator expected", what));
505 if (ckWARN_d(WARN_SYNTAX)) {
507 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
508 "\t(Missing semicolon on previous line?)\n");
509 else if (PL_oldoldbufptr && isIDFIRST_lazy_if(PL_oldoldbufptr,UTF)) {
511 for (t = PL_oldoldbufptr; (isALNUM_lazy_if(t,UTF) || *t == ':'); t++)
513 if (t < PL_bufptr && isSPACE(*t))
514 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
515 "\t(Do you need to predeclare %.*s?)\n",
516 (int)(t - PL_oldoldbufptr), PL_oldoldbufptr);
520 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
521 "\t(Missing operator before %.*s?)\n", (int)(s - oldbp), oldbp);
529 * Complain about missing quote/regexp/heredoc terminator.
530 * If it's called with NULL then it cauterizes the line buffer.
531 * If we're in a delimited string and the delimiter is a control
532 * character, it's reformatted into a two-char sequence like ^C.
537 S_missingterm(pTHX_ char *s)
543 char * const nl = strrchr(s,'\n');
547 else if (isCNTRL(PL_multi_close)) {
549 tmpbuf[1] = (char)toCTRL(PL_multi_close);
554 *tmpbuf = (char)PL_multi_close;
558 q = strchr(s,'"') ? '\'' : '"';
559 Perl_croak(aTHX_ "Can't find string terminator %c%s%c anywhere before EOF",q,s,q);
562 #define FEATURE_IS_ENABLED(name) \
563 ((0 != (PL_hints & HINT_LOCALIZE_HH)) \
564 && S_feature_is_enabled(aTHX_ STR_WITH_LEN(name)))
565 /* The longest string we pass in. */
566 #define MAX_FEATURE_LEN (sizeof("switch")-1)
569 * S_feature_is_enabled
570 * Check whether the named feature is enabled.
573 S_feature_is_enabled(pTHX_ const char *const name, STRLEN namelen)
576 HV * const hinthv = GvHV(PL_hintgv);
577 char he_name[8 + MAX_FEATURE_LEN] = "feature_";
579 PERL_ARGS_ASSERT_FEATURE_IS_ENABLED;
581 assert(namelen <= MAX_FEATURE_LEN);
582 memcpy(&he_name[8], name, namelen);
584 return (hinthv && hv_exists(hinthv, he_name, 8 + namelen));
592 Perl_deprecate(pTHX_ const char *const s)
594 PERL_ARGS_ASSERT_DEPRECATE;
596 Perl_ck_warner(aTHX_ packWARN(WARN_DEPRECATED), "Use of %s is deprecated", s);
600 S_deprecate_old(pTHX_ const char *const s)
602 /* This function should NOT be called for any new deprecated warnings */
603 /* Use Perl_deprecate instead */
605 /* It is here to maintain backward compatibility with the pre-5.8 */
606 /* warnings category hierarchy. The "deprecated" category used to */
607 /* live under the "syntax" category. It is now a top-level category */
608 /* in its own right. */
610 PERL_ARGS_ASSERT_DEPRECATE_OLD;
612 Perl_ck_warner(aTHX_ packWARN2(WARN_DEPRECATED, WARN_SYNTAX),
613 "Use of %s is deprecated", s);
617 * experimental text filters for win32 carriage-returns, utf16-to-utf8 and
618 * utf16-to-utf8-reversed.
621 #ifdef PERL_CR_FILTER
625 register const char *s = SvPVX_const(sv);
626 register const char * const e = s + SvCUR(sv);
628 PERL_ARGS_ASSERT_STRIP_RETURN;
630 /* outer loop optimized to do nothing if there are no CR-LFs */
632 if (*s++ == '\r' && *s == '\n') {
633 /* hit a CR-LF, need to copy the rest */
634 register char *d = s - 1;
637 if (*s == '\r' && s[1] == '\n')
648 S_cr_textfilter(pTHX_ int idx, SV *sv, int maxlen)
650 const I32 count = FILTER_READ(idx+1, sv, maxlen);
651 if (count > 0 && !maxlen)
662 * Create a parser object and initialise its parser and lexer fields
664 * rsfp is the opened file handle to read from (if any),
666 * line holds any initial content already read from the file (or in
667 * the case of no file, such as an eval, the whole contents);
669 * new_filter indicates that this is a new file and it shouldn't inherit
670 * the filters from the current parser (ie require).
674 Perl_lex_start(pTHX_ SV *line, PerlIO *rsfp, bool new_filter)
677 const char *s = NULL;
679 yy_parser *parser, *oparser;
681 /* create and initialise a parser */
683 Newxz(parser, 1, yy_parser);
684 parser->old_parser = oparser = PL_parser;
687 Newx(parser->stack, YYINITDEPTH, yy_stack_frame);
688 parser->ps = parser->stack;
689 parser->stack_size = YYINITDEPTH;
691 parser->stack->state = 0;
692 parser->yyerrstatus = 0;
693 parser->yychar = YYEMPTY; /* Cause a token to be read. */
695 /* on scope exit, free this parser and restore any outer one */
697 parser->saved_curcop = PL_curcop;
699 /* initialise lexer state */
702 parser->curforce = -1;
704 parser->nexttoke = 0;
706 parser->error_count = oparser ? oparser->error_count : 0;
707 parser->copline = NOLINE;
708 parser->lex_state = LEX_NORMAL;
709 parser->expect = XSTATE;
711 parser->rsfp_filters = (new_filter || !oparser) ? newAV()
712 : MUTABLE_AV(SvREFCNT_inc(oparser->rsfp_filters));
714 Newx(parser->lex_brackstack, 120, char);
715 Newx(parser->lex_casestack, 12, char);
716 *parser->lex_casestack = '\0';
719 s = SvPV_const(line, len);
725 parser->linestr = newSVpvs("\n;");
726 } else if (SvREADONLY(line) || s[len-1] != ';') {
727 parser->linestr = newSVsv(line);
729 sv_catpvs(parser->linestr, "\n;");
732 SvREFCNT_inc_simple_void_NN(line);
733 parser->linestr = line;
735 parser->oldoldbufptr =
738 parser->linestart = SvPVX(parser->linestr);
739 parser->bufend = parser->bufptr + SvCUR(parser->linestr);
740 parser->last_lop = parser->last_uni = NULL;
744 /* delete a parser object */
747 Perl_parser_free(pTHX_ const yy_parser *parser)
749 PERL_ARGS_ASSERT_PARSER_FREE;
751 PL_curcop = parser->saved_curcop;
752 SvREFCNT_dec(parser->linestr);
754 if (parser->rsfp == PerlIO_stdin())
755 PerlIO_clearerr(parser->rsfp);
756 else if (parser->rsfp && (!parser->old_parser ||
757 (parser->old_parser && parser->rsfp != parser->old_parser->rsfp)))
758 PerlIO_close(parser->rsfp);
759 SvREFCNT_dec(parser->rsfp_filters);
761 Safefree(parser->stack);
762 Safefree(parser->lex_brackstack);
763 Safefree(parser->lex_casestack);
764 PL_parser = parser->old_parser;
771 * Finalizer for lexing operations. Must be called when the parser is
772 * done with the lexer.
779 PL_doextract = FALSE;
784 * This subroutine has nothing to do with tilting, whether at windmills
785 * or pinball tables. Its name is short for "increment line". It
786 * increments the current line number in CopLINE(PL_curcop) and checks
787 * to see whether the line starts with a comment of the form
788 * # line 500 "foo.pm"
789 * If so, it sets the current line number and file to the values in the comment.
793 S_incline(pTHX_ const char *s)
800 PERL_ARGS_ASSERT_INCLINE;
802 CopLINE_inc(PL_curcop);
805 while (SPACE_OR_TAB(*s))
807 if (strnEQ(s, "line", 4))
811 if (SPACE_OR_TAB(*s))
815 while (SPACE_OR_TAB(*s))
823 if (!SPACE_OR_TAB(*s) && *s != '\r' && *s != '\n' && *s != '\0')
825 while (SPACE_OR_TAB(*s))
827 if (*s == '"' && (t = strchr(s+1, '"'))) {
837 while (SPACE_OR_TAB(*e) || *e == '\r' || *e == '\f')
839 if (*e != '\n' && *e != '\0')
840 return; /* false alarm */
843 const STRLEN len = t - s;
845 SV *const temp_sv = CopFILESV(PL_curcop);
851 tmplen = SvCUR(temp_sv);
857 if (tmplen > 7 && strnEQ(cf, "(eval ", 6)) {
858 /* must copy *{"::_<(eval N)[oldfilename:L]"}
859 * to *{"::_<newfilename"} */
860 /* However, the long form of evals is only turned on by the
861 debugger - usually they're "(eval %lu)" */
865 STRLEN tmplen2 = len;
866 if (tmplen + 2 <= sizeof smallbuf)
869 Newx(tmpbuf, tmplen + 2, char);
872 memcpy(tmpbuf + 2, cf, tmplen);
874 gvp = (GV**)hv_fetch(PL_defstash, tmpbuf, tmplen, FALSE);
879 if (tmplen2 + 2 <= sizeof smallbuf)
882 Newx(tmpbuf2, tmplen2 + 2, char);
884 if (tmpbuf2 != smallbuf || tmpbuf != smallbuf) {
885 /* Either they malloc'd it, or we malloc'd it,
886 so no prefix is present in ours. */
891 memcpy(tmpbuf2 + 2, s, tmplen2);
894 gv2 = *(GV**)hv_fetch(PL_defstash, tmpbuf2, tmplen2, TRUE);
896 gv_init(gv2, PL_defstash, tmpbuf2, tmplen2, FALSE);
897 /* adjust ${"::_<newfilename"} to store the new file name */
898 GvSV(gv2) = newSVpvn(tmpbuf2 + 2, tmplen2 - 2);
899 GvHV(gv2) = MUTABLE_HV(SvREFCNT_inc(GvHV(*gvp)));
900 GvAV(gv2) = MUTABLE_AV(SvREFCNT_inc(GvAV(*gvp)));
903 if (tmpbuf2 != smallbuf) Safefree(tmpbuf2);
905 if (tmpbuf != smallbuf) Safefree(tmpbuf);
908 CopFILE_free(PL_curcop);
909 CopFILE_setn(PL_curcop, s, len);
911 CopLINE_set(PL_curcop, atoi(n)-1);
915 /* skip space before PL_thistoken */
918 S_skipspace0(pTHX_ register char *s)
920 PERL_ARGS_ASSERT_SKIPSPACE0;
927 PL_thiswhite = newSVpvs("");
928 sv_catsv(PL_thiswhite, PL_skipwhite);
929 sv_free(PL_skipwhite);
932 PL_realtokenstart = s - SvPVX(PL_linestr);
936 /* skip space after PL_thistoken */
939 S_skipspace1(pTHX_ register char *s)
941 const char *start = s;
942 I32 startoff = start - SvPVX(PL_linestr);
944 PERL_ARGS_ASSERT_SKIPSPACE1;
949 start = SvPVX(PL_linestr) + startoff;
950 if (!PL_thistoken && PL_realtokenstart >= 0) {
951 const char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
952 PL_thistoken = newSVpvn(tstart, start - tstart);
954 PL_realtokenstart = -1;
957 PL_nextwhite = newSVpvs("");
958 sv_catsv(PL_nextwhite, PL_skipwhite);
959 sv_free(PL_skipwhite);
966 S_skipspace2(pTHX_ register char *s, SV **svp)
969 const I32 bufptroff = PL_bufptr - SvPVX(PL_linestr);
970 const I32 startoff = s - SvPVX(PL_linestr);
972 PERL_ARGS_ASSERT_SKIPSPACE2;
975 PL_bufptr = SvPVX(PL_linestr) + bufptroff;
976 if (!PL_madskills || !svp)
978 start = SvPVX(PL_linestr) + startoff;
979 if (!PL_thistoken && PL_realtokenstart >= 0) {
980 char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
981 PL_thistoken = newSVpvn(tstart, start - tstart);
982 PL_realtokenstart = -1;
987 sv_setsv(*svp, PL_skipwhite);
988 sv_free(PL_skipwhite);
997 S_update_debugger_info(pTHX_ SV *orig_sv, const char *const buf, STRLEN len)
999 AV *av = CopFILEAVx(PL_curcop);
1001 SV * const sv = newSV_type(SVt_PVMG);
1003 sv_setsv(sv, orig_sv);
1005 sv_setpvn(sv, buf, len);
1008 av_store(av, (I32)CopLINE(PL_curcop), sv);
1014 * Called to gobble the appropriate amount and type of whitespace.
1015 * Skips comments as well.
1019 S_skipspace(pTHX_ register char *s)
1024 int startoff = s - SvPVX(PL_linestr);
1026 PERL_ARGS_ASSERT_SKIPSPACE;
1029 sv_free(PL_skipwhite);
1033 PERL_ARGS_ASSERT_SKIPSPACE;
1035 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
1036 while (s < PL_bufend && SPACE_OR_TAB(*s))
1046 SSize_t oldprevlen, oldoldprevlen;
1047 SSize_t oldloplen = 0, oldunilen = 0;
1048 while (s < PL_bufend && isSPACE(*s)) {
1049 if (*s++ == '\n' && PL_in_eval && !PL_rsfp)
1054 if (s < PL_bufend && *s == '#') {
1055 while (s < PL_bufend && *s != '\n')
1057 if (s < PL_bufend) {
1059 if (PL_in_eval && !PL_rsfp) {
1066 /* only continue to recharge the buffer if we're at the end
1067 * of the buffer, we're not reading from a source filter, and
1068 * we're in normal lexing mode
1070 if (s < PL_bufend || !PL_rsfp || PL_sublex_info.sub_inwhat ||
1071 PL_lex_state == LEX_FORMLINE)
1078 /* try to recharge the buffer */
1080 curoff = s - SvPVX(PL_linestr);
1083 if ((s = filter_gets(PL_linestr, PL_rsfp,
1084 (prevlen = SvCUR(PL_linestr)))) == NULL)
1087 if (PL_madskills && curoff != startoff) {
1089 PL_skipwhite = newSVpvs("");
1090 sv_catpvn(PL_skipwhite, SvPVX(PL_linestr) + startoff,
1094 /* mustn't throw out old stuff yet if madpropping */
1095 SvCUR(PL_linestr) = curoff;
1096 s = SvPVX(PL_linestr) + curoff;
1098 if (curoff && s[-1] == '\n')
1102 /* end of file. Add on the -p or -n magic */
1103 /* XXX these shouldn't really be added here, can't set PL_faketokens */
1106 sv_catpvs(PL_linestr,
1107 ";}continue{print or die qq(-p destination: $!\\n);}");
1109 sv_setpvs(PL_linestr,
1110 ";}continue{print or die qq(-p destination: $!\\n);}");
1112 PL_minus_n = PL_minus_p = 0;
1114 else if (PL_minus_n) {
1116 sv_catpvs(PL_linestr, ";}");
1118 sv_setpvs(PL_linestr, ";}");
1124 sv_catpvs(PL_linestr,";");
1126 sv_setpvs(PL_linestr,";");
1129 /* reset variables for next time we lex */
1130 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = s = PL_linestart
1136 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
1137 PL_last_lop = PL_last_uni = NULL;
1139 /* Close the filehandle. Could be from
1140 * STDIN, or a regular file. If we were reading code from
1141 * STDIN (because the commandline held no -e or filename)
1142 * then we don't close it, we reset it so the code can
1143 * read from STDIN too.
1146 if ((PerlIO*)PL_rsfp == PerlIO_stdin())
1147 PerlIO_clearerr(PL_rsfp);
1149 (void)PerlIO_close(PL_rsfp);
1154 /* not at end of file, so we only read another line */
1155 /* make corresponding updates to old pointers, for yyerror() */
1156 oldprevlen = PL_oldbufptr - PL_bufend;
1157 oldoldprevlen = PL_oldoldbufptr - PL_bufend;
1159 oldunilen = PL_last_uni - PL_bufend;
1161 oldloplen = PL_last_lop - PL_bufend;
1162 PL_linestart = PL_bufptr = s + prevlen;
1163 PL_bufend = s + SvCUR(PL_linestr);
1165 PL_oldbufptr = s + oldprevlen;
1166 PL_oldoldbufptr = s + oldoldprevlen;
1168 PL_last_uni = s + oldunilen;
1170 PL_last_lop = s + oldloplen;
1173 /* debugger active and we're not compiling the debugger code,
1174 * so store the line into the debugger's array of lines
1176 if ((PERLDB_LINE || PERLDB_SAVESRC) && PL_curstash != PL_debstash)
1177 update_debugger_info(NULL, PL_bufptr, PL_bufend - PL_bufptr);
1184 PL_skipwhite = newSVpvs("");
1185 curoff = s - SvPVX(PL_linestr);
1186 if (curoff - startoff)
1187 sv_catpvn(PL_skipwhite, SvPVX(PL_linestr) + startoff,
1196 * Check the unary operators to ensure there's no ambiguity in how they're
1197 * used. An ambiguous piece of code would be:
1199 * This doesn't mean rand() + 5. Because rand() is a unary operator,
1200 * the +5 is its argument.
1210 if (PL_oldoldbufptr != PL_last_uni)
1212 while (isSPACE(*PL_last_uni))
1215 while (isALNUM_lazy_if(s,UTF) || *s == '-')
1217 if ((t = strchr(s, '(')) && t < PL_bufptr)
1220 Perl_ck_warner_d(aTHX_ packWARN(WARN_AMBIGUOUS),
1221 "Warning: Use of \"%.*s\" without parentheses is ambiguous",
1222 (int)(s - PL_last_uni), PL_last_uni);
1226 * LOP : macro to build a list operator. Its behaviour has been replaced
1227 * with a subroutine, S_lop() for which LOP is just another name.
1230 #define LOP(f,x) return lop(f,x,s)
1234 * Build a list operator (or something that might be one). The rules:
1235 * - if we have a next token, then it's a list operator [why?]
1236 * - if the next thing is an opening paren, then it's a function
1237 * - else it's a list operator
1241 S_lop(pTHX_ I32 f, int x, char *s)
1245 PERL_ARGS_ASSERT_LOP;
1251 PL_last_lop = PL_oldbufptr;
1252 PL_last_lop_op = (OPCODE)f;
1255 return REPORT(LSTOP);
1258 return REPORT(LSTOP);
1261 return REPORT(FUNC);
1264 return REPORT(FUNC);
1266 return REPORT(LSTOP);
1272 * Sets up for an eventual force_next(). start_force(0) basically does
1273 * an unshift, while start_force(-1) does a push. yylex removes items
1278 S_start_force(pTHX_ int where)
1282 if (where < 0) /* so people can duplicate start_force(PL_curforce) */
1283 where = PL_lasttoke;
1284 assert(PL_curforce < 0 || PL_curforce == where);
1285 if (PL_curforce != where) {
1286 for (i = PL_lasttoke; i > where; --i) {
1287 PL_nexttoke[i] = PL_nexttoke[i-1];
1291 if (PL_curforce < 0) /* in case of duplicate start_force() */
1292 Zero(&PL_nexttoke[where], 1, NEXTTOKE);
1293 PL_curforce = where;
1296 curmad('^', newSVpvs(""));
1297 CURMAD('_', PL_nextwhite);
1302 S_curmad(pTHX_ char slot, SV *sv)
1308 if (PL_curforce < 0)
1309 where = &PL_thismad;
1311 where = &PL_nexttoke[PL_curforce].next_mad;
1317 if (UTF && is_utf8_string((U8*)SvPVX(sv), SvCUR(sv)))
1319 else if (PL_encoding) {
1320 sv_recode_to_utf8(sv, PL_encoding);
1325 /* keep a slot open for the head of the list? */
1326 if (slot != '_' && *where && (*where)->mad_key == '^') {
1327 (*where)->mad_key = slot;
1328 sv_free(MUTABLE_SV(((*where)->mad_val)));
1329 (*where)->mad_val = (void*)sv;
1332 addmad(newMADsv(slot, sv), where, 0);
1335 # define start_force(where) NOOP
1336 # define curmad(slot, sv) NOOP
1341 * When the lexer realizes it knows the next token (for instance,
1342 * it is reordering tokens for the parser) then it can call S_force_next
1343 * to know what token to return the next time the lexer is called. Caller
1344 * will need to set PL_nextval[] (or PL_nexttoke[].next_val with PERL_MAD),
1345 * and possibly PL_expect to ensure the lexer handles the token correctly.
1349 S_force_next(pTHX_ I32 type)
1354 PerlIO_printf(Perl_debug_log, "### forced token:\n");
1355 tokereport(type, &NEXTVAL_NEXTTOKE);
1359 if (PL_curforce < 0)
1360 start_force(PL_lasttoke);
1361 PL_nexttoke[PL_curforce].next_type = type;
1362 if (PL_lex_state != LEX_KNOWNEXT)
1363 PL_lex_defer = PL_lex_state;
1364 PL_lex_state = LEX_KNOWNEXT;
1365 PL_lex_expect = PL_expect;
1368 PL_nexttype[PL_nexttoke] = type;
1370 if (PL_lex_state != LEX_KNOWNEXT) {
1371 PL_lex_defer = PL_lex_state;
1372 PL_lex_expect = PL_expect;
1373 PL_lex_state = LEX_KNOWNEXT;
1379 S_newSV_maybe_utf8(pTHX_ const char *const start, STRLEN len)
1382 SV * const sv = newSVpvn_utf8(start, len,
1385 && !is_ascii_string((const U8*)start, len)
1386 && is_utf8_string((const U8*)start, len));
1392 * When the lexer knows the next thing is a word (for instance, it has
1393 * just seen -> and it knows that the next char is a word char, then
1394 * it calls S_force_word to stick the next word into the PL_nexttoke/val
1398 * char *start : buffer position (must be within PL_linestr)
1399 * int token : PL_next* will be this type of bare word (e.g., METHOD,WORD)
1400 * int check_keyword : if true, Perl checks to make sure the word isn't
1401 * a keyword (do this if the word is a label, e.g. goto FOO)
1402 * int allow_pack : if true, : characters will also be allowed (require,
1403 * use, etc. do this)
1404 * int allow_initial_tick : used by the "sub" lexer only.
1408 S_force_word(pTHX_ register char *start, int token, int check_keyword, int allow_pack, int allow_initial_tick)
1414 PERL_ARGS_ASSERT_FORCE_WORD;
1416 start = SKIPSPACE1(start);
1418 if (isIDFIRST_lazy_if(s,UTF) ||
1419 (allow_pack && *s == ':') ||
1420 (allow_initial_tick && *s == '\'') )
1422 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, allow_pack, &len);
1423 if (check_keyword && keyword(PL_tokenbuf, len, 0))
1425 start_force(PL_curforce);
1427 curmad('X', newSVpvn(start,s-start));
1428 if (token == METHOD) {
1433 PL_expect = XOPERATOR;
1437 curmad('g', newSVpvs( "forced" ));
1438 NEXTVAL_NEXTTOKE.opval
1439 = (OP*)newSVOP(OP_CONST,0,
1440 S_newSV_maybe_utf8(aTHX_ PL_tokenbuf, len));
1441 NEXTVAL_NEXTTOKE.opval->op_private |= OPpCONST_BARE;
1449 * Called when the lexer wants $foo *foo &foo etc, but the program
1450 * text only contains the "foo" portion. The first argument is a pointer
1451 * to the "foo", and the second argument is the type symbol to prefix.
1452 * Forces the next token to be a "WORD".
1453 * Creates the symbol if it didn't already exist (via gv_fetchpv()).
1457 S_force_ident(pTHX_ register const char *s, int kind)
1461 PERL_ARGS_ASSERT_FORCE_IDENT;
1464 const STRLEN len = strlen(s);
1465 OP* const o = (OP*)newSVOP(OP_CONST, 0, newSVpvn(s, len));
1466 start_force(PL_curforce);
1467 NEXTVAL_NEXTTOKE.opval = o;
1470 o->op_private = OPpCONST_ENTERED;
1471 /* XXX see note in pp_entereval() for why we forgo typo
1472 warnings if the symbol must be introduced in an eval.
1474 gv_fetchpvn_flags(s, len,
1475 PL_in_eval ? (GV_ADDMULTI | GV_ADDINEVAL)
1477 kind == '$' ? SVt_PV :
1478 kind == '@' ? SVt_PVAV :
1479 kind == '%' ? SVt_PVHV :
1487 Perl_str_to_version(pTHX_ SV *sv)
1492 const char *start = SvPV_const(sv,len);
1493 const char * const end = start + len;
1494 const bool utf = SvUTF8(sv) ? TRUE : FALSE;
1496 PERL_ARGS_ASSERT_STR_TO_VERSION;
1498 while (start < end) {
1502 n = utf8n_to_uvchr((U8*)start, len, &skip, 0);
1507 retval += ((NV)n)/nshift;
1516 * Forces the next token to be a version number.
1517 * If the next token appears to be an invalid version number, (e.g. "v2b"),
1518 * and if "guessing" is TRUE, then no new token is created (and the caller
1519 * must use an alternative parsing method).
1523 S_force_version(pTHX_ char *s, int guessing)
1529 I32 startoff = s - SvPVX(PL_linestr);
1532 PERL_ARGS_ASSERT_FORCE_VERSION;
1540 while (isDIGIT(*d) || *d == '_' || *d == '.')
1544 start_force(PL_curforce);
1545 curmad('X', newSVpvn(s,d-s));
1548 if (*d == ';' || isSPACE(*d) || *d == '}' || !*d) {
1550 s = scan_num(s, &pl_yylval);
1551 version = pl_yylval.opval;
1552 ver = cSVOPx(version)->op_sv;
1553 if (SvPOK(ver) && !SvNIOK(ver)) {
1554 SvUPGRADE(ver, SVt_PVNV);
1555 SvNV_set(ver, str_to_version(ver));
1556 SvNOK_on(ver); /* hint that it is a version */
1559 else if (guessing) {
1562 sv_free(PL_nextwhite); /* let next token collect whitespace */
1564 s = SvPVX(PL_linestr) + startoff;
1572 if (PL_madskills && !version) {
1573 sv_free(PL_nextwhite); /* let next token collect whitespace */
1575 s = SvPVX(PL_linestr) + startoff;
1578 /* NOTE: The parser sees the package name and the VERSION swapped */
1579 start_force(PL_curforce);
1580 NEXTVAL_NEXTTOKE.opval = version;
1588 * Tokenize a quoted string passed in as an SV. It finds the next
1589 * chunk, up to end of string or a backslash. It may make a new
1590 * SV containing that chunk (if HINT_NEW_STRING is on). It also
1595 S_tokeq(pTHX_ SV *sv)
1599 register char *send;
1604 PERL_ARGS_ASSERT_TOKEQ;
1609 s = SvPV_force(sv, len);
1610 if (SvTYPE(sv) >= SVt_PVIV && SvIVX(sv) == -1)
1613 while (s < send && *s != '\\')
1618 if ( PL_hints & HINT_NEW_STRING ) {
1619 pv = newSVpvn_flags(SvPVX_const(pv), len, SVs_TEMP | SvUTF8(sv));
1623 if (s + 1 < send && (s[1] == '\\'))
1624 s++; /* all that, just for this */
1629 SvCUR_set(sv, d - SvPVX_const(sv));
1631 if ( PL_hints & HINT_NEW_STRING )
1632 return new_constant(NULL, 0, "q", sv, pv, "q", 1);
1637 * Now come three functions related to double-quote context,
1638 * S_sublex_start, S_sublex_push, and S_sublex_done. They're used when
1639 * converting things like "\u\Lgnat" into ucfirst(lc("gnat")). They
1640 * interact with PL_lex_state, and create fake ( ... ) argument lists
1641 * to handle functions and concatenation.
1642 * They assume that whoever calls them will be setting up a fake
1643 * join call, because each subthing puts a ',' after it. This lets
1646 * join($, , 'lower ', lcfirst( 'uPpEr', ) ,)
1648 * (I'm not sure whether the spurious commas at the end of lcfirst's
1649 * arguments and join's arguments are created or not).
1654 * Assumes that pl_yylval.ival is the op we're creating (e.g. OP_LCFIRST).
1656 * Pattern matching will set PL_lex_op to the pattern-matching op to
1657 * make (we return THING if pl_yylval.ival is OP_NULL, PMFUNC otherwise).
1659 * OP_CONST and OP_READLINE are easy--just make the new op and return.
1661 * Everything else becomes a FUNC.
1663 * Sets PL_lex_state to LEX_INTERPPUSH unless (ival was OP_NULL or we
1664 * had an OP_CONST or OP_READLINE). This just sets us up for a
1665 * call to S_sublex_push().
1669 S_sublex_start(pTHX)
1672 register const I32 op_type = pl_yylval.ival;
1674 if (op_type == OP_NULL) {
1675 pl_yylval.opval = PL_lex_op;
1679 if (op_type == OP_CONST || op_type == OP_READLINE) {
1680 SV *sv = tokeq(PL_lex_stuff);
1682 if (SvTYPE(sv) == SVt_PVIV) {
1683 /* Overloaded constants, nothing fancy: Convert to SVt_PV: */
1685 const char * const p = SvPV_const(sv, len);
1686 SV * const nsv = newSVpvn_flags(p, len, SvUTF8(sv));
1690 pl_yylval.opval = (OP*)newSVOP(op_type, 0, sv);
1691 PL_lex_stuff = NULL;
1692 /* Allow <FH> // "foo" */
1693 if (op_type == OP_READLINE)
1694 PL_expect = XTERMORDORDOR;
1697 else if (op_type == OP_BACKTICK && PL_lex_op) {
1698 /* readpipe() vas overriden */
1699 cSVOPx(cLISTOPx(cUNOPx(PL_lex_op)->op_first)->op_first->op_sibling)->op_sv = tokeq(PL_lex_stuff);
1700 pl_yylval.opval = PL_lex_op;
1702 PL_lex_stuff = NULL;
1706 PL_sublex_info.super_state = PL_lex_state;
1707 PL_sublex_info.sub_inwhat = (U16)op_type;
1708 PL_sublex_info.sub_op = PL_lex_op;
1709 PL_lex_state = LEX_INTERPPUSH;
1713 pl_yylval.opval = PL_lex_op;
1723 * Create a new scope to save the lexing state. The scope will be
1724 * ended in S_sublex_done. Returns a '(', starting the function arguments
1725 * to the uc, lc, etc. found before.
1726 * Sets PL_lex_state to LEX_INTERPCONCAT.
1735 PL_lex_state = PL_sublex_info.super_state;
1736 SAVEBOOL(PL_lex_dojoin);
1737 SAVEI32(PL_lex_brackets);
1738 SAVEI32(PL_lex_casemods);
1739 SAVEI32(PL_lex_starts);
1740 SAVEI8(PL_lex_state);
1741 SAVEVPTR(PL_lex_inpat);
1742 SAVEI16(PL_lex_inwhat);
1743 SAVECOPLINE(PL_curcop);
1744 SAVEPPTR(PL_bufptr);
1745 SAVEPPTR(PL_bufend);
1746 SAVEPPTR(PL_oldbufptr);
1747 SAVEPPTR(PL_oldoldbufptr);
1748 SAVEPPTR(PL_last_lop);
1749 SAVEPPTR(PL_last_uni);
1750 SAVEPPTR(PL_linestart);
1751 SAVESPTR(PL_linestr);
1752 SAVEGENERICPV(PL_lex_brackstack);
1753 SAVEGENERICPV(PL_lex_casestack);
1755 PL_linestr = PL_lex_stuff;
1756 PL_lex_stuff = NULL;
1758 PL_bufend = PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart
1759 = SvPVX(PL_linestr);
1760 PL_bufend += SvCUR(PL_linestr);
1761 PL_last_lop = PL_last_uni = NULL;
1762 SAVEFREESV(PL_linestr);
1764 PL_lex_dojoin = FALSE;
1765 PL_lex_brackets = 0;
1766 Newx(PL_lex_brackstack, 120, char);
1767 Newx(PL_lex_casestack, 12, char);
1768 PL_lex_casemods = 0;
1769 *PL_lex_casestack = '\0';
1771 PL_lex_state = LEX_INTERPCONCAT;
1772 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
1774 PL_lex_inwhat = PL_sublex_info.sub_inwhat;
1775 if (PL_lex_inwhat == OP_MATCH || PL_lex_inwhat == OP_QR || PL_lex_inwhat == OP_SUBST)
1776 PL_lex_inpat = PL_sublex_info.sub_op;
1778 PL_lex_inpat = NULL;
1785 * Restores lexer state after a S_sublex_push.
1792 if (!PL_lex_starts++) {
1793 SV * const sv = newSVpvs("");
1794 if (SvUTF8(PL_linestr))
1796 PL_expect = XOPERATOR;
1797 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
1801 if (PL_lex_casemods) { /* oops, we've got some unbalanced parens */
1802 PL_lex_state = LEX_INTERPCASEMOD;
1806 /* Is there a right-hand side to take care of? (s//RHS/ or tr//RHS/) */
1807 if (PL_lex_repl && (PL_lex_inwhat == OP_SUBST || PL_lex_inwhat == OP_TRANS)) {
1808 PL_linestr = PL_lex_repl;
1810 PL_bufend = PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart = SvPVX(PL_linestr);
1811 PL_bufend += SvCUR(PL_linestr);
1812 PL_last_lop = PL_last_uni = NULL;
1813 SAVEFREESV(PL_linestr);
1814 PL_lex_dojoin = FALSE;
1815 PL_lex_brackets = 0;
1816 PL_lex_casemods = 0;
1817 *PL_lex_casestack = '\0';
1819 if (SvEVALED(PL_lex_repl)) {
1820 PL_lex_state = LEX_INTERPNORMAL;
1822 /* we don't clear PL_lex_repl here, so that we can check later
1823 whether this is an evalled subst; that means we rely on the
1824 logic to ensure sublex_done() is called again only via the
1825 branch (in yylex()) that clears PL_lex_repl, else we'll loop */
1828 PL_lex_state = LEX_INTERPCONCAT;
1838 PL_endwhite = newSVpvs("");
1839 sv_catsv(PL_endwhite, PL_thiswhite);
1843 sv_setpvs(PL_thistoken,"");
1845 PL_realtokenstart = -1;
1849 PL_bufend = SvPVX(PL_linestr);
1850 PL_bufend += SvCUR(PL_linestr);
1851 PL_expect = XOPERATOR;
1852 PL_sublex_info.sub_inwhat = 0;
1860 Extracts a pattern, double-quoted string, or transliteration. This
1863 It looks at PL_lex_inwhat and PL_lex_inpat to find out whether it's
1864 processing a pattern (PL_lex_inpat is true), a transliteration
1865 (PL_lex_inwhat == OP_TRANS is true), or a double-quoted string.
1867 Returns a pointer to the character scanned up to. If this is
1868 advanced from the start pointer supplied (i.e. if anything was
1869 successfully parsed), will leave an OP for the substring scanned
1870 in pl_yylval. Caller must intuit reason for not parsing further
1871 by looking at the next characters herself.
1875 double-quoted style: \r and \n
1876 regexp special ones: \D \s
1879 case and quoting: \U \Q \E
1880 stops on @ and $, but not for $ as tail anchor
1882 In transliterations:
1883 characters are VERY literal, except for - not at the start or end
1884 of the string, which indicates a range. If the range is in bytes,
1885 scan_const expands the range to the full set of intermediate
1886 characters. If the range is in utf8, the hyphen is replaced with
1887 a certain range mark which will be handled by pmtrans() in op.c.
1889 In double-quoted strings:
1891 double-quoted style: \r and \n
1893 deprecated backrefs: \1 (in substitution replacements)
1894 case and quoting: \U \Q \E
1897 scan_const does *not* construct ops to handle interpolated strings.
1898 It stops processing as soon as it finds an embedded $ or @ variable
1899 and leaves it to the caller to work out what's going on.
1901 embedded arrays (whether in pattern or not) could be:
1902 @foo, @::foo, @'foo, @{foo}, @$foo, @+, @-.
1904 $ in double-quoted strings must be the symbol of an embedded scalar.
1906 $ in pattern could be $foo or could be tail anchor. Assumption:
1907 it's a tail anchor if $ is the last thing in the string, or if it's
1908 followed by one of "()| \r\n\t"
1910 \1 (backreferences) are turned into $1
1912 The structure of the code is
1913 while (there's a character to process) {
1914 handle transliteration ranges
1915 skip regexp comments /(?#comment)/ and codes /(?{code})/
1916 skip #-initiated comments in //x patterns
1917 check for embedded arrays
1918 check for embedded scalars
1920 leave intact backslashes from leaveit (below)
1921 deprecate \1 in substitution replacements
1922 handle string-changing backslashes \l \U \Q \E, etc.
1923 switch (what was escaped) {
1924 handle \- in a transliteration (becomes a literal -)
1925 handle \132 (octal characters)
1926 handle \x15 and \x{1234} (hex characters)
1927 handle \N{name} (named characters)
1928 handle \cV (control characters)
1929 handle printf-style backslashes (\f, \r, \n, etc)
1932 } (end if backslash)
1933 handle regular character
1934 } (end while character to read)
1939 S_scan_const(pTHX_ char *start)
1942 register char *send = PL_bufend; /* end of the constant */
1943 SV *sv = newSV(send - start); /* sv for the constant. See
1944 note below on sizing. */
1945 register char *s = start; /* start of the constant */
1946 register char *d = SvPVX(sv); /* destination for copies */
1947 bool dorange = FALSE; /* are we in a translit range? */
1948 bool didrange = FALSE; /* did we just finish a range? */
1949 I32 has_utf8 = FALSE; /* Output constant is UTF8 */
1950 I32 this_utf8 = UTF; /* Is the source string assumed
1951 to be UTF8? But, this can
1952 show as true when the source
1953 isn't utf8, as for example
1954 when it is entirely composed
1957 /* Note on sizing: The scanned constant is placed into sv, which is
1958 * initialized by newSV() assuming one byte of output for every byte of
1959 * input. This routine expects newSV() to allocate an extra byte for a
1960 * trailing NUL, which this routine will append if it gets to the end of
1961 * the input. There may be more bytes of input than output (eg., \N{LATIN
1962 * CAPITAL LETTER A}), or more output than input if the constant ends up
1963 * recoded to utf8, but each time a construct is found that might increase
1964 * the needed size, SvGROW() is called. Its size parameter each time is
1965 * based on the best guess estimate at the time, namely the length used so
1966 * far, plus the length the current construct will occupy, plus room for
1967 * the trailing NUL, plus one byte for every input byte still unscanned */
1971 UV literal_endpoint = 0;
1972 bool native_range = TRUE; /* turned to FALSE if the first endpoint is Unicode. */
1975 PERL_ARGS_ASSERT_SCAN_CONST;
1977 if (PL_lex_inwhat == OP_TRANS && PL_sublex_info.sub_op) {
1978 /* If we are doing a trans and we know we want UTF8 set expectation */
1979 has_utf8 = PL_sublex_info.sub_op->op_private & (OPpTRANS_FROM_UTF|OPpTRANS_TO_UTF);
1980 this_utf8 = PL_sublex_info.sub_op->op_private & (PL_lex_repl ? OPpTRANS_FROM_UTF : OPpTRANS_TO_UTF);
1984 while (s < send || dorange) {
1985 /* get transliterations out of the way (they're most literal) */
1986 if (PL_lex_inwhat == OP_TRANS) {
1987 /* expand a range A-Z to the full set of characters. AIE! */
1989 I32 i; /* current expanded character */
1990 I32 min; /* first character in range */
1991 I32 max; /* last character in range */
2002 char * const c = (char*)utf8_hop((U8*)d, -1);
2006 *c = (char)UTF_TO_NATIVE(0xff);
2007 /* mark the range as done, and continue */
2013 i = d - SvPVX_const(sv); /* remember current offset */
2016 SvLEN(sv) + (has_utf8 ?
2017 (512 - UTF_CONTINUATION_MARK +
2020 /* How many two-byte within 0..255: 128 in UTF-8,
2021 * 96 in UTF-8-mod. */
2023 SvGROW(sv, SvLEN(sv) + 256); /* never more than 256 chars in a range */
2025 d = SvPVX(sv) + i; /* refresh d after realloc */
2029 for (j = 0; j <= 1; j++) {
2030 char * const c = (char*)utf8_hop((U8*)d, -1);
2031 const UV uv = utf8n_to_uvchr((U8*)c, d - c, NULL, 0);
2037 max = (U8)0xff; /* only to \xff */
2038 uvmax = uv; /* \x{100} to uvmax */
2040 d = c; /* eat endpoint chars */
2045 d -= 2; /* eat the first char and the - */
2046 min = (U8)*d; /* first char in range */
2047 max = (U8)d[1]; /* last char in range */
2054 "Invalid range \"%c-%c\" in transliteration operator",
2055 (char)min, (char)max);
2059 if (literal_endpoint == 2 &&
2060 ((isLOWER(min) && isLOWER(max)) ||
2061 (isUPPER(min) && isUPPER(max)))) {
2063 for (i = min; i <= max; i++)
2065 *d++ = NATIVE_TO_NEED(has_utf8,i);
2067 for (i = min; i <= max; i++)
2069 *d++ = NATIVE_TO_NEED(has_utf8,i);
2074 for (i = min; i <= max; i++)
2077 const U8 ch = (U8)NATIVE_TO_UTF(i);
2078 if (UNI_IS_INVARIANT(ch))
2081 *d++ = (U8)UTF8_EIGHT_BIT_HI(ch);
2082 *d++ = (U8)UTF8_EIGHT_BIT_LO(ch);
2091 d = (char*)uvchr_to_utf8((U8*)d, 0x100);
2093 *d++ = (char)UTF_TO_NATIVE(0xff);
2095 d = (char*)uvchr_to_utf8((U8*)d, uvmax);
2099 /* mark the range as done, and continue */
2103 literal_endpoint = 0;
2108 /* range begins (ignore - as first or last char) */
2109 else if (*s == '-' && s+1 < send && s != start) {
2111 Perl_croak(aTHX_ "Ambiguous range in transliteration operator");
2118 *d++ = (char)UTF_TO_NATIVE(0xff); /* use illegal utf8 byte--see pmtrans */
2128 literal_endpoint = 0;
2129 native_range = TRUE;
2134 /* if we get here, we're not doing a transliteration */
2136 /* skip for regexp comments /(?#comment)/ and code /(?{code})/,
2137 except for the last char, which will be done separately. */
2138 else if (*s == '(' && PL_lex_inpat && s[1] == '?') {
2140 while (s+1 < send && *s != ')')
2141 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2143 else if (s[2] == '{' /* This should match regcomp.c */
2144 || (s[2] == '?' && s[3] == '{'))
2147 char *regparse = s + (s[2] == '{' ? 3 : 4);
2150 while (count && (c = *regparse)) {
2151 if (c == '\\' && regparse[1])
2159 if (*regparse != ')')
2160 regparse--; /* Leave one char for continuation. */
2161 while (s < regparse)
2162 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2166 /* likewise skip #-initiated comments in //x patterns */
2167 else if (*s == '#' && PL_lex_inpat &&
2168 ((PMOP*)PL_lex_inpat)->op_pmflags & PMf_EXTENDED) {
2169 while (s+1 < send && *s != '\n')
2170 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2173 /* check for embedded arrays
2174 (@foo, @::foo, @'foo, @{foo}, @$foo, @+, @-)
2176 else if (*s == '@' && s[1]) {
2177 if (isALNUM_lazy_if(s+1,UTF))
2179 if (strchr(":'{$", s[1]))
2181 if (!PL_lex_inpat && (s[1] == '+' || s[1] == '-'))
2182 break; /* in regexp, neither @+ nor @- are interpolated */
2185 /* check for embedded scalars. only stop if we're sure it's a
2188 else if (*s == '$') {
2189 if (!PL_lex_inpat) /* not a regexp, so $ must be var */
2191 if (s + 1 < send && !strchr("()| \r\n\t", s[1])) {
2193 Perl_ck_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
2194 "Possible unintended interpolation of $\\ in regex");
2196 break; /* in regexp, $ might be tail anchor */
2200 /* End of else if chain - OP_TRANS rejoin rest */
2203 if (*s == '\\' && s+1 < send) {
2206 /* deprecate \1 in strings and substitution replacements */
2207 if (PL_lex_inwhat == OP_SUBST && !PL_lex_inpat &&
2208 isDIGIT(*s) && *s != '0' && !isDIGIT(s[1]))
2210 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX), "\\%c better written as $%c", *s, *s);
2215 /* string-change backslash escapes */
2216 if (PL_lex_inwhat != OP_TRANS && *s && strchr("lLuUEQ", *s)) {
2220 /* skip any other backslash escapes in a pattern */
2221 else if (PL_lex_inpat) {
2222 *d++ = NATIVE_TO_NEED(has_utf8,'\\');
2223 goto default_action;
2226 /* if we get here, it's either a quoted -, or a digit */
2229 /* quoted - in transliterations */
2231 if (PL_lex_inwhat == OP_TRANS) {
2238 if ((isALPHA(*s) || isDIGIT(*s)))
2239 Perl_ck_warner(aTHX_ packWARN(WARN_MISC),
2240 "Unrecognized escape \\%c passed through",
2242 /* default action is to copy the quoted character */
2243 goto default_action;
2246 /* eg. \132 indicates the octal constant 0x132 */
2247 case '0': case '1': case '2': case '3':
2248 case '4': case '5': case '6': case '7':
2252 uv = NATIVE_TO_UNI(grok_oct(s, &len, &flags, NULL));
2255 goto NUM_ESCAPE_INSERT;
2257 /* eg. \x24 indicates the hex constant 0x24 */
2261 char* const e = strchr(s, '}');
2262 I32 flags = PERL_SCAN_ALLOW_UNDERSCORES |
2263 PERL_SCAN_DISALLOW_PREFIX;
2268 yyerror("Missing right brace on \\x{}");
2272 uv = NATIVE_TO_UNI(grok_hex(s, &len, &flags, NULL));
2278 I32 flags = PERL_SCAN_DISALLOW_PREFIX;
2279 uv = NATIVE_TO_UNI(grok_hex(s, &len, &flags, NULL));
2285 /* Insert oct, hex, or \N{U+...} escaped character. There will
2286 * always be enough room in sv since such escapes will be
2287 * longer than any UTF-8 sequence they can end up as, except if
2288 * they force us to recode the rest of the string into utf8 */
2290 /* Here uv is the ordinal of the next character being added in
2291 * unicode (converted from native). (It has to be done before
2292 * here because \N is interpreted as unicode, and oct and hex
2294 if (!UNI_IS_INVARIANT(uv)) {
2295 if (!has_utf8 && uv > 255) {
2296 /* Might need to recode whatever we have accumulated so
2297 * far if it contains any chars variant in utf8 or
2300 SvCUR_set(sv, d - SvPVX_const(sv));
2303 /* See Note on sizing above. */
2304 sv_utf8_upgrade_flags_grow(sv,
2305 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
2306 UNISKIP(uv) + (STRLEN)(send - s) + 1);
2307 d = SvPVX(sv) + SvCUR(sv);
2312 d = (char*)uvuni_to_utf8((U8*)d, uv);
2313 if (PL_lex_inwhat == OP_TRANS &&
2314 PL_sublex_info.sub_op) {
2315 PL_sublex_info.sub_op->op_private |=
2316 (PL_lex_repl ? OPpTRANS_FROM_UTF
2320 if (uv > 255 && !dorange)
2321 native_range = FALSE;
2333 /* \N{LATIN SMALL LETTER A} is a named character, and so is
2338 char* e = strchr(s, '}');
2344 yyerror("Missing right brace on \\N{}");
2348 if (e > s + 2 && s[1] == 'U' && s[2] == '+') {
2349 /* \N{U+...} The ... is a unicode value even on EBCDIC
2351 I32 flags = PERL_SCAN_ALLOW_UNDERSCORES |
2352 PERL_SCAN_DISALLOW_PREFIX;
2355 uv = grok_hex(s, &len, &flags, NULL);
2356 if ( e > s && len != (STRLEN)(e - s) ) {
2360 goto NUM_ESCAPE_INSERT;
2362 res = newSVpvn(s + 1, e - s - 1);
2363 res = new_constant( NULL, 0, "charnames",
2364 res, NULL, s - 2, e - s + 3 );
2366 sv_utf8_upgrade(res);
2367 str = SvPV_const(res,len);
2368 #ifdef EBCDIC_NEVER_MIND
2369 /* charnames uses pack U and that has been
2370 * recently changed to do the below uni->native
2371 * mapping, so this would be redundant (and wrong,
2372 * the code point would be doubly converted).
2373 * But leave this in just in case the pack U change
2374 * gets revoked, but the semantics is still
2375 * desireable for charnames. --jhi */
2377 UV uv = utf8_to_uvchr((const U8*)str, 0);
2380 U8 tmpbuf[UTF8_MAXBYTES+1], *d;
2382 d = uvchr_to_utf8(tmpbuf, UNI_TO_NATIVE(uv));
2383 sv_setpvn(res, (char *)tmpbuf, d - tmpbuf);
2384 str = SvPV_const(res, len);
2388 /* If destination is not in utf8 but this new character is,
2389 * recode the dest to utf8 */
2390 if (!has_utf8 && SvUTF8(res)) {
2391 SvCUR_set(sv, d - SvPVX_const(sv));
2394 /* See Note on sizing above. */
2395 sv_utf8_upgrade_flags_grow(sv,
2396 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
2397 len + (STRLEN)(send - s) + 1);
2398 d = SvPVX(sv) + SvCUR(sv);
2400 } else if (len > (STRLEN)(e - s + 4)) { /* I _guess_ 4 is \N{} --jhi */
2402 /* See Note on sizing above. (NOTE: SvCUR() is not set
2403 * correctly here). */
2404 const STRLEN off = d - SvPVX_const(sv);
2405 d = SvGROW(sv, off + len + (STRLEN)(send - s) + 1) + off;
2409 native_range = FALSE; /* \N{} is guessed to be Unicode */
2411 Copy(str, d, len, char);
2418 yyerror("Missing braces on \\N{}");
2421 /* \c is a control character */
2430 *d++ = NATIVE_TO_NEED(has_utf8,toCTRL(c));
2433 yyerror("Missing control char name in \\c");
2437 /* printf-style backslashes, formfeeds, newlines, etc */
2439 *d++ = NATIVE_TO_NEED(has_utf8,'\b');
2442 *d++ = NATIVE_TO_NEED(has_utf8,'\n');
2445 *d++ = NATIVE_TO_NEED(has_utf8,'\r');
2448 *d++ = NATIVE_TO_NEED(has_utf8,'\f');
2451 *d++ = NATIVE_TO_NEED(has_utf8,'\t');
2454 *d++ = ASCII_TO_NEED(has_utf8,'\033');
2457 *d++ = ASCII_TO_NEED(has_utf8,'\007');
2463 } /* end if (backslash) */
2470 /* If we started with encoded form, or already know we want it,
2471 then encode the next character */
2472 if (! NATIVE_IS_INVARIANT((U8)(*s)) && (this_utf8 || has_utf8)) {
2476 /* One might think that it is wasted effort in the case of the
2477 * source being utf8 (this_utf8 == TRUE) to take the next character
2478 * in the source, convert it to an unsigned value, and then convert
2479 * it back again. But the source has not been validated here. The
2480 * routine that does the conversion checks for errors like
2483 const UV nextuv = (this_utf8) ? utf8n_to_uvchr((U8*)s, send - s, &len, 0) : (UV) ((U8) *s);
2484 const STRLEN need = UNISKIP(NATIVE_TO_UNI(nextuv));
2486 SvCUR_set(sv, d - SvPVX_const(sv));
2489 /* See Note on sizing above. */
2490 sv_utf8_upgrade_flags_grow(sv,
2491 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
2492 need + (STRLEN)(send - s) + 1);
2493 d = SvPVX(sv) + SvCUR(sv);
2495 } else if (need > len) {
2496 /* encoded value larger than old, may need extra space (NOTE:
2497 * SvCUR() is not set correctly here). See Note on sizing
2499 const STRLEN off = d - SvPVX_const(sv);
2500 d = SvGROW(sv, off + need + (STRLEN)(send - s) + 1) + off;
2504 d = (char*)uvchr_to_utf8((U8*)d, nextuv);
2506 if (uv > 255 && !dorange)
2507 native_range = FALSE;
2511 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2513 } /* while loop to process each character */
2515 /* terminate the string and set up the sv */
2517 SvCUR_set(sv, d - SvPVX_const(sv));
2518 if (SvCUR(sv) >= SvLEN(sv))
2519 Perl_croak(aTHX_ "panic: constant overflowed allocated space");
2522 if (PL_encoding && !has_utf8) {
2523 sv_recode_to_utf8(sv, PL_encoding);
2529 if (PL_lex_inwhat == OP_TRANS && PL_sublex_info.sub_op) {
2530 PL_sublex_info.sub_op->op_private |=
2531 (PL_lex_repl ? OPpTRANS_FROM_UTF : OPpTRANS_TO_UTF);
2535 /* shrink the sv if we allocated more than we used */
2536 if (SvCUR(sv) + 5 < SvLEN(sv)) {
2537 SvPV_shrink_to_cur(sv);
2540 /* return the substring (via pl_yylval) only if we parsed anything */
2541 if (s > PL_bufptr) {
2542 if ( PL_hints & ( PL_lex_inpat ? HINT_NEW_RE : HINT_NEW_STRING ) ) {
2543 const char *const key = PL_lex_inpat ? "qr" : "q";
2544 const STRLEN keylen = PL_lex_inpat ? 2 : 1;
2548 if (PL_lex_inwhat == OP_TRANS) {
2551 } else if (PL_lex_inwhat == OP_SUBST && !PL_lex_inpat) {
2559 sv = S_new_constant(aTHX_ start, s - start, key, keylen, sv, NULL,
2562 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
2569 * Returns TRUE if there's more to the expression (e.g., a subscript),
2572 * It deals with "$foo[3]" and /$foo[3]/ and /$foo[0123456789$]+/
2574 * ->[ and ->{ return TRUE
2575 * { and [ outside a pattern are always subscripts, so return TRUE
2576 * if we're outside a pattern and it's not { or [, then return FALSE
2577 * if we're in a pattern and the first char is a {
2578 * {4,5} (any digits around the comma) returns FALSE
2579 * if we're in a pattern and the first char is a [
2581 * [SOMETHING] has a funky algorithm to decide whether it's a
2582 * character class or not. It has to deal with things like
2583 * /$foo[-3]/ and /$foo[$bar]/ as well as /$foo[$\d]+/
2584 * anything else returns TRUE
2587 /* This is the one truly awful dwimmer necessary to conflate C and sed. */
2590 S_intuit_more(pTHX_ register char *s)
2594 PERL_ARGS_ASSERT_INTUIT_MORE;
2596 if (PL_lex_brackets)
2598 if (*s == '-' && s[1] == '>' && (s[2] == '[' || s[2] == '{'))
2600 if (*s != '{' && *s != '[')
2605 /* In a pattern, so maybe we have {n,m}. */
2622 /* On the other hand, maybe we have a character class */
2625 if (*s == ']' || *s == '^')
2628 /* this is terrifying, and it works */
2629 int weight = 2; /* let's weigh the evidence */
2631 unsigned char un_char = 255, last_un_char;
2632 const char * const send = strchr(s,']');
2633 char tmpbuf[sizeof PL_tokenbuf * 4];
2635 if (!send) /* has to be an expression */
2638 Zero(seen,256,char);
2641 else if (isDIGIT(*s)) {
2643 if (isDIGIT(s[1]) && s[2] == ']')
2649 for (; s < send; s++) {
2650 last_un_char = un_char;
2651 un_char = (unsigned char)*s;
2656 weight -= seen[un_char] * 10;
2657 if (isALNUM_lazy_if(s+1,UTF)) {
2659 scan_ident(s, send, tmpbuf, sizeof tmpbuf, FALSE);
2660 len = (int)strlen(tmpbuf);
2661 if (len > 1 && gv_fetchpvn_flags(tmpbuf, len, 0, SVt_PV))
2666 else if (*s == '$' && s[1] &&
2667 strchr("[#!%*<>()-=",s[1])) {
2668 if (/*{*/ strchr("])} =",s[2]))
2677 if (strchr("wds]",s[1]))
2679 else if (seen[(U8)'\''] || seen[(U8)'"'])
2681 else if (strchr("rnftbxcav",s[1]))
2683 else if (isDIGIT(s[1])) {
2685 while (s[1] && isDIGIT(s[1]))
2695 if (strchr("aA01! ",last_un_char))
2697 if (strchr("zZ79~",s[1]))
2699 if (last_un_char == 255 && (isDIGIT(s[1]) || s[1] == '$'))
2700 weight -= 5; /* cope with negative subscript */
2703 if (!isALNUM(last_un_char)
2704 && !(last_un_char == '$' || last_un_char == '@'
2705 || last_un_char == '&')
2706 && isALPHA(*s) && s[1] && isALPHA(s[1])) {
2711 if (keyword(tmpbuf, d - tmpbuf, 0))
2714 if (un_char == last_un_char + 1)
2716 weight -= seen[un_char];
2721 if (weight >= 0) /* probably a character class */
2731 * Does all the checking to disambiguate
2733 * between foo(bar) and bar->foo. Returns 0 if not a method, otherwise
2734 * FUNCMETH (bar->foo(args)) or METHOD (bar->foo args).
2736 * First argument is the stuff after the first token, e.g. "bar".
2738 * Not a method if bar is a filehandle.
2739 * Not a method if foo is a subroutine prototyped to take a filehandle.
2740 * Not a method if it's really "Foo $bar"
2741 * Method if it's "foo $bar"
2742 * Not a method if it's really "print foo $bar"
2743 * Method if it's really "foo package::" (interpreted as package->foo)
2744 * Not a method if bar is known to be a subroutine ("sub bar; foo bar")
2745 * Not a method if bar is a filehandle or package, but is quoted with
2750 S_intuit_method(pTHX_ char *start, GV *gv, CV *cv)
2753 char *s = start + (*start == '$');
2754 char tmpbuf[sizeof PL_tokenbuf];
2761 PERL_ARGS_ASSERT_INTUIT_METHOD;
2764 if (SvTYPE(gv) == SVt_PVGV && GvIO(gv))
2768 const char *proto = SvPVX_const(cv);
2779 s = scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
2780 /* start is the beginning of the possible filehandle/object,
2781 * and s is the end of it
2782 * tmpbuf is a copy of it
2785 if (*start == '$') {
2786 if (gv || PL_last_lop_op == OP_PRINT || PL_last_lop_op == OP_SAY ||
2787 isUPPER(*PL_tokenbuf))
2790 len = start - SvPVX(PL_linestr);
2794 start = SvPVX(PL_linestr) + len;
2798 return *s == '(' ? FUNCMETH : METHOD;
2800 if (!keyword(tmpbuf, len, 0)) {
2801 if (len > 2 && tmpbuf[len - 2] == ':' && tmpbuf[len - 1] == ':') {
2805 soff = s - SvPVX(PL_linestr);
2809 indirgv = gv_fetchpvn_flags(tmpbuf, len, 0, SVt_PVCV);
2810 if (indirgv && GvCVu(indirgv))
2812 /* filehandle or package name makes it a method */
2813 if (!gv || GvIO(indirgv) || gv_stashpvn(tmpbuf, len, 0)) {
2815 soff = s - SvPVX(PL_linestr);
2818 if ((PL_bufend - s) >= 2 && *s == '=' && *(s+1) == '>')
2819 return 0; /* no assumptions -- "=>" quotes bearword */
2821 start_force(PL_curforce);
2822 NEXTVAL_NEXTTOKE.opval = (OP*)newSVOP(OP_CONST, 0,
2823 S_newSV_maybe_utf8(aTHX_ tmpbuf, len));
2824 NEXTVAL_NEXTTOKE.opval->op_private = OPpCONST_BARE;
2826 curmad('X', newSVpvn(start,SvPVX(PL_linestr) + soff - start));
2831 PL_bufptr = SvPVX(PL_linestr) + soff; /* restart before space */
2833 return *s == '(' ? FUNCMETH : METHOD;
2839 /* Encoded script support. filter_add() effectively inserts a
2840 * 'pre-processing' function into the current source input stream.
2841 * Note that the filter function only applies to the current source file
2842 * (e.g., it will not affect files 'require'd or 'use'd by this one).
2844 * The datasv parameter (which may be NULL) can be used to pass
2845 * private data to this instance of the filter. The filter function
2846 * can recover the SV using the FILTER_DATA macro and use it to
2847 * store private buffers and state information.
2849 * The supplied datasv parameter is upgraded to a PVIO type
2850 * and the IoDIRP/IoANY field is used to store the function pointer,
2851 * and IOf_FAKE_DIRP is enabled on datasv to mark this as such.
2852 * Note that IoTOP_NAME, IoFMT_NAME, IoBOTTOM_NAME, if set for
2853 * private use must be set using malloc'd pointers.
2857 Perl_filter_add(pTHX_ filter_t funcp, SV *datasv)
2866 if (!PL_rsfp_filters)
2867 PL_rsfp_filters = newAV();
2870 SvUPGRADE(datasv, SVt_PVIO);
2871 IoANY(datasv) = FPTR2DPTR(void *, funcp); /* stash funcp into spare field */
2872 IoFLAGS(datasv) |= IOf_FAKE_DIRP;
2873 DEBUG_P(PerlIO_printf(Perl_debug_log, "filter_add func %p (%s)\n",
2874 FPTR2DPTR(void *, IoANY(datasv)),
2875 SvPV_nolen(datasv)));
2876 av_unshift(PL_rsfp_filters, 1);
2877 av_store(PL_rsfp_filters, 0, datasv) ;
2882 /* Delete most recently added instance of this filter function. */
2884 Perl_filter_del(pTHX_ filter_t funcp)
2889 PERL_ARGS_ASSERT_FILTER_DEL;
2892 DEBUG_P(PerlIO_printf(Perl_debug_log, "filter_del func %p",
2893 FPTR2DPTR(void*, funcp)));
2895 if (!PL_parser || !PL_rsfp_filters || AvFILLp(PL_rsfp_filters)<0)
2897 /* if filter is on top of stack (usual case) just pop it off */
2898 datasv = FILTER_DATA(AvFILLp(PL_rsfp_filters));
2899 if (IoANY(datasv) == FPTR2DPTR(void *, funcp)) {
2900 IoFLAGS(datasv) &= ~IOf_FAKE_DIRP;
2901 IoANY(datasv) = (void *)NULL;
2902 sv_free(av_pop(PL_rsfp_filters));
2906 /* we need to search for the correct entry and clear it */
2907 Perl_die(aTHX_ "filter_del can only delete in reverse order (currently)");
2911 /* Invoke the idxth filter function for the current rsfp. */
2912 /* maxlen 0 = read one text line */
2914 Perl_filter_read(pTHX_ int idx, SV *buf_sv, int maxlen)
2919 /* This API is bad. It should have been using unsigned int for maxlen.
2920 Not sure if we want to change the API, but if not we should sanity
2921 check the value here. */
2922 const unsigned int correct_length
2931 PERL_ARGS_ASSERT_FILTER_READ;
2933 if (!PL_parser || !PL_rsfp_filters)
2935 if (idx > AvFILLp(PL_rsfp_filters)) { /* Any more filters? */
2936 /* Provide a default input filter to make life easy. */
2937 /* Note that we append to the line. This is handy. */
2938 DEBUG_P(PerlIO_printf(Perl_debug_log,
2939 "filter_read %d: from rsfp\n", idx));
2940 if (correct_length) {
2943 const int old_len = SvCUR(buf_sv);
2945 /* ensure buf_sv is large enough */
2946 SvGROW(buf_sv, (STRLEN)(old_len + correct_length)) ;
2947 if ((len = PerlIO_read(PL_rsfp, SvPVX(buf_sv) + old_len,
2948 correct_length)) <= 0) {
2949 if (PerlIO_error(PL_rsfp))
2950 return -1; /* error */
2952 return 0 ; /* end of file */
2954 SvCUR_set(buf_sv, old_len + len) ;
2957 if (sv_gets(buf_sv, PL_rsfp, SvCUR(buf_sv)) == NULL) {
2958 if (PerlIO_error(PL_rsfp))
2959 return -1; /* error */
2961 return 0 ; /* end of file */
2964 return SvCUR(buf_sv);
2966 /* Skip this filter slot if filter has been deleted */
2967 if ( (datasv = FILTER_DATA(idx)) == &PL_sv_undef) {
2968 DEBUG_P(PerlIO_printf(Perl_debug_log,
2969 "filter_read %d: skipped (filter deleted)\n",
2971 return FILTER_READ(idx+1, buf_sv, correct_length); /* recurse */
2973 /* Get function pointer hidden within datasv */
2974 funcp = DPTR2FPTR(filter_t, IoANY(datasv));
2975 DEBUG_P(PerlIO_printf(Perl_debug_log,
2976 "filter_read %d: via function %p (%s)\n",
2977 idx, (void*)datasv, SvPV_nolen_const(datasv)));
2978 /* Call function. The function is expected to */
2979 /* call "FILTER_READ(idx+1, buf_sv)" first. */
2980 /* Return: <0:error, =0:eof, >0:not eof */
2981 return (*funcp)(aTHX_ idx, buf_sv, correct_length);
2985 S_filter_gets(pTHX_ register SV *sv, register PerlIO *fp, STRLEN append)
2989 PERL_ARGS_ASSERT_FILTER_GETS;
2991 #ifdef PERL_CR_FILTER
2992 if (!PL_rsfp_filters) {
2993 filter_add(S_cr_textfilter,NULL);
2996 if (PL_rsfp_filters) {
2998 SvCUR_set(sv, 0); /* start with empty line */
2999 if (FILTER_READ(0, sv, 0) > 0)
3000 return ( SvPVX(sv) ) ;
3005 return (sv_gets(sv, fp, append));
3009 S_find_in_my_stash(pTHX_ const char *pkgname, STRLEN len)
3014 PERL_ARGS_ASSERT_FIND_IN_MY_STASH;
3016 if (len == 11 && *pkgname == '_' && strEQ(pkgname, "__PACKAGE__"))
3020 (pkgname[len - 2] == ':' && pkgname[len - 1] == ':') &&
3021 (gv = gv_fetchpvn_flags(pkgname, len, 0, SVt_PVHV)))
3023 return GvHV(gv); /* Foo:: */
3026 /* use constant CLASS => 'MyClass' */
3027 gv = gv_fetchpvn_flags(pkgname, len, 0, SVt_PVCV);
3028 if (gv && GvCV(gv)) {
3029 SV * const sv = cv_const_sv(GvCV(gv));
3031 pkgname = SvPV_const(sv, len);
3034 return gv_stashpvn(pkgname, len, 0);
3038 * S_readpipe_override
3039 * Check whether readpipe() is overriden, and generates the appropriate
3040 * optree, provided sublex_start() is called afterwards.
3043 S_readpipe_override(pTHX)
3046 GV *gv_readpipe = gv_fetchpvs("readpipe", GV_NOTQUAL, SVt_PVCV);
3047 pl_yylval.ival = OP_BACKTICK;
3049 && GvCVu(gv_readpipe) && GvIMPORTED_CV(gv_readpipe))
3051 ((gvp = (GV**)hv_fetchs(PL_globalstash, "readpipe", FALSE))
3052 && (gv_readpipe = *gvp) && isGV_with_GP(gv_readpipe)
3053 && GvCVu(gv_readpipe) && GvIMPORTED_CV(gv_readpipe)))
3055 PL_lex_op = (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
3056 append_elem(OP_LIST,
3057 newSVOP(OP_CONST, 0, &PL_sv_undef), /* value will be read later */
3058 newCVREF(0, newGVOP(OP_GV, 0, gv_readpipe))));
3065 * The intent of this yylex wrapper is to minimize the changes to the
3066 * tokener when we aren't interested in collecting madprops. It remains
3067 * to be seen how successful this strategy will be...
3074 char *s = PL_bufptr;
3076 /* make sure PL_thiswhite is initialized */
3080 /* just do what yylex would do on pending identifier; leave PL_thiswhite alone */
3081 if (PL_pending_ident)
3082 return S_pending_ident(aTHX);
3084 /* previous token ate up our whitespace? */
3085 if (!PL_lasttoke && PL_nextwhite) {
3086 PL_thiswhite = PL_nextwhite;
3090 /* isolate the token, and figure out where it is without whitespace */
3091 PL_realtokenstart = -1;
3095 assert(PL_curforce < 0);
3097 if (!PL_thismad || PL_thismad->mad_key == '^') { /* not forced already? */
3098 if (!PL_thistoken) {
3099 if (PL_realtokenstart < 0 || !CopLINE(PL_curcop))
3100 PL_thistoken = newSVpvs("");
3102 char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
3103 PL_thistoken = newSVpvn(tstart, s - tstart);
3106 if (PL_thismad) /* install head */
3107 CURMAD('X', PL_thistoken);
3110 /* last whitespace of a sublex? */
3111 if (optype == ')' && PL_endwhite) {
3112 CURMAD('X', PL_endwhite);
3117 /* if no whitespace and we're at EOF, bail. Otherwise fake EOF below. */
3118 if (!PL_thiswhite && !PL_endwhite && !optype) {
3119 sv_free(PL_thistoken);
3124 /* put off final whitespace till peg */
3125 if (optype == ';' && !PL_rsfp) {
3126 PL_nextwhite = PL_thiswhite;
3129 else if (PL_thisopen) {
3130 CURMAD('q', PL_thisopen);
3132 sv_free(PL_thistoken);
3136 /* Store actual token text as madprop X */
3137 CURMAD('X', PL_thistoken);
3141 /* add preceding whitespace as madprop _ */
3142 CURMAD('_', PL_thiswhite);
3146 /* add quoted material as madprop = */
3147 CURMAD('=', PL_thisstuff);
3151 /* add terminating quote as madprop Q */
3152 CURMAD('Q', PL_thisclose);
3156 /* special processing based on optype */
3160 /* opval doesn't need a TOKEN since it can already store mp */
3170 if (pl_yylval.opval)
3171 append_madprops(PL_thismad, pl_yylval.opval, 0);
3179 addmad(newMADsv('p', PL_endwhite), &PL_thismad, 0);
3188 /* remember any fake bracket that lexer is about to discard */
3189 if (PL_lex_brackets == 1 &&
3190 ((expectation)PL_lex_brackstack[0] & XFAKEBRACK))
3193 while (s < PL_bufend && (*s == ' ' || *s == '\t'))
3196 PL_thiswhite = newSVpvn(PL_bufptr, ++s - PL_bufptr);
3197 addmad(newMADsv('#', PL_thiswhite), &PL_thismad, 0);
3200 break; /* don't bother looking for trailing comment */
3209 /* attach a trailing comment to its statement instead of next token */
3213 if (PL_bufptr > PL_oldbufptr && PL_bufptr[-1] == optype) {
3215 while (s < PL_bufend && (*s == ' ' || *s == '\t'))
3217 if (*s == '\n' || *s == '#') {
3218 while (s < PL_bufend && *s != '\n')
3222 PL_thiswhite = newSVpvn(PL_bufptr, s - PL_bufptr);
3223 addmad(newMADsv('#', PL_thiswhite), &PL_thismad, 0);
3240 /* Create new token struct. Note: opvals return early above. */
3241 pl_yylval.tkval = newTOKEN(optype, pl_yylval, PL_thismad);
3248 S_tokenize_use(pTHX_ int is_use, char *s) {
3251 PERL_ARGS_ASSERT_TOKENIZE_USE;
3253 if (PL_expect != XSTATE)
3254 yyerror(Perl_form(aTHX_ "\"%s\" not allowed in expression",
3255 is_use ? "use" : "no"));
3257 if (isDIGIT(*s) || (*s == 'v' && isDIGIT(s[1]))) {
3258 s = force_version(s, TRUE);
3259 if (*s == ';' || (s = SKIPSPACE1(s), *s == ';')) {
3260 start_force(PL_curforce);
3261 NEXTVAL_NEXTTOKE.opval = NULL;
3264 else if (*s == 'v') {
3265 s = force_word(s,WORD,FALSE,TRUE,FALSE);
3266 s = force_version(s, FALSE);
3270 s = force_word(s,WORD,FALSE,TRUE,FALSE);
3271 s = force_version(s, FALSE);
3273 pl_yylval.ival = is_use;
3277 static const char* const exp_name[] =
3278 { "OPERATOR", "TERM", "REF", "STATE", "BLOCK", "ATTRBLOCK",
3279 "ATTRTERM", "TERMBLOCK", "TERMORDORDOR"
3286 Works out what to call the token just pulled out of the input
3287 stream. The yacc parser takes care of taking the ops we return and
3288 stitching them into a tree.
3294 if read an identifier
3295 if we're in a my declaration
3296 croak if they tried to say my($foo::bar)
3297 build the ops for a my() declaration
3298 if it's an access to a my() variable
3299 are we in a sort block?
3300 croak if my($a); $a <=> $b
3301 build ops for access to a my() variable
3302 if in a dq string, and they've said @foo and we can't find @foo
3304 build ops for a bareword
3305 if we already built the token before, use it.
3310 #pragma segment Perl_yylex
3316 register char *s = PL_bufptr;
3321 /* orig_keyword, gvp, and gv are initialized here because
3322 * jump to the label just_a_word_zero can bypass their
3323 * initialization later. */
3324 I32 orig_keyword = 0;
3329 SV* tmp = newSVpvs("");
3330 PerlIO_printf(Perl_debug_log, "### %"IVdf":LEX_%s/X%s %s\n",
3331 (IV)CopLINE(PL_curcop),
3332 lex_state_names[PL_lex_state],
3333 exp_name[PL_expect],
3334 pv_display(tmp, s, strlen(s), 0, 60));
3337 /* check if there's an identifier for us to look at */
3338 if (PL_pending_ident)
3339 return REPORT(S_pending_ident(aTHX));
3341 /* no identifier pending identification */
3343 switch (PL_lex_state) {
3345 case LEX_NORMAL: /* Some compilers will produce faster */
3346 case LEX_INTERPNORMAL: /* code if we comment these out. */
3350 /* when we've already built the next token, just pull it out of the queue */
3354 pl_yylval = PL_nexttoke[PL_lasttoke].next_val;
3356 PL_thismad = PL_nexttoke[PL_lasttoke].next_mad;
3357 PL_nexttoke[PL_lasttoke].next_mad = 0;
3358 if (PL_thismad && PL_thismad->mad_key == '_') {
3359 PL_thiswhite = MUTABLE_SV(PL_thismad->mad_val);
3360 PL_thismad->mad_val = 0;
3361 mad_free(PL_thismad);
3366 PL_lex_state = PL_lex_defer;
3367 PL_expect = PL_lex_expect;
3368 PL_lex_defer = LEX_NORMAL;
3369 if (!PL_nexttoke[PL_lasttoke].next_type)
3374 pl_yylval = PL_nextval[PL_nexttoke];
3376 PL_lex_state = PL_lex_defer;
3377 PL_expect = PL_lex_expect;
3378 PL_lex_defer = LEX_NORMAL;
3382 /* FIXME - can these be merged? */
3383 return(PL_nexttoke[PL_lasttoke].next_type);
3385 return REPORT(PL_nexttype[PL_nexttoke]);
3388 /* interpolated case modifiers like \L \U, including \Q and \E.
3389 when we get here, PL_bufptr is at the \
3391 case LEX_INTERPCASEMOD:
3393 if (PL_bufptr != PL_bufend && *PL_bufptr != '\\')
3394 Perl_croak(aTHX_ "panic: INTERPCASEMOD");
3396 /* handle \E or end of string */
3397 if (PL_bufptr == PL_bufend || PL_bufptr[1] == 'E') {
3399 if (PL_lex_casemods) {
3400 const char oldmod = PL_lex_casestack[--PL_lex_casemods];
3401 PL_lex_casestack[PL_lex_casemods] = '\0';
3403 if (PL_bufptr != PL_bufend
3404 && (oldmod == 'L' || oldmod == 'U' || oldmod == 'Q')) {
3406 PL_lex_state = LEX_INTERPCONCAT;
3409 PL_thistoken = newSVpvs("\\E");
3415 while (PL_bufptr != PL_bufend &&
3416 PL_bufptr[0] == '\\' && PL_bufptr[1] == 'E') {
3418 PL_thiswhite = newSVpvs("");
3419 sv_catpvn(PL_thiswhite, PL_bufptr, 2);
3423 if (PL_bufptr != PL_bufend)
3426 PL_lex_state = LEX_INTERPCONCAT;
3430 DEBUG_T({ PerlIO_printf(Perl_debug_log,
3431 "### Saw case modifier\n"); });
3433 if (s[1] == '\\' && s[2] == 'E') {
3436 PL_thiswhite = newSVpvs("");
3437 sv_catpvn(PL_thiswhite, PL_bufptr, 4);
3440 PL_lex_state = LEX_INTERPCONCAT;
3445 if (!PL_madskills) /* when just compiling don't need correct */
3446 if (strnEQ(s, "L\\u", 3) || strnEQ(s, "U\\l", 3))
3447 tmp = *s, *s = s[2], s[2] = (char)tmp; /* misordered... */
3448 if ((*s == 'L' || *s == 'U') &&
3449 (strchr(PL_lex_casestack, 'L') || strchr(PL_lex_casestack, 'U'))) {
3450 PL_lex_casestack[--PL_lex_casemods] = '\0';
3453 if (PL_lex_casemods > 10)
3454 Renew(PL_lex_casestack, PL_lex_casemods + 2, char);
3455 PL_lex_casestack[PL_lex_casemods++] = *s;
3456 PL_lex_casestack[PL_lex_casemods] = '\0';
3457 PL_lex_state = LEX_INTERPCONCAT;
3458 start_force(PL_curforce);
3459 NEXTVAL_NEXTTOKE.ival = 0;
3461 start_force(PL_curforce);
3463 NEXTVAL_NEXTTOKE.ival = OP_LCFIRST;
3465 NEXTVAL_NEXTTOKE.ival = OP_UCFIRST;
3467 NEXTVAL_NEXTTOKE.ival = OP_LC;
3469 NEXTVAL_NEXTTOKE.ival = OP_UC;
3471 NEXTVAL_NEXTTOKE.ival = OP_QUOTEMETA;
3473 Perl_croak(aTHX_ "panic: yylex");
3475 SV* const tmpsv = newSVpvs("\\ ");
3476 /* replace the space with the character we want to escape
3478 SvPVX(tmpsv)[1] = *s;
3484 if (PL_lex_starts) {
3490 sv_free(PL_thistoken);
3491 PL_thistoken = newSVpvs("");
3494 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
3495 if (PL_lex_casemods == 1 && PL_lex_inpat)
3504 case LEX_INTERPPUSH:
3505 return REPORT(sublex_push());
3507 case LEX_INTERPSTART:
3508 if (PL_bufptr == PL_bufend)
3509 return REPORT(sublex_done());
3510 DEBUG_T({ PerlIO_printf(Perl_debug_log,
3511 "### Interpolated variable\n"); });
3513 PL_lex_dojoin = (*PL_bufptr == '@');
3514 PL_lex_state = LEX_INTERPNORMAL;
3515 if (PL_lex_dojoin) {
3516 start_force(PL_curforce);
3517 NEXTVAL_NEXTTOKE.ival = 0;
3519 start_force(PL_curforce);
3520 force_ident("\"", '$');
3521 start_force(PL_curforce);
3522 NEXTVAL_NEXTTOKE.ival = 0;
3524 start_force(PL_curforce);
3525 NEXTVAL_NEXTTOKE.ival = 0;
3527 start_force(PL_curforce);
3528 NEXTVAL_NEXTTOKE.ival = OP_JOIN; /* emulate join($", ...) */
3531 if (PL_lex_starts++) {
3536 sv_free(PL_thistoken);
3537 PL_thistoken = newSVpvs("");
3540 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
3541 if (!PL_lex_casemods && PL_lex_inpat)
3548 case LEX_INTERPENDMAYBE:
3549 if (intuit_more(PL_bufptr)) {
3550 PL_lex_state = LEX_INTERPNORMAL; /* false alarm, more expr */
3556 if (PL_lex_dojoin) {
3557 PL_lex_dojoin = FALSE;
3558 PL_lex_state = LEX_INTERPCONCAT;
3562 sv_free(PL_thistoken);
3563 PL_thistoken = newSVpvs("");
3568 if (PL_lex_inwhat == OP_SUBST && PL_linestr == PL_lex_repl
3569 && SvEVALED(PL_lex_repl))
3571 if (PL_bufptr != PL_bufend)
3572 Perl_croak(aTHX_ "Bad evalled substitution pattern");
3576 case LEX_INTERPCONCAT:
3578 if (PL_lex_brackets)
3579 Perl_croak(aTHX_ "panic: INTERPCONCAT");
3581 if (PL_bufptr == PL_bufend)
3582 return REPORT(sublex_done());
3584 if (SvIVX(PL_linestr) == '\'') {
3585 SV *sv = newSVsv(PL_linestr);
3588 else if ( PL_hints & HINT_NEW_RE )
3589 sv = new_constant(NULL, 0, "qr", sv, sv, "q", 1);
3590 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
3594 s = scan_const(PL_bufptr);
3596 PL_lex_state = LEX_INTERPCASEMOD;
3598 PL_lex_state = LEX_INTERPSTART;
3601 if (s != PL_bufptr) {
3602 start_force(PL_curforce);
3604 curmad('X', newSVpvn(PL_bufptr,s-PL_bufptr));
3606 NEXTVAL_NEXTTOKE = pl_yylval;
3609 if (PL_lex_starts++) {
3613 sv_free(PL_thistoken);
3614 PL_thistoken = newSVpvs("");
3617 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
3618 if (!PL_lex_casemods && PL_lex_inpat)
3631 PL_lex_state = LEX_NORMAL;
3632 s = scan_formline(PL_bufptr);
3633 if (!PL_lex_formbrack)
3639 PL_oldoldbufptr = PL_oldbufptr;
3645 sv_free(PL_thistoken);
3648 PL_realtokenstart = s - SvPVX(PL_linestr); /* assume but undo on ws */
3652 if (isIDFIRST_lazy_if(s,UTF))
3655 unsigned char c = *s;
3656 len = UTF ? Perl_utf8_length(aTHX_ (U8 *) PL_linestart, (U8 *) s) : (STRLEN) (s - PL_linestart);
3657 if (len > UNRECOGNIZED_PRECEDE_COUNT) {
3658 d = UTF ? (char *) Perl_utf8_hop(aTHX_ (U8 *) s, -UNRECOGNIZED_PRECEDE_COUNT) : s - UNRECOGNIZED_PRECEDE_COUNT;
3663 Perl_croak(aTHX_ "Unrecognized character \\x%02X; marked by <-- HERE after %s<-- HERE near column %d", c, d, (int) len + 1);
3667 goto fake_eof; /* emulate EOF on ^D or ^Z */
3676 if (PL_lex_brackets) {
3677 yyerror((const char *)
3679 ? "Format not terminated"
3680 : "Missing right curly or square bracket"));
3682 DEBUG_T( { PerlIO_printf(Perl_debug_log,
3683 "### Tokener got EOF\n");
3687 if (s++ < PL_bufend)
3688 goto retry; /* ignore stray nulls */
3691 if (!PL_in_eval && !PL_preambled) {
3692 PL_preambled = TRUE;
3698 /* Generate a string of Perl code to load the debugger.
3699 * If PERL5DB is set, it will return the contents of that,
3700 * otherwise a compile-time require of perl5db.pl. */
3702 const char * const pdb = PerlEnv_getenv("PERL5DB");
3705 sv_setpv(PL_linestr, pdb);
3706 sv_catpvs(PL_linestr,";");
3708 SETERRNO(0,SS_NORMAL);
3709 sv_setpvs(PL_linestr, "BEGIN { require 'perl5db.pl' };");
3712 sv_setpvs(PL_linestr,"");
3713 if (PL_preambleav) {
3714 SV **svp = AvARRAY(PL_preambleav);
3715 SV **const end = svp + AvFILLp(PL_preambleav);
3717 sv_catsv(PL_linestr, *svp);
3719 sv_catpvs(PL_linestr, ";");
3721 sv_free(MUTABLE_SV(PL_preambleav));
3722 PL_preambleav = NULL;
3725 sv_catpvs(PL_linestr,
3726 "use feature ':5." STRINGIFY(PERL_VERSION) "';");
3727 if (PL_minus_n || PL_minus_p) {
3728 sv_catpvs(PL_linestr, "LINE: while (<>) {");
3730 sv_catpvs(PL_linestr,"chomp;");
3733 if ((*PL_splitstr == '/' || *PL_splitstr == '\''
3734 || *PL_splitstr == '"')
3735 && strchr(PL_splitstr + 1, *PL_splitstr))
3736 Perl_sv_catpvf(aTHX_ PL_linestr, "our @F=split(%s);", PL_splitstr);
3738 /* "q\0${splitstr}\0" is legal perl. Yes, even NUL
3739 bytes can be used as quoting characters. :-) */
3740 const char *splits = PL_splitstr;
3741 sv_catpvs(PL_linestr, "our @F=split(q\0");
3744 if (*splits == '\\')
3745 sv_catpvn(PL_linestr, splits, 1);
3746 sv_catpvn(PL_linestr, splits, 1);
3747 } while (*splits++);
3748 /* This loop will embed the trailing NUL of
3749 PL_linestr as the last thing it does before
3751 sv_catpvs(PL_linestr, ");");
3755 sv_catpvs(PL_linestr,"our @F=split(' ');");
3758 sv_catpvs(PL_linestr, "\n");
3759 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3760 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3761 PL_last_lop = PL_last_uni = NULL;
3762 if ((PERLDB_LINE || PERLDB_SAVESRC) && PL_curstash != PL_debstash)
3763 update_debugger_info(PL_linestr, NULL, 0);
3767 bof = PL_rsfp ? TRUE : FALSE;
3768 if ((s = filter_gets(PL_linestr, PL_rsfp, 0)) == NULL) {
3771 PL_realtokenstart = -1;
3774 if ((PerlIO *)PL_rsfp == PerlIO_stdin())
3775 PerlIO_clearerr(PL_rsfp);
3777 (void)PerlIO_close(PL_rsfp);
3779 PL_doextract = FALSE;
3781 if (!PL_in_eval && (PL_minus_n || PL_minus_p)) {
3787 sv_setpvs(PL_linestr, ";}continue{print;}");
3789 sv_setpvs(PL_linestr, ";}");
3790 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3791 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3792 PL_last_lop = PL_last_uni = NULL;
3793 PL_minus_n = PL_minus_p = 0;
3796 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3797 PL_last_lop = PL_last_uni = NULL;
3798 sv_setpvs(PL_linestr,"");
3799 TOKEN(';'); /* not infinite loop because rsfp is NULL now */
3801 /* If it looks like the start of a BOM or raw UTF-16,
3802 * check if it in fact is. */
3808 #ifdef PERLIO_IS_STDIO
3809 # ifdef __GNU_LIBRARY__
3810 # if __GNU_LIBRARY__ == 1 /* Linux glibc5 */
3811 # define FTELL_FOR_PIPE_IS_BROKEN
3815 # if __GLIBC__ == 1 /* maybe some glibc5 release had it like this? */
3816 # define FTELL_FOR_PIPE_IS_BROKEN
3821 bof = PerlIO_tell(PL_rsfp) == (Off_t)SvCUR(PL_linestr);
3823 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3824 s = swallow_bom((U8*)s);
3828 /* Incest with pod. */
3831 sv_catsv(PL_thiswhite, PL_linestr);
3833 if (*s == '=' && strnEQ(s, "=cut", 4) && !isALPHA(s[4])) {
3834 sv_setpvs(PL_linestr, "");
3835 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3836 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3837 PL_last_lop = PL_last_uni = NULL;
3838 PL_doextract = FALSE;
3842 } while (PL_doextract);
3843 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = s;
3844 if ((PERLDB_LINE || PERLDB_SAVESRC) && PL_curstash != PL_debstash)
3845 update_debugger_info(PL_linestr, NULL, 0);
3846 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3847 PL_last_lop = PL_last_uni = NULL;
3848 if (CopLINE(PL_curcop) == 1) {
3849 while (s < PL_bufend && isSPACE(*s))
3851 if (*s == ':' && s[1] != ':') /* for csh execing sh scripts */
3855 PL_thiswhite = newSVpvn(PL_linestart, s - PL_linestart);
3859 if (*s == '#' && *(s+1) == '!')
3861 #ifdef ALTERNATE_SHEBANG
3863 static char const as[] = ALTERNATE_SHEBANG;
3864 if (*s == as[0] && strnEQ(s, as, sizeof(as) - 1))
3865 d = s + (sizeof(as) - 1);
3867 #endif /* ALTERNATE_SHEBANG */
3876 while (*d && !isSPACE(*d))
3880 #ifdef ARG_ZERO_IS_SCRIPT
3881 if (ipathend > ipath) {
3883 * HP-UX (at least) sets argv[0] to the script name,
3884 * which makes $^X incorrect. And Digital UNIX and Linux,
3885 * at least, set argv[0] to the basename of the Perl
3886 * interpreter. So, having found "#!", we'll set it right.
3888 SV * const x = GvSV(gv_fetchpvs("\030", GV_ADD|GV_NOTQUAL,
3890 assert(SvPOK(x) || SvGMAGICAL(x));
3891 if (sv_eq(x, CopFILESV(PL_curcop))) {
3892 sv_setpvn(x, ipath, ipathend - ipath);
3898 const char *bstart = SvPV_const(CopFILESV(PL_curcop),blen);
3899 const char * const lstart = SvPV_const(x,llen);
3901 bstart += blen - llen;
3902 if (strnEQ(bstart, lstart, llen) && bstart[-1] == '/') {
3903 sv_setpvn(x, ipath, ipathend - ipath);
3908 TAINT_NOT; /* $^X is always tainted, but that's OK */
3910 #endif /* ARG_ZERO_IS_SCRIPT */
3915 d = instr(s,"perl -");
3917 d = instr(s,"perl");
3919 /* avoid getting into infinite loops when shebang
3920 * line contains "Perl" rather than "perl" */
3922 for (d = ipathend-4; d >= ipath; --d) {
3923 if ((*d == 'p' || *d == 'P')
3924 && !ibcmp(d, "perl", 4))
3934 #ifdef ALTERNATE_SHEBANG
3936 * If the ALTERNATE_SHEBANG on this system starts with a
3937 * character that can be part of a Perl expression, then if
3938 * we see it but not "perl", we're probably looking at the
3939 * start of Perl code, not a request to hand off to some
3940 * other interpreter. Similarly, if "perl" is there, but
3941 * not in the first 'word' of the line, we assume the line
3942 * contains the start of the Perl program.
3944 if (d && *s != '#') {
3945 const char *c = ipath;
3946 while (*c && !strchr("; \t\r\n\f\v#", *c))
3949 d = NULL; /* "perl" not in first word; ignore */
3951 *s = '#'; /* Don't try to parse shebang line */
3953 #endif /* ALTERNATE_SHEBANG */
3958 !instr(s,"indir") &&
3959 instr(PL_origargv[0],"perl"))
3966 while (s < PL_bufend && isSPACE(*s))
3968 if (s < PL_bufend) {
3969 Newx(newargv,PL_origargc+3,char*);
3971 while (s < PL_bufend && !isSPACE(*s))
3974 Copy(PL_origargv+1, newargv+2, PL_origargc+1, char*);
3977 newargv = PL_origargv;
3980 PerlProc_execv(ipath, EXEC_ARGV_CAST(newargv));
3982 Perl_croak(aTHX_ "Can't exec %s", ipath);
3985 while (*d && !isSPACE(*d))
3987 while (SPACE_OR_TAB(*d))
3991 const bool switches_done = PL_doswitches;
3992 const U32 oldpdb = PL_perldb;
3993 const bool oldn = PL_minus_n;
3994 const bool oldp = PL_minus_p;
3998 bool baduni = FALSE;
4000 const char *d2 = d1 + 1;
4001 if (parse_unicode_opts((const char **)&d2)
4005 if (baduni || *d1 == 'M' || *d1 == 'm') {
4006 const char * const m = d1;
4007 while (*d1 && !isSPACE(*d1))
4009 Perl_croak(aTHX_ "Too late for \"-%.*s\" option",
4012 d1 = moreswitches(d1);
4014 if (PL_doswitches && !switches_done) {
4015 int argc = PL_origargc;
4016 char **argv = PL_origargv;
4019 } while (argc && argv[0][0] == '-' && argv[0][1]);
4020 init_argv_symbols(argc,argv);
4022 if (((PERLDB_LINE || PERLDB_SAVESRC) && !oldpdb) ||
4023 ((PL_minus_n || PL_minus_p) && !(oldn || oldp)))
4024 /* if we have already added "LINE: while (<>) {",
4025 we must not do it again */
4027 sv_setpvs(PL_linestr, "");
4028 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
4029 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
4030 PL_last_lop = PL_last_uni = NULL;
4031 PL_preambled = FALSE;
4032 if (PERLDB_LINE || PERLDB_SAVESRC)
4033 (void)gv_fetchfile(PL_origfilename);
4040 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
4042 PL_lex_state = LEX_FORMLINE;
4047 #ifdef PERL_STRICT_CR
4048 Perl_warn(aTHX_ "Illegal character \\%03o (carriage return)", '\r');
4050 "\t(Maybe you didn't strip carriage returns after a network transfer?)\n");
4052 case ' ': case '\t': case '\f': case 013:
4054 PL_realtokenstart = -1;
4056 PL_thiswhite = newSVpvs("");
4057 sv_catpvn(PL_thiswhite, s, 1);
4064 PL_realtokenstart = -1;
4068 if (PL_lex_state != LEX_NORMAL || (PL_in_eval && !PL_rsfp)) {
4069 if (*s == '#' && s == PL_linestart && PL_in_eval && !PL_rsfp) {
4070 /* handle eval qq[#line 1 "foo"\n ...] */
4071 CopLINE_dec(PL_curcop);
4074 if (PL_madskills && !PL_lex_formbrack && !PL_in_eval) {
4076 if (!PL_in_eval || PL_rsfp)
4081 while (d < PL_bufend && *d != '\n')
4085 else if (d > PL_bufend) /* Found by Ilya: feed random input to Perl. */
4086 Perl_croak(aTHX_ "panic: input overflow");
4089 PL_thiswhite = newSVpvn(s, d - s);
4094 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
4096 PL_lex_state = LEX_FORMLINE;
4102 if (PL_madskills && CopLINE(PL_curcop) >= 1 && !PL_lex_formbrack) {
4103 if (CopLINE(PL_curcop) == 1 && s[0] == '#' && s[1] == '!') {
4106 TOKEN(PEG); /* make sure any #! line is accessible */
4111 /* if (PL_madskills && PL_lex_formbrack) { */
4113 while (d < PL_bufend && *d != '\n')
4117 else if (d > PL_bufend) /* Found by Ilya: feed random input to Perl. */
4118 Perl_croak(aTHX_ "panic: input overflow");
4119 if (PL_madskills && CopLINE(PL_curcop) >= 1) {
4121 PL_thiswhite = newSVpvs("");
4122 if (CopLINE(PL_curcop) == 1) {
4123 sv_setpvs(PL_thiswhite, "");
4126 sv_catpvn(PL_thiswhite, s, d - s);
4140 if (s[1] && isALPHA(s[1]) && !isALNUM(s[2])) {
4148 while (s < PL_bufend && SPACE_OR_TAB(*s))
4151 if (strnEQ(s,"=>",2)) {
4152 s = force_word(PL_bufptr,WORD,FALSE,FALSE,FALSE);
4153 DEBUG_T( { printbuf("### Saw unary minus before =>, forcing word %s\n", s); } );
4154 OPERATOR('-'); /* unary minus */
4156 PL_last_uni = PL_oldbufptr;
4158 case 'r': ftst = OP_FTEREAD; break;
4159 case 'w': ftst = OP_FTEWRITE; break;
4160 case 'x': ftst = OP_FTEEXEC; break;
4161 case 'o': ftst = OP_FTEOWNED; break;
4162 case 'R': ftst = OP_FTRREAD; break;
4163 case 'W': ftst = OP_FTRWRITE; break;
4164 case 'X': ftst = OP_FTREXEC; break;
4165 case 'O': ftst = OP_FTROWNED; break;
4166 case 'e': ftst = OP_FTIS; break;
4167 case 'z': ftst = OP_FTZERO; break;
4168 case 's': ftst = OP_FTSIZE; break;
4169 case 'f': ftst = OP_FTFILE; break;
4170 case 'd': ftst = OP_FTDIR; break;
4171 case 'l': ftst = OP_FTLINK; break;
4172 case 'p': ftst = OP_FTPIPE; break;
4173 case 'S': ftst = OP_FTSOCK; break;
4174 case 'u': ftst = OP_FTSUID; break;
4175 case 'g': ftst = OP_FTSGID; break;
4176 case 'k': ftst = OP_FTSVTX; break;
4177 case 'b': ftst = OP_FTBLK; break;
4178 case 'c': ftst = OP_FTCHR; break;
4179 case 't': ftst = OP_FTTTY; break;
4180 case 'T': ftst = OP_FTTEXT; break;
4181 case 'B': ftst = OP_FTBINARY; break;
4182 case 'M': case 'A': case 'C':
4183 gv_fetchpvs("\024", GV_ADD|GV_NOTQUAL, SVt_PV);
4185 case 'M': ftst = OP_FTMTIME; break;
4186 case 'A': ftst = OP_FTATIME; break;
4187 case 'C': ftst = OP_FTCTIME; break;
4195 PL_last_lop_op = (OPCODE)ftst;
4196 DEBUG_T( { PerlIO_printf(Perl_debug_log,
4197 "### Saw file test %c\n", (int)tmp);
4202 /* Assume it was a minus followed by a one-letter named
4203 * subroutine call (or a -bareword), then. */
4204 DEBUG_T( { PerlIO_printf(Perl_debug_log,
4205 "### '-%c' looked like a file test but was not\n",
4212 const char tmp = *s++;
4215 if (PL_expect == XOPERATOR)
4220 else if (*s == '>') {
4223 if (isIDFIRST_lazy_if(s,UTF)) {
4224 s = force_word(s,METHOD,FALSE,TRUE,FALSE);
4232 if (PL_expect == XOPERATOR)
4235 if (isSPACE(*s) || !isSPACE(*PL_bufptr))
4237 OPERATOR('-'); /* unary minus */
4243 const char tmp = *s++;
4246 if (PL_expect == XOPERATOR)
4251 if (PL_expect == XOPERATOR)
4254 if (isSPACE(*s) || !isSPACE(*PL_bufptr))
4261 if (PL_expect != XOPERATOR) {
4262 s = scan_ident(s, PL_bufend, PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
4263 PL_expect = XOPERATOR;
4264 force_ident(PL_tokenbuf, '*');
4277 if (PL_expect == XOPERATOR) {
4281 PL_tokenbuf[0] = '%';
4282 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1,
4283 sizeof PL_tokenbuf - 1, FALSE);
4284 if (!PL_tokenbuf[1]) {
4287 PL_pending_ident = '%';
4296 const char tmp = *s++;
4301 && (PL_expect == XOPERATOR || PL_expect == XTERMORDORDOR))
4308 const char tmp = *s++;
4314 goto just_a_word_zero_gv;
4317 switch (PL_expect) {
4323 if (!PL_in_my || PL_lex_state != LEX_NORMAL)
4325 PL_bufptr = s; /* update in case we back off */
4331 PL_expect = XTERMBLOCK;
4334 stuffstart = s - SvPVX(PL_linestr) - 1;
4338 while (isIDFIRST_lazy_if(s,UTF)) {
4341 d = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
4342 if (isLOWER(*s) && (tmp = keyword(PL_tokenbuf, len, 0))) {
4343 if (tmp < 0) tmp = -tmp;
4358 sv = newSVpvn(s, len);
4360 d = scan_str(d,TRUE,TRUE);
4362 /* MUST advance bufptr here to avoid bogus
4363 "at end of line" context messages from yyerror().
4365 PL_bufptr = s + len;
4366 yyerror("Unterminated attribute parameter in attribute list");
4370 return REPORT(0); /* EOF indicator */
4374 sv_catsv(sv, PL_lex_stuff);
4375 attrs = append_elem(OP_LIST, attrs,
4376 newSVOP(OP_CONST, 0, sv));
4377 SvREFCNT_dec(PL_lex_stuff);
4378 PL_lex_stuff = NULL;
4381 if (len == 6 && strnEQ(SvPVX(sv), "unique", len)) {
4383 if (PL_in_my == KEY_our) {
4384 deprecate(":unique");
4387 Perl_croak(aTHX_ "The 'unique' attribute may only be applied to 'our' variables");
4390 /* NOTE: any CV attrs applied here need to be part of
4391 the CVf_BUILTIN_ATTRS define in cv.h! */
4392 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "lvalue", len)) {
4394 CvLVALUE_on(PL_compcv);
4396 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "locked", len)) {
4398 deprecate(":locked");
4400 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "method", len)) {
4402 CvMETHOD_on(PL_compcv);
4404 /* After we've set the flags, it could be argued that
4405 we don't need to do the attributes.pm-based setting
4406 process, and shouldn't bother appending recognized
4407 flags. To experiment with that, uncomment the
4408 following "else". (Note that's already been
4409 uncommented. That keeps the above-applied built-in
4410 attributes from being intercepted (and possibly
4411 rejected) by a package's attribute routines, but is
4412 justified by the performance win for the common case
4413 of applying only built-in attributes.) */
4415 attrs = append_elem(OP_LIST, attrs,
4416 newSVOP(OP_CONST, 0,
4420 if (*s == ':' && s[1] != ':')
4423 break; /* require real whitespace or :'s */
4424 /* XXX losing whitespace on sequential attributes here */
4428 = (PL_expect == XOPERATOR ? '=' : '{'); /*'}(' for vi */
4429 if (*s != ';' && *s != '}' && *s != tmp
4430 && (tmp != '=' || *s != ')')) {
4431 const char q = ((*s == '\'') ? '"' : '\'');
4432 /* If here for an expression, and parsed no attrs, back
4434 if (tmp == '=' && !attrs) {
4438 /* MUST advance bufptr here to avoid bogus "at end of line"
4439 context messages from yyerror().
4442 yyerror( (const char *)
4444 ? Perl_form(aTHX_ "Invalid separator character "
4445 "%c%c%c in attribute list", q, *s, q)
4446 : "Unterminated attribute list" ) );
4454 start_force(PL_curforce);
4455 NEXTVAL_NEXTTOKE.opval = attrs;
4456 CURMAD('_', PL_nextwhite);
4461 PL_thistoken = newSVpvn(SvPVX(PL_linestr) + stuffstart,
4462 (s - SvPVX(PL_linestr)) - stuffstart);
4470 if (PL_last_lop == PL_oldoldbufptr || PL_last_uni == PL_oldoldbufptr)
4471 PL_oldbufptr = PL_oldoldbufptr; /* allow print(STDOUT 123) */
4479 const char tmp = *s++;
4484 const char tmp = *s++;
4492 if (PL_lex_brackets <= 0)
4493 yyerror("Unmatched right square bracket");
4496 if (PL_lex_state == LEX_INTERPNORMAL) {
4497 if (PL_lex_brackets == 0) {
4498 if (*s == '-' && s[1] == '>')
4499 PL_lex_state = LEX_INTERPENDMAYBE;
4500 else if (*s != '[' && *s != '{')
4501 PL_lex_state = LEX_INTERPEND;
4508 if (PL_lex_brackets > 100) {
4509 Renew(PL_lex_brackstack, PL_lex_brackets + 10, char);
4511 switch (PL_expect) {
4513 if (PL_lex_formbrack) {
4517 if (PL_oldoldbufptr == PL_last_lop)
4518 PL_lex_brackstack[PL_lex_brackets++] = XTERM;
4520 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
4521 OPERATOR(HASHBRACK);
4523 while (s < PL_bufend && SPACE_OR_TAB(*s))
4526 PL_tokenbuf[0] = '\0';
4527 if (d < PL_bufend && *d == '-') {
4528 PL_tokenbuf[0] = '-';
4530 while (d < PL_bufend && SPACE_OR_TAB(*d))
4533 if (d < PL_bufend && isIDFIRST_lazy_if(d,UTF)) {
4534 d = scan_word(d, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1,
4536 while (d < PL_bufend && SPACE_OR_TAB(*d))
4539 const char minus = (PL_tokenbuf[0] == '-');
4540 s = force_word(s + minus, WORD, FALSE, TRUE, FALSE);
4548 PL_lex_brackstack[PL_lex_brackets++] = XSTATE;
4553 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
4558 if (PL_oldoldbufptr == PL_last_lop)
4559 PL_lex_brackstack[PL_lex_brackets++] = XTERM;
4561 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
4564 if (PL_expect == XREF && PL_lex_state == LEX_INTERPNORMAL) {
4566 /* This hack is to get the ${} in the message. */
4568 yyerror("syntax error");
4571 OPERATOR(HASHBRACK);
4573 /* This hack serves to disambiguate a pair of curlies
4574 * as being a block or an anon hash. Normally, expectation
4575 * determines that, but in cases where we're not in a
4576 * position to expect anything in particular (like inside
4577 * eval"") we have to resolve the ambiguity. This code
4578 * covers the case where the first term in the curlies is a
4579 * quoted string. Most other cases need to be explicitly
4580 * disambiguated by prepending a "+" before the opening
4581 * curly in order to force resolution as an anon hash.
4583 * XXX should probably propagate the outer expectation
4584 * into eval"" to rely less on this hack, but that could
4585 * potentially break current behavior of eval"".
4589 if (*s == '\'' || *s == '"' || *s == '`') {
4590 /* common case: get past first string, handling escapes */
4591 for (t++; t < PL_bufend && *t != *s;)
4592 if (*t++ == '\\' && (*t == '\\' || *t == *s))
4596 else if (*s == 'q') {
4599 || ((*t == 'q' || *t == 'x') && ++t < PL_bufend
4602 /* skip q//-like construct */
4604 char open, close, term;
4607 while (t < PL_bufend && isSPACE(*t))
4609 /* check for q => */
4610 if (t+1 < PL_bufend && t[0] == '=' && t[1] == '>') {
4611 OPERATOR(HASHBRACK);
4615 if (term && (tmps = strchr("([{< )]}> )]}>",term)))
4619 for (t++; t < PL_bufend; t++) {
4620 if (*t == '\\' && t+1 < PL_bufend && open != '\\')
4622 else if (*t == open)
4626 for (t++; t < PL_bufend; t++) {
4627 if (*t == '\\' && t+1 < PL_bufend)
4629 else if (*t == close && --brackets <= 0)
4631 else if (*t == open)
4638 /* skip plain q word */
4639 while (t < PL_bufend && isALNUM_lazy_if(t,UTF))
4642 else if (isALNUM_lazy_if(t,UTF)) {
4644 while (t < PL_bufend && isALNUM_lazy_if(t,UTF))
4647 while (t < PL_bufend && isSPACE(*t))
4649 /* if comma follows first term, call it an anon hash */
4650 /* XXX it could be a comma expression with loop modifiers */
4651 if (t < PL_bufend && ((*t == ',' && (*s == 'q' || !isLOWER(*s)))
4652 || (*t == '=' && t[1] == '>')))
4653 OPERATOR(HASHBRACK);
4654 if (PL_expect == XREF)
4657 PL_lex_brackstack[PL_lex_brackets-1] = XSTATE;
4663 pl_yylval.ival = CopLINE(PL_curcop);
4664 if (isSPACE(*s) || *s == '#')
4665 PL_copline = NOLINE; /* invalidate current command line number */
4670 if (PL_lex_brackets <= 0)
4671 yyerror("Unmatched right curly bracket");
4673 PL_expect = (expectation)PL_lex_brackstack[--PL_lex_brackets];
4674 if (PL_lex_brackets < PL_lex_formbrack && PL_lex_state != LEX_INTERPNORMAL)
4675 PL_lex_formbrack = 0;
4676 if (PL_lex_state == LEX_INTERPNORMAL) {
4677 if (PL_lex_brackets == 0) {
4678 if (PL_expect & XFAKEBRACK) {
4679 PL_expect &= XENUMMASK;
4680 PL_lex_state = LEX_INTERPEND;
4685 PL_thiswhite = newSVpvs("");
4686 sv_catpvs(PL_thiswhite,"}");
4689 return yylex(); /* ignore fake brackets */
4691 if (*s == '-' && s[1] == '>')
4692 PL_lex_state = LEX_INTERPENDMAYBE;
4693 else if (*s != '[' && *s != '{')
4694 PL_lex_state = LEX_INTERPEND;
4697 if (PL_expect & XFAKEBRACK) {
4698 PL_expect &= XENUMMASK;
4700 return yylex(); /* ignore fake brackets */
4702 start_force(PL_curforce);
4704 curmad('X', newSVpvn(s-1,1));
4705 CURMAD('_', PL_thiswhite);
4710 PL_thistoken = newSVpvs("");
4718 if (PL_expect == XOPERATOR) {
4719 if (PL_bufptr == PL_linestart && ckWARN(WARN_SEMICOLON)
4720 && isIDFIRST_lazy_if(s,UTF))
4722 CopLINE_dec(PL_curcop);
4723 Perl_warner(aTHX_ packWARN(WARN_SEMICOLON), "%s", PL_warn_nosemi);
4724 CopLINE_inc(PL_curcop);
4729 s = scan_ident(s - 1, PL_bufend, PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
4731 PL_expect = XOPERATOR;
4732 force_ident(PL_tokenbuf, '&');
4736 pl_yylval.ival = (OPpENTERSUB_AMPER<<8);
4748 const char tmp = *s++;
4755 if (tmp && isSPACE(*s) && ckWARN(WARN_SYNTAX)
4756 && strchr("+-*/%.^&|<",tmp))
4757 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
4758 "Reversed %c= operator",(int)tmp);
4760 if (PL_expect == XSTATE && isALPHA(tmp) &&
4761 (s == PL_linestart+1 || s[-2] == '\n') )
4763 if (PL_in_eval && !PL_rsfp) {
4768 if (strnEQ(s,"=cut",4)) {
4784 PL_thiswhite = newSVpvs("");
4785 sv_catpvn(PL_thiswhite, PL_linestart,
4786 PL_bufend - PL_linestart);
4790 PL_doextract = TRUE;
4794 if (PL_lex_brackets < PL_lex_formbrack) {
4796 #ifdef PERL_STRICT_CR
4797 while (SPACE_OR_TAB(*t))
4799 while (SPACE_OR_TAB(*t) || *t == '\r')
4802 if (*t == '\n' || *t == '#') {
4813 const char tmp = *s++;
4815 /* was this !=~ where !~ was meant?
4816 * warn on m:!=~\s+([/?]|[msy]\W|tr\W): */
4818 if (*s == '~' && ckWARN(WARN_SYNTAX)) {
4819 const char *t = s+1;
4821 while (t < PL_bufend && isSPACE(*t))
4824 if (*t == '/' || *t == '?' ||
4825 ((*t == 'm' || *t == 's' || *t == 'y')
4826 && !isALNUM(t[1])) ||
4827 (*t == 't' && t[1] == 'r' && !isALNUM(t[2])))
4828 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
4829 "!=~ should be !~");
4839 if (PL_expect != XOPERATOR) {
4840 if (s[1] != '<' && !strchr(s,'>'))
4843 s = scan_heredoc(s);
4845 s = scan_inputsymbol(s);
4846 TERM(sublex_start());
4852 SHop(OP_LEFT_SHIFT);
4866 const char tmp = *s++;
4868 SHop(OP_RIGHT_SHIFT);
4869 else if (tmp == '=')
4878 if (PL_expect == XOPERATOR) {
4879 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
4881 deprecate_old(commaless_variable_list);
4882 return REPORT(','); /* grandfather non-comma-format format */
4886 if (s[1] == '#' && (isIDFIRST_lazy_if(s+2,UTF) || strchr("{$:+-", s[2]))) {
4887 PL_tokenbuf[0] = '@';
4888 s = scan_ident(s + 1, PL_bufend, PL_tokenbuf + 1,
4889 sizeof PL_tokenbuf - 1, FALSE);
4890 if (PL_expect == XOPERATOR)
4891 no_op("Array length", s);
4892 if (!PL_tokenbuf[1])
4894 PL_expect = XOPERATOR;
4895 PL_pending_ident = '#';
4899 PL_tokenbuf[0] = '$';
4900 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1,
4901 sizeof PL_tokenbuf - 1, FALSE);
4902 if (PL_expect == XOPERATOR)
4904 if (!PL_tokenbuf[1]) {
4906 yyerror("Final $ should be \\$ or $name");
4910 /* This kludge not intended to be bulletproof. */
4911 if (PL_tokenbuf[1] == '[' && !PL_tokenbuf[2]) {
4912 pl_yylval.opval = newSVOP(OP_CONST, 0,
4913 newSViv(CopARYBASE_get(&PL_compiling)));
4914 pl_yylval.opval->op_private = OPpCONST_ARYBASE;
4920 const char tmp = *s;
4921 if (PL_lex_state == LEX_NORMAL)
4924 if ((PL_expect != XREF || PL_oldoldbufptr == PL_last_lop)
4925 && intuit_more(s)) {
4927 PL_tokenbuf[0] = '@';
4928 if (ckWARN(WARN_SYNTAX)) {
4931 while (isSPACE(*t) || isALNUM_lazy_if(t,UTF) || *t == '$')
4934 PL_bufptr = PEEKSPACE(PL_bufptr); /* XXX can realloc */
4935 while (t < PL_bufend && *t != ']')
4937 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
4938 "Multidimensional syntax %.*s not supported",
4939 (int)((t - PL_bufptr) + 1), PL_bufptr);
4943 else if (*s == '{') {
4945 PL_tokenbuf[0] = '%';
4946 if (strEQ(PL_tokenbuf+1, "SIG") && ckWARN(WARN_SYNTAX)
4947 && (t = strchr(s, '}')) && (t = strchr(t, '=')))
4949 char tmpbuf[sizeof PL_tokenbuf];
4952 } while (isSPACE(*t));
4953 if (isIDFIRST_lazy_if(t,UTF)) {
4955 t = scan_word(t, tmpbuf, sizeof tmpbuf, TRUE,
4959 if (*t == ';' && get_cvn_flags(tmpbuf, len, 0))
4960 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
4961 "You need to quote \"%s\"",
4968 PL_expect = XOPERATOR;
4969 if (PL_lex_state == LEX_NORMAL && isSPACE((char)tmp)) {
4970 const bool islop = (PL_last_lop == PL_oldoldbufptr);
4971 if (!islop || PL_last_lop_op == OP_GREPSTART)
4972 PL_expect = XOPERATOR;
4973 else if (strchr("$@\"'`q", *s))
4974 PL_expect = XTERM; /* e.g. print $fh "foo" */
4975 else if (strchr("&*<%", *s) && isIDFIRST_lazy_if(s+1,UTF))
4976 PL_expect = XTERM; /* e.g. print $fh &sub */
4977 else if (isIDFIRST_lazy_if(s,UTF)) {
4978 char tmpbuf[sizeof PL_tokenbuf];
4980 scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
4981 if ((t2 = keyword(tmpbuf, len, 0))) {
4982 /* binary operators exclude handle interpretations */
4994 PL_expect = XTERM; /* e.g. print $fh length() */
4999 PL_expect = XTERM; /* e.g. print $fh subr() */
5002 else if (isDIGIT(*s))
5003 PL_expect = XTERM; /* e.g. print $fh 3 */
5004 else if (*s == '.' && isDIGIT(s[1]))
5005 PL_expect = XTERM; /* e.g. print $fh .3 */
5006 else if ((*s == '?' || *s == '-' || *s == '+')
5007 && !isSPACE(s[1]) && s[1] != '=')
5008 PL_expect = XTERM; /* e.g. print $fh -1 */
5009 else if (*s == '/' && !isSPACE(s[1]) && s[1] != '='
5011 PL_expect = XTERM; /* e.g. print $fh /.../
5012 XXX except DORDOR operator
5014 else if (*s == '<' && s[1] == '<' && !isSPACE(s[2])
5016 PL_expect = XTERM; /* print $fh <<"EOF" */
5019 PL_pending_ident = '$';
5023 if (PL_expect == XOPERATOR)
5025 PL_tokenbuf[0] = '@';
5026 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1, FALSE);
5027 if (!PL_tokenbuf[1]) {
5030 if (PL_lex_state == LEX_NORMAL)
5032 if ((PL_expect != XREF || PL_oldoldbufptr == PL_last_lop) && intuit_more(s)) {
5034 PL_tokenbuf[0] = '%';
5036 /* Warn about @ where they meant $. */
5037 if (*s == '[' || *s == '{') {
5038 if (ckWARN(WARN_SYNTAX)) {
5039 const char *t = s + 1;
5040 while (*t && (isALNUM_lazy_if(t,UTF) || strchr(" \t$#+-'\"", *t)))
5042 if (*t == '}' || *t == ']') {
5044 PL_bufptr = PEEKSPACE(PL_bufptr); /* XXX can realloc */
5045 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
5046 "Scalar value %.*s better written as $%.*s",
5047 (int)(t-PL_bufptr), PL_bufptr,
5048 (int)(t-PL_bufptr-1), PL_bufptr+1);
5053 PL_pending_ident = '@';
5056 case '/': /* may be division, defined-or, or pattern */
5057 if (PL_expect == XTERMORDORDOR && s[1] == '/') {
5061 case '?': /* may either be conditional or pattern */
5062 if (PL_expect == XOPERATOR) {
5070 /* A // operator. */
5080 /* Disable warning on "study /blah/" */
5081 if (PL_oldoldbufptr == PL_last_uni
5082 && (*PL_last_uni != 's' || s - PL_last_uni < 5
5083 || memNE(PL_last_uni, "study", 5)
5084 || isALNUM_lazy_if(PL_last_uni+5,UTF)
5087 s = scan_pat(s,OP_MATCH);
5088 TERM(sublex_start());
5092 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack
5093 #ifdef PERL_STRICT_CR
5096 && (s[1] == '\n' || (s[1] == '\r' && s[2] == '\n'))
5098 && (s == PL_linestart || s[-1] == '\n') )
5100 PL_lex_formbrack = 0;
5104 if (PL_expect == XSTATE && s[1] == '.' && s[2] == '.') {
5108 if (PL_expect == XOPERATOR || !isDIGIT(s[1])) {
5114 pl_yylval.ival = OPf_SPECIAL;
5120 if (PL_expect != XOPERATOR)
5125 case '0': case '1': case '2': case '3': case '4':
5126 case '5': case '6': case '7': case '8': case '9':
5127 s = scan_num(s, &pl_yylval);
5128 DEBUG_T( { printbuf("### Saw number in %s\n", s); } );
5129 if (PL_expect == XOPERATOR)
5134 s = scan_str(s,!!PL_madskills,FALSE);
5135 DEBUG_T( { printbuf("### Saw string before %s\n", s); } );
5136 if (PL_expect == XOPERATOR) {
5137 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
5139 deprecate_old(commaless_variable_list);
5140 return REPORT(','); /* grandfather non-comma-format format */
5147 pl_yylval.ival = OP_CONST;
5148 TERM(sublex_start());
5151 s = scan_str(s,!!PL_madskills,FALSE);
5152 DEBUG_T( { printbuf("### Saw string before %s\n", s); } );
5153 if (PL_expect == XOPERATOR) {
5154 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
5156 deprecate_old(commaless_variable_list);
5157 return REPORT(','); /* grandfather non-comma-format format */
5164 pl_yylval.ival = OP_CONST;
5165 /* FIXME. I think that this can be const if char *d is replaced by
5166 more localised variables. */
5167 for (d = SvPV(PL_lex_stuff, len); len; len--, d++) {
5168 if (*d == '$' || *d == '@' || *d == '\\' || !UTF8_IS_INVARIANT((U8)*d)) {
5169 pl_yylval.ival = OP_STRINGIFY;
5173 TERM(sublex_start());
5176 s = scan_str(s,!!PL_madskills,FALSE);
5177 DEBUG_T( { printbuf("### Saw backtick string before %s\n", s); } );
5178 if (PL_expect == XOPERATOR)
5179 no_op("Backticks",s);
5182 readpipe_override();
5183 TERM(sublex_start());
5187 if (PL_lex_inwhat && isDIGIT(*s))
5188 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),"Can't use \\%c to mean $%c in expression",
5190 if (PL_expect == XOPERATOR)
5191 no_op("Backslash",s);
5195 if (isDIGIT(s[1]) && PL_expect != XOPERATOR) {
5196 char *start = s + 2;
5197 while (isDIGIT(*start) || *start == '_')
5199 if (*start == '.' && isDIGIT(start[1])) {
5200 s = scan_num(s, &pl_yylval);
5203 /* avoid v123abc() or $h{v1}, allow C<print v10;> */
5204 else if (!isALPHA(*start) && (PL_expect == XTERM
5205 || PL_expect == XREF || PL_expect == XSTATE
5206 || PL_expect == XTERMORDORDOR)) {
5207 GV *const gv = gv_fetchpvn_flags(s, start - s, 0, SVt_PVCV);
5209 s = scan_num(s, &pl_yylval);
5216 if (isDIGIT(s[1]) && PL_expect == XOPERATOR) {
5258 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
5260 /* Some keywords can be followed by any delimiter, including ':' */
5261 tmp = ((len == 1 && strchr("msyq", PL_tokenbuf[0])) ||
5262 (len == 2 && ((PL_tokenbuf[0] == 't' && PL_tokenbuf[1] == 'r') ||
5263 (PL_tokenbuf[0] == 'q' &&
5264 strchr("qwxr", PL_tokenbuf[1])))));
5266 /* x::* is just a word, unless x is "CORE" */
5267 if (!tmp && *s == ':' && s[1] == ':' && strNE(PL_tokenbuf, "CORE"))
5271 while (d < PL_bufend && isSPACE(*d))
5272 d++; /* no comments skipped here, or s### is misparsed */
5274 /* Is this a label? */
5275 if (!tmp && PL_expect == XSTATE
5276 && d < PL_bufend && *d == ':' && *(d + 1) != ':') {
5277 tmp = keyword(PL_tokenbuf, len, 0);
5279 Perl_croak(aTHX_ "Can't use keyword '%s' as a label", PL_tokenbuf);
5281 pl_yylval.pval = CopLABEL_alloc(PL_tokenbuf);
5286 /* Check for keywords */
5287 tmp = keyword(PL_tokenbuf, len, 0);
5289 /* Is this a word before a => operator? */
5290 if (*d == '=' && d[1] == '>') {
5293 = (OP*)newSVOP(OP_CONST, 0,
5294 S_newSV_maybe_utf8(aTHX_ PL_tokenbuf, len));
5295 pl_yylval.opval->op_private = OPpCONST_BARE;
5299 if (tmp < 0) { /* second-class keyword? */
5300 GV *ogv = NULL; /* override (winner) */
5301 GV *hgv = NULL; /* hidden (loser) */
5302 if (PL_expect != XOPERATOR && (*s != ':' || s[1] != ':')) {
5304 if ((gv = gv_fetchpvn_flags(PL_tokenbuf, len, 0, SVt_PVCV)) &&
5307 if (GvIMPORTED_CV(gv))
5309 else if (! CvMETHOD(cv))
5313 (gvp = (GV**)hv_fetch(PL_globalstash,PL_tokenbuf,len,FALSE)) &&
5314 (gv = *gvp) && isGV_with_GP(gv) &&
5315 GvCVu(gv) && GvIMPORTED_CV(gv))
5322 tmp = 0; /* overridden by import or by GLOBAL */
5325 && -tmp==KEY_lock /* XXX generalizable kludge */
5328 tmp = 0; /* any sub overrides "weak" keyword */
5330 else { /* no override */
5332 if (tmp == KEY_dump) {
5333 Perl_ck_warner(aTHX_ packWARN(WARN_MISC),
5334 "dump() better written as CORE::dump()");
5338 if (hgv && tmp != KEY_x && tmp != KEY_CORE) /* never ambiguous */
5339 Perl_ck_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
5340 "Ambiguous call resolved as CORE::%s(), %s",
5341 GvENAME(hgv), "qualify as such or use &");
5348 default: /* not a keyword */
5349 /* Trade off - by using this evil construction we can pull the
5350 variable gv into the block labelled keylookup. If not, then
5351 we have to give it function scope so that the goto from the
5352 earlier ':' case doesn't bypass the initialisation. */
5354 just_a_word_zero_gv:
5362 const char lastchar = (PL_bufptr == PL_oldoldbufptr ? 0 : PL_bufptr[-1]);
5365 SV *nextPL_nextwhite = 0;
5369 /* Get the rest if it looks like a package qualifier */
5371 if (*s == '\'' || (*s == ':' && s[1] == ':')) {
5373 s = scan_word(s, PL_tokenbuf + len, sizeof PL_tokenbuf - len,
5376 Perl_croak(aTHX_ "Bad name after %s%s", PL_tokenbuf,
5377 *s == '\'' ? "'" : "::");
5382 if (PL_expect == XOPERATOR) {
5383 if (PL_bufptr == PL_linestart) {
5384 CopLINE_dec(PL_curcop);
5385 Perl_warner(aTHX_ packWARN(WARN_SEMICOLON), "%s", PL_warn_nosemi);
5386 CopLINE_inc(PL_curcop);
5389 no_op("Bareword",s);
5392 /* Look for a subroutine with this name in current package,
5393 unless name is "Foo::", in which case Foo is a bearword
5394 (and a package name). */
5396 if (len > 2 && !PL_madskills &&
5397 PL_tokenbuf[len - 2] == ':' && PL_tokenbuf[len - 1] == ':')
5399 if (ckWARN(WARN_BAREWORD)
5400 && ! gv_fetchpvn_flags(PL_tokenbuf, len, 0, SVt_PVHV))
5401 Perl_warner(aTHX_ packWARN(WARN_BAREWORD),
5402 "Bareword \"%s\" refers to nonexistent package",
5405 PL_tokenbuf[len] = '\0';
5411 /* Mustn't actually add anything to a symbol table.
5412 But also don't want to "initialise" any placeholder
5413 constants that might already be there into full
5414 blown PVGVs with attached PVCV. */
5415 gv = gv_fetchpvn_flags(PL_tokenbuf, len,
5416 GV_NOADD_NOINIT, SVt_PVCV);
5421 /* if we saw a global override before, get the right name */
5424 sv = newSVpvs("CORE::GLOBAL::");
5425 sv_catpv(sv,PL_tokenbuf);
5428 /* If len is 0, newSVpv does strlen(), which is correct.
5429 If len is non-zero, then it will be the true length,
5430 and so the scalar will be created correctly. */
5431 sv = newSVpv(PL_tokenbuf,len);
5434 if (PL_madskills && !PL_thistoken) {
5435 char *start = SvPVX(PL_linestr) + PL_realtokenstart;
5436 PL_thistoken = newSVpvn(start,s - start);
5437 PL_realtokenstart = s - SvPVX(PL_linestr);
5441 /* Presume this is going to be a bareword of some sort. */
5444 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
5445 pl_yylval.opval->op_private = OPpCONST_BARE;
5446 /* UTF-8 package name? */
5447 if (UTF && !IN_BYTES &&
5448 is_utf8_string((U8*)SvPVX_const(sv), SvCUR(sv)))
5451 /* And if "Foo::", then that's what it certainly is. */
5456 /* Do the explicit type check so that we don't need to force
5457 the initialisation of the symbol table to have a real GV.
5458 Beware - gv may not really be a PVGV, cv may not really be
5459 a PVCV, (because of the space optimisations that gv_init
5460 understands) But they're true if for this symbol there is
5461 respectively a typeglob and a subroutine.
5463 cv = gv ? ((SvTYPE(gv) == SVt_PVGV)
5464 /* Real typeglob, so get the real subroutine: */
5466 /* A proxy for a subroutine in this package? */
5467 : SvOK(gv) ? MUTABLE_CV(gv) : NULL)
5470 /* See if it's the indirect object for a list operator. */
5472 if (PL_oldoldbufptr &&
5473 PL_oldoldbufptr < PL_bufptr &&
5474 (PL_oldoldbufptr == PL_last_lop
5475 || PL_oldoldbufptr == PL_last_uni) &&
5476 /* NO SKIPSPACE BEFORE HERE! */
5477 (PL_expect == XREF ||
5478 ((PL_opargs[PL_last_lop_op] >> OASHIFT)& 7) == OA_FILEREF))
5480 bool immediate_paren = *s == '(';
5482 /* (Now we can afford to cross potential line boundary.) */
5483 s = SKIPSPACE2(s,nextPL_nextwhite);
5485 PL_nextwhite = nextPL_nextwhite; /* assume no & deception */
5488 /* Two barewords in a row may indicate method call. */
5490 if ((isIDFIRST_lazy_if(s,UTF) || *s == '$') &&
5491 (tmp = intuit_method(s, gv, cv)))
5494 /* If not a declared subroutine, it's an indirect object. */
5495 /* (But it's an indir obj regardless for sort.) */
5496 /* Also, if "_" follows a filetest operator, it's a bareword */
5499 ( !immediate_paren && (PL_last_lop_op == OP_SORT ||
5501 (PL_last_lop_op != OP_MAPSTART &&
5502 PL_last_lop_op != OP_GREPSTART))))
5503 || (PL_tokenbuf[0] == '_' && PL_tokenbuf[1] == '\0'
5504 && ((PL_opargs[PL_last_lop_op] & OA_CLASS_MASK) == OA_FILESTATOP))
5507 PL_expect = (PL_last_lop == PL_oldoldbufptr) ? XTERM : XOPERATOR;
5512 PL_expect = XOPERATOR;
5515 s = SKIPSPACE2(s,nextPL_nextwhite);
5516 PL_nextwhite = nextPL_nextwhite;
5521 /* Is this a word before a => operator? */
5522 if (*s == '=' && s[1] == '>' && !pkgname) {
5524 sv_setpv(((SVOP*)pl_yylval.opval)->op_sv, PL_tokenbuf);
5525 if (UTF && !IN_BYTES && is_utf8_string((U8*)PL_tokenbuf, len))
5526 SvUTF8_on(((SVOP*)pl_yylval.opval)->op_sv);
5530 /* If followed by a paren, it's certainly a subroutine. */
5535 while (SPACE_OR_TAB(*d))
5537 if (*d == ')' && (sv = gv_const_sv(gv))) {
5544 PL_nextwhite = PL_thiswhite;
5547 start_force(PL_curforce);
5549 NEXTVAL_NEXTTOKE.opval = pl_yylval.opval;
5550 PL_expect = XOPERATOR;
5553 PL_nextwhite = nextPL_nextwhite;
5554 curmad('X', PL_thistoken);
5555 PL_thistoken = newSVpvs("");
5563 /* If followed by var or block, call it a method (unless sub) */
5565 if ((*s == '$' || *s == '{') && (!gv || !cv)) {
5566 PL_last_lop = PL_oldbufptr;
5567 PL_last_lop_op = OP_METHOD;
5571 /* If followed by a bareword, see if it looks like indir obj. */
5574 && (isIDFIRST_lazy_if(s,UTF) || *s == '$')
5575 && (tmp = intuit_method(s, gv, cv)))
5578 /* Not a method, so call it a subroutine (if defined) */
5581 if (lastchar == '-')
5582 Perl_ck_warner_d(aTHX_ packWARN(WARN_AMBIGUOUS),
5583 "Ambiguous use of -%s resolved as -&%s()",
5584 PL_tokenbuf, PL_tokenbuf);
5585 /* Check for a constant sub */
5586 if ((sv = gv_const_sv(gv))) {
5588 SvREFCNT_dec(((SVOP*)pl_yylval.opval)->op_sv);
5589 ((SVOP*)pl_yylval.opval)->op_sv = SvREFCNT_inc_simple(sv);
5590 pl_yylval.opval->op_private = 0;
5594 /* Resolve to GV now. */
5595 if (SvTYPE(gv) != SVt_PVGV) {
5596 gv = gv_fetchpv(PL_tokenbuf, 0, SVt_PVCV);
5597 assert (SvTYPE(gv) == SVt_PVGV);
5598 /* cv must have been some sort of placeholder, so
5599 now needs replacing with a real code reference. */
5603 op_free(pl_yylval.opval);
5604 pl_yylval.opval = newCVREF(0, newGVOP(OP_GV, 0, gv));
5605 pl_yylval.opval->op_private |= OPpENTERSUB_NOPAREN;
5606 PL_last_lop = PL_oldbufptr;
5607 PL_last_lop_op = OP_ENTERSUB;
5608 /* Is there a prototype? */
5616 const char *proto = SvPV_const(MUTABLE_SV(cv), protolen);
5619 if ((*proto == '$' || *proto == '_') && proto[1] == '\0')
5621 while (*proto == ';')
5623 if (*proto == '&' && *s == '{') {
5625 sv_setpvs(PL_subname, "__ANON__");
5627 sv_setpvs(PL_subname, "__ANON__::__ANON__");
5634 PL_nextwhite = PL_thiswhite;
5637 start_force(PL_curforce);
5638 NEXTVAL_NEXTTOKE.opval = pl_yylval.opval;
5641 PL_nextwhite = nextPL_nextwhite;
5642 curmad('X', PL_thistoken);
5643 PL_thistoken = newSVpvs("");
5650 /* Guess harder when madskills require "best effort". */
5651 if (PL_madskills && (!gv || !GvCVu(gv))) {
5652 int probable_sub = 0;
5653 if (strchr("\"'`$@%0123456789!*+{[<", *s))
5655 else if (isALPHA(*s)) {
5659 d = scan_word(d, tmpbuf, sizeof tmpbuf, TRUE, &tmplen);
5660 if (!keyword(tmpbuf, tmplen, 0))
5663 while (d < PL_bufend && isSPACE(*d))
5665 if (*d == '=' && d[1] == '>')
5670 gv = gv_fetchpv(PL_tokenbuf, GV_ADD, SVt_PVCV);
5671 op_free(pl_yylval.opval);
5672 pl_yylval.opval = newCVREF(0, newGVOP(OP_GV, 0, gv));
5673 pl_yylval.opval->op_private |= OPpENTERSUB_NOPAREN;
5674 PL_last_lop = PL_oldbufptr;
5675 PL_last_lop_op = OP_ENTERSUB;
5676 PL_nextwhite = PL_thiswhite;
5678 start_force(PL_curforce);
5679 NEXTVAL_NEXTTOKE.opval = pl_yylval.opval;
5681 PL_nextwhite = nextPL_nextwhite;
5682 curmad('X', PL_thistoken);
5683 PL_thistoken = newSVpvs("");
5688 NEXTVAL_NEXTTOKE.opval = pl_yylval.opval;
5695 /* Call it a bare word */
5697 if (PL_hints & HINT_STRICT_SUBS)
5698 pl_yylval.opval->op_private |= OPpCONST_STRICT;
5701 /* after "print" and similar functions (corresponding to
5702 * "F? L" in opcode.pl), whatever wasn't already parsed as
5703 * a filehandle should be subject to "strict subs".
5704 * Likewise for the optional indirect-object argument to system
5705 * or exec, which can't be a bareword */
5706 if ((PL_last_lop_op == OP_PRINT
5707 || PL_last_lop_op == OP_PRTF
5708 || PL_last_lop_op == OP_SAY
5709 || PL_last_lop_op == OP_SYSTEM
5710 || PL_last_lop_op == OP_EXEC)
5711 && (PL_hints & HINT_STRICT_SUBS))
5712 pl_yylval.opval->op_private |= OPpCONST_STRICT;
5713 if (lastchar != '-') {
5714 if (ckWARN(WARN_RESERVED)) {
5718 if (!*d && !gv_stashpv(PL_tokenbuf, 0))
5719 Perl_warner(aTHX_ packWARN(WARN_RESERVED), PL_warn_reserved,
5726 if ((lastchar == '*' || lastchar == '%' || lastchar == '&')) {
5727 Perl_ck_warner_d(aTHX_ packWARN(WARN_AMBIGUOUS),
5728 "Operator or semicolon missing before %c%s",
5729 lastchar, PL_tokenbuf);
5730 Perl_ck_warner_d(aTHX_ packWARN(WARN_AMBIGUOUS),
5731 "Ambiguous use of %c resolved as operator %c",
5732 lastchar, lastchar);
5738 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0,
5739 newSVpv(CopFILE(PL_curcop),0));
5743 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0,
5744 Perl_newSVpvf(aTHX_ "%"IVdf, (IV)CopLINE(PL_curcop)));
5747 case KEY___PACKAGE__:
5748 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0,
5750 ? newSVhek(HvNAME_HEK(PL_curstash))
5757 if (PL_rsfp && (!PL_in_eval || PL_tokenbuf[2] == 'D')) {
5758 const char *pname = "main";
5759 if (PL_tokenbuf[2] == 'D')
5760 pname = HvNAME_get(PL_curstash ? PL_curstash : PL_defstash);
5761 gv = gv_fetchpv(Perl_form(aTHX_ "%s::DATA", pname), GV_ADD,
5765 GvIOp(gv) = newIO();
5766 IoIFP(GvIOp(gv)) = PL_rsfp;
5767 #if defined(HAS_FCNTL) && defined(F_SETFD)
5769 const int fd = PerlIO_fileno(PL_rsfp);
5770 fcntl(fd,F_SETFD,fd >= 3);
5773 /* Mark this internal pseudo-handle as clean */
5774 IoFLAGS(GvIOp(gv)) |= IOf_UNTAINT;
5775 if ((PerlIO*)PL_rsfp == PerlIO_stdin())
5776 IoTYPE(GvIOp(gv)) = IoTYPE_STD;
5778 IoTYPE(GvIOp(gv)) = IoTYPE_RDONLY;
5779 #if defined(WIN32) && !defined(PERL_TEXTMODE_SCRIPTS)
5780 /* if the script was opened in binmode, we need to revert
5781 * it to text mode for compatibility; but only iff it has CRs
5782 * XXX this is a questionable hack at best. */
5783 if (PL_bufend-PL_bufptr > 2
5784 && PL_bufend[-1] == '\n' && PL_bufend[-2] == '\r')
5787 if (IoTYPE(GvIOp(gv)) == IoTYPE_RDONLY) {
5788 loc = PerlIO_tell(PL_rsfp);
5789 (void)PerlIO_seek(PL_rsfp, 0L, 0);
5792 if (PerlLIO_setmode(PL_rsfp, O_TEXT) != -1) {
5794 if (PerlLIO_setmode(PerlIO_fileno(PL_rsfp), O_TEXT) != -1) {
5795 #endif /* NETWARE */
5796 #ifdef PERLIO_IS_STDIO /* really? */
5797 # if defined(__BORLANDC__)
5798 /* XXX see note in do_binmode() */
5799 ((FILE*)PL_rsfp)->flags &= ~_F_BIN;
5803 PerlIO_seek(PL_rsfp, loc, 0);
5807 #ifdef PERLIO_LAYERS
5810 PerlIO_apply_layers(aTHX_ PL_rsfp, NULL, ":utf8");
5811 else if (PL_encoding) {
5818 XPUSHs(PL_encoding);
5820 call_method("name", G_SCALAR);
5824 PerlIO_apply_layers(aTHX_ PL_rsfp, NULL,
5825 Perl_form(aTHX_ ":encoding(%"SVf")",
5834 if (PL_realtokenstart >= 0) {
5835 char *tstart = SvPVX(PL_linestr) + PL_realtokenstart;
5837 PL_endwhite = newSVpvs("");
5838 sv_catsv(PL_endwhite, PL_thiswhite);
5840 sv_catpvn(PL_endwhite, tstart, PL_bufend - tstart);
5841 PL_realtokenstart = -1;
5843 while ((s = filter_gets(PL_endwhite, PL_rsfp,
5844 SvCUR(PL_endwhite))) != NULL) ;
5859 if (PL_expect == XSTATE) {
5866 if (*s == ':' && s[1] == ':') {
5869 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
5870 if (!(tmp = keyword(PL_tokenbuf, len, 0)))
5871 Perl_croak(aTHX_ "CORE::%s is not a keyword", PL_tokenbuf);
5874 else if (tmp == KEY_require || tmp == KEY_do)
5875 /* that's a way to remember we saw "CORE::" */
5888 LOP(OP_ACCEPT,XTERM);
5894 LOP(OP_ATAN2,XTERM);
5900 LOP(OP_BINMODE,XTERM);
5903 LOP(OP_BLESS,XTERM);
5912 /* When 'use switch' is in effect, continue has a dual
5913 life as a control operator. */
5915 if (!FEATURE_IS_ENABLED("switch"))
5918 /* We have to disambiguate the two senses of
5919 "continue". If the next token is a '{' then
5920 treat it as the start of a continue block;
5921 otherwise treat it as a control operator.
5933 (void)gv_fetchpvs("ENV", GV_ADD|GV_NOTQUAL, SVt_PVHV);
5950 if (!PL_cryptseen) {
5951 PL_cryptseen = TRUE;
5955 LOP(OP_CRYPT,XTERM);
5958 LOP(OP_CHMOD,XTERM);
5961 LOP(OP_CHOWN,XTERM);
5964 LOP(OP_CONNECT,XTERM);
5983 s = force_word(s,WORD,TRUE,TRUE,FALSE);
5984 if (orig_keyword == KEY_do) {
5993 PL_hints |= HINT_BLOCK_SCOPE;
6003 gv_fetchpvs("AnyDBM_File::ISA", GV_ADDMULTI, SVt_PVAV);
6004 LOP(OP_DBMOPEN,XTERM);
6010 s = force_word(s,WORD,TRUE,FALSE,FALSE);
6017 pl_yylval.ival = CopLINE(PL_curcop);
6033 PL_expect = (*s == '{') ? XTERMBLOCK : XTERM;
6034 UNIBRACK(OP_ENTEREVAL);
6048 case KEY_endhostent:
6054 case KEY_endservent:
6057 case KEY_endprotoent:
6068 pl_yylval.ival = CopLINE(PL_curcop);
6070 if (PL_expect == XSTATE && isIDFIRST_lazy_if(s,UTF)) {
6073 int soff = s - SvPVX(PL_linestr); /* for skipspace realloc */
6076 if ((PL_bufend - p) >= 3 &&
6077 strnEQ(p, "my", 2) && isSPACE(*(p + 2)))
6079 else if ((PL_bufend - p) >= 4 &&
6080 strnEQ(p, "our", 3) && isSPACE(*(p + 3)))
6083 if (isIDFIRST_lazy_if(p,UTF)) {
6084 p = scan_ident(p, PL_bufend,
6085 PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
6089 Perl_croak(aTHX_ "Missing $ on loop variable");
6091 s = SvPVX(PL_linestr) + soff;
6097 LOP(OP_FORMLINE,XTERM);
6103 LOP(OP_FCNTL,XTERM);
6109 LOP(OP_FLOCK,XTERM);
6118 LOP(OP_GREPSTART, XREF);
6121 s = force_word(s,WORD,TRUE,FALSE,FALSE);
6136 case KEY_getpriority:
6137 LOP(OP_GETPRIORITY,XTERM);
6139 case KEY_getprotobyname:
6142 case KEY_getprotobynumber:
6143 LOP(OP_GPBYNUMBER,XTERM);
6145 case KEY_getprotoent:
6157 case KEY_getpeername:
6158 UNI(OP_GETPEERNAME);
6160 case KEY_gethostbyname:
6163 case KEY_gethostbyaddr:
6164 LOP(OP_GHBYADDR,XTERM);
6166 case KEY_gethostent:
6169 case KEY_getnetbyname:
6172 case KEY_getnetbyaddr:
6173 LOP(OP_GNBYADDR,XTERM);
6178 case KEY_getservbyname:
6179 LOP(OP_GSBYNAME,XTERM);
6181 case KEY_getservbyport:
6182 LOP(OP_GSBYPORT,XTERM);
6184 case KEY_getservent:
6187 case KEY_getsockname:
6188 UNI(OP_GETSOCKNAME);
6190 case KEY_getsockopt:
6191 LOP(OP_GSOCKOPT,XTERM);
6206 pl_yylval.ival = CopLINE(PL_curcop);
6216 pl_yylval.ival = CopLINE(PL_curcop);
6220 LOP(OP_INDEX,XTERM);
6226 LOP(OP_IOCTL,XTERM);
6238 s = force_word(s,WORD,TRUE,FALSE,FALSE);
6270 LOP(OP_LISTEN,XTERM);
6279 s = scan_pat(s,OP_MATCH);
6280 TERM(sublex_start());
6283 LOP(OP_MAPSTART, XREF);
6286 LOP(OP_MKDIR,XTERM);
6289 LOP(OP_MSGCTL,XTERM);
6292 LOP(OP_MSGGET,XTERM);
6295 LOP(OP_MSGRCV,XTERM);
6298 LOP(OP_MSGSND,XTERM);
6303 PL_in_my = (U16)tmp;
6305 if (isIDFIRST_lazy_if(s,UTF)) {
6309 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, TRUE, &len);
6310 if (len == 3 && strnEQ(PL_tokenbuf, "sub", 3))
6312 PL_in_my_stash = find_in_my_stash(PL_tokenbuf, len);
6313 if (!PL_in_my_stash) {
6316 my_snprintf(tmpbuf, sizeof(tmpbuf), "No such class %.1000s", PL_tokenbuf);
6320 if (PL_madskills) { /* just add type to declarator token */
6321 sv_catsv(PL_thistoken, PL_nextwhite);
6323 sv_catpvn(PL_thistoken, start, s - start);
6331 s = force_word(s,WORD,TRUE,FALSE,FALSE);
6338 s = tokenize_use(0, s);
6342 if (*s == '(' || (s = SKIPSPACE1(s), *s == '('))
6349 if (isIDFIRST_lazy_if(s,UTF)) {
6351 for (d = s; isALNUM_lazy_if(d,UTF);)
6353 for (t=d; isSPACE(*t);)
6355 if ( *t && strchr("|&*+-=!?:.", *t) && ckWARN_d(WARN_PRECEDENCE)
6357 && !(t[0] == '=' && t[1] == '>')
6359 int parms_len = (int)(d-s);
6360 Perl_warner(aTHX_ packWARN(WARN_PRECEDENCE),
6361 "Precedence problem: open %.*s should be open(%.*s)",
6362 parms_len, s, parms_len, s);
6368 pl_yylval.ival = OP_OR;
6378 LOP(OP_OPEN_DIR,XTERM);
6381 checkcomma(s,PL_tokenbuf,"filehandle");
6385 checkcomma(s,PL_tokenbuf,"filehandle");
6404 s = force_word(s,WORD,FALSE,TRUE,FALSE);
6405 s = force_version(s, FALSE);
6409 LOP(OP_PIPE_OP,XTERM);
6412 s = scan_str(s,!!PL_madskills,FALSE);
6415 pl_yylval.ival = OP_CONST;
6416 TERM(sublex_start());
6422 s = scan_str(s,!!PL_madskills,FALSE);
6425 PL_expect = XOPERATOR;
6427 if (SvCUR(PL_lex_stuff)) {
6430 d = SvPV_force(PL_lex_stuff, len);
6432 for (; isSPACE(*d) && len; --len, ++d)
6437 if (!warned && ckWARN(WARN_QW)) {
6438 for (; !isSPACE(*d) && len; --len, ++d) {
6440 Perl_warner(aTHX_ packWARN(WARN_QW),
6441 "Possible attempt to separate words with commas");
6444 else if (*d == '#') {
6445 Perl_warner(aTHX_ packWARN(WARN_QW),
6446 "Possible attempt to put comments in qw() list");
6452 for (; !isSPACE(*d) && len; --len, ++d)
6455 sv = newSVpvn_utf8(b, d-b, DO_UTF8(PL_lex_stuff));
6456 words = append_elem(OP_LIST, words,
6457 newSVOP(OP_CONST, 0, tokeq(sv)));
6461 start_force(PL_curforce);
6462 NEXTVAL_NEXTTOKE.opval = words;
6467 SvREFCNT_dec(PL_lex_stuff);
6468 PL_lex_stuff = NULL;
6474 s = scan_str(s,!!PL_madskills,FALSE);
6477 pl_yylval.ival = OP_STRINGIFY;
6478 if (SvIVX(PL_lex_stuff) == '\'')
6479 SvIV_set(PL_lex_stuff, 0); /* qq'$foo' should intepolate */
6480 TERM(sublex_start());
6483 s = scan_pat(s,OP_QR);
6484 TERM(sublex_start());
6487 s = scan_str(s,!!PL_madskills,FALSE);
6490 readpipe_override();
6491 TERM(sublex_start());
6499 s = force_version(s, FALSE);
6501 else if (*s != 'v' || !isDIGIT(s[1])
6502 || (s = force_version(s, TRUE), *s == 'v'))
6504 *PL_tokenbuf = '\0';
6505 s = force_word(s,WORD,TRUE,TRUE,FALSE);
6506 if (isIDFIRST_lazy_if(PL_tokenbuf,UTF))
6507 gv_stashpvn(PL_tokenbuf, strlen(PL_tokenbuf), GV_ADD);
6509 yyerror("<> should be quotes");
6511 if (orig_keyword == KEY_require) {
6519 PL_last_uni = PL_oldbufptr;
6520 PL_last_lop_op = OP_REQUIRE;
6522 return REPORT( (int)REQUIRE );
6528 s = force_word(s,WORD,TRUE,FALSE,FALSE);
6532 LOP(OP_RENAME,XTERM);
6541 LOP(OP_RINDEX,XTERM);
6550 UNIDOR(OP_READLINE);
6553 UNIDOR(OP_BACKTICK);
6562 LOP(OP_REVERSE,XTERM);
6565 UNIDOR(OP_READLINK);
6572 if (pl_yylval.opval)
6573 TERM(sublex_start());
6575 TOKEN(1); /* force error */
6578 checkcomma(s,PL_tokenbuf,"filehandle");
6588 LOP(OP_SELECT,XTERM);
6594 LOP(OP_SEMCTL,XTERM);
6597 LOP(OP_SEMGET,XTERM);
6600 LOP(OP_SEMOP,XTERM);
6606 LOP(OP_SETPGRP,XTERM);
6608 case KEY_setpriority:
6609 LOP(OP_SETPRIORITY,XTERM);
6611 case KEY_sethostent:
6617 case KEY_setservent:
6620 case KEY_setprotoent:
6630 LOP(OP_SEEKDIR,XTERM);
6632 case KEY_setsockopt:
6633 LOP(OP_SSOCKOPT,XTERM);
6639 LOP(OP_SHMCTL,XTERM);
6642 LOP(OP_SHMGET,XTERM);
6645 LOP(OP_SHMREAD,XTERM);
6648 LOP(OP_SHMWRITE,XTERM);
6651 LOP(OP_SHUTDOWN,XTERM);
6660 LOP(OP_SOCKET,XTERM);
6662 case KEY_socketpair:
6663 LOP(OP_SOCKPAIR,XTERM);
6666 checkcomma(s,PL_tokenbuf,"subroutine name");
6668 if (*s == ';' || *s == ')') /* probably a close */
6669 Perl_croak(aTHX_ "sort is now a reserved word");
6671 s = force_word(s,WORD,TRUE,TRUE,FALSE);
6675 LOP(OP_SPLIT,XTERM);
6678 LOP(OP_SPRINTF,XTERM);
6681 LOP(OP_SPLICE,XTERM);
6696 LOP(OP_SUBSTR,XTERM);
6702 char tmpbuf[sizeof PL_tokenbuf];
6703 SSize_t tboffset = 0;
6704 expectation attrful;
6705 bool have_name, have_proto;
6706 const int key = tmp;
6711 char *tstart = SvPVX(PL_linestr) + PL_realtokenstart;
6712 SV *subtoken = newSVpvn(tstart, s - tstart);
6716 s = SKIPSPACE2(s,tmpwhite);
6721 if (isIDFIRST_lazy_if(s,UTF) || *s == '\'' ||
6722 (*s == ':' && s[1] == ':'))
6725 SV *nametoke = NULL;
6729 attrful = XATTRBLOCK;
6730 /* remember buffer pos'n for later force_word */
6731 tboffset = s - PL_oldbufptr;
6732 d = scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
6735 nametoke = newSVpvn(s, d - s);
6737 if (memchr(tmpbuf, ':', len))
6738 sv_setpvn(PL_subname, tmpbuf, len);
6740 sv_setsv(PL_subname,PL_curstname);
6741 sv_catpvs(PL_subname,"::");
6742 sv_catpvn(PL_subname,tmpbuf,len);
6749 CURMAD('X', nametoke);
6750 CURMAD('_', tmpwhite);
6751 (void) force_word(PL_oldbufptr + tboffset, WORD,
6754 s = SKIPSPACE2(d,tmpwhite);
6761 Perl_croak(aTHX_ "Missing name in \"my sub\"");
6762 PL_expect = XTERMBLOCK;
6763 attrful = XATTRTERM;
6764 sv_setpvs(PL_subname,"?");
6768 if (key == KEY_format) {
6770 PL_lex_formbrack = PL_lex_brackets + 1;
6772 PL_thistoken = subtoken;
6776 (void) force_word(PL_oldbufptr + tboffset, WORD,
6782 /* Look for a prototype */
6785 bool bad_proto = FALSE;
6786 bool in_brackets = FALSE;
6787 char greedy_proto = ' ';
6788 bool proto_after_greedy_proto = FALSE;
6789 bool must_be_last = FALSE;
6790 bool underscore = FALSE;
6791 bool seen_underscore = FALSE;
6792 const bool warnsyntax = ckWARN(WARN_SYNTAX);
6794 s = scan_str(s,!!PL_madskills,FALSE);
6796 Perl_croak(aTHX_ "Prototype not terminated");
6797 /* strip spaces and check for bad characters */
6798 d = SvPVX(PL_lex_stuff);
6800 for (p = d; *p; ++p) {
6806 proto_after_greedy_proto = TRUE;
6807 if (!strchr("$@%*;[]&\\_", *p)) {
6819 else if ( *p == ']' ) {
6820 in_brackets = FALSE;
6822 else if ( (*p == '@' || *p == '%') &&
6823 ( tmp < 2 || d[tmp-2] != '\\' ) &&
6825 must_be_last = TRUE;
6828 else if ( *p == '_' ) {
6829 underscore = seen_underscore = TRUE;
6836 if (proto_after_greedy_proto)
6837 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
6838 "Prototype after '%c' for %"SVf" : %s",
6839 greedy_proto, SVfARG(PL_subname), d);
6841 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
6842 "Illegal character %sin prototype for %"SVf" : %s",
6843 seen_underscore ? "after '_' " : "",
6844 SVfARG(PL_subname), d);
6845 SvCUR_set(PL_lex_stuff, tmp);
6850 CURMAD('q', PL_thisopen);
6851 CURMAD('_', tmpwhite);
6852 CURMAD('=', PL_thisstuff);
6853 CURMAD('Q', PL_thisclose);
6854 NEXTVAL_NEXTTOKE.opval =
6855 (OP*)newSVOP(OP_CONST, 0, PL_lex_stuff);
6856 PL_lex_stuff = NULL;
6859 s = SKIPSPACE2(s,tmpwhite);
6867 if (*s == ':' && s[1] != ':')
6868 PL_expect = attrful;
6869 else if (*s != '{' && key == KEY_sub) {
6871 Perl_croak(aTHX_ "Illegal declaration of anonymous subroutine");
6873 Perl_croak(aTHX_ "Illegal declaration of subroutine %"SVf, SVfARG(PL_subname));
6880 curmad('^', newSVpvs(""));
6881 CURMAD('_', tmpwhite);
6885 PL_thistoken = subtoken;
6888 NEXTVAL_NEXTTOKE.opval =
6889 (OP*)newSVOP(OP_CONST, 0, PL_lex_stuff);
6890 PL_lex_stuff = NULL;
6896 sv_setpvs(PL_subname, "__ANON__");
6898 sv_setpvs(PL_subname, "__ANON__::__ANON__");
6902 (void) force_word(PL_oldbufptr + tboffset, WORD,
6911 LOP(OP_SYSTEM,XREF);
6914 LOP(OP_SYMLINK,XTERM);
6917 LOP(OP_SYSCALL,XTERM);
6920 LOP(OP_SYSOPEN,XTERM);
6923 LOP(OP_SYSSEEK,XTERM);
6926 LOP(OP_SYSREAD,XTERM);
6929 LOP(OP_SYSWRITE,XTERM);
6933 TERM(sublex_start());
6954 LOP(OP_TRUNCATE,XTERM);
6966 pl_yylval.ival = CopLINE(PL_curcop);
6970 pl_yylval.ival = CopLINE(PL_curcop);
6974 LOP(OP_UNLINK,XTERM);
6980 LOP(OP_UNPACK,XTERM);
6983 LOP(OP_UTIME,XTERM);
6989 LOP(OP_UNSHIFT,XTERM);
6992 s = tokenize_use(1, s);
7002 pl_yylval.ival = CopLINE(PL_curcop);
7006 pl_yylval.ival = CopLINE(PL_curcop);
7010 PL_hints |= HINT_BLOCK_SCOPE;
7017 LOP(OP_WAITPID,XTERM);
7026 ctl_l[0] = toCTRL('L');
7028 gv_fetchpvn_flags(ctl_l, 1, GV_ADD|GV_NOTQUAL, SVt_PV);
7031 /* Make sure $^L is defined */
7032 gv_fetchpvs("\f", GV_ADD|GV_NOTQUAL, SVt_PV);
7037 if (PL_expect == XOPERATOR)
7043 pl_yylval.ival = OP_XOR;
7048 TERM(sublex_start());
7053 #pragma segment Main
7057 S_pending_ident(pTHX)
7062 /* pit holds the identifier we read and pending_ident is reset */
7063 char pit = PL_pending_ident;
7064 const STRLEN tokenbuf_len = strlen(PL_tokenbuf);
7065 /* All routes through this function want to know if there is a colon. */
7066 const char *const has_colon = (const char*) memchr (PL_tokenbuf, ':', tokenbuf_len);
7067 PL_pending_ident = 0;
7069 /* PL_realtokenstart = realtokenend = PL_bufptr - SvPVX(PL_linestr); */
7070 DEBUG_T({ PerlIO_printf(Perl_debug_log,
7071 "### Pending identifier '%s'\n", PL_tokenbuf); });
7073 /* if we're in a my(), we can't allow dynamics here.
7074 $foo'bar has already been turned into $foo::bar, so
7075 just check for colons.
7077 if it's a legal name, the OP is a PADANY.
7080 if (PL_in_my == KEY_our) { /* "our" is merely analogous to "my" */
7082 yyerror(Perl_form(aTHX_ "No package name allowed for "
7083 "variable %s in \"our\"",
7085 tmp = allocmy(PL_tokenbuf);
7089 yyerror(Perl_form(aTHX_ PL_no_myglob,
7090 PL_in_my == KEY_my ? "my" : "state", PL_tokenbuf));
7092 pl_yylval.opval = newOP(OP_PADANY, 0);
7093 pl_yylval.opval->op_targ = allocmy(PL_tokenbuf);
7099 build the ops for accesses to a my() variable.
7101 Deny my($a) or my($b) in a sort block, *if* $a or $b is
7102 then used in a comparison. This catches most, but not
7103 all cases. For instance, it catches
7104 sort { my($a); $a <=> $b }
7106 sort { my($a); $a < $b ? -1 : $a == $b ? 0 : 1; }
7107 (although why you'd do that is anyone's guess).
7112 tmp = pad_findmy(PL_tokenbuf);
7113 if (tmp != NOT_IN_PAD) {
7114 /* might be an "our" variable" */
7115 if (PAD_COMPNAME_FLAGS_isOUR(tmp)) {
7116 /* build ops for a bareword */
7117 HV * const stash = PAD_COMPNAME_OURSTASH(tmp);
7118 HEK * const stashname = HvNAME_HEK(stash);
7119 SV * const sym = newSVhek(stashname);
7120 sv_catpvs(sym, "::");
7121 sv_catpvn(sym, PL_tokenbuf+1, tokenbuf_len - 1);
7122 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sym);
7123 pl_yylval.opval->op_private = OPpCONST_ENTERED;
7126 ? (GV_ADDMULTI | GV_ADDINEVAL)
7129 ((PL_tokenbuf[0] == '$') ? SVt_PV
7130 : (PL_tokenbuf[0] == '@') ? SVt_PVAV
7135 /* if it's a sort block and they're naming $a or $b */
7136 if (PL_last_lop_op == OP_SORT &&
7137 PL_tokenbuf[0] == '$' &&
7138 (PL_tokenbuf[1] == 'a' || PL_tokenbuf[1] == 'b')
7141 for (d = PL_in_eval ? PL_oldoldbufptr : PL_linestart;
7142 d < PL_bufend && *d != '\n';
7145 if (strnEQ(d,"<=>",3) || strnEQ(d,"cmp",3)) {
7146 Perl_croak(aTHX_ "Can't use \"my %s\" in sort comparison",
7152 pl_yylval.opval = newOP(OP_PADANY, 0);
7153 pl_yylval.opval->op_targ = tmp;
7159 Whine if they've said @foo in a doublequoted string,
7160 and @foo isn't a variable we can find in the symbol
7163 if (pit == '@' && PL_lex_state != LEX_NORMAL && !PL_lex_brackets) {
7164 GV *const gv = gv_fetchpvn_flags(PL_tokenbuf + 1, tokenbuf_len - 1, 0,
7166 if ((!gv || ((PL_tokenbuf[0] == '@') ? !GvAV(gv) : !GvHV(gv)))
7167 /* DO NOT warn for @- and @+ */
7168 && !( PL_tokenbuf[2] == '\0' &&
7169 ( PL_tokenbuf[1] == '-' || PL_tokenbuf[1] == '+' ))
7172 /* Downgraded from fatal to warning 20000522 mjd */
7173 Perl_ck_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
7174 "Possible unintended interpolation of %s in string",
7179 /* build ops for a bareword */
7180 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, newSVpvn(PL_tokenbuf + 1,
7182 pl_yylval.opval->op_private = OPpCONST_ENTERED;
7184 PL_tokenbuf + 1, tokenbuf_len - 1,
7185 /* If the identifier refers to a stash, don't autovivify it.
7186 * Change 24660 had the side effect of causing symbol table
7187 * hashes to always be defined, even if they were freshly
7188 * created and the only reference in the entire program was
7189 * the single statement with the defined %foo::bar:: test.
7190 * It appears that all code in the wild doing this actually
7191 * wants to know whether sub-packages have been loaded, so
7192 * by avoiding auto-vivifying symbol tables, we ensure that
7193 * defined %foo::bar:: continues to be false, and the existing
7194 * tests still give the expected answers, even though what
7195 * they're actually testing has now changed subtly.
7197 (*PL_tokenbuf == '%'
7198 && *(d = PL_tokenbuf + tokenbuf_len - 1) == ':'
7201 : PL_in_eval ? (GV_ADDMULTI | GV_ADDINEVAL) : GV_ADD),
7202 ((PL_tokenbuf[0] == '$') ? SVt_PV
7203 : (PL_tokenbuf[0] == '@') ? SVt_PVAV
7209 * The following code was generated by perl_keyword.pl.
7213 Perl_keyword (pTHX_ const char *name, I32 len, bool all_keywords)
7217 PERL_ARGS_ASSERT_KEYWORD;
7221 case 1: /* 5 tokens of length 1 */
7253 case 2: /* 18 tokens of length 2 */
7399 case 3: /* 29 tokens of length 3 */
7403 if (name[1] == 'N' &&
7466 if (name[1] == 'i' &&
7498 if (name[1] == 'o' &&
7507 if (name[1] == 'e' &&
7516 if (name[1] == 'n' &&
7525 if (name[1] == 'o' &&
7534 if (name[1] == 'a' &&
7543 if (name[1] == 'o' &&
7605 if (name[1] == 'e' &&
7619 return (all_keywords || FEATURE_IS_ENABLED("say") ? KEY_say : 0);
7645 if (name[1] == 'i' &&
7654 if (name[1] == 's' &&
7663 if (name[1] == 'e' &&
7672 if (name[1] == 'o' &&
7684 case 4: /* 41 tokens of length 4 */
7688 if (name[1] == 'O' &&
7698 if (name[1] == 'N' &&
7708 if (name[1] == 'i' &&
7718 if (name[1] == 'h' &&
7728 if (name[1] == 'u' &&
7741 if (name[2] == 'c' &&
7750 if (name[2] == 's' &&
7759 if (name[2] == 'a' &&
7795 if (name[1] == 'o' &&
7808 if (name[2] == 't' &&
7817 if (name[2] == 'o' &&
7826 if (name[2] == 't' &&
7835 if (name[2] == 'e' &&
7848 if (name[1] == 'o' &&
7861 if (name[2] == 'y' &&
7870 if (name[2] == 'l' &&
7886 if (name[2] == 's' &&
7895 if (name[2] == 'n' &&
7904 if (name[2] == 'c' &&
7917 if (name[1] == 'e' &&
7927 if (name[1] == 'p' &&
7940 if (name[2] == 'c' &&
7949 if (name[2] == 'p' &&
7958 if (name[2] == 's' &&
7974 if (name[2] == 'n' &&
8044 if (name[2] == 'r' &&
8053 if (name[2] == 'r' &&
8062 if (name[2] == 'a' &&
8078 if (name[2] == 'l' &&
8140 if (name[2] == 'e' &&
8143 return (all_keywords || FEATURE_IS_ENABLED("switch") ? KEY_when : 0);
8156 case 5: /* 39 tokens of length 5 */
8160 if (name[1] == 'E' &&
8171 if (name[1] == 'H' &&
8185 if (name[2] == 'a' &&
8195 if (name[2] == 'a' &&
8212 if (name[2] == 'e' &&
8222 if (name[2] == 'e' &&
8226 return (all_keywords || FEATURE_IS_ENABLED("switch") ? -KEY_break : 0);
8242 if (name[3] == 'i' &&
8251 if (name[3] == 'o' &&
8287 if (name[2] == 'o' &&
8297 if (name[2] == 'y' &&
8311 if (name[1] == 'l' &&
8325 if (name[2] == 'n' &&
8335 if (name[2] == 'o' &&
8349 if (name[1] == 'i' &&
8354 return (all_keywords || FEATURE_IS_ENABLED("switch") ? KEY_given : 0);
8363 if (name[2] == 'd' &&
8373 if (name[2] == 'c' &&
8390 if (name[2] == 'c' &&
8400 if (name[2] == 't' &&
8414 if (name[1] == 'k' &&
8425 if (name[1] == 'r' &&
8439 if (name[2] == 's' &&
8449 if (name[2] == 'd' &&
8466 if (name[2] == 'm' &&
8476 if (name[2] == 'i' &&
8486 if (name[2] == 'e' &&
8496 if (name[2] == 'l' &&
8506 if (name[2] == 'a' &&
8519 if (name[3] == 't' &&
8522 return (all_keywords || FEATURE_IS_ENABLED("state") ? KEY_state : 0);
8528 if (name[3] == 'd' &&
8545 if (name[1] == 'i' &&
8559 if (name[2] == 'a' &&
8572 if (name[3] == 'e' &&
8607 if (name[2] == 'i' &&
8624 if (name[2] == 'i' &&
8634 if (name[2] == 'i' &&
8651 case 6: /* 33 tokens of length 6 */
8655 if (name[1] == 'c' &&
8670 if (name[2] == 'l' &&
8681 if (name[2] == 'r' &&
8696 if (name[1] == 'e' &&
8711 if (name[2] == 's' &&
8716 Perl_ck_warner_d(aTHX_ packWARN(WARN_SYNTAX), "elseif should be elsif");
8722 if (name[2] == 'i' &&
8740 if (name[2] == 'l' &&
8751 if (name[2] == 'r' &&
8766 if (name[1] == 'm' &&
8781 if (name[2] == 'n' &&
8792 if (name[2] == 's' &&
8807 if (name[1] == 's' &&
8813 if (name[4] == 't' &&
8822 if (name[4] == 'e' &&
8831 if (name[4] == 'c' &&
8840 if (name[4] == 'n' &&
8856 if (name[1] == 'r' &&
8874 if (name[3] == 'a' &&
8884 if (name[3] == 'u' &&
8898 if (name[2] == 'n' &&
8916 if (name[2] == 'a' &&
8930 if (name[3] == 'e' &&
8943 if (name[4] == 't' &&
8952 if (name[4] == 'e' &&
8974 if (name[4] == 't' &&
8983 if (name[4] == 'e' &&
8999 if (name[2] == 'c' &&
9010 if (name[2] == 'l' &&
9021 if (name[2] == 'b' &&
9032 if (name[2] == 's' &&
9055 if (name[4] == 's' &&
9064 if (name[4] == 'n' &&
9077 if (name[3] == 'a' &&
9094 if (name[1] == 'a' &&
9109 case 7: /* 29 tokens of length 7 */
9113 if (name[1] == 'E' &&
9126 if (name[1] == '_' &&
9139 if (name[1] == 'i' &&
9146 return -KEY_binmode;
9152 if (name[1] == 'o' &&
9159 return -KEY_connect;
9168 if (name[2] == 'm' &&
9174 return -KEY_dbmopen;
9185 if (name[4] == 'u' &&
9189 return (all_keywords || FEATURE_IS_ENABLED("switch") ? KEY_default : 0);
9195 if (name[4] == 'n' &&
9216 if (name[1] == 'o' &&
9229 if (name[1] == 'e' &&
9236 if (name[5] == 'r' &&
9239 return -KEY_getpgrp;
9245 if (name[5] == 'i' &&
9248 return -KEY_getppid;
9261 if (name[1] == 'c' &&
9268 return -KEY_lcfirst;
9274 if (name[1] == 'p' &&
9281 return -KEY_opendir;
9287 if (name[1] == 'a' &&
9305 if (name[3] == 'd' &&
9310 return -KEY_readdir;
9316 if (name[3] == 'u' &&
9327 if (name[3] == 'e' &&
9332 return -KEY_reverse;
9351 if (name[3] == 'k' &&
9356 return -KEY_seekdir;
9362 if (name[3] == 'p' &&
9367 return -KEY_setpgrp;
9377 if (name[2] == 'm' &&
9383 return -KEY_shmread;
9389 if (name[2] == 'r' &&
9395 return -KEY_sprintf;
9404 if (name[3] == 'l' &&
9409 return -KEY_symlink;
9418 if (name[4] == 'a' &&
9422 return -KEY_syscall;
9428 if (name[4] == 'p' &&
9432 return -KEY_sysopen;
9438 if (name[4] == 'e' &&
9442 return -KEY_sysread;
9448 if (name[4] == 'e' &&
9452 return -KEY_sysseek;
9470 if (name[1] == 'e' &&
9477 return -KEY_telldir;
9486 if (name[2] == 'f' &&
9492 return -KEY_ucfirst;
9498 if (name[2] == 's' &&
9504 return -KEY_unshift;
9514 if (name[1] == 'a' &&
9521 return -KEY_waitpid;
9530 case 8: /* 26 tokens of length 8 */
9534 if (name[1] == 'U' &&
9542 return KEY_AUTOLOAD;
9553 if (name[3] == 'A' &&
9559 return KEY___DATA__;
9565 if (name[3] == 'I' &&
9571 return -KEY___FILE__;
9577 if (name[3] == 'I' &&
9583 return -KEY___LINE__;
9599 if (name[2] == 'o' &&
9606 return -KEY_closedir;
9612 if (name[2] == 'n' &&
9619 return -KEY_continue;
9629 if (name[1] == 'b' &&
9637 return -KEY_dbmclose;
9643 if (name[1] == 'n' &&
9649 if (name[4] == 'r' &&
9654 return -KEY_endgrent;
9660 if (name[4] == 'w' &&
9665 return -KEY_endpwent;
9678 if (name[1] == 'o' &&
9686 return -KEY_formline;
9692 if (name[1] == 'e' &&
9703 if (name[6] == 'n' &&
9706 return -KEY_getgrent;
9712 if (name[6] == 'i' &&
9715 return -KEY_getgrgid;
9721 if (name[6] == 'a' &&
9724 return -KEY_getgrnam;
9737 if (name[4] == 'o' &&
9742 return -KEY_getlogin;
9753 if (name[6] == 'n' &&
9756 return -KEY_getpwent;
9762 if (name[6] == 'a' &&
9765 return -KEY_getpwnam;
9771 if (name[6] == 'i' &&
9774 return -KEY_getpwuid;
9794 if (name[1] == 'e' &&
9801 if (name[5] == 'i' &&
9808 return -KEY_readline;
9813 return -KEY_readlink;
9824 if (name[5] == 'i' &&
9828 return -KEY_readpipe;
9849 if (name[4] == 'r' &&
9854 return -KEY_setgrent;
9860 if (name[4] == 'w' &&
9865 return -KEY_setpwent;
9881 if (name[3] == 'w' &&
9887 return -KEY_shmwrite;
9893 if (name[3] == 't' &&
9899 return -KEY_shutdown;
9909 if (name[2] == 's' &&
9916 return -KEY_syswrite;
9926 if (name[1] == 'r' &&
9934 return -KEY_truncate;
9943 case 9: /* 9 tokens of length 9 */
9947 if (name[1] == 'N' &&
9956 return KEY_UNITCHECK;
9962 if (name[1] == 'n' &&
9971 return -KEY_endnetent;
9977 if (name[1] == 'e' &&
9986 return -KEY_getnetent;
9992 if (name[1] == 'o' &&
10001 return -KEY_localtime;
10007 if (name[1] == 'r' &&
10016 return KEY_prototype;
10022 if (name[1] == 'u' &&
10031 return -KEY_quotemeta;
10037 if (name[1] == 'e' &&
10046 return -KEY_rewinddir;
10052 if (name[1] == 'e' &&
10061 return -KEY_setnetent;
10067 if (name[1] == 'a' &&
10076 return -KEY_wantarray;
10085 case 10: /* 9 tokens of length 10 */
10089 if (name[1] == 'n' &&
10095 if (name[4] == 'o' &&
10102 return -KEY_endhostent;
10108 if (name[4] == 'e' &&
10115 return -KEY_endservent;
10128 if (name[1] == 'e' &&
10134 if (name[4] == 'o' &&
10141 return -KEY_gethostent;
10150 if (name[5] == 'r' &&
10156 return -KEY_getservent;
10162 if (name[5] == 'c' &&
10168 return -KEY_getsockopt;
10188 if (name[2] == 't')
10193 if (name[4] == 'o' &&
10200 return -KEY_sethostent;
10209 if (name[5] == 'r' &&
10215 return -KEY_setservent;
10221 if (name[5] == 'c' &&
10227 return -KEY_setsockopt;
10244 if (name[2] == 'c' &&
10253 return -KEY_socketpair;
10266 case 11: /* 8 tokens of length 11 */
10270 if (name[1] == '_' &&
10280 { /* __PACKAGE__ */
10281 return -KEY___PACKAGE__;
10287 if (name[1] == 'n' &&
10297 { /* endprotoent */
10298 return -KEY_endprotoent;
10304 if (name[1] == 'e' &&
10313 if (name[5] == 'e' &&
10319 { /* getpeername */
10320 return -KEY_getpeername;
10329 if (name[6] == 'o' &&
10334 { /* getpriority */
10335 return -KEY_getpriority;
10341 if (name[6] == 't' &&
10346 { /* getprotoent */
10347 return -KEY_getprotoent;
10361 if (name[4] == 'o' &&
10368 { /* getsockname */
10369 return -KEY_getsockname;
10382 if (name[1] == 'e' &&
10390 if (name[6] == 'o' &&
10395 { /* setpriority */
10396 return -KEY_setpriority;
10402 if (name[6] == 't' &&
10407 { /* setprotoent */
10408 return -KEY_setprotoent;
10424 case 12: /* 2 tokens of length 12 */
10425 if (name[0] == 'g' &&
10437 if (name[9] == 'd' &&
10440 { /* getnetbyaddr */
10441 return -KEY_getnetbyaddr;
10447 if (name[9] == 'a' &&
10450 { /* getnetbyname */
10451 return -KEY_getnetbyname;
10463 case 13: /* 4 tokens of length 13 */
10464 if (name[0] == 'g' &&
10471 if (name[4] == 'o' &&
10480 if (name[10] == 'd' &&
10483 { /* gethostbyaddr */
10484 return -KEY_gethostbyaddr;
10490 if (name[10] == 'a' &&
10493 { /* gethostbyname */
10494 return -KEY_gethostbyname;
10507 if (name[4] == 'e' &&
10516 if (name[10] == 'a' &&
10519 { /* getservbyname */
10520 return -KEY_getservbyname;
10526 if (name[10] == 'o' &&
10529 { /* getservbyport */
10530 return -KEY_getservbyport;
10549 case 14: /* 1 tokens of length 14 */
10550 if (name[0] == 'g' &&
10564 { /* getprotobyname */
10565 return -KEY_getprotobyname;
10570 case 16: /* 1 tokens of length 16 */
10571 if (name[0] == 'g' &&
10587 { /* getprotobynumber */
10588 return -KEY_getprotobynumber;
10602 S_checkcomma(pTHX_ const char *s, const char *name, const char *what)
10606 PERL_ARGS_ASSERT_CHECKCOMMA;
10608 if (*s == ' ' && s[1] == '(') { /* XXX gotta be a better way */
10609 if (ckWARN(WARN_SYNTAX)) {
10612 for (w = s+2; *w && level; w++) {
10615 else if (*w == ')')
10618 while (isSPACE(*w))
10620 /* the list of chars below is for end of statements or
10621 * block / parens, boolean operators (&&, ||, //) and branch
10622 * constructs (or, and, if, until, unless, while, err, for).
10623 * Not a very solid hack... */
10624 if (!*w || !strchr(";&/|})]oaiuwef!=", *w))
10625 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
10626 "%s (...) interpreted as function",name);
10629 while (s < PL_bufend && isSPACE(*s))
10633 while (s < PL_bufend && isSPACE(*s))
10635 if (isIDFIRST_lazy_if(s,UTF)) {
10636 const char * const w = s++;
10637 while (isALNUM_lazy_if(s,UTF))
10639 while (s < PL_bufend && isSPACE(*s))
10643 if (keyword(w, s - w, 0))
10646 gv = gv_fetchpvn_flags(w, s - w, 0, SVt_PVCV);
10647 if (gv && GvCVu(gv))
10649 Perl_croak(aTHX_ "No comma allowed after %s", what);
10654 /* Either returns sv, or mortalizes sv and returns a new SV*.
10655 Best used as sv=new_constant(..., sv, ...).
10656 If s, pv are NULL, calls subroutine with one argument,
10657 and type is used with error messages only. */
10660 S_new_constant(pTHX_ const char *s, STRLEN len, const char *key, STRLEN keylen,
10661 SV *sv, SV *pv, const char *type, STRLEN typelen)
10664 HV * const table = GvHV(PL_hintgv); /* ^H */
10668 const char *why1 = "", *why2 = "", *why3 = "";
10670 PERL_ARGS_ASSERT_NEW_CONSTANT;
10672 if (!table || !(PL_hints & HINT_LOCALIZE_HH)) {
10675 why2 = (const char *)
10676 (strEQ(key,"charnames")
10677 ? "(possibly a missing \"use charnames ...\")"
10679 msg = Perl_newSVpvf(aTHX_ "Constant(%s) unknown: %s",
10680 (type ? type: "undef"), why2);
10682 /* This is convoluted and evil ("goto considered harmful")
10683 * but I do not understand the intricacies of all the different
10684 * failure modes of %^H in here. The goal here is to make
10685 * the most probable error message user-friendly. --jhi */
10690 msg = Perl_newSVpvf(aTHX_ "Constant(%s): %s%s%s",
10691 (type ? type: "undef"), why1, why2, why3);
10693 yyerror(SvPVX_const(msg));
10697 cvp = hv_fetch(table, key, keylen, FALSE);
10698 if (!cvp || !SvOK(*cvp)) {
10701 why3 = "} is not defined";
10704 sv_2mortal(sv); /* Parent created it permanently */
10707 pv = newSVpvn_flags(s, len, SVs_TEMP);
10709 typesv = newSVpvn_flags(type, typelen, SVs_TEMP);
10711 typesv = &PL_sv_undef;
10713 PUSHSTACKi(PERLSI_OVERLOAD);
10725 call_sv(cv, G_SCALAR | ( PL_in_eval ? 0 : G_EVAL));
10729 /* Check the eval first */
10730 if (!PL_in_eval && SvTRUE(ERRSV)) {
10731 sv_catpvs(ERRSV, "Propagated");
10732 yyerror(SvPV_nolen_const(ERRSV)); /* Duplicates the message inside eval */
10734 res = SvREFCNT_inc_simple(sv);
10738 SvREFCNT_inc_simple_void(res);
10747 why1 = "Call to &{$^H{";
10749 why3 = "}} did not return a defined value";
10757 /* Returns a NUL terminated string, with the length of the string written to
10761 S_scan_word(pTHX_ register char *s, char *dest, STRLEN destlen, int allow_package, STRLEN *slp)
10764 register char *d = dest;
10765 register char * const e = d + destlen - 3; /* two-character token, ending NUL */
10767 PERL_ARGS_ASSERT_SCAN_WORD;
10771 Perl_croak(aTHX_ ident_too_long);
10772 if (isALNUM(*s)) /* UTF handled below */
10774 else if (allow_package && (*s == '\'') && isIDFIRST_lazy_if(s+1,UTF)) {
10779 else if (allow_package && (s[0] == ':') && (s[1] == ':') && (s[2] != '$')) {
10783 else if (UTF && UTF8_IS_START(*s) && isALNUM_utf8((U8*)s)) {
10784 char *t = s + UTF8SKIP(s);
10786 while (UTF8_IS_CONTINUED(*t) && is_utf8_mark((U8*)t))
10790 Perl_croak(aTHX_ ident_too_long);
10791 Copy(s, d, len, char);
10804 S_scan_ident(pTHX_ register char *s, register const char *send, char *dest, STRLEN destlen, I32 ck_uni)
10807 char *bracket = NULL;
10809 register char *d = dest;
10810 register char * const e = d + destlen + 3; /* two-character token, ending NUL */
10812 PERL_ARGS_ASSERT_SCAN_IDENT;
10817 while (isDIGIT(*s)) {
10819 Perl_croak(aTHX_ ident_too_long);
10826 Perl_croak(aTHX_ ident_too_long);
10827 if (isALNUM(*s)) /* UTF handled below */
10829 else if (*s == '\'' && isIDFIRST_lazy_if(s+1,UTF)) {
10834 else if (*s == ':' && s[1] == ':') {
10838 else if (UTF && UTF8_IS_START(*s) && isALNUM_utf8((U8*)s)) {
10839 char *t = s + UTF8SKIP(s);
10840 while (UTF8_IS_CONTINUED(*t) && is_utf8_mark((U8*)t))
10842 if (d + (t - s) > e)
10843 Perl_croak(aTHX_ ident_too_long);
10844 Copy(s, d, t - s, char);
10855 if (PL_lex_state != LEX_NORMAL)
10856 PL_lex_state = LEX_INTERPENDMAYBE;
10859 if (*s == '$' && s[1] &&
10860 (isALNUM_lazy_if(s+1,UTF) || s[1] == '$' || s[1] == '{' || strnEQ(s+1,"::",2)) )
10873 if (*d == '^' && *s && isCONTROLVAR(*s)) {
10878 if (isSPACE(s[-1])) {
10880 const char ch = *s++;
10881 if (!SPACE_OR_TAB(ch)) {
10887 if (isIDFIRST_lazy_if(d,UTF)) {
10891 while ((end < send && isALNUM_lazy_if(end,UTF)) || *end == ':') {
10892 end += UTF8SKIP(end);
10893 while (end < send && UTF8_IS_CONTINUED(*end) && is_utf8_mark((U8*)end))
10894 end += UTF8SKIP(end);
10896 Copy(s, d, end - s, char);
10901 while ((isALNUM(*s) || *s == ':') && d < e)
10904 Perl_croak(aTHX_ ident_too_long);
10907 while (s < send && SPACE_OR_TAB(*s))
10909 if ((*s == '[' || (*s == '{' && strNE(dest, "sub")))) {
10910 if (ckWARN(WARN_AMBIGUOUS) && keyword(dest, d - dest, 0)) {
10911 const char * const brack =
10913 ((*s == '[') ? "[...]" : "{...}");
10914 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
10915 "Ambiguous use of %c{%s%s} resolved to %c%s%s",
10916 funny, dest, brack, funny, dest, brack);
10919 PL_lex_brackstack[PL_lex_brackets++] = (char)(XOPERATOR | XFAKEBRACK);
10923 /* Handle extended ${^Foo} variables
10924 * 1999-02-27 mjd-perl-patch@plover.com */
10925 else if (!isALNUM(*d) && !isPRINT(*d) /* isCTRL(d) */
10929 while (isALNUM(*s) && d < e) {
10933 Perl_croak(aTHX_ ident_too_long);
10938 if (PL_lex_state == LEX_INTERPNORMAL && !PL_lex_brackets) {
10939 PL_lex_state = LEX_INTERPEND;
10942 if (PL_lex_state == LEX_NORMAL) {
10943 if (ckWARN(WARN_AMBIGUOUS) &&
10944 (keyword(dest, d - dest, 0)
10945 || get_cvn_flags(dest, d - dest, 0)))
10949 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
10950 "Ambiguous use of %c{%s} resolved to %c%s",
10951 funny, dest, funny, dest);
10956 s = bracket; /* let the parser handle it */
10960 else if (PL_lex_state == LEX_INTERPNORMAL && !PL_lex_brackets && !intuit_more(s))
10961 PL_lex_state = LEX_INTERPEND;
10966 Perl_pmflag(pTHX_ U32* pmfl, int ch)
10968 PERL_ARGS_ASSERT_PMFLAG;
10970 PERL_UNUSED_CONTEXT;
10972 const char c = (char)ch;
10974 CASE_STD_PMMOD_FLAGS_PARSE_SET(pmfl);
10975 case GLOBAL_PAT_MOD: *pmfl |= PMf_GLOBAL; break;
10976 case CONTINUE_PAT_MOD: *pmfl |= PMf_CONTINUE; break;
10977 case ONCE_PAT_MOD: *pmfl |= PMf_KEEP; break;
10978 case KEEPCOPY_PAT_MOD: *pmfl |= PMf_KEEPCOPY; break;
10984 S_scan_pat(pTHX_ char *start, I32 type)
10988 char *s = scan_str(start,!!PL_madskills,FALSE);
10989 const char * const valid_flags =
10990 (const char *)((type == OP_QR) ? QR_PAT_MODS : M_PAT_MODS);
10995 PERL_ARGS_ASSERT_SCAN_PAT;
10998 const char * const delimiter = skipspace(start);
11002 ? "Search pattern not terminated or ternary operator parsed as search pattern"
11003 : "Search pattern not terminated" ));
11006 pm = (PMOP*)newPMOP(type, 0);
11007 if (PL_multi_open == '?') {
11008 /* This is the only point in the code that sets PMf_ONCE: */
11009 pm->op_pmflags |= PMf_ONCE;
11011 /* Hence it's safe to do this bit of PMOP book-keeping here, which
11012 allows us to restrict the list needed by reset to just the ??
11014 assert(type != OP_TRANS);
11016 MAGIC *mg = mg_find((const SV *)PL_curstash, PERL_MAGIC_symtab);
11019 mg = sv_magicext(MUTABLE_SV(PL_curstash), 0, PERL_MAGIC_symtab, 0, 0,
11022 elements = mg->mg_len / sizeof(PMOP**);
11023 Renewc(mg->mg_ptr, elements + 1, PMOP*, char);
11024 ((PMOP**)mg->mg_ptr) [elements++] = pm;
11025 mg->mg_len = elements * sizeof(PMOP**);
11026 PmopSTASH_set(pm,PL_curstash);
11032 while (*s && strchr(valid_flags, *s))
11033 pmflag(&pm->op_pmflags,*s++);
11035 if (PL_madskills && modstart != s) {
11036 SV* tmptoken = newSVpvn(modstart, s - modstart);
11037 append_madprops(newMADPROP('m', MAD_SV, tmptoken, 0), (OP*)pm, 0);
11040 /* issue a warning if /c is specified,but /g is not */
11041 if ((pm->op_pmflags & PMf_CONTINUE) && !(pm->op_pmflags & PMf_GLOBAL))
11043 Perl_ck_warner(aTHX_ packWARN(WARN_REGEXP),
11044 "Use of /c modifier is meaningless without /g" );
11047 PL_lex_op = (OP*)pm;
11048 pl_yylval.ival = OP_MATCH;
11053 S_scan_subst(pTHX_ char *start)
11064 PERL_ARGS_ASSERT_SCAN_SUBST;
11066 pl_yylval.ival = OP_NULL;
11068 s = scan_str(start,!!PL_madskills,FALSE);
11071 Perl_croak(aTHX_ "Substitution pattern not terminated");
11073 if (s[-1] == PL_multi_open)
11076 if (PL_madskills) {
11077 CURMAD('q', PL_thisopen);
11078 CURMAD('_', PL_thiswhite);
11079 CURMAD('E', PL_thisstuff);
11080 CURMAD('Q', PL_thisclose);
11081 PL_realtokenstart = s - SvPVX(PL_linestr);
11085 first_start = PL_multi_start;
11086 s = scan_str(s,!!PL_madskills,FALSE);
11088 if (PL_lex_stuff) {
11089 SvREFCNT_dec(PL_lex_stuff);
11090 PL_lex_stuff = NULL;
11092 Perl_croak(aTHX_ "Substitution replacement not terminated");
11094 PL_multi_start = first_start; /* so whole substitution is taken together */
11096 pm = (PMOP*)newPMOP(OP_SUBST, 0);
11099 if (PL_madskills) {
11100 CURMAD('z', PL_thisopen);
11101 CURMAD('R', PL_thisstuff);
11102 CURMAD('Z', PL_thisclose);
11108 if (*s == EXEC_PAT_MOD) {
11112 else if (strchr(S_PAT_MODS, *s))
11113 pmflag(&pm->op_pmflags,*s++);
11119 if (PL_madskills) {
11121 curmad('m', newSVpvn(modstart, s - modstart));
11122 append_madprops(PL_thismad, (OP*)pm, 0);
11126 if ((pm->op_pmflags & PMf_CONTINUE)) {
11127 Perl_ck_warner(aTHX_ packWARN(WARN_REGEXP), "Use of /c modifier is meaningless in s///" );
11131 SV * const repl = newSVpvs("");
11133 PL_sublex_info.super_bufptr = s;
11134 PL_sublex_info.super_bufend = PL_bufend;
11136 pm->op_pmflags |= PMf_EVAL;
11139 sv_catpvs(repl, "eval ");
11141 sv_catpvs(repl, "do ");
11143 sv_catpvs(repl, "{");
11144 sv_catsv(repl, PL_lex_repl);
11145 if (strchr(SvPVX(PL_lex_repl), '#'))
11146 sv_catpvs(repl, "\n");
11147 sv_catpvs(repl, "}");
11149 SvREFCNT_dec(PL_lex_repl);
11150 PL_lex_repl = repl;
11153 PL_lex_op = (OP*)pm;
11154 pl_yylval.ival = OP_SUBST;
11159 S_scan_trans(pTHX_ char *start)
11172 PERL_ARGS_ASSERT_SCAN_TRANS;
11174 pl_yylval.ival = OP_NULL;
11176 s = scan_str(start,!!PL_madskills,FALSE);
11178 Perl_croak(aTHX_ "Transliteration pattern not terminated");
11180 if (s[-1] == PL_multi_open)
11183 if (PL_madskills) {
11184 CURMAD('q', PL_thisopen);
11185 CURMAD('_', PL_thiswhite);
11186 CURMAD('E', PL_thisstuff);
11187 CURMAD('Q', PL_thisclose);
11188 PL_realtokenstart = s - SvPVX(PL_linestr);
11192 s = scan_str(s,!!PL_madskills,FALSE);
11194 if (PL_lex_stuff) {
11195 SvREFCNT_dec(PL_lex_stuff);
11196 PL_lex_stuff = NULL;
11198 Perl_croak(aTHX_ "Transliteration replacement not terminated");
11200 if (PL_madskills) {
11201 CURMAD('z', PL_thisopen);
11202 CURMAD('R', PL_thisstuff);
11203 CURMAD('Z', PL_thisclose);
11206 complement = del = squash = 0;
11213 complement = OPpTRANS_COMPLEMENT;
11216 del = OPpTRANS_DELETE;
11219 squash = OPpTRANS_SQUASH;
11228 tbl = (short *)PerlMemShared_calloc(complement&&!del?258:256, sizeof(short));
11229 o = newPVOP(OP_TRANS, 0, (char*)tbl);
11230 o->op_private &= ~OPpTRANS_ALL;
11231 o->op_private |= del|squash|complement|
11232 (DO_UTF8(PL_lex_stuff)? OPpTRANS_FROM_UTF : 0)|
11233 (DO_UTF8(PL_lex_repl) ? OPpTRANS_TO_UTF : 0);
11236 pl_yylval.ival = OP_TRANS;
11239 if (PL_madskills) {
11241 curmad('m', newSVpvn(modstart, s - modstart));
11242 append_madprops(PL_thismad, o, 0);
11251 S_scan_heredoc(pTHX_ register char *s)
11255 I32 op_type = OP_SCALAR;
11259 const char *found_newline;
11263 const int outer = (PL_rsfp && !(PL_lex_inwhat == OP_SCALAR));
11265 I32 stuffstart = s - SvPVX(PL_linestr);
11268 PL_realtokenstart = -1;
11271 PERL_ARGS_ASSERT_SCAN_HEREDOC;
11275 e = PL_tokenbuf + sizeof PL_tokenbuf - 1;
11279 while (SPACE_OR_TAB(*peek))
11281 if (*peek == '`' || *peek == '\'' || *peek =='"') {
11284 s = delimcpy(d, e, s, PL_bufend, term, &len);
11294 if (!isALNUM_lazy_if(s,UTF))
11295 deprecate_old("bare << to mean <<\"\"");
11296 for (; isALNUM_lazy_if(s,UTF); s++) {
11301 if (d >= PL_tokenbuf + sizeof PL_tokenbuf - 1)
11302 Perl_croak(aTHX_ "Delimiter for here document is too long");
11305 len = d - PL_tokenbuf;
11308 if (PL_madskills) {
11309 tstart = PL_tokenbuf + !outer;
11310 PL_thisclose = newSVpvn(tstart, len - !outer);
11311 tstart = SvPVX(PL_linestr) + stuffstart;
11312 PL_thisopen = newSVpvn(tstart, s - tstart);
11313 stuffstart = s - SvPVX(PL_linestr);
11316 #ifndef PERL_STRICT_CR
11317 d = strchr(s, '\r');
11319 char * const olds = s;
11321 while (s < PL_bufend) {
11327 else if (*s == '\n' && s[1] == '\r') { /* \015\013 on a mac? */
11336 SvCUR_set(PL_linestr, PL_bufend - SvPVX_const(PL_linestr));
11343 if ( outer || !(found_newline = (char*)memchr((void*)s, '\n', PL_bufend - s)) ) {
11344 herewas = newSVpvn(s,PL_bufend-s);
11348 herewas = newSVpvn(s-1,found_newline-s+1);
11351 herewas = newSVpvn(s,found_newline-s);
11355 if (PL_madskills) {
11356 tstart = SvPVX(PL_linestr) + stuffstart;
11358 sv_catpvn(PL_thisstuff, tstart, s - tstart);
11360 PL_thisstuff = newSVpvn(tstart, s - tstart);
11363 s += SvCUR(herewas);
11366 stuffstart = s - SvPVX(PL_linestr);
11372 tmpstr = newSV_type(SVt_PVIV);
11373 SvGROW(tmpstr, 80);
11374 if (term == '\'') {
11375 op_type = OP_CONST;
11376 SvIV_set(tmpstr, -1);
11378 else if (term == '`') {
11379 op_type = OP_BACKTICK;
11380 SvIV_set(tmpstr, '\\');
11384 PL_multi_start = CopLINE(PL_curcop);
11385 PL_multi_open = PL_multi_close = '<';
11386 term = *PL_tokenbuf;
11387 if (PL_lex_inwhat == OP_SUBST && PL_in_eval && !PL_rsfp) {
11388 char * const bufptr = PL_sublex_info.super_bufptr;
11389 char * const bufend = PL_sublex_info.super_bufend;
11390 char * const olds = s - SvCUR(herewas);
11391 s = strchr(bufptr, '\n');
11395 while (s < bufend &&
11396 (*s != term || memNE(s,PL_tokenbuf,len)) ) {
11398 CopLINE_inc(PL_curcop);
11401 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
11402 missingterm(PL_tokenbuf);
11404 sv_setpvn(herewas,bufptr,d-bufptr+1);
11405 sv_setpvn(tmpstr,d+1,s-d);
11407 sv_catpvn(herewas,s,bufend-s);
11408 Copy(SvPVX_const(herewas),bufptr,SvCUR(herewas) + 1,char);
11415 while (s < PL_bufend &&
11416 (*s != term || memNE(s,PL_tokenbuf,len)) ) {
11418 CopLINE_inc(PL_curcop);
11420 if (s >= PL_bufend) {
11421 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
11422 missingterm(PL_tokenbuf);
11424 sv_setpvn(tmpstr,d+1,s-d);
11426 if (PL_madskills) {
11428 sv_catpvn(PL_thisstuff, d + 1, s - d);
11430 PL_thisstuff = newSVpvn(d + 1, s - d);
11431 stuffstart = s - SvPVX(PL_linestr);
11435 CopLINE_inc(PL_curcop); /* the preceding stmt passes a newline */
11437 sv_catpvn(herewas,s,PL_bufend-s);
11438 sv_setsv(PL_linestr,herewas);
11439 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = s = PL_linestart = SvPVX(PL_linestr);
11440 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
11441 PL_last_lop = PL_last_uni = NULL;
11444 sv_setpvs(tmpstr,""); /* avoid "uninitialized" warning */
11445 while (s >= PL_bufend) { /* multiple line string? */
11447 if (PL_madskills) {
11448 tstart = SvPVX(PL_linestr) + stuffstart;
11450 sv_catpvn(PL_thisstuff, tstart, PL_bufend - tstart);
11452 PL_thisstuff = newSVpvn(tstart, PL_bufend - tstart);
11456 !(PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = filter_gets(PL_linestr, PL_rsfp, 0))) {
11457 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
11458 missingterm(PL_tokenbuf);
11461 stuffstart = s - SvPVX(PL_linestr);
11463 CopLINE_inc(PL_curcop);
11464 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
11465 PL_last_lop = PL_last_uni = NULL;
11466 #ifndef PERL_STRICT_CR
11467 if (PL_bufend - PL_linestart >= 2) {
11468 if ((PL_bufend[-2] == '\r' && PL_bufend[-1] == '\n') ||
11469 (PL_bufend[-2] == '\n' && PL_bufend[-1] == '\r'))
11471 PL_bufend[-2] = '\n';
11473 SvCUR_set(PL_linestr, PL_bufend - SvPVX_const(PL_linestr));
11475 else if (PL_bufend[-1] == '\r')
11476 PL_bufend[-1] = '\n';
11478 else if (PL_bufend - PL_linestart == 1 && PL_bufend[-1] == '\r')
11479 PL_bufend[-1] = '\n';
11481 if ((PERLDB_LINE || PERLDB_SAVESRC) && PL_curstash != PL_debstash)
11482 update_debugger_info(PL_linestr, NULL, 0);
11483 if (*s == term && memEQ(s,PL_tokenbuf,len)) {
11484 STRLEN off = PL_bufend - 1 - SvPVX_const(PL_linestr);
11485 *(SvPVX(PL_linestr) + off ) = ' ';
11486 sv_catsv(PL_linestr,herewas);
11487 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
11488 s = SvPVX(PL_linestr) + off; /* In case PV of PL_linestr moved. */
11492 sv_catsv(tmpstr,PL_linestr);
11497 PL_multi_end = CopLINE(PL_curcop);
11498 if (SvCUR(tmpstr) + 5 < SvLEN(tmpstr)) {
11499 SvPV_shrink_to_cur(tmpstr);
11501 SvREFCNT_dec(herewas);
11503 if (UTF && is_utf8_string((U8*)SvPVX_const(tmpstr), SvCUR(tmpstr)))
11505 else if (PL_encoding)
11506 sv_recode_to_utf8(tmpstr, PL_encoding);
11508 PL_lex_stuff = tmpstr;
11509 pl_yylval.ival = op_type;
11513 /* scan_inputsymbol
11514 takes: current position in input buffer
11515 returns: new position in input buffer
11516 side-effects: pl_yylval and lex_op are set.
11521 <FH> read from filehandle
11522 <pkg::FH> read from package qualified filehandle
11523 <pkg'FH> read from package qualified filehandle
11524 <$fh> read from filehandle in $fh
11525 <*.h> filename glob
11530 S_scan_inputsymbol(pTHX_ char *start)
11533 register char *s = start; /* current position in buffer */
11536 char *d = PL_tokenbuf; /* start of temp holding space */
11537 const char * const e = PL_tokenbuf + sizeof PL_tokenbuf; /* end of temp holding space */
11539 PERL_ARGS_ASSERT_SCAN_INPUTSYMBOL;
11541 end = strchr(s, '\n');
11544 s = delimcpy(d, e, s + 1, end, '>', &len); /* extract until > */
11546 /* die if we didn't have space for the contents of the <>,
11547 or if it didn't end, or if we see a newline
11550 if (len >= (I32)sizeof PL_tokenbuf)
11551 Perl_croak(aTHX_ "Excessively long <> operator");
11553 Perl_croak(aTHX_ "Unterminated <> operator");
11558 Remember, only scalar variables are interpreted as filehandles by
11559 this code. Anything more complex (e.g., <$fh{$num}>) will be
11560 treated as a glob() call.
11561 This code makes use of the fact that except for the $ at the front,
11562 a scalar variable and a filehandle look the same.
11564 if (*d == '$' && d[1]) d++;
11566 /* allow <Pkg'VALUE> or <Pkg::VALUE> */
11567 while (*d && (isALNUM_lazy_if(d,UTF) || *d == '\'' || *d == ':'))
11570 /* If we've tried to read what we allow filehandles to look like, and
11571 there's still text left, then it must be a glob() and not a getline.
11572 Use scan_str to pull out the stuff between the <> and treat it
11573 as nothing more than a string.
11576 if (d - PL_tokenbuf != len) {
11577 pl_yylval.ival = OP_GLOB;
11578 s = scan_str(start,!!PL_madskills,FALSE);
11580 Perl_croak(aTHX_ "Glob not terminated");
11584 bool readline_overriden = FALSE;
11587 /* we're in a filehandle read situation */
11590 /* turn <> into <ARGV> */
11592 Copy("ARGV",d,5,char);
11594 /* Check whether readline() is overriden */
11595 gv_readline = gv_fetchpvs("readline", GV_NOTQUAL, SVt_PVCV);
11597 && GvCVu(gv_readline) && GvIMPORTED_CV(gv_readline))
11599 ((gvp = (GV**)hv_fetchs(PL_globalstash, "readline", FALSE))
11600 && (gv_readline = *gvp) && isGV_with_GP(gv_readline)
11601 && GvCVu(gv_readline) && GvIMPORTED_CV(gv_readline)))
11602 readline_overriden = TRUE;
11604 /* if <$fh>, create the ops to turn the variable into a
11608 /* try to find it in the pad for this block, otherwise find
11609 add symbol table ops
11611 const PADOFFSET tmp = pad_findmy(d);
11612 if (tmp != NOT_IN_PAD) {
11613 if (PAD_COMPNAME_FLAGS_isOUR(tmp)) {
11614 HV * const stash = PAD_COMPNAME_OURSTASH(tmp);
11615 HEK * const stashname = HvNAME_HEK(stash);
11616 SV * const sym = sv_2mortal(newSVhek(stashname));
11617 sv_catpvs(sym, "::");
11618 sv_catpv(sym, d+1);
11623 OP * const o = newOP(OP_PADSV, 0);
11625 PL_lex_op = readline_overriden
11626 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
11627 append_elem(OP_LIST, o,
11628 newCVREF(0, newGVOP(OP_GV,0,gv_readline))))
11629 : (OP*)newUNOP(OP_READLINE, 0, o);
11638 ? (GV_ADDMULTI | GV_ADDINEVAL)
11641 PL_lex_op = readline_overriden
11642 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
11643 append_elem(OP_LIST,
11644 newUNOP(OP_RV2SV, 0, newGVOP(OP_GV, 0, gv)),
11645 newCVREF(0, newGVOP(OP_GV, 0, gv_readline))))
11646 : (OP*)newUNOP(OP_READLINE, 0,
11647 newUNOP(OP_RV2SV, 0,
11648 newGVOP(OP_GV, 0, gv)));
11650 if (!readline_overriden)
11651 PL_lex_op->op_flags |= OPf_SPECIAL;
11652 /* we created the ops in PL_lex_op, so make pl_yylval.ival a null op */
11653 pl_yylval.ival = OP_NULL;
11656 /* If it's none of the above, it must be a literal filehandle
11657 (<Foo::BAR> or <FOO>) so build a simple readline OP */
11659 GV * const gv = gv_fetchpv(d, GV_ADD, SVt_PVIO);
11660 PL_lex_op = readline_overriden
11661 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
11662 append_elem(OP_LIST,
11663 newGVOP(OP_GV, 0, gv),
11664 newCVREF(0, newGVOP(OP_GV, 0, gv_readline))))
11665 : (OP*)newUNOP(OP_READLINE, 0, newGVOP(OP_GV, 0, gv));
11666 pl_yylval.ival = OP_NULL;
11675 takes: start position in buffer
11676 keep_quoted preserve \ on the embedded delimiter(s)
11677 keep_delims preserve the delimiters around the string
11678 returns: position to continue reading from buffer
11679 side-effects: multi_start, multi_close, lex_repl or lex_stuff, and
11680 updates the read buffer.
11682 This subroutine pulls a string out of the input. It is called for:
11683 q single quotes q(literal text)
11684 ' single quotes 'literal text'
11685 qq double quotes qq(interpolate $here please)
11686 " double quotes "interpolate $here please"
11687 qx backticks qx(/bin/ls -l)
11688 ` backticks `/bin/ls -l`
11689 qw quote words @EXPORT_OK = qw( func() $spam )
11690 m// regexp match m/this/
11691 s/// regexp substitute s/this/that/
11692 tr/// string transliterate tr/this/that/
11693 y/// string transliterate y/this/that/
11694 ($*@) sub prototypes sub foo ($)
11695 (stuff) sub attr parameters sub foo : attr(stuff)
11696 <> readline or globs <FOO>, <>, <$fh>, or <*.c>
11698 In most of these cases (all but <>, patterns and transliterate)
11699 yylex() calls scan_str(). m// makes yylex() call scan_pat() which
11700 calls scan_str(). s/// makes yylex() call scan_subst() which calls
11701 scan_str(). tr/// and y/// make yylex() call scan_trans() which
11704 It skips whitespace before the string starts, and treats the first
11705 character as the delimiter. If the delimiter is one of ([{< then
11706 the corresponding "close" character )]}> is used as the closing
11707 delimiter. It allows quoting of delimiters, and if the string has
11708 balanced delimiters ([{<>}]) it allows nesting.
11710 On success, the SV with the resulting string is put into lex_stuff or,
11711 if that is already non-NULL, into lex_repl. The second case occurs only
11712 when parsing the RHS of the special constructs s/// and tr/// (y///).
11713 For convenience, the terminating delimiter character is stuffed into
11718 S_scan_str(pTHX_ char *start, int keep_quoted, int keep_delims)
11721 SV *sv; /* scalar value: string */
11722 const char *tmps; /* temp string, used for delimiter matching */
11723 register char *s = start; /* current position in the buffer */
11724 register char term; /* terminating character */
11725 register char *to; /* current position in the sv's data */
11726 I32 brackets = 1; /* bracket nesting level */
11727 bool has_utf8 = FALSE; /* is there any utf8 content? */
11728 I32 termcode; /* terminating char. code */
11729 U8 termstr[UTF8_MAXBYTES]; /* terminating string */
11730 STRLEN termlen; /* length of terminating string */
11731 int last_off = 0; /* last position for nesting bracket */
11737 PERL_ARGS_ASSERT_SCAN_STR;
11739 /* skip space before the delimiter */
11745 if (PL_realtokenstart >= 0) {
11746 stuffstart = PL_realtokenstart;
11747 PL_realtokenstart = -1;
11750 stuffstart = start - SvPVX(PL_linestr);
11752 /* mark where we are, in case we need to report errors */
11755 /* after skipping whitespace, the next character is the terminator */
11758 termcode = termstr[0] = term;
11762 termcode = utf8_to_uvchr((U8*)s, &termlen);
11763 Copy(s, termstr, termlen, U8);
11764 if (!UTF8_IS_INVARIANT(term))
11768 /* mark where we are */
11769 PL_multi_start = CopLINE(PL_curcop);
11770 PL_multi_open = term;
11772 /* find corresponding closing delimiter */
11773 if (term && (tmps = strchr("([{< )]}> )]}>",term)))
11774 termcode = termstr[0] = term = tmps[5];
11776 PL_multi_close = term;
11778 /* create a new SV to hold the contents. 79 is the SV's initial length.
11779 What a random number. */
11780 sv = newSV_type(SVt_PVIV);
11782 SvIV_set(sv, termcode);
11783 (void)SvPOK_only(sv); /* validate pointer */
11785 /* move past delimiter and try to read a complete string */
11787 sv_catpvn(sv, s, termlen);
11790 tstart = SvPVX(PL_linestr) + stuffstart;
11791 if (!PL_thisopen && !keep_delims) {
11792 PL_thisopen = newSVpvn(tstart, s - tstart);
11793 stuffstart = s - SvPVX(PL_linestr);
11797 if (PL_encoding && !UTF) {
11801 int offset = s - SvPVX_const(PL_linestr);
11802 const bool found = sv_cat_decode(sv, PL_encoding, PL_linestr,
11803 &offset, (char*)termstr, termlen);
11804 const char * const ns = SvPVX_const(PL_linestr) + offset;
11805 char * const svlast = SvEND(sv) - 1;
11807 for (; s < ns; s++) {
11808 if (*s == '\n' && !PL_rsfp)
11809 CopLINE_inc(PL_curcop);
11812 goto read_more_line;
11814 /* handle quoted delimiters */
11815 if (SvCUR(sv) > 1 && *(svlast-1) == '\\') {
11817 for (t = svlast-2; t >= SvPVX_const(sv) && *t == '\\';)
11819 if ((svlast-1 - t) % 2) {
11820 if (!keep_quoted) {
11821 *(svlast-1) = term;
11823 SvCUR_set(sv, SvCUR(sv) - 1);
11828 if (PL_multi_open == PL_multi_close) {
11834 for (t = w = SvPVX(sv)+last_off; t < svlast; w++, t++) {
11835 /* At here, all closes are "was quoted" one,
11836 so we don't check PL_multi_close. */
11838 if (!keep_quoted && *(t+1) == PL_multi_open)
11843 else if (*t == PL_multi_open)
11851 SvCUR_set(sv, w - SvPVX_const(sv));
11853 last_off = w - SvPVX(sv);
11854 if (--brackets <= 0)
11859 if (!keep_delims) {
11860 SvCUR_set(sv, SvCUR(sv) - 1);
11866 /* extend sv if need be */
11867 SvGROW(sv, SvCUR(sv) + (PL_bufend - s) + 1);
11868 /* set 'to' to the next character in the sv's string */
11869 to = SvPVX(sv)+SvCUR(sv);
11871 /* if open delimiter is the close delimiter read unbridle */
11872 if (PL_multi_open == PL_multi_close) {
11873 for (; s < PL_bufend; s++,to++) {
11874 /* embedded newlines increment the current line number */
11875 if (*s == '\n' && !PL_rsfp)
11876 CopLINE_inc(PL_curcop);
11877 /* handle quoted delimiters */
11878 if (*s == '\\' && s+1 < PL_bufend && term != '\\') {
11879 if (!keep_quoted && s[1] == term)
11881 /* any other quotes are simply copied straight through */
11885 /* terminate when run out of buffer (the for() condition), or
11886 have found the terminator */
11887 else if (*s == term) {
11890 if (s+termlen <= PL_bufend && memEQ(s, (char*)termstr, termlen))
11893 else if (!has_utf8 && !UTF8_IS_INVARIANT((U8)*s) && UTF)
11899 /* if the terminator isn't the same as the start character (e.g.,
11900 matched brackets), we have to allow more in the quoting, and
11901 be prepared for nested brackets.
11904 /* read until we run out of string, or we find the terminator */
11905 for (; s < PL_bufend; s++,to++) {
11906 /* embedded newlines increment the line count */
11907 if (*s == '\n' && !PL_rsfp)
11908 CopLINE_inc(PL_curcop);
11909 /* backslashes can escape the open or closing characters */
11910 if (*s == '\\' && s+1 < PL_bufend) {
11911 if (!keep_quoted &&
11912 ((s[1] == PL_multi_open) || (s[1] == PL_multi_close)))
11917 /* allow nested opens and closes */
11918 else if (*s == PL_multi_close && --brackets <= 0)
11920 else if (*s == PL_multi_open)
11922 else if (!has_utf8 && !UTF8_IS_INVARIANT((U8)*s) && UTF)
11927 /* terminate the copied string and update the sv's end-of-string */
11929 SvCUR_set(sv, to - SvPVX_const(sv));
11932 * this next chunk reads more into the buffer if we're not done yet
11936 break; /* handle case where we are done yet :-) */
11938 #ifndef PERL_STRICT_CR
11939 if (to - SvPVX_const(sv) >= 2) {
11940 if ((to[-2] == '\r' && to[-1] == '\n') ||
11941 (to[-2] == '\n' && to[-1] == '\r'))
11945 SvCUR_set(sv, to - SvPVX_const(sv));
11947 else if (to[-1] == '\r')
11950 else if (to - SvPVX_const(sv) == 1 && to[-1] == '\r')
11955 /* if we're out of file, or a read fails, bail and reset the current
11956 line marker so we can report where the unterminated string began
11959 if (PL_madskills) {
11960 char * const tstart = SvPVX(PL_linestr) + stuffstart;
11962 sv_catpvn(PL_thisstuff, tstart, PL_bufend - tstart);
11964 PL_thisstuff = newSVpvn(tstart, PL_bufend - tstart);
11968 !(PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = filter_gets(PL_linestr, PL_rsfp, 0))) {
11970 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
11976 /* we read a line, so increment our line counter */
11977 CopLINE_inc(PL_curcop);
11979 /* update debugger info */
11980 if ((PERLDB_LINE || PERLDB_SAVESRC) && PL_curstash != PL_debstash)
11981 update_debugger_info(PL_linestr, NULL, 0);
11983 /* having changed the buffer, we must update PL_bufend */
11984 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
11985 PL_last_lop = PL_last_uni = NULL;
11988 /* at this point, we have successfully read the delimited string */
11990 if (!PL_encoding || UTF) {
11992 if (PL_madskills) {
11993 char * const tstart = SvPVX(PL_linestr) + stuffstart;
11994 const int len = s - tstart;
11996 sv_catpvn(PL_thisstuff, tstart, len);
11998 PL_thisstuff = newSVpvn(tstart, len);
11999 if (!PL_thisclose && !keep_delims)
12000 PL_thisclose = newSVpvn(s,termlen);
12005 sv_catpvn(sv, s, termlen);
12010 if (PL_madskills) {
12011 char * const tstart = SvPVX(PL_linestr) + stuffstart;
12012 const int len = s - tstart - termlen;
12014 sv_catpvn(PL_thisstuff, tstart, len);
12016 PL_thisstuff = newSVpvn(tstart, len);
12017 if (!PL_thisclose && !keep_delims)
12018 PL_thisclose = newSVpvn(s - termlen,termlen);
12022 if (has_utf8 || PL_encoding)
12025 PL_multi_end = CopLINE(PL_curcop);
12027 /* if we allocated too much space, give some back */
12028 if (SvCUR(sv) + 5 < SvLEN(sv)) {
12029 SvLEN_set(sv, SvCUR(sv) + 1);
12030 SvPV_renew(sv, SvLEN(sv));
12033 /* decide whether this is the first or second quoted string we've read
12046 takes: pointer to position in buffer
12047 returns: pointer to new position in buffer
12048 side-effects: builds ops for the constant in pl_yylval.op
12050 Read a number in any of the formats that Perl accepts:
12052 \d(_?\d)*(\.(\d(_?\d)*)?)?[Ee][\+\-]?(\d(_?\d)*) 12 12.34 12.
12053 \.\d(_?\d)*[Ee][\+\-]?(\d(_?\d)*) .34
12056 0x[0-9A-Fa-f](_?[0-9A-Fa-f])*
12058 Like most scan_ routines, it uses the PL_tokenbuf buffer to hold the
12061 If it reads a number without a decimal point or an exponent, it will
12062 try converting the number to an integer and see if it can do so
12063 without loss of precision.
12067 Perl_scan_num(pTHX_ const char *start, YYSTYPE* lvalp)
12070 register const char *s = start; /* current position in buffer */
12071 register char *d; /* destination in temp buffer */
12072 register char *e; /* end of temp buffer */
12073 NV nv; /* number read, as a double */
12074 SV *sv = NULL; /* place to put the converted number */
12075 bool floatit; /* boolean: int or float? */
12076 const char *lastub = NULL; /* position of last underbar */
12077 static char const number_too_long[] = "Number too long";
12079 PERL_ARGS_ASSERT_SCAN_NUM;
12081 /* We use the first character to decide what type of number this is */
12085 Perl_croak(aTHX_ "panic: scan_num");
12087 /* if it starts with a 0, it could be an octal number, a decimal in
12088 0.13 disguise, or a hexadecimal number, or a binary number. */
12092 u holds the "number so far"
12093 shift the power of 2 of the base
12094 (hex == 4, octal == 3, binary == 1)
12095 overflowed was the number more than we can hold?
12097 Shift is used when we add a digit. It also serves as an "are
12098 we in octal/hex/binary?" indicator to disallow hex characters
12099 when in octal mode.
12104 bool overflowed = FALSE;
12105 bool just_zero = TRUE; /* just plain 0 or binary number? */
12106 static const NV nvshift[5] = { 1.0, 2.0, 4.0, 8.0, 16.0 };
12107 static const char* const bases[5] =
12108 { "", "binary", "", "octal", "hexadecimal" };
12109 static const char* const Bases[5] =
12110 { "", "Binary", "", "Octal", "Hexadecimal" };
12111 static const char* const maxima[5] =
12113 "0b11111111111111111111111111111111",
12117 const char *base, *Base, *max;
12119 /* check for hex */
12124 } else if (s[1] == 'b') {
12129 /* check for a decimal in disguise */
12130 else if (s[1] == '.' || s[1] == 'e' || s[1] == 'E')
12132 /* so it must be octal */
12139 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
12140 "Misplaced _ in number");
12144 base = bases[shift];
12145 Base = Bases[shift];
12146 max = maxima[shift];
12148 /* read the rest of the number */
12150 /* x is used in the overflow test,
12151 b is the digit we're adding on. */
12156 /* if we don't mention it, we're done */
12160 /* _ are ignored -- but warned about if consecutive */
12162 if (lastub && s == lastub + 1)
12163 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
12164 "Misplaced _ in number");
12168 /* 8 and 9 are not octal */
12169 case '8': case '9':
12171 yyerror(Perl_form(aTHX_ "Illegal octal digit '%c'", *s));
12175 case '2': case '3': case '4':
12176 case '5': case '6': case '7':
12178 yyerror(Perl_form(aTHX_ "Illegal binary digit '%c'", *s));
12181 case '0': case '1':
12182 b = *s++ & 15; /* ASCII digit -> value of digit */
12186 case 'a': case 'b': case 'c': case 'd': case 'e': case 'f':
12187 case 'A': case 'B': case 'C': case 'D': case 'E': case 'F':
12188 /* make sure they said 0x */
12191 b = (*s++ & 7) + 9;
12193 /* Prepare to put the digit we have onto the end
12194 of the number so far. We check for overflows.
12200 x = u << shift; /* make room for the digit */
12202 if ((x >> shift) != u
12203 && !(PL_hints & HINT_NEW_BINARY)) {
12206 Perl_ck_warner_d(aTHX_ packWARN(WARN_OVERFLOW),
12207 "Integer overflow in %s number",
12210 u = x | b; /* add the digit to the end */
12213 n *= nvshift[shift];
12214 /* If an NV has not enough bits in its
12215 * mantissa to represent an UV this summing of
12216 * small low-order numbers is a waste of time
12217 * (because the NV cannot preserve the
12218 * low-order bits anyway): we could just
12219 * remember when did we overflow and in the
12220 * end just multiply n by the right
12228 /* if we get here, we had success: make a scalar value from
12233 /* final misplaced underbar check */
12234 if (s[-1] == '_') {
12235 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX), "Misplaced _ in number");
12240 if (n > 4294967295.0)
12241 Perl_ck_warner(aTHX_ packWARN(WARN_PORTABLE),
12242 "%s number > %s non-portable",
12248 if (u > 0xffffffff)
12249 Perl_ck_warner(aTHX_ packWARN(WARN_PORTABLE),
12250 "%s number > %s non-portable",
12255 if (just_zero && (PL_hints & HINT_NEW_INTEGER))
12256 sv = new_constant(start, s - start, "integer",
12257 sv, NULL, NULL, 0);
12258 else if (PL_hints & HINT_NEW_BINARY)
12259 sv = new_constant(start, s - start, "binary", sv, NULL, NULL, 0);
12264 handle decimal numbers.
12265 we're also sent here when we read a 0 as the first digit
12267 case '1': case '2': case '3': case '4': case '5':
12268 case '6': case '7': case '8': case '9': case '.':
12271 e = PL_tokenbuf + sizeof PL_tokenbuf - 6; /* room for various punctuation */
12274 /* read next group of digits and _ and copy into d */
12275 while (isDIGIT(*s) || *s == '_') {
12276 /* skip underscores, checking for misplaced ones
12280 if (lastub && s == lastub + 1)
12281 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
12282 "Misplaced _ in number");
12286 /* check for end of fixed-length buffer */
12288 Perl_croak(aTHX_ number_too_long);
12289 /* if we're ok, copy the character */
12294 /* final misplaced underbar check */
12295 if (lastub && s == lastub + 1) {
12296 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX), "Misplaced _ in number");
12299 /* read a decimal portion if there is one. avoid
12300 3..5 being interpreted as the number 3. followed
12303 if (*s == '.' && s[1] != '.') {
12308 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
12309 "Misplaced _ in number");
12313 /* copy, ignoring underbars, until we run out of digits.
12315 for (; isDIGIT(*s) || *s == '_'; s++) {
12316 /* fixed length buffer check */
12318 Perl_croak(aTHX_ number_too_long);
12320 if (lastub && s == lastub + 1)
12321 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
12322 "Misplaced _ in number");
12328 /* fractional part ending in underbar? */
12329 if (s[-1] == '_') {
12330 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
12331 "Misplaced _ in number");
12333 if (*s == '.' && isDIGIT(s[1])) {
12334 /* oops, it's really a v-string, but without the "v" */
12340 /* read exponent part, if present */
12341 if ((*s == 'e' || *s == 'E') && strchr("+-0123456789_", s[1])) {
12345 /* regardless of whether user said 3E5 or 3e5, use lower 'e' */
12346 *d++ = 'e'; /* At least some Mach atof()s don't grok 'E' */
12348 /* stray preinitial _ */
12350 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
12351 "Misplaced _ in number");
12355 /* allow positive or negative exponent */
12356 if (*s == '+' || *s == '-')
12359 /* stray initial _ */
12361 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
12362 "Misplaced _ in number");
12366 /* read digits of exponent */
12367 while (isDIGIT(*s) || *s == '_') {
12370 Perl_croak(aTHX_ number_too_long);
12374 if (((lastub && s == lastub + 1) ||
12375 (!isDIGIT(s[1]) && s[1] != '_')))
12376 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
12377 "Misplaced _ in number");
12384 /* make an sv from the string */
12388 We try to do an integer conversion first if no characters
12389 indicating "float" have been found.
12394 const int flags = grok_number (PL_tokenbuf, d - PL_tokenbuf, &uv);
12396 if (flags == IS_NUMBER_IN_UV) {
12398 sv_setiv(sv, uv); /* Prefer IVs over UVs. */
12401 } else if (flags == (IS_NUMBER_IN_UV | IS_NUMBER_NEG)) {
12402 if (uv <= (UV) IV_MIN)
12403 sv_setiv(sv, -(IV)uv);
12410 /* terminate the string */
12412 nv = Atof(PL_tokenbuf);
12417 ? (PL_hints & HINT_NEW_FLOAT) : (PL_hints & HINT_NEW_INTEGER) ) {
12418 const char *const key = floatit ? "float" : "integer";
12419 const STRLEN keylen = floatit ? 5 : 7;
12420 sv = S_new_constant(aTHX_ PL_tokenbuf, d - PL_tokenbuf,
12421 key, keylen, sv, NULL, NULL, 0);
12425 /* if it starts with a v, it could be a v-string */
12428 sv = newSV(5); /* preallocate storage space */
12429 s = scan_vstring(s, PL_bufend, sv);
12433 /* make the op for the constant and return */
12436 lvalp->opval = newSVOP(OP_CONST, 0, sv);
12438 lvalp->opval = NULL;
12444 S_scan_formline(pTHX_ register char *s)
12447 register char *eol;
12449 SV * const stuff = newSVpvs("");
12450 bool needargs = FALSE;
12451 bool eofmt = FALSE;
12453 char *tokenstart = s;
12454 SV* savewhite = NULL;
12456 if (PL_madskills) {
12457 savewhite = PL_thiswhite;
12462 PERL_ARGS_ASSERT_SCAN_FORMLINE;
12464 while (!needargs) {
12467 #ifdef PERL_STRICT_CR
12468 while (SPACE_OR_TAB(*t))
12471 while (SPACE_OR_TAB(*t) || *t == '\r')
12474 if (*t == '\n' || t == PL_bufend) {
12479 if (PL_in_eval && !PL_rsfp) {
12480 eol = (char *) memchr(s,'\n',PL_bufend-s);
12485 eol = PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
12487 for (t = s; t < eol; t++) {
12488 if (*t == '~' && t[1] == '~' && SvCUR(stuff)) {
12490 goto enough; /* ~~ must be first line in formline */
12492 if (*t == '@' || *t == '^')
12496 sv_catpvn(stuff, s, eol-s);
12497 #ifndef PERL_STRICT_CR
12498 if (eol-s > 1 && eol[-2] == '\r' && eol[-1] == '\n') {
12499 char *end = SvPVX(stuff) + SvCUR(stuff);
12502 SvCUR_set(stuff, SvCUR(stuff) - 1);
12512 if (PL_madskills) {
12514 sv_catpvn(PL_thistoken, tokenstart, PL_bufend - tokenstart);
12516 PL_thistoken = newSVpvn(tokenstart, PL_bufend - tokenstart);
12519 s = filter_gets(PL_linestr, PL_rsfp, 0);
12521 tokenstart = PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = SvPVX(PL_linestr);
12523 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = SvPVX(PL_linestr);
12525 PL_bufend = PL_bufptr + SvCUR(PL_linestr);
12526 PL_last_lop = PL_last_uni = NULL;
12535 if (SvCUR(stuff)) {
12538 PL_lex_state = LEX_NORMAL;
12539 start_force(PL_curforce);
12540 NEXTVAL_NEXTTOKE.ival = 0;
12544 PL_lex_state = LEX_FORMLINE;
12546 if (UTF && is_utf8_string((U8*)SvPVX_const(stuff), SvCUR(stuff)))
12548 else if (PL_encoding)
12549 sv_recode_to_utf8(stuff, PL_encoding);
12551 start_force(PL_curforce);
12552 NEXTVAL_NEXTTOKE.opval = (OP*)newSVOP(OP_CONST, 0, stuff);
12554 start_force(PL_curforce);
12555 NEXTVAL_NEXTTOKE.ival = OP_FORMLINE;
12559 SvREFCNT_dec(stuff);
12561 PL_lex_formbrack = 0;
12565 if (PL_madskills) {
12567 sv_catpvn(PL_thistoken, tokenstart, s - tokenstart);
12569 PL_thistoken = newSVpvn(tokenstart, s - tokenstart);
12570 PL_thiswhite = savewhite;
12577 Perl_start_subparse(pTHX_ I32 is_format, U32 flags)
12580 const I32 oldsavestack_ix = PL_savestack_ix;
12581 CV* const outsidecv = PL_compcv;
12584 assert(SvTYPE(PL_compcv) == SVt_PVCV);
12586 SAVEI32(PL_subline);
12587 save_item(PL_subname);
12588 SAVESPTR(PL_compcv);
12590 PL_compcv = MUTABLE_CV(newSV_type(is_format ? SVt_PVFM : SVt_PVCV));
12591 CvFLAGS(PL_compcv) |= flags;
12593 PL_subline = CopLINE(PL_curcop);
12594 CvPADLIST(PL_compcv) = pad_new(padnew_SAVE|padnew_SAVESUB);
12595 CvOUTSIDE(PL_compcv) = MUTABLE_CV(SvREFCNT_inc_simple(outsidecv));
12596 CvOUTSIDE_SEQ(PL_compcv) = PL_cop_seqmax;
12598 return oldsavestack_ix;
12602 #pragma segment Perl_yylex
12605 S_yywarn(pTHX_ const char *const s)
12609 PERL_ARGS_ASSERT_YYWARN;
12611 PL_in_eval |= EVAL_WARNONLY;
12613 PL_in_eval &= ~EVAL_WARNONLY;
12618 Perl_yyerror(pTHX_ const char *const s)
12621 const char *where = NULL;
12622 const char *context = NULL;
12625 int yychar = PL_parser->yychar;
12627 PERL_ARGS_ASSERT_YYERROR;
12629 if (!yychar || (yychar == ';' && !PL_rsfp))
12631 else if (PL_oldoldbufptr && PL_bufptr > PL_oldoldbufptr &&
12632 PL_bufptr - PL_oldoldbufptr < 200 && PL_oldoldbufptr != PL_oldbufptr &&
12633 PL_oldbufptr != PL_bufptr) {
12636 The code below is removed for NetWare because it abends/crashes on NetWare
12637 when the script has error such as not having the closing quotes like:
12638 if ($var eq "value)
12639 Checking of white spaces is anyway done in NetWare code.
12642 while (isSPACE(*PL_oldoldbufptr))
12645 context = PL_oldoldbufptr;
12646 contlen = PL_bufptr - PL_oldoldbufptr;
12648 else if (PL_oldbufptr && PL_bufptr > PL_oldbufptr &&
12649 PL_bufptr - PL_oldbufptr < 200 && PL_oldbufptr != PL_bufptr) {
12652 The code below is removed for NetWare because it abends/crashes on NetWare
12653 when the script has error such as not having the closing quotes like:
12654 if ($var eq "value)
12655 Checking of white spaces is anyway done in NetWare code.
12658 while (isSPACE(*PL_oldbufptr))
12661 context = PL_oldbufptr;
12662 contlen = PL_bufptr - PL_oldbufptr;
12664 else if (yychar > 255)
12665 where = "next token ???";
12666 else if (yychar == -2) { /* YYEMPTY */
12667 if (PL_lex_state == LEX_NORMAL ||
12668 (PL_lex_state == LEX_KNOWNEXT && PL_lex_defer == LEX_NORMAL))
12669 where = "at end of line";
12670 else if (PL_lex_inpat)
12671 where = "within pattern";
12673 where = "within string";
12676 SV * const where_sv = newSVpvs_flags("next char ", SVs_TEMP);
12678 Perl_sv_catpvf(aTHX_ where_sv, "^%c", toCTRL(yychar));
12679 else if (isPRINT_LC(yychar)) {
12680 const char string = yychar;
12681 sv_catpvn(where_sv, &string, 1);
12684 Perl_sv_catpvf(aTHX_ where_sv, "\\%03o", yychar & 255);
12685 where = SvPVX_const(where_sv);
12687 msg = sv_2mortal(newSVpv(s, 0));
12688 Perl_sv_catpvf(aTHX_ msg, " at %s line %"IVdf", ",
12689 OutCopFILE(PL_curcop), (IV)CopLINE(PL_curcop));
12691 Perl_sv_catpvf(aTHX_ msg, "near \"%.*s\"\n", contlen, context);
12693 Perl_sv_catpvf(aTHX_ msg, "%s\n", where);
12694 if (PL_multi_start < PL_multi_end && (U32)(CopLINE(PL_curcop) - PL_multi_end) <= 1) {
12695 Perl_sv_catpvf(aTHX_ msg,
12696 " (Might be a runaway multi-line %c%c string starting on line %"IVdf")\n",
12697 (int)PL_multi_open,(int)PL_multi_close,(IV)PL_multi_start);
12700 if (PL_in_eval & EVAL_WARNONLY) {
12701 Perl_ck_warner_d(aTHX_ packWARN(WARN_SYNTAX), "%"SVf, SVfARG(msg));
12705 if (PL_error_count >= 10) {
12706 if (PL_in_eval && SvCUR(ERRSV))
12707 Perl_croak(aTHX_ "%"SVf"%s has too many errors.\n",
12708 SVfARG(ERRSV), OutCopFILE(PL_curcop));
12710 Perl_croak(aTHX_ "%s has too many errors.\n",
12711 OutCopFILE(PL_curcop));
12714 PL_in_my_stash = NULL;
12718 #pragma segment Main
12722 S_swallow_bom(pTHX_ U8 *s)
12725 const STRLEN slen = SvCUR(PL_linestr);
12727 PERL_ARGS_ASSERT_SWALLOW_BOM;
12731 if (s[1] == 0xFE) {
12732 /* UTF-16 little-endian? (or UTF32-LE?) */
12733 if (s[2] == 0 && s[3] == 0) /* UTF-32 little-endian */
12734 Perl_croak(aTHX_ "Unsupported script encoding UTF32-LE");
12735 #ifndef PERL_NO_UTF16_FILTER
12736 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF16-LE script encoding (BOM)\n");
12739 if (PL_bufend > (char*)s) {
12743 filter_add(utf16rev_textfilter, NULL);
12744 Newx(news, (PL_bufend - (char*)s) * 3 / 2 + 1, U8);
12745 utf16_to_utf8_reversed(s, news,
12746 PL_bufend - (char*)s - 1,
12748 sv_setpvn(PL_linestr, (const char*)news, newlen);
12750 s = (U8*)SvPVX(PL_linestr);
12751 Copy(news, s, newlen, U8);
12755 SvUTF8_on(PL_linestr);
12756 s = (U8*)SvPVX(PL_linestr);
12758 /* FIXME - is this a general bug fix? */
12761 PL_bufend = SvPVX(PL_linestr) + newlen;
12764 Perl_croak(aTHX_ "Unsupported script encoding UTF16-LE");
12769 if (s[1] == 0xFF) { /* UTF-16 big-endian? */
12770 #ifndef PERL_NO_UTF16_FILTER
12771 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16BE script encoding (BOM)\n");
12774 if (PL_bufend > (char *)s) {
12778 filter_add(utf16_textfilter, NULL);
12779 Newx(news, (PL_bufend - (char*)s) * 3 / 2 + 1, U8);
12780 utf16_to_utf8(s, news,
12781 PL_bufend - (char*)s,
12783 sv_setpvn(PL_linestr, (const char*)news, newlen);
12785 SvUTF8_on(PL_linestr);
12786 s = (U8*)SvPVX(PL_linestr);
12787 PL_bufend = SvPVX(PL_linestr) + newlen;
12790 Perl_croak(aTHX_ "Unsupported script encoding UTF16-BE");
12795 if (slen > 2 && s[1] == 0xBB && s[2] == 0xBF) {
12796 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-8 script encoding (BOM)\n");
12797 s += 3; /* UTF-8 */
12803 if (s[2] == 0xFE && s[3] == 0xFF) {
12804 /* UTF-32 big-endian */
12805 Perl_croak(aTHX_ "Unsupported script encoding UTF32-BE");
12808 else if (s[2] == 0 && s[3] != 0) {
12811 * are a good indicator of UTF-16BE. */
12812 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16BE script encoding (no BOM)\n");
12818 if (slen > 3 && s[1] == 0x73 && s[2] == 0x66 && s[3] == 0x73) {
12819 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-8 script encoding (BOM)\n");
12820 s += 4; /* UTF-8 */
12826 if (slen > 3 && s[1] == 0 && s[2] != 0 && s[3] == 0) {
12829 * are a good indicator of UTF-16LE. */
12830 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16LE script encoding (no BOM)\n");
12838 #ifndef PERL_NO_UTF16_FILTER
12840 utf16_textfilter(pTHX_ int idx, SV *sv, int maxlen)
12843 const STRLEN old = SvCUR(sv);
12844 const I32 count = FILTER_READ(idx+1, sv, maxlen);
12845 DEBUG_P(PerlIO_printf(Perl_debug_log,
12846 "utf16_textfilter(%p): %d %d (%d)\n",
12847 FPTR2DPTR(void *, utf16_textfilter),
12848 idx, maxlen, (int) count));
12852 Newx(tmps, SvCUR(sv) * 3 / 2 + 1, U8);
12853 Copy(SvPVX_const(sv), tmps, old, char);
12854 utf16_to_utf8((U8*)SvPVX_const(sv) + old, tmps + old,
12855 SvCUR(sv) - old, &newlen);
12856 sv_usepvn(sv, (char*)tmps, (STRLEN)newlen + old);
12858 DEBUG_P({sv_dump(sv);});
12863 utf16rev_textfilter(pTHX_ int idx, SV *sv, int maxlen)
12866 const STRLEN old = SvCUR(sv);
12867 const I32 count = FILTER_READ(idx+1, sv, maxlen);
12868 DEBUG_P(PerlIO_printf(Perl_debug_log,
12869 "utf16rev_textfilter(%p): %d %d (%d)\n",
12870 FPTR2DPTR(void *, utf16rev_textfilter),
12871 idx, maxlen, (int) count));
12875 Newx(tmps, SvCUR(sv) * 3 / 2 + 1, U8);
12876 Copy(SvPVX_const(sv), tmps, old, char);
12877 utf16_to_utf8((U8*)SvPVX_const(sv) + old, tmps + old,
12878 SvCUR(sv) - old, &newlen);
12879 sv_usepvn(sv, (char*)tmps, (STRLEN)newlen + old);
12881 DEBUG_P({ sv_dump(sv); });
12887 Returns a pointer to the next character after the parsed
12888 vstring, as well as updating the passed in sv.
12890 Function must be called like
12893 s = scan_vstring(s,e,sv);
12895 where s and e are the start and end of the string.
12896 The sv should already be large enough to store the vstring
12897 passed in, for performance reasons.
12902 Perl_scan_vstring(pTHX_ const char *s, const char *const e, SV *sv)
12905 const char *pos = s;
12906 const char *start = s;
12908 PERL_ARGS_ASSERT_SCAN_VSTRING;
12910 if (*pos == 'v') pos++; /* get past 'v' */
12911 while (pos < e && (isDIGIT(*pos) || *pos == '_'))
12913 if ( *pos != '.') {
12914 /* this may not be a v-string if followed by => */
12915 const char *next = pos;
12916 while (next < e && isSPACE(*next))
12918 if ((e - next) >= 2 && *next == '=' && next[1] == '>' ) {
12919 /* return string not v-string */
12920 sv_setpvn(sv,(char *)s,pos-s);
12921 return (char *)pos;
12925 if (!isALPHA(*pos)) {
12926 U8 tmpbuf[UTF8_MAXBYTES+1];
12929 s++; /* get past 'v' */
12934 /* this is atoi() that tolerates underscores */
12937 const char *end = pos;
12939 while (--end >= s) {
12941 const UV orev = rev;
12942 rev += (*end - '0') * mult;
12945 Perl_ck_warner_d(aTHX_ packWARN(WARN_OVERFLOW),
12946 "Integer overflow in decimal number");
12950 if (rev > 0x7FFFFFFF)
12951 Perl_croak(aTHX_ "In EBCDIC the v-string components cannot exceed 2147483647");
12953 /* Append native character for the rev point */
12954 tmpend = uvchr_to_utf8(tmpbuf, rev);
12955 sv_catpvn(sv, (const char*)tmpbuf, tmpend - tmpbuf);
12956 if (!UNI_IS_INVARIANT(NATIVE_TO_UNI(rev)))
12958 if (pos + 1 < e && *pos == '.' && isDIGIT(pos[1]))
12964 while (pos < e && (isDIGIT(*pos) || *pos == '_'))
12968 sv_magic(sv,NULL,PERL_MAGIC_vstring,(const char*)start, pos-start);
12976 * c-indentation-style: bsd
12977 * c-basic-offset: 4
12978 * indent-tabs-mode: t
12981 * ex: set ts=8 sts=4 sw=4 noet: