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 if (ckWARN(WARN_DEPRECATED))
597 Perl_warner(aTHX_ packWARN(WARN_DEPRECATED), "Use of %s is deprecated", s);
601 Perl_deprecate_old(pTHX_ const char *const s)
603 /* This function should NOT be called for any new deprecated warnings */
604 /* Use Perl_deprecate instead */
606 /* It is here to maintain backward compatibility with the pre-5.8 */
607 /* warnings category hierarchy. The "deprecated" category used to */
608 /* live under the "syntax" category. It is now a top-level category */
609 /* in its own right. */
611 PERL_ARGS_ASSERT_DEPRECATE_OLD;
613 if (ckWARN2(WARN_DEPRECATED, WARN_SYNTAX))
614 Perl_warner(aTHX_ packWARN2(WARN_DEPRECATED, WARN_SYNTAX),
615 "Use of %s is deprecated", s);
619 * experimental text filters for win32 carriage-returns, utf16-to-utf8 and
620 * utf16-to-utf8-reversed.
623 #ifdef PERL_CR_FILTER
627 register const char *s = SvPVX_const(sv);
628 register const char * const e = s + SvCUR(sv);
630 PERL_ARGS_ASSERT_STRIP_RETURN;
632 /* outer loop optimized to do nothing if there are no CR-LFs */
634 if (*s++ == '\r' && *s == '\n') {
635 /* hit a CR-LF, need to copy the rest */
636 register char *d = s - 1;
639 if (*s == '\r' && s[1] == '\n')
650 S_cr_textfilter(pTHX_ int idx, SV *sv, int maxlen)
652 const I32 count = FILTER_READ(idx+1, sv, maxlen);
653 if (count > 0 && !maxlen)
664 * Create a parser object and initialise its parser and lexer fields
666 * rsfp is the opened file handle to read from (if any),
668 * line holds any initial content already read from the file (or in
669 * the case of no file, such as an eval, the whole contents);
671 * new_filter indicates that this is a new file and it shouldn't inherit
672 * the filters from the current parser (ie require).
676 Perl_lex_start(pTHX_ SV *line, PerlIO *rsfp, bool new_filter)
679 const char *s = NULL;
681 yy_parser *parser, *oparser;
683 /* create and initialise a parser */
685 Newxz(parser, 1, yy_parser);
686 parser->old_parser = oparser = PL_parser;
689 Newx(parser->stack, YYINITDEPTH, yy_stack_frame);
690 parser->ps = parser->stack;
691 parser->stack_size = YYINITDEPTH;
693 parser->stack->state = 0;
694 parser->yyerrstatus = 0;
695 parser->yychar = YYEMPTY; /* Cause a token to be read. */
697 /* on scope exit, free this parser and restore any outer one */
699 parser->saved_curcop = PL_curcop;
701 /* initialise lexer state */
704 parser->curforce = -1;
706 parser->nexttoke = 0;
708 parser->error_count = oparser ? oparser->error_count : 0;
709 parser->copline = NOLINE;
710 parser->lex_state = LEX_NORMAL;
711 parser->expect = XSTATE;
713 parser->rsfp_filters = (new_filter || !oparser) ? newAV()
714 : MUTABLE_AV(SvREFCNT_inc(oparser->rsfp_filters));
716 Newx(parser->lex_brackstack, 120, char);
717 Newx(parser->lex_casestack, 12, char);
718 *parser->lex_casestack = '\0';
721 s = SvPV_const(line, len);
727 parser->linestr = newSVpvs("\n;");
728 } else if (SvREADONLY(line) || s[len-1] != ';') {
729 parser->linestr = newSVsv(line);
731 sv_catpvs(parser->linestr, "\n;");
734 SvREFCNT_inc_simple_void_NN(line);
735 parser->linestr = line;
737 parser->oldoldbufptr =
740 parser->linestart = SvPVX(parser->linestr);
741 parser->bufend = parser->bufptr + SvCUR(parser->linestr);
742 parser->last_lop = parser->last_uni = NULL;
746 /* delete a parser object */
749 Perl_parser_free(pTHX_ const yy_parser *parser)
751 PERL_ARGS_ASSERT_PARSER_FREE;
753 PL_curcop = parser->saved_curcop;
754 SvREFCNT_dec(parser->linestr);
756 if (parser->rsfp == PerlIO_stdin())
757 PerlIO_clearerr(parser->rsfp);
758 else if (parser->rsfp && (!parser->old_parser ||
759 (parser->old_parser && parser->rsfp != parser->old_parser->rsfp)))
760 PerlIO_close(parser->rsfp);
761 SvREFCNT_dec(parser->rsfp_filters);
763 Safefree(parser->stack);
764 Safefree(parser->lex_brackstack);
765 Safefree(parser->lex_casestack);
766 PL_parser = parser->old_parser;
773 * Finalizer for lexing operations. Must be called when the parser is
774 * done with the lexer.
781 PL_doextract = FALSE;
786 * This subroutine has nothing to do with tilting, whether at windmills
787 * or pinball tables. Its name is short for "increment line". It
788 * increments the current line number in CopLINE(PL_curcop) and checks
789 * to see whether the line starts with a comment of the form
790 * # line 500 "foo.pm"
791 * If so, it sets the current line number and file to the values in the comment.
795 S_incline(pTHX_ const char *s)
802 PERL_ARGS_ASSERT_INCLINE;
804 CopLINE_inc(PL_curcop);
807 while (SPACE_OR_TAB(*s))
809 if (strnEQ(s, "line", 4))
813 if (SPACE_OR_TAB(*s))
817 while (SPACE_OR_TAB(*s))
825 if (!SPACE_OR_TAB(*s) && *s != '\r' && *s != '\n' && *s != '\0')
827 while (SPACE_OR_TAB(*s))
829 if (*s == '"' && (t = strchr(s+1, '"'))) {
839 while (SPACE_OR_TAB(*e) || *e == '\r' || *e == '\f')
841 if (*e != '\n' && *e != '\0')
842 return; /* false alarm */
845 const STRLEN len = t - s;
847 SV *const temp_sv = CopFILESV(PL_curcop);
853 tmplen = SvCUR(temp_sv);
859 if (tmplen > 7 && strnEQ(cf, "(eval ", 6)) {
860 /* must copy *{"::_<(eval N)[oldfilename:L]"}
861 * to *{"::_<newfilename"} */
862 /* However, the long form of evals is only turned on by the
863 debugger - usually they're "(eval %lu)" */
867 STRLEN tmplen2 = len;
868 if (tmplen + 2 <= sizeof smallbuf)
871 Newx(tmpbuf, tmplen + 2, char);
874 memcpy(tmpbuf + 2, cf, tmplen);
876 gvp = (GV**)hv_fetch(PL_defstash, tmpbuf, tmplen, FALSE);
881 if (tmplen2 + 2 <= sizeof smallbuf)
884 Newx(tmpbuf2, tmplen2 + 2, char);
886 if (tmpbuf2 != smallbuf || tmpbuf != smallbuf) {
887 /* Either they malloc'd it, or we malloc'd it,
888 so no prefix is present in ours. */
893 memcpy(tmpbuf2 + 2, s, tmplen2);
896 gv2 = *(GV**)hv_fetch(PL_defstash, tmpbuf2, tmplen2, TRUE);
898 gv_init(gv2, PL_defstash, tmpbuf2, tmplen2, FALSE);
899 /* adjust ${"::_<newfilename"} to store the new file name */
900 GvSV(gv2) = newSVpvn(tmpbuf2 + 2, tmplen2 - 2);
901 GvHV(gv2) = MUTABLE_HV(SvREFCNT_inc(GvHV(*gvp)));
902 GvAV(gv2) = MUTABLE_AV(SvREFCNT_inc(GvAV(*gvp)));
905 if (tmpbuf2 != smallbuf) Safefree(tmpbuf2);
907 if (tmpbuf != smallbuf) Safefree(tmpbuf);
910 CopFILE_free(PL_curcop);
911 CopFILE_setn(PL_curcop, s, len);
913 CopLINE_set(PL_curcop, atoi(n)-1);
917 /* skip space before PL_thistoken */
920 S_skipspace0(pTHX_ register char *s)
922 PERL_ARGS_ASSERT_SKIPSPACE0;
929 PL_thiswhite = newSVpvs("");
930 sv_catsv(PL_thiswhite, PL_skipwhite);
931 sv_free(PL_skipwhite);
934 PL_realtokenstart = s - SvPVX(PL_linestr);
938 /* skip space after PL_thistoken */
941 S_skipspace1(pTHX_ register char *s)
943 const char *start = s;
944 I32 startoff = start - SvPVX(PL_linestr);
946 PERL_ARGS_ASSERT_SKIPSPACE1;
951 start = SvPVX(PL_linestr) + startoff;
952 if (!PL_thistoken && PL_realtokenstart >= 0) {
953 const char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
954 PL_thistoken = newSVpvn(tstart, start - tstart);
956 PL_realtokenstart = -1;
959 PL_nextwhite = newSVpvs("");
960 sv_catsv(PL_nextwhite, PL_skipwhite);
961 sv_free(PL_skipwhite);
968 S_skipspace2(pTHX_ register char *s, SV **svp)
971 const I32 bufptroff = PL_bufptr - SvPVX(PL_linestr);
972 const I32 startoff = s - SvPVX(PL_linestr);
974 PERL_ARGS_ASSERT_SKIPSPACE2;
977 PL_bufptr = SvPVX(PL_linestr) + bufptroff;
978 if (!PL_madskills || !svp)
980 start = SvPVX(PL_linestr) + startoff;
981 if (!PL_thistoken && PL_realtokenstart >= 0) {
982 char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
983 PL_thistoken = newSVpvn(tstart, start - tstart);
984 PL_realtokenstart = -1;
989 sv_setsv(*svp, PL_skipwhite);
990 sv_free(PL_skipwhite);
999 S_update_debugger_info(pTHX_ SV *orig_sv, const char *const buf, STRLEN len)
1001 AV *av = CopFILEAVx(PL_curcop);
1003 SV * const sv = newSV_type(SVt_PVMG);
1005 sv_setsv(sv, orig_sv);
1007 sv_setpvn(sv, buf, len);
1010 av_store(av, (I32)CopLINE(PL_curcop), sv);
1016 * Called to gobble the appropriate amount and type of whitespace.
1017 * Skips comments as well.
1021 S_skipspace(pTHX_ register char *s)
1026 int startoff = s - SvPVX(PL_linestr);
1028 PERL_ARGS_ASSERT_SKIPSPACE;
1031 sv_free(PL_skipwhite);
1035 PERL_ARGS_ASSERT_SKIPSPACE;
1037 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
1038 while (s < PL_bufend && SPACE_OR_TAB(*s))
1048 SSize_t oldprevlen, oldoldprevlen;
1049 SSize_t oldloplen = 0, oldunilen = 0;
1050 while (s < PL_bufend && isSPACE(*s)) {
1051 if (*s++ == '\n' && PL_in_eval && !PL_rsfp)
1056 if (s < PL_bufend && *s == '#') {
1057 while (s < PL_bufend && *s != '\n')
1059 if (s < PL_bufend) {
1061 if (PL_in_eval && !PL_rsfp) {
1068 /* only continue to recharge the buffer if we're at the end
1069 * of the buffer, we're not reading from a source filter, and
1070 * we're in normal lexing mode
1072 if (s < PL_bufend || !PL_rsfp || PL_sublex_info.sub_inwhat ||
1073 PL_lex_state == LEX_FORMLINE)
1080 /* try to recharge the buffer */
1082 curoff = s - SvPVX(PL_linestr);
1085 if ((s = filter_gets(PL_linestr, PL_rsfp,
1086 (prevlen = SvCUR(PL_linestr)))) == NULL)
1089 if (PL_madskills && curoff != startoff) {
1091 PL_skipwhite = newSVpvs("");
1092 sv_catpvn(PL_skipwhite, SvPVX(PL_linestr) + startoff,
1096 /* mustn't throw out old stuff yet if madpropping */
1097 SvCUR(PL_linestr) = curoff;
1098 s = SvPVX(PL_linestr) + curoff;
1100 if (curoff && s[-1] == '\n')
1104 /* end of file. Add on the -p or -n magic */
1105 /* XXX these shouldn't really be added here, can't set PL_faketokens */
1108 sv_catpvs(PL_linestr,
1109 ";}continue{print or die qq(-p destination: $!\\n);}");
1111 sv_setpvs(PL_linestr,
1112 ";}continue{print or die qq(-p destination: $!\\n);}");
1114 PL_minus_n = PL_minus_p = 0;
1116 else if (PL_minus_n) {
1118 sv_catpvs(PL_linestr, ";}");
1120 sv_setpvs(PL_linestr, ";}");
1126 sv_catpvs(PL_linestr,";");
1128 sv_setpvs(PL_linestr,";");
1131 /* reset variables for next time we lex */
1132 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = s = PL_linestart
1138 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
1139 PL_last_lop = PL_last_uni = NULL;
1141 /* Close the filehandle. Could be from
1142 * STDIN, or a regular file. If we were reading code from
1143 * STDIN (because the commandline held no -e or filename)
1144 * then we don't close it, we reset it so the code can
1145 * read from STDIN too.
1148 if ((PerlIO*)PL_rsfp == PerlIO_stdin())
1149 PerlIO_clearerr(PL_rsfp);
1151 (void)PerlIO_close(PL_rsfp);
1156 /* not at end of file, so we only read another line */
1157 /* make corresponding updates to old pointers, for yyerror() */
1158 oldprevlen = PL_oldbufptr - PL_bufend;
1159 oldoldprevlen = PL_oldoldbufptr - PL_bufend;
1161 oldunilen = PL_last_uni - PL_bufend;
1163 oldloplen = PL_last_lop - PL_bufend;
1164 PL_linestart = PL_bufptr = s + prevlen;
1165 PL_bufend = s + SvCUR(PL_linestr);
1167 PL_oldbufptr = s + oldprevlen;
1168 PL_oldoldbufptr = s + oldoldprevlen;
1170 PL_last_uni = s + oldunilen;
1172 PL_last_lop = s + oldloplen;
1175 /* debugger active and we're not compiling the debugger code,
1176 * so store the line into the debugger's array of lines
1178 if ((PERLDB_LINE || PERLDB_SAVESRC) && PL_curstash != PL_debstash)
1179 update_debugger_info(NULL, PL_bufptr, PL_bufend - PL_bufptr);
1186 PL_skipwhite = newSVpvs("");
1187 curoff = s - SvPVX(PL_linestr);
1188 if (curoff - startoff)
1189 sv_catpvn(PL_skipwhite, SvPVX(PL_linestr) + startoff,
1198 * Check the unary operators to ensure there's no ambiguity in how they're
1199 * used. An ambiguous piece of code would be:
1201 * This doesn't mean rand() + 5. Because rand() is a unary operator,
1202 * the +5 is its argument.
1212 if (PL_oldoldbufptr != PL_last_uni)
1214 while (isSPACE(*PL_last_uni))
1217 while (isALNUM_lazy_if(s,UTF) || *s == '-')
1219 if ((t = strchr(s, '(')) && t < PL_bufptr)
1222 if (ckWARN_d(WARN_AMBIGUOUS)){
1223 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
1224 "Warning: Use of \"%.*s\" without parentheses is ambiguous",
1225 (int)(s - PL_last_uni), PL_last_uni);
1230 * LOP : macro to build a list operator. Its behaviour has been replaced
1231 * with a subroutine, S_lop() for which LOP is just another name.
1234 #define LOP(f,x) return lop(f,x,s)
1238 * Build a list operator (or something that might be one). The rules:
1239 * - if we have a next token, then it's a list operator [why?]
1240 * - if the next thing is an opening paren, then it's a function
1241 * - else it's a list operator
1245 S_lop(pTHX_ I32 f, int x, char *s)
1249 PERL_ARGS_ASSERT_LOP;
1255 PL_last_lop = PL_oldbufptr;
1256 PL_last_lop_op = (OPCODE)f;
1259 return REPORT(LSTOP);
1262 return REPORT(LSTOP);
1265 return REPORT(FUNC);
1268 return REPORT(FUNC);
1270 return REPORT(LSTOP);
1276 * Sets up for an eventual force_next(). start_force(0) basically does
1277 * an unshift, while start_force(-1) does a push. yylex removes items
1282 S_start_force(pTHX_ int where)
1286 if (where < 0) /* so people can duplicate start_force(PL_curforce) */
1287 where = PL_lasttoke;
1288 assert(PL_curforce < 0 || PL_curforce == where);
1289 if (PL_curforce != where) {
1290 for (i = PL_lasttoke; i > where; --i) {
1291 PL_nexttoke[i] = PL_nexttoke[i-1];
1295 if (PL_curforce < 0) /* in case of duplicate start_force() */
1296 Zero(&PL_nexttoke[where], 1, NEXTTOKE);
1297 PL_curforce = where;
1300 curmad('^', newSVpvs(""));
1301 CURMAD('_', PL_nextwhite);
1306 S_curmad(pTHX_ char slot, SV *sv)
1312 if (PL_curforce < 0)
1313 where = &PL_thismad;
1315 where = &PL_nexttoke[PL_curforce].next_mad;
1321 if (UTF && is_utf8_string((U8*)SvPVX(sv), SvCUR(sv)))
1323 else if (PL_encoding) {
1324 sv_recode_to_utf8(sv, PL_encoding);
1329 /* keep a slot open for the head of the list? */
1330 if (slot != '_' && *where && (*where)->mad_key == '^') {
1331 (*where)->mad_key = slot;
1332 sv_free(MUTABLE_SV(((*where)->mad_val)));
1333 (*where)->mad_val = (void*)sv;
1336 addmad(newMADsv(slot, sv), where, 0);
1339 # define start_force(where) NOOP
1340 # define curmad(slot, sv) NOOP
1345 * When the lexer realizes it knows the next token (for instance,
1346 * it is reordering tokens for the parser) then it can call S_force_next
1347 * to know what token to return the next time the lexer is called. Caller
1348 * will need to set PL_nextval[] (or PL_nexttoke[].next_val with PERL_MAD),
1349 * and possibly PL_expect to ensure the lexer handles the token correctly.
1353 S_force_next(pTHX_ I32 type)
1358 PerlIO_printf(Perl_debug_log, "### forced token:\n");
1359 tokereport(type, &NEXTVAL_NEXTTOKE);
1363 if (PL_curforce < 0)
1364 start_force(PL_lasttoke);
1365 PL_nexttoke[PL_curforce].next_type = type;
1366 if (PL_lex_state != LEX_KNOWNEXT)
1367 PL_lex_defer = PL_lex_state;
1368 PL_lex_state = LEX_KNOWNEXT;
1369 PL_lex_expect = PL_expect;
1372 PL_nexttype[PL_nexttoke] = type;
1374 if (PL_lex_state != LEX_KNOWNEXT) {
1375 PL_lex_defer = PL_lex_state;
1376 PL_lex_expect = PL_expect;
1377 PL_lex_state = LEX_KNOWNEXT;
1383 S_newSV_maybe_utf8(pTHX_ const char *const start, STRLEN len)
1386 SV * const sv = newSVpvn_utf8(start, len,
1388 && is_utf8_string((const U8*)start, len));
1394 * When the lexer knows the next thing is a word (for instance, it has
1395 * just seen -> and it knows that the next char is a word char, then
1396 * it calls S_force_word to stick the next word into the PL_nexttoke/val
1400 * char *start : buffer position (must be within PL_linestr)
1401 * int token : PL_next* will be this type of bare word (e.g., METHOD,WORD)
1402 * int check_keyword : if true, Perl checks to make sure the word isn't
1403 * a keyword (do this if the word is a label, e.g. goto FOO)
1404 * int allow_pack : if true, : characters will also be allowed (require,
1405 * use, etc. do this)
1406 * int allow_initial_tick : used by the "sub" lexer only.
1410 S_force_word(pTHX_ register char *start, int token, int check_keyword, int allow_pack, int allow_initial_tick)
1416 PERL_ARGS_ASSERT_FORCE_WORD;
1418 start = SKIPSPACE1(start);
1420 if (isIDFIRST_lazy_if(s,UTF) ||
1421 (allow_pack && *s == ':') ||
1422 (allow_initial_tick && *s == '\'') )
1424 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, allow_pack, &len);
1425 if (check_keyword && keyword(PL_tokenbuf, len, 0))
1427 start_force(PL_curforce);
1429 curmad('X', newSVpvn(start,s-start));
1430 if (token == METHOD) {
1435 PL_expect = XOPERATOR;
1439 curmad('g', newSVpvs( "forced" ));
1440 NEXTVAL_NEXTTOKE.opval
1441 = (OP*)newSVOP(OP_CONST,0,
1442 S_newSV_maybe_utf8(aTHX_ PL_tokenbuf, len));
1443 NEXTVAL_NEXTTOKE.opval->op_private |= OPpCONST_BARE;
1451 * Called when the lexer wants $foo *foo &foo etc, but the program
1452 * text only contains the "foo" portion. The first argument is a pointer
1453 * to the "foo", and the second argument is the type symbol to prefix.
1454 * Forces the next token to be a "WORD".
1455 * Creates the symbol if it didn't already exist (via gv_fetchpv()).
1459 S_force_ident(pTHX_ register const char *s, int kind)
1463 PERL_ARGS_ASSERT_FORCE_IDENT;
1466 const STRLEN len = strlen(s);
1467 OP* const o = (OP*)newSVOP(OP_CONST, 0, newSVpvn(s, len));
1468 start_force(PL_curforce);
1469 NEXTVAL_NEXTTOKE.opval = o;
1472 o->op_private = OPpCONST_ENTERED;
1473 /* XXX see note in pp_entereval() for why we forgo typo
1474 warnings if the symbol must be introduced in an eval.
1476 gv_fetchpvn_flags(s, len,
1477 PL_in_eval ? (GV_ADDMULTI | GV_ADDINEVAL)
1479 kind == '$' ? SVt_PV :
1480 kind == '@' ? SVt_PVAV :
1481 kind == '%' ? SVt_PVHV :
1489 Perl_str_to_version(pTHX_ SV *sv)
1494 const char *start = SvPV_const(sv,len);
1495 const char * const end = start + len;
1496 const bool utf = SvUTF8(sv) ? TRUE : FALSE;
1498 PERL_ARGS_ASSERT_STR_TO_VERSION;
1500 while (start < end) {
1504 n = utf8n_to_uvchr((U8*)start, len, &skip, 0);
1509 retval += ((NV)n)/nshift;
1518 * Forces the next token to be a version number.
1519 * If the next token appears to be an invalid version number, (e.g. "v2b"),
1520 * and if "guessing" is TRUE, then no new token is created (and the caller
1521 * must use an alternative parsing method).
1525 S_force_version(pTHX_ char *s, int guessing)
1531 I32 startoff = s - SvPVX(PL_linestr);
1534 PERL_ARGS_ASSERT_FORCE_VERSION;
1542 while (isDIGIT(*d) || *d == '_' || *d == '.')
1546 start_force(PL_curforce);
1547 curmad('X', newSVpvn(s,d-s));
1550 if (*d == ';' || isSPACE(*d) || *d == '}' || !*d) {
1552 s = scan_num(s, &pl_yylval);
1553 version = pl_yylval.opval;
1554 ver = cSVOPx(version)->op_sv;
1555 if (SvPOK(ver) && !SvNIOK(ver)) {
1556 SvUPGRADE(ver, SVt_PVNV);
1557 SvNV_set(ver, str_to_version(ver));
1558 SvNOK_on(ver); /* hint that it is a version */
1561 else if (guessing) {
1564 sv_free(PL_nextwhite); /* let next token collect whitespace */
1566 s = SvPVX(PL_linestr) + startoff;
1574 if (PL_madskills && !version) {
1575 sv_free(PL_nextwhite); /* let next token collect whitespace */
1577 s = SvPVX(PL_linestr) + startoff;
1580 /* NOTE: The parser sees the package name and the VERSION swapped */
1581 start_force(PL_curforce);
1582 NEXTVAL_NEXTTOKE.opval = version;
1590 * Tokenize a quoted string passed in as an SV. It finds the next
1591 * chunk, up to end of string or a backslash. It may make a new
1592 * SV containing that chunk (if HINT_NEW_STRING is on). It also
1597 S_tokeq(pTHX_ SV *sv)
1601 register char *send;
1606 PERL_ARGS_ASSERT_TOKEQ;
1611 s = SvPV_force(sv, len);
1612 if (SvTYPE(sv) >= SVt_PVIV && SvIVX(sv) == -1)
1615 while (s < send && *s != '\\')
1620 if ( PL_hints & HINT_NEW_STRING ) {
1621 pv = newSVpvn_flags(SvPVX_const(pv), len, SVs_TEMP | SvUTF8(sv));
1625 if (s + 1 < send && (s[1] == '\\'))
1626 s++; /* all that, just for this */
1631 SvCUR_set(sv, d - SvPVX_const(sv));
1633 if ( PL_hints & HINT_NEW_STRING )
1634 return new_constant(NULL, 0, "q", sv, pv, "q", 1);
1639 * Now come three functions related to double-quote context,
1640 * S_sublex_start, S_sublex_push, and S_sublex_done. They're used when
1641 * converting things like "\u\Lgnat" into ucfirst(lc("gnat")). They
1642 * interact with PL_lex_state, and create fake ( ... ) argument lists
1643 * to handle functions and concatenation.
1644 * They assume that whoever calls them will be setting up a fake
1645 * join call, because each subthing puts a ',' after it. This lets
1648 * join($, , 'lower ', lcfirst( 'uPpEr', ) ,)
1650 * (I'm not sure whether the spurious commas at the end of lcfirst's
1651 * arguments and join's arguments are created or not).
1656 * Assumes that pl_yylval.ival is the op we're creating (e.g. OP_LCFIRST).
1658 * Pattern matching will set PL_lex_op to the pattern-matching op to
1659 * make (we return THING if pl_yylval.ival is OP_NULL, PMFUNC otherwise).
1661 * OP_CONST and OP_READLINE are easy--just make the new op and return.
1663 * Everything else becomes a FUNC.
1665 * Sets PL_lex_state to LEX_INTERPPUSH unless (ival was OP_NULL or we
1666 * had an OP_CONST or OP_READLINE). This just sets us up for a
1667 * call to S_sublex_push().
1671 S_sublex_start(pTHX)
1674 register const I32 op_type = pl_yylval.ival;
1676 if (op_type == OP_NULL) {
1677 pl_yylval.opval = PL_lex_op;
1681 if (op_type == OP_CONST || op_type == OP_READLINE) {
1682 SV *sv = tokeq(PL_lex_stuff);
1684 if (SvTYPE(sv) == SVt_PVIV) {
1685 /* Overloaded constants, nothing fancy: Convert to SVt_PV: */
1687 const char * const p = SvPV_const(sv, len);
1688 SV * const nsv = newSVpvn_flags(p, len, SvUTF8(sv));
1692 pl_yylval.opval = (OP*)newSVOP(op_type, 0, sv);
1693 PL_lex_stuff = NULL;
1694 /* Allow <FH> // "foo" */
1695 if (op_type == OP_READLINE)
1696 PL_expect = XTERMORDORDOR;
1699 else if (op_type == OP_BACKTICK && PL_lex_op) {
1700 /* readpipe() vas overriden */
1701 cSVOPx(cLISTOPx(cUNOPx(PL_lex_op)->op_first)->op_first->op_sibling)->op_sv = tokeq(PL_lex_stuff);
1702 pl_yylval.opval = PL_lex_op;
1704 PL_lex_stuff = NULL;
1708 PL_sublex_info.super_state = PL_lex_state;
1709 PL_sublex_info.sub_inwhat = (U16)op_type;
1710 PL_sublex_info.sub_op = PL_lex_op;
1711 PL_lex_state = LEX_INTERPPUSH;
1715 pl_yylval.opval = PL_lex_op;
1725 * Create a new scope to save the lexing state. The scope will be
1726 * ended in S_sublex_done. Returns a '(', starting the function arguments
1727 * to the uc, lc, etc. found before.
1728 * Sets PL_lex_state to LEX_INTERPCONCAT.
1737 PL_lex_state = PL_sublex_info.super_state;
1738 SAVEBOOL(PL_lex_dojoin);
1739 SAVEI32(PL_lex_brackets);
1740 SAVEI32(PL_lex_casemods);
1741 SAVEI32(PL_lex_starts);
1742 SAVEI8(PL_lex_state);
1743 SAVEVPTR(PL_lex_inpat);
1744 SAVEI16(PL_lex_inwhat);
1745 SAVECOPLINE(PL_curcop);
1746 SAVEPPTR(PL_bufptr);
1747 SAVEPPTR(PL_bufend);
1748 SAVEPPTR(PL_oldbufptr);
1749 SAVEPPTR(PL_oldoldbufptr);
1750 SAVEPPTR(PL_last_lop);
1751 SAVEPPTR(PL_last_uni);
1752 SAVEPPTR(PL_linestart);
1753 SAVESPTR(PL_linestr);
1754 SAVEGENERICPV(PL_lex_brackstack);
1755 SAVEGENERICPV(PL_lex_casestack);
1757 PL_linestr = PL_lex_stuff;
1758 PL_lex_stuff = NULL;
1760 PL_bufend = PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart
1761 = SvPVX(PL_linestr);
1762 PL_bufend += SvCUR(PL_linestr);
1763 PL_last_lop = PL_last_uni = NULL;
1764 SAVEFREESV(PL_linestr);
1766 PL_lex_dojoin = FALSE;
1767 PL_lex_brackets = 0;
1768 Newx(PL_lex_brackstack, 120, char);
1769 Newx(PL_lex_casestack, 12, char);
1770 PL_lex_casemods = 0;
1771 *PL_lex_casestack = '\0';
1773 PL_lex_state = LEX_INTERPCONCAT;
1774 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
1776 PL_lex_inwhat = PL_sublex_info.sub_inwhat;
1777 if (PL_lex_inwhat == OP_MATCH || PL_lex_inwhat == OP_QR || PL_lex_inwhat == OP_SUBST)
1778 PL_lex_inpat = PL_sublex_info.sub_op;
1780 PL_lex_inpat = NULL;
1787 * Restores lexer state after a S_sublex_push.
1794 if (!PL_lex_starts++) {
1795 SV * const sv = newSVpvs("");
1796 if (SvUTF8(PL_linestr))
1798 PL_expect = XOPERATOR;
1799 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
1803 if (PL_lex_casemods) { /* oops, we've got some unbalanced parens */
1804 PL_lex_state = LEX_INTERPCASEMOD;
1808 /* Is there a right-hand side to take care of? (s//RHS/ or tr//RHS/) */
1809 if (PL_lex_repl && (PL_lex_inwhat == OP_SUBST || PL_lex_inwhat == OP_TRANS)) {
1810 PL_linestr = PL_lex_repl;
1812 PL_bufend = PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart = SvPVX(PL_linestr);
1813 PL_bufend += SvCUR(PL_linestr);
1814 PL_last_lop = PL_last_uni = NULL;
1815 SAVEFREESV(PL_linestr);
1816 PL_lex_dojoin = FALSE;
1817 PL_lex_brackets = 0;
1818 PL_lex_casemods = 0;
1819 *PL_lex_casestack = '\0';
1821 if (SvEVALED(PL_lex_repl)) {
1822 PL_lex_state = LEX_INTERPNORMAL;
1824 /* we don't clear PL_lex_repl here, so that we can check later
1825 whether this is an evalled subst; that means we rely on the
1826 logic to ensure sublex_done() is called again only via the
1827 branch (in yylex()) that clears PL_lex_repl, else we'll loop */
1830 PL_lex_state = LEX_INTERPCONCAT;
1840 PL_endwhite = newSVpvs("");
1841 sv_catsv(PL_endwhite, PL_thiswhite);
1845 sv_setpvs(PL_thistoken,"");
1847 PL_realtokenstart = -1;
1851 PL_bufend = SvPVX(PL_linestr);
1852 PL_bufend += SvCUR(PL_linestr);
1853 PL_expect = XOPERATOR;
1854 PL_sublex_info.sub_inwhat = 0;
1862 Extracts a pattern, double-quoted string, or transliteration. This
1865 It looks at PL_lex_inwhat and PL_lex_inpat to find out whether it's
1866 processing a pattern (PL_lex_inpat is true), a transliteration
1867 (PL_lex_inwhat == OP_TRANS is true), or a double-quoted string.
1869 Returns a pointer to the character scanned up to. If this is
1870 advanced from the start pointer supplied (i.e. if anything was
1871 successfully parsed), will leave an OP for the substring scanned
1872 in pl_yylval. Caller must intuit reason for not parsing further
1873 by looking at the next characters herself.
1877 double-quoted style: \r and \n
1878 regexp special ones: \D \s
1881 case and quoting: \U \Q \E
1882 stops on @ and $, but not for $ as tail anchor
1884 In transliterations:
1885 characters are VERY literal, except for - not at the start or end
1886 of the string, which indicates a range. If the range is in bytes,
1887 scan_const expands the range to the full set of intermediate
1888 characters. If the range is in utf8, the hyphen is replaced with
1889 a certain range mark which will be handled by pmtrans() in op.c.
1891 In double-quoted strings:
1893 double-quoted style: \r and \n
1895 deprecated backrefs: \1 (in substitution replacements)
1896 case and quoting: \U \Q \E
1899 scan_const does *not* construct ops to handle interpolated strings.
1900 It stops processing as soon as it finds an embedded $ or @ variable
1901 and leaves it to the caller to work out what's going on.
1903 embedded arrays (whether in pattern or not) could be:
1904 @foo, @::foo, @'foo, @{foo}, @$foo, @+, @-.
1906 $ in double-quoted strings must be the symbol of an embedded scalar.
1908 $ in pattern could be $foo or could be tail anchor. Assumption:
1909 it's a tail anchor if $ is the last thing in the string, or if it's
1910 followed by one of "()| \r\n\t"
1912 \1 (backreferences) are turned into $1
1914 The structure of the code is
1915 while (there's a character to process) {
1916 handle transliteration ranges
1917 skip regexp comments /(?#comment)/ and codes /(?{code})/
1918 skip #-initiated comments in //x patterns
1919 check for embedded arrays
1920 check for embedded scalars
1922 leave intact backslashes from leaveit (below)
1923 deprecate \1 in substitution replacements
1924 handle string-changing backslashes \l \U \Q \E, etc.
1925 switch (what was escaped) {
1926 handle \- in a transliteration (becomes a literal -)
1927 handle \132 (octal characters)
1928 handle \x15 and \x{1234} (hex characters)
1929 handle \N{name} (named characters)
1930 handle \cV (control characters)
1931 handle printf-style backslashes (\f, \r, \n, etc)
1934 } (end if backslash)
1935 handle regular character
1936 } (end while character to read)
1941 S_scan_const(pTHX_ char *start)
1944 register char *send = PL_bufend; /* end of the constant */
1945 SV *sv = newSV(send - start); /* sv for the constant. See
1946 note below on sizing. */
1947 register char *s = start; /* start of the constant */
1948 register char *d = SvPVX(sv); /* destination for copies */
1949 bool dorange = FALSE; /* are we in a translit range? */
1950 bool didrange = FALSE; /* did we just finish a range? */
1951 I32 has_utf8 = FALSE; /* Output constant is UTF8 */
1952 I32 this_utf8 = UTF; /* Is the source string assumed
1953 to be UTF8? But, this can
1954 show as true when the source
1955 isn't utf8, as for example
1956 when it is entirely composed
1959 /* Note on sizing: The scanned constant is placed into sv, which is
1960 * initialized by newSV() assuming one byte of output for every byte of
1961 * input. This routine expects newSV() to allocate an extra byte for a
1962 * trailing NUL, which this routine will append if it gets to the end of
1963 * the input. There may be more bytes of input than output (eg., \N{LATIN
1964 * CAPITAL LETTER A}), or more output than input if the constant ends up
1965 * recoded to utf8, but each time a construct is found that might increase
1966 * the needed size, SvGROW() is called. Its size parameter each time is
1967 * based on the best guess estimate at the time, namely the length used so
1968 * far, plus the length the current construct will occupy, plus room for
1969 * the trailing NUL, plus one byte for every input byte still unscanned */
1973 UV literal_endpoint = 0;
1974 bool native_range = TRUE; /* turned to FALSE if the first endpoint is Unicode. */
1977 PERL_ARGS_ASSERT_SCAN_CONST;
1979 if (PL_lex_inwhat == OP_TRANS && PL_sublex_info.sub_op) {
1980 /* If we are doing a trans and we know we want UTF8 set expectation */
1981 has_utf8 = PL_sublex_info.sub_op->op_private & (OPpTRANS_FROM_UTF|OPpTRANS_TO_UTF);
1982 this_utf8 = PL_sublex_info.sub_op->op_private & (PL_lex_repl ? OPpTRANS_FROM_UTF : OPpTRANS_TO_UTF);
1986 while (s < send || dorange) {
1987 /* get transliterations out of the way (they're most literal) */
1988 if (PL_lex_inwhat == OP_TRANS) {
1989 /* expand a range A-Z to the full set of characters. AIE! */
1991 I32 i; /* current expanded character */
1992 I32 min; /* first character in range */
1993 I32 max; /* last character in range */
2004 char * const c = (char*)utf8_hop((U8*)d, -1);
2008 *c = (char)UTF_TO_NATIVE(0xff);
2009 /* mark the range as done, and continue */
2015 i = d - SvPVX_const(sv); /* remember current offset */
2018 SvLEN(sv) + (has_utf8 ?
2019 (512 - UTF_CONTINUATION_MARK +
2022 /* How many two-byte within 0..255: 128 in UTF-8,
2023 * 96 in UTF-8-mod. */
2025 SvGROW(sv, SvLEN(sv) + 256); /* never more than 256 chars in a range */
2027 d = SvPVX(sv) + i; /* refresh d after realloc */
2031 for (j = 0; j <= 1; j++) {
2032 char * const c = (char*)utf8_hop((U8*)d, -1);
2033 const UV uv = utf8n_to_uvchr((U8*)c, d - c, NULL, 0);
2039 max = (U8)0xff; /* only to \xff */
2040 uvmax = uv; /* \x{100} to uvmax */
2042 d = c; /* eat endpoint chars */
2047 d -= 2; /* eat the first char and the - */
2048 min = (U8)*d; /* first char in range */
2049 max = (U8)d[1]; /* last char in range */
2056 "Invalid range \"%c-%c\" in transliteration operator",
2057 (char)min, (char)max);
2061 if (literal_endpoint == 2 &&
2062 ((isLOWER(min) && isLOWER(max)) ||
2063 (isUPPER(min) && isUPPER(max)))) {
2065 for (i = min; i <= max; i++)
2067 *d++ = NATIVE_TO_NEED(has_utf8,i);
2069 for (i = min; i <= max; i++)
2071 *d++ = NATIVE_TO_NEED(has_utf8,i);
2076 for (i = min; i <= max; i++)
2079 const U8 ch = (U8)NATIVE_TO_UTF(i);
2080 if (UNI_IS_INVARIANT(ch))
2083 *d++ = (U8)UTF8_EIGHT_BIT_HI(ch);
2084 *d++ = (U8)UTF8_EIGHT_BIT_LO(ch);
2093 d = (char*)uvchr_to_utf8((U8*)d, 0x100);
2095 *d++ = (char)UTF_TO_NATIVE(0xff);
2097 d = (char*)uvchr_to_utf8((U8*)d, uvmax);
2101 /* mark the range as done, and continue */
2105 literal_endpoint = 0;
2110 /* range begins (ignore - as first or last char) */
2111 else if (*s == '-' && s+1 < send && s != start) {
2113 Perl_croak(aTHX_ "Ambiguous range in transliteration operator");
2120 *d++ = (char)UTF_TO_NATIVE(0xff); /* use illegal utf8 byte--see pmtrans */
2130 literal_endpoint = 0;
2131 native_range = TRUE;
2136 /* if we get here, we're not doing a transliteration */
2138 /* skip for regexp comments /(?#comment)/ and code /(?{code})/,
2139 except for the last char, which will be done separately. */
2140 else if (*s == '(' && PL_lex_inpat && s[1] == '?') {
2142 while (s+1 < send && *s != ')')
2143 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2145 else if (s[2] == '{' /* This should match regcomp.c */
2146 || (s[2] == '?' && s[3] == '{'))
2149 char *regparse = s + (s[2] == '{' ? 3 : 4);
2152 while (count && (c = *regparse)) {
2153 if (c == '\\' && regparse[1])
2161 if (*regparse != ')')
2162 regparse--; /* Leave one char for continuation. */
2163 while (s < regparse)
2164 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2168 /* likewise skip #-initiated comments in //x patterns */
2169 else if (*s == '#' && PL_lex_inpat &&
2170 ((PMOP*)PL_lex_inpat)->op_pmflags & PMf_EXTENDED) {
2171 while (s+1 < send && *s != '\n')
2172 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2175 /* check for embedded arrays
2176 (@foo, @::foo, @'foo, @{foo}, @$foo, @+, @-)
2178 else if (*s == '@' && s[1]) {
2179 if (isALNUM_lazy_if(s+1,UTF))
2181 if (strchr(":'{$", s[1]))
2183 if (!PL_lex_inpat && (s[1] == '+' || s[1] == '-'))
2184 break; /* in regexp, neither @+ nor @- are interpolated */
2187 /* check for embedded scalars. only stop if we're sure it's a
2190 else if (*s == '$') {
2191 if (!PL_lex_inpat) /* not a regexp, so $ must be var */
2193 if (s + 1 < send && !strchr("()| \r\n\t", s[1])) {
2194 if (s[1] == '\\' && ckWARN(WARN_AMBIGUOUS)) {
2195 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
2196 "Possible unintended interpolation of $\\ in regex");
2198 break; /* in regexp, $ might be tail anchor */
2202 /* End of else if chain - OP_TRANS rejoin rest */
2205 if (*s == '\\' && s+1 < send) {
2208 /* deprecate \1 in strings and substitution replacements */
2209 if (PL_lex_inwhat == OP_SUBST && !PL_lex_inpat &&
2210 isDIGIT(*s) && *s != '0' && !isDIGIT(s[1]))
2212 if (ckWARN(WARN_SYNTAX))
2213 Perl_warner(aTHX_ packWARN(WARN_SYNTAX), "\\%c better written as $%c", *s, *s);
2218 /* string-change backslash escapes */
2219 if (PL_lex_inwhat != OP_TRANS && *s && strchr("lLuUEQ", *s)) {
2223 /* skip any other backslash escapes in a pattern */
2224 else if (PL_lex_inpat) {
2225 *d++ = NATIVE_TO_NEED(has_utf8,'\\');
2226 goto default_action;
2229 /* if we get here, it's either a quoted -, or a digit */
2232 /* quoted - in transliterations */
2234 if (PL_lex_inwhat == OP_TRANS) {
2241 if ((isALPHA(*s) || isDIGIT(*s)) &&
2243 Perl_warner(aTHX_ packWARN(WARN_MISC),
2244 "Unrecognized escape \\%c passed through",
2246 /* default action is to copy the quoted character */
2247 goto default_action;
2250 /* eg. \132 indicates the octal constant 0x132 */
2251 case '0': case '1': case '2': case '3':
2252 case '4': case '5': case '6': case '7':
2256 uv = NATIVE_TO_UNI(grok_oct(s, &len, &flags, NULL));
2259 goto NUM_ESCAPE_INSERT;
2261 /* eg. \x24 indicates the hex constant 0x24 */
2265 char* const e = strchr(s, '}');
2266 I32 flags = PERL_SCAN_ALLOW_UNDERSCORES |
2267 PERL_SCAN_DISALLOW_PREFIX;
2272 yyerror("Missing right brace on \\x{}");
2276 uv = NATIVE_TO_UNI(grok_hex(s, &len, &flags, NULL));
2282 I32 flags = PERL_SCAN_DISALLOW_PREFIX;
2283 uv = NATIVE_TO_UNI(grok_hex(s, &len, &flags, NULL));
2289 /* Insert oct, hex, or \N{U+...} escaped character. There will
2290 * always be enough room in sv since such escapes will be
2291 * longer than any UTF-8 sequence they can end up as, except if
2292 * they force us to recode the rest of the string into utf8 */
2294 /* Here uv is the ordinal of the next character being added in
2295 * unicode (converted from native). (It has to be done before
2296 * here because \N is interpreted as unicode, and oct and hex
2298 if (!UNI_IS_INVARIANT(uv)) {
2299 if (!has_utf8 && uv > 255) {
2300 /* Might need to recode whatever we have accumulated so
2301 * far if it contains any chars variant in utf8 or
2304 SvCUR_set(sv, d - SvPVX_const(sv));
2307 /* See Note on sizing above. */
2308 sv_utf8_upgrade_flags_grow(sv,
2309 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
2310 UNISKIP(uv) + (STRLEN)(send - s) + 1);
2311 d = SvPVX(sv) + SvCUR(sv);
2316 d = (char*)uvuni_to_utf8((U8*)d, uv);
2317 if (PL_lex_inwhat == OP_TRANS &&
2318 PL_sublex_info.sub_op) {
2319 PL_sublex_info.sub_op->op_private |=
2320 (PL_lex_repl ? OPpTRANS_FROM_UTF
2324 if (uv > 255 && !dorange)
2325 native_range = FALSE;
2337 /* \N{LATIN SMALL LETTER A} is a named character, and so is
2342 char* e = strchr(s, '}');
2348 yyerror("Missing right brace on \\N{}");
2352 if (e > s + 2 && s[1] == 'U' && s[2] == '+') {
2353 /* \N{U+...} The ... is a unicode value even on EBCDIC
2355 I32 flags = PERL_SCAN_ALLOW_UNDERSCORES |
2356 PERL_SCAN_DISALLOW_PREFIX;
2359 uv = grok_hex(s, &len, &flags, NULL);
2360 if ( e > s && len != (STRLEN)(e - s) ) {
2364 goto NUM_ESCAPE_INSERT;
2366 res = newSVpvn(s + 1, e - s - 1);
2367 res = new_constant( NULL, 0, "charnames",
2368 res, NULL, s - 2, e - s + 3 );
2370 sv_utf8_upgrade(res);
2371 str = SvPV_const(res,len);
2372 #ifdef EBCDIC_NEVER_MIND
2373 /* charnames uses pack U and that has been
2374 * recently changed to do the below uni->native
2375 * mapping, so this would be redundant (and wrong,
2376 * the code point would be doubly converted).
2377 * But leave this in just in case the pack U change
2378 * gets revoked, but the semantics is still
2379 * desireable for charnames. --jhi */
2381 UV uv = utf8_to_uvchr((const U8*)str, 0);
2384 U8 tmpbuf[UTF8_MAXBYTES+1], *d;
2386 d = uvchr_to_utf8(tmpbuf, UNI_TO_NATIVE(uv));
2387 sv_setpvn(res, (char *)tmpbuf, d - tmpbuf);
2388 str = SvPV_const(res, len);
2392 /* If destination is not in utf8 but this new character is,
2393 * recode the dest to utf8 */
2394 if (!has_utf8 && SvUTF8(res)) {
2395 SvCUR_set(sv, d - SvPVX_const(sv));
2398 /* See Note on sizing above. */
2399 sv_utf8_upgrade_flags_grow(sv,
2400 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
2401 len + (STRLEN)(send - s) + 1);
2402 d = SvPVX(sv) + SvCUR(sv);
2404 } else if (len > (STRLEN)(e - s + 4)) { /* I _guess_ 4 is \N{} --jhi */
2406 /* See Note on sizing above. (NOTE: SvCUR() is not set
2407 * correctly here). */
2408 const STRLEN off = d - SvPVX_const(sv);
2409 d = SvGROW(sv, off + len + (STRLEN)(send - s) + 1) + off;
2413 native_range = FALSE; /* \N{} is guessed to be Unicode */
2415 Copy(str, d, len, char);
2422 yyerror("Missing braces on \\N{}");
2425 /* \c is a control character */
2434 *d++ = NATIVE_TO_NEED(has_utf8,toCTRL(c));
2437 yyerror("Missing control char name in \\c");
2441 /* printf-style backslashes, formfeeds, newlines, etc */
2443 *d++ = NATIVE_TO_NEED(has_utf8,'\b');
2446 *d++ = NATIVE_TO_NEED(has_utf8,'\n');
2449 *d++ = NATIVE_TO_NEED(has_utf8,'\r');
2452 *d++ = NATIVE_TO_NEED(has_utf8,'\f');
2455 *d++ = NATIVE_TO_NEED(has_utf8,'\t');
2458 *d++ = ASCII_TO_NEED(has_utf8,'\033');
2461 *d++ = ASCII_TO_NEED(has_utf8,'\007');
2467 } /* end if (backslash) */
2474 /* If we started with encoded form, or already know we want it,
2475 then encode the next character */
2476 if (! NATIVE_IS_INVARIANT((U8)(*s)) && (this_utf8 || has_utf8)) {
2480 /* One might think that it is wasted effort in the case of the
2481 * source being utf8 (this_utf8 == TRUE) to take the next character
2482 * in the source, convert it to an unsigned value, and then convert
2483 * it back again. But the source has not been validated here. The
2484 * routine that does the conversion checks for errors like
2487 const UV nextuv = (this_utf8) ? utf8n_to_uvchr((U8*)s, send - s, &len, 0) : (UV) ((U8) *s);
2488 const STRLEN need = UNISKIP(NATIVE_TO_UNI(nextuv));
2490 SvCUR_set(sv, d - SvPVX_const(sv));
2493 /* See Note on sizing above. */
2494 sv_utf8_upgrade_flags_grow(sv,
2495 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
2496 need + (STRLEN)(send - s) + 1);
2497 d = SvPVX(sv) + SvCUR(sv);
2499 } else if (need > len) {
2500 /* encoded value larger than old, may need extra space (NOTE:
2501 * SvCUR() is not set correctly here). See Note on sizing
2503 const STRLEN off = d - SvPVX_const(sv);
2504 d = SvGROW(sv, off + need + (STRLEN)(send - s) + 1) + off;
2508 d = (char*)uvchr_to_utf8((U8*)d, nextuv);
2510 if (uv > 255 && !dorange)
2511 native_range = FALSE;
2515 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2517 } /* while loop to process each character */
2519 /* terminate the string and set up the sv */
2521 SvCUR_set(sv, d - SvPVX_const(sv));
2522 if (SvCUR(sv) >= SvLEN(sv))
2523 Perl_croak(aTHX_ "panic: constant overflowed allocated space");
2526 if (PL_encoding && !has_utf8) {
2527 sv_recode_to_utf8(sv, PL_encoding);
2533 if (PL_lex_inwhat == OP_TRANS && PL_sublex_info.sub_op) {
2534 PL_sublex_info.sub_op->op_private |=
2535 (PL_lex_repl ? OPpTRANS_FROM_UTF : OPpTRANS_TO_UTF);
2539 /* shrink the sv if we allocated more than we used */
2540 if (SvCUR(sv) + 5 < SvLEN(sv)) {
2541 SvPV_shrink_to_cur(sv);
2544 /* return the substring (via pl_yylval) only if we parsed anything */
2545 if (s > PL_bufptr) {
2546 if ( PL_hints & ( PL_lex_inpat ? HINT_NEW_RE : HINT_NEW_STRING ) ) {
2547 const char *const key = PL_lex_inpat ? "qr" : "q";
2548 const STRLEN keylen = PL_lex_inpat ? 2 : 1;
2552 if (PL_lex_inwhat == OP_TRANS) {
2555 } else if (PL_lex_inwhat == OP_SUBST && !PL_lex_inpat) {
2563 sv = S_new_constant(aTHX_ start, s - start, key, keylen, sv, NULL,
2566 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
2573 * Returns TRUE if there's more to the expression (e.g., a subscript),
2576 * It deals with "$foo[3]" and /$foo[3]/ and /$foo[0123456789$]+/
2578 * ->[ and ->{ return TRUE
2579 * { and [ outside a pattern are always subscripts, so return TRUE
2580 * if we're outside a pattern and it's not { or [, then return FALSE
2581 * if we're in a pattern and the first char is a {
2582 * {4,5} (any digits around the comma) returns FALSE
2583 * if we're in a pattern and the first char is a [
2585 * [SOMETHING] has a funky algorithm to decide whether it's a
2586 * character class or not. It has to deal with things like
2587 * /$foo[-3]/ and /$foo[$bar]/ as well as /$foo[$\d]+/
2588 * anything else returns TRUE
2591 /* This is the one truly awful dwimmer necessary to conflate C and sed. */
2594 S_intuit_more(pTHX_ register char *s)
2598 PERL_ARGS_ASSERT_INTUIT_MORE;
2600 if (PL_lex_brackets)
2602 if (*s == '-' && s[1] == '>' && (s[2] == '[' || s[2] == '{'))
2604 if (*s != '{' && *s != '[')
2609 /* In a pattern, so maybe we have {n,m}. */
2626 /* On the other hand, maybe we have a character class */
2629 if (*s == ']' || *s == '^')
2632 /* this is terrifying, and it works */
2633 int weight = 2; /* let's weigh the evidence */
2635 unsigned char un_char = 255, last_un_char;
2636 const char * const send = strchr(s,']');
2637 char tmpbuf[sizeof PL_tokenbuf * 4];
2639 if (!send) /* has to be an expression */
2642 Zero(seen,256,char);
2645 else if (isDIGIT(*s)) {
2647 if (isDIGIT(s[1]) && s[2] == ']')
2653 for (; s < send; s++) {
2654 last_un_char = un_char;
2655 un_char = (unsigned char)*s;
2660 weight -= seen[un_char] * 10;
2661 if (isALNUM_lazy_if(s+1,UTF)) {
2663 scan_ident(s, send, tmpbuf, sizeof tmpbuf, FALSE);
2664 len = (int)strlen(tmpbuf);
2665 if (len > 1 && gv_fetchpvn_flags(tmpbuf, len, 0, SVt_PV))
2670 else if (*s == '$' && s[1] &&
2671 strchr("[#!%*<>()-=",s[1])) {
2672 if (/*{*/ strchr("])} =",s[2]))
2681 if (strchr("wds]",s[1]))
2683 else if (seen[(U8)'\''] || seen[(U8)'"'])
2685 else if (strchr("rnftbxcav",s[1]))
2687 else if (isDIGIT(s[1])) {
2689 while (s[1] && isDIGIT(s[1]))
2699 if (strchr("aA01! ",last_un_char))
2701 if (strchr("zZ79~",s[1]))
2703 if (last_un_char == 255 && (isDIGIT(s[1]) || s[1] == '$'))
2704 weight -= 5; /* cope with negative subscript */
2707 if (!isALNUM(last_un_char)
2708 && !(last_un_char == '$' || last_un_char == '@'
2709 || last_un_char == '&')
2710 && isALPHA(*s) && s[1] && isALPHA(s[1])) {
2715 if (keyword(tmpbuf, d - tmpbuf, 0))
2718 if (un_char == last_un_char + 1)
2720 weight -= seen[un_char];
2725 if (weight >= 0) /* probably a character class */
2735 * Does all the checking to disambiguate
2737 * between foo(bar) and bar->foo. Returns 0 if not a method, otherwise
2738 * FUNCMETH (bar->foo(args)) or METHOD (bar->foo args).
2740 * First argument is the stuff after the first token, e.g. "bar".
2742 * Not a method if bar is a filehandle.
2743 * Not a method if foo is a subroutine prototyped to take a filehandle.
2744 * Not a method if it's really "Foo $bar"
2745 * Method if it's "foo $bar"
2746 * Not a method if it's really "print foo $bar"
2747 * Method if it's really "foo package::" (interpreted as package->foo)
2748 * Not a method if bar is known to be a subroutine ("sub bar; foo bar")
2749 * Not a method if bar is a filehandle or package, but is quoted with
2754 S_intuit_method(pTHX_ char *start, GV *gv, CV *cv)
2757 char *s = start + (*start == '$');
2758 char tmpbuf[sizeof PL_tokenbuf];
2765 PERL_ARGS_ASSERT_INTUIT_METHOD;
2768 if (SvTYPE(gv) == SVt_PVGV && GvIO(gv))
2772 const char *proto = SvPVX_const(cv);
2783 s = scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
2784 /* start is the beginning of the possible filehandle/object,
2785 * and s is the end of it
2786 * tmpbuf is a copy of it
2789 if (*start == '$') {
2790 if (gv || PL_last_lop_op == OP_PRINT || PL_last_lop_op == OP_SAY ||
2791 isUPPER(*PL_tokenbuf))
2794 len = start - SvPVX(PL_linestr);
2798 start = SvPVX(PL_linestr) + len;
2802 return *s == '(' ? FUNCMETH : METHOD;
2804 if (!keyword(tmpbuf, len, 0)) {
2805 if (len > 2 && tmpbuf[len - 2] == ':' && tmpbuf[len - 1] == ':') {
2809 soff = s - SvPVX(PL_linestr);
2813 indirgv = gv_fetchpvn_flags(tmpbuf, len, 0, SVt_PVCV);
2814 if (indirgv && GvCVu(indirgv))
2816 /* filehandle or package name makes it a method */
2817 if (!gv || GvIO(indirgv) || gv_stashpvn(tmpbuf, len, 0)) {
2819 soff = s - SvPVX(PL_linestr);
2822 if ((PL_bufend - s) >= 2 && *s == '=' && *(s+1) == '>')
2823 return 0; /* no assumptions -- "=>" quotes bearword */
2825 start_force(PL_curforce);
2826 NEXTVAL_NEXTTOKE.opval = (OP*)newSVOP(OP_CONST, 0,
2827 newSVpvn(tmpbuf,len));
2828 NEXTVAL_NEXTTOKE.opval->op_private = OPpCONST_BARE;
2830 curmad('X', newSVpvn(start,SvPVX(PL_linestr) + soff - start));
2835 PL_bufptr = SvPVX(PL_linestr) + soff; /* restart before space */
2837 return *s == '(' ? FUNCMETH : METHOD;
2843 /* Encoded script support. filter_add() effectively inserts a
2844 * 'pre-processing' function into the current source input stream.
2845 * Note that the filter function only applies to the current source file
2846 * (e.g., it will not affect files 'require'd or 'use'd by this one).
2848 * The datasv parameter (which may be NULL) can be used to pass
2849 * private data to this instance of the filter. The filter function
2850 * can recover the SV using the FILTER_DATA macro and use it to
2851 * store private buffers and state information.
2853 * The supplied datasv parameter is upgraded to a PVIO type
2854 * and the IoDIRP/IoANY field is used to store the function pointer,
2855 * and IOf_FAKE_DIRP is enabled on datasv to mark this as such.
2856 * Note that IoTOP_NAME, IoFMT_NAME, IoBOTTOM_NAME, if set for
2857 * private use must be set using malloc'd pointers.
2861 Perl_filter_add(pTHX_ filter_t funcp, SV *datasv)
2870 if (!PL_rsfp_filters)
2871 PL_rsfp_filters = newAV();
2874 SvUPGRADE(datasv, SVt_PVIO);
2875 IoANY(datasv) = FPTR2DPTR(void *, funcp); /* stash funcp into spare field */
2876 IoFLAGS(datasv) |= IOf_FAKE_DIRP;
2877 DEBUG_P(PerlIO_printf(Perl_debug_log, "filter_add func %p (%s)\n",
2878 FPTR2DPTR(void *, IoANY(datasv)),
2879 SvPV_nolen(datasv)));
2880 av_unshift(PL_rsfp_filters, 1);
2881 av_store(PL_rsfp_filters, 0, datasv) ;
2886 /* Delete most recently added instance of this filter function. */
2888 Perl_filter_del(pTHX_ filter_t funcp)
2893 PERL_ARGS_ASSERT_FILTER_DEL;
2896 DEBUG_P(PerlIO_printf(Perl_debug_log, "filter_del func %p",
2897 FPTR2DPTR(void*, funcp)));
2899 if (!PL_parser || !PL_rsfp_filters || AvFILLp(PL_rsfp_filters)<0)
2901 /* if filter is on top of stack (usual case) just pop it off */
2902 datasv = FILTER_DATA(AvFILLp(PL_rsfp_filters));
2903 if (IoANY(datasv) == FPTR2DPTR(void *, funcp)) {
2904 IoFLAGS(datasv) &= ~IOf_FAKE_DIRP;
2905 IoANY(datasv) = (void *)NULL;
2906 sv_free(av_pop(PL_rsfp_filters));
2910 /* we need to search for the correct entry and clear it */
2911 Perl_die(aTHX_ "filter_del can only delete in reverse order (currently)");
2915 /* Invoke the idxth filter function for the current rsfp. */
2916 /* maxlen 0 = read one text line */
2918 Perl_filter_read(pTHX_ int idx, SV *buf_sv, int maxlen)
2923 /* This API is bad. It should have been using unsigned int for maxlen.
2924 Not sure if we want to change the API, but if not we should sanity
2925 check the value here. */
2926 const unsigned int correct_length
2935 PERL_ARGS_ASSERT_FILTER_READ;
2937 if (!PL_parser || !PL_rsfp_filters)
2939 if (idx > AvFILLp(PL_rsfp_filters)) { /* Any more filters? */
2940 /* Provide a default input filter to make life easy. */
2941 /* Note that we append to the line. This is handy. */
2942 DEBUG_P(PerlIO_printf(Perl_debug_log,
2943 "filter_read %d: from rsfp\n", idx));
2944 if (correct_length) {
2947 const int old_len = SvCUR(buf_sv);
2949 /* ensure buf_sv is large enough */
2950 SvGROW(buf_sv, (STRLEN)(old_len + correct_length)) ;
2951 if ((len = PerlIO_read(PL_rsfp, SvPVX(buf_sv) + old_len,
2952 correct_length)) <= 0) {
2953 if (PerlIO_error(PL_rsfp))
2954 return -1; /* error */
2956 return 0 ; /* end of file */
2958 SvCUR_set(buf_sv, old_len + len) ;
2961 if (sv_gets(buf_sv, PL_rsfp, SvCUR(buf_sv)) == NULL) {
2962 if (PerlIO_error(PL_rsfp))
2963 return -1; /* error */
2965 return 0 ; /* end of file */
2968 return SvCUR(buf_sv);
2970 /* Skip this filter slot if filter has been deleted */
2971 if ( (datasv = FILTER_DATA(idx)) == &PL_sv_undef) {
2972 DEBUG_P(PerlIO_printf(Perl_debug_log,
2973 "filter_read %d: skipped (filter deleted)\n",
2975 return FILTER_READ(idx+1, buf_sv, correct_length); /* recurse */
2977 /* Get function pointer hidden within datasv */
2978 funcp = DPTR2FPTR(filter_t, IoANY(datasv));
2979 DEBUG_P(PerlIO_printf(Perl_debug_log,
2980 "filter_read %d: via function %p (%s)\n",
2981 idx, (void*)datasv, SvPV_nolen_const(datasv)));
2982 /* Call function. The function is expected to */
2983 /* call "FILTER_READ(idx+1, buf_sv)" first. */
2984 /* Return: <0:error, =0:eof, >0:not eof */
2985 return (*funcp)(aTHX_ idx, buf_sv, correct_length);
2989 S_filter_gets(pTHX_ register SV *sv, register PerlIO *fp, STRLEN append)
2993 PERL_ARGS_ASSERT_FILTER_GETS;
2995 #ifdef PERL_CR_FILTER
2996 if (!PL_rsfp_filters) {
2997 filter_add(S_cr_textfilter,NULL);
3000 if (PL_rsfp_filters) {
3002 SvCUR_set(sv, 0); /* start with empty line */
3003 if (FILTER_READ(0, sv, 0) > 0)
3004 return ( SvPVX(sv) ) ;
3009 return (sv_gets(sv, fp, append));
3013 S_find_in_my_stash(pTHX_ const char *pkgname, STRLEN len)
3018 PERL_ARGS_ASSERT_FIND_IN_MY_STASH;
3020 if (len == 11 && *pkgname == '_' && strEQ(pkgname, "__PACKAGE__"))
3024 (pkgname[len - 2] == ':' && pkgname[len - 1] == ':') &&
3025 (gv = gv_fetchpvn_flags(pkgname, len, 0, SVt_PVHV)))
3027 return GvHV(gv); /* Foo:: */
3030 /* use constant CLASS => 'MyClass' */
3031 gv = gv_fetchpvn_flags(pkgname, len, 0, SVt_PVCV);
3032 if (gv && GvCV(gv)) {
3033 SV * const sv = cv_const_sv(GvCV(gv));
3035 pkgname = SvPV_const(sv, len);
3038 return gv_stashpvn(pkgname, len, 0);
3042 * S_readpipe_override
3043 * Check whether readpipe() is overriden, and generates the appropriate
3044 * optree, provided sublex_start() is called afterwards.
3047 S_readpipe_override(pTHX)
3050 GV *gv_readpipe = gv_fetchpvs("readpipe", GV_NOTQUAL, SVt_PVCV);
3051 pl_yylval.ival = OP_BACKTICK;
3053 && GvCVu(gv_readpipe) && GvIMPORTED_CV(gv_readpipe))
3055 ((gvp = (GV**)hv_fetchs(PL_globalstash, "readpipe", FALSE))
3056 && (gv_readpipe = *gvp) && isGV_with_GP(gv_readpipe)
3057 && GvCVu(gv_readpipe) && GvIMPORTED_CV(gv_readpipe)))
3059 PL_lex_op = (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
3060 append_elem(OP_LIST,
3061 newSVOP(OP_CONST, 0, &PL_sv_undef), /* value will be read later */
3062 newCVREF(0, newGVOP(OP_GV, 0, gv_readpipe))));
3069 * The intent of this yylex wrapper is to minimize the changes to the
3070 * tokener when we aren't interested in collecting madprops. It remains
3071 * to be seen how successful this strategy will be...
3078 char *s = PL_bufptr;
3080 /* make sure PL_thiswhite is initialized */
3084 /* just do what yylex would do on pending identifier; leave PL_thiswhite alone */
3085 if (PL_pending_ident)
3086 return S_pending_ident(aTHX);
3088 /* previous token ate up our whitespace? */
3089 if (!PL_lasttoke && PL_nextwhite) {
3090 PL_thiswhite = PL_nextwhite;
3094 /* isolate the token, and figure out where it is without whitespace */
3095 PL_realtokenstart = -1;
3099 assert(PL_curforce < 0);
3101 if (!PL_thismad || PL_thismad->mad_key == '^') { /* not forced already? */
3102 if (!PL_thistoken) {
3103 if (PL_realtokenstart < 0 || !CopLINE(PL_curcop))
3104 PL_thistoken = newSVpvs("");
3106 char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
3107 PL_thistoken = newSVpvn(tstart, s - tstart);
3110 if (PL_thismad) /* install head */
3111 CURMAD('X', PL_thistoken);
3114 /* last whitespace of a sublex? */
3115 if (optype == ')' && PL_endwhite) {
3116 CURMAD('X', PL_endwhite);
3121 /* if no whitespace and we're at EOF, bail. Otherwise fake EOF below. */
3122 if (!PL_thiswhite && !PL_endwhite && !optype) {
3123 sv_free(PL_thistoken);
3128 /* put off final whitespace till peg */
3129 if (optype == ';' && !PL_rsfp) {
3130 PL_nextwhite = PL_thiswhite;
3133 else if (PL_thisopen) {
3134 CURMAD('q', PL_thisopen);
3136 sv_free(PL_thistoken);
3140 /* Store actual token text as madprop X */
3141 CURMAD('X', PL_thistoken);
3145 /* add preceding whitespace as madprop _ */
3146 CURMAD('_', PL_thiswhite);
3150 /* add quoted material as madprop = */
3151 CURMAD('=', PL_thisstuff);
3155 /* add terminating quote as madprop Q */
3156 CURMAD('Q', PL_thisclose);
3160 /* special processing based on optype */
3164 /* opval doesn't need a TOKEN since it can already store mp */
3174 if (pl_yylval.opval)
3175 append_madprops(PL_thismad, pl_yylval.opval, 0);
3183 addmad(newMADsv('p', PL_endwhite), &PL_thismad, 0);
3192 /* remember any fake bracket that lexer is about to discard */
3193 if (PL_lex_brackets == 1 &&
3194 ((expectation)PL_lex_brackstack[0] & XFAKEBRACK))
3197 while (s < PL_bufend && (*s == ' ' || *s == '\t'))
3200 PL_thiswhite = newSVpvn(PL_bufptr, ++s - PL_bufptr);
3201 addmad(newMADsv('#', PL_thiswhite), &PL_thismad, 0);
3204 break; /* don't bother looking for trailing comment */
3213 /* attach a trailing comment to its statement instead of next token */
3217 if (PL_bufptr > PL_oldbufptr && PL_bufptr[-1] == optype) {
3219 while (s < PL_bufend && (*s == ' ' || *s == '\t'))
3221 if (*s == '\n' || *s == '#') {
3222 while (s < PL_bufend && *s != '\n')
3226 PL_thiswhite = newSVpvn(PL_bufptr, s - PL_bufptr);
3227 addmad(newMADsv('#', PL_thiswhite), &PL_thismad, 0);
3244 /* Create new token struct. Note: opvals return early above. */
3245 pl_yylval.tkval = newTOKEN(optype, pl_yylval, PL_thismad);
3252 S_tokenize_use(pTHX_ int is_use, char *s) {
3255 PERL_ARGS_ASSERT_TOKENIZE_USE;
3257 if (PL_expect != XSTATE)
3258 yyerror(Perl_form(aTHX_ "\"%s\" not allowed in expression",
3259 is_use ? "use" : "no"));
3261 if (isDIGIT(*s) || (*s == 'v' && isDIGIT(s[1]))) {
3262 s = force_version(s, TRUE);
3263 if (*s == ';' || (s = SKIPSPACE1(s), *s == ';')) {
3264 start_force(PL_curforce);
3265 NEXTVAL_NEXTTOKE.opval = NULL;
3268 else if (*s == 'v') {
3269 s = force_word(s,WORD,FALSE,TRUE,FALSE);
3270 s = force_version(s, FALSE);
3274 s = force_word(s,WORD,FALSE,TRUE,FALSE);
3275 s = force_version(s, FALSE);
3277 pl_yylval.ival = is_use;
3281 static const char* const exp_name[] =
3282 { "OPERATOR", "TERM", "REF", "STATE", "BLOCK", "ATTRBLOCK",
3283 "ATTRTERM", "TERMBLOCK", "TERMORDORDOR"
3290 Works out what to call the token just pulled out of the input
3291 stream. The yacc parser takes care of taking the ops we return and
3292 stitching them into a tree.
3298 if read an identifier
3299 if we're in a my declaration
3300 croak if they tried to say my($foo::bar)
3301 build the ops for a my() declaration
3302 if it's an access to a my() variable
3303 are we in a sort block?
3304 croak if my($a); $a <=> $b
3305 build ops for access to a my() variable
3306 if in a dq string, and they've said @foo and we can't find @foo
3308 build ops for a bareword
3309 if we already built the token before, use it.
3314 #pragma segment Perl_yylex
3320 register char *s = PL_bufptr;
3325 /* orig_keyword, gvp, and gv are initialized here because
3326 * jump to the label just_a_word_zero can bypass their
3327 * initialization later. */
3328 I32 orig_keyword = 0;
3333 SV* tmp = newSVpvs("");
3334 PerlIO_printf(Perl_debug_log, "### %"IVdf":LEX_%s/X%s %s\n",
3335 (IV)CopLINE(PL_curcop),
3336 lex_state_names[PL_lex_state],
3337 exp_name[PL_expect],
3338 pv_display(tmp, s, strlen(s), 0, 60));
3341 /* check if there's an identifier for us to look at */
3342 if (PL_pending_ident)
3343 return REPORT(S_pending_ident(aTHX));
3345 /* no identifier pending identification */
3347 switch (PL_lex_state) {
3349 case LEX_NORMAL: /* Some compilers will produce faster */
3350 case LEX_INTERPNORMAL: /* code if we comment these out. */
3354 /* when we've already built the next token, just pull it out of the queue */
3358 pl_yylval = PL_nexttoke[PL_lasttoke].next_val;
3360 PL_thismad = PL_nexttoke[PL_lasttoke].next_mad;
3361 PL_nexttoke[PL_lasttoke].next_mad = 0;
3362 if (PL_thismad && PL_thismad->mad_key == '_') {
3363 PL_thiswhite = MUTABLE_SV(PL_thismad->mad_val);
3364 PL_thismad->mad_val = 0;
3365 mad_free(PL_thismad);
3370 PL_lex_state = PL_lex_defer;
3371 PL_expect = PL_lex_expect;
3372 PL_lex_defer = LEX_NORMAL;
3373 if (!PL_nexttoke[PL_lasttoke].next_type)
3378 pl_yylval = PL_nextval[PL_nexttoke];
3380 PL_lex_state = PL_lex_defer;
3381 PL_expect = PL_lex_expect;
3382 PL_lex_defer = LEX_NORMAL;
3386 /* FIXME - can these be merged? */
3387 return(PL_nexttoke[PL_lasttoke].next_type);
3389 return REPORT(PL_nexttype[PL_nexttoke]);
3392 /* interpolated case modifiers like \L \U, including \Q and \E.
3393 when we get here, PL_bufptr is at the \
3395 case LEX_INTERPCASEMOD:
3397 if (PL_bufptr != PL_bufend && *PL_bufptr != '\\')
3398 Perl_croak(aTHX_ "panic: INTERPCASEMOD");
3400 /* handle \E or end of string */
3401 if (PL_bufptr == PL_bufend || PL_bufptr[1] == 'E') {
3403 if (PL_lex_casemods) {
3404 const char oldmod = PL_lex_casestack[--PL_lex_casemods];
3405 PL_lex_casestack[PL_lex_casemods] = '\0';
3407 if (PL_bufptr != PL_bufend
3408 && (oldmod == 'L' || oldmod == 'U' || oldmod == 'Q')) {
3410 PL_lex_state = LEX_INTERPCONCAT;
3413 PL_thistoken = newSVpvs("\\E");
3419 while (PL_bufptr != PL_bufend &&
3420 PL_bufptr[0] == '\\' && PL_bufptr[1] == 'E') {
3422 PL_thiswhite = newSVpvs("");
3423 sv_catpvn(PL_thiswhite, PL_bufptr, 2);
3427 if (PL_bufptr != PL_bufend)
3430 PL_lex_state = LEX_INTERPCONCAT;
3434 DEBUG_T({ PerlIO_printf(Perl_debug_log,
3435 "### Saw case modifier\n"); });
3437 if (s[1] == '\\' && s[2] == 'E') {
3440 PL_thiswhite = newSVpvs("");
3441 sv_catpvn(PL_thiswhite, PL_bufptr, 4);
3444 PL_lex_state = LEX_INTERPCONCAT;
3449 if (!PL_madskills) /* when just compiling don't need correct */
3450 if (strnEQ(s, "L\\u", 3) || strnEQ(s, "U\\l", 3))
3451 tmp = *s, *s = s[2], s[2] = (char)tmp; /* misordered... */
3452 if ((*s == 'L' || *s == 'U') &&
3453 (strchr(PL_lex_casestack, 'L') || strchr(PL_lex_casestack, 'U'))) {
3454 PL_lex_casestack[--PL_lex_casemods] = '\0';
3457 if (PL_lex_casemods > 10)
3458 Renew(PL_lex_casestack, PL_lex_casemods + 2, char);
3459 PL_lex_casestack[PL_lex_casemods++] = *s;
3460 PL_lex_casestack[PL_lex_casemods] = '\0';
3461 PL_lex_state = LEX_INTERPCONCAT;
3462 start_force(PL_curforce);
3463 NEXTVAL_NEXTTOKE.ival = 0;
3465 start_force(PL_curforce);
3467 NEXTVAL_NEXTTOKE.ival = OP_LCFIRST;
3469 NEXTVAL_NEXTTOKE.ival = OP_UCFIRST;
3471 NEXTVAL_NEXTTOKE.ival = OP_LC;
3473 NEXTVAL_NEXTTOKE.ival = OP_UC;
3475 NEXTVAL_NEXTTOKE.ival = OP_QUOTEMETA;
3477 Perl_croak(aTHX_ "panic: yylex");
3479 SV* const tmpsv = newSVpvs("\\ ");
3480 /* replace the space with the character we want to escape
3482 SvPVX(tmpsv)[1] = *s;
3488 if (PL_lex_starts) {
3494 sv_free(PL_thistoken);
3495 PL_thistoken = newSVpvs("");
3498 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
3499 if (PL_lex_casemods == 1 && PL_lex_inpat)
3508 case LEX_INTERPPUSH:
3509 return REPORT(sublex_push());
3511 case LEX_INTERPSTART:
3512 if (PL_bufptr == PL_bufend)
3513 return REPORT(sublex_done());
3514 DEBUG_T({ PerlIO_printf(Perl_debug_log,
3515 "### Interpolated variable\n"); });
3517 PL_lex_dojoin = (*PL_bufptr == '@');
3518 PL_lex_state = LEX_INTERPNORMAL;
3519 if (PL_lex_dojoin) {
3520 start_force(PL_curforce);
3521 NEXTVAL_NEXTTOKE.ival = 0;
3523 start_force(PL_curforce);
3524 force_ident("\"", '$');
3525 start_force(PL_curforce);
3526 NEXTVAL_NEXTTOKE.ival = 0;
3528 start_force(PL_curforce);
3529 NEXTVAL_NEXTTOKE.ival = 0;
3531 start_force(PL_curforce);
3532 NEXTVAL_NEXTTOKE.ival = OP_JOIN; /* emulate join($", ...) */
3535 if (PL_lex_starts++) {
3540 sv_free(PL_thistoken);
3541 PL_thistoken = newSVpvs("");
3544 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
3545 if (!PL_lex_casemods && PL_lex_inpat)
3552 case LEX_INTERPENDMAYBE:
3553 if (intuit_more(PL_bufptr)) {
3554 PL_lex_state = LEX_INTERPNORMAL; /* false alarm, more expr */
3560 if (PL_lex_dojoin) {
3561 PL_lex_dojoin = FALSE;
3562 PL_lex_state = LEX_INTERPCONCAT;
3566 sv_free(PL_thistoken);
3567 PL_thistoken = newSVpvs("");
3572 if (PL_lex_inwhat == OP_SUBST && PL_linestr == PL_lex_repl
3573 && SvEVALED(PL_lex_repl))
3575 if (PL_bufptr != PL_bufend)
3576 Perl_croak(aTHX_ "Bad evalled substitution pattern");
3580 case LEX_INTERPCONCAT:
3582 if (PL_lex_brackets)
3583 Perl_croak(aTHX_ "panic: INTERPCONCAT");
3585 if (PL_bufptr == PL_bufend)
3586 return REPORT(sublex_done());
3588 if (SvIVX(PL_linestr) == '\'') {
3589 SV *sv = newSVsv(PL_linestr);
3592 else if ( PL_hints & HINT_NEW_RE )
3593 sv = new_constant(NULL, 0, "qr", sv, sv, "q", 1);
3594 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
3598 s = scan_const(PL_bufptr);
3600 PL_lex_state = LEX_INTERPCASEMOD;
3602 PL_lex_state = LEX_INTERPSTART;
3605 if (s != PL_bufptr) {
3606 start_force(PL_curforce);
3608 curmad('X', newSVpvn(PL_bufptr,s-PL_bufptr));
3610 NEXTVAL_NEXTTOKE = pl_yylval;
3613 if (PL_lex_starts++) {
3617 sv_free(PL_thistoken);
3618 PL_thistoken = newSVpvs("");
3621 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
3622 if (!PL_lex_casemods && PL_lex_inpat)
3635 PL_lex_state = LEX_NORMAL;
3636 s = scan_formline(PL_bufptr);
3637 if (!PL_lex_formbrack)
3643 PL_oldoldbufptr = PL_oldbufptr;
3649 sv_free(PL_thistoken);
3652 PL_realtokenstart = s - SvPVX(PL_linestr); /* assume but undo on ws */
3656 if (isIDFIRST_lazy_if(s,UTF))
3659 unsigned char c = *s;
3660 len = UTF ? Perl_utf8_length(aTHX_ (U8 *) PL_linestart, (U8 *) s) : (STRLEN) (s - PL_linestart);
3661 if (len > UNRECOGNIZED_PRECEDE_COUNT) {
3662 d = UTF ? (char *) Perl_utf8_hop(aTHX_ (U8 *) s, -UNRECOGNIZED_PRECEDE_COUNT) : s - UNRECOGNIZED_PRECEDE_COUNT;
3667 Perl_croak(aTHX_ "Unrecognized character \\x%02X; marked by <-- HERE after %s<-- HERE near column %d", c, d, (int) len + 1);
3671 goto fake_eof; /* emulate EOF on ^D or ^Z */
3680 if (PL_lex_brackets) {
3681 yyerror((const char *)
3683 ? "Format not terminated"
3684 : "Missing right curly or square bracket"));
3686 DEBUG_T( { PerlIO_printf(Perl_debug_log,
3687 "### Tokener got EOF\n");
3691 if (s++ < PL_bufend)
3692 goto retry; /* ignore stray nulls */
3695 if (!PL_in_eval && !PL_preambled) {
3696 PL_preambled = TRUE;
3702 /* Generate a string of Perl code to load the debugger.
3703 * If PERL5DB is set, it will return the contents of that,
3704 * otherwise a compile-time require of perl5db.pl. */
3706 const char * const pdb = PerlEnv_getenv("PERL5DB");
3709 sv_setpv(PL_linestr, pdb);
3710 sv_catpvs(PL_linestr,";");
3712 SETERRNO(0,SS_NORMAL);
3713 sv_setpvs(PL_linestr, "BEGIN { require 'perl5db.pl' };");
3716 sv_setpvs(PL_linestr,"");
3717 if (PL_preambleav) {
3718 SV **svp = AvARRAY(PL_preambleav);
3719 SV **const end = svp + AvFILLp(PL_preambleav);
3721 sv_catsv(PL_linestr, *svp);
3723 sv_catpvs(PL_linestr, ";");
3725 sv_free(MUTABLE_SV(PL_preambleav));
3726 PL_preambleav = NULL;
3729 sv_catpvs(PL_linestr,
3730 "use feature ':5." STRINGIFY(PERL_VERSION) "';");
3731 if (PL_minus_n || PL_minus_p) {
3732 sv_catpvs(PL_linestr, "LINE: while (<>) {");
3734 sv_catpvs(PL_linestr,"chomp;");
3737 if ((*PL_splitstr == '/' || *PL_splitstr == '\''
3738 || *PL_splitstr == '"')
3739 && strchr(PL_splitstr + 1, *PL_splitstr))
3740 Perl_sv_catpvf(aTHX_ PL_linestr, "our @F=split(%s);", PL_splitstr);
3742 /* "q\0${splitstr}\0" is legal perl. Yes, even NUL
3743 bytes can be used as quoting characters. :-) */
3744 const char *splits = PL_splitstr;
3745 sv_catpvs(PL_linestr, "our @F=split(q\0");
3748 if (*splits == '\\')
3749 sv_catpvn(PL_linestr, splits, 1);
3750 sv_catpvn(PL_linestr, splits, 1);
3751 } while (*splits++);
3752 /* This loop will embed the trailing NUL of
3753 PL_linestr as the last thing it does before
3755 sv_catpvs(PL_linestr, ");");
3759 sv_catpvs(PL_linestr,"our @F=split(' ');");
3762 sv_catpvs(PL_linestr, "\n");
3763 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3764 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3765 PL_last_lop = PL_last_uni = NULL;
3766 if ((PERLDB_LINE || PERLDB_SAVESRC) && PL_curstash != PL_debstash)
3767 update_debugger_info(PL_linestr, NULL, 0);
3771 bof = PL_rsfp ? TRUE : FALSE;
3772 if ((s = filter_gets(PL_linestr, PL_rsfp, 0)) == NULL) {
3775 PL_realtokenstart = -1;
3778 if ((PerlIO *)PL_rsfp == PerlIO_stdin())
3779 PerlIO_clearerr(PL_rsfp);
3781 (void)PerlIO_close(PL_rsfp);
3783 PL_doextract = FALSE;
3785 if (!PL_in_eval && (PL_minus_n || PL_minus_p)) {
3791 sv_setpvs(PL_linestr, ";}continue{print;}");
3793 sv_setpvs(PL_linestr, ";}");
3794 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3795 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3796 PL_last_lop = PL_last_uni = NULL;
3797 PL_minus_n = PL_minus_p = 0;
3800 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3801 PL_last_lop = PL_last_uni = NULL;
3802 sv_setpvs(PL_linestr,"");
3803 TOKEN(';'); /* not infinite loop because rsfp is NULL now */
3805 /* If it looks like the start of a BOM or raw UTF-16,
3806 * check if it in fact is. */
3812 #ifdef PERLIO_IS_STDIO
3813 # ifdef __GNU_LIBRARY__
3814 # if __GNU_LIBRARY__ == 1 /* Linux glibc5 */
3815 # define FTELL_FOR_PIPE_IS_BROKEN
3819 # if __GLIBC__ == 1 /* maybe some glibc5 release had it like this? */
3820 # define FTELL_FOR_PIPE_IS_BROKEN
3825 bof = PerlIO_tell(PL_rsfp) == (Off_t)SvCUR(PL_linestr);
3827 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3828 s = swallow_bom((U8*)s);
3832 /* Incest with pod. */
3835 sv_catsv(PL_thiswhite, PL_linestr);
3837 if (*s == '=' && strnEQ(s, "=cut", 4) && !isALPHA(s[4])) {
3838 sv_setpvs(PL_linestr, "");
3839 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3840 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3841 PL_last_lop = PL_last_uni = NULL;
3842 PL_doextract = FALSE;
3846 } while (PL_doextract);
3847 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = s;
3848 if ((PERLDB_LINE || PERLDB_SAVESRC) && PL_curstash != PL_debstash)
3849 update_debugger_info(PL_linestr, NULL, 0);
3850 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3851 PL_last_lop = PL_last_uni = NULL;
3852 if (CopLINE(PL_curcop) == 1) {
3853 while (s < PL_bufend && isSPACE(*s))
3855 if (*s == ':' && s[1] != ':') /* for csh execing sh scripts */
3859 PL_thiswhite = newSVpvn(PL_linestart, s - PL_linestart);
3863 if (*s == '#' && *(s+1) == '!')
3865 #ifdef ALTERNATE_SHEBANG
3867 static char const as[] = ALTERNATE_SHEBANG;
3868 if (*s == as[0] && strnEQ(s, as, sizeof(as) - 1))
3869 d = s + (sizeof(as) - 1);
3871 #endif /* ALTERNATE_SHEBANG */
3880 while (*d && !isSPACE(*d))
3884 #ifdef ARG_ZERO_IS_SCRIPT
3885 if (ipathend > ipath) {
3887 * HP-UX (at least) sets argv[0] to the script name,
3888 * which makes $^X incorrect. And Digital UNIX and Linux,
3889 * at least, set argv[0] to the basename of the Perl
3890 * interpreter. So, having found "#!", we'll set it right.
3892 SV * const x = GvSV(gv_fetchpvs("\030", GV_ADD|GV_NOTQUAL,
3894 assert(SvPOK(x) || SvGMAGICAL(x));
3895 if (sv_eq(x, CopFILESV(PL_curcop))) {
3896 sv_setpvn(x, ipath, ipathend - ipath);
3902 const char *bstart = SvPV_const(CopFILESV(PL_curcop),blen);
3903 const char * const lstart = SvPV_const(x,llen);
3905 bstart += blen - llen;
3906 if (strnEQ(bstart, lstart, llen) && bstart[-1] == '/') {
3907 sv_setpvn(x, ipath, ipathend - ipath);
3912 TAINT_NOT; /* $^X is always tainted, but that's OK */
3914 #endif /* ARG_ZERO_IS_SCRIPT */
3919 d = instr(s,"perl -");
3921 d = instr(s,"perl");
3923 /* avoid getting into infinite loops when shebang
3924 * line contains "Perl" rather than "perl" */
3926 for (d = ipathend-4; d >= ipath; --d) {
3927 if ((*d == 'p' || *d == 'P')
3928 && !ibcmp(d, "perl", 4))
3938 #ifdef ALTERNATE_SHEBANG
3940 * If the ALTERNATE_SHEBANG on this system starts with a
3941 * character that can be part of a Perl expression, then if
3942 * we see it but not "perl", we're probably looking at the
3943 * start of Perl code, not a request to hand off to some
3944 * other interpreter. Similarly, if "perl" is there, but
3945 * not in the first 'word' of the line, we assume the line
3946 * contains the start of the Perl program.
3948 if (d && *s != '#') {
3949 const char *c = ipath;
3950 while (*c && !strchr("; \t\r\n\f\v#", *c))
3953 d = NULL; /* "perl" not in first word; ignore */
3955 *s = '#'; /* Don't try to parse shebang line */
3957 #endif /* ALTERNATE_SHEBANG */
3962 !instr(s,"indir") &&
3963 instr(PL_origargv[0],"perl"))
3970 while (s < PL_bufend && isSPACE(*s))
3972 if (s < PL_bufend) {
3973 Newx(newargv,PL_origargc+3,char*);
3975 while (s < PL_bufend && !isSPACE(*s))
3978 Copy(PL_origargv+1, newargv+2, PL_origargc+1, char*);
3981 newargv = PL_origargv;
3984 PerlProc_execv(ipath, EXEC_ARGV_CAST(newargv));
3986 Perl_croak(aTHX_ "Can't exec %s", ipath);
3989 while (*d && !isSPACE(*d))
3991 while (SPACE_OR_TAB(*d))
3995 const bool switches_done = PL_doswitches;
3996 const U32 oldpdb = PL_perldb;
3997 const bool oldn = PL_minus_n;
3998 const bool oldp = PL_minus_p;
4002 if (*d1 == 'M' || *d1 == 'm' || *d1 == 'C') {
4003 const char * const m = d1;
4004 while (*d1 && !isSPACE(*d1))
4006 Perl_croak(aTHX_ "Too late for \"-%.*s\" option",
4009 d1 = moreswitches(d1);
4011 if (PL_doswitches && !switches_done) {
4012 int argc = PL_origargc;
4013 char **argv = PL_origargv;
4016 } while (argc && argv[0][0] == '-' && argv[0][1]);
4017 init_argv_symbols(argc,argv);
4019 if (((PERLDB_LINE || PERLDB_SAVESRC) && !oldpdb) ||
4020 ((PL_minus_n || PL_minus_p) && !(oldn || oldp)))
4021 /* if we have already added "LINE: while (<>) {",
4022 we must not do it again */
4024 sv_setpvs(PL_linestr, "");
4025 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
4026 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
4027 PL_last_lop = PL_last_uni = NULL;
4028 PL_preambled = FALSE;
4029 if (PERLDB_LINE || PERLDB_SAVESRC)
4030 (void)gv_fetchfile(PL_origfilename);
4037 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
4039 PL_lex_state = LEX_FORMLINE;
4044 #ifdef PERL_STRICT_CR
4045 Perl_warn(aTHX_ "Illegal character \\%03o (carriage return)", '\r');
4047 "\t(Maybe you didn't strip carriage returns after a network transfer?)\n");
4049 case ' ': case '\t': case '\f': case 013:
4051 PL_realtokenstart = -1;
4053 PL_thiswhite = newSVpvs("");
4054 sv_catpvn(PL_thiswhite, s, 1);
4061 PL_realtokenstart = -1;
4065 if (PL_lex_state != LEX_NORMAL || (PL_in_eval && !PL_rsfp)) {
4066 if (*s == '#' && s == PL_linestart && PL_in_eval && !PL_rsfp) {
4067 /* handle eval qq[#line 1 "foo"\n ...] */
4068 CopLINE_dec(PL_curcop);
4071 if (PL_madskills && !PL_lex_formbrack && !PL_in_eval) {
4073 if (!PL_in_eval || PL_rsfp)
4078 while (d < PL_bufend && *d != '\n')
4082 else if (d > PL_bufend) /* Found by Ilya: feed random input to Perl. */
4083 Perl_croak(aTHX_ "panic: input overflow");
4086 PL_thiswhite = newSVpvn(s, d - s);
4091 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
4093 PL_lex_state = LEX_FORMLINE;
4099 if (PL_madskills && CopLINE(PL_curcop) >= 1 && !PL_lex_formbrack) {
4100 if (CopLINE(PL_curcop) == 1 && s[0] == '#' && s[1] == '!') {
4103 TOKEN(PEG); /* make sure any #! line is accessible */
4108 /* if (PL_madskills && PL_lex_formbrack) { */
4110 while (d < PL_bufend && *d != '\n')
4114 else if (d > PL_bufend) /* Found by Ilya: feed random input to Perl. */
4115 Perl_croak(aTHX_ "panic: input overflow");
4116 if (PL_madskills && CopLINE(PL_curcop) >= 1) {
4118 PL_thiswhite = newSVpvs("");
4119 if (CopLINE(PL_curcop) == 1) {
4120 sv_setpvs(PL_thiswhite, "");
4123 sv_catpvn(PL_thiswhite, s, d - s);
4137 if (s[1] && isALPHA(s[1]) && !isALNUM(s[2])) {
4145 while (s < PL_bufend && SPACE_OR_TAB(*s))
4148 if (strnEQ(s,"=>",2)) {
4149 s = force_word(PL_bufptr,WORD,FALSE,FALSE,FALSE);
4150 DEBUG_T( { printbuf("### Saw unary minus before =>, forcing word %s\n", s); } );
4151 OPERATOR('-'); /* unary minus */
4153 PL_last_uni = PL_oldbufptr;
4155 case 'r': ftst = OP_FTEREAD; break;
4156 case 'w': ftst = OP_FTEWRITE; break;
4157 case 'x': ftst = OP_FTEEXEC; break;
4158 case 'o': ftst = OP_FTEOWNED; break;
4159 case 'R': ftst = OP_FTRREAD; break;
4160 case 'W': ftst = OP_FTRWRITE; break;
4161 case 'X': ftst = OP_FTREXEC; break;
4162 case 'O': ftst = OP_FTROWNED; break;
4163 case 'e': ftst = OP_FTIS; break;
4164 case 'z': ftst = OP_FTZERO; break;
4165 case 's': ftst = OP_FTSIZE; break;
4166 case 'f': ftst = OP_FTFILE; break;
4167 case 'd': ftst = OP_FTDIR; break;
4168 case 'l': ftst = OP_FTLINK; break;
4169 case 'p': ftst = OP_FTPIPE; break;
4170 case 'S': ftst = OP_FTSOCK; break;
4171 case 'u': ftst = OP_FTSUID; break;
4172 case 'g': ftst = OP_FTSGID; break;
4173 case 'k': ftst = OP_FTSVTX; break;
4174 case 'b': ftst = OP_FTBLK; break;
4175 case 'c': ftst = OP_FTCHR; break;
4176 case 't': ftst = OP_FTTTY; break;
4177 case 'T': ftst = OP_FTTEXT; break;
4178 case 'B': ftst = OP_FTBINARY; break;
4179 case 'M': case 'A': case 'C':
4180 gv_fetchpvs("\024", GV_ADD|GV_NOTQUAL, SVt_PV);
4182 case 'M': ftst = OP_FTMTIME; break;
4183 case 'A': ftst = OP_FTATIME; break;
4184 case 'C': ftst = OP_FTCTIME; break;
4192 PL_last_lop_op = (OPCODE)ftst;
4193 DEBUG_T( { PerlIO_printf(Perl_debug_log,
4194 "### Saw file test %c\n", (int)tmp);
4199 /* Assume it was a minus followed by a one-letter named
4200 * subroutine call (or a -bareword), then. */
4201 DEBUG_T( { PerlIO_printf(Perl_debug_log,
4202 "### '-%c' looked like a file test but was not\n",
4209 const char tmp = *s++;
4212 if (PL_expect == XOPERATOR)
4217 else if (*s == '>') {
4220 if (isIDFIRST_lazy_if(s,UTF)) {
4221 s = force_word(s,METHOD,FALSE,TRUE,FALSE);
4229 if (PL_expect == XOPERATOR)
4232 if (isSPACE(*s) || !isSPACE(*PL_bufptr))
4234 OPERATOR('-'); /* unary minus */
4240 const char tmp = *s++;
4243 if (PL_expect == XOPERATOR)
4248 if (PL_expect == XOPERATOR)
4251 if (isSPACE(*s) || !isSPACE(*PL_bufptr))
4258 if (PL_expect != XOPERATOR) {
4259 s = scan_ident(s, PL_bufend, PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
4260 PL_expect = XOPERATOR;
4261 force_ident(PL_tokenbuf, '*');
4274 if (PL_expect == XOPERATOR) {
4278 PL_tokenbuf[0] = '%';
4279 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1,
4280 sizeof PL_tokenbuf - 1, FALSE);
4281 if (!PL_tokenbuf[1]) {
4284 PL_pending_ident = '%';
4293 const char tmp = *s++;
4298 && (PL_expect == XOPERATOR || PL_expect == XTERMORDORDOR))
4305 const char tmp = *s++;
4311 goto just_a_word_zero_gv;
4314 switch (PL_expect) {
4320 if (!PL_in_my || PL_lex_state != LEX_NORMAL)
4322 PL_bufptr = s; /* update in case we back off */
4328 PL_expect = XTERMBLOCK;
4331 stuffstart = s - SvPVX(PL_linestr) - 1;
4335 while (isIDFIRST_lazy_if(s,UTF)) {
4338 d = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
4339 if (isLOWER(*s) && (tmp = keyword(PL_tokenbuf, len, 0))) {
4340 if (tmp < 0) tmp = -tmp;
4355 sv = newSVpvn(s, len);
4357 d = scan_str(d,TRUE,TRUE);
4359 /* MUST advance bufptr here to avoid bogus
4360 "at end of line" context messages from yyerror().
4362 PL_bufptr = s + len;
4363 yyerror("Unterminated attribute parameter in attribute list");
4367 return REPORT(0); /* EOF indicator */
4371 sv_catsv(sv, PL_lex_stuff);
4372 attrs = append_elem(OP_LIST, attrs,
4373 newSVOP(OP_CONST, 0, sv));
4374 SvREFCNT_dec(PL_lex_stuff);
4375 PL_lex_stuff = NULL;
4378 if (len == 6 && strnEQ(SvPVX(sv), "unique", len)) {
4380 if (PL_in_my == KEY_our) {
4381 deprecate(":unique");
4384 Perl_croak(aTHX_ "The 'unique' attribute may only be applied to 'our' variables");
4387 /* NOTE: any CV attrs applied here need to be part of
4388 the CVf_BUILTIN_ATTRS define in cv.h! */
4389 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "lvalue", len)) {
4391 CvLVALUE_on(PL_compcv);
4393 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "locked", len)) {
4395 deprecate(":locked");
4397 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "method", len)) {
4399 CvMETHOD_on(PL_compcv);
4401 /* After we've set the flags, it could be argued that
4402 we don't need to do the attributes.pm-based setting
4403 process, and shouldn't bother appending recognized
4404 flags. To experiment with that, uncomment the
4405 following "else". (Note that's already been
4406 uncommented. That keeps the above-applied built-in
4407 attributes from being intercepted (and possibly
4408 rejected) by a package's attribute routines, but is
4409 justified by the performance win for the common case
4410 of applying only built-in attributes.) */
4412 attrs = append_elem(OP_LIST, attrs,
4413 newSVOP(OP_CONST, 0,
4417 if (*s == ':' && s[1] != ':')
4420 break; /* require real whitespace or :'s */
4421 /* XXX losing whitespace on sequential attributes here */
4425 = (PL_expect == XOPERATOR ? '=' : '{'); /*'}(' for vi */
4426 if (*s != ';' && *s != '}' && *s != tmp
4427 && (tmp != '=' || *s != ')')) {
4428 const char q = ((*s == '\'') ? '"' : '\'');
4429 /* If here for an expression, and parsed no attrs, back
4431 if (tmp == '=' && !attrs) {
4435 /* MUST advance bufptr here to avoid bogus "at end of line"
4436 context messages from yyerror().
4439 yyerror( (const char *)
4441 ? Perl_form(aTHX_ "Invalid separator character "
4442 "%c%c%c in attribute list", q, *s, q)
4443 : "Unterminated attribute list" ) );
4451 start_force(PL_curforce);
4452 NEXTVAL_NEXTTOKE.opval = attrs;
4453 CURMAD('_', PL_nextwhite);
4458 PL_thistoken = newSVpvn(SvPVX(PL_linestr) + stuffstart,
4459 (s - SvPVX(PL_linestr)) - stuffstart);
4467 if (PL_last_lop == PL_oldoldbufptr || PL_last_uni == PL_oldoldbufptr)
4468 PL_oldbufptr = PL_oldoldbufptr; /* allow print(STDOUT 123) */
4476 const char tmp = *s++;
4481 const char tmp = *s++;
4489 if (PL_lex_brackets <= 0)
4490 yyerror("Unmatched right square bracket");
4493 if (PL_lex_state == LEX_INTERPNORMAL) {
4494 if (PL_lex_brackets == 0) {
4495 if (*s == '-' && s[1] == '>')
4496 PL_lex_state = LEX_INTERPENDMAYBE;
4497 else if (*s != '[' && *s != '{')
4498 PL_lex_state = LEX_INTERPEND;
4505 if (PL_lex_brackets > 100) {
4506 Renew(PL_lex_brackstack, PL_lex_brackets + 10, char);
4508 switch (PL_expect) {
4510 if (PL_lex_formbrack) {
4514 if (PL_oldoldbufptr == PL_last_lop)
4515 PL_lex_brackstack[PL_lex_brackets++] = XTERM;
4517 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
4518 OPERATOR(HASHBRACK);
4520 while (s < PL_bufend && SPACE_OR_TAB(*s))
4523 PL_tokenbuf[0] = '\0';
4524 if (d < PL_bufend && *d == '-') {
4525 PL_tokenbuf[0] = '-';
4527 while (d < PL_bufend && SPACE_OR_TAB(*d))
4530 if (d < PL_bufend && isIDFIRST_lazy_if(d,UTF)) {
4531 d = scan_word(d, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1,
4533 while (d < PL_bufend && SPACE_OR_TAB(*d))
4536 const char minus = (PL_tokenbuf[0] == '-');
4537 s = force_word(s + minus, WORD, FALSE, TRUE, FALSE);
4545 PL_lex_brackstack[PL_lex_brackets++] = XSTATE;
4550 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
4555 if (PL_oldoldbufptr == PL_last_lop)
4556 PL_lex_brackstack[PL_lex_brackets++] = XTERM;
4558 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
4561 if (PL_expect == XREF && PL_lex_state == LEX_INTERPNORMAL) {
4563 /* This hack is to get the ${} in the message. */
4565 yyerror("syntax error");
4568 OPERATOR(HASHBRACK);
4570 /* This hack serves to disambiguate a pair of curlies
4571 * as being a block or an anon hash. Normally, expectation
4572 * determines that, but in cases where we're not in a
4573 * position to expect anything in particular (like inside
4574 * eval"") we have to resolve the ambiguity. This code
4575 * covers the case where the first term in the curlies is a
4576 * quoted string. Most other cases need to be explicitly
4577 * disambiguated by prepending a "+" before the opening
4578 * curly in order to force resolution as an anon hash.
4580 * XXX should probably propagate the outer expectation
4581 * into eval"" to rely less on this hack, but that could
4582 * potentially break current behavior of eval"".
4586 if (*s == '\'' || *s == '"' || *s == '`') {
4587 /* common case: get past first string, handling escapes */
4588 for (t++; t < PL_bufend && *t != *s;)
4589 if (*t++ == '\\' && (*t == '\\' || *t == *s))
4593 else if (*s == 'q') {
4596 || ((*t == 'q' || *t == 'x') && ++t < PL_bufend
4599 /* skip q//-like construct */
4601 char open, close, term;
4604 while (t < PL_bufend && isSPACE(*t))
4606 /* check for q => */
4607 if (t+1 < PL_bufend && t[0] == '=' && t[1] == '>') {
4608 OPERATOR(HASHBRACK);
4612 if (term && (tmps = strchr("([{< )]}> )]}>",term)))
4616 for (t++; t < PL_bufend; t++) {
4617 if (*t == '\\' && t+1 < PL_bufend && open != '\\')
4619 else if (*t == open)
4623 for (t++; t < PL_bufend; t++) {
4624 if (*t == '\\' && t+1 < PL_bufend)
4626 else if (*t == close && --brackets <= 0)
4628 else if (*t == open)
4635 /* skip plain q word */
4636 while (t < PL_bufend && isALNUM_lazy_if(t,UTF))
4639 else if (isALNUM_lazy_if(t,UTF)) {
4641 while (t < PL_bufend && isALNUM_lazy_if(t,UTF))
4644 while (t < PL_bufend && isSPACE(*t))
4646 /* if comma follows first term, call it an anon hash */
4647 /* XXX it could be a comma expression with loop modifiers */
4648 if (t < PL_bufend && ((*t == ',' && (*s == 'q' || !isLOWER(*s)))
4649 || (*t == '=' && t[1] == '>')))
4650 OPERATOR(HASHBRACK);
4651 if (PL_expect == XREF)
4654 PL_lex_brackstack[PL_lex_brackets-1] = XSTATE;
4660 pl_yylval.ival = CopLINE(PL_curcop);
4661 if (isSPACE(*s) || *s == '#')
4662 PL_copline = NOLINE; /* invalidate current command line number */
4667 if (PL_lex_brackets <= 0)
4668 yyerror("Unmatched right curly bracket");
4670 PL_expect = (expectation)PL_lex_brackstack[--PL_lex_brackets];
4671 if (PL_lex_brackets < PL_lex_formbrack && PL_lex_state != LEX_INTERPNORMAL)
4672 PL_lex_formbrack = 0;
4673 if (PL_lex_state == LEX_INTERPNORMAL) {
4674 if (PL_lex_brackets == 0) {
4675 if (PL_expect & XFAKEBRACK) {
4676 PL_expect &= XENUMMASK;
4677 PL_lex_state = LEX_INTERPEND;
4682 PL_thiswhite = newSVpvs("");
4683 sv_catpvs(PL_thiswhite,"}");
4686 return yylex(); /* ignore fake brackets */
4688 if (*s == '-' && s[1] == '>')
4689 PL_lex_state = LEX_INTERPENDMAYBE;
4690 else if (*s != '[' && *s != '{')
4691 PL_lex_state = LEX_INTERPEND;
4694 if (PL_expect & XFAKEBRACK) {
4695 PL_expect &= XENUMMASK;
4697 return yylex(); /* ignore fake brackets */
4699 start_force(PL_curforce);
4701 curmad('X', newSVpvn(s-1,1));
4702 CURMAD('_', PL_thiswhite);
4707 PL_thistoken = newSVpvs("");
4715 if (PL_expect == XOPERATOR) {
4716 if (PL_bufptr == PL_linestart && ckWARN(WARN_SEMICOLON)
4717 && isIDFIRST_lazy_if(s,UTF))
4719 CopLINE_dec(PL_curcop);
4720 Perl_warner(aTHX_ packWARN(WARN_SEMICOLON), "%s", PL_warn_nosemi);
4721 CopLINE_inc(PL_curcop);
4726 s = scan_ident(s - 1, PL_bufend, PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
4728 PL_expect = XOPERATOR;
4729 force_ident(PL_tokenbuf, '&');
4733 pl_yylval.ival = (OPpENTERSUB_AMPER<<8);
4745 const char tmp = *s++;
4752 if (tmp && isSPACE(*s) && ckWARN(WARN_SYNTAX)
4753 && strchr("+-*/%.^&|<",tmp))
4754 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
4755 "Reversed %c= operator",(int)tmp);
4757 if (PL_expect == XSTATE && isALPHA(tmp) &&
4758 (s == PL_linestart+1 || s[-2] == '\n') )
4760 if (PL_in_eval && !PL_rsfp) {
4765 if (strnEQ(s,"=cut",4)) {
4781 PL_thiswhite = newSVpvs("");
4782 sv_catpvn(PL_thiswhite, PL_linestart,
4783 PL_bufend - PL_linestart);
4787 PL_doextract = TRUE;
4791 if (PL_lex_brackets < PL_lex_formbrack) {
4793 #ifdef PERL_STRICT_CR
4794 while (SPACE_OR_TAB(*t))
4796 while (SPACE_OR_TAB(*t) || *t == '\r')
4799 if (*t == '\n' || *t == '#') {
4810 const char tmp = *s++;
4812 /* was this !=~ where !~ was meant?
4813 * warn on m:!=~\s+([/?]|[msy]\W|tr\W): */
4815 if (*s == '~' && ckWARN(WARN_SYNTAX)) {
4816 const char *t = s+1;
4818 while (t < PL_bufend && isSPACE(*t))
4821 if (*t == '/' || *t == '?' ||
4822 ((*t == 'm' || *t == 's' || *t == 'y')
4823 && !isALNUM(t[1])) ||
4824 (*t == 't' && t[1] == 'r' && !isALNUM(t[2])))
4825 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
4826 "!=~ should be !~");
4836 if (PL_expect != XOPERATOR) {
4837 if (s[1] != '<' && !strchr(s,'>'))
4840 s = scan_heredoc(s);
4842 s = scan_inputsymbol(s);
4843 TERM(sublex_start());
4849 SHop(OP_LEFT_SHIFT);
4863 const char tmp = *s++;
4865 SHop(OP_RIGHT_SHIFT);
4866 else if (tmp == '=')
4875 if (PL_expect == XOPERATOR) {
4876 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
4878 deprecate_old(commaless_variable_list);
4879 return REPORT(','); /* grandfather non-comma-format format */
4883 if (s[1] == '#' && (isIDFIRST_lazy_if(s+2,UTF) || strchr("{$:+-", s[2]))) {
4884 PL_tokenbuf[0] = '@';
4885 s = scan_ident(s + 1, PL_bufend, PL_tokenbuf + 1,
4886 sizeof PL_tokenbuf - 1, FALSE);
4887 if (PL_expect == XOPERATOR)
4888 no_op("Array length", s);
4889 if (!PL_tokenbuf[1])
4891 PL_expect = XOPERATOR;
4892 PL_pending_ident = '#';
4896 PL_tokenbuf[0] = '$';
4897 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1,
4898 sizeof PL_tokenbuf - 1, FALSE);
4899 if (PL_expect == XOPERATOR)
4901 if (!PL_tokenbuf[1]) {
4903 yyerror("Final $ should be \\$ or $name");
4907 /* This kludge not intended to be bulletproof. */
4908 if (PL_tokenbuf[1] == '[' && !PL_tokenbuf[2]) {
4909 pl_yylval.opval = newSVOP(OP_CONST, 0,
4910 newSViv(CopARYBASE_get(&PL_compiling)));
4911 pl_yylval.opval->op_private = OPpCONST_ARYBASE;
4917 const char tmp = *s;
4918 if (PL_lex_state == LEX_NORMAL)
4921 if ((PL_expect != XREF || PL_oldoldbufptr == PL_last_lop)
4922 && intuit_more(s)) {
4924 PL_tokenbuf[0] = '@';
4925 if (ckWARN(WARN_SYNTAX)) {
4928 while (isSPACE(*t) || isALNUM_lazy_if(t,UTF) || *t == '$')
4931 PL_bufptr = PEEKSPACE(PL_bufptr); /* XXX can realloc */
4932 while (t < PL_bufend && *t != ']')
4934 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
4935 "Multidimensional syntax %.*s not supported",
4936 (int)((t - PL_bufptr) + 1), PL_bufptr);
4940 else if (*s == '{') {
4942 PL_tokenbuf[0] = '%';
4943 if (strEQ(PL_tokenbuf+1, "SIG") && ckWARN(WARN_SYNTAX)
4944 && (t = strchr(s, '}')) && (t = strchr(t, '=')))
4946 char tmpbuf[sizeof PL_tokenbuf];
4949 } while (isSPACE(*t));
4950 if (isIDFIRST_lazy_if(t,UTF)) {
4952 t = scan_word(t, tmpbuf, sizeof tmpbuf, TRUE,
4956 if (*t == ';' && get_cvn_flags(tmpbuf, len, 0))
4957 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
4958 "You need to quote \"%s\"",
4965 PL_expect = XOPERATOR;
4966 if (PL_lex_state == LEX_NORMAL && isSPACE((char)tmp)) {
4967 const bool islop = (PL_last_lop == PL_oldoldbufptr);
4968 if (!islop || PL_last_lop_op == OP_GREPSTART)
4969 PL_expect = XOPERATOR;
4970 else if (strchr("$@\"'`q", *s))
4971 PL_expect = XTERM; /* e.g. print $fh "foo" */
4972 else if (strchr("&*<%", *s) && isIDFIRST_lazy_if(s+1,UTF))
4973 PL_expect = XTERM; /* e.g. print $fh &sub */
4974 else if (isIDFIRST_lazy_if(s,UTF)) {
4975 char tmpbuf[sizeof PL_tokenbuf];
4977 scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
4978 if ((t2 = keyword(tmpbuf, len, 0))) {
4979 /* binary operators exclude handle interpretations */
4991 PL_expect = XTERM; /* e.g. print $fh length() */
4996 PL_expect = XTERM; /* e.g. print $fh subr() */
4999 else if (isDIGIT(*s))
5000 PL_expect = XTERM; /* e.g. print $fh 3 */
5001 else if (*s == '.' && isDIGIT(s[1]))
5002 PL_expect = XTERM; /* e.g. print $fh .3 */
5003 else if ((*s == '?' || *s == '-' || *s == '+')
5004 && !isSPACE(s[1]) && s[1] != '=')
5005 PL_expect = XTERM; /* e.g. print $fh -1 */
5006 else if (*s == '/' && !isSPACE(s[1]) && s[1] != '='
5008 PL_expect = XTERM; /* e.g. print $fh /.../
5009 XXX except DORDOR operator
5011 else if (*s == '<' && s[1] == '<' && !isSPACE(s[2])
5013 PL_expect = XTERM; /* print $fh <<"EOF" */
5016 PL_pending_ident = '$';
5020 if (PL_expect == XOPERATOR)
5022 PL_tokenbuf[0] = '@';
5023 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1, FALSE);
5024 if (!PL_tokenbuf[1]) {
5027 if (PL_lex_state == LEX_NORMAL)
5029 if ((PL_expect != XREF || PL_oldoldbufptr == PL_last_lop) && intuit_more(s)) {
5031 PL_tokenbuf[0] = '%';
5033 /* Warn about @ where they meant $. */
5034 if (*s == '[' || *s == '{') {
5035 if (ckWARN(WARN_SYNTAX)) {
5036 const char *t = s + 1;
5037 while (*t && (isALNUM_lazy_if(t,UTF) || strchr(" \t$#+-'\"", *t)))
5039 if (*t == '}' || *t == ']') {
5041 PL_bufptr = PEEKSPACE(PL_bufptr); /* XXX can realloc */
5042 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
5043 "Scalar value %.*s better written as $%.*s",
5044 (int)(t-PL_bufptr), PL_bufptr,
5045 (int)(t-PL_bufptr-1), PL_bufptr+1);
5050 PL_pending_ident = '@';
5053 case '/': /* may be division, defined-or, or pattern */
5054 if (PL_expect == XTERMORDORDOR && s[1] == '/') {
5058 case '?': /* may either be conditional or pattern */
5059 if (PL_expect == XOPERATOR) {
5067 /* A // operator. */
5077 /* Disable warning on "study /blah/" */
5078 if (PL_oldoldbufptr == PL_last_uni
5079 && (*PL_last_uni != 's' || s - PL_last_uni < 5
5080 || memNE(PL_last_uni, "study", 5)
5081 || isALNUM_lazy_if(PL_last_uni+5,UTF)
5084 s = scan_pat(s,OP_MATCH);
5085 TERM(sublex_start());
5089 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack
5090 #ifdef PERL_STRICT_CR
5093 && (s[1] == '\n' || (s[1] == '\r' && s[2] == '\n'))
5095 && (s == PL_linestart || s[-1] == '\n') )
5097 PL_lex_formbrack = 0;
5101 if (PL_expect == XSTATE && s[1] == '.' && s[2] == '.') {
5105 if (PL_expect == XOPERATOR || !isDIGIT(s[1])) {
5111 pl_yylval.ival = OPf_SPECIAL;
5117 if (PL_expect != XOPERATOR)
5122 case '0': case '1': case '2': case '3': case '4':
5123 case '5': case '6': case '7': case '8': case '9':
5124 s = scan_num(s, &pl_yylval);
5125 DEBUG_T( { printbuf("### Saw number in %s\n", s); } );
5126 if (PL_expect == XOPERATOR)
5131 s = scan_str(s,!!PL_madskills,FALSE);
5132 DEBUG_T( { printbuf("### Saw string before %s\n", s); } );
5133 if (PL_expect == XOPERATOR) {
5134 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
5136 deprecate_old(commaless_variable_list);
5137 return REPORT(','); /* grandfather non-comma-format format */
5144 pl_yylval.ival = OP_CONST;
5145 TERM(sublex_start());
5148 s = scan_str(s,!!PL_madskills,FALSE);
5149 DEBUG_T( { printbuf("### Saw string before %s\n", s); } );
5150 if (PL_expect == XOPERATOR) {
5151 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
5153 deprecate_old(commaless_variable_list);
5154 return REPORT(','); /* grandfather non-comma-format format */
5161 pl_yylval.ival = OP_CONST;
5162 /* FIXME. I think that this can be const if char *d is replaced by
5163 more localised variables. */
5164 for (d = SvPV(PL_lex_stuff, len); len; len--, d++) {
5165 if (*d == '$' || *d == '@' || *d == '\\' || !UTF8_IS_INVARIANT((U8)*d)) {
5166 pl_yylval.ival = OP_STRINGIFY;
5170 TERM(sublex_start());
5173 s = scan_str(s,!!PL_madskills,FALSE);
5174 DEBUG_T( { printbuf("### Saw backtick string before %s\n", s); } );
5175 if (PL_expect == XOPERATOR)
5176 no_op("Backticks",s);
5179 readpipe_override();
5180 TERM(sublex_start());
5184 if (PL_lex_inwhat && isDIGIT(*s) && ckWARN(WARN_SYNTAX))
5185 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),"Can't use \\%c to mean $%c in expression",
5187 if (PL_expect == XOPERATOR)
5188 no_op("Backslash",s);
5192 if (isDIGIT(s[1]) && PL_expect != XOPERATOR) {
5193 char *start = s + 2;
5194 while (isDIGIT(*start) || *start == '_')
5196 if (*start == '.' && isDIGIT(start[1])) {
5197 s = scan_num(s, &pl_yylval);
5200 /* avoid v123abc() or $h{v1}, allow C<print v10;> */
5201 else if (!isALPHA(*start) && (PL_expect == XTERM
5202 || PL_expect == XREF || PL_expect == XSTATE
5203 || PL_expect == XTERMORDORDOR)) {
5204 GV *const gv = gv_fetchpvn_flags(s, start - s, 0, SVt_PVCV);
5206 s = scan_num(s, &pl_yylval);
5213 if (isDIGIT(s[1]) && PL_expect == XOPERATOR) {
5255 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
5257 /* Some keywords can be followed by any delimiter, including ':' */
5258 tmp = ((len == 1 && strchr("msyq", PL_tokenbuf[0])) ||
5259 (len == 2 && ((PL_tokenbuf[0] == 't' && PL_tokenbuf[1] == 'r') ||
5260 (PL_tokenbuf[0] == 'q' &&
5261 strchr("qwxr", PL_tokenbuf[1])))));
5263 /* x::* is just a word, unless x is "CORE" */
5264 if (!tmp && *s == ':' && s[1] == ':' && strNE(PL_tokenbuf, "CORE"))
5268 while (d < PL_bufend && isSPACE(*d))
5269 d++; /* no comments skipped here, or s### is misparsed */
5271 /* Is this a label? */
5272 if (!tmp && PL_expect == XSTATE
5273 && d < PL_bufend && *d == ':' && *(d + 1) != ':') {
5275 pl_yylval.pval = CopLABEL_alloc(PL_tokenbuf);
5280 /* Check for keywords */
5281 tmp = keyword(PL_tokenbuf, len, 0);
5283 /* Is this a word before a => operator? */
5284 if (*d == '=' && d[1] == '>') {
5287 = (OP*)newSVOP(OP_CONST, 0,
5288 S_newSV_maybe_utf8(aTHX_ PL_tokenbuf, len));
5289 pl_yylval.opval->op_private = OPpCONST_BARE;
5293 if (tmp < 0) { /* second-class keyword? */
5294 GV *ogv = NULL; /* override (winner) */
5295 GV *hgv = NULL; /* hidden (loser) */
5296 if (PL_expect != XOPERATOR && (*s != ':' || s[1] != ':')) {
5298 if ((gv = gv_fetchpvn_flags(PL_tokenbuf, len, 0, SVt_PVCV)) &&
5301 if (GvIMPORTED_CV(gv))
5303 else if (! CvMETHOD(cv))
5307 (gvp = (GV**)hv_fetch(PL_globalstash,PL_tokenbuf,len,FALSE)) &&
5308 (gv = *gvp) && isGV_with_GP(gv) &&
5309 GvCVu(gv) && GvIMPORTED_CV(gv))
5316 tmp = 0; /* overridden by import or by GLOBAL */
5319 && -tmp==KEY_lock /* XXX generalizable kludge */
5322 tmp = 0; /* any sub overrides "weak" keyword */
5324 else { /* no override */
5326 if (tmp == KEY_dump && ckWARN(WARN_MISC)) {
5327 Perl_warner(aTHX_ packWARN(WARN_MISC),
5328 "dump() better written as CORE::dump()");
5332 if (hgv && tmp != KEY_x && tmp != KEY_CORE
5333 && ckWARN(WARN_AMBIGUOUS)) /* never ambiguous */
5334 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
5335 "Ambiguous call resolved as CORE::%s(), %s",
5336 GvENAME(hgv), "qualify as such or use &");
5343 default: /* not a keyword */
5344 /* Trade off - by using this evil construction we can pull the
5345 variable gv into the block labelled keylookup. If not, then
5346 we have to give it function scope so that the goto from the
5347 earlier ':' case doesn't bypass the initialisation. */
5349 just_a_word_zero_gv:
5357 const char lastchar = (PL_bufptr == PL_oldoldbufptr ? 0 : PL_bufptr[-1]);
5360 SV *nextPL_nextwhite = 0;
5364 /* Get the rest if it looks like a package qualifier */
5366 if (*s == '\'' || (*s == ':' && s[1] == ':')) {
5368 s = scan_word(s, PL_tokenbuf + len, sizeof PL_tokenbuf - len,
5371 Perl_croak(aTHX_ "Bad name after %s%s", PL_tokenbuf,
5372 *s == '\'' ? "'" : "::");
5377 if (PL_expect == XOPERATOR) {
5378 if (PL_bufptr == PL_linestart) {
5379 CopLINE_dec(PL_curcop);
5380 Perl_warner(aTHX_ packWARN(WARN_SEMICOLON), "%s", PL_warn_nosemi);
5381 CopLINE_inc(PL_curcop);
5384 no_op("Bareword",s);
5387 /* Look for a subroutine with this name in current package,
5388 unless name is "Foo::", in which case Foo is a bearword
5389 (and a package name). */
5391 if (len > 2 && !PL_madskills &&
5392 PL_tokenbuf[len - 2] == ':' && PL_tokenbuf[len - 1] == ':')
5394 if (ckWARN(WARN_BAREWORD)
5395 && ! gv_fetchpvn_flags(PL_tokenbuf, len, 0, SVt_PVHV))
5396 Perl_warner(aTHX_ packWARN(WARN_BAREWORD),
5397 "Bareword \"%s\" refers to nonexistent package",
5400 PL_tokenbuf[len] = '\0';
5406 /* Mustn't actually add anything to a symbol table.
5407 But also don't want to "initialise" any placeholder
5408 constants that might already be there into full
5409 blown PVGVs with attached PVCV. */
5410 gv = gv_fetchpvn_flags(PL_tokenbuf, len,
5411 GV_NOADD_NOINIT, SVt_PVCV);
5416 /* if we saw a global override before, get the right name */
5419 sv = newSVpvs("CORE::GLOBAL::");
5420 sv_catpv(sv,PL_tokenbuf);
5423 /* If len is 0, newSVpv does strlen(), which is correct.
5424 If len is non-zero, then it will be the true length,
5425 and so the scalar will be created correctly. */
5426 sv = newSVpv(PL_tokenbuf,len);
5429 if (PL_madskills && !PL_thistoken) {
5430 char *start = SvPVX(PL_linestr) + PL_realtokenstart;
5431 PL_thistoken = newSVpvn(start,s - start);
5432 PL_realtokenstart = s - SvPVX(PL_linestr);
5436 /* Presume this is going to be a bareword of some sort. */
5439 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
5440 pl_yylval.opval->op_private = OPpCONST_BARE;
5441 /* UTF-8 package name? */
5442 if (UTF && !IN_BYTES &&
5443 is_utf8_string((U8*)SvPVX_const(sv), SvCUR(sv)))
5446 /* And if "Foo::", then that's what it certainly is. */
5451 /* Do the explicit type check so that we don't need to force
5452 the initialisation of the symbol table to have a real GV.
5453 Beware - gv may not really be a PVGV, cv may not really be
5454 a PVCV, (because of the space optimisations that gv_init
5455 understands) But they're true if for this symbol there is
5456 respectively a typeglob and a subroutine.
5458 cv = gv ? ((SvTYPE(gv) == SVt_PVGV)
5459 /* Real typeglob, so get the real subroutine: */
5461 /* A proxy for a subroutine in this package? */
5462 : SvOK(gv) ? MUTABLE_CV(gv) : NULL)
5465 /* See if it's the indirect object for a list operator. */
5467 if (PL_oldoldbufptr &&
5468 PL_oldoldbufptr < PL_bufptr &&
5469 (PL_oldoldbufptr == PL_last_lop
5470 || PL_oldoldbufptr == PL_last_uni) &&
5471 /* NO SKIPSPACE BEFORE HERE! */
5472 (PL_expect == XREF ||
5473 ((PL_opargs[PL_last_lop_op] >> OASHIFT)& 7) == OA_FILEREF))
5475 bool immediate_paren = *s == '(';
5477 /* (Now we can afford to cross potential line boundary.) */
5478 s = SKIPSPACE2(s,nextPL_nextwhite);
5480 PL_nextwhite = nextPL_nextwhite; /* assume no & deception */
5483 /* Two barewords in a row may indicate method call. */
5485 if ((isIDFIRST_lazy_if(s,UTF) || *s == '$') &&
5486 (tmp = intuit_method(s, gv, cv)))
5489 /* If not a declared subroutine, it's an indirect object. */
5490 /* (But it's an indir obj regardless for sort.) */
5491 /* Also, if "_" follows a filetest operator, it's a bareword */
5494 ( !immediate_paren && (PL_last_lop_op == OP_SORT ||
5496 (PL_last_lop_op != OP_MAPSTART &&
5497 PL_last_lop_op != OP_GREPSTART))))
5498 || (PL_tokenbuf[0] == '_' && PL_tokenbuf[1] == '\0'
5499 && ((PL_opargs[PL_last_lop_op] & OA_CLASS_MASK) == OA_FILESTATOP))
5502 PL_expect = (PL_last_lop == PL_oldoldbufptr) ? XTERM : XOPERATOR;
5507 PL_expect = XOPERATOR;
5510 s = SKIPSPACE2(s,nextPL_nextwhite);
5511 PL_nextwhite = nextPL_nextwhite;
5516 /* Is this a word before a => operator? */
5517 if (*s == '=' && s[1] == '>' && !pkgname) {
5519 sv_setpv(((SVOP*)pl_yylval.opval)->op_sv, PL_tokenbuf);
5520 if (UTF && !IN_BYTES && is_utf8_string((U8*)PL_tokenbuf, len))
5521 SvUTF8_on(((SVOP*)pl_yylval.opval)->op_sv);
5525 /* If followed by a paren, it's certainly a subroutine. */
5530 while (SPACE_OR_TAB(*d))
5532 if (*d == ')' && (sv = gv_const_sv(gv))) {
5539 PL_nextwhite = PL_thiswhite;
5542 start_force(PL_curforce);
5544 NEXTVAL_NEXTTOKE.opval = pl_yylval.opval;
5545 PL_expect = XOPERATOR;
5548 PL_nextwhite = nextPL_nextwhite;
5549 curmad('X', PL_thistoken);
5550 PL_thistoken = newSVpvs("");
5558 /* If followed by var or block, call it a method (unless sub) */
5560 if ((*s == '$' || *s == '{') && (!gv || !cv)) {
5561 PL_last_lop = PL_oldbufptr;
5562 PL_last_lop_op = OP_METHOD;
5566 /* If followed by a bareword, see if it looks like indir obj. */
5569 && (isIDFIRST_lazy_if(s,UTF) || *s == '$')
5570 && (tmp = intuit_method(s, gv, cv)))
5573 /* Not a method, so call it a subroutine (if defined) */
5576 if (lastchar == '-' && ckWARN_d(WARN_AMBIGUOUS))
5577 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
5578 "Ambiguous use of -%s resolved as -&%s()",
5579 PL_tokenbuf, PL_tokenbuf);
5580 /* Check for a constant sub */
5581 if ((sv = gv_const_sv(gv))) {
5583 SvREFCNT_dec(((SVOP*)pl_yylval.opval)->op_sv);
5584 ((SVOP*)pl_yylval.opval)->op_sv = SvREFCNT_inc_simple(sv);
5585 pl_yylval.opval->op_private = 0;
5589 /* Resolve to GV now. */
5590 if (SvTYPE(gv) != SVt_PVGV) {
5591 gv = gv_fetchpv(PL_tokenbuf, 0, SVt_PVCV);
5592 assert (SvTYPE(gv) == SVt_PVGV);
5593 /* cv must have been some sort of placeholder, so
5594 now needs replacing with a real code reference. */
5598 op_free(pl_yylval.opval);
5599 pl_yylval.opval = newCVREF(0, newGVOP(OP_GV, 0, gv));
5600 pl_yylval.opval->op_private |= OPpENTERSUB_NOPAREN;
5601 PL_last_lop = PL_oldbufptr;
5602 PL_last_lop_op = OP_ENTERSUB;
5603 /* Is there a prototype? */
5611 const char *proto = SvPV_const(MUTABLE_SV(cv), protolen);
5614 if ((*proto == '$' || *proto == '_') && proto[1] == '\0')
5616 while (*proto == ';')
5618 if (*proto == '&' && *s == '{') {
5620 sv_setpvs(PL_subname, "__ANON__");
5622 sv_setpvs(PL_subname, "__ANON__::__ANON__");
5629 PL_nextwhite = PL_thiswhite;
5632 start_force(PL_curforce);
5633 NEXTVAL_NEXTTOKE.opval = pl_yylval.opval;
5636 PL_nextwhite = nextPL_nextwhite;
5637 curmad('X', PL_thistoken);
5638 PL_thistoken = newSVpvs("");
5645 /* Guess harder when madskills require "best effort". */
5646 if (PL_madskills && (!gv || !GvCVu(gv))) {
5647 int probable_sub = 0;
5648 if (strchr("\"'`$@%0123456789!*+{[<", *s))
5650 else if (isALPHA(*s)) {
5654 d = scan_word(d, tmpbuf, sizeof tmpbuf, TRUE, &tmplen);
5655 if (!keyword(tmpbuf, tmplen, 0))
5658 while (d < PL_bufend && isSPACE(*d))
5660 if (*d == '=' && d[1] == '>')
5665 gv = gv_fetchpv(PL_tokenbuf, GV_ADD, SVt_PVCV);
5666 op_free(pl_yylval.opval);
5667 pl_yylval.opval = newCVREF(0, newGVOP(OP_GV, 0, gv));
5668 pl_yylval.opval->op_private |= OPpENTERSUB_NOPAREN;
5669 PL_last_lop = PL_oldbufptr;
5670 PL_last_lop_op = OP_ENTERSUB;
5671 PL_nextwhite = PL_thiswhite;
5673 start_force(PL_curforce);
5674 NEXTVAL_NEXTTOKE.opval = pl_yylval.opval;
5676 PL_nextwhite = nextPL_nextwhite;
5677 curmad('X', PL_thistoken);
5678 PL_thistoken = newSVpvs("");
5683 NEXTVAL_NEXTTOKE.opval = pl_yylval.opval;
5690 /* Call it a bare word */
5693 if (PL_hints & HINT_STRICT_SUBS)
5694 pl_yylval.opval->op_private |= OPpCONST_STRICT;
5696 if (lastchar != '-') {
5697 if (ckWARN(WARN_RESERVED)) {
5701 if (!*d && !gv_stashpv(PL_tokenbuf, 0))
5702 Perl_warner(aTHX_ packWARN(WARN_RESERVED), PL_warn_reserved,
5709 if ((lastchar == '*' || lastchar == '%' || lastchar == '&')
5710 && ckWARN_d(WARN_AMBIGUOUS)) {
5711 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
5712 "Operator or semicolon missing before %c%s",
5713 lastchar, PL_tokenbuf);
5714 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
5715 "Ambiguous use of %c resolved as operator %c",
5716 lastchar, lastchar);
5722 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0,
5723 newSVpv(CopFILE(PL_curcop),0));
5727 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0,
5728 Perl_newSVpvf(aTHX_ "%"IVdf, (IV)CopLINE(PL_curcop)));
5731 case KEY___PACKAGE__:
5732 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0,
5734 ? newSVhek(HvNAME_HEK(PL_curstash))
5741 if (PL_rsfp && (!PL_in_eval || PL_tokenbuf[2] == 'D')) {
5742 const char *pname = "main";
5743 if (PL_tokenbuf[2] == 'D')
5744 pname = HvNAME_get(PL_curstash ? PL_curstash : PL_defstash);
5745 gv = gv_fetchpv(Perl_form(aTHX_ "%s::DATA", pname), GV_ADD,
5749 GvIOp(gv) = newIO();
5750 IoIFP(GvIOp(gv)) = PL_rsfp;
5751 #if defined(HAS_FCNTL) && defined(F_SETFD)
5753 const int fd = PerlIO_fileno(PL_rsfp);
5754 fcntl(fd,F_SETFD,fd >= 3);
5757 /* Mark this internal pseudo-handle as clean */
5758 IoFLAGS(GvIOp(gv)) |= IOf_UNTAINT;
5759 if ((PerlIO*)PL_rsfp == PerlIO_stdin())
5760 IoTYPE(GvIOp(gv)) = IoTYPE_STD;
5762 IoTYPE(GvIOp(gv)) = IoTYPE_RDONLY;
5763 #if defined(WIN32) && !defined(PERL_TEXTMODE_SCRIPTS)
5764 /* if the script was opened in binmode, we need to revert
5765 * it to text mode for compatibility; but only iff it has CRs
5766 * XXX this is a questionable hack at best. */
5767 if (PL_bufend-PL_bufptr > 2
5768 && PL_bufend[-1] == '\n' && PL_bufend[-2] == '\r')
5771 if (IoTYPE(GvIOp(gv)) == IoTYPE_RDONLY) {
5772 loc = PerlIO_tell(PL_rsfp);
5773 (void)PerlIO_seek(PL_rsfp, 0L, 0);
5776 if (PerlLIO_setmode(PL_rsfp, O_TEXT) != -1) {
5778 if (PerlLIO_setmode(PerlIO_fileno(PL_rsfp), O_TEXT) != -1) {
5779 #endif /* NETWARE */
5780 #ifdef PERLIO_IS_STDIO /* really? */
5781 # if defined(__BORLANDC__)
5782 /* XXX see note in do_binmode() */
5783 ((FILE*)PL_rsfp)->flags &= ~_F_BIN;
5787 PerlIO_seek(PL_rsfp, loc, 0);
5791 #ifdef PERLIO_LAYERS
5794 PerlIO_apply_layers(aTHX_ PL_rsfp, NULL, ":utf8");
5795 else if (PL_encoding) {
5802 XPUSHs(PL_encoding);
5804 call_method("name", G_SCALAR);
5808 PerlIO_apply_layers(aTHX_ PL_rsfp, NULL,
5809 Perl_form(aTHX_ ":encoding(%"SVf")",
5818 if (PL_realtokenstart >= 0) {
5819 char *tstart = SvPVX(PL_linestr) + PL_realtokenstart;
5821 PL_endwhite = newSVpvs("");
5822 sv_catsv(PL_endwhite, PL_thiswhite);
5824 sv_catpvn(PL_endwhite, tstart, PL_bufend - tstart);
5825 PL_realtokenstart = -1;
5827 while ((s = filter_gets(PL_endwhite, PL_rsfp,
5828 SvCUR(PL_endwhite))) != NULL) ;
5843 if (PL_expect == XSTATE) {
5850 if (*s == ':' && s[1] == ':') {
5853 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
5854 if (!(tmp = keyword(PL_tokenbuf, len, 0)))
5855 Perl_croak(aTHX_ "CORE::%s is not a keyword", PL_tokenbuf);
5858 else if (tmp == KEY_require || tmp == KEY_do)
5859 /* that's a way to remember we saw "CORE::" */
5872 LOP(OP_ACCEPT,XTERM);
5878 LOP(OP_ATAN2,XTERM);
5884 LOP(OP_BINMODE,XTERM);
5887 LOP(OP_BLESS,XTERM);
5896 /* When 'use switch' is in effect, continue has a dual
5897 life as a control operator. */
5899 if (!FEATURE_IS_ENABLED("switch"))
5902 /* We have to disambiguate the two senses of
5903 "continue". If the next token is a '{' then
5904 treat it as the start of a continue block;
5905 otherwise treat it as a control operator.
5917 (void)gv_fetchpvs("ENV", GV_ADD|GV_NOTQUAL, SVt_PVHV);
5934 if (!PL_cryptseen) {
5935 PL_cryptseen = TRUE;
5939 LOP(OP_CRYPT,XTERM);
5942 LOP(OP_CHMOD,XTERM);
5945 LOP(OP_CHOWN,XTERM);
5948 LOP(OP_CONNECT,XTERM);
5967 s = force_word(s,WORD,TRUE,TRUE,FALSE);
5968 if (orig_keyword == KEY_do) {
5977 PL_hints |= HINT_BLOCK_SCOPE;
5987 gv_fetchpvs("AnyDBM_File::ISA", GV_ADDMULTI, SVt_PVAV);
5988 LOP(OP_DBMOPEN,XTERM);
5994 s = force_word(s,WORD,TRUE,FALSE,FALSE);
6001 pl_yylval.ival = CopLINE(PL_curcop);
6017 PL_expect = (*s == '{') ? XTERMBLOCK : XTERM;
6018 UNIBRACK(OP_ENTEREVAL);
6032 case KEY_endhostent:
6038 case KEY_endservent:
6041 case KEY_endprotoent:
6052 pl_yylval.ival = CopLINE(PL_curcop);
6054 if (PL_expect == XSTATE && isIDFIRST_lazy_if(s,UTF)) {
6057 int soff = s - SvPVX(PL_linestr); /* for skipspace realloc */
6060 if ((PL_bufend - p) >= 3 &&
6061 strnEQ(p, "my", 2) && isSPACE(*(p + 2)))
6063 else if ((PL_bufend - p) >= 4 &&
6064 strnEQ(p, "our", 3) && isSPACE(*(p + 3)))
6067 if (isIDFIRST_lazy_if(p,UTF)) {
6068 p = scan_ident(p, PL_bufend,
6069 PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
6073 Perl_croak(aTHX_ "Missing $ on loop variable");
6075 s = SvPVX(PL_linestr) + soff;
6081 LOP(OP_FORMLINE,XTERM);
6087 LOP(OP_FCNTL,XTERM);
6093 LOP(OP_FLOCK,XTERM);
6102 LOP(OP_GREPSTART, XREF);
6105 s = force_word(s,WORD,TRUE,FALSE,FALSE);
6120 case KEY_getpriority:
6121 LOP(OP_GETPRIORITY,XTERM);
6123 case KEY_getprotobyname:
6126 case KEY_getprotobynumber:
6127 LOP(OP_GPBYNUMBER,XTERM);
6129 case KEY_getprotoent:
6141 case KEY_getpeername:
6142 UNI(OP_GETPEERNAME);
6144 case KEY_gethostbyname:
6147 case KEY_gethostbyaddr:
6148 LOP(OP_GHBYADDR,XTERM);
6150 case KEY_gethostent:
6153 case KEY_getnetbyname:
6156 case KEY_getnetbyaddr:
6157 LOP(OP_GNBYADDR,XTERM);
6162 case KEY_getservbyname:
6163 LOP(OP_GSBYNAME,XTERM);
6165 case KEY_getservbyport:
6166 LOP(OP_GSBYPORT,XTERM);
6168 case KEY_getservent:
6171 case KEY_getsockname:
6172 UNI(OP_GETSOCKNAME);
6174 case KEY_getsockopt:
6175 LOP(OP_GSOCKOPT,XTERM);
6190 pl_yylval.ival = CopLINE(PL_curcop);
6200 pl_yylval.ival = CopLINE(PL_curcop);
6204 LOP(OP_INDEX,XTERM);
6210 LOP(OP_IOCTL,XTERM);
6222 s = force_word(s,WORD,TRUE,FALSE,FALSE);
6254 LOP(OP_LISTEN,XTERM);
6263 s = scan_pat(s,OP_MATCH);
6264 TERM(sublex_start());
6267 LOP(OP_MAPSTART, XREF);
6270 LOP(OP_MKDIR,XTERM);
6273 LOP(OP_MSGCTL,XTERM);
6276 LOP(OP_MSGGET,XTERM);
6279 LOP(OP_MSGRCV,XTERM);
6282 LOP(OP_MSGSND,XTERM);
6287 PL_in_my = (U16)tmp;
6289 if (isIDFIRST_lazy_if(s,UTF)) {
6293 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, TRUE, &len);
6294 if (len == 3 && strnEQ(PL_tokenbuf, "sub", 3))
6296 PL_in_my_stash = find_in_my_stash(PL_tokenbuf, len);
6297 if (!PL_in_my_stash) {
6300 my_snprintf(tmpbuf, sizeof(tmpbuf), "No such class %.1000s", PL_tokenbuf);
6304 if (PL_madskills) { /* just add type to declarator token */
6305 sv_catsv(PL_thistoken, PL_nextwhite);
6307 sv_catpvn(PL_thistoken, start, s - start);
6315 s = force_word(s,WORD,TRUE,FALSE,FALSE);
6322 s = tokenize_use(0, s);
6326 if (*s == '(' || (s = SKIPSPACE1(s), *s == '('))
6333 if (isIDFIRST_lazy_if(s,UTF)) {
6335 for (d = s; isALNUM_lazy_if(d,UTF);)
6337 for (t=d; isSPACE(*t);)
6339 if ( *t && strchr("|&*+-=!?:.", *t) && ckWARN_d(WARN_PRECEDENCE)
6341 && !(t[0] == '=' && t[1] == '>')
6343 int parms_len = (int)(d-s);
6344 Perl_warner(aTHX_ packWARN(WARN_PRECEDENCE),
6345 "Precedence problem: open %.*s should be open(%.*s)",
6346 parms_len, s, parms_len, s);
6352 pl_yylval.ival = OP_OR;
6362 LOP(OP_OPEN_DIR,XTERM);
6365 checkcomma(s,PL_tokenbuf,"filehandle");
6369 checkcomma(s,PL_tokenbuf,"filehandle");
6388 s = force_word(s,WORD,FALSE,TRUE,FALSE);
6392 LOP(OP_PIPE_OP,XTERM);
6395 s = scan_str(s,!!PL_madskills,FALSE);
6398 pl_yylval.ival = OP_CONST;
6399 TERM(sublex_start());
6405 s = scan_str(s,!!PL_madskills,FALSE);
6408 PL_expect = XOPERATOR;
6410 if (SvCUR(PL_lex_stuff)) {
6413 d = SvPV_force(PL_lex_stuff, len);
6415 for (; isSPACE(*d) && len; --len, ++d)
6420 if (!warned && ckWARN(WARN_QW)) {
6421 for (; !isSPACE(*d) && len; --len, ++d) {
6423 Perl_warner(aTHX_ packWARN(WARN_QW),
6424 "Possible attempt to separate words with commas");
6427 else if (*d == '#') {
6428 Perl_warner(aTHX_ packWARN(WARN_QW),
6429 "Possible attempt to put comments in qw() list");
6435 for (; !isSPACE(*d) && len; --len, ++d)
6438 sv = newSVpvn_utf8(b, d-b, DO_UTF8(PL_lex_stuff));
6439 words = append_elem(OP_LIST, words,
6440 newSVOP(OP_CONST, 0, tokeq(sv)));
6444 start_force(PL_curforce);
6445 NEXTVAL_NEXTTOKE.opval = words;
6450 SvREFCNT_dec(PL_lex_stuff);
6451 PL_lex_stuff = NULL;
6457 s = scan_str(s,!!PL_madskills,FALSE);
6460 pl_yylval.ival = OP_STRINGIFY;
6461 if (SvIVX(PL_lex_stuff) == '\'')
6462 SvIV_set(PL_lex_stuff, 0); /* qq'$foo' should intepolate */
6463 TERM(sublex_start());
6466 s = scan_pat(s,OP_QR);
6467 TERM(sublex_start());
6470 s = scan_str(s,!!PL_madskills,FALSE);
6473 readpipe_override();
6474 TERM(sublex_start());
6482 s = force_version(s, FALSE);
6484 else if (*s != 'v' || !isDIGIT(s[1])
6485 || (s = force_version(s, TRUE), *s == 'v'))
6487 *PL_tokenbuf = '\0';
6488 s = force_word(s,WORD,TRUE,TRUE,FALSE);
6489 if (isIDFIRST_lazy_if(PL_tokenbuf,UTF))
6490 gv_stashpvn(PL_tokenbuf, strlen(PL_tokenbuf), GV_ADD);
6492 yyerror("<> should be quotes");
6494 if (orig_keyword == KEY_require) {
6502 PL_last_uni = PL_oldbufptr;
6503 PL_last_lop_op = OP_REQUIRE;
6505 return REPORT( (int)REQUIRE );
6511 s = force_word(s,WORD,TRUE,FALSE,FALSE);
6515 LOP(OP_RENAME,XTERM);
6524 LOP(OP_RINDEX,XTERM);
6533 UNIDOR(OP_READLINE);
6536 UNIDOR(OP_BACKTICK);
6545 LOP(OP_REVERSE,XTERM);
6548 UNIDOR(OP_READLINK);
6555 if (pl_yylval.opval)
6556 TERM(sublex_start());
6558 TOKEN(1); /* force error */
6561 checkcomma(s,PL_tokenbuf,"filehandle");
6571 LOP(OP_SELECT,XTERM);
6577 LOP(OP_SEMCTL,XTERM);
6580 LOP(OP_SEMGET,XTERM);
6583 LOP(OP_SEMOP,XTERM);
6589 LOP(OP_SETPGRP,XTERM);
6591 case KEY_setpriority:
6592 LOP(OP_SETPRIORITY,XTERM);
6594 case KEY_sethostent:
6600 case KEY_setservent:
6603 case KEY_setprotoent:
6613 LOP(OP_SEEKDIR,XTERM);
6615 case KEY_setsockopt:
6616 LOP(OP_SSOCKOPT,XTERM);
6622 LOP(OP_SHMCTL,XTERM);
6625 LOP(OP_SHMGET,XTERM);
6628 LOP(OP_SHMREAD,XTERM);
6631 LOP(OP_SHMWRITE,XTERM);
6634 LOP(OP_SHUTDOWN,XTERM);
6643 LOP(OP_SOCKET,XTERM);
6645 case KEY_socketpair:
6646 LOP(OP_SOCKPAIR,XTERM);
6649 checkcomma(s,PL_tokenbuf,"subroutine name");
6651 if (*s == ';' || *s == ')') /* probably a close */
6652 Perl_croak(aTHX_ "sort is now a reserved word");
6654 s = force_word(s,WORD,TRUE,TRUE,FALSE);
6658 LOP(OP_SPLIT,XTERM);
6661 LOP(OP_SPRINTF,XTERM);
6664 LOP(OP_SPLICE,XTERM);
6679 LOP(OP_SUBSTR,XTERM);
6685 char tmpbuf[sizeof PL_tokenbuf];
6686 SSize_t tboffset = 0;
6687 expectation attrful;
6688 bool have_name, have_proto;
6689 const int key = tmp;
6694 char *tstart = SvPVX(PL_linestr) + PL_realtokenstart;
6695 SV *subtoken = newSVpvn(tstart, s - tstart);
6699 s = SKIPSPACE2(s,tmpwhite);
6704 if (isIDFIRST_lazy_if(s,UTF) || *s == '\'' ||
6705 (*s == ':' && s[1] == ':'))
6708 SV *nametoke = NULL;
6712 attrful = XATTRBLOCK;
6713 /* remember buffer pos'n for later force_word */
6714 tboffset = s - PL_oldbufptr;
6715 d = scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
6718 nametoke = newSVpvn(s, d - s);
6720 if (memchr(tmpbuf, ':', len))
6721 sv_setpvn(PL_subname, tmpbuf, len);
6723 sv_setsv(PL_subname,PL_curstname);
6724 sv_catpvs(PL_subname,"::");
6725 sv_catpvn(PL_subname,tmpbuf,len);
6732 CURMAD('X', nametoke);
6733 CURMAD('_', tmpwhite);
6734 (void) force_word(PL_oldbufptr + tboffset, WORD,
6737 s = SKIPSPACE2(d,tmpwhite);
6744 Perl_croak(aTHX_ "Missing name in \"my sub\"");
6745 PL_expect = XTERMBLOCK;
6746 attrful = XATTRTERM;
6747 sv_setpvs(PL_subname,"?");
6751 if (key == KEY_format) {
6753 PL_lex_formbrack = PL_lex_brackets + 1;
6755 PL_thistoken = subtoken;
6759 (void) force_word(PL_oldbufptr + tboffset, WORD,
6765 /* Look for a prototype */
6768 bool bad_proto = FALSE;
6769 bool in_brackets = FALSE;
6770 char greedy_proto = ' ';
6771 bool proto_after_greedy_proto = FALSE;
6772 bool must_be_last = FALSE;
6773 bool underscore = FALSE;
6774 bool seen_underscore = FALSE;
6775 const bool warnsyntax = ckWARN(WARN_SYNTAX);
6777 s = scan_str(s,!!PL_madskills,FALSE);
6779 Perl_croak(aTHX_ "Prototype not terminated");
6780 /* strip spaces and check for bad characters */
6781 d = SvPVX(PL_lex_stuff);
6783 for (p = d; *p; ++p) {
6789 proto_after_greedy_proto = TRUE;
6790 if (!strchr("$@%*;[]&\\_", *p)) {
6802 else if ( *p == ']' ) {
6803 in_brackets = FALSE;
6805 else if ( (*p == '@' || *p == '%') &&
6806 ( tmp < 2 || d[tmp-2] != '\\' ) &&
6808 must_be_last = TRUE;
6811 else if ( *p == '_' ) {
6812 underscore = seen_underscore = TRUE;
6819 if (proto_after_greedy_proto)
6820 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
6821 "Prototype after '%c' for %"SVf" : %s",
6822 greedy_proto, SVfARG(PL_subname), d);
6824 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
6825 "Illegal character %sin prototype for %"SVf" : %s",
6826 seen_underscore ? "after '_' " : "",
6827 SVfARG(PL_subname), d);
6828 SvCUR_set(PL_lex_stuff, tmp);
6833 CURMAD('q', PL_thisopen);
6834 CURMAD('_', tmpwhite);
6835 CURMAD('=', PL_thisstuff);
6836 CURMAD('Q', PL_thisclose);
6837 NEXTVAL_NEXTTOKE.opval =
6838 (OP*)newSVOP(OP_CONST, 0, PL_lex_stuff);
6839 PL_lex_stuff = NULL;
6842 s = SKIPSPACE2(s,tmpwhite);
6850 if (*s == ':' && s[1] != ':')
6851 PL_expect = attrful;
6852 else if (*s != '{' && key == KEY_sub) {
6854 Perl_croak(aTHX_ "Illegal declaration of anonymous subroutine");
6856 Perl_croak(aTHX_ "Illegal declaration of subroutine %"SVf, SVfARG(PL_subname));
6863 curmad('^', newSVpvs(""));
6864 CURMAD('_', tmpwhite);
6868 PL_thistoken = subtoken;
6871 NEXTVAL_NEXTTOKE.opval =
6872 (OP*)newSVOP(OP_CONST, 0, PL_lex_stuff);
6873 PL_lex_stuff = NULL;
6879 sv_setpvs(PL_subname, "__ANON__");
6881 sv_setpvs(PL_subname, "__ANON__::__ANON__");
6885 (void) force_word(PL_oldbufptr + tboffset, WORD,
6894 LOP(OP_SYSTEM,XREF);
6897 LOP(OP_SYMLINK,XTERM);
6900 LOP(OP_SYSCALL,XTERM);
6903 LOP(OP_SYSOPEN,XTERM);
6906 LOP(OP_SYSSEEK,XTERM);
6909 LOP(OP_SYSREAD,XTERM);
6912 LOP(OP_SYSWRITE,XTERM);
6916 TERM(sublex_start());
6937 LOP(OP_TRUNCATE,XTERM);
6949 pl_yylval.ival = CopLINE(PL_curcop);
6953 pl_yylval.ival = CopLINE(PL_curcop);
6957 LOP(OP_UNLINK,XTERM);
6963 LOP(OP_UNPACK,XTERM);
6966 LOP(OP_UTIME,XTERM);
6972 LOP(OP_UNSHIFT,XTERM);
6975 s = tokenize_use(1, s);
6985 pl_yylval.ival = CopLINE(PL_curcop);
6989 pl_yylval.ival = CopLINE(PL_curcop);
6993 PL_hints |= HINT_BLOCK_SCOPE;
7000 LOP(OP_WAITPID,XTERM);
7009 ctl_l[0] = toCTRL('L');
7011 gv_fetchpvn_flags(ctl_l, 1, GV_ADD|GV_NOTQUAL, SVt_PV);
7014 /* Make sure $^L is defined */
7015 gv_fetchpvs("\f", GV_ADD|GV_NOTQUAL, SVt_PV);
7020 if (PL_expect == XOPERATOR)
7026 pl_yylval.ival = OP_XOR;
7031 TERM(sublex_start());
7036 #pragma segment Main
7040 S_pending_ident(pTHX)
7045 /* pit holds the identifier we read and pending_ident is reset */
7046 char pit = PL_pending_ident;
7047 const STRLEN tokenbuf_len = strlen(PL_tokenbuf);
7048 /* All routes through this function want to know if there is a colon. */
7049 const char *const has_colon = (const char*) memchr (PL_tokenbuf, ':', tokenbuf_len);
7050 PL_pending_ident = 0;
7052 /* PL_realtokenstart = realtokenend = PL_bufptr - SvPVX(PL_linestr); */
7053 DEBUG_T({ PerlIO_printf(Perl_debug_log,
7054 "### Pending identifier '%s'\n", PL_tokenbuf); });
7056 /* if we're in a my(), we can't allow dynamics here.
7057 $foo'bar has already been turned into $foo::bar, so
7058 just check for colons.
7060 if it's a legal name, the OP is a PADANY.
7063 if (PL_in_my == KEY_our) { /* "our" is merely analogous to "my" */
7065 yyerror(Perl_form(aTHX_ "No package name allowed for "
7066 "variable %s in \"our\"",
7068 tmp = allocmy(PL_tokenbuf);
7072 yyerror(Perl_form(aTHX_ PL_no_myglob,
7073 PL_in_my == KEY_my ? "my" : "state", PL_tokenbuf));
7075 pl_yylval.opval = newOP(OP_PADANY, 0);
7076 pl_yylval.opval->op_targ = allocmy(PL_tokenbuf);
7082 build the ops for accesses to a my() variable.
7084 Deny my($a) or my($b) in a sort block, *if* $a or $b is
7085 then used in a comparison. This catches most, but not
7086 all cases. For instance, it catches
7087 sort { my($a); $a <=> $b }
7089 sort { my($a); $a < $b ? -1 : $a == $b ? 0 : 1; }
7090 (although why you'd do that is anyone's guess).
7095 tmp = pad_findmy(PL_tokenbuf);
7096 if (tmp != NOT_IN_PAD) {
7097 /* might be an "our" variable" */
7098 if (PAD_COMPNAME_FLAGS_isOUR(tmp)) {
7099 /* build ops for a bareword */
7100 HV * const stash = PAD_COMPNAME_OURSTASH(tmp);
7101 HEK * const stashname = HvNAME_HEK(stash);
7102 SV * const sym = newSVhek(stashname);
7103 sv_catpvs(sym, "::");
7104 sv_catpvn(sym, PL_tokenbuf+1, tokenbuf_len - 1);
7105 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sym);
7106 pl_yylval.opval->op_private = OPpCONST_ENTERED;
7109 ? (GV_ADDMULTI | GV_ADDINEVAL)
7112 ((PL_tokenbuf[0] == '$') ? SVt_PV
7113 : (PL_tokenbuf[0] == '@') ? SVt_PVAV
7118 /* if it's a sort block and they're naming $a or $b */
7119 if (PL_last_lop_op == OP_SORT &&
7120 PL_tokenbuf[0] == '$' &&
7121 (PL_tokenbuf[1] == 'a' || PL_tokenbuf[1] == 'b')
7124 for (d = PL_in_eval ? PL_oldoldbufptr : PL_linestart;
7125 d < PL_bufend && *d != '\n';
7128 if (strnEQ(d,"<=>",3) || strnEQ(d,"cmp",3)) {
7129 Perl_croak(aTHX_ "Can't use \"my %s\" in sort comparison",
7135 pl_yylval.opval = newOP(OP_PADANY, 0);
7136 pl_yylval.opval->op_targ = tmp;
7142 Whine if they've said @foo in a doublequoted string,
7143 and @foo isn't a variable we can find in the symbol
7146 if (pit == '@' && PL_lex_state != LEX_NORMAL && !PL_lex_brackets) {
7147 GV *const gv = gv_fetchpvn_flags(PL_tokenbuf + 1, tokenbuf_len - 1, 0,
7149 if ((!gv || ((PL_tokenbuf[0] == '@') ? !GvAV(gv) : !GvHV(gv)))
7150 && ckWARN(WARN_AMBIGUOUS)
7151 /* DO NOT warn for @- and @+ */
7152 && !( PL_tokenbuf[2] == '\0' &&
7153 ( PL_tokenbuf[1] == '-' || PL_tokenbuf[1] == '+' ))
7156 /* Downgraded from fatal to warning 20000522 mjd */
7157 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
7158 "Possible unintended interpolation of %s in string",
7163 /* build ops for a bareword */
7164 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, newSVpvn(PL_tokenbuf + 1,
7166 pl_yylval.opval->op_private = OPpCONST_ENTERED;
7168 PL_tokenbuf + 1, tokenbuf_len - 1,
7169 /* If the identifier refers to a stash, don't autovivify it.
7170 * Change 24660 had the side effect of causing symbol table
7171 * hashes to always be defined, even if they were freshly
7172 * created and the only reference in the entire program was
7173 * the single statement with the defined %foo::bar:: test.
7174 * It appears that all code in the wild doing this actually
7175 * wants to know whether sub-packages have been loaded, so
7176 * by avoiding auto-vivifying symbol tables, we ensure that
7177 * defined %foo::bar:: continues to be false, and the existing
7178 * tests still give the expected answers, even though what
7179 * they're actually testing has now changed subtly.
7181 (*PL_tokenbuf == '%'
7182 && *(d = PL_tokenbuf + tokenbuf_len - 1) == ':'
7185 : PL_in_eval ? (GV_ADDMULTI | GV_ADDINEVAL) : GV_ADD),
7186 ((PL_tokenbuf[0] == '$') ? SVt_PV
7187 : (PL_tokenbuf[0] == '@') ? SVt_PVAV
7193 * The following code was generated by perl_keyword.pl.
7197 Perl_keyword (pTHX_ const char *name, I32 len, bool all_keywords)
7201 PERL_ARGS_ASSERT_KEYWORD;
7205 case 1: /* 5 tokens of length 1 */
7237 case 2: /* 18 tokens of length 2 */
7383 case 3: /* 29 tokens of length 3 */
7387 if (name[1] == 'N' &&
7450 if (name[1] == 'i' &&
7482 if (name[1] == 'o' &&
7491 if (name[1] == 'e' &&
7500 if (name[1] == 'n' &&
7509 if (name[1] == 'o' &&
7518 if (name[1] == 'a' &&
7527 if (name[1] == 'o' &&
7589 if (name[1] == 'e' &&
7603 return (all_keywords || FEATURE_IS_ENABLED("say") ? KEY_say : 0);
7629 if (name[1] == 'i' &&
7638 if (name[1] == 's' &&
7647 if (name[1] == 'e' &&
7656 if (name[1] == 'o' &&
7668 case 4: /* 41 tokens of length 4 */
7672 if (name[1] == 'O' &&
7682 if (name[1] == 'N' &&
7692 if (name[1] == 'i' &&
7702 if (name[1] == 'h' &&
7712 if (name[1] == 'u' &&
7725 if (name[2] == 'c' &&
7734 if (name[2] == 's' &&
7743 if (name[2] == 'a' &&
7779 if (name[1] == 'o' &&
7792 if (name[2] == 't' &&
7801 if (name[2] == 'o' &&
7810 if (name[2] == 't' &&
7819 if (name[2] == 'e' &&
7832 if (name[1] == 'o' &&
7845 if (name[2] == 'y' &&
7854 if (name[2] == 'l' &&
7870 if (name[2] == 's' &&
7879 if (name[2] == 'n' &&
7888 if (name[2] == 'c' &&
7901 if (name[1] == 'e' &&
7911 if (name[1] == 'p' &&
7924 if (name[2] == 'c' &&
7933 if (name[2] == 'p' &&
7942 if (name[2] == 's' &&
7958 if (name[2] == 'n' &&
8028 if (name[2] == 'r' &&
8037 if (name[2] == 'r' &&
8046 if (name[2] == 'a' &&
8062 if (name[2] == 'l' &&
8124 if (name[2] == 'e' &&
8127 return (all_keywords || FEATURE_IS_ENABLED("switch") ? KEY_when : 0);
8140 case 5: /* 39 tokens of length 5 */
8144 if (name[1] == 'E' &&
8155 if (name[1] == 'H' &&
8169 if (name[2] == 'a' &&
8179 if (name[2] == 'a' &&
8196 if (name[2] == 'e' &&
8206 if (name[2] == 'e' &&
8210 return (all_keywords || FEATURE_IS_ENABLED("switch") ? -KEY_break : 0);
8226 if (name[3] == 'i' &&
8235 if (name[3] == 'o' &&
8271 if (name[2] == 'o' &&
8281 if (name[2] == 'y' &&
8295 if (name[1] == 'l' &&
8309 if (name[2] == 'n' &&
8319 if (name[2] == 'o' &&
8333 if (name[1] == 'i' &&
8338 return (all_keywords || FEATURE_IS_ENABLED("switch") ? KEY_given : 0);
8347 if (name[2] == 'd' &&
8357 if (name[2] == 'c' &&
8374 if (name[2] == 'c' &&
8384 if (name[2] == 't' &&
8398 if (name[1] == 'k' &&
8409 if (name[1] == 'r' &&
8423 if (name[2] == 's' &&
8433 if (name[2] == 'd' &&
8450 if (name[2] == 'm' &&
8460 if (name[2] == 'i' &&
8470 if (name[2] == 'e' &&
8480 if (name[2] == 'l' &&
8490 if (name[2] == 'a' &&
8503 if (name[3] == 't' &&
8506 return (all_keywords || FEATURE_IS_ENABLED("state") ? KEY_state : 0);
8512 if (name[3] == 'd' &&
8529 if (name[1] == 'i' &&
8543 if (name[2] == 'a' &&
8556 if (name[3] == 'e' &&
8591 if (name[2] == 'i' &&
8608 if (name[2] == 'i' &&
8618 if (name[2] == 'i' &&
8635 case 6: /* 33 tokens of length 6 */
8639 if (name[1] == 'c' &&
8654 if (name[2] == 'l' &&
8665 if (name[2] == 'r' &&
8680 if (name[1] == 'e' &&
8695 if (name[2] == 's' &&
8700 if(ckWARN_d(WARN_SYNTAX))
8701 Perl_warner(aTHX_ packWARN(WARN_SYNTAX), "elseif should be elsif");
8707 if (name[2] == 'i' &&
8725 if (name[2] == 'l' &&
8736 if (name[2] == 'r' &&
8751 if (name[1] == 'm' &&
8766 if (name[2] == 'n' &&
8777 if (name[2] == 's' &&
8792 if (name[1] == 's' &&
8798 if (name[4] == 't' &&
8807 if (name[4] == 'e' &&
8816 if (name[4] == 'c' &&
8825 if (name[4] == 'n' &&
8841 if (name[1] == 'r' &&
8859 if (name[3] == 'a' &&
8869 if (name[3] == 'u' &&
8883 if (name[2] == 'n' &&
8901 if (name[2] == 'a' &&
8915 if (name[3] == 'e' &&
8928 if (name[4] == 't' &&
8937 if (name[4] == 'e' &&
8959 if (name[4] == 't' &&
8968 if (name[4] == 'e' &&
8984 if (name[2] == 'c' &&
8995 if (name[2] == 'l' &&
9006 if (name[2] == 'b' &&
9017 if (name[2] == 's' &&
9040 if (name[4] == 's' &&
9049 if (name[4] == 'n' &&
9062 if (name[3] == 'a' &&
9079 if (name[1] == 'a' &&
9094 case 7: /* 29 tokens of length 7 */
9098 if (name[1] == 'E' &&
9111 if (name[1] == '_' &&
9124 if (name[1] == 'i' &&
9131 return -KEY_binmode;
9137 if (name[1] == 'o' &&
9144 return -KEY_connect;
9153 if (name[2] == 'm' &&
9159 return -KEY_dbmopen;
9170 if (name[4] == 'u' &&
9174 return (all_keywords || FEATURE_IS_ENABLED("switch") ? KEY_default : 0);
9180 if (name[4] == 'n' &&
9201 if (name[1] == 'o' &&
9214 if (name[1] == 'e' &&
9221 if (name[5] == 'r' &&
9224 return -KEY_getpgrp;
9230 if (name[5] == 'i' &&
9233 return -KEY_getppid;
9246 if (name[1] == 'c' &&
9253 return -KEY_lcfirst;
9259 if (name[1] == 'p' &&
9266 return -KEY_opendir;
9272 if (name[1] == 'a' &&
9290 if (name[3] == 'd' &&
9295 return -KEY_readdir;
9301 if (name[3] == 'u' &&
9312 if (name[3] == 'e' &&
9317 return -KEY_reverse;
9336 if (name[3] == 'k' &&
9341 return -KEY_seekdir;
9347 if (name[3] == 'p' &&
9352 return -KEY_setpgrp;
9362 if (name[2] == 'm' &&
9368 return -KEY_shmread;
9374 if (name[2] == 'r' &&
9380 return -KEY_sprintf;
9389 if (name[3] == 'l' &&
9394 return -KEY_symlink;
9403 if (name[4] == 'a' &&
9407 return -KEY_syscall;
9413 if (name[4] == 'p' &&
9417 return -KEY_sysopen;
9423 if (name[4] == 'e' &&
9427 return -KEY_sysread;
9433 if (name[4] == 'e' &&
9437 return -KEY_sysseek;
9455 if (name[1] == 'e' &&
9462 return -KEY_telldir;
9471 if (name[2] == 'f' &&
9477 return -KEY_ucfirst;
9483 if (name[2] == 's' &&
9489 return -KEY_unshift;
9499 if (name[1] == 'a' &&
9506 return -KEY_waitpid;
9515 case 8: /* 26 tokens of length 8 */
9519 if (name[1] == 'U' &&
9527 return KEY_AUTOLOAD;
9538 if (name[3] == 'A' &&
9544 return KEY___DATA__;
9550 if (name[3] == 'I' &&
9556 return -KEY___FILE__;
9562 if (name[3] == 'I' &&
9568 return -KEY___LINE__;
9584 if (name[2] == 'o' &&
9591 return -KEY_closedir;
9597 if (name[2] == 'n' &&
9604 return -KEY_continue;
9614 if (name[1] == 'b' &&
9622 return -KEY_dbmclose;
9628 if (name[1] == 'n' &&
9634 if (name[4] == 'r' &&
9639 return -KEY_endgrent;
9645 if (name[4] == 'w' &&
9650 return -KEY_endpwent;
9663 if (name[1] == 'o' &&
9671 return -KEY_formline;
9677 if (name[1] == 'e' &&
9688 if (name[6] == 'n' &&
9691 return -KEY_getgrent;
9697 if (name[6] == 'i' &&
9700 return -KEY_getgrgid;
9706 if (name[6] == 'a' &&
9709 return -KEY_getgrnam;
9722 if (name[4] == 'o' &&
9727 return -KEY_getlogin;
9738 if (name[6] == 'n' &&
9741 return -KEY_getpwent;
9747 if (name[6] == 'a' &&
9750 return -KEY_getpwnam;
9756 if (name[6] == 'i' &&
9759 return -KEY_getpwuid;
9779 if (name[1] == 'e' &&
9786 if (name[5] == 'i' &&
9793 return -KEY_readline;
9798 return -KEY_readlink;
9809 if (name[5] == 'i' &&
9813 return -KEY_readpipe;
9834 if (name[4] == 'r' &&
9839 return -KEY_setgrent;
9845 if (name[4] == 'w' &&
9850 return -KEY_setpwent;
9866 if (name[3] == 'w' &&
9872 return -KEY_shmwrite;
9878 if (name[3] == 't' &&
9884 return -KEY_shutdown;
9894 if (name[2] == 's' &&
9901 return -KEY_syswrite;
9911 if (name[1] == 'r' &&
9919 return -KEY_truncate;
9928 case 9: /* 9 tokens of length 9 */
9932 if (name[1] == 'N' &&
9941 return KEY_UNITCHECK;
9947 if (name[1] == 'n' &&
9956 return -KEY_endnetent;
9962 if (name[1] == 'e' &&
9971 return -KEY_getnetent;
9977 if (name[1] == 'o' &&
9986 return -KEY_localtime;
9992 if (name[1] == 'r' &&
10001 return KEY_prototype;
10007 if (name[1] == 'u' &&
10016 return -KEY_quotemeta;
10022 if (name[1] == 'e' &&
10031 return -KEY_rewinddir;
10037 if (name[1] == 'e' &&
10046 return -KEY_setnetent;
10052 if (name[1] == 'a' &&
10061 return -KEY_wantarray;
10070 case 10: /* 9 tokens of length 10 */
10074 if (name[1] == 'n' &&
10080 if (name[4] == 'o' &&
10087 return -KEY_endhostent;
10093 if (name[4] == 'e' &&
10100 return -KEY_endservent;
10113 if (name[1] == 'e' &&
10119 if (name[4] == 'o' &&
10126 return -KEY_gethostent;
10135 if (name[5] == 'r' &&
10141 return -KEY_getservent;
10147 if (name[5] == 'c' &&
10153 return -KEY_getsockopt;
10173 if (name[2] == 't')
10178 if (name[4] == 'o' &&
10185 return -KEY_sethostent;
10194 if (name[5] == 'r' &&
10200 return -KEY_setservent;
10206 if (name[5] == 'c' &&
10212 return -KEY_setsockopt;
10229 if (name[2] == 'c' &&
10238 return -KEY_socketpair;
10251 case 11: /* 8 tokens of length 11 */
10255 if (name[1] == '_' &&
10265 { /* __PACKAGE__ */
10266 return -KEY___PACKAGE__;
10272 if (name[1] == 'n' &&
10282 { /* endprotoent */
10283 return -KEY_endprotoent;
10289 if (name[1] == 'e' &&
10298 if (name[5] == 'e' &&
10304 { /* getpeername */
10305 return -KEY_getpeername;
10314 if (name[6] == 'o' &&
10319 { /* getpriority */
10320 return -KEY_getpriority;
10326 if (name[6] == 't' &&
10331 { /* getprotoent */
10332 return -KEY_getprotoent;
10346 if (name[4] == 'o' &&
10353 { /* getsockname */
10354 return -KEY_getsockname;
10367 if (name[1] == 'e' &&
10375 if (name[6] == 'o' &&
10380 { /* setpriority */
10381 return -KEY_setpriority;
10387 if (name[6] == 't' &&
10392 { /* setprotoent */
10393 return -KEY_setprotoent;
10409 case 12: /* 2 tokens of length 12 */
10410 if (name[0] == 'g' &&
10422 if (name[9] == 'd' &&
10425 { /* getnetbyaddr */
10426 return -KEY_getnetbyaddr;
10432 if (name[9] == 'a' &&
10435 { /* getnetbyname */
10436 return -KEY_getnetbyname;
10448 case 13: /* 4 tokens of length 13 */
10449 if (name[0] == 'g' &&
10456 if (name[4] == 'o' &&
10465 if (name[10] == 'd' &&
10468 { /* gethostbyaddr */
10469 return -KEY_gethostbyaddr;
10475 if (name[10] == 'a' &&
10478 { /* gethostbyname */
10479 return -KEY_gethostbyname;
10492 if (name[4] == 'e' &&
10501 if (name[10] == 'a' &&
10504 { /* getservbyname */
10505 return -KEY_getservbyname;
10511 if (name[10] == 'o' &&
10514 { /* getservbyport */
10515 return -KEY_getservbyport;
10534 case 14: /* 1 tokens of length 14 */
10535 if (name[0] == 'g' &&
10549 { /* getprotobyname */
10550 return -KEY_getprotobyname;
10555 case 16: /* 1 tokens of length 16 */
10556 if (name[0] == 'g' &&
10572 { /* getprotobynumber */
10573 return -KEY_getprotobynumber;
10587 S_checkcomma(pTHX_ const char *s, const char *name, const char *what)
10591 PERL_ARGS_ASSERT_CHECKCOMMA;
10593 if (*s == ' ' && s[1] == '(') { /* XXX gotta be a better way */
10594 if (ckWARN(WARN_SYNTAX)) {
10597 for (w = s+2; *w && level; w++) {
10600 else if (*w == ')')
10603 while (isSPACE(*w))
10605 /* the list of chars below is for end of statements or
10606 * block / parens, boolean operators (&&, ||, //) and branch
10607 * constructs (or, and, if, until, unless, while, err, for).
10608 * Not a very solid hack... */
10609 if (!*w || !strchr(";&/|})]oaiuwef!=", *w))
10610 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
10611 "%s (...) interpreted as function",name);
10614 while (s < PL_bufend && isSPACE(*s))
10618 while (s < PL_bufend && isSPACE(*s))
10620 if (isIDFIRST_lazy_if(s,UTF)) {
10621 const char * const w = s++;
10622 while (isALNUM_lazy_if(s,UTF))
10624 while (s < PL_bufend && isSPACE(*s))
10628 if (keyword(w, s - w, 0))
10631 gv = gv_fetchpvn_flags(w, s - w, 0, SVt_PVCV);
10632 if (gv && GvCVu(gv))
10634 Perl_croak(aTHX_ "No comma allowed after %s", what);
10639 /* Either returns sv, or mortalizes sv and returns a new SV*.
10640 Best used as sv=new_constant(..., sv, ...).
10641 If s, pv are NULL, calls subroutine with one argument,
10642 and type is used with error messages only. */
10645 S_new_constant(pTHX_ const char *s, STRLEN len, const char *key, STRLEN keylen,
10646 SV *sv, SV *pv, const char *type, STRLEN typelen)
10649 HV * const table = GvHV(PL_hintgv); /* ^H */
10653 const char *why1 = "", *why2 = "", *why3 = "";
10655 PERL_ARGS_ASSERT_NEW_CONSTANT;
10657 if (!table || !(PL_hints & HINT_LOCALIZE_HH)) {
10660 why2 = (const char *)
10661 (strEQ(key,"charnames")
10662 ? "(possibly a missing \"use charnames ...\")"
10664 msg = Perl_newSVpvf(aTHX_ "Constant(%s) unknown: %s",
10665 (type ? type: "undef"), why2);
10667 /* This is convoluted and evil ("goto considered harmful")
10668 * but I do not understand the intricacies of all the different
10669 * failure modes of %^H in here. The goal here is to make
10670 * the most probable error message user-friendly. --jhi */
10675 msg = Perl_newSVpvf(aTHX_ "Constant(%s): %s%s%s",
10676 (type ? type: "undef"), why1, why2, why3);
10678 yyerror(SvPVX_const(msg));
10682 cvp = hv_fetch(table, key, keylen, FALSE);
10683 if (!cvp || !SvOK(*cvp)) {
10686 why3 = "} is not defined";
10689 sv_2mortal(sv); /* Parent created it permanently */
10692 pv = newSVpvn_flags(s, len, SVs_TEMP);
10694 typesv = newSVpvn_flags(type, typelen, SVs_TEMP);
10696 typesv = &PL_sv_undef;
10698 PUSHSTACKi(PERLSI_OVERLOAD);
10710 call_sv(cv, G_SCALAR | ( PL_in_eval ? 0 : G_EVAL));
10714 /* Check the eval first */
10715 if (!PL_in_eval && SvTRUE(ERRSV)) {
10716 sv_catpvs(ERRSV, "Propagated");
10717 yyerror(SvPV_nolen_const(ERRSV)); /* Duplicates the message inside eval */
10719 res = SvREFCNT_inc_simple(sv);
10723 SvREFCNT_inc_simple_void(res);
10732 why1 = "Call to &{$^H{";
10734 why3 = "}} did not return a defined value";
10742 /* Returns a NUL terminated string, with the length of the string written to
10746 S_scan_word(pTHX_ register char *s, char *dest, STRLEN destlen, int allow_package, STRLEN *slp)
10749 register char *d = dest;
10750 register char * const e = d + destlen - 3; /* two-character token, ending NUL */
10752 PERL_ARGS_ASSERT_SCAN_WORD;
10756 Perl_croak(aTHX_ ident_too_long);
10757 if (isALNUM(*s)) /* UTF handled below */
10759 else if (allow_package && (*s == '\'') && isIDFIRST_lazy_if(s+1,UTF)) {
10764 else if (allow_package && (s[0] == ':') && (s[1] == ':') && (s[2] != '$')) {
10768 else if (UTF && UTF8_IS_START(*s) && isALNUM_utf8((U8*)s)) {
10769 char *t = s + UTF8SKIP(s);
10771 while (UTF8_IS_CONTINUED(*t) && is_utf8_mark((U8*)t))
10775 Perl_croak(aTHX_ ident_too_long);
10776 Copy(s, d, len, char);
10789 S_scan_ident(pTHX_ register char *s, register const char *send, char *dest, STRLEN destlen, I32 ck_uni)
10792 char *bracket = NULL;
10794 register char *d = dest;
10795 register char * const e = d + destlen + 3; /* two-character token, ending NUL */
10797 PERL_ARGS_ASSERT_SCAN_IDENT;
10802 while (isDIGIT(*s)) {
10804 Perl_croak(aTHX_ ident_too_long);
10811 Perl_croak(aTHX_ ident_too_long);
10812 if (isALNUM(*s)) /* UTF handled below */
10814 else if (*s == '\'' && isIDFIRST_lazy_if(s+1,UTF)) {
10819 else if (*s == ':' && s[1] == ':') {
10823 else if (UTF && UTF8_IS_START(*s) && isALNUM_utf8((U8*)s)) {
10824 char *t = s + UTF8SKIP(s);
10825 while (UTF8_IS_CONTINUED(*t) && is_utf8_mark((U8*)t))
10827 if (d + (t - s) > e)
10828 Perl_croak(aTHX_ ident_too_long);
10829 Copy(s, d, t - s, char);
10840 if (PL_lex_state != LEX_NORMAL)
10841 PL_lex_state = LEX_INTERPENDMAYBE;
10844 if (*s == '$' && s[1] &&
10845 (isALNUM_lazy_if(s+1,UTF) || s[1] == '$' || s[1] == '{' || strnEQ(s+1,"::",2)) )
10858 if (*d == '^' && *s && isCONTROLVAR(*s)) {
10863 if (isSPACE(s[-1])) {
10865 const char ch = *s++;
10866 if (!SPACE_OR_TAB(ch)) {
10872 if (isIDFIRST_lazy_if(d,UTF)) {
10876 while ((end < send && isALNUM_lazy_if(end,UTF)) || *end == ':') {
10877 end += UTF8SKIP(end);
10878 while (end < send && UTF8_IS_CONTINUED(*end) && is_utf8_mark((U8*)end))
10879 end += UTF8SKIP(end);
10881 Copy(s, d, end - s, char);
10886 while ((isALNUM(*s) || *s == ':') && d < e)
10889 Perl_croak(aTHX_ ident_too_long);
10892 while (s < send && SPACE_OR_TAB(*s))
10894 if ((*s == '[' || (*s == '{' && strNE(dest, "sub")))) {
10895 if (ckWARN(WARN_AMBIGUOUS) && keyword(dest, d - dest, 0)) {
10896 const char * const brack =
10898 ((*s == '[') ? "[...]" : "{...}");
10899 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
10900 "Ambiguous use of %c{%s%s} resolved to %c%s%s",
10901 funny, dest, brack, funny, dest, brack);
10904 PL_lex_brackstack[PL_lex_brackets++] = (char)(XOPERATOR | XFAKEBRACK);
10908 /* Handle extended ${^Foo} variables
10909 * 1999-02-27 mjd-perl-patch@plover.com */
10910 else if (!isALNUM(*d) && !isPRINT(*d) /* isCTRL(d) */
10914 while (isALNUM(*s) && d < e) {
10918 Perl_croak(aTHX_ ident_too_long);
10923 if (PL_lex_state == LEX_INTERPNORMAL && !PL_lex_brackets) {
10924 PL_lex_state = LEX_INTERPEND;
10927 if (PL_lex_state == LEX_NORMAL) {
10928 if (ckWARN(WARN_AMBIGUOUS) &&
10929 (keyword(dest, d - dest, 0)
10930 || get_cvn_flags(dest, d - dest, 0)))
10934 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
10935 "Ambiguous use of %c{%s} resolved to %c%s",
10936 funny, dest, funny, dest);
10941 s = bracket; /* let the parser handle it */
10945 else if (PL_lex_state == LEX_INTERPNORMAL && !PL_lex_brackets && !intuit_more(s))
10946 PL_lex_state = LEX_INTERPEND;
10951 Perl_pmflag(pTHX_ U32* pmfl, int ch)
10953 PERL_ARGS_ASSERT_PMFLAG;
10955 PERL_UNUSED_CONTEXT;
10957 const char c = (char)ch;
10959 CASE_STD_PMMOD_FLAGS_PARSE_SET(pmfl);
10960 case GLOBAL_PAT_MOD: *pmfl |= PMf_GLOBAL; break;
10961 case CONTINUE_PAT_MOD: *pmfl |= PMf_CONTINUE; break;
10962 case ONCE_PAT_MOD: *pmfl |= PMf_KEEP; break;
10963 case KEEPCOPY_PAT_MOD: *pmfl |= PMf_KEEPCOPY; break;
10969 S_scan_pat(pTHX_ char *start, I32 type)
10973 char *s = scan_str(start,!!PL_madskills,FALSE);
10974 const char * const valid_flags =
10975 (const char *)((type == OP_QR) ? QR_PAT_MODS : M_PAT_MODS);
10980 PERL_ARGS_ASSERT_SCAN_PAT;
10983 const char * const delimiter = skipspace(start);
10987 ? "Search pattern not terminated or ternary operator parsed as search pattern"
10988 : "Search pattern not terminated" ));
10991 pm = (PMOP*)newPMOP(type, 0);
10992 if (PL_multi_open == '?') {
10993 /* This is the only point in the code that sets PMf_ONCE: */
10994 pm->op_pmflags |= PMf_ONCE;
10996 /* Hence it's safe to do this bit of PMOP book-keeping here, which
10997 allows us to restrict the list needed by reset to just the ??
10999 assert(type != OP_TRANS);
11001 MAGIC *mg = mg_find((const SV *)PL_curstash, PERL_MAGIC_symtab);
11004 mg = sv_magicext(MUTABLE_SV(PL_curstash), 0, PERL_MAGIC_symtab, 0, 0,
11007 elements = mg->mg_len / sizeof(PMOP**);
11008 Renewc(mg->mg_ptr, elements + 1, PMOP*, char);
11009 ((PMOP**)mg->mg_ptr) [elements++] = pm;
11010 mg->mg_len = elements * sizeof(PMOP**);
11011 PmopSTASH_set(pm,PL_curstash);
11017 while (*s && strchr(valid_flags, *s))
11018 pmflag(&pm->op_pmflags,*s++);
11020 if (PL_madskills && modstart != s) {
11021 SV* tmptoken = newSVpvn(modstart, s - modstart);
11022 append_madprops(newMADPROP('m', MAD_SV, tmptoken, 0), (OP*)pm, 0);
11025 /* issue a warning if /c is specified,but /g is not */
11026 if ((pm->op_pmflags & PMf_CONTINUE) && !(pm->op_pmflags & PMf_GLOBAL)
11027 && ckWARN(WARN_REGEXP))
11029 Perl_warner(aTHX_ packWARN(WARN_REGEXP),
11030 "Use of /c modifier is meaningless without /g" );
11033 PL_lex_op = (OP*)pm;
11034 pl_yylval.ival = OP_MATCH;
11039 S_scan_subst(pTHX_ char *start)
11050 PERL_ARGS_ASSERT_SCAN_SUBST;
11052 pl_yylval.ival = OP_NULL;
11054 s = scan_str(start,!!PL_madskills,FALSE);
11057 Perl_croak(aTHX_ "Substitution pattern not terminated");
11059 if (s[-1] == PL_multi_open)
11062 if (PL_madskills) {
11063 CURMAD('q', PL_thisopen);
11064 CURMAD('_', PL_thiswhite);
11065 CURMAD('E', PL_thisstuff);
11066 CURMAD('Q', PL_thisclose);
11067 PL_realtokenstart = s - SvPVX(PL_linestr);
11071 first_start = PL_multi_start;
11072 s = scan_str(s,!!PL_madskills,FALSE);
11074 if (PL_lex_stuff) {
11075 SvREFCNT_dec(PL_lex_stuff);
11076 PL_lex_stuff = NULL;
11078 Perl_croak(aTHX_ "Substitution replacement not terminated");
11080 PL_multi_start = first_start; /* so whole substitution is taken together */
11082 pm = (PMOP*)newPMOP(OP_SUBST, 0);
11085 if (PL_madskills) {
11086 CURMAD('z', PL_thisopen);
11087 CURMAD('R', PL_thisstuff);
11088 CURMAD('Z', PL_thisclose);
11094 if (*s == EXEC_PAT_MOD) {
11098 else if (strchr(S_PAT_MODS, *s))
11099 pmflag(&pm->op_pmflags,*s++);
11105 if (PL_madskills) {
11107 curmad('m', newSVpvn(modstart, s - modstart));
11108 append_madprops(PL_thismad, (OP*)pm, 0);
11112 if ((pm->op_pmflags & PMf_CONTINUE) && ckWARN(WARN_REGEXP)) {
11113 Perl_warner(aTHX_ packWARN(WARN_REGEXP), "Use of /c modifier is meaningless in s///" );
11117 SV * const repl = newSVpvs("");
11119 PL_sublex_info.super_bufptr = s;
11120 PL_sublex_info.super_bufend = PL_bufend;
11122 pm->op_pmflags |= PMf_EVAL;
11125 sv_catpvs(repl, "eval ");
11127 sv_catpvs(repl, "do ");
11129 sv_catpvs(repl, "{");
11130 sv_catsv(repl, PL_lex_repl);
11131 if (strchr(SvPVX(PL_lex_repl), '#'))
11132 sv_catpvs(repl, "\n");
11133 sv_catpvs(repl, "}");
11135 SvREFCNT_dec(PL_lex_repl);
11136 PL_lex_repl = repl;
11139 PL_lex_op = (OP*)pm;
11140 pl_yylval.ival = OP_SUBST;
11145 S_scan_trans(pTHX_ char *start)
11158 PERL_ARGS_ASSERT_SCAN_TRANS;
11160 pl_yylval.ival = OP_NULL;
11162 s = scan_str(start,!!PL_madskills,FALSE);
11164 Perl_croak(aTHX_ "Transliteration pattern not terminated");
11166 if (s[-1] == PL_multi_open)
11169 if (PL_madskills) {
11170 CURMAD('q', PL_thisopen);
11171 CURMAD('_', PL_thiswhite);
11172 CURMAD('E', PL_thisstuff);
11173 CURMAD('Q', PL_thisclose);
11174 PL_realtokenstart = s - SvPVX(PL_linestr);
11178 s = scan_str(s,!!PL_madskills,FALSE);
11180 if (PL_lex_stuff) {
11181 SvREFCNT_dec(PL_lex_stuff);
11182 PL_lex_stuff = NULL;
11184 Perl_croak(aTHX_ "Transliteration replacement not terminated");
11186 if (PL_madskills) {
11187 CURMAD('z', PL_thisopen);
11188 CURMAD('R', PL_thisstuff);
11189 CURMAD('Z', PL_thisclose);
11192 complement = del = squash = 0;
11199 complement = OPpTRANS_COMPLEMENT;
11202 del = OPpTRANS_DELETE;
11205 squash = OPpTRANS_SQUASH;
11214 tbl = (short *)PerlMemShared_calloc(complement&&!del?258:256, sizeof(short));
11215 o = newPVOP(OP_TRANS, 0, (char*)tbl);
11216 o->op_private &= ~OPpTRANS_ALL;
11217 o->op_private |= del|squash|complement|
11218 (DO_UTF8(PL_lex_stuff)? OPpTRANS_FROM_UTF : 0)|
11219 (DO_UTF8(PL_lex_repl) ? OPpTRANS_TO_UTF : 0);
11222 pl_yylval.ival = OP_TRANS;
11225 if (PL_madskills) {
11227 curmad('m', newSVpvn(modstart, s - modstart));
11228 append_madprops(PL_thismad, o, 0);
11237 S_scan_heredoc(pTHX_ register char *s)
11241 I32 op_type = OP_SCALAR;
11245 const char *found_newline;
11249 const int outer = (PL_rsfp && !(PL_lex_inwhat == OP_SCALAR));
11251 I32 stuffstart = s - SvPVX(PL_linestr);
11254 PL_realtokenstart = -1;
11257 PERL_ARGS_ASSERT_SCAN_HEREDOC;
11261 e = PL_tokenbuf + sizeof PL_tokenbuf - 1;
11265 while (SPACE_OR_TAB(*peek))
11267 if (*peek == '`' || *peek == '\'' || *peek =='"') {
11270 s = delimcpy(d, e, s, PL_bufend, term, &len);
11280 if (!isALNUM_lazy_if(s,UTF))
11281 deprecate_old("bare << to mean <<\"\"");
11282 for (; isALNUM_lazy_if(s,UTF); s++) {
11287 if (d >= PL_tokenbuf + sizeof PL_tokenbuf - 1)
11288 Perl_croak(aTHX_ "Delimiter for here document is too long");
11291 len = d - PL_tokenbuf;
11294 if (PL_madskills) {
11295 tstart = PL_tokenbuf + !outer;
11296 PL_thisclose = newSVpvn(tstart, len - !outer);
11297 tstart = SvPVX(PL_linestr) + stuffstart;
11298 PL_thisopen = newSVpvn(tstart, s - tstart);
11299 stuffstart = s - SvPVX(PL_linestr);
11302 #ifndef PERL_STRICT_CR
11303 d = strchr(s, '\r');
11305 char * const olds = s;
11307 while (s < PL_bufend) {
11313 else if (*s == '\n' && s[1] == '\r') { /* \015\013 on a mac? */
11322 SvCUR_set(PL_linestr, PL_bufend - SvPVX_const(PL_linestr));
11329 if ( outer || !(found_newline = (char*)memchr((void*)s, '\n', PL_bufend - s)) ) {
11330 herewas = newSVpvn(s,PL_bufend-s);
11334 herewas = newSVpvn(s-1,found_newline-s+1);
11337 herewas = newSVpvn(s,found_newline-s);
11341 if (PL_madskills) {
11342 tstart = SvPVX(PL_linestr) + stuffstart;
11344 sv_catpvn(PL_thisstuff, tstart, s - tstart);
11346 PL_thisstuff = newSVpvn(tstart, s - tstart);
11349 s += SvCUR(herewas);
11352 stuffstart = s - SvPVX(PL_linestr);
11358 tmpstr = newSV_type(SVt_PVIV);
11359 SvGROW(tmpstr, 80);
11360 if (term == '\'') {
11361 op_type = OP_CONST;
11362 SvIV_set(tmpstr, -1);
11364 else if (term == '`') {
11365 op_type = OP_BACKTICK;
11366 SvIV_set(tmpstr, '\\');
11370 PL_multi_start = CopLINE(PL_curcop);
11371 PL_multi_open = PL_multi_close = '<';
11372 term = *PL_tokenbuf;
11373 if (PL_lex_inwhat == OP_SUBST && PL_in_eval && !PL_rsfp) {
11374 char * const bufptr = PL_sublex_info.super_bufptr;
11375 char * const bufend = PL_sublex_info.super_bufend;
11376 char * const olds = s - SvCUR(herewas);
11377 s = strchr(bufptr, '\n');
11381 while (s < bufend &&
11382 (*s != term || memNE(s,PL_tokenbuf,len)) ) {
11384 CopLINE_inc(PL_curcop);
11387 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
11388 missingterm(PL_tokenbuf);
11390 sv_setpvn(herewas,bufptr,d-bufptr+1);
11391 sv_setpvn(tmpstr,d+1,s-d);
11393 sv_catpvn(herewas,s,bufend-s);
11394 Copy(SvPVX_const(herewas),bufptr,SvCUR(herewas) + 1,char);
11401 while (s < PL_bufend &&
11402 (*s != term || memNE(s,PL_tokenbuf,len)) ) {
11404 CopLINE_inc(PL_curcop);
11406 if (s >= PL_bufend) {
11407 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
11408 missingterm(PL_tokenbuf);
11410 sv_setpvn(tmpstr,d+1,s-d);
11412 if (PL_madskills) {
11414 sv_catpvn(PL_thisstuff, d + 1, s - d);
11416 PL_thisstuff = newSVpvn(d + 1, s - d);
11417 stuffstart = s - SvPVX(PL_linestr);
11421 CopLINE_inc(PL_curcop); /* the preceding stmt passes a newline */
11423 sv_catpvn(herewas,s,PL_bufend-s);
11424 sv_setsv(PL_linestr,herewas);
11425 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = s = PL_linestart = SvPVX(PL_linestr);
11426 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
11427 PL_last_lop = PL_last_uni = NULL;
11430 sv_setpvs(tmpstr,""); /* avoid "uninitialized" warning */
11431 while (s >= PL_bufend) { /* multiple line string? */
11433 if (PL_madskills) {
11434 tstart = SvPVX(PL_linestr) + stuffstart;
11436 sv_catpvn(PL_thisstuff, tstart, PL_bufend - tstart);
11438 PL_thisstuff = newSVpvn(tstart, PL_bufend - tstart);
11442 !(PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = filter_gets(PL_linestr, PL_rsfp, 0))) {
11443 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
11444 missingterm(PL_tokenbuf);
11447 stuffstart = s - SvPVX(PL_linestr);
11449 CopLINE_inc(PL_curcop);
11450 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
11451 PL_last_lop = PL_last_uni = NULL;
11452 #ifndef PERL_STRICT_CR
11453 if (PL_bufend - PL_linestart >= 2) {
11454 if ((PL_bufend[-2] == '\r' && PL_bufend[-1] == '\n') ||
11455 (PL_bufend[-2] == '\n' && PL_bufend[-1] == '\r'))
11457 PL_bufend[-2] = '\n';
11459 SvCUR_set(PL_linestr, PL_bufend - SvPVX_const(PL_linestr));
11461 else if (PL_bufend[-1] == '\r')
11462 PL_bufend[-1] = '\n';
11464 else if (PL_bufend - PL_linestart == 1 && PL_bufend[-1] == '\r')
11465 PL_bufend[-1] = '\n';
11467 if ((PERLDB_LINE || PERLDB_SAVESRC) && PL_curstash != PL_debstash)
11468 update_debugger_info(PL_linestr, NULL, 0);
11469 if (*s == term && memEQ(s,PL_tokenbuf,len)) {
11470 STRLEN off = PL_bufend - 1 - SvPVX_const(PL_linestr);
11471 *(SvPVX(PL_linestr) + off ) = ' ';
11472 sv_catsv(PL_linestr,herewas);
11473 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
11474 s = SvPVX(PL_linestr) + off; /* In case PV of PL_linestr moved. */
11478 sv_catsv(tmpstr,PL_linestr);
11483 PL_multi_end = CopLINE(PL_curcop);
11484 if (SvCUR(tmpstr) + 5 < SvLEN(tmpstr)) {
11485 SvPV_shrink_to_cur(tmpstr);
11487 SvREFCNT_dec(herewas);
11489 if (UTF && is_utf8_string((U8*)SvPVX_const(tmpstr), SvCUR(tmpstr)))
11491 else if (PL_encoding)
11492 sv_recode_to_utf8(tmpstr, PL_encoding);
11494 PL_lex_stuff = tmpstr;
11495 pl_yylval.ival = op_type;
11499 /* scan_inputsymbol
11500 takes: current position in input buffer
11501 returns: new position in input buffer
11502 side-effects: pl_yylval and lex_op are set.
11507 <FH> read from filehandle
11508 <pkg::FH> read from package qualified filehandle
11509 <pkg'FH> read from package qualified filehandle
11510 <$fh> read from filehandle in $fh
11511 <*.h> filename glob
11516 S_scan_inputsymbol(pTHX_ char *start)
11519 register char *s = start; /* current position in buffer */
11522 char *d = PL_tokenbuf; /* start of temp holding space */
11523 const char * const e = PL_tokenbuf + sizeof PL_tokenbuf; /* end of temp holding space */
11525 PERL_ARGS_ASSERT_SCAN_INPUTSYMBOL;
11527 end = strchr(s, '\n');
11530 s = delimcpy(d, e, s + 1, end, '>', &len); /* extract until > */
11532 /* die if we didn't have space for the contents of the <>,
11533 or if it didn't end, or if we see a newline
11536 if (len >= (I32)sizeof PL_tokenbuf)
11537 Perl_croak(aTHX_ "Excessively long <> operator");
11539 Perl_croak(aTHX_ "Unterminated <> operator");
11544 Remember, only scalar variables are interpreted as filehandles by
11545 this code. Anything more complex (e.g., <$fh{$num}>) will be
11546 treated as a glob() call.
11547 This code makes use of the fact that except for the $ at the front,
11548 a scalar variable and a filehandle look the same.
11550 if (*d == '$' && d[1]) d++;
11552 /* allow <Pkg'VALUE> or <Pkg::VALUE> */
11553 while (*d && (isALNUM_lazy_if(d,UTF) || *d == '\'' || *d == ':'))
11556 /* If we've tried to read what we allow filehandles to look like, and
11557 there's still text left, then it must be a glob() and not a getline.
11558 Use scan_str to pull out the stuff between the <> and treat it
11559 as nothing more than a string.
11562 if (d - PL_tokenbuf != len) {
11563 pl_yylval.ival = OP_GLOB;
11564 s = scan_str(start,!!PL_madskills,FALSE);
11566 Perl_croak(aTHX_ "Glob not terminated");
11570 bool readline_overriden = FALSE;
11573 /* we're in a filehandle read situation */
11576 /* turn <> into <ARGV> */
11578 Copy("ARGV",d,5,char);
11580 /* Check whether readline() is overriden */
11581 gv_readline = gv_fetchpvs("readline", GV_NOTQUAL, SVt_PVCV);
11583 && GvCVu(gv_readline) && GvIMPORTED_CV(gv_readline))
11585 ((gvp = (GV**)hv_fetchs(PL_globalstash, "readline", FALSE))
11586 && (gv_readline = *gvp) && isGV_with_GP(gv_readline)
11587 && GvCVu(gv_readline) && GvIMPORTED_CV(gv_readline)))
11588 readline_overriden = TRUE;
11590 /* if <$fh>, create the ops to turn the variable into a
11594 /* try to find it in the pad for this block, otherwise find
11595 add symbol table ops
11597 const PADOFFSET tmp = pad_findmy(d);
11598 if (tmp != NOT_IN_PAD) {
11599 if (PAD_COMPNAME_FLAGS_isOUR(tmp)) {
11600 HV * const stash = PAD_COMPNAME_OURSTASH(tmp);
11601 HEK * const stashname = HvNAME_HEK(stash);
11602 SV * const sym = sv_2mortal(newSVhek(stashname));
11603 sv_catpvs(sym, "::");
11604 sv_catpv(sym, d+1);
11609 OP * const o = newOP(OP_PADSV, 0);
11611 PL_lex_op = readline_overriden
11612 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
11613 append_elem(OP_LIST, o,
11614 newCVREF(0, newGVOP(OP_GV,0,gv_readline))))
11615 : (OP*)newUNOP(OP_READLINE, 0, o);
11624 ? (GV_ADDMULTI | GV_ADDINEVAL)
11627 PL_lex_op = readline_overriden
11628 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
11629 append_elem(OP_LIST,
11630 newUNOP(OP_RV2SV, 0, newGVOP(OP_GV, 0, gv)),
11631 newCVREF(0, newGVOP(OP_GV, 0, gv_readline))))
11632 : (OP*)newUNOP(OP_READLINE, 0,
11633 newUNOP(OP_RV2SV, 0,
11634 newGVOP(OP_GV, 0, gv)));
11636 if (!readline_overriden)
11637 PL_lex_op->op_flags |= OPf_SPECIAL;
11638 /* we created the ops in PL_lex_op, so make pl_yylval.ival a null op */
11639 pl_yylval.ival = OP_NULL;
11642 /* If it's none of the above, it must be a literal filehandle
11643 (<Foo::BAR> or <FOO>) so build a simple readline OP */
11645 GV * const gv = gv_fetchpv(d, GV_ADD, SVt_PVIO);
11646 PL_lex_op = readline_overriden
11647 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
11648 append_elem(OP_LIST,
11649 newGVOP(OP_GV, 0, gv),
11650 newCVREF(0, newGVOP(OP_GV, 0, gv_readline))))
11651 : (OP*)newUNOP(OP_READLINE, 0, newGVOP(OP_GV, 0, gv));
11652 pl_yylval.ival = OP_NULL;
11661 takes: start position in buffer
11662 keep_quoted preserve \ on the embedded delimiter(s)
11663 keep_delims preserve the delimiters around the string
11664 returns: position to continue reading from buffer
11665 side-effects: multi_start, multi_close, lex_repl or lex_stuff, and
11666 updates the read buffer.
11668 This subroutine pulls a string out of the input. It is called for:
11669 q single quotes q(literal text)
11670 ' single quotes 'literal text'
11671 qq double quotes qq(interpolate $here please)
11672 " double quotes "interpolate $here please"
11673 qx backticks qx(/bin/ls -l)
11674 ` backticks `/bin/ls -l`
11675 qw quote words @EXPORT_OK = qw( func() $spam )
11676 m// regexp match m/this/
11677 s/// regexp substitute s/this/that/
11678 tr/// string transliterate tr/this/that/
11679 y/// string transliterate y/this/that/
11680 ($*@) sub prototypes sub foo ($)
11681 (stuff) sub attr parameters sub foo : attr(stuff)
11682 <> readline or globs <FOO>, <>, <$fh>, or <*.c>
11684 In most of these cases (all but <>, patterns and transliterate)
11685 yylex() calls scan_str(). m// makes yylex() call scan_pat() which
11686 calls scan_str(). s/// makes yylex() call scan_subst() which calls
11687 scan_str(). tr/// and y/// make yylex() call scan_trans() which
11690 It skips whitespace before the string starts, and treats the first
11691 character as the delimiter. If the delimiter is one of ([{< then
11692 the corresponding "close" character )]}> is used as the closing
11693 delimiter. It allows quoting of delimiters, and if the string has
11694 balanced delimiters ([{<>}]) it allows nesting.
11696 On success, the SV with the resulting string is put into lex_stuff or,
11697 if that is already non-NULL, into lex_repl. The second case occurs only
11698 when parsing the RHS of the special constructs s/// and tr/// (y///).
11699 For convenience, the terminating delimiter character is stuffed into
11704 S_scan_str(pTHX_ char *start, int keep_quoted, int keep_delims)
11707 SV *sv; /* scalar value: string */
11708 const char *tmps; /* temp string, used for delimiter matching */
11709 register char *s = start; /* current position in the buffer */
11710 register char term; /* terminating character */
11711 register char *to; /* current position in the sv's data */
11712 I32 brackets = 1; /* bracket nesting level */
11713 bool has_utf8 = FALSE; /* is there any utf8 content? */
11714 I32 termcode; /* terminating char. code */
11715 U8 termstr[UTF8_MAXBYTES]; /* terminating string */
11716 STRLEN termlen; /* length of terminating string */
11717 int last_off = 0; /* last position for nesting bracket */
11723 PERL_ARGS_ASSERT_SCAN_STR;
11725 /* skip space before the delimiter */
11731 if (PL_realtokenstart >= 0) {
11732 stuffstart = PL_realtokenstart;
11733 PL_realtokenstart = -1;
11736 stuffstart = start - SvPVX(PL_linestr);
11738 /* mark where we are, in case we need to report errors */
11741 /* after skipping whitespace, the next character is the terminator */
11744 termcode = termstr[0] = term;
11748 termcode = utf8_to_uvchr((U8*)s, &termlen);
11749 Copy(s, termstr, termlen, U8);
11750 if (!UTF8_IS_INVARIANT(term))
11754 /* mark where we are */
11755 PL_multi_start = CopLINE(PL_curcop);
11756 PL_multi_open = term;
11758 /* find corresponding closing delimiter */
11759 if (term && (tmps = strchr("([{< )]}> )]}>",term)))
11760 termcode = termstr[0] = term = tmps[5];
11762 PL_multi_close = term;
11764 /* create a new SV to hold the contents. 79 is the SV's initial length.
11765 What a random number. */
11766 sv = newSV_type(SVt_PVIV);
11768 SvIV_set(sv, termcode);
11769 (void)SvPOK_only(sv); /* validate pointer */
11771 /* move past delimiter and try to read a complete string */
11773 sv_catpvn(sv, s, termlen);
11776 tstart = SvPVX(PL_linestr) + stuffstart;
11777 if (!PL_thisopen && !keep_delims) {
11778 PL_thisopen = newSVpvn(tstart, s - tstart);
11779 stuffstart = s - SvPVX(PL_linestr);
11783 if (PL_encoding && !UTF) {
11787 int offset = s - SvPVX_const(PL_linestr);
11788 const bool found = sv_cat_decode(sv, PL_encoding, PL_linestr,
11789 &offset, (char*)termstr, termlen);
11790 const char * const ns = SvPVX_const(PL_linestr) + offset;
11791 char * const svlast = SvEND(sv) - 1;
11793 for (; s < ns; s++) {
11794 if (*s == '\n' && !PL_rsfp)
11795 CopLINE_inc(PL_curcop);
11798 goto read_more_line;
11800 /* handle quoted delimiters */
11801 if (SvCUR(sv) > 1 && *(svlast-1) == '\\') {
11803 for (t = svlast-2; t >= SvPVX_const(sv) && *t == '\\';)
11805 if ((svlast-1 - t) % 2) {
11806 if (!keep_quoted) {
11807 *(svlast-1) = term;
11809 SvCUR_set(sv, SvCUR(sv) - 1);
11814 if (PL_multi_open == PL_multi_close) {
11820 for (t = w = SvPVX(sv)+last_off; t < svlast; w++, t++) {
11821 /* At here, all closes are "was quoted" one,
11822 so we don't check PL_multi_close. */
11824 if (!keep_quoted && *(t+1) == PL_multi_open)
11829 else if (*t == PL_multi_open)
11837 SvCUR_set(sv, w - SvPVX_const(sv));
11839 last_off = w - SvPVX(sv);
11840 if (--brackets <= 0)
11845 if (!keep_delims) {
11846 SvCUR_set(sv, SvCUR(sv) - 1);
11852 /* extend sv if need be */
11853 SvGROW(sv, SvCUR(sv) + (PL_bufend - s) + 1);
11854 /* set 'to' to the next character in the sv's string */
11855 to = SvPVX(sv)+SvCUR(sv);
11857 /* if open delimiter is the close delimiter read unbridle */
11858 if (PL_multi_open == PL_multi_close) {
11859 for (; s < PL_bufend; s++,to++) {
11860 /* embedded newlines increment the current line number */
11861 if (*s == '\n' && !PL_rsfp)
11862 CopLINE_inc(PL_curcop);
11863 /* handle quoted delimiters */
11864 if (*s == '\\' && s+1 < PL_bufend && term != '\\') {
11865 if (!keep_quoted && s[1] == term)
11867 /* any other quotes are simply copied straight through */
11871 /* terminate when run out of buffer (the for() condition), or
11872 have found the terminator */
11873 else if (*s == term) {
11876 if (s+termlen <= PL_bufend && memEQ(s, (char*)termstr, termlen))
11879 else if (!has_utf8 && !UTF8_IS_INVARIANT((U8)*s) && UTF)
11885 /* if the terminator isn't the same as the start character (e.g.,
11886 matched brackets), we have to allow more in the quoting, and
11887 be prepared for nested brackets.
11890 /* read until we run out of string, or we find the terminator */
11891 for (; s < PL_bufend; s++,to++) {
11892 /* embedded newlines increment the line count */
11893 if (*s == '\n' && !PL_rsfp)
11894 CopLINE_inc(PL_curcop);
11895 /* backslashes can escape the open or closing characters */
11896 if (*s == '\\' && s+1 < PL_bufend) {
11897 if (!keep_quoted &&
11898 ((s[1] == PL_multi_open) || (s[1] == PL_multi_close)))
11903 /* allow nested opens and closes */
11904 else if (*s == PL_multi_close && --brackets <= 0)
11906 else if (*s == PL_multi_open)
11908 else if (!has_utf8 && !UTF8_IS_INVARIANT((U8)*s) && UTF)
11913 /* terminate the copied string and update the sv's end-of-string */
11915 SvCUR_set(sv, to - SvPVX_const(sv));
11918 * this next chunk reads more into the buffer if we're not done yet
11922 break; /* handle case where we are done yet :-) */
11924 #ifndef PERL_STRICT_CR
11925 if (to - SvPVX_const(sv) >= 2) {
11926 if ((to[-2] == '\r' && to[-1] == '\n') ||
11927 (to[-2] == '\n' && to[-1] == '\r'))
11931 SvCUR_set(sv, to - SvPVX_const(sv));
11933 else if (to[-1] == '\r')
11936 else if (to - SvPVX_const(sv) == 1 && to[-1] == '\r')
11941 /* if we're out of file, or a read fails, bail and reset the current
11942 line marker so we can report where the unterminated string began
11945 if (PL_madskills) {
11946 char * const tstart = SvPVX(PL_linestr) + stuffstart;
11948 sv_catpvn(PL_thisstuff, tstart, PL_bufend - tstart);
11950 PL_thisstuff = newSVpvn(tstart, PL_bufend - tstart);
11954 !(PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = filter_gets(PL_linestr, PL_rsfp, 0))) {
11956 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
11962 /* we read a line, so increment our line counter */
11963 CopLINE_inc(PL_curcop);
11965 /* update debugger info */
11966 if ((PERLDB_LINE || PERLDB_SAVESRC) && PL_curstash != PL_debstash)
11967 update_debugger_info(PL_linestr, NULL, 0);
11969 /* having changed the buffer, we must update PL_bufend */
11970 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
11971 PL_last_lop = PL_last_uni = NULL;
11974 /* at this point, we have successfully read the delimited string */
11976 if (!PL_encoding || UTF) {
11978 if (PL_madskills) {
11979 char * const tstart = SvPVX(PL_linestr) + stuffstart;
11980 const int len = s - tstart;
11982 sv_catpvn(PL_thisstuff, tstart, len);
11984 PL_thisstuff = newSVpvn(tstart, len);
11985 if (!PL_thisclose && !keep_delims)
11986 PL_thisclose = newSVpvn(s,termlen);
11991 sv_catpvn(sv, s, termlen);
11996 if (PL_madskills) {
11997 char * const tstart = SvPVX(PL_linestr) + stuffstart;
11998 const int len = s - tstart - termlen;
12000 sv_catpvn(PL_thisstuff, tstart, len);
12002 PL_thisstuff = newSVpvn(tstart, len);
12003 if (!PL_thisclose && !keep_delims)
12004 PL_thisclose = newSVpvn(s - termlen,termlen);
12008 if (has_utf8 || PL_encoding)
12011 PL_multi_end = CopLINE(PL_curcop);
12013 /* if we allocated too much space, give some back */
12014 if (SvCUR(sv) + 5 < SvLEN(sv)) {
12015 SvLEN_set(sv, SvCUR(sv) + 1);
12016 SvPV_renew(sv, SvLEN(sv));
12019 /* decide whether this is the first or second quoted string we've read
12032 takes: pointer to position in buffer
12033 returns: pointer to new position in buffer
12034 side-effects: builds ops for the constant in pl_yylval.op
12036 Read a number in any of the formats that Perl accepts:
12038 \d(_?\d)*(\.(\d(_?\d)*)?)?[Ee][\+\-]?(\d(_?\d)*) 12 12.34 12.
12039 \.\d(_?\d)*[Ee][\+\-]?(\d(_?\d)*) .34
12042 0x[0-9A-Fa-f](_?[0-9A-Fa-f])*
12044 Like most scan_ routines, it uses the PL_tokenbuf buffer to hold the
12047 If it reads a number without a decimal point or an exponent, it will
12048 try converting the number to an integer and see if it can do so
12049 without loss of precision.
12053 Perl_scan_num(pTHX_ const char *start, YYSTYPE* lvalp)
12056 register const char *s = start; /* current position in buffer */
12057 register char *d; /* destination in temp buffer */
12058 register char *e; /* end of temp buffer */
12059 NV nv; /* number read, as a double */
12060 SV *sv = NULL; /* place to put the converted number */
12061 bool floatit; /* boolean: int or float? */
12062 const char *lastub = NULL; /* position of last underbar */
12063 static char const number_too_long[] = "Number too long";
12065 PERL_ARGS_ASSERT_SCAN_NUM;
12067 /* We use the first character to decide what type of number this is */
12071 Perl_croak(aTHX_ "panic: scan_num");
12073 /* if it starts with a 0, it could be an octal number, a decimal in
12074 0.13 disguise, or a hexadecimal number, or a binary number. */
12078 u holds the "number so far"
12079 shift the power of 2 of the base
12080 (hex == 4, octal == 3, binary == 1)
12081 overflowed was the number more than we can hold?
12083 Shift is used when we add a digit. It also serves as an "are
12084 we in octal/hex/binary?" indicator to disallow hex characters
12085 when in octal mode.
12090 bool overflowed = FALSE;
12091 bool just_zero = TRUE; /* just plain 0 or binary number? */
12092 static const NV nvshift[5] = { 1.0, 2.0, 4.0, 8.0, 16.0 };
12093 static const char* const bases[5] =
12094 { "", "binary", "", "octal", "hexadecimal" };
12095 static const char* const Bases[5] =
12096 { "", "Binary", "", "Octal", "Hexadecimal" };
12097 static const char* const maxima[5] =
12099 "0b11111111111111111111111111111111",
12103 const char *base, *Base, *max;
12105 /* check for hex */
12110 } else if (s[1] == 'b') {
12115 /* check for a decimal in disguise */
12116 else if (s[1] == '.' || s[1] == 'e' || s[1] == 'E')
12118 /* so it must be octal */
12125 if (ckWARN(WARN_SYNTAX))
12126 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
12127 "Misplaced _ in number");
12131 base = bases[shift];
12132 Base = Bases[shift];
12133 max = maxima[shift];
12135 /* read the rest of the number */
12137 /* x is used in the overflow test,
12138 b is the digit we're adding on. */
12143 /* if we don't mention it, we're done */
12147 /* _ are ignored -- but warned about if consecutive */
12149 if (lastub && s == lastub + 1 && ckWARN(WARN_SYNTAX))
12150 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
12151 "Misplaced _ in number");
12155 /* 8 and 9 are not octal */
12156 case '8': case '9':
12158 yyerror(Perl_form(aTHX_ "Illegal octal digit '%c'", *s));
12162 case '2': case '3': case '4':
12163 case '5': case '6': case '7':
12165 yyerror(Perl_form(aTHX_ "Illegal binary digit '%c'", *s));
12168 case '0': case '1':
12169 b = *s++ & 15; /* ASCII digit -> value of digit */
12173 case 'a': case 'b': case 'c': case 'd': case 'e': case 'f':
12174 case 'A': case 'B': case 'C': case 'D': case 'E': case 'F':
12175 /* make sure they said 0x */
12178 b = (*s++ & 7) + 9;
12180 /* Prepare to put the digit we have onto the end
12181 of the number so far. We check for overflows.
12187 x = u << shift; /* make room for the digit */
12189 if ((x >> shift) != u
12190 && !(PL_hints & HINT_NEW_BINARY)) {
12193 if (ckWARN_d(WARN_OVERFLOW))
12194 Perl_warner(aTHX_ packWARN(WARN_OVERFLOW),
12195 "Integer overflow in %s number",
12198 u = x | b; /* add the digit to the end */
12201 n *= nvshift[shift];
12202 /* If an NV has not enough bits in its
12203 * mantissa to represent an UV this summing of
12204 * small low-order numbers is a waste of time
12205 * (because the NV cannot preserve the
12206 * low-order bits anyway): we could just
12207 * remember when did we overflow and in the
12208 * end just multiply n by the right
12216 /* if we get here, we had success: make a scalar value from
12221 /* final misplaced underbar check */
12222 if (s[-1] == '_') {
12223 if (ckWARN(WARN_SYNTAX))
12224 Perl_warner(aTHX_ packWARN(WARN_SYNTAX), "Misplaced _ in number");
12229 if (n > 4294967295.0 && ckWARN(WARN_PORTABLE))
12230 Perl_warner(aTHX_ packWARN(WARN_PORTABLE),
12231 "%s number > %s non-portable",
12237 if (u > 0xffffffff && ckWARN(WARN_PORTABLE))
12238 Perl_warner(aTHX_ packWARN(WARN_PORTABLE),
12239 "%s number > %s non-portable",
12244 if (just_zero && (PL_hints & HINT_NEW_INTEGER))
12245 sv = new_constant(start, s - start, "integer",
12246 sv, NULL, NULL, 0);
12247 else if (PL_hints & HINT_NEW_BINARY)
12248 sv = new_constant(start, s - start, "binary", sv, NULL, NULL, 0);
12253 handle decimal numbers.
12254 we're also sent here when we read a 0 as the first digit
12256 case '1': case '2': case '3': case '4': case '5':
12257 case '6': case '7': case '8': case '9': case '.':
12260 e = PL_tokenbuf + sizeof PL_tokenbuf - 6; /* room for various punctuation */
12263 /* read next group of digits and _ and copy into d */
12264 while (isDIGIT(*s) || *s == '_') {
12265 /* skip underscores, checking for misplaced ones
12269 if (lastub && s == lastub + 1 && ckWARN(WARN_SYNTAX))
12270 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
12271 "Misplaced _ in number");
12275 /* check for end of fixed-length buffer */
12277 Perl_croak(aTHX_ number_too_long);
12278 /* if we're ok, copy the character */
12283 /* final misplaced underbar check */
12284 if (lastub && s == lastub + 1) {
12285 if (ckWARN(WARN_SYNTAX))
12286 Perl_warner(aTHX_ packWARN(WARN_SYNTAX), "Misplaced _ in number");
12289 /* read a decimal portion if there is one. avoid
12290 3..5 being interpreted as the number 3. followed
12293 if (*s == '.' && s[1] != '.') {
12298 if (ckWARN(WARN_SYNTAX))
12299 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
12300 "Misplaced _ in number");
12304 /* copy, ignoring underbars, until we run out of digits.
12306 for (; isDIGIT(*s) || *s == '_'; s++) {
12307 /* fixed length buffer check */
12309 Perl_croak(aTHX_ number_too_long);
12311 if (lastub && s == lastub + 1 && ckWARN(WARN_SYNTAX))
12312 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
12313 "Misplaced _ in number");
12319 /* fractional part ending in underbar? */
12320 if (s[-1] == '_') {
12321 if (ckWARN(WARN_SYNTAX))
12322 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
12323 "Misplaced _ in number");
12325 if (*s == '.' && isDIGIT(s[1])) {
12326 /* oops, it's really a v-string, but without the "v" */
12332 /* read exponent part, if present */
12333 if ((*s == 'e' || *s == 'E') && strchr("+-0123456789_", s[1])) {
12337 /* regardless of whether user said 3E5 or 3e5, use lower 'e' */
12338 *d++ = 'e'; /* At least some Mach atof()s don't grok 'E' */
12340 /* stray preinitial _ */
12342 if (ckWARN(WARN_SYNTAX))
12343 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
12344 "Misplaced _ in number");
12348 /* allow positive or negative exponent */
12349 if (*s == '+' || *s == '-')
12352 /* stray initial _ */
12354 if (ckWARN(WARN_SYNTAX))
12355 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
12356 "Misplaced _ in number");
12360 /* read digits of exponent */
12361 while (isDIGIT(*s) || *s == '_') {
12364 Perl_croak(aTHX_ number_too_long);
12368 if (((lastub && s == lastub + 1) ||
12369 (!isDIGIT(s[1]) && s[1] != '_'))
12370 && ckWARN(WARN_SYNTAX))
12371 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
12372 "Misplaced _ in number");
12379 /* make an sv from the string */
12383 We try to do an integer conversion first if no characters
12384 indicating "float" have been found.
12389 const int flags = grok_number (PL_tokenbuf, d - PL_tokenbuf, &uv);
12391 if (flags == IS_NUMBER_IN_UV) {
12393 sv_setiv(sv, uv); /* Prefer IVs over UVs. */
12396 } else if (flags == (IS_NUMBER_IN_UV | IS_NUMBER_NEG)) {
12397 if (uv <= (UV) IV_MIN)
12398 sv_setiv(sv, -(IV)uv);
12405 /* terminate the string */
12407 nv = Atof(PL_tokenbuf);
12412 ? (PL_hints & HINT_NEW_FLOAT) : (PL_hints & HINT_NEW_INTEGER) ) {
12413 const char *const key = floatit ? "float" : "integer";
12414 const STRLEN keylen = floatit ? 5 : 7;
12415 sv = S_new_constant(aTHX_ PL_tokenbuf, d - PL_tokenbuf,
12416 key, keylen, sv, NULL, NULL, 0);
12420 /* if it starts with a v, it could be a v-string */
12423 sv = newSV(5); /* preallocate storage space */
12424 s = scan_vstring(s, PL_bufend, sv);
12428 /* make the op for the constant and return */
12431 lvalp->opval = newSVOP(OP_CONST, 0, sv);
12433 lvalp->opval = NULL;
12439 S_scan_formline(pTHX_ register char *s)
12442 register char *eol;
12444 SV * const stuff = newSVpvs("");
12445 bool needargs = FALSE;
12446 bool eofmt = FALSE;
12448 char *tokenstart = s;
12449 SV* savewhite = NULL;
12451 if (PL_madskills) {
12452 savewhite = PL_thiswhite;
12457 PERL_ARGS_ASSERT_SCAN_FORMLINE;
12459 while (!needargs) {
12462 #ifdef PERL_STRICT_CR
12463 while (SPACE_OR_TAB(*t))
12466 while (SPACE_OR_TAB(*t) || *t == '\r')
12469 if (*t == '\n' || t == PL_bufend) {
12474 if (PL_in_eval && !PL_rsfp) {
12475 eol = (char *) memchr(s,'\n',PL_bufend-s);
12480 eol = PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
12482 for (t = s; t < eol; t++) {
12483 if (*t == '~' && t[1] == '~' && SvCUR(stuff)) {
12485 goto enough; /* ~~ must be first line in formline */
12487 if (*t == '@' || *t == '^')
12491 sv_catpvn(stuff, s, eol-s);
12492 #ifndef PERL_STRICT_CR
12493 if (eol-s > 1 && eol[-2] == '\r' && eol[-1] == '\n') {
12494 char *end = SvPVX(stuff) + SvCUR(stuff);
12497 SvCUR_set(stuff, SvCUR(stuff) - 1);
12507 if (PL_madskills) {
12509 sv_catpvn(PL_thistoken, tokenstart, PL_bufend - tokenstart);
12511 PL_thistoken = newSVpvn(tokenstart, PL_bufend - tokenstart);
12514 s = filter_gets(PL_linestr, PL_rsfp, 0);
12516 tokenstart = PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = SvPVX(PL_linestr);
12518 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = SvPVX(PL_linestr);
12520 PL_bufend = PL_bufptr + SvCUR(PL_linestr);
12521 PL_last_lop = PL_last_uni = NULL;
12530 if (SvCUR(stuff)) {
12533 PL_lex_state = LEX_NORMAL;
12534 start_force(PL_curforce);
12535 NEXTVAL_NEXTTOKE.ival = 0;
12539 PL_lex_state = LEX_FORMLINE;
12541 if (UTF && is_utf8_string((U8*)SvPVX_const(stuff), SvCUR(stuff)))
12543 else if (PL_encoding)
12544 sv_recode_to_utf8(stuff, PL_encoding);
12546 start_force(PL_curforce);
12547 NEXTVAL_NEXTTOKE.opval = (OP*)newSVOP(OP_CONST, 0, stuff);
12549 start_force(PL_curforce);
12550 NEXTVAL_NEXTTOKE.ival = OP_FORMLINE;
12554 SvREFCNT_dec(stuff);
12556 PL_lex_formbrack = 0;
12560 if (PL_madskills) {
12562 sv_catpvn(PL_thistoken, tokenstart, s - tokenstart);
12564 PL_thistoken = newSVpvn(tokenstart, s - tokenstart);
12565 PL_thiswhite = savewhite;
12572 Perl_start_subparse(pTHX_ I32 is_format, U32 flags)
12575 const I32 oldsavestack_ix = PL_savestack_ix;
12576 CV* const outsidecv = PL_compcv;
12579 assert(SvTYPE(PL_compcv) == SVt_PVCV);
12581 SAVEI32(PL_subline);
12582 save_item(PL_subname);
12583 SAVESPTR(PL_compcv);
12585 PL_compcv = MUTABLE_CV(newSV_type(is_format ? SVt_PVFM : SVt_PVCV));
12586 CvFLAGS(PL_compcv) |= flags;
12588 PL_subline = CopLINE(PL_curcop);
12589 CvPADLIST(PL_compcv) = pad_new(padnew_SAVE|padnew_SAVESUB);
12590 CvOUTSIDE(PL_compcv) = MUTABLE_CV(SvREFCNT_inc_simple(outsidecv));
12591 CvOUTSIDE_SEQ(PL_compcv) = PL_cop_seqmax;
12593 return oldsavestack_ix;
12597 #pragma segment Perl_yylex
12600 S_yywarn(pTHX_ const char *const s)
12604 PERL_ARGS_ASSERT_YYWARN;
12606 PL_in_eval |= EVAL_WARNONLY;
12608 PL_in_eval &= ~EVAL_WARNONLY;
12613 Perl_yyerror(pTHX_ const char *const s)
12616 const char *where = NULL;
12617 const char *context = NULL;
12620 int yychar = PL_parser->yychar;
12622 PERL_ARGS_ASSERT_YYERROR;
12624 if (!yychar || (yychar == ';' && !PL_rsfp))
12626 else if (PL_oldoldbufptr && PL_bufptr > PL_oldoldbufptr &&
12627 PL_bufptr - PL_oldoldbufptr < 200 && PL_oldoldbufptr != PL_oldbufptr &&
12628 PL_oldbufptr != PL_bufptr) {
12631 The code below is removed for NetWare because it abends/crashes on NetWare
12632 when the script has error such as not having the closing quotes like:
12633 if ($var eq "value)
12634 Checking of white spaces is anyway done in NetWare code.
12637 while (isSPACE(*PL_oldoldbufptr))
12640 context = PL_oldoldbufptr;
12641 contlen = PL_bufptr - PL_oldoldbufptr;
12643 else if (PL_oldbufptr && PL_bufptr > PL_oldbufptr &&
12644 PL_bufptr - PL_oldbufptr < 200 && PL_oldbufptr != PL_bufptr) {
12647 The code below is removed for NetWare because it abends/crashes on NetWare
12648 when the script has error such as not having the closing quotes like:
12649 if ($var eq "value)
12650 Checking of white spaces is anyway done in NetWare code.
12653 while (isSPACE(*PL_oldbufptr))
12656 context = PL_oldbufptr;
12657 contlen = PL_bufptr - PL_oldbufptr;
12659 else if (yychar > 255)
12660 where = "next token ???";
12661 else if (yychar == -2) { /* YYEMPTY */
12662 if (PL_lex_state == LEX_NORMAL ||
12663 (PL_lex_state == LEX_KNOWNEXT && PL_lex_defer == LEX_NORMAL))
12664 where = "at end of line";
12665 else if (PL_lex_inpat)
12666 where = "within pattern";
12668 where = "within string";
12671 SV * const where_sv = newSVpvs_flags("next char ", SVs_TEMP);
12673 Perl_sv_catpvf(aTHX_ where_sv, "^%c", toCTRL(yychar));
12674 else if (isPRINT_LC(yychar)) {
12675 const char string = yychar;
12676 sv_catpvn(where_sv, &string, 1);
12679 Perl_sv_catpvf(aTHX_ where_sv, "\\%03o", yychar & 255);
12680 where = SvPVX_const(where_sv);
12682 msg = sv_2mortal(newSVpv(s, 0));
12683 Perl_sv_catpvf(aTHX_ msg, " at %s line %"IVdf", ",
12684 OutCopFILE(PL_curcop), (IV)CopLINE(PL_curcop));
12686 Perl_sv_catpvf(aTHX_ msg, "near \"%.*s\"\n", contlen, context);
12688 Perl_sv_catpvf(aTHX_ msg, "%s\n", where);
12689 if (PL_multi_start < PL_multi_end && (U32)(CopLINE(PL_curcop) - PL_multi_end) <= 1) {
12690 Perl_sv_catpvf(aTHX_ msg,
12691 " (Might be a runaway multi-line %c%c string starting on line %"IVdf")\n",
12692 (int)PL_multi_open,(int)PL_multi_close,(IV)PL_multi_start);
12695 if (PL_in_eval & EVAL_WARNONLY) {
12696 if (ckWARN_d(WARN_SYNTAX))
12697 Perl_warner(aTHX_ packWARN(WARN_SYNTAX), "%"SVf, SVfARG(msg));
12701 if (PL_error_count >= 10) {
12702 if (PL_in_eval && SvCUR(ERRSV))
12703 Perl_croak(aTHX_ "%"SVf"%s has too many errors.\n",
12704 SVfARG(ERRSV), OutCopFILE(PL_curcop));
12706 Perl_croak(aTHX_ "%s has too many errors.\n",
12707 OutCopFILE(PL_curcop));
12710 PL_in_my_stash = NULL;
12714 #pragma segment Main
12718 S_swallow_bom(pTHX_ U8 *s)
12721 const STRLEN slen = SvCUR(PL_linestr);
12723 PERL_ARGS_ASSERT_SWALLOW_BOM;
12727 if (s[1] == 0xFE) {
12728 /* UTF-16 little-endian? (or UTF32-LE?) */
12729 if (s[2] == 0 && s[3] == 0) /* UTF-32 little-endian */
12730 Perl_croak(aTHX_ "Unsupported script encoding UTF32-LE");
12731 #ifndef PERL_NO_UTF16_FILTER
12732 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF16-LE script encoding (BOM)\n");
12735 if (PL_bufend > (char*)s) {
12739 filter_add(utf16rev_textfilter, NULL);
12740 Newx(news, (PL_bufend - (char*)s) * 3 / 2 + 1, U8);
12741 utf16_to_utf8_reversed(s, news,
12742 PL_bufend - (char*)s - 1,
12744 sv_setpvn(PL_linestr, (const char*)news, newlen);
12746 s = (U8*)SvPVX(PL_linestr);
12747 Copy(news, s, newlen, U8);
12751 SvUTF8_on(PL_linestr);
12752 s = (U8*)SvPVX(PL_linestr);
12754 /* FIXME - is this a general bug fix? */
12757 PL_bufend = SvPVX(PL_linestr) + newlen;
12760 Perl_croak(aTHX_ "Unsupported script encoding UTF16-LE");
12765 if (s[1] == 0xFF) { /* UTF-16 big-endian? */
12766 #ifndef PERL_NO_UTF16_FILTER
12767 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16BE script encoding (BOM)\n");
12770 if (PL_bufend > (char *)s) {
12774 filter_add(utf16_textfilter, NULL);
12775 Newx(news, (PL_bufend - (char*)s) * 3 / 2 + 1, U8);
12776 utf16_to_utf8(s, news,
12777 PL_bufend - (char*)s,
12779 sv_setpvn(PL_linestr, (const char*)news, newlen);
12781 SvUTF8_on(PL_linestr);
12782 s = (U8*)SvPVX(PL_linestr);
12783 PL_bufend = SvPVX(PL_linestr) + newlen;
12786 Perl_croak(aTHX_ "Unsupported script encoding UTF16-BE");
12791 if (slen > 2 && s[1] == 0xBB && s[2] == 0xBF) {
12792 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-8 script encoding (BOM)\n");
12793 s += 3; /* UTF-8 */
12799 if (s[2] == 0xFE && s[3] == 0xFF) {
12800 /* UTF-32 big-endian */
12801 Perl_croak(aTHX_ "Unsupported script encoding UTF32-BE");
12804 else if (s[2] == 0 && s[3] != 0) {
12807 * are a good indicator of UTF-16BE. */
12808 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16BE script encoding (no BOM)\n");
12814 if (slen > 3 && s[1] == 0x73 && s[2] == 0x66 && s[3] == 0x73) {
12815 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-8 script encoding (BOM)\n");
12816 s += 4; /* UTF-8 */
12822 if (slen > 3 && s[1] == 0 && s[2] != 0 && s[3] == 0) {
12825 * are a good indicator of UTF-16LE. */
12826 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16LE script encoding (no BOM)\n");
12834 #ifndef PERL_NO_UTF16_FILTER
12836 utf16_textfilter(pTHX_ int idx, SV *sv, int maxlen)
12839 const STRLEN old = SvCUR(sv);
12840 const I32 count = FILTER_READ(idx+1, sv, maxlen);
12841 DEBUG_P(PerlIO_printf(Perl_debug_log,
12842 "utf16_textfilter(%p): %d %d (%d)\n",
12843 FPTR2DPTR(void *, utf16_textfilter),
12844 idx, maxlen, (int) count));
12848 Newx(tmps, SvCUR(sv) * 3 / 2 + 1, U8);
12849 Copy(SvPVX_const(sv), tmps, old, char);
12850 utf16_to_utf8((U8*)SvPVX_const(sv) + old, tmps + old,
12851 SvCUR(sv) - old, &newlen);
12852 sv_usepvn(sv, (char*)tmps, (STRLEN)newlen + old);
12854 DEBUG_P({sv_dump(sv);});
12859 utf16rev_textfilter(pTHX_ int idx, SV *sv, int maxlen)
12862 const STRLEN old = SvCUR(sv);
12863 const I32 count = FILTER_READ(idx+1, sv, maxlen);
12864 DEBUG_P(PerlIO_printf(Perl_debug_log,
12865 "utf16rev_textfilter(%p): %d %d (%d)\n",
12866 FPTR2DPTR(void *, utf16rev_textfilter),
12867 idx, maxlen, (int) count));
12871 Newx(tmps, SvCUR(sv) * 3 / 2 + 1, U8);
12872 Copy(SvPVX_const(sv), tmps, old, char);
12873 utf16_to_utf8((U8*)SvPVX_const(sv) + old, tmps + old,
12874 SvCUR(sv) - old, &newlen);
12875 sv_usepvn(sv, (char*)tmps, (STRLEN)newlen + old);
12877 DEBUG_P({ sv_dump(sv); });
12883 Returns a pointer to the next character after the parsed
12884 vstring, as well as updating the passed in sv.
12886 Function must be called like
12889 s = scan_vstring(s,e,sv);
12891 where s and e are the start and end of the string.
12892 The sv should already be large enough to store the vstring
12893 passed in, for performance reasons.
12898 Perl_scan_vstring(pTHX_ const char *s, const char *const e, SV *sv)
12901 const char *pos = s;
12902 const char *start = s;
12904 PERL_ARGS_ASSERT_SCAN_VSTRING;
12906 if (*pos == 'v') pos++; /* get past 'v' */
12907 while (pos < e && (isDIGIT(*pos) || *pos == '_'))
12909 if ( *pos != '.') {
12910 /* this may not be a v-string if followed by => */
12911 const char *next = pos;
12912 while (next < e && isSPACE(*next))
12914 if ((e - next) >= 2 && *next == '=' && next[1] == '>' ) {
12915 /* return string not v-string */
12916 sv_setpvn(sv,(char *)s,pos-s);
12917 return (char *)pos;
12921 if (!isALPHA(*pos)) {
12922 U8 tmpbuf[UTF8_MAXBYTES+1];
12925 s++; /* get past 'v' */
12930 /* this is atoi() that tolerates underscores */
12933 const char *end = pos;
12935 while (--end >= s) {
12937 const UV orev = rev;
12938 rev += (*end - '0') * mult;
12940 if (orev > rev && ckWARN_d(WARN_OVERFLOW))
12941 Perl_warner(aTHX_ packWARN(WARN_OVERFLOW),
12942 "Integer overflow in decimal number");
12946 if (rev > 0x7FFFFFFF)
12947 Perl_croak(aTHX_ "In EBCDIC the v-string components cannot exceed 2147483647");
12949 /* Append native character for the rev point */
12950 tmpend = uvchr_to_utf8(tmpbuf, rev);
12951 sv_catpvn(sv, (const char*)tmpbuf, tmpend - tmpbuf);
12952 if (!UNI_IS_INVARIANT(NATIVE_TO_UNI(rev)))
12954 if (pos + 1 < e && *pos == '.' && isDIGIT(pos[1]))
12960 while (pos < e && (isDIGIT(*pos) || *pos == '_'))
12964 sv_magic(sv,NULL,PERL_MAGIC_vstring,(const char*)start, pos-start);
12972 * c-indentation-style: bsd
12973 * c-basic-offset: 4
12974 * indent-tabs-mode: t
12977 * ex: set ts=8 sts=4 sw=4 noet: