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 /* In variables named $^X, these are the legal values for X.
128 * 1999-02-27 mjd-perl-patch@plover.com */
129 #define isCONTROLVAR(x) (isUPPER(x) || strchr("[\\]^_?", (x)))
131 /* On MacOS, respect nonbreaking spaces */
132 #ifdef MACOS_TRADITIONAL
133 #define SPACE_OR_TAB(c) ((c)==' '||(c)=='\312'||(c)=='\t')
135 #define SPACE_OR_TAB(c) ((c)==' '||(c)=='\t')
138 /* LEX_* are values for PL_lex_state, the state of the lexer.
139 * They are arranged oddly so that the guard on the switch statement
140 * can get by with a single comparison (if the compiler is smart enough).
143 /* #define LEX_NOTPARSING 11 is done in perl.h. */
145 #define LEX_NORMAL 10 /* normal code (ie not within "...") */
146 #define LEX_INTERPNORMAL 9 /* code within a string, eg "$foo[$x+1]" */
147 #define LEX_INTERPCASEMOD 8 /* expecting a \U, \Q or \E etc */
148 #define LEX_INTERPPUSH 7 /* starting a new sublex parse level */
149 #define LEX_INTERPSTART 6 /* expecting the start of a $var */
151 /* at end of code, eg "$x" followed by: */
152 #define LEX_INTERPEND 5 /* ... eg not one of [, { or -> */
153 #define LEX_INTERPENDMAYBE 4 /* ... eg one of [, { or -> */
155 #define LEX_INTERPCONCAT 3 /* expecting anything, eg at start of
156 string or after \E, $foo, etc */
157 #define LEX_INTERPCONST 2 /* NOT USED */
158 #define LEX_FORMLINE 1 /* expecting a format line */
159 #define LEX_KNOWNEXT 0 /* next token known; just return it */
163 static const char* const lex_state_names[] = {
182 #include "keywords.h"
184 /* CLINE is a macro that ensures PL_copline has a sane value */
189 #define CLINE (PL_copline = (CopLINE(PL_curcop) < PL_copline ? CopLINE(PL_curcop) : PL_copline))
192 # define SKIPSPACE0(s) skipspace0(s)
193 # define SKIPSPACE1(s) skipspace1(s)
194 # define SKIPSPACE2(s,tsv) skipspace2(s,&tsv)
195 # define PEEKSPACE(s) skipspace2(s,0)
197 # define SKIPSPACE0(s) skipspace(s)
198 # define SKIPSPACE1(s) skipspace(s)
199 # define SKIPSPACE2(s,tsv) skipspace(s)
200 # define PEEKSPACE(s) skipspace(s)
204 * Convenience functions to return different tokens and prime the
205 * lexer for the next token. They all take an argument.
207 * TOKEN : generic token (used for '(', DOLSHARP, etc)
208 * OPERATOR : generic operator
209 * AOPERATOR : assignment operator
210 * PREBLOCK : beginning the block after an if, while, foreach, ...
211 * PRETERMBLOCK : beginning a non-code-defining {} block (eg, hash ref)
212 * PREREF : *EXPR where EXPR is not a simple identifier
213 * TERM : expression term
214 * LOOPX : loop exiting command (goto, last, dump, etc)
215 * FTST : file test operator
216 * FUN0 : zero-argument function
217 * FUN1 : not used, except for not, which isn't a UNIOP
218 * BOop : bitwise or or xor
220 * SHop : shift operator
221 * PWop : power operator
222 * PMop : pattern-matching operator
223 * Aop : addition-level operator
224 * Mop : multiplication-level operator
225 * Eop : equality-testing operator
226 * Rop : relational operator <= != gt
228 * Also see LOP and lop() below.
231 #ifdef DEBUGGING /* Serve -DT. */
232 # define REPORT(retval) tokereport((I32)retval, &pl_yylval)
234 # define REPORT(retval) (retval)
237 #define TOKEN(retval) return ( PL_bufptr = s, REPORT(retval))
238 #define OPERATOR(retval) return (PL_expect = XTERM, PL_bufptr = s, REPORT(retval))
239 #define AOPERATOR(retval) return ao((PL_expect = XTERM, PL_bufptr = s, REPORT(retval)))
240 #define PREBLOCK(retval) return (PL_expect = XBLOCK,PL_bufptr = s, REPORT(retval))
241 #define PRETERMBLOCK(retval) return (PL_expect = XTERMBLOCK,PL_bufptr = s, REPORT(retval))
242 #define PREREF(retval) return (PL_expect = XREF,PL_bufptr = s, REPORT(retval))
243 #define TERM(retval) return (CLINE, PL_expect = XOPERATOR, PL_bufptr = s, REPORT(retval))
244 #define LOOPX(f) return (pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)LOOPEX))
245 #define FTST(f) return (pl_yylval.ival=f, PL_expect=XTERMORDORDOR, PL_bufptr=s, REPORT((int)UNIOP))
246 #define FUN0(f) return (pl_yylval.ival=f, PL_expect=XOPERATOR, PL_bufptr=s, REPORT((int)FUNC0))
247 #define FUN1(f) return (pl_yylval.ival=f, PL_expect=XOPERATOR, PL_bufptr=s, REPORT((int)FUNC1))
248 #define BOop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)BITOROP)))
249 #define BAop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)BITANDOP)))
250 #define SHop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)SHIFTOP)))
251 #define PWop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)POWOP)))
252 #define PMop(f) return(pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)MATCHOP))
253 #define Aop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)ADDOP)))
254 #define Mop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)MULOP)))
255 #define Eop(f) return (pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)EQOP))
256 #define Rop(f) return (pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)RELOP))
258 /* This bit of chicanery makes a unary function followed by
259 * a parenthesis into a function with one argument, highest precedence.
260 * The UNIDOR macro is for unary functions that can be followed by the //
261 * operator (such as C<shift // 0>).
263 #define UNI2(f,x) { \
264 pl_yylval.ival = f; \
267 PL_last_uni = PL_oldbufptr; \
268 PL_last_lop_op = f; \
270 return REPORT( (int)FUNC1 ); \
272 return REPORT( *s=='(' ? (int)FUNC1 : (int)UNIOP ); \
274 #define UNI(f) UNI2(f,XTERM)
275 #define UNIDOR(f) UNI2(f,XTERMORDORDOR)
277 #define UNIBRACK(f) { \
278 pl_yylval.ival = f; \
280 PL_last_uni = PL_oldbufptr; \
282 return REPORT( (int)FUNC1 ); \
284 return REPORT( (*s == '(') ? (int)FUNC1 : (int)UNIOP ); \
287 /* grandfather return to old style */
288 #define OLDLOP(f) return(pl_yylval.ival=f,PL_expect = XTERM,PL_bufptr = s,(int)LSTOP)
292 /* how to interpret the pl_yylval associated with the token */
296 TOKENTYPE_OPNUM, /* pl_yylval.ival contains an opcode number */
302 static struct debug_tokens {
304 enum token_type type;
306 } const debug_tokens[] =
308 { ADDOP, TOKENTYPE_OPNUM, "ADDOP" },
309 { ANDAND, TOKENTYPE_NONE, "ANDAND" },
310 { ANDOP, TOKENTYPE_NONE, "ANDOP" },
311 { ANONSUB, TOKENTYPE_IVAL, "ANONSUB" },
312 { ARROW, TOKENTYPE_NONE, "ARROW" },
313 { ASSIGNOP, TOKENTYPE_OPNUM, "ASSIGNOP" },
314 { BITANDOP, TOKENTYPE_OPNUM, "BITANDOP" },
315 { BITOROP, TOKENTYPE_OPNUM, "BITOROP" },
316 { COLONATTR, TOKENTYPE_NONE, "COLONATTR" },
317 { CONTINUE, TOKENTYPE_NONE, "CONTINUE" },
318 { DEFAULT, TOKENTYPE_NONE, "DEFAULT" },
319 { DO, TOKENTYPE_NONE, "DO" },
320 { DOLSHARP, TOKENTYPE_NONE, "DOLSHARP" },
321 { DORDOR, TOKENTYPE_NONE, "DORDOR" },
322 { DOROP, TOKENTYPE_OPNUM, "DOROP" },
323 { DOTDOT, TOKENTYPE_IVAL, "DOTDOT" },
324 { ELSE, TOKENTYPE_NONE, "ELSE" },
325 { ELSIF, TOKENTYPE_IVAL, "ELSIF" },
326 { EQOP, TOKENTYPE_OPNUM, "EQOP" },
327 { FOR, TOKENTYPE_IVAL, "FOR" },
328 { FORMAT, TOKENTYPE_NONE, "FORMAT" },
329 { FUNC, TOKENTYPE_OPNUM, "FUNC" },
330 { FUNC0, TOKENTYPE_OPNUM, "FUNC0" },
331 { FUNC0SUB, TOKENTYPE_OPVAL, "FUNC0SUB" },
332 { FUNC1, TOKENTYPE_OPNUM, "FUNC1" },
333 { FUNCMETH, TOKENTYPE_OPVAL, "FUNCMETH" },
334 { GIVEN, TOKENTYPE_IVAL, "GIVEN" },
335 { HASHBRACK, TOKENTYPE_NONE, "HASHBRACK" },
336 { IF, TOKENTYPE_IVAL, "IF" },
337 { LABEL, TOKENTYPE_PVAL, "LABEL" },
338 { LOCAL, TOKENTYPE_IVAL, "LOCAL" },
339 { LOOPEX, TOKENTYPE_OPNUM, "LOOPEX" },
340 { LSTOP, TOKENTYPE_OPNUM, "LSTOP" },
341 { LSTOPSUB, TOKENTYPE_OPVAL, "LSTOPSUB" },
342 { MATCHOP, TOKENTYPE_OPNUM, "MATCHOP" },
343 { METHOD, TOKENTYPE_OPVAL, "METHOD" },
344 { MULOP, TOKENTYPE_OPNUM, "MULOP" },
345 { MY, TOKENTYPE_IVAL, "MY" },
346 { MYSUB, TOKENTYPE_NONE, "MYSUB" },
347 { NOAMP, TOKENTYPE_NONE, "NOAMP" },
348 { NOTOP, TOKENTYPE_NONE, "NOTOP" },
349 { OROP, TOKENTYPE_IVAL, "OROP" },
350 { OROR, TOKENTYPE_NONE, "OROR" },
351 { PACKAGE, TOKENTYPE_NONE, "PACKAGE" },
352 { PMFUNC, TOKENTYPE_OPVAL, "PMFUNC" },
353 { POSTDEC, TOKENTYPE_NONE, "POSTDEC" },
354 { POSTINC, TOKENTYPE_NONE, "POSTINC" },
355 { POWOP, TOKENTYPE_OPNUM, "POWOP" },
356 { PREDEC, TOKENTYPE_NONE, "PREDEC" },
357 { PREINC, TOKENTYPE_NONE, "PREINC" },
358 { PRIVATEREF, TOKENTYPE_OPVAL, "PRIVATEREF" },
359 { REFGEN, TOKENTYPE_NONE, "REFGEN" },
360 { RELOP, TOKENTYPE_OPNUM, "RELOP" },
361 { SHIFTOP, TOKENTYPE_OPNUM, "SHIFTOP" },
362 { SUB, TOKENTYPE_NONE, "SUB" },
363 { THING, TOKENTYPE_OPVAL, "THING" },
364 { UMINUS, TOKENTYPE_NONE, "UMINUS" },
365 { UNIOP, TOKENTYPE_OPNUM, "UNIOP" },
366 { UNIOPSUB, TOKENTYPE_OPVAL, "UNIOPSUB" },
367 { UNLESS, TOKENTYPE_IVAL, "UNLESS" },
368 { UNTIL, TOKENTYPE_IVAL, "UNTIL" },
369 { USE, TOKENTYPE_IVAL, "USE" },
370 { WHEN, TOKENTYPE_IVAL, "WHEN" },
371 { WHILE, TOKENTYPE_IVAL, "WHILE" },
372 { WORD, TOKENTYPE_OPVAL, "WORD" },
373 { YADAYADA, TOKENTYPE_IVAL, "YADAYADA" },
374 { 0, TOKENTYPE_NONE, NULL }
377 /* dump the returned token in rv, plus any optional arg in pl_yylval */
380 S_tokereport(pTHX_ I32 rv, const YYSTYPE* lvalp)
384 PERL_ARGS_ASSERT_TOKEREPORT;
387 const char *name = NULL;
388 enum token_type type = TOKENTYPE_NONE;
389 const struct debug_tokens *p;
390 SV* const report = newSVpvs("<== ");
392 for (p = debug_tokens; p->token; p++) {
393 if (p->token == (int)rv) {
400 Perl_sv_catpv(aTHX_ report, name);
401 else if ((char)rv > ' ' && (char)rv < '~')
402 Perl_sv_catpvf(aTHX_ report, "'%c'", (char)rv);
404 sv_catpvs(report, "EOF");
406 Perl_sv_catpvf(aTHX_ report, "?? %"IVdf, (IV)rv);
409 case TOKENTYPE_GVVAL: /* doesn't appear to be used */
412 Perl_sv_catpvf(aTHX_ report, "(ival=%"IVdf")", (IV)lvalp->ival);
414 case TOKENTYPE_OPNUM:
415 Perl_sv_catpvf(aTHX_ report, "(ival=op_%s)",
416 PL_op_name[lvalp->ival]);
419 Perl_sv_catpvf(aTHX_ report, "(pval=\"%s\")", lvalp->pval);
421 case TOKENTYPE_OPVAL:
423 Perl_sv_catpvf(aTHX_ report, "(opval=op_%s)",
424 PL_op_name[lvalp->opval->op_type]);
425 if (lvalp->opval->op_type == OP_CONST) {
426 Perl_sv_catpvf(aTHX_ report, " %s",
427 SvPEEK(cSVOPx_sv(lvalp->opval)));
432 sv_catpvs(report, "(opval=null)");
435 PerlIO_printf(Perl_debug_log, "### %s\n\n", SvPV_nolen_const(report));
441 /* print the buffer with suitable escapes */
444 S_printbuf(pTHX_ const char *const fmt, const char *const s)
446 SV* const tmp = newSVpvs("");
448 PERL_ARGS_ASSERT_PRINTBUF;
450 PerlIO_printf(Perl_debug_log, fmt, pv_display(tmp, s, strlen(s), 0, 60));
459 * This subroutine detects &&=, ||=, and //= and turns an ANDAND, OROR or DORDOR
460 * into an OP_ANDASSIGN, OP_ORASSIGN, or OP_DORASSIGN
464 S_ao(pTHX_ int toketype)
467 if (*PL_bufptr == '=') {
469 if (toketype == ANDAND)
470 pl_yylval.ival = OP_ANDASSIGN;
471 else if (toketype == OROR)
472 pl_yylval.ival = OP_ORASSIGN;
473 else if (toketype == DORDOR)
474 pl_yylval.ival = OP_DORASSIGN;
482 * When Perl expects an operator and finds something else, no_op
483 * prints the warning. It always prints "<something> found where
484 * operator expected. It prints "Missing semicolon on previous line?"
485 * if the surprise occurs at the start of the line. "do you need to
486 * predeclare ..." is printed out for code like "sub bar; foo bar $x"
487 * where the compiler doesn't know if foo is a method call or a function.
488 * It prints "Missing operator before end of line" if there's nothing
489 * after the missing operator, or "... before <...>" if there is something
490 * after the missing operator.
494 S_no_op(pTHX_ const char *const what, char *s)
497 char * const oldbp = PL_bufptr;
498 const bool is_first = (PL_oldbufptr == PL_linestart);
500 PERL_ARGS_ASSERT_NO_OP;
506 yywarn(Perl_form(aTHX_ "%s found where operator expected", what));
507 if (ckWARN_d(WARN_SYNTAX)) {
509 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
510 "\t(Missing semicolon on previous line?)\n");
511 else if (PL_oldoldbufptr && isIDFIRST_lazy_if(PL_oldoldbufptr,UTF)) {
513 for (t = PL_oldoldbufptr; (isALNUM_lazy_if(t,UTF) || *t == ':'); t++)
515 if (t < PL_bufptr && isSPACE(*t))
516 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
517 "\t(Do you need to predeclare %.*s?)\n",
518 (int)(t - PL_oldoldbufptr), PL_oldoldbufptr);
522 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
523 "\t(Missing operator before %.*s?)\n", (int)(s - oldbp), oldbp);
531 * Complain about missing quote/regexp/heredoc terminator.
532 * If it's called with NULL then it cauterizes the line buffer.
533 * If we're in a delimited string and the delimiter is a control
534 * character, it's reformatted into a two-char sequence like ^C.
539 S_missingterm(pTHX_ char *s)
545 char * const nl = strrchr(s,'\n');
549 else if (isCNTRL(PL_multi_close)) {
551 tmpbuf[1] = (char)toCTRL(PL_multi_close);
556 *tmpbuf = (char)PL_multi_close;
560 q = strchr(s,'"') ? '\'' : '"';
561 Perl_croak(aTHX_ "Can't find string terminator %c%s%c anywhere before EOF",q,s,q);
564 #define FEATURE_IS_ENABLED(name) \
565 ((0 != (PL_hints & HINT_LOCALIZE_HH)) \
566 && S_feature_is_enabled(aTHX_ STR_WITH_LEN(name)))
567 /* The longest string we pass in. */
568 #define MAX_FEATURE_LEN (sizeof("switch")-1)
571 * S_feature_is_enabled
572 * Check whether the named feature is enabled.
575 S_feature_is_enabled(pTHX_ const char *const name, STRLEN namelen)
578 HV * const hinthv = GvHV(PL_hintgv);
579 char he_name[8 + MAX_FEATURE_LEN] = "feature_";
581 PERL_ARGS_ASSERT_FEATURE_IS_ENABLED;
583 assert(namelen <= MAX_FEATURE_LEN);
584 memcpy(&he_name[8], name, namelen);
586 return (hinthv && hv_exists(hinthv, he_name, 8 + namelen));
594 Perl_deprecate(pTHX_ const char *const s)
596 PERL_ARGS_ASSERT_DEPRECATE;
598 if (ckWARN(WARN_DEPRECATED))
599 Perl_warner(aTHX_ packWARN(WARN_DEPRECATED), "Use of %s is deprecated", s);
603 Perl_deprecate_old(pTHX_ const char *const s)
605 /* This function should NOT be called for any new deprecated warnings */
606 /* Use Perl_deprecate instead */
608 /* It is here to maintain backward compatibility with the pre-5.8 */
609 /* warnings category hierarchy. The "deprecated" category used to */
610 /* live under the "syntax" category. It is now a top-level category */
611 /* in its own right. */
613 PERL_ARGS_ASSERT_DEPRECATE_OLD;
615 if (ckWARN2(WARN_DEPRECATED, WARN_SYNTAX))
616 Perl_warner(aTHX_ packWARN2(WARN_DEPRECATED, WARN_SYNTAX),
617 "Use of %s is deprecated", s);
621 * experimental text filters for win32 carriage-returns, utf16-to-utf8 and
622 * utf16-to-utf8-reversed.
625 #ifdef PERL_CR_FILTER
629 register const char *s = SvPVX_const(sv);
630 register const char * const e = s + SvCUR(sv);
632 PERL_ARGS_ASSERT_STRIP_RETURN;
634 /* outer loop optimized to do nothing if there are no CR-LFs */
636 if (*s++ == '\r' && *s == '\n') {
637 /* hit a CR-LF, need to copy the rest */
638 register char *d = s - 1;
641 if (*s == '\r' && s[1] == '\n')
652 S_cr_textfilter(pTHX_ int idx, SV *sv, int maxlen)
654 const I32 count = FILTER_READ(idx+1, sv, maxlen);
655 if (count > 0 && !maxlen)
666 * Create a parser object and initialise its parser and lexer fields
668 * rsfp is the opened file handle to read from (if any),
670 * line holds any initial content already read from the file (or in
671 * the case of no file, such as an eval, the whole contents);
673 * new_filter indicates that this is a new file and it shouldn't inherit
674 * the filters from the current parser (ie require).
678 Perl_lex_start(pTHX_ SV *line, PerlIO *rsfp, bool new_filter)
681 const char *s = NULL;
683 yy_parser *parser, *oparser;
685 /* create and initialise a parser */
687 Newxz(parser, 1, yy_parser);
688 parser->old_parser = oparser = PL_parser;
691 Newx(parser->stack, YYINITDEPTH, yy_stack_frame);
692 parser->ps = parser->stack;
693 parser->stack_size = YYINITDEPTH;
695 parser->stack->state = 0;
696 parser->yyerrstatus = 0;
697 parser->yychar = YYEMPTY; /* Cause a token to be read. */
699 /* on scope exit, free this parser and restore any outer one */
701 parser->saved_curcop = PL_curcop;
703 /* initialise lexer state */
706 parser->curforce = -1;
708 parser->nexttoke = 0;
710 parser->error_count = oparser ? oparser->error_count : 0;
711 parser->copline = NOLINE;
712 parser->lex_state = LEX_NORMAL;
713 parser->expect = XSTATE;
715 parser->rsfp_filters = (new_filter || !oparser) ? newAV()
716 : MUTABLE_AV(SvREFCNT_inc(oparser->rsfp_filters));
718 Newx(parser->lex_brackstack, 120, char);
719 Newx(parser->lex_casestack, 12, char);
720 *parser->lex_casestack = '\0';
723 s = SvPV_const(line, len);
729 parser->linestr = newSVpvs("\n;");
730 } else if (SvREADONLY(line) || s[len-1] != ';') {
731 parser->linestr = newSVsv(line);
733 sv_catpvs(parser->linestr, "\n;");
736 SvREFCNT_inc_simple_void_NN(line);
737 parser->linestr = line;
739 parser->oldoldbufptr =
742 parser->linestart = SvPVX(parser->linestr);
743 parser->bufend = parser->bufptr + SvCUR(parser->linestr);
744 parser->last_lop = parser->last_uni = NULL;
748 /* delete a parser object */
751 Perl_parser_free(pTHX_ const yy_parser *parser)
753 PERL_ARGS_ASSERT_PARSER_FREE;
755 PL_curcop = parser->saved_curcop;
756 SvREFCNT_dec(parser->linestr);
758 if (parser->rsfp == PerlIO_stdin())
759 PerlIO_clearerr(parser->rsfp);
760 else if (parser->rsfp && (!parser->old_parser ||
761 (parser->old_parser && parser->rsfp != parser->old_parser->rsfp)))
762 PerlIO_close(parser->rsfp);
763 SvREFCNT_dec(parser->rsfp_filters);
765 Safefree(parser->stack);
766 Safefree(parser->lex_brackstack);
767 Safefree(parser->lex_casestack);
768 PL_parser = parser->old_parser;
775 * Finalizer for lexing operations. Must be called when the parser is
776 * done with the lexer.
783 PL_doextract = FALSE;
788 * This subroutine has nothing to do with tilting, whether at windmills
789 * or pinball tables. Its name is short for "increment line". It
790 * increments the current line number in CopLINE(PL_curcop) and checks
791 * to see whether the line starts with a comment of the form
792 * # line 500 "foo.pm"
793 * If so, it sets the current line number and file to the values in the comment.
797 S_incline(pTHX_ const char *s)
804 PERL_ARGS_ASSERT_INCLINE;
806 CopLINE_inc(PL_curcop);
809 while (SPACE_OR_TAB(*s))
811 if (strnEQ(s, "line", 4))
815 if (SPACE_OR_TAB(*s))
819 while (SPACE_OR_TAB(*s))
827 if (!SPACE_OR_TAB(*s) && *s != '\r' && *s != '\n' && *s != '\0')
829 while (SPACE_OR_TAB(*s))
831 if (*s == '"' && (t = strchr(s+1, '"'))) {
841 while (SPACE_OR_TAB(*e) || *e == '\r' || *e == '\f')
843 if (*e != '\n' && *e != '\0')
844 return; /* false alarm */
847 const STRLEN len = t - s;
849 SV *const temp_sv = CopFILESV(PL_curcop);
855 tmplen = SvCUR(temp_sv);
861 if (tmplen > 7 && strnEQ(cf, "(eval ", 6)) {
862 /* must copy *{"::_<(eval N)[oldfilename:L]"}
863 * to *{"::_<newfilename"} */
864 /* However, the long form of evals is only turned on by the
865 debugger - usually they're "(eval %lu)" */
869 STRLEN tmplen2 = len;
870 if (tmplen + 2 <= sizeof smallbuf)
873 Newx(tmpbuf, tmplen + 2, char);
876 memcpy(tmpbuf + 2, cf, tmplen);
878 gvp = (GV**)hv_fetch(PL_defstash, tmpbuf, tmplen, FALSE);
883 if (tmplen2 + 2 <= sizeof smallbuf)
886 Newx(tmpbuf2, tmplen2 + 2, char);
888 if (tmpbuf2 != smallbuf || tmpbuf != smallbuf) {
889 /* Either they malloc'd it, or we malloc'd it,
890 so no prefix is present in ours. */
895 memcpy(tmpbuf2 + 2, s, tmplen2);
898 gv2 = *(GV**)hv_fetch(PL_defstash, tmpbuf2, tmplen2, TRUE);
900 gv_init(gv2, PL_defstash, tmpbuf2, tmplen2, FALSE);
901 /* adjust ${"::_<newfilename"} to store the new file name */
902 GvSV(gv2) = newSVpvn(tmpbuf2 + 2, tmplen2 - 2);
903 GvHV(gv2) = MUTABLE_HV(SvREFCNT_inc(GvHV(*gvp)));
904 GvAV(gv2) = MUTABLE_AV(SvREFCNT_inc(GvAV(*gvp)));
907 if (tmpbuf2 != smallbuf) Safefree(tmpbuf2);
909 if (tmpbuf != smallbuf) Safefree(tmpbuf);
912 CopFILE_free(PL_curcop);
913 CopFILE_setn(PL_curcop, s, len);
915 CopLINE_set(PL_curcop, atoi(n)-1);
919 /* skip space before PL_thistoken */
922 S_skipspace0(pTHX_ register char *s)
924 PERL_ARGS_ASSERT_SKIPSPACE0;
931 PL_thiswhite = newSVpvs("");
932 sv_catsv(PL_thiswhite, PL_skipwhite);
933 sv_free(PL_skipwhite);
936 PL_realtokenstart = s - SvPVX(PL_linestr);
940 /* skip space after PL_thistoken */
943 S_skipspace1(pTHX_ register char *s)
945 const char *start = s;
946 I32 startoff = start - SvPVX(PL_linestr);
948 PERL_ARGS_ASSERT_SKIPSPACE1;
953 start = SvPVX(PL_linestr) + startoff;
954 if (!PL_thistoken && PL_realtokenstart >= 0) {
955 const char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
956 PL_thistoken = newSVpvn(tstart, start - tstart);
958 PL_realtokenstart = -1;
961 PL_nextwhite = newSVpvs("");
962 sv_catsv(PL_nextwhite, PL_skipwhite);
963 sv_free(PL_skipwhite);
970 S_skipspace2(pTHX_ register char *s, SV **svp)
973 const I32 bufptroff = PL_bufptr - SvPVX(PL_linestr);
974 const I32 startoff = s - SvPVX(PL_linestr);
976 PERL_ARGS_ASSERT_SKIPSPACE2;
979 PL_bufptr = SvPVX(PL_linestr) + bufptroff;
980 if (!PL_madskills || !svp)
982 start = SvPVX(PL_linestr) + startoff;
983 if (!PL_thistoken && PL_realtokenstart >= 0) {
984 char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
985 PL_thistoken = newSVpvn(tstart, start - tstart);
986 PL_realtokenstart = -1;
991 sv_setsv(*svp, PL_skipwhite);
992 sv_free(PL_skipwhite);
1001 S_update_debugger_info(pTHX_ SV *orig_sv, const char *const buf, STRLEN len)
1003 AV *av = CopFILEAVx(PL_curcop);
1005 SV * const sv = newSV_type(SVt_PVMG);
1007 sv_setsv(sv, orig_sv);
1009 sv_setpvn(sv, buf, len);
1012 av_store(av, (I32)CopLINE(PL_curcop), sv);
1018 * Called to gobble the appropriate amount and type of whitespace.
1019 * Skips comments as well.
1023 S_skipspace(pTHX_ register char *s)
1028 int startoff = s - SvPVX(PL_linestr);
1030 PERL_ARGS_ASSERT_SKIPSPACE;
1033 sv_free(PL_skipwhite);
1037 PERL_ARGS_ASSERT_SKIPSPACE;
1039 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
1040 while (s < PL_bufend && SPACE_OR_TAB(*s))
1050 SSize_t oldprevlen, oldoldprevlen;
1051 SSize_t oldloplen = 0, oldunilen = 0;
1052 while (s < PL_bufend && isSPACE(*s)) {
1053 if (*s++ == '\n' && PL_in_eval && !PL_rsfp)
1058 if (s < PL_bufend && *s == '#') {
1059 while (s < PL_bufend && *s != '\n')
1061 if (s < PL_bufend) {
1063 if (PL_in_eval && !PL_rsfp) {
1070 /* only continue to recharge the buffer if we're at the end
1071 * of the buffer, we're not reading from a source filter, and
1072 * we're in normal lexing mode
1074 if (s < PL_bufend || !PL_rsfp || PL_sublex_info.sub_inwhat ||
1075 PL_lex_state == LEX_FORMLINE)
1082 /* try to recharge the buffer */
1084 curoff = s - SvPVX(PL_linestr);
1087 if ((s = filter_gets(PL_linestr, PL_rsfp,
1088 (prevlen = SvCUR(PL_linestr)))) == NULL)
1091 if (PL_madskills && curoff != startoff) {
1093 PL_skipwhite = newSVpvs("");
1094 sv_catpvn(PL_skipwhite, SvPVX(PL_linestr) + startoff,
1098 /* mustn't throw out old stuff yet if madpropping */
1099 SvCUR(PL_linestr) = curoff;
1100 s = SvPVX(PL_linestr) + curoff;
1102 if (curoff && s[-1] == '\n')
1106 /* end of file. Add on the -p or -n magic */
1107 /* XXX these shouldn't really be added here, can't set PL_faketokens */
1110 sv_catpvs(PL_linestr,
1111 ";}continue{print or die qq(-p destination: $!\\n);}");
1113 sv_setpvs(PL_linestr,
1114 ";}continue{print or die qq(-p destination: $!\\n);}");
1116 PL_minus_n = PL_minus_p = 0;
1118 else if (PL_minus_n) {
1120 sv_catpvs(PL_linestr, ";}");
1122 sv_setpvs(PL_linestr, ";}");
1128 sv_catpvs(PL_linestr,";");
1130 sv_setpvs(PL_linestr,";");
1133 /* reset variables for next time we lex */
1134 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = s = PL_linestart
1140 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
1141 PL_last_lop = PL_last_uni = NULL;
1143 /* Close the filehandle. Could be from
1144 * STDIN, or a regular file. If we were reading code from
1145 * STDIN (because the commandline held no -e or filename)
1146 * then we don't close it, we reset it so the code can
1147 * read from STDIN too.
1150 if ((PerlIO*)PL_rsfp == PerlIO_stdin())
1151 PerlIO_clearerr(PL_rsfp);
1153 (void)PerlIO_close(PL_rsfp);
1158 /* not at end of file, so we only read another line */
1159 /* make corresponding updates to old pointers, for yyerror() */
1160 oldprevlen = PL_oldbufptr - PL_bufend;
1161 oldoldprevlen = PL_oldoldbufptr - PL_bufend;
1163 oldunilen = PL_last_uni - PL_bufend;
1165 oldloplen = PL_last_lop - PL_bufend;
1166 PL_linestart = PL_bufptr = s + prevlen;
1167 PL_bufend = s + SvCUR(PL_linestr);
1169 PL_oldbufptr = s + oldprevlen;
1170 PL_oldoldbufptr = s + oldoldprevlen;
1172 PL_last_uni = s + oldunilen;
1174 PL_last_lop = s + oldloplen;
1177 /* debugger active and we're not compiling the debugger code,
1178 * so store the line into the debugger's array of lines
1180 if ((PERLDB_LINE || PERLDB_SAVESRC) && PL_curstash != PL_debstash)
1181 update_debugger_info(NULL, PL_bufptr, PL_bufend - PL_bufptr);
1188 PL_skipwhite = newSVpvs("");
1189 curoff = s - SvPVX(PL_linestr);
1190 if (curoff - startoff)
1191 sv_catpvn(PL_skipwhite, SvPVX(PL_linestr) + startoff,
1200 * Check the unary operators to ensure there's no ambiguity in how they're
1201 * used. An ambiguous piece of code would be:
1203 * This doesn't mean rand() + 5. Because rand() is a unary operator,
1204 * the +5 is its argument.
1214 if (PL_oldoldbufptr != PL_last_uni)
1216 while (isSPACE(*PL_last_uni))
1219 while (isALNUM_lazy_if(s,UTF) || *s == '-')
1221 if ((t = strchr(s, '(')) && t < PL_bufptr)
1224 if (ckWARN_d(WARN_AMBIGUOUS)){
1225 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
1226 "Warning: Use of \"%.*s\" without parentheses is ambiguous",
1227 (int)(s - PL_last_uni), PL_last_uni);
1232 * LOP : macro to build a list operator. Its behaviour has been replaced
1233 * with a subroutine, S_lop() for which LOP is just another name.
1236 #define LOP(f,x) return lop(f,x,s)
1240 * Build a list operator (or something that might be one). The rules:
1241 * - if we have a next token, then it's a list operator [why?]
1242 * - if the next thing is an opening paren, then it's a function
1243 * - else it's a list operator
1247 S_lop(pTHX_ I32 f, int x, char *s)
1251 PERL_ARGS_ASSERT_LOP;
1257 PL_last_lop = PL_oldbufptr;
1258 PL_last_lop_op = (OPCODE)f;
1261 return REPORT(LSTOP);
1264 return REPORT(LSTOP);
1267 return REPORT(FUNC);
1270 return REPORT(FUNC);
1272 return REPORT(LSTOP);
1278 * Sets up for an eventual force_next(). start_force(0) basically does
1279 * an unshift, while start_force(-1) does a push. yylex removes items
1284 S_start_force(pTHX_ int where)
1288 if (where < 0) /* so people can duplicate start_force(PL_curforce) */
1289 where = PL_lasttoke;
1290 assert(PL_curforce < 0 || PL_curforce == where);
1291 if (PL_curforce != where) {
1292 for (i = PL_lasttoke; i > where; --i) {
1293 PL_nexttoke[i] = PL_nexttoke[i-1];
1297 if (PL_curforce < 0) /* in case of duplicate start_force() */
1298 Zero(&PL_nexttoke[where], 1, NEXTTOKE);
1299 PL_curforce = where;
1302 curmad('^', newSVpvs(""));
1303 CURMAD('_', PL_nextwhite);
1308 S_curmad(pTHX_ char slot, SV *sv)
1314 if (PL_curforce < 0)
1315 where = &PL_thismad;
1317 where = &PL_nexttoke[PL_curforce].next_mad;
1323 if (UTF && is_utf8_string((U8*)SvPVX(sv), SvCUR(sv)))
1325 else if (PL_encoding) {
1326 sv_recode_to_utf8(sv, PL_encoding);
1331 /* keep a slot open for the head of the list? */
1332 if (slot != '_' && *where && (*where)->mad_key == '^') {
1333 (*where)->mad_key = slot;
1334 sv_free(MUTABLE_SV(((*where)->mad_val)));
1335 (*where)->mad_val = (void*)sv;
1338 addmad(newMADsv(slot, sv), where, 0);
1341 # define start_force(where) NOOP
1342 # define curmad(slot, sv) NOOP
1347 * When the lexer realizes it knows the next token (for instance,
1348 * it is reordering tokens for the parser) then it can call S_force_next
1349 * to know what token to return the next time the lexer is called. Caller
1350 * will need to set PL_nextval[] (or PL_nexttoke[].next_val with PERL_MAD),
1351 * and possibly PL_expect to ensure the lexer handles the token correctly.
1355 S_force_next(pTHX_ I32 type)
1360 PerlIO_printf(Perl_debug_log, "### forced token:\n");
1361 tokereport(type, &NEXTVAL_NEXTTOKE);
1365 if (PL_curforce < 0)
1366 start_force(PL_lasttoke);
1367 PL_nexttoke[PL_curforce].next_type = type;
1368 if (PL_lex_state != LEX_KNOWNEXT)
1369 PL_lex_defer = PL_lex_state;
1370 PL_lex_state = LEX_KNOWNEXT;
1371 PL_lex_expect = PL_expect;
1374 PL_nexttype[PL_nexttoke] = type;
1376 if (PL_lex_state != LEX_KNOWNEXT) {
1377 PL_lex_defer = PL_lex_state;
1378 PL_lex_expect = PL_expect;
1379 PL_lex_state = LEX_KNOWNEXT;
1385 S_newSV_maybe_utf8(pTHX_ const char *const start, STRLEN len)
1388 SV * const sv = newSVpvn_utf8(start, len,
1390 && is_utf8_string((const U8*)start, len));
1396 * When the lexer knows the next thing is a word (for instance, it has
1397 * just seen -> and it knows that the next char is a word char, then
1398 * it calls S_force_word to stick the next word into the PL_nexttoke/val
1402 * char *start : buffer position (must be within PL_linestr)
1403 * int token : PL_next* will be this type of bare word (e.g., METHOD,WORD)
1404 * int check_keyword : if true, Perl checks to make sure the word isn't
1405 * a keyword (do this if the word is a label, e.g. goto FOO)
1406 * int allow_pack : if true, : characters will also be allowed (require,
1407 * use, etc. do this)
1408 * int allow_initial_tick : used by the "sub" lexer only.
1412 S_force_word(pTHX_ register char *start, int token, int check_keyword, int allow_pack, int allow_initial_tick)
1418 PERL_ARGS_ASSERT_FORCE_WORD;
1420 start = SKIPSPACE1(start);
1422 if (isIDFIRST_lazy_if(s,UTF) ||
1423 (allow_pack && *s == ':') ||
1424 (allow_initial_tick && *s == '\'') )
1426 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, allow_pack, &len);
1427 if (check_keyword && keyword(PL_tokenbuf, len, 0))
1429 start_force(PL_curforce);
1431 curmad('X', newSVpvn(start,s-start));
1432 if (token == METHOD) {
1437 PL_expect = XOPERATOR;
1441 curmad('g', newSVpvs( "forced" ));
1442 NEXTVAL_NEXTTOKE.opval
1443 = (OP*)newSVOP(OP_CONST,0,
1444 S_newSV_maybe_utf8(aTHX_ PL_tokenbuf, len));
1445 NEXTVAL_NEXTTOKE.opval->op_private |= OPpCONST_BARE;
1453 * Called when the lexer wants $foo *foo &foo etc, but the program
1454 * text only contains the "foo" portion. The first argument is a pointer
1455 * to the "foo", and the second argument is the type symbol to prefix.
1456 * Forces the next token to be a "WORD".
1457 * Creates the symbol if it didn't already exist (via gv_fetchpv()).
1461 S_force_ident(pTHX_ register const char *s, int kind)
1465 PERL_ARGS_ASSERT_FORCE_IDENT;
1468 const STRLEN len = strlen(s);
1469 OP* const o = (OP*)newSVOP(OP_CONST, 0, newSVpvn(s, len));
1470 start_force(PL_curforce);
1471 NEXTVAL_NEXTTOKE.opval = o;
1474 o->op_private = OPpCONST_ENTERED;
1475 /* XXX see note in pp_entereval() for why we forgo typo
1476 warnings if the symbol must be introduced in an eval.
1478 gv_fetchpvn_flags(s, len,
1479 PL_in_eval ? (GV_ADDMULTI | GV_ADDINEVAL)
1481 kind == '$' ? SVt_PV :
1482 kind == '@' ? SVt_PVAV :
1483 kind == '%' ? SVt_PVHV :
1491 Perl_str_to_version(pTHX_ SV *sv)
1496 const char *start = SvPV_const(sv,len);
1497 const char * const end = start + len;
1498 const bool utf = SvUTF8(sv) ? TRUE : FALSE;
1500 PERL_ARGS_ASSERT_STR_TO_VERSION;
1502 while (start < end) {
1506 n = utf8n_to_uvchr((U8*)start, len, &skip, 0);
1511 retval += ((NV)n)/nshift;
1520 * Forces the next token to be a version number.
1521 * If the next token appears to be an invalid version number, (e.g. "v2b"),
1522 * and if "guessing" is TRUE, then no new token is created (and the caller
1523 * must use an alternative parsing method).
1527 S_force_version(pTHX_ char *s, int guessing)
1533 I32 startoff = s - SvPVX(PL_linestr);
1536 PERL_ARGS_ASSERT_FORCE_VERSION;
1544 while (isDIGIT(*d) || *d == '_' || *d == '.')
1548 start_force(PL_curforce);
1549 curmad('X', newSVpvn(s,d-s));
1552 if (*d == ';' || isSPACE(*d) || *d == '}' || !*d) {
1554 s = scan_num(s, &pl_yylval);
1555 version = pl_yylval.opval;
1556 ver = cSVOPx(version)->op_sv;
1557 if (SvPOK(ver) && !SvNIOK(ver)) {
1558 SvUPGRADE(ver, SVt_PVNV);
1559 SvNV_set(ver, str_to_version(ver));
1560 SvNOK_on(ver); /* hint that it is a version */
1563 else if (guessing) {
1566 sv_free(PL_nextwhite); /* let next token collect whitespace */
1568 s = SvPVX(PL_linestr) + startoff;
1576 if (PL_madskills && !version) {
1577 sv_free(PL_nextwhite); /* let next token collect whitespace */
1579 s = SvPVX(PL_linestr) + startoff;
1582 /* NOTE: The parser sees the package name and the VERSION swapped */
1583 start_force(PL_curforce);
1584 NEXTVAL_NEXTTOKE.opval = version;
1592 * Tokenize a quoted string passed in as an SV. It finds the next
1593 * chunk, up to end of string or a backslash. It may make a new
1594 * SV containing that chunk (if HINT_NEW_STRING is on). It also
1599 S_tokeq(pTHX_ SV *sv)
1603 register char *send;
1608 PERL_ARGS_ASSERT_TOKEQ;
1613 s = SvPV_force(sv, len);
1614 if (SvTYPE(sv) >= SVt_PVIV && SvIVX(sv) == -1)
1617 while (s < send && *s != '\\')
1622 if ( PL_hints & HINT_NEW_STRING ) {
1623 pv = newSVpvn_flags(SvPVX_const(pv), len, SVs_TEMP | SvUTF8(sv));
1627 if (s + 1 < send && (s[1] == '\\'))
1628 s++; /* all that, just for this */
1633 SvCUR_set(sv, d - SvPVX_const(sv));
1635 if ( PL_hints & HINT_NEW_STRING )
1636 return new_constant(NULL, 0, "q", sv, pv, "q", 1);
1641 * Now come three functions related to double-quote context,
1642 * S_sublex_start, S_sublex_push, and S_sublex_done. They're used when
1643 * converting things like "\u\Lgnat" into ucfirst(lc("gnat")). They
1644 * interact with PL_lex_state, and create fake ( ... ) argument lists
1645 * to handle functions and concatenation.
1646 * They assume that whoever calls them will be setting up a fake
1647 * join call, because each subthing puts a ',' after it. This lets
1650 * join($, , 'lower ', lcfirst( 'uPpEr', ) ,)
1652 * (I'm not sure whether the spurious commas at the end of lcfirst's
1653 * arguments and join's arguments are created or not).
1658 * Assumes that pl_yylval.ival is the op we're creating (e.g. OP_LCFIRST).
1660 * Pattern matching will set PL_lex_op to the pattern-matching op to
1661 * make (we return THING if pl_yylval.ival is OP_NULL, PMFUNC otherwise).
1663 * OP_CONST and OP_READLINE are easy--just make the new op and return.
1665 * Everything else becomes a FUNC.
1667 * Sets PL_lex_state to LEX_INTERPPUSH unless (ival was OP_NULL or we
1668 * had an OP_CONST or OP_READLINE). This just sets us up for a
1669 * call to S_sublex_push().
1673 S_sublex_start(pTHX)
1676 register const I32 op_type = pl_yylval.ival;
1678 if (op_type == OP_NULL) {
1679 pl_yylval.opval = PL_lex_op;
1683 if (op_type == OP_CONST || op_type == OP_READLINE) {
1684 SV *sv = tokeq(PL_lex_stuff);
1686 if (SvTYPE(sv) == SVt_PVIV) {
1687 /* Overloaded constants, nothing fancy: Convert to SVt_PV: */
1689 const char * const p = SvPV_const(sv, len);
1690 SV * const nsv = newSVpvn_flags(p, len, SvUTF8(sv));
1694 pl_yylval.opval = (OP*)newSVOP(op_type, 0, sv);
1695 PL_lex_stuff = NULL;
1696 /* Allow <FH> // "foo" */
1697 if (op_type == OP_READLINE)
1698 PL_expect = XTERMORDORDOR;
1701 else if (op_type == OP_BACKTICK && PL_lex_op) {
1702 /* readpipe() vas overriden */
1703 cSVOPx(cLISTOPx(cUNOPx(PL_lex_op)->op_first)->op_first->op_sibling)->op_sv = tokeq(PL_lex_stuff);
1704 pl_yylval.opval = PL_lex_op;
1706 PL_lex_stuff = NULL;
1710 PL_sublex_info.super_state = PL_lex_state;
1711 PL_sublex_info.sub_inwhat = (U16)op_type;
1712 PL_sublex_info.sub_op = PL_lex_op;
1713 PL_lex_state = LEX_INTERPPUSH;
1717 pl_yylval.opval = PL_lex_op;
1727 * Create a new scope to save the lexing state. The scope will be
1728 * ended in S_sublex_done. Returns a '(', starting the function arguments
1729 * to the uc, lc, etc. found before.
1730 * Sets PL_lex_state to LEX_INTERPCONCAT.
1739 PL_lex_state = PL_sublex_info.super_state;
1740 SAVEBOOL(PL_lex_dojoin);
1741 SAVEI32(PL_lex_brackets);
1742 SAVEI32(PL_lex_casemods);
1743 SAVEI32(PL_lex_starts);
1744 SAVEI8(PL_lex_state);
1745 SAVEVPTR(PL_lex_inpat);
1746 SAVEI16(PL_lex_inwhat);
1747 SAVECOPLINE(PL_curcop);
1748 SAVEPPTR(PL_bufptr);
1749 SAVEPPTR(PL_bufend);
1750 SAVEPPTR(PL_oldbufptr);
1751 SAVEPPTR(PL_oldoldbufptr);
1752 SAVEPPTR(PL_last_lop);
1753 SAVEPPTR(PL_last_uni);
1754 SAVEPPTR(PL_linestart);
1755 SAVESPTR(PL_linestr);
1756 SAVEGENERICPV(PL_lex_brackstack);
1757 SAVEGENERICPV(PL_lex_casestack);
1759 PL_linestr = PL_lex_stuff;
1760 PL_lex_stuff = NULL;
1762 PL_bufend = PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart
1763 = SvPVX(PL_linestr);
1764 PL_bufend += SvCUR(PL_linestr);
1765 PL_last_lop = PL_last_uni = NULL;
1766 SAVEFREESV(PL_linestr);
1768 PL_lex_dojoin = FALSE;
1769 PL_lex_brackets = 0;
1770 Newx(PL_lex_brackstack, 120, char);
1771 Newx(PL_lex_casestack, 12, char);
1772 PL_lex_casemods = 0;
1773 *PL_lex_casestack = '\0';
1775 PL_lex_state = LEX_INTERPCONCAT;
1776 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
1778 PL_lex_inwhat = PL_sublex_info.sub_inwhat;
1779 if (PL_lex_inwhat == OP_MATCH || PL_lex_inwhat == OP_QR || PL_lex_inwhat == OP_SUBST)
1780 PL_lex_inpat = PL_sublex_info.sub_op;
1782 PL_lex_inpat = NULL;
1789 * Restores lexer state after a S_sublex_push.
1796 if (!PL_lex_starts++) {
1797 SV * const sv = newSVpvs("");
1798 if (SvUTF8(PL_linestr))
1800 PL_expect = XOPERATOR;
1801 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
1805 if (PL_lex_casemods) { /* oops, we've got some unbalanced parens */
1806 PL_lex_state = LEX_INTERPCASEMOD;
1810 /* Is there a right-hand side to take care of? (s//RHS/ or tr//RHS/) */
1811 if (PL_lex_repl && (PL_lex_inwhat == OP_SUBST || PL_lex_inwhat == OP_TRANS)) {
1812 PL_linestr = PL_lex_repl;
1814 PL_bufend = PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart = SvPVX(PL_linestr);
1815 PL_bufend += SvCUR(PL_linestr);
1816 PL_last_lop = PL_last_uni = NULL;
1817 SAVEFREESV(PL_linestr);
1818 PL_lex_dojoin = FALSE;
1819 PL_lex_brackets = 0;
1820 PL_lex_casemods = 0;
1821 *PL_lex_casestack = '\0';
1823 if (SvEVALED(PL_lex_repl)) {
1824 PL_lex_state = LEX_INTERPNORMAL;
1826 /* we don't clear PL_lex_repl here, so that we can check later
1827 whether this is an evalled subst; that means we rely on the
1828 logic to ensure sublex_done() is called again only via the
1829 branch (in yylex()) that clears PL_lex_repl, else we'll loop */
1832 PL_lex_state = LEX_INTERPCONCAT;
1842 PL_endwhite = newSVpvs("");
1843 sv_catsv(PL_endwhite, PL_thiswhite);
1847 sv_setpvs(PL_thistoken,"");
1849 PL_realtokenstart = -1;
1853 PL_bufend = SvPVX(PL_linestr);
1854 PL_bufend += SvCUR(PL_linestr);
1855 PL_expect = XOPERATOR;
1856 PL_sublex_info.sub_inwhat = 0;
1864 Extracts a pattern, double-quoted string, or transliteration. This
1867 It looks at PL_lex_inwhat and PL_lex_inpat to find out whether it's
1868 processing a pattern (PL_lex_inpat is true), a transliteration
1869 (PL_lex_inwhat == OP_TRANS is true), or a double-quoted string.
1871 Returns a pointer to the character scanned up to. If this is
1872 advanced from the start pointer supplied (i.e. if anything was
1873 successfully parsed), will leave an OP for the substring scanned
1874 in pl_yylval. Caller must intuit reason for not parsing further
1875 by looking at the next characters herself.
1879 double-quoted style: \r and \n
1880 regexp special ones: \D \s
1883 case and quoting: \U \Q \E
1884 stops on @ and $, but not for $ as tail anchor
1886 In transliterations:
1887 characters are VERY literal, except for - not at the start or end
1888 of the string, which indicates a range. If the range is in bytes,
1889 scan_const expands the range to the full set of intermediate
1890 characters. If the range is in utf8, the hyphen is replaced with
1891 a certain range mark which will be handled by pmtrans() in op.c.
1893 In double-quoted strings:
1895 double-quoted style: \r and \n
1897 deprecated backrefs: \1 (in substitution replacements)
1898 case and quoting: \U \Q \E
1901 scan_const does *not* construct ops to handle interpolated strings.
1902 It stops processing as soon as it finds an embedded $ or @ variable
1903 and leaves it to the caller to work out what's going on.
1905 embedded arrays (whether in pattern or not) could be:
1906 @foo, @::foo, @'foo, @{foo}, @$foo, @+, @-.
1908 $ in double-quoted strings must be the symbol of an embedded scalar.
1910 $ in pattern could be $foo or could be tail anchor. Assumption:
1911 it's a tail anchor if $ is the last thing in the string, or if it's
1912 followed by one of "()| \r\n\t"
1914 \1 (backreferences) are turned into $1
1916 The structure of the code is
1917 while (there's a character to process) {
1918 handle transliteration ranges
1919 skip regexp comments /(?#comment)/ and codes /(?{code})/
1920 skip #-initiated comments in //x patterns
1921 check for embedded arrays
1922 check for embedded scalars
1924 leave intact backslashes from leaveit (below)
1925 deprecate \1 in substitution replacements
1926 handle string-changing backslashes \l \U \Q \E, etc.
1927 switch (what was escaped) {
1928 handle \- in a transliteration (becomes a literal -)
1929 handle \132 (octal characters)
1930 handle \x15 and \x{1234} (hex characters)
1931 handle \N{name} (named characters)
1932 handle \cV (control characters)
1933 handle printf-style backslashes (\f, \r, \n, etc)
1936 } (end if backslash)
1937 handle regular character
1938 } (end while character to read)
1943 S_scan_const(pTHX_ char *start)
1946 register char *send = PL_bufend; /* end of the constant */
1947 SV *sv = newSV(send - start); /* sv for the constant. See
1948 note below on sizing. */
1949 register char *s = start; /* start of the constant */
1950 register char *d = SvPVX(sv); /* destination for copies */
1951 bool dorange = FALSE; /* are we in a translit range? */
1952 bool didrange = FALSE; /* did we just finish a range? */
1953 I32 has_utf8 = FALSE; /* Output constant is UTF8 */
1954 I32 this_utf8 = UTF; /* Is the source string assumed
1955 to be UTF8? But, this can
1956 show as true when the source
1957 isn't utf8, as for example
1958 when it is entirely composed
1961 /* Note on sizing: The scanned constant is placed into sv, which is
1962 * initialized by newSV() assuming one byte of output for every byte of
1963 * input. This routine expects newSV() to allocate an extra byte for a
1964 * trailing NUL, which this routine will append if it gets to the end of
1965 * the input. There may be more bytes of input than output (eg., \N{LATIN
1966 * CAPITAL LETTER A}), or more output than input if the constant ends up
1967 * recoded to utf8, but each time a construct is found that might increase
1968 * the needed size, SvGROW() is called. Its size parameter each time is
1969 * based on the best guess estimate at the time, namely the length used so
1970 * far, plus the length the current construct will occupy, plus room for
1971 * the trailing NUL, plus one byte for every input byte still unscanned */
1975 UV literal_endpoint = 0;
1976 bool native_range = TRUE; /* turned to FALSE if the first endpoint is Unicode. */
1979 PERL_ARGS_ASSERT_SCAN_CONST;
1981 if (PL_lex_inwhat == OP_TRANS && PL_sublex_info.sub_op) {
1982 /* If we are doing a trans and we know we want UTF8 set expectation */
1983 has_utf8 = PL_sublex_info.sub_op->op_private & (OPpTRANS_FROM_UTF|OPpTRANS_TO_UTF);
1984 this_utf8 = PL_sublex_info.sub_op->op_private & (PL_lex_repl ? OPpTRANS_FROM_UTF : OPpTRANS_TO_UTF);
1988 while (s < send || dorange) {
1989 /* get transliterations out of the way (they're most literal) */
1990 if (PL_lex_inwhat == OP_TRANS) {
1991 /* expand a range A-Z to the full set of characters. AIE! */
1993 I32 i; /* current expanded character */
1994 I32 min; /* first character in range */
1995 I32 max; /* last character in range */
2006 char * const c = (char*)utf8_hop((U8*)d, -1);
2010 *c = (char)UTF_TO_NATIVE(0xff);
2011 /* mark the range as done, and continue */
2017 i = d - SvPVX_const(sv); /* remember current offset */
2020 SvLEN(sv) + (has_utf8 ?
2021 (512 - UTF_CONTINUATION_MARK +
2024 /* How many two-byte within 0..255: 128 in UTF-8,
2025 * 96 in UTF-8-mod. */
2027 SvGROW(sv, SvLEN(sv) + 256); /* never more than 256 chars in a range */
2029 d = SvPVX(sv) + i; /* refresh d after realloc */
2033 for (j = 0; j <= 1; j++) {
2034 char * const c = (char*)utf8_hop((U8*)d, -1);
2035 const UV uv = utf8n_to_uvchr((U8*)c, d - c, NULL, 0);
2041 max = (U8)0xff; /* only to \xff */
2042 uvmax = uv; /* \x{100} to uvmax */
2044 d = c; /* eat endpoint chars */
2049 d -= 2; /* eat the first char and the - */
2050 min = (U8)*d; /* first char in range */
2051 max = (U8)d[1]; /* last char in range */
2058 "Invalid range \"%c-%c\" in transliteration operator",
2059 (char)min, (char)max);
2063 if (literal_endpoint == 2 &&
2064 ((isLOWER(min) && isLOWER(max)) ||
2065 (isUPPER(min) && isUPPER(max)))) {
2067 for (i = min; i <= max; i++)
2069 *d++ = NATIVE_TO_NEED(has_utf8,i);
2071 for (i = min; i <= max; i++)
2073 *d++ = NATIVE_TO_NEED(has_utf8,i);
2078 for (i = min; i <= max; i++)
2081 const U8 ch = (U8)NATIVE_TO_UTF(i);
2082 if (UNI_IS_INVARIANT(ch))
2085 *d++ = (U8)UTF8_EIGHT_BIT_HI(ch);
2086 *d++ = (U8)UTF8_EIGHT_BIT_LO(ch);
2095 d = (char*)uvchr_to_utf8((U8*)d, 0x100);
2097 *d++ = (char)UTF_TO_NATIVE(0xff);
2099 d = (char*)uvchr_to_utf8((U8*)d, uvmax);
2103 /* mark the range as done, and continue */
2107 literal_endpoint = 0;
2112 /* range begins (ignore - as first or last char) */
2113 else if (*s == '-' && s+1 < send && s != start) {
2115 Perl_croak(aTHX_ "Ambiguous range in transliteration operator");
2122 *d++ = (char)UTF_TO_NATIVE(0xff); /* use illegal utf8 byte--see pmtrans */
2132 literal_endpoint = 0;
2133 native_range = TRUE;
2138 /* if we get here, we're not doing a transliteration */
2140 /* skip for regexp comments /(?#comment)/ and code /(?{code})/,
2141 except for the last char, which will be done separately. */
2142 else if (*s == '(' && PL_lex_inpat && s[1] == '?') {
2144 while (s+1 < send && *s != ')')
2145 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2147 else if (s[2] == '{' /* This should match regcomp.c */
2148 || (s[2] == '?' && s[3] == '{'))
2151 char *regparse = s + (s[2] == '{' ? 3 : 4);
2154 while (count && (c = *regparse)) {
2155 if (c == '\\' && regparse[1])
2163 if (*regparse != ')')
2164 regparse--; /* Leave one char for continuation. */
2165 while (s < regparse)
2166 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2170 /* likewise skip #-initiated comments in //x patterns */
2171 else if (*s == '#' && PL_lex_inpat &&
2172 ((PMOP*)PL_lex_inpat)->op_pmflags & PMf_EXTENDED) {
2173 while (s+1 < send && *s != '\n')
2174 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2177 /* check for embedded arrays
2178 (@foo, @::foo, @'foo, @{foo}, @$foo, @+, @-)
2180 else if (*s == '@' && s[1]) {
2181 if (isALNUM_lazy_if(s+1,UTF))
2183 if (strchr(":'{$", s[1]))
2185 if (!PL_lex_inpat && (s[1] == '+' || s[1] == '-'))
2186 break; /* in regexp, neither @+ nor @- are interpolated */
2189 /* check for embedded scalars. only stop if we're sure it's a
2192 else if (*s == '$') {
2193 if (!PL_lex_inpat) /* not a regexp, so $ must be var */
2195 if (s + 1 < send && !strchr("()| \r\n\t", s[1])) {
2196 if (s[1] == '\\' && ckWARN(WARN_AMBIGUOUS)) {
2197 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
2198 "Possible unintended interpolation of $\\ in regex");
2200 break; /* in regexp, $ might be tail anchor */
2204 /* End of else if chain - OP_TRANS rejoin rest */
2207 if (*s == '\\' && s+1 < send) {
2210 /* deprecate \1 in strings and substitution replacements */
2211 if (PL_lex_inwhat == OP_SUBST && !PL_lex_inpat &&
2212 isDIGIT(*s) && *s != '0' && !isDIGIT(s[1]))
2214 if (ckWARN(WARN_SYNTAX))
2215 Perl_warner(aTHX_ packWARN(WARN_SYNTAX), "\\%c better written as $%c", *s, *s);
2220 /* string-change backslash escapes */
2221 if (PL_lex_inwhat != OP_TRANS && *s && strchr("lLuUEQ", *s)) {
2225 /* skip any other backslash escapes in a pattern */
2226 else if (PL_lex_inpat) {
2227 *d++ = NATIVE_TO_NEED(has_utf8,'\\');
2228 goto default_action;
2231 /* if we get here, it's either a quoted -, or a digit */
2234 /* quoted - in transliterations */
2236 if (PL_lex_inwhat == OP_TRANS) {
2243 if ((isALPHA(*s) || isDIGIT(*s)) &&
2245 Perl_warner(aTHX_ packWARN(WARN_MISC),
2246 "Unrecognized escape \\%c passed through",
2248 /* default action is to copy the quoted character */
2249 goto default_action;
2252 /* eg. \132 indicates the octal constant 0x132 */
2253 case '0': case '1': case '2': case '3':
2254 case '4': case '5': case '6': case '7':
2258 uv = NATIVE_TO_UNI(grok_oct(s, &len, &flags, NULL));
2261 goto NUM_ESCAPE_INSERT;
2263 /* eg. \x24 indicates the hex constant 0x24 */
2267 char* const e = strchr(s, '}');
2268 I32 flags = PERL_SCAN_ALLOW_UNDERSCORES |
2269 PERL_SCAN_DISALLOW_PREFIX;
2274 yyerror("Missing right brace on \\x{}");
2278 uv = NATIVE_TO_UNI(grok_hex(s, &len, &flags, NULL));
2284 I32 flags = PERL_SCAN_DISALLOW_PREFIX;
2285 uv = NATIVE_TO_UNI(grok_hex(s, &len, &flags, NULL));
2291 /* Insert oct, hex, or \N{U+...} escaped character. There will
2292 * always be enough room in sv since such escapes will be
2293 * longer than any UTF-8 sequence they can end up as, except if
2294 * they force us to recode the rest of the string into utf8 */
2296 /* Here uv is the ordinal of the next character being added in
2297 * unicode (converted from native). (It has to be done before
2298 * here because \N is interpreted as unicode, and oct and hex
2300 if (!UNI_IS_INVARIANT(uv)) {
2301 if (!has_utf8 && uv > 255) {
2302 /* Might need to recode whatever we have accumulated so
2303 * far if it contains any chars variant in utf8 or
2306 SvCUR_set(sv, d - SvPVX_const(sv));
2309 /* See Note on sizing above. */
2310 sv_utf8_upgrade_flags_grow(sv,
2311 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
2312 UNISKIP(uv) + (STRLEN)(send - s) + 1);
2313 d = SvPVX(sv) + SvCUR(sv);
2318 d = (char*)uvuni_to_utf8((U8*)d, uv);
2319 if (PL_lex_inwhat == OP_TRANS &&
2320 PL_sublex_info.sub_op) {
2321 PL_sublex_info.sub_op->op_private |=
2322 (PL_lex_repl ? OPpTRANS_FROM_UTF
2326 if (uv > 255 && !dorange)
2327 native_range = FALSE;
2339 /* \N{LATIN SMALL LETTER A} is a named character, and so is
2344 char* e = strchr(s, '}');
2350 yyerror("Missing right brace on \\N{}");
2354 if (e > s + 2 && s[1] == 'U' && s[2] == '+') {
2355 /* \N{U+...} The ... is a unicode value even on EBCDIC
2357 I32 flags = PERL_SCAN_ALLOW_UNDERSCORES |
2358 PERL_SCAN_DISALLOW_PREFIX;
2361 uv = grok_hex(s, &len, &flags, NULL);
2362 if ( e > s && len != (STRLEN)(e - s) ) {
2366 goto NUM_ESCAPE_INSERT;
2368 res = newSVpvn(s + 1, e - s - 1);
2369 res = new_constant( NULL, 0, "charnames",
2370 res, NULL, s - 2, e - s + 3 );
2372 sv_utf8_upgrade(res);
2373 str = SvPV_const(res,len);
2374 #ifdef EBCDIC_NEVER_MIND
2375 /* charnames uses pack U and that has been
2376 * recently changed to do the below uni->native
2377 * mapping, so this would be redundant (and wrong,
2378 * the code point would be doubly converted).
2379 * But leave this in just in case the pack U change
2380 * gets revoked, but the semantics is still
2381 * desireable for charnames. --jhi */
2383 UV uv = utf8_to_uvchr((const U8*)str, 0);
2386 U8 tmpbuf[UTF8_MAXBYTES+1], *d;
2388 d = uvchr_to_utf8(tmpbuf, UNI_TO_NATIVE(uv));
2389 sv_setpvn(res, (char *)tmpbuf, d - tmpbuf);
2390 str = SvPV_const(res, len);
2394 /* If destination is not in utf8 but this new character is,
2395 * recode the dest to utf8 */
2396 if (!has_utf8 && SvUTF8(res)) {
2397 SvCUR_set(sv, d - SvPVX_const(sv));
2400 /* See Note on sizing above. */
2401 sv_utf8_upgrade_flags_grow(sv,
2402 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
2403 len + (STRLEN)(send - s) + 1);
2404 d = SvPVX(sv) + SvCUR(sv);
2406 } else if (len > (STRLEN)(e - s + 4)) { /* I _guess_ 4 is \N{} --jhi */
2408 /* See Note on sizing above. (NOTE: SvCUR() is not set
2409 * correctly here). */
2410 const STRLEN off = d - SvPVX_const(sv);
2411 d = SvGROW(sv, off + len + (STRLEN)(send - s) + 1) + off;
2415 native_range = FALSE; /* \N{} is guessed to be Unicode */
2417 Copy(str, d, len, char);
2424 yyerror("Missing braces on \\N{}");
2427 /* \c is a control character */
2436 *d++ = NATIVE_TO_NEED(has_utf8,toCTRL(c));
2439 yyerror("Missing control char name in \\c");
2443 /* printf-style backslashes, formfeeds, newlines, etc */
2445 *d++ = NATIVE_TO_NEED(has_utf8,'\b');
2448 *d++ = NATIVE_TO_NEED(has_utf8,'\n');
2451 *d++ = NATIVE_TO_NEED(has_utf8,'\r');
2454 *d++ = NATIVE_TO_NEED(has_utf8,'\f');
2457 *d++ = NATIVE_TO_NEED(has_utf8,'\t');
2460 *d++ = ASCII_TO_NEED(has_utf8,'\033');
2463 *d++ = ASCII_TO_NEED(has_utf8,'\007');
2469 } /* end if (backslash) */
2476 /* If we started with encoded form, or already know we want it,
2477 then encode the next character */
2478 if (! NATIVE_IS_INVARIANT((U8)(*s)) && (this_utf8 || has_utf8)) {
2482 /* One might think that it is wasted effort in the case of the
2483 * source being utf8 (this_utf8 == TRUE) to take the next character
2484 * in the source, convert it to an unsigned value, and then convert
2485 * it back again. But the source has not been validated here. The
2486 * routine that does the conversion checks for errors like
2489 const UV nextuv = (this_utf8) ? utf8n_to_uvchr((U8*)s, send - s, &len, 0) : (UV) ((U8) *s);
2490 const STRLEN need = UNISKIP(NATIVE_TO_UNI(nextuv));
2492 SvCUR_set(sv, d - SvPVX_const(sv));
2495 /* See Note on sizing above. */
2496 sv_utf8_upgrade_flags_grow(sv,
2497 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
2498 need + (STRLEN)(send - s) + 1);
2499 d = SvPVX(sv) + SvCUR(sv);
2501 } else if (need > len) {
2502 /* encoded value larger than old, may need extra space (NOTE:
2503 * SvCUR() is not set correctly here). See Note on sizing
2505 const STRLEN off = d - SvPVX_const(sv);
2506 d = SvGROW(sv, off + need + (STRLEN)(send - s) + 1) + off;
2510 d = (char*)uvchr_to_utf8((U8*)d, nextuv);
2512 if (uv > 255 && !dorange)
2513 native_range = FALSE;
2517 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2519 } /* while loop to process each character */
2521 /* terminate the string and set up the sv */
2523 SvCUR_set(sv, d - SvPVX_const(sv));
2524 if (SvCUR(sv) >= SvLEN(sv))
2525 Perl_croak(aTHX_ "panic: constant overflowed allocated space");
2528 if (PL_encoding && !has_utf8) {
2529 sv_recode_to_utf8(sv, PL_encoding);
2535 if (PL_lex_inwhat == OP_TRANS && PL_sublex_info.sub_op) {
2536 PL_sublex_info.sub_op->op_private |=
2537 (PL_lex_repl ? OPpTRANS_FROM_UTF : OPpTRANS_TO_UTF);
2541 /* shrink the sv if we allocated more than we used */
2542 if (SvCUR(sv) + 5 < SvLEN(sv)) {
2543 SvPV_shrink_to_cur(sv);
2546 /* return the substring (via pl_yylval) only if we parsed anything */
2547 if (s > PL_bufptr) {
2548 if ( PL_hints & ( PL_lex_inpat ? HINT_NEW_RE : HINT_NEW_STRING ) ) {
2549 const char *const key = PL_lex_inpat ? "qr" : "q";
2550 const STRLEN keylen = PL_lex_inpat ? 2 : 1;
2554 if (PL_lex_inwhat == OP_TRANS) {
2557 } else if (PL_lex_inwhat == OP_SUBST && !PL_lex_inpat) {
2565 sv = S_new_constant(aTHX_ start, s - start, key, keylen, sv, NULL,
2568 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
2575 * Returns TRUE if there's more to the expression (e.g., a subscript),
2578 * It deals with "$foo[3]" and /$foo[3]/ and /$foo[0123456789$]+/
2580 * ->[ and ->{ return TRUE
2581 * { and [ outside a pattern are always subscripts, so return TRUE
2582 * if we're outside a pattern and it's not { or [, then return FALSE
2583 * if we're in a pattern and the first char is a {
2584 * {4,5} (any digits around the comma) returns FALSE
2585 * if we're in a pattern and the first char is a [
2587 * [SOMETHING] has a funky algorithm to decide whether it's a
2588 * character class or not. It has to deal with things like
2589 * /$foo[-3]/ and /$foo[$bar]/ as well as /$foo[$\d]+/
2590 * anything else returns TRUE
2593 /* This is the one truly awful dwimmer necessary to conflate C and sed. */
2596 S_intuit_more(pTHX_ register char *s)
2600 PERL_ARGS_ASSERT_INTUIT_MORE;
2602 if (PL_lex_brackets)
2604 if (*s == '-' && s[1] == '>' && (s[2] == '[' || s[2] == '{'))
2606 if (*s != '{' && *s != '[')
2611 /* In a pattern, so maybe we have {n,m}. */
2628 /* On the other hand, maybe we have a character class */
2631 if (*s == ']' || *s == '^')
2634 /* this is terrifying, and it works */
2635 int weight = 2; /* let's weigh the evidence */
2637 unsigned char un_char = 255, last_un_char;
2638 const char * const send = strchr(s,']');
2639 char tmpbuf[sizeof PL_tokenbuf * 4];
2641 if (!send) /* has to be an expression */
2644 Zero(seen,256,char);
2647 else if (isDIGIT(*s)) {
2649 if (isDIGIT(s[1]) && s[2] == ']')
2655 for (; s < send; s++) {
2656 last_un_char = un_char;
2657 un_char = (unsigned char)*s;
2662 weight -= seen[un_char] * 10;
2663 if (isALNUM_lazy_if(s+1,UTF)) {
2665 scan_ident(s, send, tmpbuf, sizeof tmpbuf, FALSE);
2666 len = (int)strlen(tmpbuf);
2667 if (len > 1 && gv_fetchpvn_flags(tmpbuf, len, 0, SVt_PV))
2672 else if (*s == '$' && s[1] &&
2673 strchr("[#!%*<>()-=",s[1])) {
2674 if (/*{*/ strchr("])} =",s[2]))
2683 if (strchr("wds]",s[1]))
2685 else if (seen[(U8)'\''] || seen[(U8)'"'])
2687 else if (strchr("rnftbxcav",s[1]))
2689 else if (isDIGIT(s[1])) {
2691 while (s[1] && isDIGIT(s[1]))
2701 if (strchr("aA01! ",last_un_char))
2703 if (strchr("zZ79~",s[1]))
2705 if (last_un_char == 255 && (isDIGIT(s[1]) || s[1] == '$'))
2706 weight -= 5; /* cope with negative subscript */
2709 if (!isALNUM(last_un_char)
2710 && !(last_un_char == '$' || last_un_char == '@'
2711 || last_un_char == '&')
2712 && isALPHA(*s) && s[1] && isALPHA(s[1])) {
2717 if (keyword(tmpbuf, d - tmpbuf, 0))
2720 if (un_char == last_un_char + 1)
2722 weight -= seen[un_char];
2727 if (weight >= 0) /* probably a character class */
2737 * Does all the checking to disambiguate
2739 * between foo(bar) and bar->foo. Returns 0 if not a method, otherwise
2740 * FUNCMETH (bar->foo(args)) or METHOD (bar->foo args).
2742 * First argument is the stuff after the first token, e.g. "bar".
2744 * Not a method if bar is a filehandle.
2745 * Not a method if foo is a subroutine prototyped to take a filehandle.
2746 * Not a method if it's really "Foo $bar"
2747 * Method if it's "foo $bar"
2748 * Not a method if it's really "print foo $bar"
2749 * Method if it's really "foo package::" (interpreted as package->foo)
2750 * Not a method if bar is known to be a subroutine ("sub bar; foo bar")
2751 * Not a method if bar is a filehandle or package, but is quoted with
2756 S_intuit_method(pTHX_ char *start, GV *gv, CV *cv)
2759 char *s = start + (*start == '$');
2760 char tmpbuf[sizeof PL_tokenbuf];
2767 PERL_ARGS_ASSERT_INTUIT_METHOD;
2770 if (SvTYPE(gv) == SVt_PVGV && GvIO(gv))
2774 const char *proto = SvPVX_const(cv);
2785 s = scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
2786 /* start is the beginning of the possible filehandle/object,
2787 * and s is the end of it
2788 * tmpbuf is a copy of it
2791 if (*start == '$') {
2792 if (gv || PL_last_lop_op == OP_PRINT || PL_last_lop_op == OP_SAY ||
2793 isUPPER(*PL_tokenbuf))
2796 len = start - SvPVX(PL_linestr);
2800 start = SvPVX(PL_linestr) + len;
2804 return *s == '(' ? FUNCMETH : METHOD;
2806 if (!keyword(tmpbuf, len, 0)) {
2807 if (len > 2 && tmpbuf[len - 2] == ':' && tmpbuf[len - 1] == ':') {
2811 soff = s - SvPVX(PL_linestr);
2815 indirgv = gv_fetchpvn_flags(tmpbuf, len, 0, SVt_PVCV);
2816 if (indirgv && GvCVu(indirgv))
2818 /* filehandle or package name makes it a method */
2819 if (!gv || GvIO(indirgv) || gv_stashpvn(tmpbuf, len, 0)) {
2821 soff = s - SvPVX(PL_linestr);
2824 if ((PL_bufend - s) >= 2 && *s == '=' && *(s+1) == '>')
2825 return 0; /* no assumptions -- "=>" quotes bearword */
2827 start_force(PL_curforce);
2828 NEXTVAL_NEXTTOKE.opval = (OP*)newSVOP(OP_CONST, 0,
2829 newSVpvn(tmpbuf,len));
2830 NEXTVAL_NEXTTOKE.opval->op_private = OPpCONST_BARE;
2832 curmad('X', newSVpvn(start,SvPVX(PL_linestr) + soff - start));
2837 PL_bufptr = SvPVX(PL_linestr) + soff; /* restart before space */
2839 return *s == '(' ? FUNCMETH : METHOD;
2845 /* Encoded script support. filter_add() effectively inserts a
2846 * 'pre-processing' function into the current source input stream.
2847 * Note that the filter function only applies to the current source file
2848 * (e.g., it will not affect files 'require'd or 'use'd by this one).
2850 * The datasv parameter (which may be NULL) can be used to pass
2851 * private data to this instance of the filter. The filter function
2852 * can recover the SV using the FILTER_DATA macro and use it to
2853 * store private buffers and state information.
2855 * The supplied datasv parameter is upgraded to a PVIO type
2856 * and the IoDIRP/IoANY field is used to store the function pointer,
2857 * and IOf_FAKE_DIRP is enabled on datasv to mark this as such.
2858 * Note that IoTOP_NAME, IoFMT_NAME, IoBOTTOM_NAME, if set for
2859 * private use must be set using malloc'd pointers.
2863 Perl_filter_add(pTHX_ filter_t funcp, SV *datasv)
2872 if (!PL_rsfp_filters)
2873 PL_rsfp_filters = newAV();
2876 SvUPGRADE(datasv, SVt_PVIO);
2877 IoANY(datasv) = FPTR2DPTR(void *, funcp); /* stash funcp into spare field */
2878 IoFLAGS(datasv) |= IOf_FAKE_DIRP;
2879 DEBUG_P(PerlIO_printf(Perl_debug_log, "filter_add func %p (%s)\n",
2880 FPTR2DPTR(void *, IoANY(datasv)),
2881 SvPV_nolen(datasv)));
2882 av_unshift(PL_rsfp_filters, 1);
2883 av_store(PL_rsfp_filters, 0, datasv) ;
2888 /* Delete most recently added instance of this filter function. */
2890 Perl_filter_del(pTHX_ filter_t funcp)
2895 PERL_ARGS_ASSERT_FILTER_DEL;
2898 DEBUG_P(PerlIO_printf(Perl_debug_log, "filter_del func %p",
2899 FPTR2DPTR(void*, funcp)));
2901 if (!PL_parser || !PL_rsfp_filters || AvFILLp(PL_rsfp_filters)<0)
2903 /* if filter is on top of stack (usual case) just pop it off */
2904 datasv = FILTER_DATA(AvFILLp(PL_rsfp_filters));
2905 if (IoANY(datasv) == FPTR2DPTR(void *, funcp)) {
2906 IoFLAGS(datasv) &= ~IOf_FAKE_DIRP;
2907 IoANY(datasv) = (void *)NULL;
2908 sv_free(av_pop(PL_rsfp_filters));
2912 /* we need to search for the correct entry and clear it */
2913 Perl_die(aTHX_ "filter_del can only delete in reverse order (currently)");
2917 /* Invoke the idxth filter function for the current rsfp. */
2918 /* maxlen 0 = read one text line */
2920 Perl_filter_read(pTHX_ int idx, SV *buf_sv, int maxlen)
2925 /* This API is bad. It should have been using unsigned int for maxlen.
2926 Not sure if we want to change the API, but if not we should sanity
2927 check the value here. */
2928 const unsigned int correct_length
2937 PERL_ARGS_ASSERT_FILTER_READ;
2939 if (!PL_parser || !PL_rsfp_filters)
2941 if (idx > AvFILLp(PL_rsfp_filters)) { /* Any more filters? */
2942 /* Provide a default input filter to make life easy. */
2943 /* Note that we append to the line. This is handy. */
2944 DEBUG_P(PerlIO_printf(Perl_debug_log,
2945 "filter_read %d: from rsfp\n", idx));
2946 if (correct_length) {
2949 const int old_len = SvCUR(buf_sv);
2951 /* ensure buf_sv is large enough */
2952 SvGROW(buf_sv, (STRLEN)(old_len + correct_length)) ;
2953 if ((len = PerlIO_read(PL_rsfp, SvPVX(buf_sv) + old_len,
2954 correct_length)) <= 0) {
2955 if (PerlIO_error(PL_rsfp))
2956 return -1; /* error */
2958 return 0 ; /* end of file */
2960 SvCUR_set(buf_sv, old_len + len) ;
2963 if (sv_gets(buf_sv, PL_rsfp, SvCUR(buf_sv)) == NULL) {
2964 if (PerlIO_error(PL_rsfp))
2965 return -1; /* error */
2967 return 0 ; /* end of file */
2970 return SvCUR(buf_sv);
2972 /* Skip this filter slot if filter has been deleted */
2973 if ( (datasv = FILTER_DATA(idx)) == &PL_sv_undef) {
2974 DEBUG_P(PerlIO_printf(Perl_debug_log,
2975 "filter_read %d: skipped (filter deleted)\n",
2977 return FILTER_READ(idx+1, buf_sv, correct_length); /* recurse */
2979 /* Get function pointer hidden within datasv */
2980 funcp = DPTR2FPTR(filter_t, IoANY(datasv));
2981 DEBUG_P(PerlIO_printf(Perl_debug_log,
2982 "filter_read %d: via function %p (%s)\n",
2983 idx, (void*)datasv, SvPV_nolen_const(datasv)));
2984 /* Call function. The function is expected to */
2985 /* call "FILTER_READ(idx+1, buf_sv)" first. */
2986 /* Return: <0:error, =0:eof, >0:not eof */
2987 return (*funcp)(aTHX_ idx, buf_sv, correct_length);
2991 S_filter_gets(pTHX_ register SV *sv, register PerlIO *fp, STRLEN append)
2995 PERL_ARGS_ASSERT_FILTER_GETS;
2997 #ifdef PERL_CR_FILTER
2998 if (!PL_rsfp_filters) {
2999 filter_add(S_cr_textfilter,NULL);
3002 if (PL_rsfp_filters) {
3004 SvCUR_set(sv, 0); /* start with empty line */
3005 if (FILTER_READ(0, sv, 0) > 0)
3006 return ( SvPVX(sv) ) ;
3011 return (sv_gets(sv, fp, append));
3015 S_find_in_my_stash(pTHX_ const char *pkgname, STRLEN len)
3020 PERL_ARGS_ASSERT_FIND_IN_MY_STASH;
3022 if (len == 11 && *pkgname == '_' && strEQ(pkgname, "__PACKAGE__"))
3026 (pkgname[len - 2] == ':' && pkgname[len - 1] == ':') &&
3027 (gv = gv_fetchpvn_flags(pkgname, len, 0, SVt_PVHV)))
3029 return GvHV(gv); /* Foo:: */
3032 /* use constant CLASS => 'MyClass' */
3033 gv = gv_fetchpvn_flags(pkgname, len, 0, SVt_PVCV);
3034 if (gv && GvCV(gv)) {
3035 SV * const sv = cv_const_sv(GvCV(gv));
3037 pkgname = SvPV_const(sv, len);
3040 return gv_stashpvn(pkgname, len, 0);
3044 * S_readpipe_override
3045 * Check whether readpipe() is overriden, and generates the appropriate
3046 * optree, provided sublex_start() is called afterwards.
3049 S_readpipe_override(pTHX)
3052 GV *gv_readpipe = gv_fetchpvs("readpipe", GV_NOTQUAL, SVt_PVCV);
3053 pl_yylval.ival = OP_BACKTICK;
3055 && GvCVu(gv_readpipe) && GvIMPORTED_CV(gv_readpipe))
3057 ((gvp = (GV**)hv_fetchs(PL_globalstash, "readpipe", FALSE))
3058 && (gv_readpipe = *gvp) && isGV_with_GP(gv_readpipe)
3059 && GvCVu(gv_readpipe) && GvIMPORTED_CV(gv_readpipe)))
3061 PL_lex_op = (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
3062 append_elem(OP_LIST,
3063 newSVOP(OP_CONST, 0, &PL_sv_undef), /* value will be read later */
3064 newCVREF(0, newGVOP(OP_GV, 0, gv_readpipe))));
3071 * The intent of this yylex wrapper is to minimize the changes to the
3072 * tokener when we aren't interested in collecting madprops. It remains
3073 * to be seen how successful this strategy will be...
3080 char *s = PL_bufptr;
3082 /* make sure PL_thiswhite is initialized */
3086 /* just do what yylex would do on pending identifier; leave PL_thiswhite alone */
3087 if (PL_pending_ident)
3088 return S_pending_ident(aTHX);
3090 /* previous token ate up our whitespace? */
3091 if (!PL_lasttoke && PL_nextwhite) {
3092 PL_thiswhite = PL_nextwhite;
3096 /* isolate the token, and figure out where it is without whitespace */
3097 PL_realtokenstart = -1;
3101 assert(PL_curforce < 0);
3103 if (!PL_thismad || PL_thismad->mad_key == '^') { /* not forced already? */
3104 if (!PL_thistoken) {
3105 if (PL_realtokenstart < 0 || !CopLINE(PL_curcop))
3106 PL_thistoken = newSVpvs("");
3108 char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
3109 PL_thistoken = newSVpvn(tstart, s - tstart);
3112 if (PL_thismad) /* install head */
3113 CURMAD('X', PL_thistoken);
3116 /* last whitespace of a sublex? */
3117 if (optype == ')' && PL_endwhite) {
3118 CURMAD('X', PL_endwhite);
3123 /* if no whitespace and we're at EOF, bail. Otherwise fake EOF below. */
3124 if (!PL_thiswhite && !PL_endwhite && !optype) {
3125 sv_free(PL_thistoken);
3130 /* put off final whitespace till peg */
3131 if (optype == ';' && !PL_rsfp) {
3132 PL_nextwhite = PL_thiswhite;
3135 else if (PL_thisopen) {
3136 CURMAD('q', PL_thisopen);
3138 sv_free(PL_thistoken);
3142 /* Store actual token text as madprop X */
3143 CURMAD('X', PL_thistoken);
3147 /* add preceding whitespace as madprop _ */
3148 CURMAD('_', PL_thiswhite);
3152 /* add quoted material as madprop = */
3153 CURMAD('=', PL_thisstuff);
3157 /* add terminating quote as madprop Q */
3158 CURMAD('Q', PL_thisclose);
3162 /* special processing based on optype */
3166 /* opval doesn't need a TOKEN since it can already store mp */
3176 if (pl_yylval.opval)
3177 append_madprops(PL_thismad, pl_yylval.opval, 0);
3185 addmad(newMADsv('p', PL_endwhite), &PL_thismad, 0);
3194 /* remember any fake bracket that lexer is about to discard */
3195 if (PL_lex_brackets == 1 &&
3196 ((expectation)PL_lex_brackstack[0] & XFAKEBRACK))
3199 while (s < PL_bufend && (*s == ' ' || *s == '\t'))
3202 PL_thiswhite = newSVpvn(PL_bufptr, ++s - PL_bufptr);
3203 addmad(newMADsv('#', PL_thiswhite), &PL_thismad, 0);
3206 break; /* don't bother looking for trailing comment */
3215 /* attach a trailing comment to its statement instead of next token */
3219 if (PL_bufptr > PL_oldbufptr && PL_bufptr[-1] == optype) {
3221 while (s < PL_bufend && (*s == ' ' || *s == '\t'))
3223 if (*s == '\n' || *s == '#') {
3224 while (s < PL_bufend && *s != '\n')
3228 PL_thiswhite = newSVpvn(PL_bufptr, s - PL_bufptr);
3229 addmad(newMADsv('#', PL_thiswhite), &PL_thismad, 0);
3246 /* Create new token struct. Note: opvals return early above. */
3247 pl_yylval.tkval = newTOKEN(optype, pl_yylval, PL_thismad);
3254 S_tokenize_use(pTHX_ int is_use, char *s) {
3257 PERL_ARGS_ASSERT_TOKENIZE_USE;
3259 if (PL_expect != XSTATE)
3260 yyerror(Perl_form(aTHX_ "\"%s\" not allowed in expression",
3261 is_use ? "use" : "no"));
3263 if (isDIGIT(*s) || (*s == 'v' && isDIGIT(s[1]))) {
3264 s = force_version(s, TRUE);
3265 if (*s == ';' || (s = SKIPSPACE1(s), *s == ';')) {
3266 start_force(PL_curforce);
3267 NEXTVAL_NEXTTOKE.opval = NULL;
3270 else if (*s == 'v') {
3271 s = force_word(s,WORD,FALSE,TRUE,FALSE);
3272 s = force_version(s, FALSE);
3276 s = force_word(s,WORD,FALSE,TRUE,FALSE);
3277 s = force_version(s, FALSE);
3279 pl_yylval.ival = is_use;
3283 static const char* const exp_name[] =
3284 { "OPERATOR", "TERM", "REF", "STATE", "BLOCK", "ATTRBLOCK",
3285 "ATTRTERM", "TERMBLOCK", "TERMORDORDOR"
3292 Works out what to call the token just pulled out of the input
3293 stream. The yacc parser takes care of taking the ops we return and
3294 stitching them into a tree.
3300 if read an identifier
3301 if we're in a my declaration
3302 croak if they tried to say my($foo::bar)
3303 build the ops for a my() declaration
3304 if it's an access to a my() variable
3305 are we in a sort block?
3306 croak if my($a); $a <=> $b
3307 build ops for access to a my() variable
3308 if in a dq string, and they've said @foo and we can't find @foo
3310 build ops for a bareword
3311 if we already built the token before, use it.
3316 #pragma segment Perl_yylex
3322 register char *s = PL_bufptr;
3327 /* orig_keyword, gvp, and gv are initialized here because
3328 * jump to the label just_a_word_zero can bypass their
3329 * initialization later. */
3330 I32 orig_keyword = 0;
3335 SV* tmp = newSVpvs("");
3336 PerlIO_printf(Perl_debug_log, "### %"IVdf":LEX_%s/X%s %s\n",
3337 (IV)CopLINE(PL_curcop),
3338 lex_state_names[PL_lex_state],
3339 exp_name[PL_expect],
3340 pv_display(tmp, s, strlen(s), 0, 60));
3343 /* check if there's an identifier for us to look at */
3344 if (PL_pending_ident)
3345 return REPORT(S_pending_ident(aTHX));
3347 /* no identifier pending identification */
3349 switch (PL_lex_state) {
3351 case LEX_NORMAL: /* Some compilers will produce faster */
3352 case LEX_INTERPNORMAL: /* code if we comment these out. */
3356 /* when we've already built the next token, just pull it out of the queue */
3360 pl_yylval = PL_nexttoke[PL_lasttoke].next_val;
3362 PL_thismad = PL_nexttoke[PL_lasttoke].next_mad;
3363 PL_nexttoke[PL_lasttoke].next_mad = 0;
3364 if (PL_thismad && PL_thismad->mad_key == '_') {
3365 PL_thiswhite = MUTABLE_SV(PL_thismad->mad_val);
3366 PL_thismad->mad_val = 0;
3367 mad_free(PL_thismad);
3372 PL_lex_state = PL_lex_defer;
3373 PL_expect = PL_lex_expect;
3374 PL_lex_defer = LEX_NORMAL;
3375 if (!PL_nexttoke[PL_lasttoke].next_type)
3380 pl_yylval = PL_nextval[PL_nexttoke];
3382 PL_lex_state = PL_lex_defer;
3383 PL_expect = PL_lex_expect;
3384 PL_lex_defer = LEX_NORMAL;
3388 /* FIXME - can these be merged? */
3389 return(PL_nexttoke[PL_lasttoke].next_type);
3391 return REPORT(PL_nexttype[PL_nexttoke]);
3394 /* interpolated case modifiers like \L \U, including \Q and \E.
3395 when we get here, PL_bufptr is at the \
3397 case LEX_INTERPCASEMOD:
3399 if (PL_bufptr != PL_bufend && *PL_bufptr != '\\')
3400 Perl_croak(aTHX_ "panic: INTERPCASEMOD");
3402 /* handle \E or end of string */
3403 if (PL_bufptr == PL_bufend || PL_bufptr[1] == 'E') {
3405 if (PL_lex_casemods) {
3406 const char oldmod = PL_lex_casestack[--PL_lex_casemods];
3407 PL_lex_casestack[PL_lex_casemods] = '\0';
3409 if (PL_bufptr != PL_bufend
3410 && (oldmod == 'L' || oldmod == 'U' || oldmod == 'Q')) {
3412 PL_lex_state = LEX_INTERPCONCAT;
3415 PL_thistoken = newSVpvs("\\E");
3421 while (PL_bufptr != PL_bufend &&
3422 PL_bufptr[0] == '\\' && PL_bufptr[1] == 'E') {
3424 PL_thiswhite = newSVpvs("");
3425 sv_catpvn(PL_thiswhite, PL_bufptr, 2);
3429 if (PL_bufptr != PL_bufend)
3432 PL_lex_state = LEX_INTERPCONCAT;
3436 DEBUG_T({ PerlIO_printf(Perl_debug_log,
3437 "### Saw case modifier\n"); });
3439 if (s[1] == '\\' && s[2] == 'E') {
3442 PL_thiswhite = newSVpvs("");
3443 sv_catpvn(PL_thiswhite, PL_bufptr, 4);
3446 PL_lex_state = LEX_INTERPCONCAT;
3451 if (!PL_madskills) /* when just compiling don't need correct */
3452 if (strnEQ(s, "L\\u", 3) || strnEQ(s, "U\\l", 3))
3453 tmp = *s, *s = s[2], s[2] = (char)tmp; /* misordered... */
3454 if ((*s == 'L' || *s == 'U') &&
3455 (strchr(PL_lex_casestack, 'L') || strchr(PL_lex_casestack, 'U'))) {
3456 PL_lex_casestack[--PL_lex_casemods] = '\0';
3459 if (PL_lex_casemods > 10)
3460 Renew(PL_lex_casestack, PL_lex_casemods + 2, char);
3461 PL_lex_casestack[PL_lex_casemods++] = *s;
3462 PL_lex_casestack[PL_lex_casemods] = '\0';
3463 PL_lex_state = LEX_INTERPCONCAT;
3464 start_force(PL_curforce);
3465 NEXTVAL_NEXTTOKE.ival = 0;
3467 start_force(PL_curforce);
3469 NEXTVAL_NEXTTOKE.ival = OP_LCFIRST;
3471 NEXTVAL_NEXTTOKE.ival = OP_UCFIRST;
3473 NEXTVAL_NEXTTOKE.ival = OP_LC;
3475 NEXTVAL_NEXTTOKE.ival = OP_UC;
3477 NEXTVAL_NEXTTOKE.ival = OP_QUOTEMETA;
3479 Perl_croak(aTHX_ "panic: yylex");
3481 SV* const tmpsv = newSVpvs("\\ ");
3482 /* replace the space with the character we want to escape
3484 SvPVX(tmpsv)[1] = *s;
3490 if (PL_lex_starts) {
3496 sv_free(PL_thistoken);
3497 PL_thistoken = newSVpvs("");
3500 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
3501 if (PL_lex_casemods == 1 && PL_lex_inpat)
3510 case LEX_INTERPPUSH:
3511 return REPORT(sublex_push());
3513 case LEX_INTERPSTART:
3514 if (PL_bufptr == PL_bufend)
3515 return REPORT(sublex_done());
3516 DEBUG_T({ PerlIO_printf(Perl_debug_log,
3517 "### Interpolated variable\n"); });
3519 PL_lex_dojoin = (*PL_bufptr == '@');
3520 PL_lex_state = LEX_INTERPNORMAL;
3521 if (PL_lex_dojoin) {
3522 start_force(PL_curforce);
3523 NEXTVAL_NEXTTOKE.ival = 0;
3525 start_force(PL_curforce);
3526 force_ident("\"", '$');
3527 start_force(PL_curforce);
3528 NEXTVAL_NEXTTOKE.ival = 0;
3530 start_force(PL_curforce);
3531 NEXTVAL_NEXTTOKE.ival = 0;
3533 start_force(PL_curforce);
3534 NEXTVAL_NEXTTOKE.ival = OP_JOIN; /* emulate join($", ...) */
3537 if (PL_lex_starts++) {
3542 sv_free(PL_thistoken);
3543 PL_thistoken = newSVpvs("");
3546 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
3547 if (!PL_lex_casemods && PL_lex_inpat)
3554 case LEX_INTERPENDMAYBE:
3555 if (intuit_more(PL_bufptr)) {
3556 PL_lex_state = LEX_INTERPNORMAL; /* false alarm, more expr */
3562 if (PL_lex_dojoin) {
3563 PL_lex_dojoin = FALSE;
3564 PL_lex_state = LEX_INTERPCONCAT;
3568 sv_free(PL_thistoken);
3569 PL_thistoken = newSVpvs("");
3574 if (PL_lex_inwhat == OP_SUBST && PL_linestr == PL_lex_repl
3575 && SvEVALED(PL_lex_repl))
3577 if (PL_bufptr != PL_bufend)
3578 Perl_croak(aTHX_ "Bad evalled substitution pattern");
3582 case LEX_INTERPCONCAT:
3584 if (PL_lex_brackets)
3585 Perl_croak(aTHX_ "panic: INTERPCONCAT");
3587 if (PL_bufptr == PL_bufend)
3588 return REPORT(sublex_done());
3590 if (SvIVX(PL_linestr) == '\'') {
3591 SV *sv = newSVsv(PL_linestr);
3594 else if ( PL_hints & HINT_NEW_RE )
3595 sv = new_constant(NULL, 0, "qr", sv, sv, "q", 1);
3596 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
3600 s = scan_const(PL_bufptr);
3602 PL_lex_state = LEX_INTERPCASEMOD;
3604 PL_lex_state = LEX_INTERPSTART;
3607 if (s != PL_bufptr) {
3608 start_force(PL_curforce);
3610 curmad('X', newSVpvn(PL_bufptr,s-PL_bufptr));
3612 NEXTVAL_NEXTTOKE = pl_yylval;
3615 if (PL_lex_starts++) {
3619 sv_free(PL_thistoken);
3620 PL_thistoken = newSVpvs("");
3623 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
3624 if (!PL_lex_casemods && PL_lex_inpat)
3637 PL_lex_state = LEX_NORMAL;
3638 s = scan_formline(PL_bufptr);
3639 if (!PL_lex_formbrack)
3645 PL_oldoldbufptr = PL_oldbufptr;
3651 sv_free(PL_thistoken);
3654 PL_realtokenstart = s - SvPVX(PL_linestr); /* assume but undo on ws */
3658 if (isIDFIRST_lazy_if(s,UTF))
3660 len = UTF ? Perl_utf8_length(aTHX_ (U8 *) PL_linestart, (U8 *) s) : (STRLEN) (s - PL_linestart);
3661 Perl_croak(aTHX_ "Unrecognized character \\x%02X in column %d", *s & 255, (int) len + 1);
3664 goto fake_eof; /* emulate EOF on ^D or ^Z */
3673 if (PL_lex_brackets) {
3674 yyerror((const char *)
3676 ? "Format not terminated"
3677 : "Missing right curly or square bracket"));
3679 DEBUG_T( { PerlIO_printf(Perl_debug_log,
3680 "### Tokener got EOF\n");
3684 if (s++ < PL_bufend)
3685 goto retry; /* ignore stray nulls */
3688 if (!PL_in_eval && !PL_preambled) {
3689 PL_preambled = TRUE;
3695 /* Generate a string of Perl code to load the debugger.
3696 * If PERL5DB is set, it will return the contents of that,
3697 * otherwise a compile-time require of perl5db.pl. */
3699 const char * const pdb = PerlEnv_getenv("PERL5DB");
3702 sv_setpv(PL_linestr, pdb);
3703 sv_catpvs(PL_linestr,";");
3705 SETERRNO(0,SS_NORMAL);
3706 sv_setpvs(PL_linestr, "BEGIN { require 'perl5db.pl' };");
3709 sv_setpvs(PL_linestr,"");
3710 if (PL_preambleav) {
3711 SV **svp = AvARRAY(PL_preambleav);
3712 SV **const end = svp + AvFILLp(PL_preambleav);
3714 sv_catsv(PL_linestr, *svp);
3716 sv_catpvs(PL_linestr, ";");
3718 sv_free(MUTABLE_SV(PL_preambleav));
3719 PL_preambleav = NULL;
3722 sv_catpvs(PL_linestr,
3723 "use feature ':5." STRINGIFY(PERL_VERSION) "';");
3724 if (PL_minus_n || PL_minus_p) {
3725 sv_catpvs(PL_linestr, "LINE: while (<>) {");
3727 sv_catpvs(PL_linestr,"chomp;");
3730 if ((*PL_splitstr == '/' || *PL_splitstr == '\''
3731 || *PL_splitstr == '"')
3732 && strchr(PL_splitstr + 1, *PL_splitstr))
3733 Perl_sv_catpvf(aTHX_ PL_linestr, "our @F=split(%s);", PL_splitstr);
3735 /* "q\0${splitstr}\0" is legal perl. Yes, even NUL
3736 bytes can be used as quoting characters. :-) */
3737 const char *splits = PL_splitstr;
3738 sv_catpvs(PL_linestr, "our @F=split(q\0");
3741 if (*splits == '\\')
3742 sv_catpvn(PL_linestr, splits, 1);
3743 sv_catpvn(PL_linestr, splits, 1);
3744 } while (*splits++);
3745 /* This loop will embed the trailing NUL of
3746 PL_linestr as the last thing it does before
3748 sv_catpvs(PL_linestr, ");");
3752 sv_catpvs(PL_linestr,"our @F=split(' ');");
3755 sv_catpvs(PL_linestr, "\n");
3756 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3757 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3758 PL_last_lop = PL_last_uni = NULL;
3759 if ((PERLDB_LINE || PERLDB_SAVESRC) && PL_curstash != PL_debstash)
3760 update_debugger_info(PL_linestr, NULL, 0);
3764 bof = PL_rsfp ? TRUE : FALSE;
3765 if ((s = filter_gets(PL_linestr, PL_rsfp, 0)) == NULL) {
3768 PL_realtokenstart = -1;
3771 if ((PerlIO *)PL_rsfp == PerlIO_stdin())
3772 PerlIO_clearerr(PL_rsfp);
3774 (void)PerlIO_close(PL_rsfp);
3776 PL_doextract = FALSE;
3778 if (!PL_in_eval && (PL_minus_n || PL_minus_p)) {
3784 sv_setpvs(PL_linestr, ";}continue{print;}");
3786 sv_setpvs(PL_linestr, ";}");
3787 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3788 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3789 PL_last_lop = PL_last_uni = NULL;
3790 PL_minus_n = PL_minus_p = 0;
3793 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3794 PL_last_lop = PL_last_uni = NULL;
3795 sv_setpvs(PL_linestr,"");
3796 TOKEN(';'); /* not infinite loop because rsfp is NULL now */
3798 /* If it looks like the start of a BOM or raw UTF-16,
3799 * check if it in fact is. */
3805 #ifdef PERLIO_IS_STDIO
3806 # ifdef __GNU_LIBRARY__
3807 # if __GNU_LIBRARY__ == 1 /* Linux glibc5 */
3808 # define FTELL_FOR_PIPE_IS_BROKEN
3812 # if __GLIBC__ == 1 /* maybe some glibc5 release had it like this? */
3813 # define FTELL_FOR_PIPE_IS_BROKEN
3818 bof = PerlIO_tell(PL_rsfp) == (Off_t)SvCUR(PL_linestr);
3820 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3821 s = swallow_bom((U8*)s);
3825 /* Incest with pod. */
3828 sv_catsv(PL_thiswhite, PL_linestr);
3830 if (*s == '=' && strnEQ(s, "=cut", 4) && !isALPHA(s[4])) {
3831 sv_setpvs(PL_linestr, "");
3832 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3833 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3834 PL_last_lop = PL_last_uni = NULL;
3835 PL_doextract = FALSE;
3839 } while (PL_doextract);
3840 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = s;
3841 if ((PERLDB_LINE || PERLDB_SAVESRC) && PL_curstash != PL_debstash)
3842 update_debugger_info(PL_linestr, NULL, 0);
3843 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3844 PL_last_lop = PL_last_uni = NULL;
3845 if (CopLINE(PL_curcop) == 1) {
3846 while (s < PL_bufend && isSPACE(*s))
3848 if (*s == ':' && s[1] != ':') /* for csh execing sh scripts */
3852 PL_thiswhite = newSVpvn(PL_linestart, s - PL_linestart);
3856 if (*s == '#' && *(s+1) == '!')
3858 #ifdef ALTERNATE_SHEBANG
3860 static char const as[] = ALTERNATE_SHEBANG;
3861 if (*s == as[0] && strnEQ(s, as, sizeof(as) - 1))
3862 d = s + (sizeof(as) - 1);
3864 #endif /* ALTERNATE_SHEBANG */
3873 while (*d && !isSPACE(*d))
3877 #ifdef ARG_ZERO_IS_SCRIPT
3878 if (ipathend > ipath) {
3880 * HP-UX (at least) sets argv[0] to the script name,
3881 * which makes $^X incorrect. And Digital UNIX and Linux,
3882 * at least, set argv[0] to the basename of the Perl
3883 * interpreter. So, having found "#!", we'll set it right.
3885 SV * const x = GvSV(gv_fetchpvs("\030", GV_ADD|GV_NOTQUAL,
3887 assert(SvPOK(x) || SvGMAGICAL(x));
3888 if (sv_eq(x, CopFILESV(PL_curcop))) {
3889 sv_setpvn(x, ipath, ipathend - ipath);
3895 const char *bstart = SvPV_const(CopFILESV(PL_curcop),blen);
3896 const char * const lstart = SvPV_const(x,llen);
3898 bstart += blen - llen;
3899 if (strnEQ(bstart, lstart, llen) && bstart[-1] == '/') {
3900 sv_setpvn(x, ipath, ipathend - ipath);
3905 TAINT_NOT; /* $^X is always tainted, but that's OK */
3907 #endif /* ARG_ZERO_IS_SCRIPT */
3912 d = instr(s,"perl -");
3914 d = instr(s,"perl");
3916 /* avoid getting into infinite loops when shebang
3917 * line contains "Perl" rather than "perl" */
3919 for (d = ipathend-4; d >= ipath; --d) {
3920 if ((*d == 'p' || *d == 'P')
3921 && !ibcmp(d, "perl", 4))
3931 #ifdef ALTERNATE_SHEBANG
3933 * If the ALTERNATE_SHEBANG on this system starts with a
3934 * character that can be part of a Perl expression, then if
3935 * we see it but not "perl", we're probably looking at the
3936 * start of Perl code, not a request to hand off to some
3937 * other interpreter. Similarly, if "perl" is there, but
3938 * not in the first 'word' of the line, we assume the line
3939 * contains the start of the Perl program.
3941 if (d && *s != '#') {
3942 const char *c = ipath;
3943 while (*c && !strchr("; \t\r\n\f\v#", *c))
3946 d = NULL; /* "perl" not in first word; ignore */
3948 *s = '#'; /* Don't try to parse shebang line */
3950 #endif /* ALTERNATE_SHEBANG */
3951 #ifndef MACOS_TRADITIONAL
3956 !instr(s,"indir") &&
3957 instr(PL_origargv[0],"perl"))
3964 while (s < PL_bufend && isSPACE(*s))
3966 if (s < PL_bufend) {
3967 Newxz(newargv,PL_origargc+3,char*);
3969 while (s < PL_bufend && !isSPACE(*s))
3972 Copy(PL_origargv+1, newargv+2, PL_origargc+1, char*);
3975 newargv = PL_origargv;
3978 PerlProc_execv(ipath, EXEC_ARGV_CAST(newargv));
3980 Perl_croak(aTHX_ "Can't exec %s", ipath);
3984 while (*d && !isSPACE(*d))
3986 while (SPACE_OR_TAB(*d))
3990 const bool switches_done = PL_doswitches;
3991 const U32 oldpdb = PL_perldb;
3992 const bool oldn = PL_minus_n;
3993 const bool oldp = PL_minus_p;
3997 if (*d1 == 'M' || *d1 == 'm' || *d1 == 'C') {
3998 const char * const m = d1;
3999 while (*d1 && !isSPACE(*d1))
4001 Perl_croak(aTHX_ "Too late for \"-%.*s\" option",
4004 d1 = moreswitches(d1);
4006 if (PL_doswitches && !switches_done) {
4007 int argc = PL_origargc;
4008 char **argv = PL_origargv;
4011 } while (argc && argv[0][0] == '-' && argv[0][1]);
4012 init_argv_symbols(argc,argv);
4014 if (((PERLDB_LINE || PERLDB_SAVESRC) && !oldpdb) ||
4015 ((PL_minus_n || PL_minus_p) && !(oldn || oldp)))
4016 /* if we have already added "LINE: while (<>) {",
4017 we must not do it again */
4019 sv_setpvs(PL_linestr, "");
4020 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
4021 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
4022 PL_last_lop = PL_last_uni = NULL;
4023 PL_preambled = FALSE;
4024 if (PERLDB_LINE || PERLDB_SAVESRC)
4025 (void)gv_fetchfile(PL_origfilename);
4032 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
4034 PL_lex_state = LEX_FORMLINE;
4039 #ifdef PERL_STRICT_CR
4040 Perl_warn(aTHX_ "Illegal character \\%03o (carriage return)", '\r');
4042 "\t(Maybe you didn't strip carriage returns after a network transfer?)\n");
4044 case ' ': case '\t': case '\f': case 013:
4045 #ifdef MACOS_TRADITIONAL
4049 PL_realtokenstart = -1;
4051 PL_thiswhite = newSVpvs("");
4052 sv_catpvn(PL_thiswhite, s, 1);
4059 PL_realtokenstart = -1;
4063 if (PL_lex_state != LEX_NORMAL || (PL_in_eval && !PL_rsfp)) {
4064 if (*s == '#' && s == PL_linestart && PL_in_eval && !PL_rsfp) {
4065 /* handle eval qq[#line 1 "foo"\n ...] */
4066 CopLINE_dec(PL_curcop);
4069 if (PL_madskills && !PL_lex_formbrack && !PL_in_eval) {
4071 if (!PL_in_eval || PL_rsfp)
4076 while (d < PL_bufend && *d != '\n')
4080 else if (d > PL_bufend) /* Found by Ilya: feed random input to Perl. */
4081 Perl_croak(aTHX_ "panic: input overflow");
4084 PL_thiswhite = newSVpvn(s, d - s);
4089 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
4091 PL_lex_state = LEX_FORMLINE;
4097 if (PL_madskills && CopLINE(PL_curcop) >= 1 && !PL_lex_formbrack) {
4098 if (CopLINE(PL_curcop) == 1 && s[0] == '#' && s[1] == '!') {
4101 TOKEN(PEG); /* make sure any #! line is accessible */
4106 /* if (PL_madskills && PL_lex_formbrack) { */
4108 while (d < PL_bufend && *d != '\n')
4112 else if (d > PL_bufend) /* Found by Ilya: feed random input to Perl. */
4113 Perl_croak(aTHX_ "panic: input overflow");
4114 if (PL_madskills && CopLINE(PL_curcop) >= 1) {
4116 PL_thiswhite = newSVpvs("");
4117 if (CopLINE(PL_curcop) == 1) {
4118 sv_setpvs(PL_thiswhite, "");
4121 sv_catpvn(PL_thiswhite, s, d - s);
4135 if (s[1] && isALPHA(s[1]) && !isALNUM(s[2])) {
4143 while (s < PL_bufend && SPACE_OR_TAB(*s))
4146 if (strnEQ(s,"=>",2)) {
4147 s = force_word(PL_bufptr,WORD,FALSE,FALSE,FALSE);
4148 DEBUG_T( { printbuf("### Saw unary minus before =>, forcing word %s\n", s); } );
4149 OPERATOR('-'); /* unary minus */
4151 PL_last_uni = PL_oldbufptr;
4153 case 'r': ftst = OP_FTEREAD; break;
4154 case 'w': ftst = OP_FTEWRITE; break;
4155 case 'x': ftst = OP_FTEEXEC; break;
4156 case 'o': ftst = OP_FTEOWNED; break;
4157 case 'R': ftst = OP_FTRREAD; break;
4158 case 'W': ftst = OP_FTRWRITE; break;
4159 case 'X': ftst = OP_FTREXEC; break;
4160 case 'O': ftst = OP_FTROWNED; break;
4161 case 'e': ftst = OP_FTIS; break;
4162 case 'z': ftst = OP_FTZERO; break;
4163 case 's': ftst = OP_FTSIZE; break;
4164 case 'f': ftst = OP_FTFILE; break;
4165 case 'd': ftst = OP_FTDIR; break;
4166 case 'l': ftst = OP_FTLINK; break;
4167 case 'p': ftst = OP_FTPIPE; break;
4168 case 'S': ftst = OP_FTSOCK; break;
4169 case 'u': ftst = OP_FTSUID; break;
4170 case 'g': ftst = OP_FTSGID; break;
4171 case 'k': ftst = OP_FTSVTX; break;
4172 case 'b': ftst = OP_FTBLK; break;
4173 case 'c': ftst = OP_FTCHR; break;
4174 case 't': ftst = OP_FTTTY; break;
4175 case 'T': ftst = OP_FTTEXT; break;
4176 case 'B': ftst = OP_FTBINARY; break;
4177 case 'M': case 'A': case 'C':
4178 gv_fetchpvs("\024", GV_ADD|GV_NOTQUAL, SVt_PV);
4180 case 'M': ftst = OP_FTMTIME; break;
4181 case 'A': ftst = OP_FTATIME; break;
4182 case 'C': ftst = OP_FTCTIME; break;
4190 PL_last_lop_op = (OPCODE)ftst;
4191 DEBUG_T( { PerlIO_printf(Perl_debug_log,
4192 "### Saw file test %c\n", (int)tmp);
4197 /* Assume it was a minus followed by a one-letter named
4198 * subroutine call (or a -bareword), then. */
4199 DEBUG_T( { PerlIO_printf(Perl_debug_log,
4200 "### '-%c' looked like a file test but was not\n",
4207 const char tmp = *s++;
4210 if (PL_expect == XOPERATOR)
4215 else if (*s == '>') {
4218 if (isIDFIRST_lazy_if(s,UTF)) {
4219 s = force_word(s,METHOD,FALSE,TRUE,FALSE);
4227 if (PL_expect == XOPERATOR)
4230 if (isSPACE(*s) || !isSPACE(*PL_bufptr))
4232 OPERATOR('-'); /* unary minus */
4238 const char tmp = *s++;
4241 if (PL_expect == XOPERATOR)
4246 if (PL_expect == XOPERATOR)
4249 if (isSPACE(*s) || !isSPACE(*PL_bufptr))
4256 if (PL_expect != XOPERATOR) {
4257 s = scan_ident(s, PL_bufend, PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
4258 PL_expect = XOPERATOR;
4259 force_ident(PL_tokenbuf, '*');
4272 if (PL_expect == XOPERATOR) {
4276 PL_tokenbuf[0] = '%';
4277 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1,
4278 sizeof PL_tokenbuf - 1, FALSE);
4279 if (!PL_tokenbuf[1]) {
4282 PL_pending_ident = '%';
4293 && (PL_expect == XOPERATOR || PL_expect == XTERMORDORDOR))
4300 const char tmp = *s++;
4306 goto just_a_word_zero_gv;
4309 switch (PL_expect) {
4315 if (!PL_in_my || PL_lex_state != LEX_NORMAL)
4317 PL_bufptr = s; /* update in case we back off */
4323 PL_expect = XTERMBLOCK;
4326 stuffstart = s - SvPVX(PL_linestr) - 1;
4330 while (isIDFIRST_lazy_if(s,UTF)) {
4333 d = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
4334 if (isLOWER(*s) && (tmp = keyword(PL_tokenbuf, len, 0))) {
4335 if (tmp < 0) tmp = -tmp;
4349 sv = newSVpvn(s, len);
4351 d = scan_str(d,TRUE,TRUE);
4353 /* MUST advance bufptr here to avoid bogus
4354 "at end of line" context messages from yyerror().
4356 PL_bufptr = s + len;
4357 yyerror("Unterminated attribute parameter in attribute list");
4361 return REPORT(0); /* EOF indicator */
4365 sv_catsv(sv, PL_lex_stuff);
4366 attrs = append_elem(OP_LIST, attrs,
4367 newSVOP(OP_CONST, 0, sv));
4368 SvREFCNT_dec(PL_lex_stuff);
4369 PL_lex_stuff = NULL;
4372 if (len == 6 && strnEQ(SvPVX(sv), "unique", len)) {
4374 if (PL_in_my == KEY_our) {
4376 GvUNIQUE_on(cGVOPx_gv(pl_yylval.opval));
4378 /* skip to avoid loading attributes.pm */
4380 deprecate(":unique");
4383 Perl_croak(aTHX_ "The 'unique' attribute may only be applied to 'our' variables");
4386 /* NOTE: any CV attrs applied here need to be part of
4387 the CVf_BUILTIN_ATTRS define in cv.h! */
4388 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "lvalue", len)) {
4390 CvLVALUE_on(PL_compcv);
4392 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "locked", len)) {
4394 CvLOCKED_on(PL_compcv);
4396 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "method", len)) {
4398 CvMETHOD_on(PL_compcv);
4400 /* After we've set the flags, it could be argued that
4401 we don't need to do the attributes.pm-based setting
4402 process, and shouldn't bother appending recognized
4403 flags. To experiment with that, uncomment the
4404 following "else". (Note that's already been
4405 uncommented. That keeps the above-applied built-in
4406 attributes from being intercepted (and possibly
4407 rejected) by a package's attribute routines, but is
4408 justified by the performance win for the common case
4409 of applying only built-in attributes.) */
4411 attrs = append_elem(OP_LIST, attrs,
4412 newSVOP(OP_CONST, 0,
4416 if (*s == ':' && s[1] != ':')
4419 break; /* require real whitespace or :'s */
4420 /* XXX losing whitespace on sequential attributes here */
4424 = (PL_expect == XOPERATOR ? '=' : '{'); /*'}(' for vi */
4425 if (*s != ';' && *s != '}' && *s != tmp
4426 && (tmp != '=' || *s != ')')) {
4427 const char q = ((*s == '\'') ? '"' : '\'');
4428 /* If here for an expression, and parsed no attrs, back
4430 if (tmp == '=' && !attrs) {
4434 /* MUST advance bufptr here to avoid bogus "at end of line"
4435 context messages from yyerror().
4438 yyerror( (const char *)
4440 ? Perl_form(aTHX_ "Invalid separator character "
4441 "%c%c%c in attribute list", q, *s, q)
4442 : "Unterminated attribute list" ) );
4450 start_force(PL_curforce);
4451 NEXTVAL_NEXTTOKE.opval = attrs;
4452 CURMAD('_', PL_nextwhite);
4457 PL_thistoken = newSVpvn(SvPVX(PL_linestr) + stuffstart,
4458 (s - SvPVX(PL_linestr)) - stuffstart);
4466 if (PL_last_lop == PL_oldoldbufptr || PL_last_uni == PL_oldoldbufptr)
4467 PL_oldbufptr = PL_oldoldbufptr; /* allow print(STDOUT 123) */
4475 const char tmp = *s++;
4480 const char tmp = *s++;
4488 if (PL_lex_brackets <= 0)
4489 yyerror("Unmatched right square bracket");
4492 if (PL_lex_state == LEX_INTERPNORMAL) {
4493 if (PL_lex_brackets == 0) {
4494 if (*s == '-' && s[1] == '>')
4495 PL_lex_state = LEX_INTERPENDMAYBE;
4496 else if (*s != '[' && *s != '{')
4497 PL_lex_state = LEX_INTERPEND;
4504 if (PL_lex_brackets > 100) {
4505 Renew(PL_lex_brackstack, PL_lex_brackets + 10, char);
4507 switch (PL_expect) {
4509 if (PL_lex_formbrack) {
4513 if (PL_oldoldbufptr == PL_last_lop)
4514 PL_lex_brackstack[PL_lex_brackets++] = XTERM;
4516 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
4517 OPERATOR(HASHBRACK);
4519 while (s < PL_bufend && SPACE_OR_TAB(*s))
4522 PL_tokenbuf[0] = '\0';
4523 if (d < PL_bufend && *d == '-') {
4524 PL_tokenbuf[0] = '-';
4526 while (d < PL_bufend && SPACE_OR_TAB(*d))
4529 if (d < PL_bufend && isIDFIRST_lazy_if(d,UTF)) {
4530 d = scan_word(d, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1,
4532 while (d < PL_bufend && SPACE_OR_TAB(*d))
4535 const char minus = (PL_tokenbuf[0] == '-');
4536 s = force_word(s + minus, WORD, FALSE, TRUE, FALSE);
4544 PL_lex_brackstack[PL_lex_brackets++] = XSTATE;
4549 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
4554 if (PL_oldoldbufptr == PL_last_lop)
4555 PL_lex_brackstack[PL_lex_brackets++] = XTERM;
4557 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
4560 if (PL_expect == XREF && PL_lex_state == LEX_INTERPNORMAL) {
4562 /* This hack is to get the ${} in the message. */
4564 yyerror("syntax error");
4567 OPERATOR(HASHBRACK);
4569 /* This hack serves to disambiguate a pair of curlies
4570 * as being a block or an anon hash. Normally, expectation
4571 * determines that, but in cases where we're not in a
4572 * position to expect anything in particular (like inside
4573 * eval"") we have to resolve the ambiguity. This code
4574 * covers the case where the first term in the curlies is a
4575 * quoted string. Most other cases need to be explicitly
4576 * disambiguated by prepending a "+" before the opening
4577 * curly in order to force resolution as an anon hash.
4579 * XXX should probably propagate the outer expectation
4580 * into eval"" to rely less on this hack, but that could
4581 * potentially break current behavior of eval"".
4585 if (*s == '\'' || *s == '"' || *s == '`') {
4586 /* common case: get past first string, handling escapes */
4587 for (t++; t < PL_bufend && *t != *s;)
4588 if (*t++ == '\\' && (*t == '\\' || *t == *s))
4592 else if (*s == 'q') {
4595 || ((*t == 'q' || *t == 'x') && ++t < PL_bufend
4598 /* skip q//-like construct */
4600 char open, close, term;
4603 while (t < PL_bufend && isSPACE(*t))
4605 /* check for q => */
4606 if (t+1 < PL_bufend && t[0] == '=' && t[1] == '>') {
4607 OPERATOR(HASHBRACK);
4611 if (term && (tmps = strchr("([{< )]}> )]}>",term)))
4615 for (t++; t < PL_bufend; t++) {
4616 if (*t == '\\' && t+1 < PL_bufend && open != '\\')
4618 else if (*t == open)
4622 for (t++; t < PL_bufend; t++) {
4623 if (*t == '\\' && t+1 < PL_bufend)
4625 else if (*t == close && --brackets <= 0)
4627 else if (*t == open)
4634 /* skip plain q word */
4635 while (t < PL_bufend && isALNUM_lazy_if(t,UTF))
4638 else if (isALNUM_lazy_if(t,UTF)) {
4640 while (t < PL_bufend && isALNUM_lazy_if(t,UTF))
4643 while (t < PL_bufend && isSPACE(*t))
4645 /* if comma follows first term, call it an anon hash */
4646 /* XXX it could be a comma expression with loop modifiers */
4647 if (t < PL_bufend && ((*t == ',' && (*s == 'q' || !isLOWER(*s)))
4648 || (*t == '=' && t[1] == '>')))
4649 OPERATOR(HASHBRACK);
4650 if (PL_expect == XREF)
4653 PL_lex_brackstack[PL_lex_brackets-1] = XSTATE;
4659 pl_yylval.ival = CopLINE(PL_curcop);
4660 if (isSPACE(*s) || *s == '#')
4661 PL_copline = NOLINE; /* invalidate current command line number */
4666 if (PL_lex_brackets <= 0)
4667 yyerror("Unmatched right curly bracket");
4669 PL_expect = (expectation)PL_lex_brackstack[--PL_lex_brackets];
4670 if (PL_lex_brackets < PL_lex_formbrack && PL_lex_state != LEX_INTERPNORMAL)
4671 PL_lex_formbrack = 0;
4672 if (PL_lex_state == LEX_INTERPNORMAL) {
4673 if (PL_lex_brackets == 0) {
4674 if (PL_expect & XFAKEBRACK) {
4675 PL_expect &= XENUMMASK;
4676 PL_lex_state = LEX_INTERPEND;
4681 PL_thiswhite = newSVpvs("");
4682 sv_catpvs(PL_thiswhite,"}");
4685 return yylex(); /* ignore fake brackets */
4687 if (*s == '-' && s[1] == '>')
4688 PL_lex_state = LEX_INTERPENDMAYBE;
4689 else if (*s != '[' && *s != '{')
4690 PL_lex_state = LEX_INTERPEND;
4693 if (PL_expect & XFAKEBRACK) {
4694 PL_expect &= XENUMMASK;
4696 return yylex(); /* ignore fake brackets */
4698 start_force(PL_curforce);
4700 curmad('X', newSVpvn(s-1,1));
4701 CURMAD('_', PL_thiswhite);
4706 PL_thistoken = newSVpvs("");
4714 if (PL_expect == XOPERATOR) {
4715 if (PL_bufptr == PL_linestart && ckWARN(WARN_SEMICOLON)
4716 && isIDFIRST_lazy_if(s,UTF))
4718 CopLINE_dec(PL_curcop);
4719 Perl_warner(aTHX_ packWARN(WARN_SEMICOLON), "%s", PL_warn_nosemi);
4720 CopLINE_inc(PL_curcop);
4725 s = scan_ident(s - 1, PL_bufend, PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
4727 PL_expect = XOPERATOR;
4728 force_ident(PL_tokenbuf, '&');
4732 pl_yylval.ival = (OPpENTERSUB_AMPER<<8);
4744 const char tmp = *s++;
4751 if (tmp && isSPACE(*s) && ckWARN(WARN_SYNTAX)
4752 && strchr("+-*/%.^&|<",tmp))
4753 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
4754 "Reversed %c= operator",(int)tmp);
4756 if (PL_expect == XSTATE && isALPHA(tmp) &&
4757 (s == PL_linestart+1 || s[-2] == '\n') )
4759 if (PL_in_eval && !PL_rsfp) {
4764 if (strnEQ(s,"=cut",4)) {
4780 PL_thiswhite = newSVpvs("");
4781 sv_catpvn(PL_thiswhite, PL_linestart,
4782 PL_bufend - PL_linestart);
4786 PL_doextract = TRUE;
4790 if (PL_lex_brackets < PL_lex_formbrack) {
4792 #ifdef PERL_STRICT_CR
4793 while (SPACE_OR_TAB(*t))
4795 while (SPACE_OR_TAB(*t) || *t == '\r')
4798 if (*t == '\n' || *t == '#') {
4807 if (PL_expect == XSTATE && s[1] == '!' && s[2] == '!') {
4813 const char tmp = *s++;
4815 /* was this !=~ where !~ was meant?
4816 * warn on m:!=~\s+([/?]|[msy]\W|tr\W): */
4818 if (*s == '~' && ckWARN(WARN_SYNTAX)) {
4819 const char *t = s+1;
4821 while (t < PL_bufend && isSPACE(*t))
4824 if (*t == '/' || *t == '?' ||
4825 ((*t == 'm' || *t == 's' || *t == 'y')
4826 && !isALNUM(t[1])) ||
4827 (*t == 't' && t[1] == 'r' && !isALNUM(t[2])))
4828 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
4829 "!=~ should be !~");
4839 if (PL_expect != XOPERATOR) {
4840 if (s[1] != '<' && !strchr(s,'>'))
4843 s = scan_heredoc(s);
4845 s = scan_inputsymbol(s);
4846 TERM(sublex_start());
4852 SHop(OP_LEFT_SHIFT);
4866 const char tmp = *s++;
4868 SHop(OP_RIGHT_SHIFT);
4869 else if (tmp == '=')
4878 if (PL_expect == XOPERATOR) {
4879 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
4881 deprecate_old(commaless_variable_list);
4882 return REPORT(','); /* grandfather non-comma-format format */
4886 if (s[1] == '#' && (isIDFIRST_lazy_if(s+2,UTF) || strchr("{$:+-", s[2]))) {
4887 PL_tokenbuf[0] = '@';
4888 s = scan_ident(s + 1, PL_bufend, PL_tokenbuf + 1,
4889 sizeof PL_tokenbuf - 1, FALSE);
4890 if (PL_expect == XOPERATOR)
4891 no_op("Array length", s);
4892 if (!PL_tokenbuf[1])
4894 PL_expect = XOPERATOR;
4895 PL_pending_ident = '#';
4899 PL_tokenbuf[0] = '$';
4900 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1,
4901 sizeof PL_tokenbuf - 1, FALSE);
4902 if (PL_expect == XOPERATOR)
4904 if (!PL_tokenbuf[1]) {
4906 yyerror("Final $ should be \\$ or $name");
4910 /* This kludge not intended to be bulletproof. */
4911 if (PL_tokenbuf[1] == '[' && !PL_tokenbuf[2]) {
4912 pl_yylval.opval = newSVOP(OP_CONST, 0,
4913 newSViv(CopARYBASE_get(&PL_compiling)));
4914 pl_yylval.opval->op_private = OPpCONST_ARYBASE;
4920 const char tmp = *s;
4921 if (PL_lex_state == LEX_NORMAL)
4924 if ((PL_expect != XREF || PL_oldoldbufptr == PL_last_lop)
4925 && intuit_more(s)) {
4927 PL_tokenbuf[0] = '@';
4928 if (ckWARN(WARN_SYNTAX)) {
4931 while (isSPACE(*t) || isALNUM_lazy_if(t,UTF) || *t == '$')
4934 PL_bufptr = PEEKSPACE(PL_bufptr); /* XXX can realloc */
4935 while (t < PL_bufend && *t != ']')
4937 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
4938 "Multidimensional syntax %.*s not supported",
4939 (int)((t - PL_bufptr) + 1), PL_bufptr);
4943 else if (*s == '{') {
4945 PL_tokenbuf[0] = '%';
4946 if (strEQ(PL_tokenbuf+1, "SIG") && ckWARN(WARN_SYNTAX)
4947 && (t = strchr(s, '}')) && (t = strchr(t, '=')))
4949 char tmpbuf[sizeof PL_tokenbuf];
4952 } while (isSPACE(*t));
4953 if (isIDFIRST_lazy_if(t,UTF)) {
4955 t = scan_word(t, tmpbuf, sizeof tmpbuf, TRUE,
4959 if (*t == ';' && get_cvn_flags(tmpbuf, len, 0))
4960 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
4961 "You need to quote \"%s\"",
4968 PL_expect = XOPERATOR;
4969 if (PL_lex_state == LEX_NORMAL && isSPACE((char)tmp)) {
4970 const bool islop = (PL_last_lop == PL_oldoldbufptr);
4971 if (!islop || PL_last_lop_op == OP_GREPSTART)
4972 PL_expect = XOPERATOR;
4973 else if (strchr("$@\"'`q", *s))
4974 PL_expect = XTERM; /* e.g. print $fh "foo" */
4975 else if (strchr("&*<%", *s) && isIDFIRST_lazy_if(s+1,UTF))
4976 PL_expect = XTERM; /* e.g. print $fh &sub */
4977 else if (isIDFIRST_lazy_if(s,UTF)) {
4978 char tmpbuf[sizeof PL_tokenbuf];
4980 scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
4981 if ((t2 = keyword(tmpbuf, len, 0))) {
4982 /* binary operators exclude handle interpretations */
4994 PL_expect = XTERM; /* e.g. print $fh length() */
4999 PL_expect = XTERM; /* e.g. print $fh subr() */
5002 else if (isDIGIT(*s))
5003 PL_expect = XTERM; /* e.g. print $fh 3 */
5004 else if (*s == '.' && isDIGIT(s[1]))
5005 PL_expect = XTERM; /* e.g. print $fh .3 */
5006 else if ((*s == '?' || *s == '-' || *s == '+')
5007 && !isSPACE(s[1]) && s[1] != '=')
5008 PL_expect = XTERM; /* e.g. print $fh -1 */
5009 else if (*s == '/' && !isSPACE(s[1]) && s[1] != '='
5011 PL_expect = XTERM; /* e.g. print $fh /.../
5012 XXX except DORDOR operator
5014 else if (*s == '<' && s[1] == '<' && !isSPACE(s[2])
5016 PL_expect = XTERM; /* print $fh <<"EOF" */
5019 PL_pending_ident = '$';
5023 if (PL_expect == XOPERATOR)
5025 PL_tokenbuf[0] = '@';
5026 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1, FALSE);
5027 if (!PL_tokenbuf[1]) {
5030 if (PL_lex_state == LEX_NORMAL)
5032 if ((PL_expect != XREF || PL_oldoldbufptr == PL_last_lop) && intuit_more(s)) {
5034 PL_tokenbuf[0] = '%';
5036 /* Warn about @ where they meant $. */
5037 if (*s == '[' || *s == '{') {
5038 if (ckWARN(WARN_SYNTAX)) {
5039 const char *t = s + 1;
5040 while (*t && (isALNUM_lazy_if(t,UTF) || strchr(" \t$#+-'\"", *t)))
5042 if (*t == '}' || *t == ']') {
5044 PL_bufptr = PEEKSPACE(PL_bufptr); /* XXX can realloc */
5045 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
5046 "Scalar value %.*s better written as $%.*s",
5047 (int)(t-PL_bufptr), PL_bufptr,
5048 (int)(t-PL_bufptr-1), PL_bufptr+1);
5053 PL_pending_ident = '@';
5056 case '/': /* may be division, defined-or, or pattern */
5057 if (PL_expect == XTERMORDORDOR && s[1] == '/') {
5061 case '?': /* may either be conditional or pattern */
5062 if (PL_expect == XSTATE && s[1] == '?' && s[2] == '?') {
5066 if (PL_expect == XOPERATOR) {
5074 /* A // operator. */
5084 /* Disable warning on "study /blah/" */
5085 if (PL_oldoldbufptr == PL_last_uni
5086 && (*PL_last_uni != 's' || s - PL_last_uni < 5
5087 || memNE(PL_last_uni, "study", 5)
5088 || isALNUM_lazy_if(PL_last_uni+5,UTF)
5091 s = scan_pat(s,OP_MATCH);
5092 TERM(sublex_start());
5096 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack
5097 #ifdef PERL_STRICT_CR
5100 && (s[1] == '\n' || (s[1] == '\r' && s[2] == '\n'))
5102 && (s == PL_linestart || s[-1] == '\n') )
5104 PL_lex_formbrack = 0;
5108 if (PL_expect == XSTATE && s[1] == '.' && s[2] == '.') {
5112 if (PL_expect == XOPERATOR || !isDIGIT(s[1])) {
5118 pl_yylval.ival = OPf_SPECIAL;
5124 if (PL_expect != XOPERATOR)
5129 case '0': case '1': case '2': case '3': case '4':
5130 case '5': case '6': case '7': case '8': case '9':
5131 s = scan_num(s, &pl_yylval);
5132 DEBUG_T( { printbuf("### Saw number in %s\n", s); } );
5133 if (PL_expect == XOPERATOR)
5138 s = scan_str(s,!!PL_madskills,FALSE);
5139 DEBUG_T( { printbuf("### Saw string before %s\n", s); } );
5140 if (PL_expect == XOPERATOR) {
5141 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
5143 deprecate_old(commaless_variable_list);
5144 return REPORT(','); /* grandfather non-comma-format format */
5151 pl_yylval.ival = OP_CONST;
5152 TERM(sublex_start());
5155 s = scan_str(s,!!PL_madskills,FALSE);
5156 DEBUG_T( { printbuf("### Saw string before %s\n", s); } );
5157 if (PL_expect == XOPERATOR) {
5158 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
5160 deprecate_old(commaless_variable_list);
5161 return REPORT(','); /* grandfather non-comma-format format */
5168 pl_yylval.ival = OP_CONST;
5169 /* FIXME. I think that this can be const if char *d is replaced by
5170 more localised variables. */
5171 for (d = SvPV(PL_lex_stuff, len); len; len--, d++) {
5172 if (*d == '$' || *d == '@' || *d == '\\' || !UTF8_IS_INVARIANT((U8)*d)) {
5173 pl_yylval.ival = OP_STRINGIFY;
5177 TERM(sublex_start());
5180 s = scan_str(s,!!PL_madskills,FALSE);
5181 DEBUG_T( { printbuf("### Saw backtick string before %s\n", s); } );
5182 if (PL_expect == XOPERATOR)
5183 no_op("Backticks",s);
5186 readpipe_override();
5187 TERM(sublex_start());
5191 if (PL_lex_inwhat && isDIGIT(*s) && ckWARN(WARN_SYNTAX))
5192 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),"Can't use \\%c to mean $%c in expression",
5194 if (PL_expect == XOPERATOR)
5195 no_op("Backslash",s);
5199 if (isDIGIT(s[1]) && PL_expect != XOPERATOR) {
5200 char *start = s + 2;
5201 while (isDIGIT(*start) || *start == '_')
5203 if (*start == '.' && isDIGIT(start[1])) {
5204 s = scan_num(s, &pl_yylval);
5207 /* avoid v123abc() or $h{v1}, allow C<print v10;> */
5208 else if (!isALPHA(*start) && (PL_expect == XTERM
5209 || PL_expect == XREF || PL_expect == XSTATE
5210 || PL_expect == XTERMORDORDOR)) {
5211 GV *const gv = gv_fetchpvn_flags(s, start - s, 0, SVt_PVCV);
5213 s = scan_num(s, &pl_yylval);
5220 if (isDIGIT(s[1]) && PL_expect == XOPERATOR) {
5262 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
5264 /* Some keywords can be followed by any delimiter, including ':' */
5265 tmp = ((len == 1 && strchr("msyq", PL_tokenbuf[0])) ||
5266 (len == 2 && ((PL_tokenbuf[0] == 't' && PL_tokenbuf[1] == 'r') ||
5267 (PL_tokenbuf[0] == 'q' &&
5268 strchr("qwxr", PL_tokenbuf[1])))));
5270 /* x::* is just a word, unless x is "CORE" */
5271 if (!tmp && *s == ':' && s[1] == ':' && strNE(PL_tokenbuf, "CORE"))
5275 while (d < PL_bufend && isSPACE(*d))
5276 d++; /* no comments skipped here, or s### is misparsed */
5278 /* Is this a label? */
5279 if (!tmp && PL_expect == XSTATE
5280 && d < PL_bufend && *d == ':' && *(d + 1) != ':') {
5282 pl_yylval.pval = CopLABEL_alloc(PL_tokenbuf);
5287 /* Check for keywords */
5288 tmp = keyword(PL_tokenbuf, len, 0);
5290 /* Is this a word before a => operator? */
5291 if (*d == '=' && d[1] == '>') {
5294 = (OP*)newSVOP(OP_CONST, 0,
5295 S_newSV_maybe_utf8(aTHX_ PL_tokenbuf, len));
5296 pl_yylval.opval->op_private = OPpCONST_BARE;
5300 if (tmp < 0) { /* second-class keyword? */
5301 GV *ogv = NULL; /* override (winner) */
5302 GV *hgv = NULL; /* hidden (loser) */
5303 if (PL_expect != XOPERATOR && (*s != ':' || s[1] != ':')) {
5305 if ((gv = gv_fetchpvn_flags(PL_tokenbuf, len, 0, SVt_PVCV)) &&
5308 if (GvIMPORTED_CV(gv))
5310 else if (! CvMETHOD(cv))
5314 (gvp = (GV**)hv_fetch(PL_globalstash,PL_tokenbuf,len,FALSE)) &&
5315 (gv = *gvp) && isGV_with_GP(gv) &&
5316 GvCVu(gv) && GvIMPORTED_CV(gv))
5323 tmp = 0; /* overridden by import or by GLOBAL */
5326 && -tmp==KEY_lock /* XXX generalizable kludge */
5329 tmp = 0; /* any sub overrides "weak" keyword */
5331 else { /* no override */
5333 if (tmp == KEY_dump && ckWARN(WARN_MISC)) {
5334 Perl_warner(aTHX_ packWARN(WARN_MISC),
5335 "dump() better written as CORE::dump()");
5339 if (hgv && tmp != KEY_x && tmp != KEY_CORE
5340 && ckWARN(WARN_AMBIGUOUS)) /* never ambiguous */
5341 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
5342 "Ambiguous call resolved as CORE::%s(), %s",
5343 GvENAME(hgv), "qualify as such or use &");
5350 default: /* not a keyword */
5351 /* Trade off - by using this evil construction we can pull the
5352 variable gv into the block labelled keylookup. If not, then
5353 we have to give it function scope so that the goto from the
5354 earlier ':' case doesn't bypass the initialisation. */
5356 just_a_word_zero_gv:
5364 const char lastchar = (PL_bufptr == PL_oldoldbufptr ? 0 : PL_bufptr[-1]);
5367 SV *nextPL_nextwhite = 0;
5371 /* Get the rest if it looks like a package qualifier */
5373 if (*s == '\'' || (*s == ':' && s[1] == ':')) {
5375 s = scan_word(s, PL_tokenbuf + len, sizeof PL_tokenbuf - len,
5378 Perl_croak(aTHX_ "Bad name after %s%s", PL_tokenbuf,
5379 *s == '\'' ? "'" : "::");
5384 if (PL_expect == XOPERATOR) {
5385 if (PL_bufptr == PL_linestart) {
5386 CopLINE_dec(PL_curcop);
5387 Perl_warner(aTHX_ packWARN(WARN_SEMICOLON), "%s", PL_warn_nosemi);
5388 CopLINE_inc(PL_curcop);
5391 no_op("Bareword",s);
5394 /* Look for a subroutine with this name in current package,
5395 unless name is "Foo::", in which case Foo is a bearword
5396 (and a package name). */
5398 if (len > 2 && !PL_madskills &&
5399 PL_tokenbuf[len - 2] == ':' && PL_tokenbuf[len - 1] == ':')
5401 if (ckWARN(WARN_BAREWORD)
5402 && ! gv_fetchpvn_flags(PL_tokenbuf, len, 0, SVt_PVHV))
5403 Perl_warner(aTHX_ packWARN(WARN_BAREWORD),
5404 "Bareword \"%s\" refers to nonexistent package",
5407 PL_tokenbuf[len] = '\0';
5413 /* Mustn't actually add anything to a symbol table.
5414 But also don't want to "initialise" any placeholder
5415 constants that might already be there into full
5416 blown PVGVs with attached PVCV. */
5417 gv = gv_fetchpvn_flags(PL_tokenbuf, len,
5418 GV_NOADD_NOINIT, SVt_PVCV);
5423 /* if we saw a global override before, get the right name */
5426 sv = newSVpvs("CORE::GLOBAL::");
5427 sv_catpv(sv,PL_tokenbuf);
5430 /* If len is 0, newSVpv does strlen(), which is correct.
5431 If len is non-zero, then it will be the true length,
5432 and so the scalar will be created correctly. */
5433 sv = newSVpv(PL_tokenbuf,len);
5436 if (PL_madskills && !PL_thistoken) {
5437 char *start = SvPVX(PL_linestr) + PL_realtokenstart;
5438 PL_thistoken = newSVpvn(start,s - start);
5439 PL_realtokenstart = s - SvPVX(PL_linestr);
5443 /* Presume this is going to be a bareword of some sort. */
5446 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
5447 pl_yylval.opval->op_private = OPpCONST_BARE;
5448 /* UTF-8 package name? */
5449 if (UTF && !IN_BYTES &&
5450 is_utf8_string((U8*)SvPVX_const(sv), SvCUR(sv)))
5453 /* And if "Foo::", then that's what it certainly is. */
5458 /* Do the explicit type check so that we don't need to force
5459 the initialisation of the symbol table to have a real GV.
5460 Beware - gv may not really be a PVGV, cv may not really be
5461 a PVCV, (because of the space optimisations that gv_init
5462 understands) But they're true if for this symbol there is
5463 respectively a typeglob and a subroutine.
5465 cv = gv ? ((SvTYPE(gv) == SVt_PVGV)
5466 /* Real typeglob, so get the real subroutine: */
5468 /* A proxy for a subroutine in this package? */
5469 : SvOK(gv) ? MUTABLE_CV(gv) : NULL)
5472 /* See if it's the indirect object for a list operator. */
5474 if (PL_oldoldbufptr &&
5475 PL_oldoldbufptr < PL_bufptr &&
5476 (PL_oldoldbufptr == PL_last_lop
5477 || PL_oldoldbufptr == PL_last_uni) &&
5478 /* NO SKIPSPACE BEFORE HERE! */
5479 (PL_expect == XREF ||
5480 ((PL_opargs[PL_last_lop_op] >> OASHIFT)& 7) == OA_FILEREF))
5482 bool immediate_paren = *s == '(';
5484 /* (Now we can afford to cross potential line boundary.) */
5485 s = SKIPSPACE2(s,nextPL_nextwhite);
5487 PL_nextwhite = nextPL_nextwhite; /* assume no & deception */
5490 /* Two barewords in a row may indicate method call. */
5492 if ((isIDFIRST_lazy_if(s,UTF) || *s == '$') &&
5493 (tmp = intuit_method(s, gv, cv)))
5496 /* If not a declared subroutine, it's an indirect object. */
5497 /* (But it's an indir obj regardless for sort.) */
5498 /* Also, if "_" follows a filetest operator, it's a bareword */
5501 ( !immediate_paren && (PL_last_lop_op == OP_SORT ||
5503 (PL_last_lop_op != OP_MAPSTART &&
5504 PL_last_lop_op != OP_GREPSTART))))
5505 || (PL_tokenbuf[0] == '_' && PL_tokenbuf[1] == '\0'
5506 && ((PL_opargs[PL_last_lop_op] & OA_CLASS_MASK) == OA_FILESTATOP))
5509 PL_expect = (PL_last_lop == PL_oldoldbufptr) ? XTERM : XOPERATOR;
5514 PL_expect = XOPERATOR;
5517 s = SKIPSPACE2(s,nextPL_nextwhite);
5518 PL_nextwhite = nextPL_nextwhite;
5523 /* Is this a word before a => operator? */
5524 if (*s == '=' && s[1] == '>' && !pkgname) {
5526 sv_setpv(((SVOP*)pl_yylval.opval)->op_sv, PL_tokenbuf);
5527 if (UTF && !IN_BYTES && is_utf8_string((U8*)PL_tokenbuf, len))
5528 SvUTF8_on(((SVOP*)pl_yylval.opval)->op_sv);
5532 /* If followed by a paren, it's certainly a subroutine. */
5537 while (SPACE_OR_TAB(*d))
5539 if (*d == ')' && (sv = gv_const_sv(gv))) {
5546 PL_nextwhite = PL_thiswhite;
5549 start_force(PL_curforce);
5551 NEXTVAL_NEXTTOKE.opval = pl_yylval.opval;
5552 PL_expect = XOPERATOR;
5555 PL_nextwhite = nextPL_nextwhite;
5556 curmad('X', PL_thistoken);
5557 PL_thistoken = newSVpvs("");
5565 /* If followed by var or block, call it a method (unless sub) */
5567 if ((*s == '$' || *s == '{') && (!gv || !cv)) {
5568 PL_last_lop = PL_oldbufptr;
5569 PL_last_lop_op = OP_METHOD;
5573 /* If followed by a bareword, see if it looks like indir obj. */
5576 && (isIDFIRST_lazy_if(s,UTF) || *s == '$')
5577 && (tmp = intuit_method(s, gv, cv)))
5580 /* Not a method, so call it a subroutine (if defined) */
5583 if (lastchar == '-' && ckWARN_d(WARN_AMBIGUOUS))
5584 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
5585 "Ambiguous use of -%s resolved as -&%s()",
5586 PL_tokenbuf, PL_tokenbuf);
5587 /* Check for a constant sub */
5588 if ((sv = gv_const_sv(gv))) {
5590 SvREFCNT_dec(((SVOP*)pl_yylval.opval)->op_sv);
5591 ((SVOP*)pl_yylval.opval)->op_sv = SvREFCNT_inc_simple(sv);
5592 pl_yylval.opval->op_private = 0;
5596 /* Resolve to GV now. */
5597 if (SvTYPE(gv) != SVt_PVGV) {
5598 gv = gv_fetchpv(PL_tokenbuf, 0, SVt_PVCV);
5599 assert (SvTYPE(gv) == SVt_PVGV);
5600 /* cv must have been some sort of placeholder, so
5601 now needs replacing with a real code reference. */
5605 op_free(pl_yylval.opval);
5606 pl_yylval.opval = newCVREF(0, newGVOP(OP_GV, 0, gv));
5607 pl_yylval.opval->op_private |= OPpENTERSUB_NOPAREN;
5608 PL_last_lop = PL_oldbufptr;
5609 PL_last_lop_op = OP_ENTERSUB;
5610 /* Is there a prototype? */
5618 const char *proto = SvPV_const(MUTABLE_SV(cv), protolen);
5621 if ((*proto == '$' || *proto == '_') && proto[1] == '\0')
5623 while (*proto == ';')
5625 if (*proto == '&' && *s == '{') {
5627 sv_setpvs(PL_subname, "__ANON__");
5629 sv_setpvs(PL_subname, "__ANON__::__ANON__");
5636 PL_nextwhite = PL_thiswhite;
5639 start_force(PL_curforce);
5640 NEXTVAL_NEXTTOKE.opval = pl_yylval.opval;
5643 PL_nextwhite = nextPL_nextwhite;
5644 curmad('X', PL_thistoken);
5645 PL_thistoken = newSVpvs("");
5652 /* Guess harder when madskills require "best effort". */
5653 if (PL_madskills && (!gv || !GvCVu(gv))) {
5654 int probable_sub = 0;
5655 if (strchr("\"'`$@%0123456789!*+{[<", *s))
5657 else if (isALPHA(*s)) {
5661 d = scan_word(d, tmpbuf, sizeof tmpbuf, TRUE, &tmplen);
5662 if (!keyword(tmpbuf, tmplen, 0))
5665 while (d < PL_bufend && isSPACE(*d))
5667 if (*d == '=' && d[1] == '>')
5672 gv = gv_fetchpv(PL_tokenbuf, GV_ADD, SVt_PVCV);
5673 op_free(pl_yylval.opval);
5674 pl_yylval.opval = newCVREF(0, newGVOP(OP_GV, 0, gv));
5675 pl_yylval.opval->op_private |= OPpENTERSUB_NOPAREN;
5676 PL_last_lop = PL_oldbufptr;
5677 PL_last_lop_op = OP_ENTERSUB;
5678 PL_nextwhite = PL_thiswhite;
5680 start_force(PL_curforce);
5681 NEXTVAL_NEXTTOKE.opval = pl_yylval.opval;
5683 PL_nextwhite = nextPL_nextwhite;
5684 curmad('X', PL_thistoken);
5685 PL_thistoken = newSVpvs("");
5690 NEXTVAL_NEXTTOKE.opval = pl_yylval.opval;
5697 /* Call it a bare word */
5700 if (PL_hints & HINT_STRICT_SUBS)
5701 pl_yylval.opval->op_private |= OPpCONST_STRICT;
5703 if (lastchar != '-') {
5704 if (ckWARN(WARN_RESERVED)) {
5708 if (!*d && !gv_stashpv(PL_tokenbuf, 0))
5709 Perl_warner(aTHX_ packWARN(WARN_RESERVED), PL_warn_reserved,
5716 if ((lastchar == '*' || lastchar == '%' || lastchar == '&')
5717 && ckWARN_d(WARN_AMBIGUOUS)) {
5718 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
5719 "Operator or semicolon missing before %c%s",
5720 lastchar, PL_tokenbuf);
5721 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
5722 "Ambiguous use of %c resolved as operator %c",
5723 lastchar, lastchar);
5729 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0,
5730 newSVpv(CopFILE(PL_curcop),0));
5734 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0,
5735 Perl_newSVpvf(aTHX_ "%"IVdf, (IV)CopLINE(PL_curcop)));
5738 case KEY___PACKAGE__:
5739 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0,
5741 ? newSVhek(HvNAME_HEK(PL_curstash))
5748 if (PL_rsfp && (!PL_in_eval || PL_tokenbuf[2] == 'D')) {
5749 const char *pname = "main";
5750 if (PL_tokenbuf[2] == 'D')
5751 pname = HvNAME_get(PL_curstash ? PL_curstash : PL_defstash);
5752 gv = gv_fetchpv(Perl_form(aTHX_ "%s::DATA", pname), GV_ADD,
5756 GvIOp(gv) = newIO();
5757 IoIFP(GvIOp(gv)) = PL_rsfp;
5758 #if defined(HAS_FCNTL) && defined(F_SETFD)
5760 const int fd = PerlIO_fileno(PL_rsfp);
5761 fcntl(fd,F_SETFD,fd >= 3);
5764 /* Mark this internal pseudo-handle as clean */
5765 IoFLAGS(GvIOp(gv)) |= IOf_UNTAINT;
5766 if ((PerlIO*)PL_rsfp == PerlIO_stdin())
5767 IoTYPE(GvIOp(gv)) = IoTYPE_STD;
5769 IoTYPE(GvIOp(gv)) = IoTYPE_RDONLY;
5770 #if defined(WIN32) && !defined(PERL_TEXTMODE_SCRIPTS)
5771 /* if the script was opened in binmode, we need to revert
5772 * it to text mode for compatibility; but only iff it has CRs
5773 * XXX this is a questionable hack at best. */
5774 if (PL_bufend-PL_bufptr > 2
5775 && PL_bufend[-1] == '\n' && PL_bufend[-2] == '\r')
5778 if (IoTYPE(GvIOp(gv)) == IoTYPE_RDONLY) {
5779 loc = PerlIO_tell(PL_rsfp);
5780 (void)PerlIO_seek(PL_rsfp, 0L, 0);
5783 if (PerlLIO_setmode(PL_rsfp, O_TEXT) != -1) {
5785 if (PerlLIO_setmode(PerlIO_fileno(PL_rsfp), O_TEXT) != -1) {
5786 #endif /* NETWARE */
5787 #ifdef PERLIO_IS_STDIO /* really? */
5788 # if defined(__BORLANDC__)
5789 /* XXX see note in do_binmode() */
5790 ((FILE*)PL_rsfp)->flags &= ~_F_BIN;
5794 PerlIO_seek(PL_rsfp, loc, 0);
5798 #ifdef PERLIO_LAYERS
5801 PerlIO_apply_layers(aTHX_ PL_rsfp, NULL, ":utf8");
5802 else if (PL_encoding) {
5809 XPUSHs(PL_encoding);
5811 call_method("name", G_SCALAR);
5815 PerlIO_apply_layers(aTHX_ PL_rsfp, NULL,
5816 Perl_form(aTHX_ ":encoding(%"SVf")",
5825 if (PL_realtokenstart >= 0) {
5826 char *tstart = SvPVX(PL_linestr) + PL_realtokenstart;
5828 PL_endwhite = newSVpvs("");
5829 sv_catsv(PL_endwhite, PL_thiswhite);
5831 sv_catpvn(PL_endwhite, tstart, PL_bufend - tstart);
5832 PL_realtokenstart = -1;
5834 while ((s = filter_gets(PL_endwhite, PL_rsfp,
5835 SvCUR(PL_endwhite))) != NULL) ;
5850 if (PL_expect == XSTATE) {
5857 if (*s == ':' && s[1] == ':') {
5860 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
5861 if (!(tmp = keyword(PL_tokenbuf, len, 0)))
5862 Perl_croak(aTHX_ "CORE::%s is not a keyword", PL_tokenbuf);
5865 else if (tmp == KEY_require || tmp == KEY_do)
5866 /* that's a way to remember we saw "CORE::" */
5879 LOP(OP_ACCEPT,XTERM);
5885 LOP(OP_ATAN2,XTERM);
5891 LOP(OP_BINMODE,XTERM);
5894 LOP(OP_BLESS,XTERM);
5903 /* When 'use switch' is in effect, continue has a dual
5904 life as a control operator. */
5906 if (!FEATURE_IS_ENABLED("switch"))
5909 /* We have to disambiguate the two senses of
5910 "continue". If the next token is a '{' then
5911 treat it as the start of a continue block;
5912 otherwise treat it as a control operator.
5924 (void)gv_fetchpvs("ENV", GV_ADD|GV_NOTQUAL, SVt_PVHV);
5941 if (!PL_cryptseen) {
5942 PL_cryptseen = TRUE;
5946 LOP(OP_CRYPT,XTERM);
5949 LOP(OP_CHMOD,XTERM);
5952 LOP(OP_CHOWN,XTERM);
5955 LOP(OP_CONNECT,XTERM);
5974 s = force_word(s,WORD,TRUE,TRUE,FALSE);
5975 if (orig_keyword == KEY_do) {
5984 PL_hints |= HINT_BLOCK_SCOPE;
5994 gv_fetchpvs("AnyDBM_File::ISA", GV_ADDMULTI, SVt_PVAV);
5995 LOP(OP_DBMOPEN,XTERM);
6001 s = force_word(s,WORD,TRUE,FALSE,FALSE);
6008 pl_yylval.ival = CopLINE(PL_curcop);
6024 PL_expect = (*s == '{') ? XTERMBLOCK : XTERM;
6025 UNIBRACK(OP_ENTEREVAL);
6039 case KEY_endhostent:
6045 case KEY_endservent:
6048 case KEY_endprotoent:
6059 pl_yylval.ival = CopLINE(PL_curcop);
6061 if (PL_expect == XSTATE && isIDFIRST_lazy_if(s,UTF)) {
6064 int soff = s - SvPVX(PL_linestr); /* for skipspace realloc */
6067 if ((PL_bufend - p) >= 3 &&
6068 strnEQ(p, "my", 2) && isSPACE(*(p + 2)))
6070 else if ((PL_bufend - p) >= 4 &&
6071 strnEQ(p, "our", 3) && isSPACE(*(p + 3)))
6074 if (isIDFIRST_lazy_if(p,UTF)) {
6075 p = scan_ident(p, PL_bufend,
6076 PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
6080 Perl_croak(aTHX_ "Missing $ on loop variable");
6082 s = SvPVX(PL_linestr) + soff;
6088 LOP(OP_FORMLINE,XTERM);
6094 LOP(OP_FCNTL,XTERM);
6100 LOP(OP_FLOCK,XTERM);
6109 LOP(OP_GREPSTART, XREF);
6112 s = force_word(s,WORD,TRUE,FALSE,FALSE);
6127 case KEY_getpriority:
6128 LOP(OP_GETPRIORITY,XTERM);
6130 case KEY_getprotobyname:
6133 case KEY_getprotobynumber:
6134 LOP(OP_GPBYNUMBER,XTERM);
6136 case KEY_getprotoent:
6148 case KEY_getpeername:
6149 UNI(OP_GETPEERNAME);
6151 case KEY_gethostbyname:
6154 case KEY_gethostbyaddr:
6155 LOP(OP_GHBYADDR,XTERM);
6157 case KEY_gethostent:
6160 case KEY_getnetbyname:
6163 case KEY_getnetbyaddr:
6164 LOP(OP_GNBYADDR,XTERM);
6169 case KEY_getservbyname:
6170 LOP(OP_GSBYNAME,XTERM);
6172 case KEY_getservbyport:
6173 LOP(OP_GSBYPORT,XTERM);
6175 case KEY_getservent:
6178 case KEY_getsockname:
6179 UNI(OP_GETSOCKNAME);
6181 case KEY_getsockopt:
6182 LOP(OP_GSOCKOPT,XTERM);
6197 pl_yylval.ival = CopLINE(PL_curcop);
6207 pl_yylval.ival = CopLINE(PL_curcop);
6211 LOP(OP_INDEX,XTERM);
6217 LOP(OP_IOCTL,XTERM);
6229 s = force_word(s,WORD,TRUE,FALSE,FALSE);
6261 LOP(OP_LISTEN,XTERM);
6270 s = scan_pat(s,OP_MATCH);
6271 TERM(sublex_start());
6274 LOP(OP_MAPSTART, XREF);
6277 LOP(OP_MKDIR,XTERM);
6280 LOP(OP_MSGCTL,XTERM);
6283 LOP(OP_MSGGET,XTERM);
6286 LOP(OP_MSGRCV,XTERM);
6289 LOP(OP_MSGSND,XTERM);
6294 PL_in_my = (U16)tmp;
6296 if (isIDFIRST_lazy_if(s,UTF)) {
6300 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, TRUE, &len);
6301 if (len == 3 && strnEQ(PL_tokenbuf, "sub", 3))
6303 PL_in_my_stash = find_in_my_stash(PL_tokenbuf, len);
6304 if (!PL_in_my_stash) {
6307 my_snprintf(tmpbuf, sizeof(tmpbuf), "No such class %.1000s", PL_tokenbuf);
6311 if (PL_madskills) { /* just add type to declarator token */
6312 sv_catsv(PL_thistoken, PL_nextwhite);
6314 sv_catpvn(PL_thistoken, start, s - start);
6322 s = force_word(s,WORD,TRUE,FALSE,FALSE);
6329 s = tokenize_use(0, s);
6333 if (*s == '(' || (s = SKIPSPACE1(s), *s == '('))
6340 if (isIDFIRST_lazy_if(s,UTF)) {
6342 for (d = s; isALNUM_lazy_if(d,UTF);)
6344 for (t=d; isSPACE(*t);)
6346 if ( *t && strchr("|&*+-=!?:.", *t) && ckWARN_d(WARN_PRECEDENCE)
6348 && !(t[0] == '=' && t[1] == '>')
6350 int parms_len = (int)(d-s);
6351 Perl_warner(aTHX_ packWARN(WARN_PRECEDENCE),
6352 "Precedence problem: open %.*s should be open(%.*s)",
6353 parms_len, s, parms_len, s);
6359 pl_yylval.ival = OP_OR;
6369 LOP(OP_OPEN_DIR,XTERM);
6372 checkcomma(s,PL_tokenbuf,"filehandle");
6376 checkcomma(s,PL_tokenbuf,"filehandle");
6395 s = force_word(s,WORD,FALSE,TRUE,FALSE);
6399 LOP(OP_PIPE_OP,XTERM);
6402 s = scan_str(s,!!PL_madskills,FALSE);
6405 pl_yylval.ival = OP_CONST;
6406 TERM(sublex_start());
6412 s = scan_str(s,!!PL_madskills,FALSE);
6415 PL_expect = XOPERATOR;
6417 if (SvCUR(PL_lex_stuff)) {
6420 d = SvPV_force(PL_lex_stuff, len);
6422 for (; isSPACE(*d) && len; --len, ++d)
6427 if (!warned && ckWARN(WARN_QW)) {
6428 for (; !isSPACE(*d) && len; --len, ++d) {
6430 Perl_warner(aTHX_ packWARN(WARN_QW),
6431 "Possible attempt to separate words with commas");
6434 else if (*d == '#') {
6435 Perl_warner(aTHX_ packWARN(WARN_QW),
6436 "Possible attempt to put comments in qw() list");
6442 for (; !isSPACE(*d) && len; --len, ++d)
6445 sv = newSVpvn_utf8(b, d-b, DO_UTF8(PL_lex_stuff));
6446 words = append_elem(OP_LIST, words,
6447 newSVOP(OP_CONST, 0, tokeq(sv)));
6451 start_force(PL_curforce);
6452 NEXTVAL_NEXTTOKE.opval = words;
6457 SvREFCNT_dec(PL_lex_stuff);
6458 PL_lex_stuff = NULL;
6464 s = scan_str(s,!!PL_madskills,FALSE);
6467 pl_yylval.ival = OP_STRINGIFY;
6468 if (SvIVX(PL_lex_stuff) == '\'')
6469 SvIV_set(PL_lex_stuff, 0); /* qq'$foo' should intepolate */
6470 TERM(sublex_start());
6473 s = scan_pat(s,OP_QR);
6474 TERM(sublex_start());
6477 s = scan_str(s,!!PL_madskills,FALSE);
6480 readpipe_override();
6481 TERM(sublex_start());
6489 s = force_version(s, FALSE);
6491 else if (*s != 'v' || !isDIGIT(s[1])
6492 || (s = force_version(s, TRUE), *s == 'v'))
6494 *PL_tokenbuf = '\0';
6495 s = force_word(s,WORD,TRUE,TRUE,FALSE);
6496 if (isIDFIRST_lazy_if(PL_tokenbuf,UTF))
6497 gv_stashpvn(PL_tokenbuf, strlen(PL_tokenbuf), GV_ADD);
6499 yyerror("<> should be quotes");
6501 if (orig_keyword == KEY_require) {
6509 PL_last_uni = PL_oldbufptr;
6510 PL_last_lop_op = OP_REQUIRE;
6512 return REPORT( (int)REQUIRE );
6518 s = force_word(s,WORD,TRUE,FALSE,FALSE);
6522 LOP(OP_RENAME,XTERM);
6531 LOP(OP_RINDEX,XTERM);
6540 UNIDOR(OP_READLINE);
6543 UNIDOR(OP_BACKTICK);
6552 LOP(OP_REVERSE,XTERM);
6555 UNIDOR(OP_READLINK);
6562 if (pl_yylval.opval)
6563 TERM(sublex_start());
6565 TOKEN(1); /* force error */
6568 checkcomma(s,PL_tokenbuf,"filehandle");
6578 LOP(OP_SELECT,XTERM);
6584 LOP(OP_SEMCTL,XTERM);
6587 LOP(OP_SEMGET,XTERM);
6590 LOP(OP_SEMOP,XTERM);
6596 LOP(OP_SETPGRP,XTERM);
6598 case KEY_setpriority:
6599 LOP(OP_SETPRIORITY,XTERM);
6601 case KEY_sethostent:
6607 case KEY_setservent:
6610 case KEY_setprotoent:
6620 LOP(OP_SEEKDIR,XTERM);
6622 case KEY_setsockopt:
6623 LOP(OP_SSOCKOPT,XTERM);
6629 LOP(OP_SHMCTL,XTERM);
6632 LOP(OP_SHMGET,XTERM);
6635 LOP(OP_SHMREAD,XTERM);
6638 LOP(OP_SHMWRITE,XTERM);
6641 LOP(OP_SHUTDOWN,XTERM);
6650 LOP(OP_SOCKET,XTERM);
6652 case KEY_socketpair:
6653 LOP(OP_SOCKPAIR,XTERM);
6656 checkcomma(s,PL_tokenbuf,"subroutine name");
6658 if (*s == ';' || *s == ')') /* probably a close */
6659 Perl_croak(aTHX_ "sort is now a reserved word");
6661 s = force_word(s,WORD,TRUE,TRUE,FALSE);
6665 LOP(OP_SPLIT,XTERM);
6668 LOP(OP_SPRINTF,XTERM);
6671 LOP(OP_SPLICE,XTERM);
6686 LOP(OP_SUBSTR,XTERM);
6692 char tmpbuf[sizeof PL_tokenbuf];
6693 SSize_t tboffset = 0;
6694 expectation attrful;
6695 bool have_name, have_proto;
6696 const int key = tmp;
6701 char *tstart = SvPVX(PL_linestr) + PL_realtokenstart;
6702 SV *subtoken = newSVpvn(tstart, s - tstart);
6706 s = SKIPSPACE2(s,tmpwhite);
6711 if (isIDFIRST_lazy_if(s,UTF) || *s == '\'' ||
6712 (*s == ':' && s[1] == ':'))
6715 SV *nametoke = NULL;
6719 attrful = XATTRBLOCK;
6720 /* remember buffer pos'n for later force_word */
6721 tboffset = s - PL_oldbufptr;
6722 d = scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
6725 nametoke = newSVpvn(s, d - s);
6727 if (memchr(tmpbuf, ':', len))
6728 sv_setpvn(PL_subname, tmpbuf, len);
6730 sv_setsv(PL_subname,PL_curstname);
6731 sv_catpvs(PL_subname,"::");
6732 sv_catpvn(PL_subname,tmpbuf,len);
6739 CURMAD('X', nametoke);
6740 CURMAD('_', tmpwhite);
6741 (void) force_word(PL_oldbufptr + tboffset, WORD,
6744 s = SKIPSPACE2(d,tmpwhite);
6751 Perl_croak(aTHX_ "Missing name in \"my sub\"");
6752 PL_expect = XTERMBLOCK;
6753 attrful = XATTRTERM;
6754 sv_setpvs(PL_subname,"?");
6758 if (key == KEY_format) {
6760 PL_lex_formbrack = PL_lex_brackets + 1;
6762 PL_thistoken = subtoken;
6766 (void) force_word(PL_oldbufptr + tboffset, WORD,
6772 /* Look for a prototype */
6775 bool bad_proto = FALSE;
6776 bool in_brackets = FALSE;
6777 char greedy_proto = ' ';
6778 bool proto_after_greedy_proto = FALSE;
6779 bool must_be_last = FALSE;
6780 bool underscore = FALSE;
6781 bool seen_underscore = FALSE;
6782 const bool warnsyntax = ckWARN(WARN_SYNTAX);
6784 s = scan_str(s,!!PL_madskills,FALSE);
6786 Perl_croak(aTHX_ "Prototype not terminated");
6787 /* strip spaces and check for bad characters */
6788 d = SvPVX(PL_lex_stuff);
6790 for (p = d; *p; ++p) {
6796 proto_after_greedy_proto = TRUE;
6797 if (!strchr("$@%*;[]&\\_", *p)) {
6809 else if ( *p == ']' ) {
6810 in_brackets = FALSE;
6812 else if ( (*p == '@' || *p == '%') &&
6813 ( tmp < 2 || d[tmp-2] != '\\' ) &&
6815 must_be_last = TRUE;
6818 else if ( *p == '_' ) {
6819 underscore = seen_underscore = TRUE;
6826 if (proto_after_greedy_proto)
6827 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
6828 "Prototype after '%c' for %"SVf" : %s",
6829 greedy_proto, SVfARG(PL_subname), d);
6831 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
6832 "Illegal character %sin prototype for %"SVf" : %s",
6833 seen_underscore ? "after '_' " : "",
6834 SVfARG(PL_subname), d);
6835 SvCUR_set(PL_lex_stuff, tmp);
6840 CURMAD('q', PL_thisopen);
6841 CURMAD('_', tmpwhite);
6842 CURMAD('=', PL_thisstuff);
6843 CURMAD('Q', PL_thisclose);
6844 NEXTVAL_NEXTTOKE.opval =
6845 (OP*)newSVOP(OP_CONST, 0, PL_lex_stuff);
6846 PL_lex_stuff = NULL;
6849 s = SKIPSPACE2(s,tmpwhite);
6857 if (*s == ':' && s[1] != ':')
6858 PL_expect = attrful;
6859 else if (*s != '{' && key == KEY_sub) {
6861 Perl_croak(aTHX_ "Illegal declaration of anonymous subroutine");
6863 Perl_croak(aTHX_ "Illegal declaration of subroutine %"SVf, SVfARG(PL_subname));
6870 curmad('^', newSVpvs(""));
6871 CURMAD('_', tmpwhite);
6875 PL_thistoken = subtoken;
6878 NEXTVAL_NEXTTOKE.opval =
6879 (OP*)newSVOP(OP_CONST, 0, PL_lex_stuff);
6880 PL_lex_stuff = NULL;
6886 sv_setpvs(PL_subname, "__ANON__");
6888 sv_setpvs(PL_subname, "__ANON__::__ANON__");
6892 (void) force_word(PL_oldbufptr + tboffset, WORD,
6901 LOP(OP_SYSTEM,XREF);
6904 LOP(OP_SYMLINK,XTERM);
6907 LOP(OP_SYSCALL,XTERM);
6910 LOP(OP_SYSOPEN,XTERM);
6913 LOP(OP_SYSSEEK,XTERM);
6916 LOP(OP_SYSREAD,XTERM);
6919 LOP(OP_SYSWRITE,XTERM);
6923 TERM(sublex_start());
6944 LOP(OP_TRUNCATE,XTERM);
6956 pl_yylval.ival = CopLINE(PL_curcop);
6960 pl_yylval.ival = CopLINE(PL_curcop);
6964 LOP(OP_UNLINK,XTERM);
6970 LOP(OP_UNPACK,XTERM);
6973 LOP(OP_UTIME,XTERM);
6979 LOP(OP_UNSHIFT,XTERM);
6982 s = tokenize_use(1, s);
6992 pl_yylval.ival = CopLINE(PL_curcop);
6996 pl_yylval.ival = CopLINE(PL_curcop);
7000 PL_hints |= HINT_BLOCK_SCOPE;
7007 LOP(OP_WAITPID,XTERM);
7016 ctl_l[0] = toCTRL('L');
7018 gv_fetchpvn_flags(ctl_l, 1, GV_ADD|GV_NOTQUAL, SVt_PV);
7021 /* Make sure $^L is defined */
7022 gv_fetchpvs("\f", GV_ADD|GV_NOTQUAL, SVt_PV);
7027 if (PL_expect == XOPERATOR)
7033 pl_yylval.ival = OP_XOR;
7038 TERM(sublex_start());
7043 #pragma segment Main
7047 S_pending_ident(pTHX)
7052 /* pit holds the identifier we read and pending_ident is reset */
7053 char pit = PL_pending_ident;
7054 const STRLEN tokenbuf_len = strlen(PL_tokenbuf);
7055 /* All routes through this function want to know if there is a colon. */
7056 const char *const has_colon = (const char*) memchr (PL_tokenbuf, ':', tokenbuf_len);
7057 PL_pending_ident = 0;
7059 /* PL_realtokenstart = realtokenend = PL_bufptr - SvPVX(PL_linestr); */
7060 DEBUG_T({ PerlIO_printf(Perl_debug_log,
7061 "### Pending identifier '%s'\n", PL_tokenbuf); });
7063 /* if we're in a my(), we can't allow dynamics here.
7064 $foo'bar has already been turned into $foo::bar, so
7065 just check for colons.
7067 if it's a legal name, the OP is a PADANY.
7070 if (PL_in_my == KEY_our) { /* "our" is merely analogous to "my" */
7072 yyerror(Perl_form(aTHX_ "No package name allowed for "
7073 "variable %s in \"our\"",
7075 tmp = allocmy(PL_tokenbuf);
7079 yyerror(Perl_form(aTHX_ PL_no_myglob,
7080 PL_in_my == KEY_my ? "my" : "state", PL_tokenbuf));
7082 pl_yylval.opval = newOP(OP_PADANY, 0);
7083 pl_yylval.opval->op_targ = allocmy(PL_tokenbuf);
7089 build the ops for accesses to a my() variable.
7091 Deny my($a) or my($b) in a sort block, *if* $a or $b is
7092 then used in a comparison. This catches most, but not
7093 all cases. For instance, it catches
7094 sort { my($a); $a <=> $b }
7096 sort { my($a); $a < $b ? -1 : $a == $b ? 0 : 1; }
7097 (although why you'd do that is anyone's guess).
7102 tmp = pad_findmy(PL_tokenbuf);
7103 if (tmp != NOT_IN_PAD) {
7104 /* might be an "our" variable" */
7105 if (PAD_COMPNAME_FLAGS_isOUR(tmp)) {
7106 /* build ops for a bareword */
7107 HV * const stash = PAD_COMPNAME_OURSTASH(tmp);
7108 HEK * const stashname = HvNAME_HEK(stash);
7109 SV * const sym = newSVhek(stashname);
7110 sv_catpvs(sym, "::");
7111 sv_catpvn(sym, PL_tokenbuf+1, tokenbuf_len - 1);
7112 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sym);
7113 pl_yylval.opval->op_private = OPpCONST_ENTERED;
7116 ? (GV_ADDMULTI | GV_ADDINEVAL)
7119 ((PL_tokenbuf[0] == '$') ? SVt_PV
7120 : (PL_tokenbuf[0] == '@') ? SVt_PVAV
7125 /* if it's a sort block and they're naming $a or $b */
7126 if (PL_last_lop_op == OP_SORT &&
7127 PL_tokenbuf[0] == '$' &&
7128 (PL_tokenbuf[1] == 'a' || PL_tokenbuf[1] == 'b')
7131 for (d = PL_in_eval ? PL_oldoldbufptr : PL_linestart;
7132 d < PL_bufend && *d != '\n';
7135 if (strnEQ(d,"<=>",3) || strnEQ(d,"cmp",3)) {
7136 Perl_croak(aTHX_ "Can't use \"my %s\" in sort comparison",
7142 pl_yylval.opval = newOP(OP_PADANY, 0);
7143 pl_yylval.opval->op_targ = tmp;
7149 Whine if they've said @foo in a doublequoted string,
7150 and @foo isn't a variable we can find in the symbol
7153 if (pit == '@' && PL_lex_state != LEX_NORMAL && !PL_lex_brackets) {
7154 GV *const gv = gv_fetchpvn_flags(PL_tokenbuf + 1, tokenbuf_len - 1, 0,
7156 if ((!gv || ((PL_tokenbuf[0] == '@') ? !GvAV(gv) : !GvHV(gv)))
7157 && ckWARN(WARN_AMBIGUOUS)
7158 /* DO NOT warn for @- and @+ */
7159 && !( PL_tokenbuf[2] == '\0' &&
7160 ( PL_tokenbuf[1] == '-' || PL_tokenbuf[1] == '+' ))
7163 /* Downgraded from fatal to warning 20000522 mjd */
7164 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
7165 "Possible unintended interpolation of %s in string",
7170 /* build ops for a bareword */
7171 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, newSVpvn(PL_tokenbuf + 1,
7173 pl_yylval.opval->op_private = OPpCONST_ENTERED;
7175 PL_tokenbuf + 1, tokenbuf_len - 1,
7176 /* If the identifier refers to a stash, don't autovivify it.
7177 * Change 24660 had the side effect of causing symbol table
7178 * hashes to always be defined, even if they were freshly
7179 * created and the only reference in the entire program was
7180 * the single statement with the defined %foo::bar:: test.
7181 * It appears that all code in the wild doing this actually
7182 * wants to know whether sub-packages have been loaded, so
7183 * by avoiding auto-vivifying symbol tables, we ensure that
7184 * defined %foo::bar:: continues to be false, and the existing
7185 * tests still give the expected answers, even though what
7186 * they're actually testing has now changed subtly.
7188 (*PL_tokenbuf == '%'
7189 && *(d = PL_tokenbuf + tokenbuf_len - 1) == ':'
7192 : PL_in_eval ? (GV_ADDMULTI | GV_ADDINEVAL) : GV_ADD),
7193 ((PL_tokenbuf[0] == '$') ? SVt_PV
7194 : (PL_tokenbuf[0] == '@') ? SVt_PVAV
7200 * The following code was generated by perl_keyword.pl.
7204 Perl_keyword (pTHX_ const char *name, I32 len, bool all_keywords)
7208 PERL_ARGS_ASSERT_KEYWORD;
7212 case 1: /* 5 tokens of length 1 */
7244 case 2: /* 18 tokens of length 2 */
7390 case 3: /* 29 tokens of length 3 */
7394 if (name[1] == 'N' &&
7457 if (name[1] == 'i' &&
7489 if (name[1] == 'o' &&
7498 if (name[1] == 'e' &&
7507 if (name[1] == 'n' &&
7516 if (name[1] == 'o' &&
7525 if (name[1] == 'a' &&
7534 if (name[1] == 'o' &&
7596 if (name[1] == 'e' &&
7610 return (all_keywords || FEATURE_IS_ENABLED("say") ? KEY_say : 0);
7636 if (name[1] == 'i' &&
7645 if (name[1] == 's' &&
7654 if (name[1] == 'e' &&
7663 if (name[1] == 'o' &&
7675 case 4: /* 41 tokens of length 4 */
7679 if (name[1] == 'O' &&
7689 if (name[1] == 'N' &&
7699 if (name[1] == 'i' &&
7709 if (name[1] == 'h' &&
7719 if (name[1] == 'u' &&
7732 if (name[2] == 'c' &&
7741 if (name[2] == 's' &&
7750 if (name[2] == 'a' &&
7786 if (name[1] == 'o' &&
7799 if (name[2] == 't' &&
7808 if (name[2] == 'o' &&
7817 if (name[2] == 't' &&
7826 if (name[2] == 'e' &&
7839 if (name[1] == 'o' &&
7852 if (name[2] == 'y' &&
7861 if (name[2] == 'l' &&
7877 if (name[2] == 's' &&
7886 if (name[2] == 'n' &&
7895 if (name[2] == 'c' &&
7908 if (name[1] == 'e' &&
7918 if (name[1] == 'p' &&
7931 if (name[2] == 'c' &&
7940 if (name[2] == 'p' &&
7949 if (name[2] == 's' &&
7965 if (name[2] == 'n' &&
8035 if (name[2] == 'r' &&
8044 if (name[2] == 'r' &&
8053 if (name[2] == 'a' &&
8069 if (name[2] == 'l' &&
8131 if (name[2] == 'e' &&
8134 return (all_keywords || FEATURE_IS_ENABLED("switch") ? KEY_when : 0);
8147 case 5: /* 39 tokens of length 5 */
8151 if (name[1] == 'E' &&
8162 if (name[1] == 'H' &&
8176 if (name[2] == 'a' &&
8186 if (name[2] == 'a' &&
8203 if (name[2] == 'e' &&
8213 if (name[2] == 'e' &&
8217 return (all_keywords || FEATURE_IS_ENABLED("switch") ? -KEY_break : 0);
8233 if (name[3] == 'i' &&
8242 if (name[3] == 'o' &&
8278 if (name[2] == 'o' &&
8288 if (name[2] == 'y' &&
8302 if (name[1] == 'l' &&
8316 if (name[2] == 'n' &&
8326 if (name[2] == 'o' &&
8340 if (name[1] == 'i' &&
8345 return (all_keywords || FEATURE_IS_ENABLED("switch") ? KEY_given : 0);
8354 if (name[2] == 'd' &&
8364 if (name[2] == 'c' &&
8381 if (name[2] == 'c' &&
8391 if (name[2] == 't' &&
8405 if (name[1] == 'k' &&
8416 if (name[1] == 'r' &&
8430 if (name[2] == 's' &&
8440 if (name[2] == 'd' &&
8457 if (name[2] == 'm' &&
8467 if (name[2] == 'i' &&
8477 if (name[2] == 'e' &&
8487 if (name[2] == 'l' &&
8497 if (name[2] == 'a' &&
8510 if (name[3] == 't' &&
8513 return (all_keywords || FEATURE_IS_ENABLED("state") ? KEY_state : 0);
8519 if (name[3] == 'd' &&
8536 if (name[1] == 'i' &&
8550 if (name[2] == 'a' &&
8563 if (name[3] == 'e' &&
8598 if (name[2] == 'i' &&
8615 if (name[2] == 'i' &&
8625 if (name[2] == 'i' &&
8642 case 6: /* 33 tokens of length 6 */
8646 if (name[1] == 'c' &&
8661 if (name[2] == 'l' &&
8672 if (name[2] == 'r' &&
8687 if (name[1] == 'e' &&
8702 if (name[2] == 's' &&
8707 if(ckWARN_d(WARN_SYNTAX))
8708 Perl_warner(aTHX_ packWARN(WARN_SYNTAX), "elseif should be elsif");
8714 if (name[2] == 'i' &&
8732 if (name[2] == 'l' &&
8743 if (name[2] == 'r' &&
8758 if (name[1] == 'm' &&
8773 if (name[2] == 'n' &&
8784 if (name[2] == 's' &&
8799 if (name[1] == 's' &&
8805 if (name[4] == 't' &&
8814 if (name[4] == 'e' &&
8823 if (name[4] == 'c' &&
8832 if (name[4] == 'n' &&
8848 if (name[1] == 'r' &&
8866 if (name[3] == 'a' &&
8876 if (name[3] == 'u' &&
8890 if (name[2] == 'n' &&
8908 if (name[2] == 'a' &&
8922 if (name[3] == 'e' &&
8935 if (name[4] == 't' &&
8944 if (name[4] == 'e' &&
8966 if (name[4] == 't' &&
8975 if (name[4] == 'e' &&
8991 if (name[2] == 'c' &&
9002 if (name[2] == 'l' &&
9013 if (name[2] == 'b' &&
9024 if (name[2] == 's' &&
9047 if (name[4] == 's' &&
9056 if (name[4] == 'n' &&
9069 if (name[3] == 'a' &&
9086 if (name[1] == 'a' &&
9101 case 7: /* 29 tokens of length 7 */
9105 if (name[1] == 'E' &&
9118 if (name[1] == '_' &&
9131 if (name[1] == 'i' &&
9138 return -KEY_binmode;
9144 if (name[1] == 'o' &&
9151 return -KEY_connect;
9160 if (name[2] == 'm' &&
9166 return -KEY_dbmopen;
9177 if (name[4] == 'u' &&
9181 return (all_keywords || FEATURE_IS_ENABLED("switch") ? KEY_default : 0);
9187 if (name[4] == 'n' &&
9208 if (name[1] == 'o' &&
9221 if (name[1] == 'e' &&
9228 if (name[5] == 'r' &&
9231 return -KEY_getpgrp;
9237 if (name[5] == 'i' &&
9240 return -KEY_getppid;
9253 if (name[1] == 'c' &&
9260 return -KEY_lcfirst;
9266 if (name[1] == 'p' &&
9273 return -KEY_opendir;
9279 if (name[1] == 'a' &&
9297 if (name[3] == 'd' &&
9302 return -KEY_readdir;
9308 if (name[3] == 'u' &&
9319 if (name[3] == 'e' &&
9324 return -KEY_reverse;
9343 if (name[3] == 'k' &&
9348 return -KEY_seekdir;
9354 if (name[3] == 'p' &&
9359 return -KEY_setpgrp;
9369 if (name[2] == 'm' &&
9375 return -KEY_shmread;
9381 if (name[2] == 'r' &&
9387 return -KEY_sprintf;
9396 if (name[3] == 'l' &&
9401 return -KEY_symlink;
9410 if (name[4] == 'a' &&
9414 return -KEY_syscall;
9420 if (name[4] == 'p' &&
9424 return -KEY_sysopen;
9430 if (name[4] == 'e' &&
9434 return -KEY_sysread;
9440 if (name[4] == 'e' &&
9444 return -KEY_sysseek;
9462 if (name[1] == 'e' &&
9469 return -KEY_telldir;
9478 if (name[2] == 'f' &&
9484 return -KEY_ucfirst;
9490 if (name[2] == 's' &&
9496 return -KEY_unshift;
9506 if (name[1] == 'a' &&
9513 return -KEY_waitpid;
9522 case 8: /* 26 tokens of length 8 */
9526 if (name[1] == 'U' &&
9534 return KEY_AUTOLOAD;
9545 if (name[3] == 'A' &&
9551 return KEY___DATA__;
9557 if (name[3] == 'I' &&
9563 return -KEY___FILE__;
9569 if (name[3] == 'I' &&
9575 return -KEY___LINE__;
9591 if (name[2] == 'o' &&
9598 return -KEY_closedir;
9604 if (name[2] == 'n' &&
9611 return -KEY_continue;
9621 if (name[1] == 'b' &&
9629 return -KEY_dbmclose;
9635 if (name[1] == 'n' &&
9641 if (name[4] == 'r' &&
9646 return -KEY_endgrent;
9652 if (name[4] == 'w' &&
9657 return -KEY_endpwent;
9670 if (name[1] == 'o' &&
9678 return -KEY_formline;
9684 if (name[1] == 'e' &&
9695 if (name[6] == 'n' &&
9698 return -KEY_getgrent;
9704 if (name[6] == 'i' &&
9707 return -KEY_getgrgid;
9713 if (name[6] == 'a' &&
9716 return -KEY_getgrnam;
9729 if (name[4] == 'o' &&
9734 return -KEY_getlogin;
9745 if (name[6] == 'n' &&
9748 return -KEY_getpwent;
9754 if (name[6] == 'a' &&
9757 return -KEY_getpwnam;
9763 if (name[6] == 'i' &&
9766 return -KEY_getpwuid;
9786 if (name[1] == 'e' &&
9793 if (name[5] == 'i' &&
9800 return -KEY_readline;
9805 return -KEY_readlink;
9816 if (name[5] == 'i' &&
9820 return -KEY_readpipe;
9841 if (name[4] == 'r' &&
9846 return -KEY_setgrent;
9852 if (name[4] == 'w' &&
9857 return -KEY_setpwent;
9873 if (name[3] == 'w' &&
9879 return -KEY_shmwrite;
9885 if (name[3] == 't' &&
9891 return -KEY_shutdown;
9901 if (name[2] == 's' &&
9908 return -KEY_syswrite;
9918 if (name[1] == 'r' &&
9926 return -KEY_truncate;
9935 case 9: /* 9 tokens of length 9 */
9939 if (name[1] == 'N' &&
9948 return KEY_UNITCHECK;
9954 if (name[1] == 'n' &&
9963 return -KEY_endnetent;
9969 if (name[1] == 'e' &&
9978 return -KEY_getnetent;
9984 if (name[1] == 'o' &&
9993 return -KEY_localtime;
9999 if (name[1] == 'r' &&
10008 return KEY_prototype;
10014 if (name[1] == 'u' &&
10023 return -KEY_quotemeta;
10029 if (name[1] == 'e' &&
10038 return -KEY_rewinddir;
10044 if (name[1] == 'e' &&
10053 return -KEY_setnetent;
10059 if (name[1] == 'a' &&
10068 return -KEY_wantarray;
10077 case 10: /* 9 tokens of length 10 */
10081 if (name[1] == 'n' &&
10087 if (name[4] == 'o' &&
10094 return -KEY_endhostent;
10100 if (name[4] == 'e' &&
10107 return -KEY_endservent;
10120 if (name[1] == 'e' &&
10126 if (name[4] == 'o' &&
10133 return -KEY_gethostent;
10142 if (name[5] == 'r' &&
10148 return -KEY_getservent;
10154 if (name[5] == 'c' &&
10160 return -KEY_getsockopt;
10180 if (name[2] == 't')
10185 if (name[4] == 'o' &&
10192 return -KEY_sethostent;
10201 if (name[5] == 'r' &&
10207 return -KEY_setservent;
10213 if (name[5] == 'c' &&
10219 return -KEY_setsockopt;
10236 if (name[2] == 'c' &&
10245 return -KEY_socketpair;
10258 case 11: /* 8 tokens of length 11 */
10262 if (name[1] == '_' &&
10272 { /* __PACKAGE__ */
10273 return -KEY___PACKAGE__;
10279 if (name[1] == 'n' &&
10289 { /* endprotoent */
10290 return -KEY_endprotoent;
10296 if (name[1] == 'e' &&
10305 if (name[5] == 'e' &&
10311 { /* getpeername */
10312 return -KEY_getpeername;
10321 if (name[6] == 'o' &&
10326 { /* getpriority */
10327 return -KEY_getpriority;
10333 if (name[6] == 't' &&
10338 { /* getprotoent */
10339 return -KEY_getprotoent;
10353 if (name[4] == 'o' &&
10360 { /* getsockname */
10361 return -KEY_getsockname;
10374 if (name[1] == 'e' &&
10382 if (name[6] == 'o' &&
10387 { /* setpriority */
10388 return -KEY_setpriority;
10394 if (name[6] == 't' &&
10399 { /* setprotoent */
10400 return -KEY_setprotoent;
10416 case 12: /* 2 tokens of length 12 */
10417 if (name[0] == 'g' &&
10429 if (name[9] == 'd' &&
10432 { /* getnetbyaddr */
10433 return -KEY_getnetbyaddr;
10439 if (name[9] == 'a' &&
10442 { /* getnetbyname */
10443 return -KEY_getnetbyname;
10455 case 13: /* 4 tokens of length 13 */
10456 if (name[0] == 'g' &&
10463 if (name[4] == 'o' &&
10472 if (name[10] == 'd' &&
10475 { /* gethostbyaddr */
10476 return -KEY_gethostbyaddr;
10482 if (name[10] == 'a' &&
10485 { /* gethostbyname */
10486 return -KEY_gethostbyname;
10499 if (name[4] == 'e' &&
10508 if (name[10] == 'a' &&
10511 { /* getservbyname */
10512 return -KEY_getservbyname;
10518 if (name[10] == 'o' &&
10521 { /* getservbyport */
10522 return -KEY_getservbyport;
10541 case 14: /* 1 tokens of length 14 */
10542 if (name[0] == 'g' &&
10556 { /* getprotobyname */
10557 return -KEY_getprotobyname;
10562 case 16: /* 1 tokens of length 16 */
10563 if (name[0] == 'g' &&
10579 { /* getprotobynumber */
10580 return -KEY_getprotobynumber;
10594 S_checkcomma(pTHX_ const char *s, const char *name, const char *what)
10598 PERL_ARGS_ASSERT_CHECKCOMMA;
10600 if (*s == ' ' && s[1] == '(') { /* XXX gotta be a better way */
10601 if (ckWARN(WARN_SYNTAX)) {
10604 for (w = s+2; *w && level; w++) {
10607 else if (*w == ')')
10610 while (isSPACE(*w))
10612 /* the list of chars below is for end of statements or
10613 * block / parens, boolean operators (&&, ||, //) and branch
10614 * constructs (or, and, if, until, unless, while, err, for).
10615 * Not a very solid hack... */
10616 if (!*w || !strchr(";&/|})]oaiuwef!=", *w))
10617 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
10618 "%s (...) interpreted as function",name);
10621 while (s < PL_bufend && isSPACE(*s))
10625 while (s < PL_bufend && isSPACE(*s))
10627 if (isIDFIRST_lazy_if(s,UTF)) {
10628 const char * const w = s++;
10629 while (isALNUM_lazy_if(s,UTF))
10631 while (s < PL_bufend && isSPACE(*s))
10635 if (keyword(w, s - w, 0))
10638 gv = gv_fetchpvn_flags(w, s - w, 0, SVt_PVCV);
10639 if (gv && GvCVu(gv))
10641 Perl_croak(aTHX_ "No comma allowed after %s", what);
10646 /* Either returns sv, or mortalizes sv and returns a new SV*.
10647 Best used as sv=new_constant(..., sv, ...).
10648 If s, pv are NULL, calls subroutine with one argument,
10649 and type is used with error messages only. */
10652 S_new_constant(pTHX_ const char *s, STRLEN len, const char *key, STRLEN keylen,
10653 SV *sv, SV *pv, const char *type, STRLEN typelen)
10656 HV * const table = GvHV(PL_hintgv); /* ^H */
10660 const char *why1 = "", *why2 = "", *why3 = "";
10662 PERL_ARGS_ASSERT_NEW_CONSTANT;
10664 if (!table || !(PL_hints & HINT_LOCALIZE_HH)) {
10667 why2 = (const char *)
10668 (strEQ(key,"charnames")
10669 ? "(possibly a missing \"use charnames ...\")"
10671 msg = Perl_newSVpvf(aTHX_ "Constant(%s) unknown: %s",
10672 (type ? type: "undef"), why2);
10674 /* This is convoluted and evil ("goto considered harmful")
10675 * but I do not understand the intricacies of all the different
10676 * failure modes of %^H in here. The goal here is to make
10677 * the most probable error message user-friendly. --jhi */
10682 msg = Perl_newSVpvf(aTHX_ "Constant(%s): %s%s%s",
10683 (type ? type: "undef"), why1, why2, why3);
10685 yyerror(SvPVX_const(msg));
10689 cvp = hv_fetch(table, key, keylen, FALSE);
10690 if (!cvp || !SvOK(*cvp)) {
10693 why3 = "} is not defined";
10696 sv_2mortal(sv); /* Parent created it permanently */
10699 pv = newSVpvn_flags(s, len, SVs_TEMP);
10701 typesv = newSVpvn_flags(type, typelen, SVs_TEMP);
10703 typesv = &PL_sv_undef;
10705 PUSHSTACKi(PERLSI_OVERLOAD);
10717 call_sv(cv, G_SCALAR | ( PL_in_eval ? 0 : G_EVAL));
10721 /* Check the eval first */
10722 if (!PL_in_eval && SvTRUE(ERRSV)) {
10723 sv_catpvs(ERRSV, "Propagated");
10724 yyerror(SvPV_nolen_const(ERRSV)); /* Duplicates the message inside eval */
10726 res = SvREFCNT_inc_simple(sv);
10730 SvREFCNT_inc_simple_void(res);
10739 why1 = "Call to &{$^H{";
10741 why3 = "}} did not return a defined value";
10749 /* Returns a NUL terminated string, with the length of the string written to
10753 S_scan_word(pTHX_ register char *s, char *dest, STRLEN destlen, int allow_package, STRLEN *slp)
10756 register char *d = dest;
10757 register char * const e = d + destlen - 3; /* two-character token, ending NUL */
10759 PERL_ARGS_ASSERT_SCAN_WORD;
10763 Perl_croak(aTHX_ ident_too_long);
10764 if (isALNUM(*s)) /* UTF handled below */
10766 else if (allow_package && (*s == '\'') && isIDFIRST_lazy_if(s+1,UTF)) {
10771 else if (allow_package && (s[0] == ':') && (s[1] == ':') && (s[2] != '$')) {
10775 else if (UTF && UTF8_IS_START(*s) && isALNUM_utf8((U8*)s)) {
10776 char *t = s + UTF8SKIP(s);
10778 while (UTF8_IS_CONTINUED(*t) && is_utf8_mark((U8*)t))
10782 Perl_croak(aTHX_ ident_too_long);
10783 Copy(s, d, len, char);
10796 S_scan_ident(pTHX_ register char *s, register const char *send, char *dest, STRLEN destlen, I32 ck_uni)
10799 char *bracket = NULL;
10801 register char *d = dest;
10802 register char * const e = d + destlen + 3; /* two-character token, ending NUL */
10804 PERL_ARGS_ASSERT_SCAN_IDENT;
10809 while (isDIGIT(*s)) {
10811 Perl_croak(aTHX_ ident_too_long);
10818 Perl_croak(aTHX_ ident_too_long);
10819 if (isALNUM(*s)) /* UTF handled below */
10821 else if (*s == '\'' && isIDFIRST_lazy_if(s+1,UTF)) {
10826 else if (*s == ':' && s[1] == ':') {
10830 else if (UTF && UTF8_IS_START(*s) && isALNUM_utf8((U8*)s)) {
10831 char *t = s + UTF8SKIP(s);
10832 while (UTF8_IS_CONTINUED(*t) && is_utf8_mark((U8*)t))
10834 if (d + (t - s) > e)
10835 Perl_croak(aTHX_ ident_too_long);
10836 Copy(s, d, t - s, char);
10847 if (PL_lex_state != LEX_NORMAL)
10848 PL_lex_state = LEX_INTERPENDMAYBE;
10851 if (*s == '$' && s[1] &&
10852 (isALNUM_lazy_if(s+1,UTF) || s[1] == '$' || s[1] == '{' || strnEQ(s+1,"::",2)) )
10865 if (*d == '^' && *s && isCONTROLVAR(*s)) {
10870 if (isSPACE(s[-1])) {
10872 const char ch = *s++;
10873 if (!SPACE_OR_TAB(ch)) {
10879 if (isIDFIRST_lazy_if(d,UTF)) {
10883 while ((end < send && isALNUM_lazy_if(end,UTF)) || *end == ':') {
10884 end += UTF8SKIP(end);
10885 while (end < send && UTF8_IS_CONTINUED(*end) && is_utf8_mark((U8*)end))
10886 end += UTF8SKIP(end);
10888 Copy(s, d, end - s, char);
10893 while ((isALNUM(*s) || *s == ':') && d < e)
10896 Perl_croak(aTHX_ ident_too_long);
10899 while (s < send && SPACE_OR_TAB(*s))
10901 if ((*s == '[' || (*s == '{' && strNE(dest, "sub")))) {
10902 if (ckWARN(WARN_AMBIGUOUS) && keyword(dest, d - dest, 0)) {
10903 const char * const brack =
10905 ((*s == '[') ? "[...]" : "{...}");
10906 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
10907 "Ambiguous use of %c{%s%s} resolved to %c%s%s",
10908 funny, dest, brack, funny, dest, brack);
10911 PL_lex_brackstack[PL_lex_brackets++] = (char)(XOPERATOR | XFAKEBRACK);
10915 /* Handle extended ${^Foo} variables
10916 * 1999-02-27 mjd-perl-patch@plover.com */
10917 else if (!isALNUM(*d) && !isPRINT(*d) /* isCTRL(d) */
10921 while (isALNUM(*s) && d < e) {
10925 Perl_croak(aTHX_ ident_too_long);
10930 if (PL_lex_state == LEX_INTERPNORMAL && !PL_lex_brackets) {
10931 PL_lex_state = LEX_INTERPEND;
10934 if (PL_lex_state == LEX_NORMAL) {
10935 if (ckWARN(WARN_AMBIGUOUS) &&
10936 (keyword(dest, d - dest, 0)
10937 || get_cvn_flags(dest, d - dest, 0)))
10941 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
10942 "Ambiguous use of %c{%s} resolved to %c%s",
10943 funny, dest, funny, dest);
10948 s = bracket; /* let the parser handle it */
10952 else if (PL_lex_state == LEX_INTERPNORMAL && !PL_lex_brackets && !intuit_more(s))
10953 PL_lex_state = LEX_INTERPEND;
10958 Perl_pmflag(pTHX_ U32* pmfl, int ch)
10960 PERL_ARGS_ASSERT_PMFLAG;
10962 PERL_UNUSED_CONTEXT;
10964 const char c = (char)ch;
10966 CASE_STD_PMMOD_FLAGS_PARSE_SET(pmfl);
10967 case GLOBAL_PAT_MOD: *pmfl |= PMf_GLOBAL; break;
10968 case CONTINUE_PAT_MOD: *pmfl |= PMf_CONTINUE; break;
10969 case ONCE_PAT_MOD: *pmfl |= PMf_KEEP; break;
10970 case KEEPCOPY_PAT_MOD: *pmfl |= PMf_KEEPCOPY; break;
10976 S_scan_pat(pTHX_ char *start, I32 type)
10980 char *s = scan_str(start,!!PL_madskills,FALSE);
10981 const char * const valid_flags =
10982 (const char *)((type == OP_QR) ? QR_PAT_MODS : M_PAT_MODS);
10987 PERL_ARGS_ASSERT_SCAN_PAT;
10990 const char * const delimiter = skipspace(start);
10994 ? "Search pattern not terminated or ternary operator parsed as search pattern"
10995 : "Search pattern not terminated" ));
10998 pm = (PMOP*)newPMOP(type, 0);
10999 if (PL_multi_open == '?') {
11000 /* This is the only point in the code that sets PMf_ONCE: */
11001 pm->op_pmflags |= PMf_ONCE;
11003 /* Hence it's safe to do this bit of PMOP book-keeping here, which
11004 allows us to restrict the list needed by reset to just the ??
11006 assert(type != OP_TRANS);
11008 MAGIC *mg = mg_find((const SV *)PL_curstash, PERL_MAGIC_symtab);
11011 mg = sv_magicext(MUTABLE_SV(PL_curstash), 0, PERL_MAGIC_symtab, 0, 0,
11014 elements = mg->mg_len / sizeof(PMOP**);
11015 Renewc(mg->mg_ptr, elements + 1, PMOP*, char);
11016 ((PMOP**)mg->mg_ptr) [elements++] = pm;
11017 mg->mg_len = elements * sizeof(PMOP**);
11018 PmopSTASH_set(pm,PL_curstash);
11024 while (*s && strchr(valid_flags, *s))
11025 pmflag(&pm->op_pmflags,*s++);
11027 if (PL_madskills && modstart != s) {
11028 SV* tmptoken = newSVpvn(modstart, s - modstart);
11029 append_madprops(newMADPROP('m', MAD_SV, tmptoken, 0), (OP*)pm, 0);
11032 /* issue a warning if /c is specified,but /g is not */
11033 if ((pm->op_pmflags & PMf_CONTINUE) && !(pm->op_pmflags & PMf_GLOBAL)
11034 && ckWARN(WARN_REGEXP))
11036 Perl_warner(aTHX_ packWARN(WARN_REGEXP),
11037 "Use of /c modifier is meaningless without /g" );
11040 PL_lex_op = (OP*)pm;
11041 pl_yylval.ival = OP_MATCH;
11046 S_scan_subst(pTHX_ char *start)
11057 PERL_ARGS_ASSERT_SCAN_SUBST;
11059 pl_yylval.ival = OP_NULL;
11061 s = scan_str(start,!!PL_madskills,FALSE);
11064 Perl_croak(aTHX_ "Substitution pattern not terminated");
11066 if (s[-1] == PL_multi_open)
11069 if (PL_madskills) {
11070 CURMAD('q', PL_thisopen);
11071 CURMAD('_', PL_thiswhite);
11072 CURMAD('E', PL_thisstuff);
11073 CURMAD('Q', PL_thisclose);
11074 PL_realtokenstart = s - SvPVX(PL_linestr);
11078 first_start = PL_multi_start;
11079 s = scan_str(s,!!PL_madskills,FALSE);
11081 if (PL_lex_stuff) {
11082 SvREFCNT_dec(PL_lex_stuff);
11083 PL_lex_stuff = NULL;
11085 Perl_croak(aTHX_ "Substitution replacement not terminated");
11087 PL_multi_start = first_start; /* so whole substitution is taken together */
11089 pm = (PMOP*)newPMOP(OP_SUBST, 0);
11092 if (PL_madskills) {
11093 CURMAD('z', PL_thisopen);
11094 CURMAD('R', PL_thisstuff);
11095 CURMAD('Z', PL_thisclose);
11101 if (*s == EXEC_PAT_MOD) {
11105 else if (strchr(S_PAT_MODS, *s))
11106 pmflag(&pm->op_pmflags,*s++);
11112 if (PL_madskills) {
11114 curmad('m', newSVpvn(modstart, s - modstart));
11115 append_madprops(PL_thismad, (OP*)pm, 0);
11119 if ((pm->op_pmflags & PMf_CONTINUE) && ckWARN(WARN_REGEXP)) {
11120 Perl_warner(aTHX_ packWARN(WARN_REGEXP), "Use of /c modifier is meaningless in s///" );
11124 SV * const repl = newSVpvs("");
11126 PL_sublex_info.super_bufptr = s;
11127 PL_sublex_info.super_bufend = PL_bufend;
11129 pm->op_pmflags |= PMf_EVAL;
11132 sv_catpvs(repl, "eval ");
11134 sv_catpvs(repl, "do ");
11136 sv_catpvs(repl, "{");
11137 sv_catsv(repl, PL_lex_repl);
11138 if (strchr(SvPVX(PL_lex_repl), '#'))
11139 sv_catpvs(repl, "\n");
11140 sv_catpvs(repl, "}");
11142 SvREFCNT_dec(PL_lex_repl);
11143 PL_lex_repl = repl;
11146 PL_lex_op = (OP*)pm;
11147 pl_yylval.ival = OP_SUBST;
11152 S_scan_trans(pTHX_ char *start)
11165 PERL_ARGS_ASSERT_SCAN_TRANS;
11167 pl_yylval.ival = OP_NULL;
11169 s = scan_str(start,!!PL_madskills,FALSE);
11171 Perl_croak(aTHX_ "Transliteration pattern not terminated");
11173 if (s[-1] == PL_multi_open)
11176 if (PL_madskills) {
11177 CURMAD('q', PL_thisopen);
11178 CURMAD('_', PL_thiswhite);
11179 CURMAD('E', PL_thisstuff);
11180 CURMAD('Q', PL_thisclose);
11181 PL_realtokenstart = s - SvPVX(PL_linestr);
11185 s = scan_str(s,!!PL_madskills,FALSE);
11187 if (PL_lex_stuff) {
11188 SvREFCNT_dec(PL_lex_stuff);
11189 PL_lex_stuff = NULL;
11191 Perl_croak(aTHX_ "Transliteration replacement not terminated");
11193 if (PL_madskills) {
11194 CURMAD('z', PL_thisopen);
11195 CURMAD('R', PL_thisstuff);
11196 CURMAD('Z', PL_thisclose);
11199 complement = del = squash = 0;
11206 complement = OPpTRANS_COMPLEMENT;
11209 del = OPpTRANS_DELETE;
11212 squash = OPpTRANS_SQUASH;
11221 tbl = (short *)PerlMemShared_calloc(complement&&!del?258:256, sizeof(short));
11222 o = newPVOP(OP_TRANS, 0, (char*)tbl);
11223 o->op_private &= ~OPpTRANS_ALL;
11224 o->op_private |= del|squash|complement|
11225 (DO_UTF8(PL_lex_stuff)? OPpTRANS_FROM_UTF : 0)|
11226 (DO_UTF8(PL_lex_repl) ? OPpTRANS_TO_UTF : 0);
11229 pl_yylval.ival = OP_TRANS;
11232 if (PL_madskills) {
11234 curmad('m', newSVpvn(modstart, s - modstart));
11235 append_madprops(PL_thismad, o, 0);
11244 S_scan_heredoc(pTHX_ register char *s)
11248 I32 op_type = OP_SCALAR;
11252 const char *found_newline;
11256 const int outer = (PL_rsfp && !(PL_lex_inwhat == OP_SCALAR));
11258 I32 stuffstart = s - SvPVX(PL_linestr);
11261 PL_realtokenstart = -1;
11264 PERL_ARGS_ASSERT_SCAN_HEREDOC;
11268 e = PL_tokenbuf + sizeof PL_tokenbuf - 1;
11272 while (SPACE_OR_TAB(*peek))
11274 if (*peek == '`' || *peek == '\'' || *peek =='"') {
11277 s = delimcpy(d, e, s, PL_bufend, term, &len);
11287 if (!isALNUM_lazy_if(s,UTF))
11288 deprecate_old("bare << to mean <<\"\"");
11289 for (; isALNUM_lazy_if(s,UTF); s++) {
11294 if (d >= PL_tokenbuf + sizeof PL_tokenbuf - 1)
11295 Perl_croak(aTHX_ "Delimiter for here document is too long");
11298 len = d - PL_tokenbuf;
11301 if (PL_madskills) {
11302 tstart = PL_tokenbuf + !outer;
11303 PL_thisclose = newSVpvn(tstart, len - !outer);
11304 tstart = SvPVX(PL_linestr) + stuffstart;
11305 PL_thisopen = newSVpvn(tstart, s - tstart);
11306 stuffstart = s - SvPVX(PL_linestr);
11309 #ifndef PERL_STRICT_CR
11310 d = strchr(s, '\r');
11312 char * const olds = s;
11314 while (s < PL_bufend) {
11320 else if (*s == '\n' && s[1] == '\r') { /* \015\013 on a mac? */
11329 SvCUR_set(PL_linestr, PL_bufend - SvPVX_const(PL_linestr));
11336 if ( outer || !(found_newline = (char*)memchr((void*)s, '\n', PL_bufend - s)) ) {
11337 herewas = newSVpvn(s,PL_bufend-s);
11341 herewas = newSVpvn(s-1,found_newline-s+1);
11344 herewas = newSVpvn(s,found_newline-s);
11348 if (PL_madskills) {
11349 tstart = SvPVX(PL_linestr) + stuffstart;
11351 sv_catpvn(PL_thisstuff, tstart, s - tstart);
11353 PL_thisstuff = newSVpvn(tstart, s - tstart);
11356 s += SvCUR(herewas);
11359 stuffstart = s - SvPVX(PL_linestr);
11365 tmpstr = newSV_type(SVt_PVIV);
11366 SvGROW(tmpstr, 80);
11367 if (term == '\'') {
11368 op_type = OP_CONST;
11369 SvIV_set(tmpstr, -1);
11371 else if (term == '`') {
11372 op_type = OP_BACKTICK;
11373 SvIV_set(tmpstr, '\\');
11377 PL_multi_start = CopLINE(PL_curcop);
11378 PL_multi_open = PL_multi_close = '<';
11379 term = *PL_tokenbuf;
11380 if (PL_lex_inwhat == OP_SUBST && PL_in_eval && !PL_rsfp) {
11381 char * const bufptr = PL_sublex_info.super_bufptr;
11382 char * const bufend = PL_sublex_info.super_bufend;
11383 char * const olds = s - SvCUR(herewas);
11384 s = strchr(bufptr, '\n');
11388 while (s < bufend &&
11389 (*s != term || memNE(s,PL_tokenbuf,len)) ) {
11391 CopLINE_inc(PL_curcop);
11394 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
11395 missingterm(PL_tokenbuf);
11397 sv_setpvn(herewas,bufptr,d-bufptr+1);
11398 sv_setpvn(tmpstr,d+1,s-d);
11400 sv_catpvn(herewas,s,bufend-s);
11401 Copy(SvPVX_const(herewas),bufptr,SvCUR(herewas) + 1,char);
11408 while (s < PL_bufend &&
11409 (*s != term || memNE(s,PL_tokenbuf,len)) ) {
11411 CopLINE_inc(PL_curcop);
11413 if (s >= PL_bufend) {
11414 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
11415 missingterm(PL_tokenbuf);
11417 sv_setpvn(tmpstr,d+1,s-d);
11419 if (PL_madskills) {
11421 sv_catpvn(PL_thisstuff, d + 1, s - d);
11423 PL_thisstuff = newSVpvn(d + 1, s - d);
11424 stuffstart = s - SvPVX(PL_linestr);
11428 CopLINE_inc(PL_curcop); /* the preceding stmt passes a newline */
11430 sv_catpvn(herewas,s,PL_bufend-s);
11431 sv_setsv(PL_linestr,herewas);
11432 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = s = PL_linestart = SvPVX(PL_linestr);
11433 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
11434 PL_last_lop = PL_last_uni = NULL;
11437 sv_setpvs(tmpstr,""); /* avoid "uninitialized" warning */
11438 while (s >= PL_bufend) { /* multiple line string? */
11440 if (PL_madskills) {
11441 tstart = SvPVX(PL_linestr) + stuffstart;
11443 sv_catpvn(PL_thisstuff, tstart, PL_bufend - tstart);
11445 PL_thisstuff = newSVpvn(tstart, PL_bufend - tstart);
11449 !(PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = filter_gets(PL_linestr, PL_rsfp, 0))) {
11450 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
11451 missingterm(PL_tokenbuf);
11454 stuffstart = s - SvPVX(PL_linestr);
11456 CopLINE_inc(PL_curcop);
11457 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
11458 PL_last_lop = PL_last_uni = NULL;
11459 #ifndef PERL_STRICT_CR
11460 if (PL_bufend - PL_linestart >= 2) {
11461 if ((PL_bufend[-2] == '\r' && PL_bufend[-1] == '\n') ||
11462 (PL_bufend[-2] == '\n' && PL_bufend[-1] == '\r'))
11464 PL_bufend[-2] = '\n';
11466 SvCUR_set(PL_linestr, PL_bufend - SvPVX_const(PL_linestr));
11468 else if (PL_bufend[-1] == '\r')
11469 PL_bufend[-1] = '\n';
11471 else if (PL_bufend - PL_linestart == 1 && PL_bufend[-1] == '\r')
11472 PL_bufend[-1] = '\n';
11474 if ((PERLDB_LINE || PERLDB_SAVESRC) && PL_curstash != PL_debstash)
11475 update_debugger_info(PL_linestr, NULL, 0);
11476 if (*s == term && memEQ(s,PL_tokenbuf,len)) {
11477 STRLEN off = PL_bufend - 1 - SvPVX_const(PL_linestr);
11478 *(SvPVX(PL_linestr) + off ) = ' ';
11479 sv_catsv(PL_linestr,herewas);
11480 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
11481 s = SvPVX(PL_linestr) + off; /* In case PV of PL_linestr moved. */
11485 sv_catsv(tmpstr,PL_linestr);
11490 PL_multi_end = CopLINE(PL_curcop);
11491 if (SvCUR(tmpstr) + 5 < SvLEN(tmpstr)) {
11492 SvPV_shrink_to_cur(tmpstr);
11494 SvREFCNT_dec(herewas);
11496 if (UTF && is_utf8_string((U8*)SvPVX_const(tmpstr), SvCUR(tmpstr)))
11498 else if (PL_encoding)
11499 sv_recode_to_utf8(tmpstr, PL_encoding);
11501 PL_lex_stuff = tmpstr;
11502 pl_yylval.ival = op_type;
11506 /* scan_inputsymbol
11507 takes: current position in input buffer
11508 returns: new position in input buffer
11509 side-effects: pl_yylval and lex_op are set.
11514 <FH> read from filehandle
11515 <pkg::FH> read from package qualified filehandle
11516 <pkg'FH> read from package qualified filehandle
11517 <$fh> read from filehandle in $fh
11518 <*.h> filename glob
11523 S_scan_inputsymbol(pTHX_ char *start)
11526 register char *s = start; /* current position in buffer */
11529 char *d = PL_tokenbuf; /* start of temp holding space */
11530 const char * const e = PL_tokenbuf + sizeof PL_tokenbuf; /* end of temp holding space */
11532 PERL_ARGS_ASSERT_SCAN_INPUTSYMBOL;
11534 end = strchr(s, '\n');
11537 s = delimcpy(d, e, s + 1, end, '>', &len); /* extract until > */
11539 /* die if we didn't have space for the contents of the <>,
11540 or if it didn't end, or if we see a newline
11543 if (len >= (I32)sizeof PL_tokenbuf)
11544 Perl_croak(aTHX_ "Excessively long <> operator");
11546 Perl_croak(aTHX_ "Unterminated <> operator");
11551 Remember, only scalar variables are interpreted as filehandles by
11552 this code. Anything more complex (e.g., <$fh{$num}>) will be
11553 treated as a glob() call.
11554 This code makes use of the fact that except for the $ at the front,
11555 a scalar variable and a filehandle look the same.
11557 if (*d == '$' && d[1]) d++;
11559 /* allow <Pkg'VALUE> or <Pkg::VALUE> */
11560 while (*d && (isALNUM_lazy_if(d,UTF) || *d == '\'' || *d == ':'))
11563 /* If we've tried to read what we allow filehandles to look like, and
11564 there's still text left, then it must be a glob() and not a getline.
11565 Use scan_str to pull out the stuff between the <> and treat it
11566 as nothing more than a string.
11569 if (d - PL_tokenbuf != len) {
11570 pl_yylval.ival = OP_GLOB;
11571 s = scan_str(start,!!PL_madskills,FALSE);
11573 Perl_croak(aTHX_ "Glob not terminated");
11577 bool readline_overriden = FALSE;
11580 /* we're in a filehandle read situation */
11583 /* turn <> into <ARGV> */
11585 Copy("ARGV",d,5,char);
11587 /* Check whether readline() is overriden */
11588 gv_readline = gv_fetchpvs("readline", GV_NOTQUAL, SVt_PVCV);
11590 && GvCVu(gv_readline) && GvIMPORTED_CV(gv_readline))
11592 ((gvp = (GV**)hv_fetchs(PL_globalstash, "readline", FALSE))
11593 && (gv_readline = *gvp) && isGV_with_GP(gv_readline)
11594 && GvCVu(gv_readline) && GvIMPORTED_CV(gv_readline)))
11595 readline_overriden = TRUE;
11597 /* if <$fh>, create the ops to turn the variable into a
11601 /* try to find it in the pad for this block, otherwise find
11602 add symbol table ops
11604 const PADOFFSET tmp = pad_findmy(d);
11605 if (tmp != NOT_IN_PAD) {
11606 if (PAD_COMPNAME_FLAGS_isOUR(tmp)) {
11607 HV * const stash = PAD_COMPNAME_OURSTASH(tmp);
11608 HEK * const stashname = HvNAME_HEK(stash);
11609 SV * const sym = sv_2mortal(newSVhek(stashname));
11610 sv_catpvs(sym, "::");
11611 sv_catpv(sym, d+1);
11616 OP * const o = newOP(OP_PADSV, 0);
11618 PL_lex_op = readline_overriden
11619 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
11620 append_elem(OP_LIST, o,
11621 newCVREF(0, newGVOP(OP_GV,0,gv_readline))))
11622 : (OP*)newUNOP(OP_READLINE, 0, o);
11631 ? (GV_ADDMULTI | GV_ADDINEVAL)
11634 PL_lex_op = readline_overriden
11635 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
11636 append_elem(OP_LIST,
11637 newUNOP(OP_RV2SV, 0, newGVOP(OP_GV, 0, gv)),
11638 newCVREF(0, newGVOP(OP_GV, 0, gv_readline))))
11639 : (OP*)newUNOP(OP_READLINE, 0,
11640 newUNOP(OP_RV2SV, 0,
11641 newGVOP(OP_GV, 0, gv)));
11643 if (!readline_overriden)
11644 PL_lex_op->op_flags |= OPf_SPECIAL;
11645 /* we created the ops in PL_lex_op, so make pl_yylval.ival a null op */
11646 pl_yylval.ival = OP_NULL;
11649 /* If it's none of the above, it must be a literal filehandle
11650 (<Foo::BAR> or <FOO>) so build a simple readline OP */
11652 GV * const gv = gv_fetchpv(d, GV_ADD, SVt_PVIO);
11653 PL_lex_op = readline_overriden
11654 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
11655 append_elem(OP_LIST,
11656 newGVOP(OP_GV, 0, gv),
11657 newCVREF(0, newGVOP(OP_GV, 0, gv_readline))))
11658 : (OP*)newUNOP(OP_READLINE, 0, newGVOP(OP_GV, 0, gv));
11659 pl_yylval.ival = OP_NULL;
11668 takes: start position in buffer
11669 keep_quoted preserve \ on the embedded delimiter(s)
11670 keep_delims preserve the delimiters around the string
11671 returns: position to continue reading from buffer
11672 side-effects: multi_start, multi_close, lex_repl or lex_stuff, and
11673 updates the read buffer.
11675 This subroutine pulls a string out of the input. It is called for:
11676 q single quotes q(literal text)
11677 ' single quotes 'literal text'
11678 qq double quotes qq(interpolate $here please)
11679 " double quotes "interpolate $here please"
11680 qx backticks qx(/bin/ls -l)
11681 ` backticks `/bin/ls -l`
11682 qw quote words @EXPORT_OK = qw( func() $spam )
11683 m// regexp match m/this/
11684 s/// regexp substitute s/this/that/
11685 tr/// string transliterate tr/this/that/
11686 y/// string transliterate y/this/that/
11687 ($*@) sub prototypes sub foo ($)
11688 (stuff) sub attr parameters sub foo : attr(stuff)
11689 <> readline or globs <FOO>, <>, <$fh>, or <*.c>
11691 In most of these cases (all but <>, patterns and transliterate)
11692 yylex() calls scan_str(). m// makes yylex() call scan_pat() which
11693 calls scan_str(). s/// makes yylex() call scan_subst() which calls
11694 scan_str(). tr/// and y/// make yylex() call scan_trans() which
11697 It skips whitespace before the string starts, and treats the first
11698 character as the delimiter. If the delimiter is one of ([{< then
11699 the corresponding "close" character )]}> is used as the closing
11700 delimiter. It allows quoting of delimiters, and if the string has
11701 balanced delimiters ([{<>}]) it allows nesting.
11703 On success, the SV with the resulting string is put into lex_stuff or,
11704 if that is already non-NULL, into lex_repl. The second case occurs only
11705 when parsing the RHS of the special constructs s/// and tr/// (y///).
11706 For convenience, the terminating delimiter character is stuffed into
11711 S_scan_str(pTHX_ char *start, int keep_quoted, int keep_delims)
11714 SV *sv; /* scalar value: string */
11715 const char *tmps; /* temp string, used for delimiter matching */
11716 register char *s = start; /* current position in the buffer */
11717 register char term; /* terminating character */
11718 register char *to; /* current position in the sv's data */
11719 I32 brackets = 1; /* bracket nesting level */
11720 bool has_utf8 = FALSE; /* is there any utf8 content? */
11721 I32 termcode; /* terminating char. code */
11722 U8 termstr[UTF8_MAXBYTES]; /* terminating string */
11723 STRLEN termlen; /* length of terminating string */
11724 int last_off = 0; /* last position for nesting bracket */
11730 PERL_ARGS_ASSERT_SCAN_STR;
11732 /* skip space before the delimiter */
11738 if (PL_realtokenstart >= 0) {
11739 stuffstart = PL_realtokenstart;
11740 PL_realtokenstart = -1;
11743 stuffstart = start - SvPVX(PL_linestr);
11745 /* mark where we are, in case we need to report errors */
11748 /* after skipping whitespace, the next character is the terminator */
11751 termcode = termstr[0] = term;
11755 termcode = utf8_to_uvchr((U8*)s, &termlen);
11756 Copy(s, termstr, termlen, U8);
11757 if (!UTF8_IS_INVARIANT(term))
11761 /* mark where we are */
11762 PL_multi_start = CopLINE(PL_curcop);
11763 PL_multi_open = term;
11765 /* find corresponding closing delimiter */
11766 if (term && (tmps = strchr("([{< )]}> )]}>",term)))
11767 termcode = termstr[0] = term = tmps[5];
11769 PL_multi_close = term;
11771 /* create a new SV to hold the contents. 79 is the SV's initial length.
11772 What a random number. */
11773 sv = newSV_type(SVt_PVIV);
11775 SvIV_set(sv, termcode);
11776 (void)SvPOK_only(sv); /* validate pointer */
11778 /* move past delimiter and try to read a complete string */
11780 sv_catpvn(sv, s, termlen);
11783 tstart = SvPVX(PL_linestr) + stuffstart;
11784 if (!PL_thisopen && !keep_delims) {
11785 PL_thisopen = newSVpvn(tstart, s - tstart);
11786 stuffstart = s - SvPVX(PL_linestr);
11790 if (PL_encoding && !UTF) {
11794 int offset = s - SvPVX_const(PL_linestr);
11795 const bool found = sv_cat_decode(sv, PL_encoding, PL_linestr,
11796 &offset, (char*)termstr, termlen);
11797 const char * const ns = SvPVX_const(PL_linestr) + offset;
11798 char * const svlast = SvEND(sv) - 1;
11800 for (; s < ns; s++) {
11801 if (*s == '\n' && !PL_rsfp)
11802 CopLINE_inc(PL_curcop);
11805 goto read_more_line;
11807 /* handle quoted delimiters */
11808 if (SvCUR(sv) > 1 && *(svlast-1) == '\\') {
11810 for (t = svlast-2; t >= SvPVX_const(sv) && *t == '\\';)
11812 if ((svlast-1 - t) % 2) {
11813 if (!keep_quoted) {
11814 *(svlast-1) = term;
11816 SvCUR_set(sv, SvCUR(sv) - 1);
11821 if (PL_multi_open == PL_multi_close) {
11827 for (t = w = SvPVX(sv)+last_off; t < svlast; w++, t++) {
11828 /* At here, all closes are "was quoted" one,
11829 so we don't check PL_multi_close. */
11831 if (!keep_quoted && *(t+1) == PL_multi_open)
11836 else if (*t == PL_multi_open)
11844 SvCUR_set(sv, w - SvPVX_const(sv));
11846 last_off = w - SvPVX(sv);
11847 if (--brackets <= 0)
11852 if (!keep_delims) {
11853 SvCUR_set(sv, SvCUR(sv) - 1);
11859 /* extend sv if need be */
11860 SvGROW(sv, SvCUR(sv) + (PL_bufend - s) + 1);
11861 /* set 'to' to the next character in the sv's string */
11862 to = SvPVX(sv)+SvCUR(sv);
11864 /* if open delimiter is the close delimiter read unbridle */
11865 if (PL_multi_open == PL_multi_close) {
11866 for (; s < PL_bufend; s++,to++) {
11867 /* embedded newlines increment the current line number */
11868 if (*s == '\n' && !PL_rsfp)
11869 CopLINE_inc(PL_curcop);
11870 /* handle quoted delimiters */
11871 if (*s == '\\' && s+1 < PL_bufend && term != '\\') {
11872 if (!keep_quoted && s[1] == term)
11874 /* any other quotes are simply copied straight through */
11878 /* terminate when run out of buffer (the for() condition), or
11879 have found the terminator */
11880 else if (*s == term) {
11883 if (s+termlen <= PL_bufend && memEQ(s, (char*)termstr, termlen))
11886 else if (!has_utf8 && !UTF8_IS_INVARIANT((U8)*s) && UTF)
11892 /* if the terminator isn't the same as the start character (e.g.,
11893 matched brackets), we have to allow more in the quoting, and
11894 be prepared for nested brackets.
11897 /* read until we run out of string, or we find the terminator */
11898 for (; s < PL_bufend; s++,to++) {
11899 /* embedded newlines increment the line count */
11900 if (*s == '\n' && !PL_rsfp)
11901 CopLINE_inc(PL_curcop);
11902 /* backslashes can escape the open or closing characters */
11903 if (*s == '\\' && s+1 < PL_bufend) {
11904 if (!keep_quoted &&
11905 ((s[1] == PL_multi_open) || (s[1] == PL_multi_close)))
11910 /* allow nested opens and closes */
11911 else if (*s == PL_multi_close && --brackets <= 0)
11913 else if (*s == PL_multi_open)
11915 else if (!has_utf8 && !UTF8_IS_INVARIANT((U8)*s) && UTF)
11920 /* terminate the copied string and update the sv's end-of-string */
11922 SvCUR_set(sv, to - SvPVX_const(sv));
11925 * this next chunk reads more into the buffer if we're not done yet
11929 break; /* handle case where we are done yet :-) */
11931 #ifndef PERL_STRICT_CR
11932 if (to - SvPVX_const(sv) >= 2) {
11933 if ((to[-2] == '\r' && to[-1] == '\n') ||
11934 (to[-2] == '\n' && to[-1] == '\r'))
11938 SvCUR_set(sv, to - SvPVX_const(sv));
11940 else if (to[-1] == '\r')
11943 else if (to - SvPVX_const(sv) == 1 && to[-1] == '\r')
11948 /* if we're out of file, or a read fails, bail and reset the current
11949 line marker so we can report where the unterminated string began
11952 if (PL_madskills) {
11953 char * const tstart = SvPVX(PL_linestr) + stuffstart;
11955 sv_catpvn(PL_thisstuff, tstart, PL_bufend - tstart);
11957 PL_thisstuff = newSVpvn(tstart, PL_bufend - tstart);
11961 !(PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = filter_gets(PL_linestr, PL_rsfp, 0))) {
11963 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
11969 /* we read a line, so increment our line counter */
11970 CopLINE_inc(PL_curcop);
11972 /* update debugger info */
11973 if ((PERLDB_LINE || PERLDB_SAVESRC) && PL_curstash != PL_debstash)
11974 update_debugger_info(PL_linestr, NULL, 0);
11976 /* having changed the buffer, we must update PL_bufend */
11977 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
11978 PL_last_lop = PL_last_uni = NULL;
11981 /* at this point, we have successfully read the delimited string */
11983 if (!PL_encoding || UTF) {
11985 if (PL_madskills) {
11986 char * const tstart = SvPVX(PL_linestr) + stuffstart;
11987 const int len = s - tstart;
11989 sv_catpvn(PL_thisstuff, tstart, len);
11991 PL_thisstuff = newSVpvn(tstart, len);
11992 if (!PL_thisclose && !keep_delims)
11993 PL_thisclose = newSVpvn(s,termlen);
11998 sv_catpvn(sv, s, termlen);
12003 if (PL_madskills) {
12004 char * const tstart = SvPVX(PL_linestr) + stuffstart;
12005 const int len = s - tstart - termlen;
12007 sv_catpvn(PL_thisstuff, tstart, len);
12009 PL_thisstuff = newSVpvn(tstart, len);
12010 if (!PL_thisclose && !keep_delims)
12011 PL_thisclose = newSVpvn(s - termlen,termlen);
12015 if (has_utf8 || PL_encoding)
12018 PL_multi_end = CopLINE(PL_curcop);
12020 /* if we allocated too much space, give some back */
12021 if (SvCUR(sv) + 5 < SvLEN(sv)) {
12022 SvLEN_set(sv, SvCUR(sv) + 1);
12023 SvPV_renew(sv, SvLEN(sv));
12026 /* decide whether this is the first or second quoted string we've read
12039 takes: pointer to position in buffer
12040 returns: pointer to new position in buffer
12041 side-effects: builds ops for the constant in pl_yylval.op
12043 Read a number in any of the formats that Perl accepts:
12045 \d(_?\d)*(\.(\d(_?\d)*)?)?[Ee][\+\-]?(\d(_?\d)*) 12 12.34 12.
12046 \.\d(_?\d)*[Ee][\+\-]?(\d(_?\d)*) .34
12049 0x[0-9A-Fa-f](_?[0-9A-Fa-f])*
12051 Like most scan_ routines, it uses the PL_tokenbuf buffer to hold the
12054 If it reads a number without a decimal point or an exponent, it will
12055 try converting the number to an integer and see if it can do so
12056 without loss of precision.
12060 Perl_scan_num(pTHX_ const char *start, YYSTYPE* lvalp)
12063 register const char *s = start; /* current position in buffer */
12064 register char *d; /* destination in temp buffer */
12065 register char *e; /* end of temp buffer */
12066 NV nv; /* number read, as a double */
12067 SV *sv = NULL; /* place to put the converted number */
12068 bool floatit; /* boolean: int or float? */
12069 const char *lastub = NULL; /* position of last underbar */
12070 static char const number_too_long[] = "Number too long";
12072 PERL_ARGS_ASSERT_SCAN_NUM;
12074 /* We use the first character to decide what type of number this is */
12078 Perl_croak(aTHX_ "panic: scan_num");
12080 /* if it starts with a 0, it could be an octal number, a decimal in
12081 0.13 disguise, or a hexadecimal number, or a binary number. */
12085 u holds the "number so far"
12086 shift the power of 2 of the base
12087 (hex == 4, octal == 3, binary == 1)
12088 overflowed was the number more than we can hold?
12090 Shift is used when we add a digit. It also serves as an "are
12091 we in octal/hex/binary?" indicator to disallow hex characters
12092 when in octal mode.
12097 bool overflowed = FALSE;
12098 bool just_zero = TRUE; /* just plain 0 or binary number? */
12099 static const NV nvshift[5] = { 1.0, 2.0, 4.0, 8.0, 16.0 };
12100 static const char* const bases[5] =
12101 { "", "binary", "", "octal", "hexadecimal" };
12102 static const char* const Bases[5] =
12103 { "", "Binary", "", "Octal", "Hexadecimal" };
12104 static const char* const maxima[5] =
12106 "0b11111111111111111111111111111111",
12110 const char *base, *Base, *max;
12112 /* check for hex */
12117 } else if (s[1] == 'b') {
12122 /* check for a decimal in disguise */
12123 else if (s[1] == '.' || s[1] == 'e' || s[1] == 'E')
12125 /* so it must be octal */
12132 if (ckWARN(WARN_SYNTAX))
12133 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
12134 "Misplaced _ in number");
12138 base = bases[shift];
12139 Base = Bases[shift];
12140 max = maxima[shift];
12142 /* read the rest of the number */
12144 /* x is used in the overflow test,
12145 b is the digit we're adding on. */
12150 /* if we don't mention it, we're done */
12154 /* _ are ignored -- but warned about if consecutive */
12156 if (lastub && s == lastub + 1 && ckWARN(WARN_SYNTAX))
12157 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
12158 "Misplaced _ in number");
12162 /* 8 and 9 are not octal */
12163 case '8': case '9':
12165 yyerror(Perl_form(aTHX_ "Illegal octal digit '%c'", *s));
12169 case '2': case '3': case '4':
12170 case '5': case '6': case '7':
12172 yyerror(Perl_form(aTHX_ "Illegal binary digit '%c'", *s));
12175 case '0': case '1':
12176 b = *s++ & 15; /* ASCII digit -> value of digit */
12180 case 'a': case 'b': case 'c': case 'd': case 'e': case 'f':
12181 case 'A': case 'B': case 'C': case 'D': case 'E': case 'F':
12182 /* make sure they said 0x */
12185 b = (*s++ & 7) + 9;
12187 /* Prepare to put the digit we have onto the end
12188 of the number so far. We check for overflows.
12194 x = u << shift; /* make room for the digit */
12196 if ((x >> shift) != u
12197 && !(PL_hints & HINT_NEW_BINARY)) {
12200 if (ckWARN_d(WARN_OVERFLOW))
12201 Perl_warner(aTHX_ packWARN(WARN_OVERFLOW),
12202 "Integer overflow in %s number",
12205 u = x | b; /* add the digit to the end */
12208 n *= nvshift[shift];
12209 /* If an NV has not enough bits in its
12210 * mantissa to represent an UV this summing of
12211 * small low-order numbers is a waste of time
12212 * (because the NV cannot preserve the
12213 * low-order bits anyway): we could just
12214 * remember when did we overflow and in the
12215 * end just multiply n by the right
12223 /* if we get here, we had success: make a scalar value from
12228 /* final misplaced underbar check */
12229 if (s[-1] == '_') {
12230 if (ckWARN(WARN_SYNTAX))
12231 Perl_warner(aTHX_ packWARN(WARN_SYNTAX), "Misplaced _ in number");
12236 if (n > 4294967295.0 && ckWARN(WARN_PORTABLE))
12237 Perl_warner(aTHX_ packWARN(WARN_PORTABLE),
12238 "%s number > %s non-portable",
12244 if (u > 0xffffffff && ckWARN(WARN_PORTABLE))
12245 Perl_warner(aTHX_ packWARN(WARN_PORTABLE),
12246 "%s number > %s non-portable",
12251 if (just_zero && (PL_hints & HINT_NEW_INTEGER))
12252 sv = new_constant(start, s - start, "integer",
12253 sv, NULL, NULL, 0);
12254 else if (PL_hints & HINT_NEW_BINARY)
12255 sv = new_constant(start, s - start, "binary", sv, NULL, NULL, 0);
12260 handle decimal numbers.
12261 we're also sent here when we read a 0 as the first digit
12263 case '1': case '2': case '3': case '4': case '5':
12264 case '6': case '7': case '8': case '9': case '.':
12267 e = PL_tokenbuf + sizeof PL_tokenbuf - 6; /* room for various punctuation */
12270 /* read next group of digits and _ and copy into d */
12271 while (isDIGIT(*s) || *s == '_') {
12272 /* skip underscores, checking for misplaced ones
12276 if (lastub && s == lastub + 1 && ckWARN(WARN_SYNTAX))
12277 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
12278 "Misplaced _ in number");
12282 /* check for end of fixed-length buffer */
12284 Perl_croak(aTHX_ number_too_long);
12285 /* if we're ok, copy the character */
12290 /* final misplaced underbar check */
12291 if (lastub && s == lastub + 1) {
12292 if (ckWARN(WARN_SYNTAX))
12293 Perl_warner(aTHX_ packWARN(WARN_SYNTAX), "Misplaced _ in number");
12296 /* read a decimal portion if there is one. avoid
12297 3..5 being interpreted as the number 3. followed
12300 if (*s == '.' && s[1] != '.') {
12305 if (ckWARN(WARN_SYNTAX))
12306 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
12307 "Misplaced _ in number");
12311 /* copy, ignoring underbars, until we run out of digits.
12313 for (; isDIGIT(*s) || *s == '_'; s++) {
12314 /* fixed length buffer check */
12316 Perl_croak(aTHX_ number_too_long);
12318 if (lastub && s == lastub + 1 && ckWARN(WARN_SYNTAX))
12319 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
12320 "Misplaced _ in number");
12326 /* fractional part ending in underbar? */
12327 if (s[-1] == '_') {
12328 if (ckWARN(WARN_SYNTAX))
12329 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
12330 "Misplaced _ in number");
12332 if (*s == '.' && isDIGIT(s[1])) {
12333 /* oops, it's really a v-string, but without the "v" */
12339 /* read exponent part, if present */
12340 if ((*s == 'e' || *s == 'E') && strchr("+-0123456789_", s[1])) {
12344 /* regardless of whether user said 3E5 or 3e5, use lower 'e' */
12345 *d++ = 'e'; /* At least some Mach atof()s don't grok 'E' */
12347 /* stray preinitial _ */
12349 if (ckWARN(WARN_SYNTAX))
12350 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
12351 "Misplaced _ in number");
12355 /* allow positive or negative exponent */
12356 if (*s == '+' || *s == '-')
12359 /* stray initial _ */
12361 if (ckWARN(WARN_SYNTAX))
12362 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
12363 "Misplaced _ in number");
12367 /* read digits of exponent */
12368 while (isDIGIT(*s) || *s == '_') {
12371 Perl_croak(aTHX_ number_too_long);
12375 if (((lastub && s == lastub + 1) ||
12376 (!isDIGIT(s[1]) && s[1] != '_'))
12377 && ckWARN(WARN_SYNTAX))
12378 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
12379 "Misplaced _ in number");
12386 /* make an sv from the string */
12390 We try to do an integer conversion first if no characters
12391 indicating "float" have been found.
12396 const int flags = grok_number (PL_tokenbuf, d - PL_tokenbuf, &uv);
12398 if (flags == IS_NUMBER_IN_UV) {
12400 sv_setiv(sv, uv); /* Prefer IVs over UVs. */
12403 } else if (flags == (IS_NUMBER_IN_UV | IS_NUMBER_NEG)) {
12404 if (uv <= (UV) IV_MIN)
12405 sv_setiv(sv, -(IV)uv);
12412 /* terminate the string */
12414 nv = Atof(PL_tokenbuf);
12419 ? (PL_hints & HINT_NEW_FLOAT) : (PL_hints & HINT_NEW_INTEGER) ) {
12420 const char *const key = floatit ? "float" : "integer";
12421 const STRLEN keylen = floatit ? 5 : 7;
12422 sv = S_new_constant(aTHX_ PL_tokenbuf, d - PL_tokenbuf,
12423 key, keylen, sv, NULL, NULL, 0);
12427 /* if it starts with a v, it could be a v-string */
12430 sv = newSV(5); /* preallocate storage space */
12431 s = scan_vstring(s, PL_bufend, sv);
12435 /* make the op for the constant and return */
12438 lvalp->opval = newSVOP(OP_CONST, 0, sv);
12440 lvalp->opval = NULL;
12446 S_scan_formline(pTHX_ register char *s)
12449 register char *eol;
12451 SV * const stuff = newSVpvs("");
12452 bool needargs = FALSE;
12453 bool eofmt = FALSE;
12455 char *tokenstart = s;
12456 SV* savewhite = NULL;
12458 if (PL_madskills) {
12459 savewhite = PL_thiswhite;
12464 PERL_ARGS_ASSERT_SCAN_FORMLINE;
12466 while (!needargs) {
12469 #ifdef PERL_STRICT_CR
12470 while (SPACE_OR_TAB(*t))
12473 while (SPACE_OR_TAB(*t) || *t == '\r')
12476 if (*t == '\n' || t == PL_bufend) {
12481 if (PL_in_eval && !PL_rsfp) {
12482 eol = (char *) memchr(s,'\n',PL_bufend-s);
12487 eol = PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
12489 for (t = s; t < eol; t++) {
12490 if (*t == '~' && t[1] == '~' && SvCUR(stuff)) {
12492 goto enough; /* ~~ must be first line in formline */
12494 if (*t == '@' || *t == '^')
12498 sv_catpvn(stuff, s, eol-s);
12499 #ifndef PERL_STRICT_CR
12500 if (eol-s > 1 && eol[-2] == '\r' && eol[-1] == '\n') {
12501 char *end = SvPVX(stuff) + SvCUR(stuff);
12504 SvCUR_set(stuff, SvCUR(stuff) - 1);
12514 if (PL_madskills) {
12516 sv_catpvn(PL_thistoken, tokenstart, PL_bufend - tokenstart);
12518 PL_thistoken = newSVpvn(tokenstart, PL_bufend - tokenstart);
12521 s = filter_gets(PL_linestr, PL_rsfp, 0);
12523 tokenstart = PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = SvPVX(PL_linestr);
12525 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = SvPVX(PL_linestr);
12527 PL_bufend = PL_bufptr + SvCUR(PL_linestr);
12528 PL_last_lop = PL_last_uni = NULL;
12537 if (SvCUR(stuff)) {
12540 PL_lex_state = LEX_NORMAL;
12541 start_force(PL_curforce);
12542 NEXTVAL_NEXTTOKE.ival = 0;
12546 PL_lex_state = LEX_FORMLINE;
12548 if (UTF && is_utf8_string((U8*)SvPVX_const(stuff), SvCUR(stuff)))
12550 else if (PL_encoding)
12551 sv_recode_to_utf8(stuff, PL_encoding);
12553 start_force(PL_curforce);
12554 NEXTVAL_NEXTTOKE.opval = (OP*)newSVOP(OP_CONST, 0, stuff);
12556 start_force(PL_curforce);
12557 NEXTVAL_NEXTTOKE.ival = OP_FORMLINE;
12561 SvREFCNT_dec(stuff);
12563 PL_lex_formbrack = 0;
12567 if (PL_madskills) {
12569 sv_catpvn(PL_thistoken, tokenstart, s - tokenstart);
12571 PL_thistoken = newSVpvn(tokenstart, s - tokenstart);
12572 PL_thiswhite = savewhite;
12579 Perl_start_subparse(pTHX_ I32 is_format, U32 flags)
12582 const I32 oldsavestack_ix = PL_savestack_ix;
12583 CV* const outsidecv = PL_compcv;
12586 assert(SvTYPE(PL_compcv) == SVt_PVCV);
12588 SAVEI32(PL_subline);
12589 save_item(PL_subname);
12590 SAVESPTR(PL_compcv);
12592 PL_compcv = MUTABLE_CV(newSV_type(is_format ? SVt_PVFM : SVt_PVCV));
12593 CvFLAGS(PL_compcv) |= flags;
12595 PL_subline = CopLINE(PL_curcop);
12596 CvPADLIST(PL_compcv) = pad_new(padnew_SAVE|padnew_SAVESUB);
12597 CvOUTSIDE(PL_compcv) = MUTABLE_CV(SvREFCNT_inc_simple(outsidecv));
12598 CvOUTSIDE_SEQ(PL_compcv) = PL_cop_seqmax;
12600 return oldsavestack_ix;
12604 #pragma segment Perl_yylex
12607 S_yywarn(pTHX_ const char *const s)
12611 PERL_ARGS_ASSERT_YYWARN;
12613 PL_in_eval |= EVAL_WARNONLY;
12615 PL_in_eval &= ~EVAL_WARNONLY;
12620 Perl_yyerror(pTHX_ const char *const s)
12623 const char *where = NULL;
12624 const char *context = NULL;
12627 int yychar = PL_parser->yychar;
12629 PERL_ARGS_ASSERT_YYERROR;
12631 if (!yychar || (yychar == ';' && !PL_rsfp))
12633 else if (PL_oldoldbufptr && PL_bufptr > PL_oldoldbufptr &&
12634 PL_bufptr - PL_oldoldbufptr < 200 && PL_oldoldbufptr != PL_oldbufptr &&
12635 PL_oldbufptr != PL_bufptr) {
12638 The code below is removed for NetWare because it abends/crashes on NetWare
12639 when the script has error such as not having the closing quotes like:
12640 if ($var eq "value)
12641 Checking of white spaces is anyway done in NetWare code.
12644 while (isSPACE(*PL_oldoldbufptr))
12647 context = PL_oldoldbufptr;
12648 contlen = PL_bufptr - PL_oldoldbufptr;
12650 else if (PL_oldbufptr && PL_bufptr > PL_oldbufptr &&
12651 PL_bufptr - PL_oldbufptr < 200 && PL_oldbufptr != PL_bufptr) {
12654 The code below is removed for NetWare because it abends/crashes on NetWare
12655 when the script has error such as not having the closing quotes like:
12656 if ($var eq "value)
12657 Checking of white spaces is anyway done in NetWare code.
12660 while (isSPACE(*PL_oldbufptr))
12663 context = PL_oldbufptr;
12664 contlen = PL_bufptr - PL_oldbufptr;
12666 else if (yychar > 255)
12667 where = "next token ???";
12668 else if (yychar == -2) { /* YYEMPTY */
12669 if (PL_lex_state == LEX_NORMAL ||
12670 (PL_lex_state == LEX_KNOWNEXT && PL_lex_defer == LEX_NORMAL))
12671 where = "at end of line";
12672 else if (PL_lex_inpat)
12673 where = "within pattern";
12675 where = "within string";
12678 SV * const where_sv = newSVpvs_flags("next char ", SVs_TEMP);
12680 Perl_sv_catpvf(aTHX_ where_sv, "^%c", toCTRL(yychar));
12681 else if (isPRINT_LC(yychar)) {
12682 const char string = yychar;
12683 sv_catpvn(where_sv, &string, 1);
12686 Perl_sv_catpvf(aTHX_ where_sv, "\\%03o", yychar & 255);
12687 where = SvPVX_const(where_sv);
12689 msg = sv_2mortal(newSVpv(s, 0));
12690 Perl_sv_catpvf(aTHX_ msg, " at %s line %"IVdf", ",
12691 OutCopFILE(PL_curcop), (IV)CopLINE(PL_curcop));
12693 Perl_sv_catpvf(aTHX_ msg, "near \"%.*s\"\n", contlen, context);
12695 Perl_sv_catpvf(aTHX_ msg, "%s\n", where);
12696 if (PL_multi_start < PL_multi_end && (U32)(CopLINE(PL_curcop) - PL_multi_end) <= 1) {
12697 Perl_sv_catpvf(aTHX_ msg,
12698 " (Might be a runaway multi-line %c%c string starting on line %"IVdf")\n",
12699 (int)PL_multi_open,(int)PL_multi_close,(IV)PL_multi_start);
12702 if (PL_in_eval & EVAL_WARNONLY) {
12703 if (ckWARN_d(WARN_SYNTAX))
12704 Perl_warner(aTHX_ packWARN(WARN_SYNTAX), "%"SVf, SVfARG(msg));
12708 if (PL_error_count >= 10) {
12709 if (PL_in_eval && SvCUR(ERRSV))
12710 Perl_croak(aTHX_ "%"SVf"%s has too many errors.\n",
12711 SVfARG(ERRSV), OutCopFILE(PL_curcop));
12713 Perl_croak(aTHX_ "%s has too many errors.\n",
12714 OutCopFILE(PL_curcop));
12717 PL_in_my_stash = NULL;
12721 #pragma segment Main
12725 S_swallow_bom(pTHX_ U8 *s)
12728 const STRLEN slen = SvCUR(PL_linestr);
12730 PERL_ARGS_ASSERT_SWALLOW_BOM;
12734 if (s[1] == 0xFE) {
12735 /* UTF-16 little-endian? (or UTF32-LE?) */
12736 if (s[2] == 0 && s[3] == 0) /* UTF-32 little-endian */
12737 Perl_croak(aTHX_ "Unsupported script encoding UTF32-LE");
12738 #ifndef PERL_NO_UTF16_FILTER
12739 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF16-LE script encoding (BOM)\n");
12742 if (PL_bufend > (char*)s) {
12746 filter_add(utf16rev_textfilter, NULL);
12747 Newx(news, (PL_bufend - (char*)s) * 3 / 2 + 1, U8);
12748 utf16_to_utf8_reversed(s, news,
12749 PL_bufend - (char*)s - 1,
12751 sv_setpvn(PL_linestr, (const char*)news, newlen);
12753 s = (U8*)SvPVX(PL_linestr);
12754 Copy(news, s, newlen, U8);
12758 SvUTF8_on(PL_linestr);
12759 s = (U8*)SvPVX(PL_linestr);
12761 /* FIXME - is this a general bug fix? */
12764 PL_bufend = SvPVX(PL_linestr) + newlen;
12767 Perl_croak(aTHX_ "Unsupported script encoding UTF16-LE");
12772 if (s[1] == 0xFF) { /* UTF-16 big-endian? */
12773 #ifndef PERL_NO_UTF16_FILTER
12774 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16BE script encoding (BOM)\n");
12777 if (PL_bufend > (char *)s) {
12781 filter_add(utf16_textfilter, NULL);
12782 Newx(news, (PL_bufend - (char*)s) * 3 / 2 + 1, U8);
12783 utf16_to_utf8(s, news,
12784 PL_bufend - (char*)s,
12786 sv_setpvn(PL_linestr, (const char*)news, newlen);
12788 SvUTF8_on(PL_linestr);
12789 s = (U8*)SvPVX(PL_linestr);
12790 PL_bufend = SvPVX(PL_linestr) + newlen;
12793 Perl_croak(aTHX_ "Unsupported script encoding UTF16-BE");
12798 if (slen > 2 && s[1] == 0xBB && s[2] == 0xBF) {
12799 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-8 script encoding (BOM)\n");
12800 s += 3; /* UTF-8 */
12806 if (s[2] == 0xFE && s[3] == 0xFF) {
12807 /* UTF-32 big-endian */
12808 Perl_croak(aTHX_ "Unsupported script encoding UTF32-BE");
12811 else if (s[2] == 0 && s[3] != 0) {
12814 * are a good indicator of UTF-16BE. */
12815 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16BE script encoding (no BOM)\n");
12821 if (slen > 3 && s[1] == 0x73 && s[2] == 0x66 && s[3] == 0x73) {
12822 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-8 script encoding (BOM)\n");
12823 s += 4; /* UTF-8 */
12829 if (slen > 3 && s[1] == 0 && s[2] != 0 && s[3] == 0) {
12832 * are a good indicator of UTF-16LE. */
12833 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16LE script encoding (no BOM)\n");
12841 #ifndef PERL_NO_UTF16_FILTER
12843 utf16_textfilter(pTHX_ int idx, SV *sv, int maxlen)
12846 const STRLEN old = SvCUR(sv);
12847 const I32 count = FILTER_READ(idx+1, sv, maxlen);
12848 DEBUG_P(PerlIO_printf(Perl_debug_log,
12849 "utf16_textfilter(%p): %d %d (%d)\n",
12850 FPTR2DPTR(void *, utf16_textfilter),
12851 idx, maxlen, (int) count));
12855 Newx(tmps, SvCUR(sv) * 3 / 2 + 1, U8);
12856 Copy(SvPVX_const(sv), tmps, old, char);
12857 utf16_to_utf8((U8*)SvPVX_const(sv) + old, tmps + old,
12858 SvCUR(sv) - old, &newlen);
12859 sv_usepvn(sv, (char*)tmps, (STRLEN)newlen + old);
12861 DEBUG_P({sv_dump(sv);});
12866 utf16rev_textfilter(pTHX_ int idx, SV *sv, int maxlen)
12869 const STRLEN old = SvCUR(sv);
12870 const I32 count = FILTER_READ(idx+1, sv, maxlen);
12871 DEBUG_P(PerlIO_printf(Perl_debug_log,
12872 "utf16rev_textfilter(%p): %d %d (%d)\n",
12873 FPTR2DPTR(void *, utf16rev_textfilter),
12874 idx, maxlen, (int) count));
12878 Newx(tmps, SvCUR(sv) * 3 / 2 + 1, U8);
12879 Copy(SvPVX_const(sv), tmps, old, char);
12880 utf16_to_utf8((U8*)SvPVX_const(sv) + old, tmps + old,
12881 SvCUR(sv) - old, &newlen);
12882 sv_usepvn(sv, (char*)tmps, (STRLEN)newlen + old);
12884 DEBUG_P({ sv_dump(sv); });
12890 Returns a pointer to the next character after the parsed
12891 vstring, as well as updating the passed in sv.
12893 Function must be called like
12896 s = scan_vstring(s,e,sv);
12898 where s and e are the start and end of the string.
12899 The sv should already be large enough to store the vstring
12900 passed in, for performance reasons.
12905 Perl_scan_vstring(pTHX_ const char *s, const char *const e, SV *sv)
12908 const char *pos = s;
12909 const char *start = s;
12911 PERL_ARGS_ASSERT_SCAN_VSTRING;
12913 if (*pos == 'v') pos++; /* get past 'v' */
12914 while (pos < e && (isDIGIT(*pos) || *pos == '_'))
12916 if ( *pos != '.') {
12917 /* this may not be a v-string if followed by => */
12918 const char *next = pos;
12919 while (next < e && isSPACE(*next))
12921 if ((e - next) >= 2 && *next == '=' && next[1] == '>' ) {
12922 /* return string not v-string */
12923 sv_setpvn(sv,(char *)s,pos-s);
12924 return (char *)pos;
12928 if (!isALPHA(*pos)) {
12929 U8 tmpbuf[UTF8_MAXBYTES+1];
12932 s++; /* get past 'v' */
12937 /* this is atoi() that tolerates underscores */
12940 const char *end = pos;
12942 while (--end >= s) {
12944 const UV orev = rev;
12945 rev += (*end - '0') * mult;
12947 if (orev > rev && ckWARN_d(WARN_OVERFLOW))
12948 Perl_warner(aTHX_ packWARN(WARN_OVERFLOW),
12949 "Integer overflow in decimal number");
12953 if (rev > 0x7FFFFFFF)
12954 Perl_croak(aTHX_ "In EBCDIC the v-string components cannot exceed 2147483647");
12956 /* Append native character for the rev point */
12957 tmpend = uvchr_to_utf8(tmpbuf, rev);
12958 sv_catpvn(sv, (const char*)tmpbuf, tmpend - tmpbuf);
12959 if (!UNI_IS_INVARIANT(NATIVE_TO_UNI(rev)))
12961 if (pos + 1 < e && *pos == '.' && isDIGIT(pos[1]))
12967 while (pos < e && (isDIGIT(*pos) || *pos == '_'))
12971 sv_magic(sv,NULL,PERL_MAGIC_vstring,(const char*)start, pos-start);
12979 * c-indentation-style: bsd
12980 * c-basic-offset: 4
12981 * indent-tabs-mode: t
12984 * ex: set ts=8 sts=4 sw=4 noet: