3 * Copyright (C) 1991, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999,
4 * 2000, 2001, 2002, 2003, 2004, 2005, 2006, 2007, 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
16 * This file is the lexer for Perl. It's closely linked to the
19 * The main routine is yylex(), which returns the next token.
23 #define PERL_IN_TOKE_C
26 #define yylval (PL_parser->yylval)
28 /* YYINITDEPTH -- initial size of the parser's stacks. */
29 #define YYINITDEPTH 200
31 /* XXX temporary backwards compatibility */
32 #define PL_lex_brackets (PL_parser->lex_brackets)
33 #define PL_lex_brackstack (PL_parser->lex_brackstack)
34 #define PL_lex_casemods (PL_parser->lex_casemods)
35 #define PL_lex_casestack (PL_parser->lex_casestack)
36 #define PL_lex_defer (PL_parser->lex_defer)
37 #define PL_lex_dojoin (PL_parser->lex_dojoin)
38 #define PL_lex_expect (PL_parser->lex_expect)
39 #define PL_lex_formbrack (PL_parser->lex_formbrack)
40 #define PL_lex_inpat (PL_parser->lex_inpat)
41 #define PL_lex_inwhat (PL_parser->lex_inwhat)
42 #define PL_lex_op (PL_parser->lex_op)
43 #define PL_lex_repl (PL_parser->lex_repl)
44 #define PL_lex_starts (PL_parser->lex_starts)
45 #define PL_lex_stuff (PL_parser->lex_stuff)
46 #define PL_multi_start (PL_parser->multi_start)
47 #define PL_multi_open (PL_parser->multi_open)
48 #define PL_multi_close (PL_parser->multi_close)
49 #define PL_pending_ident (PL_parser->pending_ident)
50 #define PL_preambled (PL_parser->preambled)
51 #define PL_sublex_info (PL_parser->sublex_info)
54 # define PL_endwhite (PL_parser->endwhite)
55 # define PL_faketokens (PL_parser->faketokens)
56 # define PL_lasttoke (PL_parser->lasttoke)
57 # define PL_nextwhite (PL_parser->nextwhite)
58 # define PL_realtokenstart (PL_parser->realtokenstart)
59 # define PL_skipwhite (PL_parser->skipwhite)
60 # define PL_thisclose (PL_parser->thisclose)
61 # define PL_thismad (PL_parser->thismad)
62 # define PL_thisopen (PL_parser->thisopen)
63 # define PL_thisstuff (PL_parser->thisstuff)
64 # define PL_thistoken (PL_parser->thistoken)
65 # define PL_thiswhite (PL_parser->thiswhite)
69 S_pending_ident(pTHX);
71 static const char ident_too_long[] = "Identifier too long";
72 static const char commaless_variable_list[] = "comma-less variable list";
74 static void restore_rsfp(pTHX_ void *f);
75 #ifndef PERL_NO_UTF16_FILTER
76 static I32 utf16_textfilter(pTHX_ int idx, SV *sv, int maxlen);
77 static I32 utf16rev_textfilter(pTHX_ int idx, SV *sv, int maxlen);
81 # define CURMAD(slot,sv) if (PL_madskills) { curmad(slot,sv); sv = 0; }
82 # define NEXTVAL_NEXTTOKE PL_nexttoke[PL_curforce].next_val
84 # define CURMAD(slot,sv)
85 # define NEXTVAL_NEXTTOKE PL_nextval[PL_nexttoke]
88 #define XFAKEBRACK 128
91 #ifdef USE_UTF8_SCRIPTS
92 # define UTF (!IN_BYTES)
94 # define UTF ((PL_linestr && DO_UTF8(PL_linestr)) || (PL_hints & HINT_UTF8))
97 /* In variables named $^X, these are the legal values for X.
98 * 1999-02-27 mjd-perl-patch@plover.com */
99 #define isCONTROLVAR(x) (isUPPER(x) || strchr("[\\]^_?", (x)))
101 /* On MacOS, respect nonbreaking spaces */
102 #ifdef MACOS_TRADITIONAL
103 #define SPACE_OR_TAB(c) ((c)==' '||(c)=='\312'||(c)=='\t')
105 #define SPACE_OR_TAB(c) ((c)==' '||(c)=='\t')
108 /* LEX_* are values for PL_lex_state, the state of the lexer.
109 * They are arranged oddly so that the guard on the switch statement
110 * can get by with a single comparison (if the compiler is smart enough).
113 /* #define LEX_NOTPARSING 11 is done in perl.h. */
115 #define LEX_NORMAL 10 /* normal code (ie not within "...") */
116 #define LEX_INTERPNORMAL 9 /* code within a string, eg "$foo[$x+1]" */
117 #define LEX_INTERPCASEMOD 8 /* expecting a \U, \Q or \E etc */
118 #define LEX_INTERPPUSH 7 /* starting a new sublex parse level */
119 #define LEX_INTERPSTART 6 /* expecting the start of a $var */
121 /* at end of code, eg "$x" followed by: */
122 #define LEX_INTERPEND 5 /* ... eg not one of [, { or -> */
123 #define LEX_INTERPENDMAYBE 4 /* ... eg one of [, { or -> */
125 #define LEX_INTERPCONCAT 3 /* expecting anything, eg at start of
126 string or after \E, $foo, etc */
127 #define LEX_INTERPCONST 2 /* NOT USED */
128 #define LEX_FORMLINE 1 /* expecting a format line */
129 #define LEX_KNOWNEXT 0 /* next token known; just return it */
133 static const char* const lex_state_names[] = {
152 #include "keywords.h"
154 /* CLINE is a macro that ensures PL_copline has a sane value */
159 #define CLINE (PL_copline = (CopLINE(PL_curcop) < PL_copline ? CopLINE(PL_curcop) : PL_copline))
162 # define SKIPSPACE0(s) skipspace0(s)
163 # define SKIPSPACE1(s) skipspace1(s)
164 # define SKIPSPACE2(s,tsv) skipspace2(s,&tsv)
165 # define PEEKSPACE(s) skipspace2(s,0)
167 # define SKIPSPACE0(s) skipspace(s)
168 # define SKIPSPACE1(s) skipspace(s)
169 # define SKIPSPACE2(s,tsv) skipspace(s)
170 # define PEEKSPACE(s) skipspace(s)
174 * Convenience functions to return different tokens and prime the
175 * lexer for the next token. They all take an argument.
177 * TOKEN : generic token (used for '(', DOLSHARP, etc)
178 * OPERATOR : generic operator
179 * AOPERATOR : assignment operator
180 * PREBLOCK : beginning the block after an if, while, foreach, ...
181 * PRETERMBLOCK : beginning a non-code-defining {} block (eg, hash ref)
182 * PREREF : *EXPR where EXPR is not a simple identifier
183 * TERM : expression term
184 * LOOPX : loop exiting command (goto, last, dump, etc)
185 * FTST : file test operator
186 * FUN0 : zero-argument function
187 * FUN1 : not used, except for not, which isn't a UNIOP
188 * BOop : bitwise or or xor
190 * SHop : shift operator
191 * PWop : power operator
192 * PMop : pattern-matching operator
193 * Aop : addition-level operator
194 * Mop : multiplication-level operator
195 * Eop : equality-testing operator
196 * Rop : relational operator <= != gt
198 * Also see LOP and lop() below.
201 #ifdef DEBUGGING /* Serve -DT. */
202 # define REPORT(retval) tokereport((I32)retval)
204 # define REPORT(retval) (retval)
207 #define TOKEN(retval) return ( PL_bufptr = s, REPORT(retval))
208 #define OPERATOR(retval) return (PL_expect = XTERM, PL_bufptr = s, REPORT(retval))
209 #define AOPERATOR(retval) return ao((PL_expect = XTERM, PL_bufptr = s, REPORT(retval)))
210 #define PREBLOCK(retval) return (PL_expect = XBLOCK,PL_bufptr = s, REPORT(retval))
211 #define PRETERMBLOCK(retval) return (PL_expect = XTERMBLOCK,PL_bufptr = s, REPORT(retval))
212 #define PREREF(retval) return (PL_expect = XREF,PL_bufptr = s, REPORT(retval))
213 #define TERM(retval) return (CLINE, PL_expect = XOPERATOR, PL_bufptr = s, REPORT(retval))
214 #define LOOPX(f) return (yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)LOOPEX))
215 #define FTST(f) return (yylval.ival=f, PL_expect=XTERMORDORDOR, PL_bufptr=s, REPORT((int)UNIOP))
216 #define FUN0(f) return (yylval.ival=f, PL_expect=XOPERATOR, PL_bufptr=s, REPORT((int)FUNC0))
217 #define FUN1(f) return (yylval.ival=f, PL_expect=XOPERATOR, PL_bufptr=s, REPORT((int)FUNC1))
218 #define BOop(f) return ao((yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)BITOROP)))
219 #define BAop(f) return ao((yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)BITANDOP)))
220 #define SHop(f) return ao((yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)SHIFTOP)))
221 #define PWop(f) return ao((yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)POWOP)))
222 #define PMop(f) return(yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)MATCHOP))
223 #define Aop(f) return ao((yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)ADDOP)))
224 #define Mop(f) return ao((yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)MULOP)))
225 #define Eop(f) return (yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)EQOP))
226 #define Rop(f) return (yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)RELOP))
228 /* This bit of chicanery makes a unary function followed by
229 * a parenthesis into a function with one argument, highest precedence.
230 * The UNIDOR macro is for unary functions that can be followed by the //
231 * operator (such as C<shift // 0>).
233 #define UNI2(f,x) { \
237 PL_last_uni = PL_oldbufptr; \
238 PL_last_lop_op = f; \
240 return REPORT( (int)FUNC1 ); \
242 return REPORT( *s=='(' ? (int)FUNC1 : (int)UNIOP ); \
244 #define UNI(f) UNI2(f,XTERM)
245 #define UNIDOR(f) UNI2(f,XTERMORDORDOR)
247 #define UNIBRACK(f) { \
250 PL_last_uni = PL_oldbufptr; \
252 return REPORT( (int)FUNC1 ); \
254 return REPORT( (*s == '(') ? (int)FUNC1 : (int)UNIOP ); \
257 /* grandfather return to old style */
258 #define OLDLOP(f) return(yylval.ival=f,PL_expect = XTERM,PL_bufptr = s,(int)LSTOP)
262 /* how to interpret the yylval associated with the token */
266 TOKENTYPE_OPNUM, /* yylval.ival contains an opcode number */
272 static struct debug_tokens {
274 enum token_type type;
276 } const debug_tokens[] =
278 { ADDOP, TOKENTYPE_OPNUM, "ADDOP" },
279 { ANDAND, TOKENTYPE_NONE, "ANDAND" },
280 { ANDOP, TOKENTYPE_NONE, "ANDOP" },
281 { ANONSUB, TOKENTYPE_IVAL, "ANONSUB" },
282 { ARROW, TOKENTYPE_NONE, "ARROW" },
283 { ASSIGNOP, TOKENTYPE_OPNUM, "ASSIGNOP" },
284 { BITANDOP, TOKENTYPE_OPNUM, "BITANDOP" },
285 { BITOROP, TOKENTYPE_OPNUM, "BITOROP" },
286 { COLONATTR, TOKENTYPE_NONE, "COLONATTR" },
287 { CONTINUE, TOKENTYPE_NONE, "CONTINUE" },
288 { DEFAULT, TOKENTYPE_NONE, "DEFAULT" },
289 { DO, TOKENTYPE_NONE, "DO" },
290 { DOLSHARP, TOKENTYPE_NONE, "DOLSHARP" },
291 { DORDOR, TOKENTYPE_NONE, "DORDOR" },
292 { DOROP, TOKENTYPE_OPNUM, "DOROP" },
293 { DOTDOT, TOKENTYPE_IVAL, "DOTDOT" },
294 { ELSE, TOKENTYPE_NONE, "ELSE" },
295 { ELSIF, TOKENTYPE_IVAL, "ELSIF" },
296 { EQOP, TOKENTYPE_OPNUM, "EQOP" },
297 { FOR, TOKENTYPE_IVAL, "FOR" },
298 { FORMAT, TOKENTYPE_NONE, "FORMAT" },
299 { FUNC, TOKENTYPE_OPNUM, "FUNC" },
300 { FUNC0, TOKENTYPE_OPNUM, "FUNC0" },
301 { FUNC0SUB, TOKENTYPE_OPVAL, "FUNC0SUB" },
302 { FUNC1, TOKENTYPE_OPNUM, "FUNC1" },
303 { FUNCMETH, TOKENTYPE_OPVAL, "FUNCMETH" },
304 { GIVEN, TOKENTYPE_IVAL, "GIVEN" },
305 { HASHBRACK, TOKENTYPE_NONE, "HASHBRACK" },
306 { IF, TOKENTYPE_IVAL, "IF" },
307 { LABEL, TOKENTYPE_PVAL, "LABEL" },
308 { LOCAL, TOKENTYPE_IVAL, "LOCAL" },
309 { LOOPEX, TOKENTYPE_OPNUM, "LOOPEX" },
310 { LSTOP, TOKENTYPE_OPNUM, "LSTOP" },
311 { LSTOPSUB, TOKENTYPE_OPVAL, "LSTOPSUB" },
312 { MATCHOP, TOKENTYPE_OPNUM, "MATCHOP" },
313 { METHOD, TOKENTYPE_OPVAL, "METHOD" },
314 { MULOP, TOKENTYPE_OPNUM, "MULOP" },
315 { MY, TOKENTYPE_IVAL, "MY" },
316 { MYSUB, TOKENTYPE_NONE, "MYSUB" },
317 { NOAMP, TOKENTYPE_NONE, "NOAMP" },
318 { NOTOP, TOKENTYPE_NONE, "NOTOP" },
319 { OROP, TOKENTYPE_IVAL, "OROP" },
320 { OROR, TOKENTYPE_NONE, "OROR" },
321 { PACKAGE, TOKENTYPE_NONE, "PACKAGE" },
322 { PMFUNC, TOKENTYPE_OPVAL, "PMFUNC" },
323 { POSTDEC, TOKENTYPE_NONE, "POSTDEC" },
324 { POSTINC, TOKENTYPE_NONE, "POSTINC" },
325 { POWOP, TOKENTYPE_OPNUM, "POWOP" },
326 { PREDEC, TOKENTYPE_NONE, "PREDEC" },
327 { PREINC, TOKENTYPE_NONE, "PREINC" },
328 { PRIVATEREF, TOKENTYPE_OPVAL, "PRIVATEREF" },
329 { REFGEN, TOKENTYPE_NONE, "REFGEN" },
330 { RELOP, TOKENTYPE_OPNUM, "RELOP" },
331 { SHIFTOP, TOKENTYPE_OPNUM, "SHIFTOP" },
332 { SUB, TOKENTYPE_NONE, "SUB" },
333 { THING, TOKENTYPE_OPVAL, "THING" },
334 { UMINUS, TOKENTYPE_NONE, "UMINUS" },
335 { UNIOP, TOKENTYPE_OPNUM, "UNIOP" },
336 { UNIOPSUB, TOKENTYPE_OPVAL, "UNIOPSUB" },
337 { UNLESS, TOKENTYPE_IVAL, "UNLESS" },
338 { UNTIL, TOKENTYPE_IVAL, "UNTIL" },
339 { USE, TOKENTYPE_IVAL, "USE" },
340 { WHEN, TOKENTYPE_IVAL, "WHEN" },
341 { WHILE, TOKENTYPE_IVAL, "WHILE" },
342 { WORD, TOKENTYPE_OPVAL, "WORD" },
343 { 0, TOKENTYPE_NONE, NULL }
346 /* dump the returned token in rv, plus any optional arg in yylval */
349 S_tokereport(pTHX_ I32 rv)
353 const char *name = NULL;
354 enum token_type type = TOKENTYPE_NONE;
355 const struct debug_tokens *p;
356 SV* const report = newSVpvs("<== ");
358 for (p = debug_tokens; p->token; p++) {
359 if (p->token == (int)rv) {
366 Perl_sv_catpv(aTHX_ report, name);
367 else if ((char)rv > ' ' && (char)rv < '~')
368 Perl_sv_catpvf(aTHX_ report, "'%c'", (char)rv);
370 sv_catpvs(report, "EOF");
372 Perl_sv_catpvf(aTHX_ report, "?? %"IVdf, (IV)rv);
375 case TOKENTYPE_GVVAL: /* doesn't appear to be used */
378 Perl_sv_catpvf(aTHX_ report, "(ival=%"IVdf")", (IV)yylval.ival);
380 case TOKENTYPE_OPNUM:
381 Perl_sv_catpvf(aTHX_ report, "(ival=op_%s)",
382 PL_op_name[yylval.ival]);
385 Perl_sv_catpvf(aTHX_ report, "(pval=\"%s\")", yylval.pval);
387 case TOKENTYPE_OPVAL:
389 Perl_sv_catpvf(aTHX_ report, "(opval=op_%s)",
390 PL_op_name[yylval.opval->op_type]);
391 if (yylval.opval->op_type == OP_CONST) {
392 Perl_sv_catpvf(aTHX_ report, " %s",
393 SvPEEK(cSVOPx_sv(yylval.opval)));
398 sv_catpvs(report, "(opval=null)");
401 PerlIO_printf(Perl_debug_log, "### %s\n\n", SvPV_nolen_const(report));
407 /* print the buffer with suitable escapes */
410 S_printbuf(pTHX_ const char* fmt, const char* s)
412 SV* const tmp = newSVpvs("");
413 PerlIO_printf(Perl_debug_log, fmt, pv_display(tmp, s, strlen(s), 0, 60));
422 * This subroutine detects &&=, ||=, and //= and turns an ANDAND, OROR or DORDOR
423 * into an OP_ANDASSIGN, OP_ORASSIGN, or OP_DORASSIGN
427 S_ao(pTHX_ int toketype)
430 if (*PL_bufptr == '=') {
432 if (toketype == ANDAND)
433 yylval.ival = OP_ANDASSIGN;
434 else if (toketype == OROR)
435 yylval.ival = OP_ORASSIGN;
436 else if (toketype == DORDOR)
437 yylval.ival = OP_DORASSIGN;
445 * When Perl expects an operator and finds something else, no_op
446 * prints the warning. It always prints "<something> found where
447 * operator expected. It prints "Missing semicolon on previous line?"
448 * if the surprise occurs at the start of the line. "do you need to
449 * predeclare ..." is printed out for code like "sub bar; foo bar $x"
450 * where the compiler doesn't know if foo is a method call or a function.
451 * It prints "Missing operator before end of line" if there's nothing
452 * after the missing operator, or "... before <...>" if there is something
453 * after the missing operator.
457 S_no_op(pTHX_ const char *what, char *s)
460 char * const oldbp = PL_bufptr;
461 const bool is_first = (PL_oldbufptr == PL_linestart);
467 yywarn(Perl_form(aTHX_ "%s found where operator expected", what));
468 if (ckWARN_d(WARN_SYNTAX)) {
470 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
471 "\t(Missing semicolon on previous line?)\n");
472 else if (PL_oldoldbufptr && isIDFIRST_lazy_if(PL_oldoldbufptr,UTF)) {
474 for (t = PL_oldoldbufptr; (isALNUM_lazy_if(t,UTF) || *t == ':'); t++)
476 if (t < PL_bufptr && isSPACE(*t))
477 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
478 "\t(Do you need to predeclare %.*s?)\n",
479 (int)(t - PL_oldoldbufptr), PL_oldoldbufptr);
483 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
484 "\t(Missing operator before %.*s?)\n", (int)(s - oldbp), oldbp);
492 * Complain about missing quote/regexp/heredoc terminator.
493 * If it's called with NULL then it cauterizes the line buffer.
494 * If we're in a delimited string and the delimiter is a control
495 * character, it's reformatted into a two-char sequence like ^C.
500 S_missingterm(pTHX_ char *s)
506 char * const nl = strrchr(s,'\n');
512 iscntrl(PL_multi_close)
514 PL_multi_close < 32 || PL_multi_close == 127
518 tmpbuf[1] = (char)toCTRL(PL_multi_close);
523 *tmpbuf = (char)PL_multi_close;
527 q = strchr(s,'"') ? '\'' : '"';
528 Perl_croak(aTHX_ "Can't find string terminator %c%s%c anywhere before EOF",q,s,q);
531 #define FEATURE_IS_ENABLED(name) \
532 ((0 != (PL_hints & HINT_LOCALIZE_HH)) \
533 && S_feature_is_enabled(aTHX_ STR_WITH_LEN(name)))
535 * S_feature_is_enabled
536 * Check whether the named feature is enabled.
539 S_feature_is_enabled(pTHX_ const char *name, STRLEN namelen)
542 HV * const hinthv = GvHV(PL_hintgv);
543 char he_name[32] = "feature_";
544 (void) my_strlcpy(&he_name[8], name, 24);
546 return (hinthv && hv_exists(hinthv, he_name, 8 + namelen));
554 Perl_deprecate(pTHX_ const char *s)
556 if (ckWARN(WARN_DEPRECATED))
557 Perl_warner(aTHX_ packWARN(WARN_DEPRECATED), "Use of %s is deprecated", s);
561 Perl_deprecate_old(pTHX_ const char *s)
563 /* This function should NOT be called for any new deprecated warnings */
564 /* Use Perl_deprecate instead */
566 /* It is here to maintain backward compatibility with the pre-5.8 */
567 /* warnings category hierarchy. The "deprecated" category used to */
568 /* live under the "syntax" category. It is now a top-level category */
569 /* in its own right. */
571 if (ckWARN2(WARN_DEPRECATED, WARN_SYNTAX))
572 Perl_warner(aTHX_ packWARN2(WARN_DEPRECATED, WARN_SYNTAX),
573 "Use of %s is deprecated", s);
577 * experimental text filters for win32 carriage-returns, utf16-to-utf8 and
578 * utf16-to-utf8-reversed.
581 #ifdef PERL_CR_FILTER
585 register const char *s = SvPVX_const(sv);
586 register const char * const e = s + SvCUR(sv);
587 /* outer loop optimized to do nothing if there are no CR-LFs */
589 if (*s++ == '\r' && *s == '\n') {
590 /* hit a CR-LF, need to copy the rest */
591 register char *d = s - 1;
594 if (*s == '\r' && s[1] == '\n')
605 S_cr_textfilter(pTHX_ int idx, SV *sv, int maxlen)
607 const I32 count = FILTER_READ(idx+1, sv, maxlen);
608 if (count > 0 && !maxlen)
618 * Initialize variables. Uses the Perl save_stack to save its state (for
619 * recursive calls to the parser).
623 Perl_lex_start(pTHX_ SV *line)
626 const char *s = NULL;
630 /* create and initialise a parser */
632 Newxz(parser, 1, yy_parser);
633 parser->old_parser = PL_parser;
636 Newx(parser->stack, YYINITDEPTH, yy_stack_frame);
637 parser->ps = parser->stack;
638 parser->stack_size = YYINITDEPTH;
640 parser->stack->state = 0;
641 parser->yyerrstatus = 0;
642 parser->yychar = YYEMPTY; /* Cause a token to be read. */
644 /* initialise lexer state */
646 SAVEI32(PL_lex_state);
648 if (PL_lex_state == LEX_KNOWNEXT) {
649 I32 toke = parser->old_parser->lasttoke;
650 while (--toke >= 0) {
651 SAVEI32(PL_nexttoke[toke].next_type);
652 SAVEVPTR(PL_nexttoke[toke].next_val);
654 SAVEVPTR(PL_nexttoke[toke].next_mad);
657 SAVEI32(PL_curforce);
659 if (PL_lex_state == LEX_KNOWNEXT) {
660 I32 toke = PL_nexttoke;
661 while (--toke >= 0) {
662 SAVEI32(PL_nexttype[toke]);
663 SAVEVPTR(PL_nextval[toke]);
665 SAVEI32(PL_nexttoke);
668 SAVECOPLINE(PL_curcop);
671 SAVEPPTR(PL_oldbufptr);
672 SAVEPPTR(PL_oldoldbufptr);
673 SAVEPPTR(PL_last_lop);
674 SAVEPPTR(PL_last_uni);
675 SAVEPPTR(PL_linestart);
676 SAVESPTR(PL_linestr);
677 SAVEDESTRUCTOR_X(restore_rsfp, PL_rsfp);
680 PL_lex_state = LEX_NORMAL;
682 Newx(parser->lex_brackstack, 120, char);
683 Newx(parser->lex_casestack, 12, char);
684 *parser->lex_casestack = '\0';
690 s = SvPV_const(line, len);
695 PL_linestr = newSVpvs("\n;");
696 } else if (SvREADONLY(line) || s[len-1] != ';') {
697 PL_linestr = newSVsv(line);
699 sv_catpvs(PL_linestr, "\n;");
702 SvREFCNT_inc_simple_void_NN(line);
705 /* PL_linestr needs to survive until end of scope, not just the next
706 FREETMPS. See changes 17505 and 17546 which fixed the symptoms only. */
707 SAVEFREESV(PL_linestr);
708 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = SvPVX(PL_linestr);
709 PL_bufend = PL_bufptr + SvCUR(PL_linestr);
710 PL_last_lop = PL_last_uni = NULL;
716 * Finalizer for lexing operations. Must be called when the parser is
717 * done with the lexer.
724 PL_doextract = FALSE;
729 * This subroutine has nothing to do with tilting, whether at windmills
730 * or pinball tables. Its name is short for "increment line". It
731 * increments the current line number in CopLINE(PL_curcop) and checks
732 * to see whether the line starts with a comment of the form
733 * # line 500 "foo.pm"
734 * If so, it sets the current line number and file to the values in the comment.
738 S_incline(pTHX_ const char *s)
745 CopLINE_inc(PL_curcop);
748 while (SPACE_OR_TAB(*s))
750 if (strnEQ(s, "line", 4))
754 if (SPACE_OR_TAB(*s))
758 while (SPACE_OR_TAB(*s))
766 while (SPACE_OR_TAB(*s))
768 if (*s == '"' && (t = strchr(s+1, '"'))) {
778 while (SPACE_OR_TAB(*e) || *e == '\r' || *e == '\f')
780 if (*e != '\n' && *e != '\0')
781 return; /* false alarm */
784 const STRLEN len = t - s;
786 const char * const cf = CopFILE(PL_curcop);
787 STRLEN tmplen = cf ? strlen(cf) : 0;
788 if (tmplen > 7 && strnEQ(cf, "(eval ", 6)) {
789 /* must copy *{"::_<(eval N)[oldfilename:L]"}
790 * to *{"::_<newfilename"} */
791 /* However, the long form of evals is only turned on by the
792 debugger - usually they're "(eval %lu)" */
796 STRLEN tmplen2 = len;
797 if (tmplen + 2 <= sizeof smallbuf)
800 Newx(tmpbuf, tmplen + 2, char);
803 memcpy(tmpbuf + 2, cf, tmplen);
805 gvp = (GV**)hv_fetch(PL_defstash, tmpbuf, tmplen, FALSE);
810 if (tmplen2 + 2 <= sizeof smallbuf)
813 Newx(tmpbuf2, tmplen2 + 2, char);
815 if (tmpbuf2 != smallbuf || tmpbuf != smallbuf) {
816 /* Either they malloc'd it, or we malloc'd it,
817 so no prefix is present in ours. */
822 memcpy(tmpbuf2 + 2, s, tmplen2);
825 gv2 = *(GV**)hv_fetch(PL_defstash, tmpbuf2, tmplen2, TRUE);
827 gv_init(gv2, PL_defstash, tmpbuf2, tmplen2, FALSE);
828 /* adjust ${"::_<newfilename"} to store the new file name */
829 GvSV(gv2) = newSVpvn(tmpbuf2 + 2, tmplen2 - 2);
830 GvHV(gv2) = (HV*)SvREFCNT_inc(GvHV(*gvp));
831 GvAV(gv2) = (AV*)SvREFCNT_inc(GvAV(*gvp));
834 if (tmpbuf2 != smallbuf) Safefree(tmpbuf2);
836 if (tmpbuf != smallbuf) Safefree(tmpbuf);
839 CopFILE_free(PL_curcop);
840 CopFILE_setn(PL_curcop, s, len);
842 CopLINE_set(PL_curcop, atoi(n)-1);
846 /* skip space before PL_thistoken */
849 S_skipspace0(pTHX_ register char *s)
856 PL_thiswhite = newSVpvs("");
857 sv_catsv(PL_thiswhite, PL_skipwhite);
858 sv_free(PL_skipwhite);
861 PL_realtokenstart = s - SvPVX(PL_linestr);
865 /* skip space after PL_thistoken */
868 S_skipspace1(pTHX_ register char *s)
870 const char *start = s;
871 I32 startoff = start - SvPVX(PL_linestr);
876 start = SvPVX(PL_linestr) + startoff;
877 if (!PL_thistoken && PL_realtokenstart >= 0) {
878 const char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
879 PL_thistoken = newSVpvn(tstart, start - tstart);
881 PL_realtokenstart = -1;
884 PL_nextwhite = newSVpvs("");
885 sv_catsv(PL_nextwhite, PL_skipwhite);
886 sv_free(PL_skipwhite);
893 S_skipspace2(pTHX_ register char *s, SV **svp)
896 const I32 bufptroff = PL_bufptr - SvPVX(PL_linestr);
897 const I32 startoff = s - SvPVX(PL_linestr);
900 PL_bufptr = SvPVX(PL_linestr) + bufptroff;
901 if (!PL_madskills || !svp)
903 start = SvPVX(PL_linestr) + startoff;
904 if (!PL_thistoken && PL_realtokenstart >= 0) {
905 char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
906 PL_thistoken = newSVpvn(tstart, start - tstart);
907 PL_realtokenstart = -1;
912 sv_setsv(*svp, PL_skipwhite);
913 sv_free(PL_skipwhite);
922 S_update_debugger_info(pTHX_ SV *orig_sv, const char *buf, STRLEN len)
924 AV *av = CopFILEAVx(PL_curcop);
926 SV * const sv = newSV(0);
927 sv_upgrade(sv, SVt_PVMG);
929 sv_setsv(sv, orig_sv);
931 sv_setpvn(sv, buf, len);
934 av_store(av, (I32)CopLINE(PL_curcop), sv);
940 * Called to gobble the appropriate amount and type of whitespace.
941 * Skips comments as well.
945 S_skipspace(pTHX_ register char *s)
950 int startoff = s - SvPVX(PL_linestr);
953 sv_free(PL_skipwhite);
958 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
959 while (s < PL_bufend && SPACE_OR_TAB(*s))
969 SSize_t oldprevlen, oldoldprevlen;
970 SSize_t oldloplen = 0, oldunilen = 0;
971 while (s < PL_bufend && isSPACE(*s)) {
972 if (*s++ == '\n' && PL_in_eval && !PL_rsfp)
977 if (s < PL_bufend && *s == '#') {
978 while (s < PL_bufend && *s != '\n')
982 if (PL_in_eval && !PL_rsfp) {
989 /* only continue to recharge the buffer if we're at the end
990 * of the buffer, we're not reading from a source filter, and
991 * we're in normal lexing mode
993 if (s < PL_bufend || !PL_rsfp || PL_sublex_info.sub_inwhat ||
994 PL_lex_state == LEX_FORMLINE)
1001 /* try to recharge the buffer */
1003 curoff = s - SvPVX(PL_linestr);
1006 if ((s = filter_gets(PL_linestr, PL_rsfp,
1007 (prevlen = SvCUR(PL_linestr)))) == NULL)
1010 if (PL_madskills && curoff != startoff) {
1012 PL_skipwhite = newSVpvs("");
1013 sv_catpvn(PL_skipwhite, SvPVX(PL_linestr) + startoff,
1017 /* mustn't throw out old stuff yet if madpropping */
1018 SvCUR(PL_linestr) = curoff;
1019 s = SvPVX(PL_linestr) + curoff;
1021 if (curoff && s[-1] == '\n')
1025 /* end of file. Add on the -p or -n magic */
1026 /* XXX these shouldn't really be added here, can't set PL_faketokens */
1029 sv_catpv(PL_linestr,
1030 ";}continue{print or die qq(-p destination: $!\\n);}");
1032 sv_setpv(PL_linestr,
1033 ";}continue{print or die qq(-p destination: $!\\n);}");
1035 PL_minus_n = PL_minus_p = 0;
1037 else if (PL_minus_n) {
1039 sv_catpvn(PL_linestr, ";}", 2);
1041 sv_setpvn(PL_linestr, ";}", 2);
1047 sv_catpvn(PL_linestr,";", 1);
1049 sv_setpvn(PL_linestr,";", 1);
1052 /* reset variables for next time we lex */
1053 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = s = PL_linestart
1059 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
1060 PL_last_lop = PL_last_uni = NULL;
1062 /* Close the filehandle. Could be from -P preprocessor,
1063 * STDIN, or a regular file. If we were reading code from
1064 * STDIN (because the commandline held no -e or filename)
1065 * then we don't close it, we reset it so the code can
1066 * read from STDIN too.
1069 if (PL_preprocess && !PL_in_eval)
1070 (void)PerlProc_pclose(PL_rsfp);
1071 else if ((PerlIO*)PL_rsfp == PerlIO_stdin())
1072 PerlIO_clearerr(PL_rsfp);
1074 (void)PerlIO_close(PL_rsfp);
1079 /* not at end of file, so we only read another line */
1080 /* make corresponding updates to old pointers, for yyerror() */
1081 oldprevlen = PL_oldbufptr - PL_bufend;
1082 oldoldprevlen = PL_oldoldbufptr - PL_bufend;
1084 oldunilen = PL_last_uni - PL_bufend;
1086 oldloplen = PL_last_lop - PL_bufend;
1087 PL_linestart = PL_bufptr = s + prevlen;
1088 PL_bufend = s + SvCUR(PL_linestr);
1090 PL_oldbufptr = s + oldprevlen;
1091 PL_oldoldbufptr = s + oldoldprevlen;
1093 PL_last_uni = s + oldunilen;
1095 PL_last_lop = s + oldloplen;
1098 /* debugger active and we're not compiling the debugger code,
1099 * so store the line into the debugger's array of lines
1101 if (PERLDB_LINE && PL_curstash != PL_debstash)
1102 update_debugger_info(NULL, PL_bufptr, PL_bufend - PL_bufptr);
1109 PL_skipwhite = newSVpvs("");
1110 curoff = s - SvPVX(PL_linestr);
1111 if (curoff - startoff)
1112 sv_catpvn(PL_skipwhite, SvPVX(PL_linestr) + startoff,
1121 * Check the unary operators to ensure there's no ambiguity in how they're
1122 * used. An ambiguous piece of code would be:
1124 * This doesn't mean rand() + 5. Because rand() is a unary operator,
1125 * the +5 is its argument.
1135 if (PL_oldoldbufptr != PL_last_uni)
1137 while (isSPACE(*PL_last_uni))
1140 while (isALNUM_lazy_if(s,UTF) || *s == '-')
1142 if ((t = strchr(s, '(')) && t < PL_bufptr)
1145 if (ckWARN_d(WARN_AMBIGUOUS)){
1146 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
1147 "Warning: Use of \"%.*s\" without parentheses is ambiguous",
1148 (int)(s - PL_last_uni), PL_last_uni);
1153 * LOP : macro to build a list operator. Its behaviour has been replaced
1154 * with a subroutine, S_lop() for which LOP is just another name.
1157 #define LOP(f,x) return lop(f,x,s)
1161 * Build a list operator (or something that might be one). The rules:
1162 * - if we have a next token, then it's a list operator [why?]
1163 * - if the next thing is an opening paren, then it's a function
1164 * - else it's a list operator
1168 S_lop(pTHX_ I32 f, int x, char *s)
1175 PL_last_lop = PL_oldbufptr;
1176 PL_last_lop_op = (OPCODE)f;
1179 return REPORT(LSTOP);
1182 return REPORT(LSTOP);
1185 return REPORT(FUNC);
1188 return REPORT(FUNC);
1190 return REPORT(LSTOP);
1196 * Sets up for an eventual force_next(). start_force(0) basically does
1197 * an unshift, while start_force(-1) does a push. yylex removes items
1202 S_start_force(pTHX_ int where)
1206 if (where < 0) /* so people can duplicate start_force(PL_curforce) */
1207 where = PL_lasttoke;
1208 assert(PL_curforce < 0 || PL_curforce == where);
1209 if (PL_curforce != where) {
1210 for (i = PL_lasttoke; i > where; --i) {
1211 PL_nexttoke[i] = PL_nexttoke[i-1];
1215 if (PL_curforce < 0) /* in case of duplicate start_force() */
1216 Zero(&PL_nexttoke[where], 1, NEXTTOKE);
1217 PL_curforce = where;
1220 curmad('^', newSVpvs(""));
1221 CURMAD('_', PL_nextwhite);
1226 S_curmad(pTHX_ char slot, SV *sv)
1232 if (PL_curforce < 0)
1233 where = &PL_thismad;
1235 where = &PL_nexttoke[PL_curforce].next_mad;
1238 sv_setpvn(sv, "", 0);
1241 if (UTF && is_utf8_string((U8*)SvPVX(sv), SvCUR(sv)))
1243 else if (PL_encoding) {
1244 sv_recode_to_utf8(sv, PL_encoding);
1249 /* keep a slot open for the head of the list? */
1250 if (slot != '_' && *where && (*where)->mad_key == '^') {
1251 (*where)->mad_key = slot;
1252 sv_free((*where)->mad_val);
1253 (*where)->mad_val = (void*)sv;
1256 addmad(newMADsv(slot, sv), where, 0);
1259 # define start_force(where) NOOP
1260 # define curmad(slot, sv) NOOP
1265 * When the lexer realizes it knows the next token (for instance,
1266 * it is reordering tokens for the parser) then it can call S_force_next
1267 * to know what token to return the next time the lexer is called. Caller
1268 * will need to set PL_nextval[] (or PL_nexttoke[].next_val with PERL_MAD),
1269 * and possibly PL_expect to ensure the lexer handles the token correctly.
1273 S_force_next(pTHX_ I32 type)
1277 if (PL_curforce < 0)
1278 start_force(PL_lasttoke);
1279 PL_nexttoke[PL_curforce].next_type = type;
1280 if (PL_lex_state != LEX_KNOWNEXT)
1281 PL_lex_defer = PL_lex_state;
1282 PL_lex_state = LEX_KNOWNEXT;
1283 PL_lex_expect = PL_expect;
1286 PL_nexttype[PL_nexttoke] = type;
1288 if (PL_lex_state != LEX_KNOWNEXT) {
1289 PL_lex_defer = PL_lex_state;
1290 PL_lex_expect = PL_expect;
1291 PL_lex_state = LEX_KNOWNEXT;
1297 S_newSV_maybe_utf8(pTHX_ const char *start, STRLEN len)
1300 SV * const sv = newSVpvn(start,len);
1301 if (UTF && !IN_BYTES && is_utf8_string((const U8*)start, len))
1308 * When the lexer knows the next thing is a word (for instance, it has
1309 * just seen -> and it knows that the next char is a word char, then
1310 * it calls S_force_word to stick the next word into the PL_nexttoke/val
1314 * char *start : buffer position (must be within PL_linestr)
1315 * int token : PL_next* will be this type of bare word (e.g., METHOD,WORD)
1316 * int check_keyword : if true, Perl checks to make sure the word isn't
1317 * a keyword (do this if the word is a label, e.g. goto FOO)
1318 * int allow_pack : if true, : characters will also be allowed (require,
1319 * use, etc. do this)
1320 * int allow_initial_tick : used by the "sub" lexer only.
1324 S_force_word(pTHX_ register char *start, int token, int check_keyword, int allow_pack, int allow_initial_tick)
1330 start = SKIPSPACE1(start);
1332 if (isIDFIRST_lazy_if(s,UTF) ||
1333 (allow_pack && *s == ':') ||
1334 (allow_initial_tick && *s == '\'') )
1336 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, allow_pack, &len);
1337 if (check_keyword && keyword(PL_tokenbuf, len, 0))
1339 start_force(PL_curforce);
1341 curmad('X', newSVpvn(start,s-start));
1342 if (token == METHOD) {
1347 PL_expect = XOPERATOR;
1350 NEXTVAL_NEXTTOKE.opval
1351 = (OP*)newSVOP(OP_CONST,0,
1352 S_newSV_maybe_utf8(aTHX_ PL_tokenbuf, len));
1353 NEXTVAL_NEXTTOKE.opval->op_private |= OPpCONST_BARE;
1361 * Called when the lexer wants $foo *foo &foo etc, but the program
1362 * text only contains the "foo" portion. The first argument is a pointer
1363 * to the "foo", and the second argument is the type symbol to prefix.
1364 * Forces the next token to be a "WORD".
1365 * Creates the symbol if it didn't already exist (via gv_fetchpv()).
1369 S_force_ident(pTHX_ register const char *s, int kind)
1373 const STRLEN len = strlen(s);
1374 OP* const o = (OP*)newSVOP(OP_CONST, 0, newSVpvn(s, len));
1375 start_force(PL_curforce);
1376 NEXTVAL_NEXTTOKE.opval = o;
1379 o->op_private = OPpCONST_ENTERED;
1380 /* XXX see note in pp_entereval() for why we forgo typo
1381 warnings if the symbol must be introduced in an eval.
1383 gv_fetchpvn_flags(s, len,
1384 PL_in_eval ? (GV_ADDMULTI | GV_ADDINEVAL)
1386 kind == '$' ? SVt_PV :
1387 kind == '@' ? SVt_PVAV :
1388 kind == '%' ? SVt_PVHV :
1396 Perl_str_to_version(pTHX_ SV *sv)
1401 const char *start = SvPV_const(sv,len);
1402 const char * const end = start + len;
1403 const bool utf = SvUTF8(sv) ? TRUE : FALSE;
1404 while (start < end) {
1408 n = utf8n_to_uvchr((U8*)start, len, &skip, 0);
1413 retval += ((NV)n)/nshift;
1422 * Forces the next token to be a version number.
1423 * If the next token appears to be an invalid version number, (e.g. "v2b"),
1424 * and if "guessing" is TRUE, then no new token is created (and the caller
1425 * must use an alternative parsing method).
1429 S_force_version(pTHX_ char *s, int guessing)
1435 I32 startoff = s - SvPVX(PL_linestr);
1444 while (isDIGIT(*d) || *d == '_' || *d == '.')
1448 start_force(PL_curforce);
1449 curmad('X', newSVpvn(s,d-s));
1452 if (*d == ';' || isSPACE(*d) || *d == '}' || !*d) {
1454 s = scan_num(s, &yylval);
1455 version = yylval.opval;
1456 ver = cSVOPx(version)->op_sv;
1457 if (SvPOK(ver) && !SvNIOK(ver)) {
1458 SvUPGRADE(ver, SVt_PVNV);
1459 SvNV_set(ver, str_to_version(ver));
1460 SvNOK_on(ver); /* hint that it is a version */
1463 else if (guessing) {
1466 sv_free(PL_nextwhite); /* let next token collect whitespace */
1468 s = SvPVX(PL_linestr) + startoff;
1476 if (PL_madskills && !version) {
1477 sv_free(PL_nextwhite); /* let next token collect whitespace */
1479 s = SvPVX(PL_linestr) + startoff;
1482 /* NOTE: The parser sees the package name and the VERSION swapped */
1483 start_force(PL_curforce);
1484 NEXTVAL_NEXTTOKE.opval = version;
1492 * Tokenize a quoted string passed in as an SV. It finds the next
1493 * chunk, up to end of string or a backslash. It may make a new
1494 * SV containing that chunk (if HINT_NEW_STRING is on). It also
1499 S_tokeq(pTHX_ SV *sv)
1503 register char *send;
1511 s = SvPV_force(sv, len);
1512 if (SvTYPE(sv) >= SVt_PVIV && SvIVX(sv) == -1)
1515 while (s < send && *s != '\\')
1520 if ( PL_hints & HINT_NEW_STRING ) {
1521 pv = sv_2mortal(newSVpvn(SvPVX_const(pv), len));
1527 if (s + 1 < send && (s[1] == '\\'))
1528 s++; /* all that, just for this */
1533 SvCUR_set(sv, d - SvPVX_const(sv));
1535 if ( PL_hints & HINT_NEW_STRING )
1536 return new_constant(NULL, 0, "q", sv, pv, "q");
1541 * Now come three functions related to double-quote context,
1542 * S_sublex_start, S_sublex_push, and S_sublex_done. They're used when
1543 * converting things like "\u\Lgnat" into ucfirst(lc("gnat")). They
1544 * interact with PL_lex_state, and create fake ( ... ) argument lists
1545 * to handle functions and concatenation.
1546 * They assume that whoever calls them will be setting up a fake
1547 * join call, because each subthing puts a ',' after it. This lets
1550 * join($, , 'lower ', lcfirst( 'uPpEr', ) ,)
1552 * (I'm not sure whether the spurious commas at the end of lcfirst's
1553 * arguments and join's arguments are created or not).
1558 * Assumes that yylval.ival is the op we're creating (e.g. OP_LCFIRST).
1560 * Pattern matching will set PL_lex_op to the pattern-matching op to
1561 * make (we return THING if yylval.ival is OP_NULL, PMFUNC otherwise).
1563 * OP_CONST and OP_READLINE are easy--just make the new op and return.
1565 * Everything else becomes a FUNC.
1567 * Sets PL_lex_state to LEX_INTERPPUSH unless (ival was OP_NULL or we
1568 * had an OP_CONST or OP_READLINE). This just sets us up for a
1569 * call to S_sublex_push().
1573 S_sublex_start(pTHX)
1576 register const I32 op_type = yylval.ival;
1578 if (op_type == OP_NULL) {
1579 yylval.opval = PL_lex_op;
1583 if (op_type == OP_CONST || op_type == OP_READLINE) {
1584 SV *sv = tokeq(PL_lex_stuff);
1586 if (SvTYPE(sv) == SVt_PVIV) {
1587 /* Overloaded constants, nothing fancy: Convert to SVt_PV: */
1589 const char * const p = SvPV_const(sv, len);
1590 SV * const nsv = newSVpvn(p, len);
1596 yylval.opval = (OP*)newSVOP(op_type, 0, sv);
1597 PL_lex_stuff = NULL;
1598 /* Allow <FH> // "foo" */
1599 if (op_type == OP_READLINE)
1600 PL_expect = XTERMORDORDOR;
1603 else if (op_type == OP_BACKTICK && PL_lex_op) {
1604 /* readpipe() vas overriden */
1605 cSVOPx(cLISTOPx(cUNOPx(PL_lex_op)->op_first)->op_first->op_sibling)->op_sv = tokeq(PL_lex_stuff);
1606 yylval.opval = PL_lex_op;
1608 PL_lex_stuff = NULL;
1612 PL_sublex_info.super_state = PL_lex_state;
1613 PL_sublex_info.sub_inwhat = op_type;
1614 PL_sublex_info.sub_op = PL_lex_op;
1615 PL_lex_state = LEX_INTERPPUSH;
1619 yylval.opval = PL_lex_op;
1629 * Create a new scope to save the lexing state. The scope will be
1630 * ended in S_sublex_done. Returns a '(', starting the function arguments
1631 * to the uc, lc, etc. found before.
1632 * Sets PL_lex_state to LEX_INTERPCONCAT.
1641 PL_lex_state = PL_sublex_info.super_state;
1642 SAVEI32(PL_lex_dojoin);
1643 SAVEI32(PL_lex_brackets);
1644 SAVEI32(PL_lex_casemods);
1645 SAVEI32(PL_lex_starts);
1646 SAVEI32(PL_lex_state);
1647 SAVEVPTR(PL_lex_inpat);
1648 SAVEI32(PL_lex_inwhat);
1649 SAVECOPLINE(PL_curcop);
1650 SAVEPPTR(PL_bufptr);
1651 SAVEPPTR(PL_bufend);
1652 SAVEPPTR(PL_oldbufptr);
1653 SAVEPPTR(PL_oldoldbufptr);
1654 SAVEPPTR(PL_last_lop);
1655 SAVEPPTR(PL_last_uni);
1656 SAVEPPTR(PL_linestart);
1657 SAVESPTR(PL_linestr);
1658 SAVEGENERICPV(PL_lex_brackstack);
1659 SAVEGENERICPV(PL_lex_casestack);
1661 PL_linestr = PL_lex_stuff;
1662 PL_lex_stuff = NULL;
1664 PL_bufend = PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart
1665 = SvPVX(PL_linestr);
1666 PL_bufend += SvCUR(PL_linestr);
1667 PL_last_lop = PL_last_uni = NULL;
1668 SAVEFREESV(PL_linestr);
1670 PL_lex_dojoin = FALSE;
1671 PL_lex_brackets = 0;
1672 Newx(PL_lex_brackstack, 120, char);
1673 Newx(PL_lex_casestack, 12, char);
1674 PL_lex_casemods = 0;
1675 *PL_lex_casestack = '\0';
1677 PL_lex_state = LEX_INTERPCONCAT;
1678 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
1680 PL_lex_inwhat = PL_sublex_info.sub_inwhat;
1681 if (PL_lex_inwhat == OP_MATCH || PL_lex_inwhat == OP_QR || PL_lex_inwhat == OP_SUBST)
1682 PL_lex_inpat = PL_sublex_info.sub_op;
1684 PL_lex_inpat = NULL;
1691 * Restores lexer state after a S_sublex_push.
1698 if (!PL_lex_starts++) {
1699 SV * const sv = newSVpvs("");
1700 if (SvUTF8(PL_linestr))
1702 PL_expect = XOPERATOR;
1703 yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
1707 if (PL_lex_casemods) { /* oops, we've got some unbalanced parens */
1708 PL_lex_state = LEX_INTERPCASEMOD;
1712 /* Is there a right-hand side to take care of? (s//RHS/ or tr//RHS/) */
1713 if (PL_lex_repl && (PL_lex_inwhat == OP_SUBST || PL_lex_inwhat == OP_TRANS)) {
1714 PL_linestr = PL_lex_repl;
1716 PL_bufend = PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart = SvPVX(PL_linestr);
1717 PL_bufend += SvCUR(PL_linestr);
1718 PL_last_lop = PL_last_uni = NULL;
1719 SAVEFREESV(PL_linestr);
1720 PL_lex_dojoin = FALSE;
1721 PL_lex_brackets = 0;
1722 PL_lex_casemods = 0;
1723 *PL_lex_casestack = '\0';
1725 if (SvEVALED(PL_lex_repl)) {
1726 PL_lex_state = LEX_INTERPNORMAL;
1728 /* we don't clear PL_lex_repl here, so that we can check later
1729 whether this is an evalled subst; that means we rely on the
1730 logic to ensure sublex_done() is called again only via the
1731 branch (in yylex()) that clears PL_lex_repl, else we'll loop */
1734 PL_lex_state = LEX_INTERPCONCAT;
1744 PL_endwhite = newSVpvs("");
1745 sv_catsv(PL_endwhite, PL_thiswhite);
1749 sv_setpvn(PL_thistoken,"",0);
1751 PL_realtokenstart = -1;
1755 PL_bufend = SvPVX(PL_linestr);
1756 PL_bufend += SvCUR(PL_linestr);
1757 PL_expect = XOPERATOR;
1758 PL_sublex_info.sub_inwhat = 0;
1766 Extracts a pattern, double-quoted string, or transliteration. This
1769 It looks at PL_lex_inwhat and PL_lex_inpat to find out whether it's
1770 processing a pattern (PL_lex_inpat is true), a transliteration
1771 (PL_lex_inwhat == OP_TRANS is true), or a double-quoted string.
1773 Returns a pointer to the character scanned up to. If this is
1774 advanced from the start pointer supplied (i.e. if anything was
1775 successfully parsed), will leave an OP for the substring scanned
1776 in yylval. Caller must intuit reason for not parsing further
1777 by looking at the next characters herself.
1781 double-quoted style: \r and \n
1782 regexp special ones: \D \s
1785 case and quoting: \U \Q \E
1786 stops on @ and $, but not for $ as tail anchor
1788 In transliterations:
1789 characters are VERY literal, except for - not at the start or end
1790 of the string, which indicates a range. If the range is in bytes,
1791 scan_const expands the range to the full set of intermediate
1792 characters. If the range is in utf8, the hyphen is replaced with
1793 a certain range mark which will be handled by pmtrans() in op.c.
1795 In double-quoted strings:
1797 double-quoted style: \r and \n
1799 deprecated backrefs: \1 (in substitution replacements)
1800 case and quoting: \U \Q \E
1803 scan_const does *not* construct ops to handle interpolated strings.
1804 It stops processing as soon as it finds an embedded $ or @ variable
1805 and leaves it to the caller to work out what's going on.
1807 embedded arrays (whether in pattern or not) could be:
1808 @foo, @::foo, @'foo, @{foo}, @$foo, @+, @-.
1810 $ in double-quoted strings must be the symbol of an embedded scalar.
1812 $ in pattern could be $foo or could be tail anchor. Assumption:
1813 it's a tail anchor if $ is the last thing in the string, or if it's
1814 followed by one of "()| \r\n\t"
1816 \1 (backreferences) are turned into $1
1818 The structure of the code is
1819 while (there's a character to process) {
1820 handle transliteration ranges
1821 skip regexp comments /(?#comment)/ and codes /(?{code})/
1822 skip #-initiated comments in //x patterns
1823 check for embedded arrays
1824 check for embedded scalars
1826 leave intact backslashes from leaveit (below)
1827 deprecate \1 in substitution replacements
1828 handle string-changing backslashes \l \U \Q \E, etc.
1829 switch (what was escaped) {
1830 handle \- in a transliteration (becomes a literal -)
1831 handle \132 (octal characters)
1832 handle \x15 and \x{1234} (hex characters)
1833 handle \N{name} (named characters)
1834 handle \cV (control characters)
1835 handle printf-style backslashes (\f, \r, \n, etc)
1837 } (end if backslash)
1838 } (end while character to read)
1843 S_scan_const(pTHX_ char *start)
1846 register char *send = PL_bufend; /* end of the constant */
1847 SV *sv = newSV(send - start); /* sv for the constant */
1848 register char *s = start; /* start of the constant */
1849 register char *d = SvPVX(sv); /* destination for copies */
1850 bool dorange = FALSE; /* are we in a translit range? */
1851 bool didrange = FALSE; /* did we just finish a range? */
1852 I32 has_utf8 = FALSE; /* Output constant is UTF8 */
1853 I32 this_utf8 = UTF; /* The source string is assumed to be UTF8 */
1856 UV literal_endpoint = 0;
1857 bool native_range = TRUE; /* turned to FALSE if the first endpoint is Unicode. */
1860 if (PL_lex_inwhat == OP_TRANS && PL_sublex_info.sub_op) {
1861 /* If we are doing a trans and we know we want UTF8 set expectation */
1862 has_utf8 = PL_sublex_info.sub_op->op_private & (OPpTRANS_FROM_UTF|OPpTRANS_TO_UTF);
1863 this_utf8 = PL_sublex_info.sub_op->op_private & (PL_lex_repl ? OPpTRANS_FROM_UTF : OPpTRANS_TO_UTF);
1867 while (s < send || dorange) {
1868 /* get transliterations out of the way (they're most literal) */
1869 if (PL_lex_inwhat == OP_TRANS) {
1870 /* expand a range A-Z to the full set of characters. AIE! */
1872 I32 i; /* current expanded character */
1873 I32 min; /* first character in range */
1874 I32 max; /* last character in range */
1885 char * const c = (char*)utf8_hop((U8*)d, -1);
1889 *c = (char)UTF_TO_NATIVE(0xff);
1890 /* mark the range as done, and continue */
1896 i = d - SvPVX_const(sv); /* remember current offset */
1899 SvLEN(sv) + (has_utf8 ?
1900 (512 - UTF_CONTINUATION_MARK +
1903 /* How many two-byte within 0..255: 128 in UTF-8,
1904 * 96 in UTF-8-mod. */
1906 SvGROW(sv, SvLEN(sv) + 256); /* never more than 256 chars in a range */
1908 d = SvPVX(sv) + i; /* refresh d after realloc */
1912 for (j = 0; j <= 1; j++) {
1913 char * const c = (char*)utf8_hop((U8*)d, -1);
1914 const UV uv = utf8n_to_uvchr((U8*)c, d - c, NULL, 0);
1920 max = (U8)0xff; /* only to \xff */
1921 uvmax = uv; /* \x{100} to uvmax */
1923 d = c; /* eat endpoint chars */
1928 d -= 2; /* eat the first char and the - */
1929 min = (U8)*d; /* first char in range */
1930 max = (U8)d[1]; /* last char in range */
1937 "Invalid range \"%c-%c\" in transliteration operator",
1938 (char)min, (char)max);
1942 if (literal_endpoint == 2 &&
1943 ((isLOWER(min) && isLOWER(max)) ||
1944 (isUPPER(min) && isUPPER(max)))) {
1946 for (i = min; i <= max; i++)
1948 *d++ = NATIVE_TO_NEED(has_utf8,i);
1950 for (i = min; i <= max; i++)
1952 *d++ = NATIVE_TO_NEED(has_utf8,i);
1957 for (i = min; i <= max; i++)
1960 const U8 ch = (U8)NATIVE_TO_UTF(i);
1961 if (UNI_IS_INVARIANT(ch))
1964 *d++ = (U8)UTF8_EIGHT_BIT_HI(ch);
1965 *d++ = (U8)UTF8_EIGHT_BIT_LO(ch);
1974 d = (char*)uvchr_to_utf8((U8*)d, 0x100);
1976 *d++ = (char)UTF_TO_NATIVE(0xff);
1978 d = (char*)uvchr_to_utf8((U8*)d, uvmax);
1982 /* mark the range as done, and continue */
1986 literal_endpoint = 0;
1991 /* range begins (ignore - as first or last char) */
1992 else if (*s == '-' && s+1 < send && s != start) {
1994 Perl_croak(aTHX_ "Ambiguous range in transliteration operator");
2001 *d++ = (char)UTF_TO_NATIVE(0xff); /* use illegal utf8 byte--see pmtrans */
2011 literal_endpoint = 0;
2012 native_range = TRUE;
2017 /* if we get here, we're not doing a transliteration */
2019 /* skip for regexp comments /(?#comment)/ and code /(?{code})/,
2020 except for the last char, which will be done separately. */
2021 else if (*s == '(' && PL_lex_inpat && s[1] == '?') {
2023 while (s+1 < send && *s != ')')
2024 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2026 else if (s[2] == '{' /* This should match regcomp.c */
2027 || ((s[2] == 'p' || s[2] == '?') && s[3] == '{'))
2030 char *regparse = s + (s[2] == '{' ? 3 : 4);
2033 while (count && (c = *regparse)) {
2034 if (c == '\\' && regparse[1])
2042 if (*regparse != ')')
2043 regparse--; /* Leave one char for continuation. */
2044 while (s < regparse)
2045 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2049 /* likewise skip #-initiated comments in //x patterns */
2050 else if (*s == '#' && PL_lex_inpat &&
2051 ((PMOP*)PL_lex_inpat)->op_pmflags & PMf_EXTENDED) {
2052 while (s+1 < send && *s != '\n')
2053 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2056 /* check for embedded arrays
2057 (@foo, @::foo, @'foo, @{foo}, @$foo, @+, @-)
2059 else if (*s == '@' && s[1]) {
2060 if (isALNUM_lazy_if(s+1,UTF))
2062 if (strchr(":'{$", s[1]))
2064 if (!PL_lex_inpat && (s[1] == '+' || s[1] == '-'))
2065 break; /* in regexp, neither @+ nor @- are interpolated */
2068 /* check for embedded scalars. only stop if we're sure it's a
2071 else if (*s == '$') {
2072 if (!PL_lex_inpat) /* not a regexp, so $ must be var */
2074 if (s + 1 < send && !strchr("()| \r\n\t", s[1]))
2075 break; /* in regexp, $ might be tail anchor */
2078 /* End of else if chain - OP_TRANS rejoin rest */
2081 if (*s == '\\' && s+1 < send) {
2084 /* deprecate \1 in strings and substitution replacements */
2085 if (PL_lex_inwhat == OP_SUBST && !PL_lex_inpat &&
2086 isDIGIT(*s) && *s != '0' && !isDIGIT(s[1]))
2088 if (ckWARN(WARN_SYNTAX))
2089 Perl_warner(aTHX_ packWARN(WARN_SYNTAX), "\\%c better written as $%c", *s, *s);
2094 /* string-change backslash escapes */
2095 if (PL_lex_inwhat != OP_TRANS && *s && strchr("lLuUEQ", *s)) {
2099 /* skip any other backslash escapes in a pattern */
2100 else if (PL_lex_inpat) {
2101 *d++ = NATIVE_TO_NEED(has_utf8,'\\');
2102 goto default_action;
2105 /* if we get here, it's either a quoted -, or a digit */
2108 /* quoted - in transliterations */
2110 if (PL_lex_inwhat == OP_TRANS) {
2117 if ((isALPHA(*s) || isDIGIT(*s)) &&
2119 Perl_warner(aTHX_ packWARN(WARN_MISC),
2120 "Unrecognized escape \\%c passed through",
2122 /* default action is to copy the quoted character */
2123 goto default_action;
2126 /* \132 indicates an octal constant */
2127 case '0': case '1': case '2': case '3':
2128 case '4': case '5': case '6': case '7':
2132 uv = grok_oct(s, &len, &flags, NULL);
2135 goto NUM_ESCAPE_INSERT;
2137 /* \x24 indicates a hex constant */
2141 char* const e = strchr(s, '}');
2142 I32 flags = PERL_SCAN_ALLOW_UNDERSCORES |
2143 PERL_SCAN_DISALLOW_PREFIX;
2148 yyerror("Missing right brace on \\x{}");
2152 uv = grok_hex(s, &len, &flags, NULL);
2158 I32 flags = PERL_SCAN_DISALLOW_PREFIX;
2159 uv = grok_hex(s, &len, &flags, NULL);
2165 /* Insert oct or hex escaped character.
2166 * There will always enough room in sv since such
2167 * escapes will be longer than any UTF-8 sequence
2168 * they can end up as. */
2170 /* We need to map to chars to ASCII before doing the tests
2173 if (!UNI_IS_INVARIANT(NATIVE_TO_UNI(uv))) {
2174 if (!has_utf8 && uv > 255) {
2175 /* Might need to recode whatever we have
2176 * accumulated so far if it contains any
2179 * (Can't we keep track of that and avoid
2180 * this rescan? --jhi)
2184 for (c = (U8 *) SvPVX(sv); c < (U8 *)d; c++) {
2185 if (!NATIVE_IS_INVARIANT(*c)) {
2190 const STRLEN offset = d - SvPVX_const(sv);
2192 d = SvGROW(sv, SvLEN(sv) + hicount + 1) + offset;
2196 while (src >= (const U8 *)SvPVX_const(sv)) {
2197 if (!NATIVE_IS_INVARIANT(*src)) {
2198 const U8 ch = NATIVE_TO_ASCII(*src);
2199 *dst-- = (U8)UTF8_EIGHT_BIT_LO(ch);
2200 *dst-- = (U8)UTF8_EIGHT_BIT_HI(ch);
2210 if (has_utf8 || uv > 255) {
2211 d = (char*)uvchr_to_utf8((U8*)d, uv);
2213 if (PL_lex_inwhat == OP_TRANS &&
2214 PL_sublex_info.sub_op) {
2215 PL_sublex_info.sub_op->op_private |=
2216 (PL_lex_repl ? OPpTRANS_FROM_UTF
2220 if (uv > 255 && !dorange)
2221 native_range = FALSE;
2233 /* \N{LATIN SMALL LETTER A} is a named character */
2237 char* e = strchr(s, '}');
2244 yyerror("Missing right brace on \\N{}");
2248 if (e > s + 2 && s[1] == 'U' && s[2] == '+') {
2250 I32 flags = PERL_SCAN_ALLOW_UNDERSCORES |
2251 PERL_SCAN_DISALLOW_PREFIX;
2254 uv = grok_hex(s, &len, &flags, NULL);
2255 if ( e > s && len != (STRLEN)(e - s) ) {
2259 goto NUM_ESCAPE_INSERT;
2261 res = newSVpvn(s + 1, e - s - 1);
2262 type = newSVpvn(s - 2,e - s + 3);
2263 res = new_constant( NULL, 0, "charnames",
2264 res, NULL, SvPVX(type) );
2267 sv_utf8_upgrade(res);
2268 str = SvPV_const(res,len);
2269 #ifdef EBCDIC_NEVER_MIND
2270 /* charnames uses pack U and that has been
2271 * recently changed to do the below uni->native
2272 * mapping, so this would be redundant (and wrong,
2273 * the code point would be doubly converted).
2274 * But leave this in just in case the pack U change
2275 * gets revoked, but the semantics is still
2276 * desireable for charnames. --jhi */
2278 UV uv = utf8_to_uvchr((const U8*)str, 0);
2281 U8 tmpbuf[UTF8_MAXBYTES+1], *d;
2283 d = uvchr_to_utf8(tmpbuf, UNI_TO_NATIVE(uv));
2284 sv_setpvn(res, (char *)tmpbuf, d - tmpbuf);
2285 str = SvPV_const(res, len);
2289 if (!has_utf8 && SvUTF8(res)) {
2290 const char * const ostart = SvPVX_const(sv);
2291 SvCUR_set(sv, d - ostart);
2294 sv_utf8_upgrade(sv);
2295 /* this just broke our allocation above... */
2296 SvGROW(sv, (STRLEN)(send - start));
2297 d = SvPVX(sv) + SvCUR(sv);
2300 if (len > (STRLEN)(e - s + 4)) { /* I _guess_ 4 is \N{} --jhi */
2301 const char * const odest = SvPVX_const(sv);
2303 SvGROW(sv, (SvLEN(sv) + len - (e - s + 4)));
2304 d = SvPVX(sv) + (d - odest);
2308 native_range = FALSE; /* \N{} is guessed to be Unicode */
2310 Copy(str, d, len, char);
2317 yyerror("Missing braces on \\N{}");
2320 /* \c is a control character */
2329 *d++ = NATIVE_TO_NEED(has_utf8,toCTRL(c));
2332 yyerror("Missing control char name in \\c");
2336 /* printf-style backslashes, formfeeds, newlines, etc */
2338 *d++ = NATIVE_TO_NEED(has_utf8,'\b');
2341 *d++ = NATIVE_TO_NEED(has_utf8,'\n');
2344 *d++ = NATIVE_TO_NEED(has_utf8,'\r');
2347 *d++ = NATIVE_TO_NEED(has_utf8,'\f');
2350 *d++ = NATIVE_TO_NEED(has_utf8,'\t');
2353 *d++ = ASCII_TO_NEED(has_utf8,'\033');
2356 *d++ = ASCII_TO_NEED(has_utf8,'\007');
2362 } /* end if (backslash) */
2369 /* If we started with encoded form, or already know we want it
2370 and then encode the next character */
2371 if ((has_utf8 || this_utf8) && !NATIVE_IS_INVARIANT((U8)(*s))) {
2373 const UV nextuv = (this_utf8) ? utf8n_to_uvchr((U8*)s, send - s, &len, 0) : (UV) ((U8) *s);
2374 const STRLEN need = UNISKIP(NATIVE_TO_UNI(nextuv));
2377 /* encoded value larger than old, need extra space (NOTE: SvCUR() not set here) */
2378 const STRLEN off = d - SvPVX_const(sv);
2379 d = SvGROW(sv, SvLEN(sv) + (need-len)) + off;
2381 d = (char*)uvchr_to_utf8((U8*)d, nextuv);
2384 if (uv > 255 && !dorange)
2385 native_range = FALSE;
2389 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2391 } /* while loop to process each character */
2393 /* terminate the string and set up the sv */
2395 SvCUR_set(sv, d - SvPVX_const(sv));
2396 if (SvCUR(sv) >= SvLEN(sv))
2397 Perl_croak(aTHX_ "panic: constant overflowed allocated space");
2400 if (PL_encoding && !has_utf8) {
2401 sv_recode_to_utf8(sv, PL_encoding);
2407 if (PL_lex_inwhat == OP_TRANS && PL_sublex_info.sub_op) {
2408 PL_sublex_info.sub_op->op_private |=
2409 (PL_lex_repl ? OPpTRANS_FROM_UTF : OPpTRANS_TO_UTF);
2413 /* shrink the sv if we allocated more than we used */
2414 if (SvCUR(sv) + 5 < SvLEN(sv)) {
2415 SvPV_shrink_to_cur(sv);
2418 /* return the substring (via yylval) only if we parsed anything */
2419 if (s > PL_bufptr) {
2420 if ( PL_hints & ( PL_lex_inpat ? HINT_NEW_RE : HINT_NEW_STRING ) )
2421 sv = new_constant(start, s - start,
2422 (const char *)(PL_lex_inpat ? "qr" : "q"),
2425 (( PL_lex_inwhat == OP_TRANS
2427 : ( (PL_lex_inwhat == OP_SUBST && !PL_lex_inpat)
2430 yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
2437 * Returns TRUE if there's more to the expression (e.g., a subscript),
2440 * It deals with "$foo[3]" and /$foo[3]/ and /$foo[0123456789$]+/
2442 * ->[ and ->{ return TRUE
2443 * { and [ outside a pattern are always subscripts, so return TRUE
2444 * if we're outside a pattern and it's not { or [, then return FALSE
2445 * if we're in a pattern and the first char is a {
2446 * {4,5} (any digits around the comma) returns FALSE
2447 * if we're in a pattern and the first char is a [
2449 * [SOMETHING] has a funky algorithm to decide whether it's a
2450 * character class or not. It has to deal with things like
2451 * /$foo[-3]/ and /$foo[$bar]/ as well as /$foo[$\d]+/
2452 * anything else returns TRUE
2455 /* This is the one truly awful dwimmer necessary to conflate C and sed. */
2458 S_intuit_more(pTHX_ register char *s)
2461 if (PL_lex_brackets)
2463 if (*s == '-' && s[1] == '>' && (s[2] == '[' || s[2] == '{'))
2465 if (*s != '{' && *s != '[')
2470 /* In a pattern, so maybe we have {n,m}. */
2487 /* On the other hand, maybe we have a character class */
2490 if (*s == ']' || *s == '^')
2493 /* this is terrifying, and it works */
2494 int weight = 2; /* let's weigh the evidence */
2496 unsigned char un_char = 255, last_un_char;
2497 const char * const send = strchr(s,']');
2498 char tmpbuf[sizeof PL_tokenbuf * 4];
2500 if (!send) /* has to be an expression */
2503 Zero(seen,256,char);
2506 else if (isDIGIT(*s)) {
2508 if (isDIGIT(s[1]) && s[2] == ']')
2514 for (; s < send; s++) {
2515 last_un_char = un_char;
2516 un_char = (unsigned char)*s;
2521 weight -= seen[un_char] * 10;
2522 if (isALNUM_lazy_if(s+1,UTF)) {
2524 scan_ident(s, send, tmpbuf, sizeof tmpbuf, FALSE);
2525 len = (int)strlen(tmpbuf);
2526 if (len > 1 && gv_fetchpvn_flags(tmpbuf, len, 0, SVt_PV))
2531 else if (*s == '$' && s[1] &&
2532 strchr("[#!%*<>()-=",s[1])) {
2533 if (/*{*/ strchr("])} =",s[2]))
2542 if (strchr("wds]",s[1]))
2544 else if (seen[(U8)'\''] || seen[(U8)'"'])
2546 else if (strchr("rnftbxcav",s[1]))
2548 else if (isDIGIT(s[1])) {
2550 while (s[1] && isDIGIT(s[1]))
2560 if (strchr("aA01! ",last_un_char))
2562 if (strchr("zZ79~",s[1]))
2564 if (last_un_char == 255 && (isDIGIT(s[1]) || s[1] == '$'))
2565 weight -= 5; /* cope with negative subscript */
2568 if (!isALNUM(last_un_char)
2569 && !(last_un_char == '$' || last_un_char == '@'
2570 || last_un_char == '&')
2571 && isALPHA(*s) && s[1] && isALPHA(s[1])) {
2576 if (keyword(tmpbuf, d - tmpbuf, 0))
2579 if (un_char == last_un_char + 1)
2581 weight -= seen[un_char];
2586 if (weight >= 0) /* probably a character class */
2596 * Does all the checking to disambiguate
2598 * between foo(bar) and bar->foo. Returns 0 if not a method, otherwise
2599 * FUNCMETH (bar->foo(args)) or METHOD (bar->foo args).
2601 * First argument is the stuff after the first token, e.g. "bar".
2603 * Not a method if bar is a filehandle.
2604 * Not a method if foo is a subroutine prototyped to take a filehandle.
2605 * Not a method if it's really "Foo $bar"
2606 * Method if it's "foo $bar"
2607 * Not a method if it's really "print foo $bar"
2608 * Method if it's really "foo package::" (interpreted as package->foo)
2609 * Not a method if bar is known to be a subroutine ("sub bar; foo bar")
2610 * Not a method if bar is a filehandle or package, but is quoted with
2615 S_intuit_method(pTHX_ char *start, GV *gv, CV *cv)
2618 char *s = start + (*start == '$');
2619 char tmpbuf[sizeof PL_tokenbuf];
2627 if (SvTYPE(gv) == SVt_PVGV && GvIO(gv))
2631 const char *proto = SvPVX_const(cv);
2642 s = scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
2643 /* start is the beginning of the possible filehandle/object,
2644 * and s is the end of it
2645 * tmpbuf is a copy of it
2648 if (*start == '$') {
2649 if (gv || PL_last_lop_op == OP_PRINT || isUPPER(*PL_tokenbuf))
2652 len = start - SvPVX(PL_linestr);
2656 start = SvPVX(PL_linestr) + len;
2660 return *s == '(' ? FUNCMETH : METHOD;
2662 if (!keyword(tmpbuf, len, 0)) {
2663 if (len > 2 && tmpbuf[len - 2] == ':' && tmpbuf[len - 1] == ':') {
2667 soff = s - SvPVX(PL_linestr);
2671 indirgv = gv_fetchpvn_flags(tmpbuf, len, 0, SVt_PVCV);
2672 if (indirgv && GvCVu(indirgv))
2674 /* filehandle or package name makes it a method */
2675 if (!gv || GvIO(indirgv) || gv_stashpvn(tmpbuf, len, 0)) {
2677 soff = s - SvPVX(PL_linestr);
2680 if ((PL_bufend - s) >= 2 && *s == '=' && *(s+1) == '>')
2681 return 0; /* no assumptions -- "=>" quotes bearword */
2683 start_force(PL_curforce);
2684 NEXTVAL_NEXTTOKE.opval = (OP*)newSVOP(OP_CONST, 0,
2685 newSVpvn(tmpbuf,len));
2686 NEXTVAL_NEXTTOKE.opval->op_private = OPpCONST_BARE;
2688 curmad('X', newSVpvn(start,SvPVX(PL_linestr) + soff - start));
2693 PL_bufptr = SvPVX(PL_linestr) + soff; /* restart before space */
2695 return *s == '(' ? FUNCMETH : METHOD;
2703 * Return a string of Perl code to load the debugger. If PERL5DB
2704 * is set, it will return the contents of that, otherwise a
2705 * compile-time require of perl5db.pl.
2713 const char * const pdb = PerlEnv_getenv("PERL5DB");
2717 SETERRNO(0,SS_NORMAL);
2718 return "BEGIN { require 'perl5db.pl' }";
2724 /* Encoded script support. filter_add() effectively inserts a
2725 * 'pre-processing' function into the current source input stream.
2726 * Note that the filter function only applies to the current source file
2727 * (e.g., it will not affect files 'require'd or 'use'd by this one).
2729 * The datasv parameter (which may be NULL) can be used to pass
2730 * private data to this instance of the filter. The filter function
2731 * can recover the SV using the FILTER_DATA macro and use it to
2732 * store private buffers and state information.
2734 * The supplied datasv parameter is upgraded to a PVIO type
2735 * and the IoDIRP/IoANY field is used to store the function pointer,
2736 * and IOf_FAKE_DIRP is enabled on datasv to mark this as such.
2737 * Note that IoTOP_NAME, IoFMT_NAME, IoBOTTOM_NAME, if set for
2738 * private use must be set using malloc'd pointers.
2742 Perl_filter_add(pTHX_ filter_t funcp, SV *datasv)
2748 if (!PL_rsfp_filters)
2749 PL_rsfp_filters = newAV();
2752 SvUPGRADE(datasv, SVt_PVIO);
2753 IoANY(datasv) = FPTR2DPTR(void *, funcp); /* stash funcp into spare field */
2754 IoFLAGS(datasv) |= IOf_FAKE_DIRP;
2755 DEBUG_P(PerlIO_printf(Perl_debug_log, "filter_add func %p (%s)\n",
2756 FPTR2DPTR(void *, IoANY(datasv)),
2757 SvPV_nolen(datasv)));
2758 av_unshift(PL_rsfp_filters, 1);
2759 av_store(PL_rsfp_filters, 0, datasv) ;
2764 /* Delete most recently added instance of this filter function. */
2766 Perl_filter_del(pTHX_ filter_t funcp)
2772 DEBUG_P(PerlIO_printf(Perl_debug_log, "filter_del func %p",
2773 FPTR2DPTR(void*, funcp)));
2775 if (!PL_rsfp_filters || AvFILLp(PL_rsfp_filters)<0)
2777 /* if filter is on top of stack (usual case) just pop it off */
2778 datasv = FILTER_DATA(AvFILLp(PL_rsfp_filters));
2779 if (IoANY(datasv) == FPTR2DPTR(void *, funcp)) {
2780 IoFLAGS(datasv) &= ~IOf_FAKE_DIRP;
2781 IoANY(datasv) = (void *)NULL;
2782 sv_free(av_pop(PL_rsfp_filters));
2786 /* we need to search for the correct entry and clear it */
2787 Perl_die(aTHX_ "filter_del can only delete in reverse order (currently)");
2791 /* Invoke the idxth filter function for the current rsfp. */
2792 /* maxlen 0 = read one text line */
2794 Perl_filter_read(pTHX_ int idx, SV *buf_sv, int maxlen)
2799 /* This API is bad. It should have been using unsigned int for maxlen.
2800 Not sure if we want to change the API, but if not we should sanity
2801 check the value here. */
2802 const unsigned int correct_length
2811 if (!PL_rsfp_filters)
2813 if (idx > AvFILLp(PL_rsfp_filters)) { /* Any more filters? */
2814 /* Provide a default input filter to make life easy. */
2815 /* Note that we append to the line. This is handy. */
2816 DEBUG_P(PerlIO_printf(Perl_debug_log,
2817 "filter_read %d: from rsfp\n", idx));
2818 if (correct_length) {
2821 const int old_len = SvCUR(buf_sv);
2823 /* ensure buf_sv is large enough */
2824 SvGROW(buf_sv, (STRLEN)(old_len + correct_length)) ;
2825 if ((len = PerlIO_read(PL_rsfp, SvPVX(buf_sv) + old_len,
2826 correct_length)) <= 0) {
2827 if (PerlIO_error(PL_rsfp))
2828 return -1; /* error */
2830 return 0 ; /* end of file */
2832 SvCUR_set(buf_sv, old_len + len) ;
2835 if (sv_gets(buf_sv, PL_rsfp, SvCUR(buf_sv)) == NULL) {
2836 if (PerlIO_error(PL_rsfp))
2837 return -1; /* error */
2839 return 0 ; /* end of file */
2842 return SvCUR(buf_sv);
2844 /* Skip this filter slot if filter has been deleted */
2845 if ( (datasv = FILTER_DATA(idx)) == &PL_sv_undef) {
2846 DEBUG_P(PerlIO_printf(Perl_debug_log,
2847 "filter_read %d: skipped (filter deleted)\n",
2849 return FILTER_READ(idx+1, buf_sv, correct_length); /* recurse */
2851 /* Get function pointer hidden within datasv */
2852 funcp = DPTR2FPTR(filter_t, IoANY(datasv));
2853 DEBUG_P(PerlIO_printf(Perl_debug_log,
2854 "filter_read %d: via function %p (%s)\n",
2855 idx, (void*)datasv, SvPV_nolen_const(datasv)));
2856 /* Call function. The function is expected to */
2857 /* call "FILTER_READ(idx+1, buf_sv)" first. */
2858 /* Return: <0:error, =0:eof, >0:not eof */
2859 return (*funcp)(aTHX_ idx, buf_sv, correct_length);
2863 S_filter_gets(pTHX_ register SV *sv, register PerlIO *fp, STRLEN append)
2866 #ifdef PERL_CR_FILTER
2867 if (!PL_rsfp_filters) {
2868 filter_add(S_cr_textfilter,NULL);
2871 if (PL_rsfp_filters) {
2873 SvCUR_set(sv, 0); /* start with empty line */
2874 if (FILTER_READ(0, sv, 0) > 0)
2875 return ( SvPVX(sv) ) ;
2880 return (sv_gets(sv, fp, append));
2884 S_find_in_my_stash(pTHX_ const char *pkgname, I32 len)
2889 if (len == 11 && *pkgname == '_' && strEQ(pkgname, "__PACKAGE__"))
2893 (pkgname[len - 2] == ':' && pkgname[len - 1] == ':') &&
2894 (gv = gv_fetchpvn_flags(pkgname, len, 0, SVt_PVHV)))
2896 return GvHV(gv); /* Foo:: */
2899 /* use constant CLASS => 'MyClass' */
2900 gv = gv_fetchpvn_flags(pkgname, len, 0, SVt_PVCV);
2901 if (gv && GvCV(gv)) {
2902 SV * const sv = cv_const_sv(GvCV(gv));
2904 pkgname = SvPV_nolen_const(sv);
2907 return gv_stashpv(pkgname, 0);
2911 * S_readpipe_override
2912 * Check whether readpipe() is overriden, and generates the appropriate
2913 * optree, provided sublex_start() is called afterwards.
2916 S_readpipe_override(pTHX)
2919 GV *gv_readpipe = gv_fetchpvs("readpipe", GV_NOTQUAL, SVt_PVCV);
2920 yylval.ival = OP_BACKTICK;
2922 && GvCVu(gv_readpipe) && GvIMPORTED_CV(gv_readpipe))
2924 ((gvp = (GV**)hv_fetchs(PL_globalstash, "readpipe", FALSE))
2925 && (gv_readpipe = *gvp) != (GV*)&PL_sv_undef
2926 && GvCVu(gv_readpipe) && GvIMPORTED_CV(gv_readpipe)))
2928 PL_lex_op = (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
2929 append_elem(OP_LIST,
2930 newSVOP(OP_CONST, 0, &PL_sv_undef), /* value will be read later */
2931 newCVREF(0, newGVOP(OP_GV, 0, gv_readpipe))));
2941 * The intent of this yylex wrapper is to minimize the changes to the
2942 * tokener when we aren't interested in collecting madprops. It remains
2943 * to be seen how successful this strategy will be...
2950 char *s = PL_bufptr;
2952 /* make sure PL_thiswhite is initialized */
2956 /* just do what yylex would do on pending identifier; leave PL_thiswhite alone */
2957 if (PL_pending_ident)
2958 return S_pending_ident(aTHX);
2960 /* previous token ate up our whitespace? */
2961 if (!PL_lasttoke && PL_nextwhite) {
2962 PL_thiswhite = PL_nextwhite;
2966 /* isolate the token, and figure out where it is without whitespace */
2967 PL_realtokenstart = -1;
2971 assert(PL_curforce < 0);
2973 if (!PL_thismad || PL_thismad->mad_key == '^') { /* not forced already? */
2974 if (!PL_thistoken) {
2975 if (PL_realtokenstart < 0 || !CopLINE(PL_curcop))
2976 PL_thistoken = newSVpvs("");
2978 char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
2979 PL_thistoken = newSVpvn(tstart, s - tstart);
2982 if (PL_thismad) /* install head */
2983 CURMAD('X', PL_thistoken);
2986 /* last whitespace of a sublex? */
2987 if (optype == ')' && PL_endwhite) {
2988 CURMAD('X', PL_endwhite);
2993 /* if no whitespace and we're at EOF, bail. Otherwise fake EOF below. */
2994 if (!PL_thiswhite && !PL_endwhite && !optype) {
2995 sv_free(PL_thistoken);
3000 /* put off final whitespace till peg */
3001 if (optype == ';' && !PL_rsfp) {
3002 PL_nextwhite = PL_thiswhite;
3005 else if (PL_thisopen) {
3006 CURMAD('q', PL_thisopen);
3008 sv_free(PL_thistoken);
3012 /* Store actual token text as madprop X */
3013 CURMAD('X', PL_thistoken);
3017 /* add preceding whitespace as madprop _ */
3018 CURMAD('_', PL_thiswhite);
3022 /* add quoted material as madprop = */
3023 CURMAD('=', PL_thisstuff);
3027 /* add terminating quote as madprop Q */
3028 CURMAD('Q', PL_thisclose);
3032 /* special processing based on optype */
3036 /* opval doesn't need a TOKEN since it can already store mp */
3047 append_madprops(PL_thismad, yylval.opval, 0);
3055 addmad(newMADsv('p', PL_endwhite), &PL_thismad, 0);
3064 /* remember any fake bracket that lexer is about to discard */
3065 if (PL_lex_brackets == 1 &&
3066 ((expectation)PL_lex_brackstack[0] & XFAKEBRACK))
3069 while (s < PL_bufend && (*s == ' ' || *s == '\t'))
3072 PL_thiswhite = newSVpvn(PL_bufptr, ++s - PL_bufptr);
3073 addmad(newMADsv('#', PL_thiswhite), &PL_thismad, 0);
3076 break; /* don't bother looking for trailing comment */
3085 /* attach a trailing comment to its statement instead of next token */
3089 if (PL_bufptr > PL_oldbufptr && PL_bufptr[-1] == optype) {
3091 while (s < PL_bufend && (*s == ' ' || *s == '\t'))
3093 if (*s == '\n' || *s == '#') {
3094 while (s < PL_bufend && *s != '\n')
3098 PL_thiswhite = newSVpvn(PL_bufptr, s - PL_bufptr);
3099 addmad(newMADsv('#', PL_thiswhite), &PL_thismad, 0);
3116 /* Create new token struct. Note: opvals return early above. */
3117 yylval.tkval = newTOKEN(optype, yylval, PL_thismad);
3124 S_tokenize_use(pTHX_ int is_use, char *s) {
3126 if (PL_expect != XSTATE)
3127 yyerror(Perl_form(aTHX_ "\"%s\" not allowed in expression",
3128 is_use ? "use" : "no"));
3130 if (isDIGIT(*s) || (*s == 'v' && isDIGIT(s[1]))) {
3131 s = force_version(s, TRUE);
3132 if (*s == ';' || (s = SKIPSPACE1(s), *s == ';')) {
3133 start_force(PL_curforce);
3134 NEXTVAL_NEXTTOKE.opval = NULL;
3137 else if (*s == 'v') {
3138 s = force_word(s,WORD,FALSE,TRUE,FALSE);
3139 s = force_version(s, FALSE);
3143 s = force_word(s,WORD,FALSE,TRUE,FALSE);
3144 s = force_version(s, FALSE);
3146 yylval.ival = is_use;
3150 static const char* const exp_name[] =
3151 { "OPERATOR", "TERM", "REF", "STATE", "BLOCK", "ATTRBLOCK",
3152 "ATTRTERM", "TERMBLOCK", "TERMORDORDOR"
3159 Works out what to call the token just pulled out of the input
3160 stream. The yacc parser takes care of taking the ops we return and
3161 stitching them into a tree.
3167 if read an identifier
3168 if we're in a my declaration
3169 croak if they tried to say my($foo::bar)
3170 build the ops for a my() declaration
3171 if it's an access to a my() variable
3172 are we in a sort block?
3173 croak if my($a); $a <=> $b
3174 build ops for access to a my() variable
3175 if in a dq string, and they've said @foo and we can't find @foo
3177 build ops for a bareword
3178 if we already built the token before, use it.
3183 #pragma segment Perl_yylex
3189 register char *s = PL_bufptr;
3194 /* orig_keyword, gvp, and gv are initialized here because
3195 * jump to the label just_a_word_zero can bypass their
3196 * initialization later. */
3197 I32 orig_keyword = 0;
3202 SV* tmp = newSVpvs("");
3203 PerlIO_printf(Perl_debug_log, "### %"IVdf":LEX_%s/X%s %s\n",
3204 (IV)CopLINE(PL_curcop),
3205 lex_state_names[PL_lex_state],
3206 exp_name[PL_expect],
3207 pv_display(tmp, s, strlen(s), 0, 60));
3210 /* check if there's an identifier for us to look at */
3211 if (PL_pending_ident)
3212 return REPORT(S_pending_ident(aTHX));
3214 /* no identifier pending identification */
3216 switch (PL_lex_state) {
3218 case LEX_NORMAL: /* Some compilers will produce faster */
3219 case LEX_INTERPNORMAL: /* code if we comment these out. */
3223 /* when we've already built the next token, just pull it out of the queue */
3227 yylval = PL_nexttoke[PL_lasttoke].next_val;
3229 PL_thismad = PL_nexttoke[PL_lasttoke].next_mad;
3230 PL_nexttoke[PL_lasttoke].next_mad = 0;
3231 if (PL_thismad && PL_thismad->mad_key == '_') {
3232 PL_thiswhite = (SV*)PL_thismad->mad_val;
3233 PL_thismad->mad_val = 0;
3234 mad_free(PL_thismad);
3239 PL_lex_state = PL_lex_defer;
3240 PL_expect = PL_lex_expect;
3241 PL_lex_defer = LEX_NORMAL;
3242 if (!PL_nexttoke[PL_lasttoke].next_type)
3247 yylval = PL_nextval[PL_nexttoke];
3249 PL_lex_state = PL_lex_defer;
3250 PL_expect = PL_lex_expect;
3251 PL_lex_defer = LEX_NORMAL;
3255 /* FIXME - can these be merged? */
3256 return(PL_nexttoke[PL_lasttoke].next_type);
3258 return REPORT(PL_nexttype[PL_nexttoke]);
3261 /* interpolated case modifiers like \L \U, including \Q and \E.
3262 when we get here, PL_bufptr is at the \
3264 case LEX_INTERPCASEMOD:
3266 if (PL_bufptr != PL_bufend && *PL_bufptr != '\\')
3267 Perl_croak(aTHX_ "panic: INTERPCASEMOD");
3269 /* handle \E or end of string */
3270 if (PL_bufptr == PL_bufend || PL_bufptr[1] == 'E') {
3272 if (PL_lex_casemods) {
3273 const char oldmod = PL_lex_casestack[--PL_lex_casemods];
3274 PL_lex_casestack[PL_lex_casemods] = '\0';
3276 if (PL_bufptr != PL_bufend
3277 && (oldmod == 'L' || oldmod == 'U' || oldmod == 'Q')) {
3279 PL_lex_state = LEX_INTERPCONCAT;
3282 PL_thistoken = newSVpvs("\\E");
3288 while (PL_bufptr != PL_bufend &&
3289 PL_bufptr[0] == '\\' && PL_bufptr[1] == 'E') {
3291 PL_thiswhite = newSVpvs("");
3292 sv_catpvn(PL_thiswhite, PL_bufptr, 2);
3296 if (PL_bufptr != PL_bufend)
3299 PL_lex_state = LEX_INTERPCONCAT;
3303 DEBUG_T({ PerlIO_printf(Perl_debug_log,
3304 "### Saw case modifier\n"); });
3306 if (s[1] == '\\' && s[2] == 'E') {
3309 PL_thiswhite = newSVpvs("");
3310 sv_catpvn(PL_thiswhite, PL_bufptr, 4);
3313 PL_lex_state = LEX_INTERPCONCAT;
3318 if (!PL_madskills) /* when just compiling don't need correct */
3319 if (strnEQ(s, "L\\u", 3) || strnEQ(s, "U\\l", 3))
3320 tmp = *s, *s = s[2], s[2] = (char)tmp; /* misordered... */
3321 if ((*s == 'L' || *s == 'U') &&
3322 (strchr(PL_lex_casestack, 'L') || strchr(PL_lex_casestack, 'U'))) {
3323 PL_lex_casestack[--PL_lex_casemods] = '\0';
3326 if (PL_lex_casemods > 10)
3327 Renew(PL_lex_casestack, PL_lex_casemods + 2, char);
3328 PL_lex_casestack[PL_lex_casemods++] = *s;
3329 PL_lex_casestack[PL_lex_casemods] = '\0';
3330 PL_lex_state = LEX_INTERPCONCAT;
3331 start_force(PL_curforce);
3332 NEXTVAL_NEXTTOKE.ival = 0;
3334 start_force(PL_curforce);
3336 NEXTVAL_NEXTTOKE.ival = OP_LCFIRST;
3338 NEXTVAL_NEXTTOKE.ival = OP_UCFIRST;
3340 NEXTVAL_NEXTTOKE.ival = OP_LC;
3342 NEXTVAL_NEXTTOKE.ival = OP_UC;
3344 NEXTVAL_NEXTTOKE.ival = OP_QUOTEMETA;
3346 Perl_croak(aTHX_ "panic: yylex");
3348 SV* const tmpsv = newSVpvs("");
3349 Perl_sv_catpvf(aTHX_ tmpsv, "\\%c", *s);
3355 if (PL_lex_starts) {
3361 sv_free(PL_thistoken);
3362 PL_thistoken = newSVpvs("");
3365 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
3366 if (PL_lex_casemods == 1 && PL_lex_inpat)
3375 case LEX_INTERPPUSH:
3376 return REPORT(sublex_push());
3378 case LEX_INTERPSTART:
3379 if (PL_bufptr == PL_bufend)
3380 return REPORT(sublex_done());
3381 DEBUG_T({ PerlIO_printf(Perl_debug_log,
3382 "### Interpolated variable\n"); });
3384 PL_lex_dojoin = (*PL_bufptr == '@');
3385 PL_lex_state = LEX_INTERPNORMAL;
3386 if (PL_lex_dojoin) {
3387 start_force(PL_curforce);
3388 NEXTVAL_NEXTTOKE.ival = 0;
3390 start_force(PL_curforce);
3391 force_ident("\"", '$');
3392 start_force(PL_curforce);
3393 NEXTVAL_NEXTTOKE.ival = 0;
3395 start_force(PL_curforce);
3396 NEXTVAL_NEXTTOKE.ival = 0;
3398 start_force(PL_curforce);
3399 NEXTVAL_NEXTTOKE.ival = OP_JOIN; /* emulate join($", ...) */
3402 if (PL_lex_starts++) {
3407 sv_free(PL_thistoken);
3408 PL_thistoken = newSVpvs("");
3411 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
3412 if (!PL_lex_casemods && PL_lex_inpat)
3419 case LEX_INTERPENDMAYBE:
3420 if (intuit_more(PL_bufptr)) {
3421 PL_lex_state = LEX_INTERPNORMAL; /* false alarm, more expr */
3427 if (PL_lex_dojoin) {
3428 PL_lex_dojoin = FALSE;
3429 PL_lex_state = LEX_INTERPCONCAT;
3433 sv_free(PL_thistoken);
3434 PL_thistoken = newSVpvs("");
3439 if (PL_lex_inwhat == OP_SUBST && PL_linestr == PL_lex_repl
3440 && SvEVALED(PL_lex_repl))
3442 if (PL_bufptr != PL_bufend)
3443 Perl_croak(aTHX_ "Bad evalled substitution pattern");
3447 case LEX_INTERPCONCAT:
3449 if (PL_lex_brackets)
3450 Perl_croak(aTHX_ "panic: INTERPCONCAT");
3452 if (PL_bufptr == PL_bufend)
3453 return REPORT(sublex_done());
3455 if (SvIVX(PL_linestr) == '\'') {
3456 SV *sv = newSVsv(PL_linestr);
3459 else if ( PL_hints & HINT_NEW_RE )
3460 sv = new_constant(NULL, 0, "qr", sv, sv, "q");
3461 yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
3465 s = scan_const(PL_bufptr);
3467 PL_lex_state = LEX_INTERPCASEMOD;
3469 PL_lex_state = LEX_INTERPSTART;
3472 if (s != PL_bufptr) {
3473 start_force(PL_curforce);
3475 curmad('X', newSVpvn(PL_bufptr,s-PL_bufptr));
3477 NEXTVAL_NEXTTOKE = yylval;
3480 if (PL_lex_starts++) {
3484 sv_free(PL_thistoken);
3485 PL_thistoken = newSVpvs("");
3488 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
3489 if (!PL_lex_casemods && PL_lex_inpat)
3502 PL_lex_state = LEX_NORMAL;
3503 s = scan_formline(PL_bufptr);
3504 if (!PL_lex_formbrack)
3510 PL_oldoldbufptr = PL_oldbufptr;
3516 sv_free(PL_thistoken);
3519 PL_realtokenstart = s - SvPVX(PL_linestr); /* assume but undo on ws */
3523 if (isIDFIRST_lazy_if(s,UTF))
3525 Perl_croak(aTHX_ "Unrecognized character \\x%02X", *s & 255);
3528 goto fake_eof; /* emulate EOF on ^D or ^Z */
3537 if (PL_lex_brackets) {
3538 yyerror((const char *)
3540 ? "Format not terminated"
3541 : "Missing right curly or square bracket"));
3543 DEBUG_T( { PerlIO_printf(Perl_debug_log,
3544 "### Tokener got EOF\n");
3548 if (s++ < PL_bufend)
3549 goto retry; /* ignore stray nulls */
3552 if (!PL_in_eval && !PL_preambled) {
3553 PL_preambled = TRUE;
3558 sv_setpv(PL_linestr,incl_perldb());
3559 if (SvCUR(PL_linestr))
3560 sv_catpvs(PL_linestr,";");
3562 while(AvFILLp(PL_preambleav) >= 0) {
3563 SV *tmpsv = av_shift(PL_preambleav);
3564 sv_catsv(PL_linestr, tmpsv);
3565 sv_catpvs(PL_linestr, ";");
3568 sv_free((SV*)PL_preambleav);
3569 PL_preambleav = NULL;
3571 if (PL_minus_n || PL_minus_p) {
3572 sv_catpvs(PL_linestr, "LINE: while (<>) {");
3574 sv_catpvs(PL_linestr,"chomp;");
3577 if ((*PL_splitstr == '/' || *PL_splitstr == '\''
3578 || *PL_splitstr == '"')
3579 && strchr(PL_splitstr + 1, *PL_splitstr))
3580 Perl_sv_catpvf(aTHX_ PL_linestr, "our @F=split(%s);", PL_splitstr);
3582 /* "q\0${splitstr}\0" is legal perl. Yes, even NUL
3583 bytes can be used as quoting characters. :-) */
3584 const char *splits = PL_splitstr;
3585 sv_catpvs(PL_linestr, "our @F=split(q\0");
3588 if (*splits == '\\')
3589 sv_catpvn(PL_linestr, splits, 1);
3590 sv_catpvn(PL_linestr, splits, 1);
3591 } while (*splits++);
3592 /* This loop will embed the trailing NUL of
3593 PL_linestr as the last thing it does before
3595 sv_catpvs(PL_linestr, ");");
3599 sv_catpvs(PL_linestr,"our @F=split(' ');");
3603 sv_catpvs(PL_linestr,"use feature ':5.10';");
3604 sv_catpvs(PL_linestr, "\n");
3605 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3606 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3607 PL_last_lop = PL_last_uni = NULL;
3608 if (PERLDB_LINE && PL_curstash != PL_debstash)
3609 update_debugger_info(PL_linestr, NULL, 0);
3613 bof = PL_rsfp ? TRUE : FALSE;
3614 if ((s = filter_gets(PL_linestr, PL_rsfp, 0)) == NULL) {
3617 PL_realtokenstart = -1;
3620 if (PL_preprocess && !PL_in_eval)
3621 (void)PerlProc_pclose(PL_rsfp);
3622 else if ((PerlIO *)PL_rsfp == PerlIO_stdin())
3623 PerlIO_clearerr(PL_rsfp);
3625 (void)PerlIO_close(PL_rsfp);
3627 PL_doextract = FALSE;
3629 if (!PL_in_eval && (PL_minus_n || PL_minus_p)) {
3634 sv_setpv(PL_linestr,
3637 ? ";}continue{print;}" : ";}"));
3638 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3639 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3640 PL_last_lop = PL_last_uni = NULL;
3641 PL_minus_n = PL_minus_p = 0;
3644 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3645 PL_last_lop = PL_last_uni = NULL;
3646 sv_setpvn(PL_linestr,"",0);
3647 TOKEN(';'); /* not infinite loop because rsfp is NULL now */
3649 /* If it looks like the start of a BOM or raw UTF-16,
3650 * check if it in fact is. */
3656 #ifdef PERLIO_IS_STDIO
3657 # ifdef __GNU_LIBRARY__
3658 # if __GNU_LIBRARY__ == 1 /* Linux glibc5 */
3659 # define FTELL_FOR_PIPE_IS_BROKEN
3663 # if __GLIBC__ == 1 /* maybe some glibc5 release had it like this? */
3664 # define FTELL_FOR_PIPE_IS_BROKEN
3669 #ifdef FTELL_FOR_PIPE_IS_BROKEN
3670 /* This loses the possibility to detect the bof
3671 * situation on perl -P when the libc5 is being used.
3672 * Workaround? Maybe attach some extra state to PL_rsfp?
3675 bof = PerlIO_tell(PL_rsfp) == SvCUR(PL_linestr);
3677 bof = PerlIO_tell(PL_rsfp) == (Off_t)SvCUR(PL_linestr);
3680 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3681 s = swallow_bom((U8*)s);
3685 /* Incest with pod. */
3688 sv_catsv(PL_thiswhite, PL_linestr);
3690 if (*s == '=' && strnEQ(s, "=cut", 4) && !isALPHA(s[4])) {
3691 sv_setpvn(PL_linestr, "", 0);
3692 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3693 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3694 PL_last_lop = PL_last_uni = NULL;
3695 PL_doextract = FALSE;
3699 } while (PL_doextract);
3700 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = s;
3701 if (PERLDB_LINE && PL_curstash != PL_debstash)
3702 update_debugger_info(PL_linestr, NULL, 0);
3703 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3704 PL_last_lop = PL_last_uni = NULL;
3705 if (CopLINE(PL_curcop) == 1) {
3706 while (s < PL_bufend && isSPACE(*s))
3708 if (*s == ':' && s[1] != ':') /* for csh execing sh scripts */
3712 PL_thiswhite = newSVpvn(PL_linestart, s - PL_linestart);
3716 if (*s == '#' && *(s+1) == '!')
3718 #ifdef ALTERNATE_SHEBANG
3720 static char const as[] = ALTERNATE_SHEBANG;
3721 if (*s == as[0] && strnEQ(s, as, sizeof(as) - 1))
3722 d = s + (sizeof(as) - 1);
3724 #endif /* ALTERNATE_SHEBANG */
3733 while (*d && !isSPACE(*d))
3737 #ifdef ARG_ZERO_IS_SCRIPT
3738 if (ipathend > ipath) {
3740 * HP-UX (at least) sets argv[0] to the script name,
3741 * which makes $^X incorrect. And Digital UNIX and Linux,
3742 * at least, set argv[0] to the basename of the Perl
3743 * interpreter. So, having found "#!", we'll set it right.
3745 SV * const x = GvSV(gv_fetchpvs("\030", GV_ADD|GV_NOTQUAL,
3747 assert(SvPOK(x) || SvGMAGICAL(x));
3748 if (sv_eq(x, CopFILESV(PL_curcop))) {
3749 sv_setpvn(x, ipath, ipathend - ipath);
3755 const char *bstart = SvPV_const(CopFILESV(PL_curcop),blen);
3756 const char * const lstart = SvPV_const(x,llen);
3758 bstart += blen - llen;
3759 if (strnEQ(bstart, lstart, llen) && bstart[-1] == '/') {
3760 sv_setpvn(x, ipath, ipathend - ipath);
3765 TAINT_NOT; /* $^X is always tainted, but that's OK */
3767 #endif /* ARG_ZERO_IS_SCRIPT */
3772 d = instr(s,"perl -");
3774 d = instr(s,"perl");
3776 /* avoid getting into infinite loops when shebang
3777 * line contains "Perl" rather than "perl" */
3779 for (d = ipathend-4; d >= ipath; --d) {
3780 if ((*d == 'p' || *d == 'P')
3781 && !ibcmp(d, "perl", 4))
3791 #ifdef ALTERNATE_SHEBANG
3793 * If the ALTERNATE_SHEBANG on this system starts with a
3794 * character that can be part of a Perl expression, then if
3795 * we see it but not "perl", we're probably looking at the
3796 * start of Perl code, not a request to hand off to some
3797 * other interpreter. Similarly, if "perl" is there, but
3798 * not in the first 'word' of the line, we assume the line
3799 * contains the start of the Perl program.
3801 if (d && *s != '#') {
3802 const char *c = ipath;
3803 while (*c && !strchr("; \t\r\n\f\v#", *c))
3806 d = NULL; /* "perl" not in first word; ignore */
3808 *s = '#'; /* Don't try to parse shebang line */
3810 #endif /* ALTERNATE_SHEBANG */
3811 #ifndef MACOS_TRADITIONAL
3816 !instr(s,"indir") &&
3817 instr(PL_origargv[0],"perl"))
3824 while (s < PL_bufend && isSPACE(*s))
3826 if (s < PL_bufend) {
3827 Newxz(newargv,PL_origargc+3,char*);
3829 while (s < PL_bufend && !isSPACE(*s))
3832 Copy(PL_origargv+1, newargv+2, PL_origargc+1, char*);
3835 newargv = PL_origargv;
3838 PerlProc_execv(ipath, EXEC_ARGV_CAST(newargv));
3840 Perl_croak(aTHX_ "Can't exec %s", ipath);
3844 while (*d && !isSPACE(*d))
3846 while (SPACE_OR_TAB(*d))
3850 const bool switches_done = PL_doswitches;
3851 const U32 oldpdb = PL_perldb;
3852 const bool oldn = PL_minus_n;
3853 const bool oldp = PL_minus_p;
3856 if (*d == 'M' || *d == 'm' || *d == 'C') {
3857 const char * const m = d;
3858 while (*d && !isSPACE(*d))
3860 Perl_croak(aTHX_ "Too late for \"-%.*s\" option",
3863 d = moreswitches(d);
3865 if (PL_doswitches && !switches_done) {
3866 int argc = PL_origargc;
3867 char **argv = PL_origargv;
3870 } while (argc && argv[0][0] == '-' && argv[0][1]);
3871 init_argv_symbols(argc,argv);
3873 if ((PERLDB_LINE && !oldpdb) ||
3874 ((PL_minus_n || PL_minus_p) && !(oldn || oldp)))
3875 /* if we have already added "LINE: while (<>) {",
3876 we must not do it again */
3878 sv_setpvn(PL_linestr, "", 0);
3879 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3880 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3881 PL_last_lop = PL_last_uni = NULL;
3882 PL_preambled = FALSE;
3884 (void)gv_fetchfile(PL_origfilename);
3891 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
3893 PL_lex_state = LEX_FORMLINE;
3898 #ifdef PERL_STRICT_CR
3899 Perl_warn(aTHX_ "Illegal character \\%03o (carriage return)", '\r');
3901 "\t(Maybe you didn't strip carriage returns after a network transfer?)\n");
3903 case ' ': case '\t': case '\f': case 013:
3904 #ifdef MACOS_TRADITIONAL
3908 PL_realtokenstart = -1;
3917 PL_realtokenstart = -1;
3921 if (PL_lex_state != LEX_NORMAL || (PL_in_eval && !PL_rsfp)) {
3922 if (*s == '#' && s == PL_linestart && PL_in_eval && !PL_rsfp) {
3923 /* handle eval qq[#line 1 "foo"\n ...] */
3924 CopLINE_dec(PL_curcop);
3927 if (PL_madskills && !PL_lex_formbrack && !PL_in_eval) {
3929 if (!PL_in_eval || PL_rsfp)
3934 while (d < PL_bufend && *d != '\n')
3938 else if (d > PL_bufend) /* Found by Ilya: feed random input to Perl. */
3939 Perl_croak(aTHX_ "panic: input overflow");
3942 PL_thiswhite = newSVpvn(s, d - s);
3947 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
3949 PL_lex_state = LEX_FORMLINE;
3955 if (PL_madskills && CopLINE(PL_curcop) >= 1 && !PL_lex_formbrack) {
3956 if (CopLINE(PL_curcop) == 1 && s[0] == '#' && s[1] == '!') {
3959 TOKEN(PEG); /* make sure any #! line is accessible */
3964 /* if (PL_madskills && PL_lex_formbrack) { */
3966 while (d < PL_bufend && *d != '\n')
3970 else if (d > PL_bufend) /* Found by Ilya: feed random input to Perl. */
3971 Perl_croak(aTHX_ "panic: input overflow");
3972 if (PL_madskills && CopLINE(PL_curcop) >= 1) {
3974 PL_thiswhite = newSVpvs("");
3975 if (CopLINE(PL_curcop) == 1) {
3976 sv_setpvn(PL_thiswhite, "", 0);
3979 sv_catpvn(PL_thiswhite, s, d - s);
3993 if (s[1] && isALPHA(s[1]) && !isALNUM(s[2])) {
4001 while (s < PL_bufend && SPACE_OR_TAB(*s))
4004 if (strnEQ(s,"=>",2)) {
4005 s = force_word(PL_bufptr,WORD,FALSE,FALSE,FALSE);
4006 DEBUG_T( { printbuf("### Saw unary minus before =>, forcing word %s\n", s); } );
4007 OPERATOR('-'); /* unary minus */
4009 PL_last_uni = PL_oldbufptr;
4011 case 'r': ftst = OP_FTEREAD; break;
4012 case 'w': ftst = OP_FTEWRITE; break;
4013 case 'x': ftst = OP_FTEEXEC; break;
4014 case 'o': ftst = OP_FTEOWNED; break;
4015 case 'R': ftst = OP_FTRREAD; break;
4016 case 'W': ftst = OP_FTRWRITE; break;
4017 case 'X': ftst = OP_FTREXEC; break;
4018 case 'O': ftst = OP_FTROWNED; break;
4019 case 'e': ftst = OP_FTIS; break;
4020 case 'z': ftst = OP_FTZERO; break;
4021 case 's': ftst = OP_FTSIZE; break;
4022 case 'f': ftst = OP_FTFILE; break;
4023 case 'd': ftst = OP_FTDIR; break;
4024 case 'l': ftst = OP_FTLINK; break;
4025 case 'p': ftst = OP_FTPIPE; break;
4026 case 'S': ftst = OP_FTSOCK; break;
4027 case 'u': ftst = OP_FTSUID; break;
4028 case 'g': ftst = OP_FTSGID; break;
4029 case 'k': ftst = OP_FTSVTX; break;
4030 case 'b': ftst = OP_FTBLK; break;
4031 case 'c': ftst = OP_FTCHR; break;
4032 case 't': ftst = OP_FTTTY; break;
4033 case 'T': ftst = OP_FTTEXT; break;
4034 case 'B': ftst = OP_FTBINARY; break;
4035 case 'M': case 'A': case 'C':
4036 gv_fetchpvs("\024", GV_ADD|GV_NOTQUAL, SVt_PV);
4038 case 'M': ftst = OP_FTMTIME; break;
4039 case 'A': ftst = OP_FTATIME; break;
4040 case 'C': ftst = OP_FTCTIME; break;
4048 PL_last_lop_op = (OPCODE)ftst;
4049 DEBUG_T( { PerlIO_printf(Perl_debug_log,
4050 "### Saw file test %c\n", (int)tmp);
4055 /* Assume it was a minus followed by a one-letter named
4056 * subroutine call (or a -bareword), then. */
4057 DEBUG_T( { PerlIO_printf(Perl_debug_log,
4058 "### '-%c' looked like a file test but was not\n",
4065 const char tmp = *s++;
4068 if (PL_expect == XOPERATOR)
4073 else if (*s == '>') {
4076 if (isIDFIRST_lazy_if(s,UTF)) {
4077 s = force_word(s,METHOD,FALSE,TRUE,FALSE);
4085 if (PL_expect == XOPERATOR)
4088 if (isSPACE(*s) || !isSPACE(*PL_bufptr))
4090 OPERATOR('-'); /* unary minus */
4096 const char tmp = *s++;
4099 if (PL_expect == XOPERATOR)
4104 if (PL_expect == XOPERATOR)
4107 if (isSPACE(*s) || !isSPACE(*PL_bufptr))
4114 if (PL_expect != XOPERATOR) {
4115 s = scan_ident(s, PL_bufend, PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
4116 PL_expect = XOPERATOR;
4117 force_ident(PL_tokenbuf, '*');
4130 if (PL_expect == XOPERATOR) {
4134 PL_tokenbuf[0] = '%';
4135 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1, TRUE);
4136 if (!PL_tokenbuf[1]) {
4139 PL_pending_ident = '%';
4150 && (PL_expect == XOPERATOR || PL_expect == XTERMORDORDOR))
4157 const char tmp = *s++;
4163 goto just_a_word_zero_gv;
4166 switch (PL_expect) {
4172 if (!PL_in_my || PL_lex_state != LEX_NORMAL)
4174 PL_bufptr = s; /* update in case we back off */
4180 PL_expect = XTERMBLOCK;
4183 stuffstart = s - SvPVX(PL_linestr) - 1;
4187 while (isIDFIRST_lazy_if(s,UTF)) {
4190 d = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
4191 if (isLOWER(*s) && (tmp = keyword(PL_tokenbuf, len, 0))) {
4192 if (tmp < 0) tmp = -tmp;
4207 sv = newSVpvn(s, len);
4209 d = scan_str(d,TRUE,TRUE);
4211 /* MUST advance bufptr here to avoid bogus
4212 "at end of line" context messages from yyerror().
4214 PL_bufptr = s + len;
4215 yyerror("Unterminated attribute parameter in attribute list");
4219 return REPORT(0); /* EOF indicator */
4223 sv_catsv(sv, PL_lex_stuff);
4224 attrs = append_elem(OP_LIST, attrs,
4225 newSVOP(OP_CONST, 0, sv));
4226 SvREFCNT_dec(PL_lex_stuff);
4227 PL_lex_stuff = NULL;
4230 if (len == 6 && strnEQ(SvPVX(sv), "unique", len)) {
4232 if (PL_in_my == KEY_our) {
4234 GvUNIQUE_on(cGVOPx_gv(yylval.opval));
4236 /* skip to avoid loading attributes.pm */
4238 deprecate(":unique");
4241 Perl_croak(aTHX_ "The 'unique' attribute may only be applied to 'our' variables");
4244 /* NOTE: any CV attrs applied here need to be part of
4245 the CVf_BUILTIN_ATTRS define in cv.h! */
4246 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "lvalue", len)) {
4248 CvLVALUE_on(PL_compcv);
4250 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "locked", len)) {
4252 CvLOCKED_on(PL_compcv);
4254 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "method", len)) {
4256 CvMETHOD_on(PL_compcv);
4258 else if (!PL_in_my && len == 9 && strnEQ(SvPVX(sv), "assertion", len)) {
4260 CvASSERTION_on(PL_compcv);
4262 /* After we've set the flags, it could be argued that
4263 we don't need to do the attributes.pm-based setting
4264 process, and shouldn't bother appending recognized
4265 flags. To experiment with that, uncomment the
4266 following "else". (Note that's already been
4267 uncommented. That keeps the above-applied built-in
4268 attributes from being intercepted (and possibly
4269 rejected) by a package's attribute routines, but is
4270 justified by the performance win for the common case
4271 of applying only built-in attributes.) */
4273 attrs = append_elem(OP_LIST, attrs,
4274 newSVOP(OP_CONST, 0,
4278 if (*s == ':' && s[1] != ':')
4281 break; /* require real whitespace or :'s */
4282 /* XXX losing whitespace on sequential attributes here */
4286 = (PL_expect == XOPERATOR ? '=' : '{'); /*'}(' for vi */
4287 if (*s != ';' && *s != '}' && *s != tmp
4288 && (tmp != '=' || *s != ')')) {
4289 const char q = ((*s == '\'') ? '"' : '\'');
4290 /* If here for an expression, and parsed no attrs, back
4292 if (tmp == '=' && !attrs) {
4296 /* MUST advance bufptr here to avoid bogus "at end of line"
4297 context messages from yyerror().
4300 yyerror( (const char *)
4302 ? Perl_form(aTHX_ "Invalid separator character "
4303 "%c%c%c in attribute list", q, *s, q)
4304 : "Unterminated attribute list" ) );
4312 start_force(PL_curforce);
4313 NEXTVAL_NEXTTOKE.opval = attrs;
4314 CURMAD('_', PL_nextwhite);
4319 PL_thistoken = newSVpvn(SvPVX(PL_linestr) + stuffstart,
4320 (s - SvPVX(PL_linestr)) - stuffstart);
4328 if (PL_last_lop == PL_oldoldbufptr || PL_last_uni == PL_oldoldbufptr)
4329 PL_oldbufptr = PL_oldoldbufptr; /* allow print(STDOUT 123) */
4337 const char tmp = *s++;
4342 const char tmp = *s++;
4350 if (PL_lex_brackets <= 0)
4351 yyerror("Unmatched right square bracket");
4354 if (PL_lex_state == LEX_INTERPNORMAL) {
4355 if (PL_lex_brackets == 0) {
4356 if (*s != '[' && *s != '{' && (*s != '-' || s[1] != '>'))
4357 PL_lex_state = LEX_INTERPEND;
4364 if (PL_lex_brackets > 100) {
4365 Renew(PL_lex_brackstack, PL_lex_brackets + 10, char);
4367 switch (PL_expect) {
4369 if (PL_lex_formbrack) {
4373 if (PL_oldoldbufptr == PL_last_lop)
4374 PL_lex_brackstack[PL_lex_brackets++] = XTERM;
4376 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
4377 OPERATOR(HASHBRACK);
4379 while (s < PL_bufend && SPACE_OR_TAB(*s))
4382 PL_tokenbuf[0] = '\0';
4383 if (d < PL_bufend && *d == '-') {
4384 PL_tokenbuf[0] = '-';
4386 while (d < PL_bufend && SPACE_OR_TAB(*d))
4389 if (d < PL_bufend && isIDFIRST_lazy_if(d,UTF)) {
4390 d = scan_word(d, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1,
4392 while (d < PL_bufend && SPACE_OR_TAB(*d))
4395 const char minus = (PL_tokenbuf[0] == '-');
4396 s = force_word(s + minus, WORD, FALSE, TRUE, FALSE);
4404 PL_lex_brackstack[PL_lex_brackets++] = XSTATE;
4409 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
4414 if (PL_oldoldbufptr == PL_last_lop)
4415 PL_lex_brackstack[PL_lex_brackets++] = XTERM;
4417 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
4420 if (PL_expect == XREF && PL_lex_state == LEX_INTERPNORMAL) {
4422 /* This hack is to get the ${} in the message. */
4424 yyerror("syntax error");
4427 OPERATOR(HASHBRACK);
4429 /* This hack serves to disambiguate a pair of curlies
4430 * as being a block or an anon hash. Normally, expectation
4431 * determines that, but in cases where we're not in a
4432 * position to expect anything in particular (like inside
4433 * eval"") we have to resolve the ambiguity. This code
4434 * covers the case where the first term in the curlies is a
4435 * quoted string. Most other cases need to be explicitly
4436 * disambiguated by prepending a "+" before the opening
4437 * curly in order to force resolution as an anon hash.
4439 * XXX should probably propagate the outer expectation
4440 * into eval"" to rely less on this hack, but that could
4441 * potentially break current behavior of eval"".
4445 if (*s == '\'' || *s == '"' || *s == '`') {
4446 /* common case: get past first string, handling escapes */
4447 for (t++; t < PL_bufend && *t != *s;)
4448 if (*t++ == '\\' && (*t == '\\' || *t == *s))
4452 else if (*s == 'q') {
4455 || ((*t == 'q' || *t == 'x') && ++t < PL_bufend
4458 /* skip q//-like construct */
4460 char open, close, term;
4463 while (t < PL_bufend && isSPACE(*t))
4465 /* check for q => */
4466 if (t+1 < PL_bufend && t[0] == '=' && t[1] == '>') {
4467 OPERATOR(HASHBRACK);
4471 if (term && (tmps = strchr("([{< )]}> )]}>",term)))
4475 for (t++; t < PL_bufend; t++) {
4476 if (*t == '\\' && t+1 < PL_bufend && open != '\\')
4478 else if (*t == open)
4482 for (t++; t < PL_bufend; t++) {
4483 if (*t == '\\' && t+1 < PL_bufend)
4485 else if (*t == close && --brackets <= 0)
4487 else if (*t == open)
4494 /* skip plain q word */
4495 while (t < PL_bufend && isALNUM_lazy_if(t,UTF))
4498 else if (isALNUM_lazy_if(t,UTF)) {
4500 while (t < PL_bufend && isALNUM_lazy_if(t,UTF))
4503 while (t < PL_bufend && isSPACE(*t))
4505 /* if comma follows first term, call it an anon hash */
4506 /* XXX it could be a comma expression with loop modifiers */
4507 if (t < PL_bufend && ((*t == ',' && (*s == 'q' || !isLOWER(*s)))
4508 || (*t == '=' && t[1] == '>')))
4509 OPERATOR(HASHBRACK);
4510 if (PL_expect == XREF)
4513 PL_lex_brackstack[PL_lex_brackets-1] = XSTATE;
4519 yylval.ival = CopLINE(PL_curcop);
4520 if (isSPACE(*s) || *s == '#')
4521 PL_copline = NOLINE; /* invalidate current command line number */
4526 if (PL_lex_brackets <= 0)
4527 yyerror("Unmatched right curly bracket");
4529 PL_expect = (expectation)PL_lex_brackstack[--PL_lex_brackets];
4530 if (PL_lex_brackets < PL_lex_formbrack && PL_lex_state != LEX_INTERPNORMAL)
4531 PL_lex_formbrack = 0;
4532 if (PL_lex_state == LEX_INTERPNORMAL) {
4533 if (PL_lex_brackets == 0) {
4534 if (PL_expect & XFAKEBRACK) {
4535 PL_expect &= XENUMMASK;
4536 PL_lex_state = LEX_INTERPEND;
4541 PL_thiswhite = newSVpvs("");
4542 sv_catpvn(PL_thiswhite,"}",1);
4545 return yylex(); /* ignore fake brackets */
4547 if (*s == '-' && s[1] == '>')
4548 PL_lex_state = LEX_INTERPENDMAYBE;
4549 else if (*s != '[' && *s != '{')
4550 PL_lex_state = LEX_INTERPEND;
4553 if (PL_expect & XFAKEBRACK) {
4554 PL_expect &= XENUMMASK;
4556 return yylex(); /* ignore fake brackets */
4558 start_force(PL_curforce);
4560 curmad('X', newSVpvn(s-1,1));
4561 CURMAD('_', PL_thiswhite);
4566 PL_thistoken = newSVpvs("");
4574 if (PL_expect == XOPERATOR) {
4575 if (PL_bufptr == PL_linestart && ckWARN(WARN_SEMICOLON)
4576 && isIDFIRST_lazy_if(s,UTF))
4578 CopLINE_dec(PL_curcop);
4579 Perl_warner(aTHX_ packWARN(WARN_SEMICOLON), PL_warn_nosemi);
4580 CopLINE_inc(PL_curcop);
4585 s = scan_ident(s - 1, PL_bufend, PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
4587 PL_expect = XOPERATOR;
4588 force_ident(PL_tokenbuf, '&');
4592 yylval.ival = (OPpENTERSUB_AMPER<<8);
4604 const char tmp = *s++;
4611 if (tmp && isSPACE(*s) && ckWARN(WARN_SYNTAX)
4612 && strchr("+-*/%.^&|<",tmp))
4613 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
4614 "Reversed %c= operator",(int)tmp);
4616 if (PL_expect == XSTATE && isALPHA(tmp) &&
4617 (s == PL_linestart+1 || s[-2] == '\n') )
4619 if (PL_in_eval && !PL_rsfp) {
4624 if (strnEQ(s,"=cut",4)) {
4640 PL_thiswhite = newSVpvs("");
4641 sv_catpvn(PL_thiswhite, PL_linestart,
4642 PL_bufend - PL_linestart);
4646 PL_doextract = TRUE;
4650 if (PL_lex_brackets < PL_lex_formbrack) {
4652 #ifdef PERL_STRICT_CR
4653 while (SPACE_OR_TAB(*t))
4655 while (SPACE_OR_TAB(*t) || *t == '\r')
4658 if (*t == '\n' || *t == '#') {
4669 const char tmp = *s++;
4671 /* was this !=~ where !~ was meant?
4672 * warn on m:!=~\s+([/?]|[msy]\W|tr\W): */
4674 if (*s == '~' && ckWARN(WARN_SYNTAX)) {
4675 const char *t = s+1;
4677 while (t < PL_bufend && isSPACE(*t))
4680 if (*t == '/' || *t == '?' ||
4681 ((*t == 'm' || *t == 's' || *t == 'y')
4682 && !isALNUM(t[1])) ||
4683 (*t == 't' && t[1] == 'r' && !isALNUM(t[2])))
4684 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
4685 "!=~ should be !~");
4695 if (PL_expect != XOPERATOR) {
4696 if (s[1] != '<' && !strchr(s,'>'))
4699 s = scan_heredoc(s);
4701 s = scan_inputsymbol(s);
4702 TERM(sublex_start());
4708 SHop(OP_LEFT_SHIFT);
4722 const char tmp = *s++;
4724 SHop(OP_RIGHT_SHIFT);
4725 else if (tmp == '=')
4734 if (PL_expect == XOPERATOR) {
4735 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
4737 deprecate_old(commaless_variable_list);
4738 return REPORT(','); /* grandfather non-comma-format format */
4742 if (s[1] == '#' && (isIDFIRST_lazy_if(s+2,UTF) || strchr("{$:+-", s[2]))) {
4743 PL_tokenbuf[0] = '@';
4744 s = scan_ident(s + 1, PL_bufend, PL_tokenbuf + 1,
4745 sizeof PL_tokenbuf - 1, FALSE);
4746 if (PL_expect == XOPERATOR)
4747 no_op("Array length", s);
4748 if (!PL_tokenbuf[1])
4750 PL_expect = XOPERATOR;
4751 PL_pending_ident = '#';
4755 PL_tokenbuf[0] = '$';
4756 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1,
4757 sizeof PL_tokenbuf - 1, FALSE);
4758 if (PL_expect == XOPERATOR)
4760 if (!PL_tokenbuf[1]) {
4762 yyerror("Final $ should be \\$ or $name");
4766 /* This kludge not intended to be bulletproof. */
4767 if (PL_tokenbuf[1] == '[' && !PL_tokenbuf[2]) {
4768 yylval.opval = newSVOP(OP_CONST, 0,
4769 newSViv(CopARYBASE_get(&PL_compiling)));
4770 yylval.opval->op_private = OPpCONST_ARYBASE;
4776 const char tmp = *s;
4777 if (PL_lex_state == LEX_NORMAL)
4780 if ((PL_expect != XREF || PL_oldoldbufptr == PL_last_lop)
4781 && intuit_more(s)) {
4783 PL_tokenbuf[0] = '@';
4784 if (ckWARN(WARN_SYNTAX)) {
4787 while (isSPACE(*t) || isALNUM_lazy_if(t,UTF) || *t == '$')
4790 PL_bufptr = PEEKSPACE(PL_bufptr); /* XXX can realloc */
4791 while (t < PL_bufend && *t != ']')
4793 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
4794 "Multidimensional syntax %.*s not supported",
4795 (int)((t - PL_bufptr) + 1), PL_bufptr);
4799 else if (*s == '{') {
4801 PL_tokenbuf[0] = '%';
4802 if (strEQ(PL_tokenbuf+1, "SIG") && ckWARN(WARN_SYNTAX)
4803 && (t = strchr(s, '}')) && (t = strchr(t, '=')))
4805 char tmpbuf[sizeof PL_tokenbuf];
4808 } while (isSPACE(*t));
4809 if (isIDFIRST_lazy_if(t,UTF)) {
4811 t = scan_word(t, tmpbuf, sizeof tmpbuf, TRUE,
4815 if (*t == ';' && get_cvn_flags(tmpbuf, len, 0))
4816 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
4817 "You need to quote \"%s\"",
4824 PL_expect = XOPERATOR;
4825 if (PL_lex_state == LEX_NORMAL && isSPACE((char)tmp)) {
4826 const bool islop = (PL_last_lop == PL_oldoldbufptr);
4827 if (!islop || PL_last_lop_op == OP_GREPSTART)
4828 PL_expect = XOPERATOR;
4829 else if (strchr("$@\"'`q", *s))
4830 PL_expect = XTERM; /* e.g. print $fh "foo" */
4831 else if (strchr("&*<%", *s) && isIDFIRST_lazy_if(s+1,UTF))
4832 PL_expect = XTERM; /* e.g. print $fh &sub */
4833 else if (isIDFIRST_lazy_if(s,UTF)) {
4834 char tmpbuf[sizeof PL_tokenbuf];
4836 scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
4837 if ((t2 = keyword(tmpbuf, len, 0))) {
4838 /* binary operators exclude handle interpretations */
4850 PL_expect = XTERM; /* e.g. print $fh length() */
4855 PL_expect = XTERM; /* e.g. print $fh subr() */
4858 else if (isDIGIT(*s))
4859 PL_expect = XTERM; /* e.g. print $fh 3 */
4860 else if (*s == '.' && isDIGIT(s[1]))
4861 PL_expect = XTERM; /* e.g. print $fh .3 */
4862 else if ((*s == '?' || *s == '-' || *s == '+')
4863 && !isSPACE(s[1]) && s[1] != '=')
4864 PL_expect = XTERM; /* e.g. print $fh -1 */
4865 else if (*s == '/' && !isSPACE(s[1]) && s[1] != '='
4867 PL_expect = XTERM; /* e.g. print $fh /.../
4868 XXX except DORDOR operator
4870 else if (*s == '<' && s[1] == '<' && !isSPACE(s[2])
4872 PL_expect = XTERM; /* print $fh <<"EOF" */
4875 PL_pending_ident = '$';
4879 if (PL_expect == XOPERATOR)
4881 PL_tokenbuf[0] = '@';
4882 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1, FALSE);
4883 if (!PL_tokenbuf[1]) {
4886 if (PL_lex_state == LEX_NORMAL)
4888 if ((PL_expect != XREF || PL_oldoldbufptr == PL_last_lop) && intuit_more(s)) {
4890 PL_tokenbuf[0] = '%';
4892 /* Warn about @ where they meant $. */
4893 if (*s == '[' || *s == '{') {
4894 if (ckWARN(WARN_SYNTAX)) {
4895 const char *t = s + 1;
4896 while (*t && (isALNUM_lazy_if(t,UTF) || strchr(" \t$#+-'\"", *t)))
4898 if (*t == '}' || *t == ']') {
4900 PL_bufptr = PEEKSPACE(PL_bufptr); /* XXX can realloc */
4901 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
4902 "Scalar value %.*s better written as $%.*s",
4903 (int)(t-PL_bufptr), PL_bufptr,
4904 (int)(t-PL_bufptr-1), PL_bufptr+1);
4909 PL_pending_ident = '@';
4912 case '/': /* may be division, defined-or, or pattern */
4913 if (PL_expect == XTERMORDORDOR && s[1] == '/') {
4917 case '?': /* may either be conditional or pattern */
4918 if(PL_expect == XOPERATOR) {
4926 /* A // operator. */
4936 /* Disable warning on "study /blah/" */
4937 if (PL_oldoldbufptr == PL_last_uni
4938 && (*PL_last_uni != 's' || s - PL_last_uni < 5
4939 || memNE(PL_last_uni, "study", 5)
4940 || isALNUM_lazy_if(PL_last_uni+5,UTF)
4943 s = scan_pat(s,OP_MATCH);
4944 TERM(sublex_start());
4948 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack
4949 #ifdef PERL_STRICT_CR
4952 && (s[1] == '\n' || (s[1] == '\r' && s[2] == '\n'))
4954 && (s == PL_linestart || s[-1] == '\n') )
4956 PL_lex_formbrack = 0;
4960 if (PL_expect == XOPERATOR || !isDIGIT(s[1])) {
4966 yylval.ival = OPf_SPECIAL;
4972 if (PL_expect != XOPERATOR)
4977 case '0': case '1': case '2': case '3': case '4':
4978 case '5': case '6': case '7': case '8': case '9':
4979 s = scan_num(s, &yylval);
4980 DEBUG_T( { printbuf("### Saw number in %s\n", s); } );
4981 if (PL_expect == XOPERATOR)
4986 s = scan_str(s,!!PL_madskills,FALSE);
4987 DEBUG_T( { printbuf("### Saw string before %s\n", s); } );
4988 if (PL_expect == XOPERATOR) {
4989 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
4991 deprecate_old(commaless_variable_list);
4992 return REPORT(','); /* grandfather non-comma-format format */
4999 yylval.ival = OP_CONST;
5000 TERM(sublex_start());
5003 s = scan_str(s,!!PL_madskills,FALSE);
5004 DEBUG_T( { printbuf("### Saw string before %s\n", s); } );
5005 if (PL_expect == XOPERATOR) {
5006 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
5008 deprecate_old(commaless_variable_list);
5009 return REPORT(','); /* grandfather non-comma-format format */
5016 yylval.ival = OP_CONST;
5017 /* FIXME. I think that this can be const if char *d is replaced by
5018 more localised variables. */
5019 for (d = SvPV(PL_lex_stuff, len); len; len--, d++) {
5020 if (*d == '$' || *d == '@' || *d == '\\' || !UTF8_IS_INVARIANT((U8)*d)) {
5021 yylval.ival = OP_STRINGIFY;
5025 TERM(sublex_start());
5028 s = scan_str(s,!!PL_madskills,FALSE);
5029 DEBUG_T( { printbuf("### Saw backtick string before %s\n", s); } );
5030 if (PL_expect == XOPERATOR)
5031 no_op("Backticks",s);
5034 readpipe_override();
5035 TERM(sublex_start());
5039 if (PL_lex_inwhat && isDIGIT(*s) && ckWARN(WARN_SYNTAX))
5040 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),"Can't use \\%c to mean $%c in expression",
5042 if (PL_expect == XOPERATOR)
5043 no_op("Backslash",s);
5047 if (isDIGIT(s[1]) && PL_expect != XOPERATOR) {
5048 char *start = s + 2;
5049 while (isDIGIT(*start) || *start == '_')
5051 if (*start == '.' && isDIGIT(start[1])) {
5052 s = scan_num(s, &yylval);
5055 /* avoid v123abc() or $h{v1}, allow C<print v10;> */
5056 else if (!isALPHA(*start) && (PL_expect == XTERM
5057 || PL_expect == XREF || PL_expect == XSTATE
5058 || PL_expect == XTERMORDORDOR)) {
5059 /* XXX Use gv_fetchpvn rather than stomping on a const string */
5060 const char c = *start;
5063 gv = gv_fetchpv(s, 0, SVt_PVCV);
5066 s = scan_num(s, &yylval);
5073 if (isDIGIT(s[1]) && PL_expect == XOPERATOR) {
5115 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
5117 /* Some keywords can be followed by any delimiter, including ':' */
5118 tmp = ((len == 1 && strchr("msyq", PL_tokenbuf[0])) ||
5119 (len == 2 && ((PL_tokenbuf[0] == 't' && PL_tokenbuf[1] == 'r') ||
5120 (PL_tokenbuf[0] == 'q' &&
5121 strchr("qwxr", PL_tokenbuf[1])))));
5123 /* x::* is just a word, unless x is "CORE" */
5124 if (!tmp && *s == ':' && s[1] == ':' && strNE(PL_tokenbuf, "CORE"))
5128 while (d < PL_bufend && isSPACE(*d))
5129 d++; /* no comments skipped here, or s### is misparsed */
5131 /* Is this a label? */
5132 if (!tmp && PL_expect == XSTATE
5133 && d < PL_bufend && *d == ':' && *(d + 1) != ':') {
5135 yylval.pval = CopLABEL_alloc(PL_tokenbuf);
5140 /* Check for keywords */
5141 tmp = keyword(PL_tokenbuf, len, 0);
5143 /* Is this a word before a => operator? */
5144 if (*d == '=' && d[1] == '>') {
5147 = (OP*)newSVOP(OP_CONST, 0,
5148 S_newSV_maybe_utf8(aTHX_ PL_tokenbuf, len));
5149 yylval.opval->op_private = OPpCONST_BARE;
5153 if (tmp < 0) { /* second-class keyword? */
5154 GV *ogv = NULL; /* override (winner) */
5155 GV *hgv = NULL; /* hidden (loser) */
5156 if (PL_expect != XOPERATOR && (*s != ':' || s[1] != ':')) {
5158 if ((gv = gv_fetchpvn_flags(PL_tokenbuf, len, 0, SVt_PVCV)) &&
5161 if (GvIMPORTED_CV(gv))
5163 else if (! CvMETHOD(cv))
5167 (gvp = (GV**)hv_fetch(PL_globalstash,PL_tokenbuf,len,FALSE)) &&
5168 (gv = *gvp) != (GV*)&PL_sv_undef &&
5169 GvCVu(gv) && GvIMPORTED_CV(gv))
5176 tmp = 0; /* overridden by import or by GLOBAL */
5179 && -tmp==KEY_lock /* XXX generalizable kludge */
5181 && !hv_fetchs(GvHVn(PL_incgv), "Thread.pm", FALSE))
5183 tmp = 0; /* any sub overrides "weak" keyword */
5185 else { /* no override */
5187 if (tmp == KEY_dump && ckWARN(WARN_MISC)) {
5188 Perl_warner(aTHX_ packWARN(WARN_MISC),
5189 "dump() better written as CORE::dump()");
5193 if (hgv && tmp != KEY_x && tmp != KEY_CORE
5194 && ckWARN(WARN_AMBIGUOUS)) /* never ambiguous */
5195 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
5196 "Ambiguous call resolved as CORE::%s(), %s",
5197 GvENAME(hgv), "qualify as such or use &");
5204 default: /* not a keyword */
5205 /* Trade off - by using this evil construction we can pull the
5206 variable gv into the block labelled keylookup. If not, then
5207 we have to give it function scope so that the goto from the
5208 earlier ':' case doesn't bypass the initialisation. */
5210 just_a_word_zero_gv:
5218 const char lastchar = (PL_bufptr == PL_oldoldbufptr ? 0 : PL_bufptr[-1]);
5221 SV *nextPL_nextwhite = 0;
5225 /* Get the rest if it looks like a package qualifier */
5227 if (*s == '\'' || (*s == ':' && s[1] == ':')) {
5229 s = scan_word(s, PL_tokenbuf + len, sizeof PL_tokenbuf - len,
5232 Perl_croak(aTHX_ "Bad name after %s%s", PL_tokenbuf,
5233 *s == '\'' ? "'" : "::");
5238 if (PL_expect == XOPERATOR) {
5239 if (PL_bufptr == PL_linestart) {
5240 CopLINE_dec(PL_curcop);
5241 Perl_warner(aTHX_ packWARN(WARN_SEMICOLON), PL_warn_nosemi);
5242 CopLINE_inc(PL_curcop);
5245 no_op("Bareword",s);
5248 /* Look for a subroutine with this name in current package,
5249 unless name is "Foo::", in which case Foo is a bearword
5250 (and a package name). */
5252 if (len > 2 && !PL_madskills &&
5253 PL_tokenbuf[len - 2] == ':' && PL_tokenbuf[len - 1] == ':')
5255 if (ckWARN(WARN_BAREWORD)
5256 && ! gv_fetchpvn_flags(PL_tokenbuf, len, 0, SVt_PVHV))
5257 Perl_warner(aTHX_ packWARN(WARN_BAREWORD),
5258 "Bareword \"%s\" refers to nonexistent package",
5261 PL_tokenbuf[len] = '\0';
5267 /* Mustn't actually add anything to a symbol table.
5268 But also don't want to "initialise" any placeholder
5269 constants that might already be there into full
5270 blown PVGVs with attached PVCV. */
5271 gv = gv_fetchpvn_flags(PL_tokenbuf, len,
5272 GV_NOADD_NOINIT, SVt_PVCV);
5277 /* if we saw a global override before, get the right name */
5280 sv = newSVpvs("CORE::GLOBAL::");
5281 sv_catpv(sv,PL_tokenbuf);
5284 /* If len is 0, newSVpv does strlen(), which is correct.
5285 If len is non-zero, then it will be the true length,
5286 and so the scalar will be created correctly. */
5287 sv = newSVpv(PL_tokenbuf,len);
5290 if (PL_madskills && !PL_thistoken) {
5291 char *start = SvPVX(PL_linestr) + PL_realtokenstart;
5292 PL_thistoken = newSVpv(start,s - start);
5293 PL_realtokenstart = s - SvPVX(PL_linestr);
5297 /* Presume this is going to be a bareword of some sort. */
5300 yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
5301 yylval.opval->op_private = OPpCONST_BARE;
5302 /* UTF-8 package name? */
5303 if (UTF && !IN_BYTES &&
5304 is_utf8_string((U8*)SvPVX_const(sv), SvCUR(sv)))
5307 /* And if "Foo::", then that's what it certainly is. */
5312 /* Do the explicit type check so that we don't need to force
5313 the initialisation of the symbol table to have a real GV.
5314 Beware - gv may not really be a PVGV, cv may not really be
5315 a PVCV, (because of the space optimisations that gv_init
5316 understands) But they're true if for this symbol there is
5317 respectively a typeglob and a subroutine.
5319 cv = gv ? ((SvTYPE(gv) == SVt_PVGV)
5320 /* Real typeglob, so get the real subroutine: */
5322 /* A proxy for a subroutine in this package? */
5323 : SvOK(gv) ? (CV *) gv : NULL)
5326 /* See if it's the indirect object for a list operator. */
5328 if (PL_oldoldbufptr &&
5329 PL_oldoldbufptr < PL_bufptr &&
5330 (PL_oldoldbufptr == PL_last_lop
5331 || PL_oldoldbufptr == PL_last_uni) &&
5332 /* NO SKIPSPACE BEFORE HERE! */
5333 (PL_expect == XREF ||
5334 ((PL_opargs[PL_last_lop_op] >> OASHIFT)& 7) == OA_FILEREF))
5336 bool immediate_paren = *s == '(';
5338 /* (Now we can afford to cross potential line boundary.) */
5339 s = SKIPSPACE2(s,nextPL_nextwhite);
5341 PL_nextwhite = nextPL_nextwhite; /* assume no & deception */
5344 /* Two barewords in a row may indicate method call. */
5346 if ((isIDFIRST_lazy_if(s,UTF) || *s == '$') &&
5347 (tmp = intuit_method(s, gv, cv)))
5350 /* If not a declared subroutine, it's an indirect object. */
5351 /* (But it's an indir obj regardless for sort.) */
5352 /* Also, if "_" follows a filetest operator, it's a bareword */
5355 ( !immediate_paren && (PL_last_lop_op == OP_SORT ||
5357 (PL_last_lop_op != OP_MAPSTART &&
5358 PL_last_lop_op != OP_GREPSTART))))
5359 || (PL_tokenbuf[0] == '_' && PL_tokenbuf[1] == '\0'
5360 && ((PL_opargs[PL_last_lop_op] & OA_CLASS_MASK) == OA_FILESTATOP))
5363 PL_expect = (PL_last_lop == PL_oldoldbufptr) ? XTERM : XOPERATOR;
5368 PL_expect = XOPERATOR;
5371 s = SKIPSPACE2(s,nextPL_nextwhite);
5372 PL_nextwhite = nextPL_nextwhite;
5377 /* Is this a word before a => operator? */
5378 if (*s == '=' && s[1] == '>' && !pkgname) {
5380 sv_setpv(((SVOP*)yylval.opval)->op_sv, PL_tokenbuf);
5381 if (UTF && !IN_BYTES && is_utf8_string((U8*)PL_tokenbuf, len))
5382 SvUTF8_on(((SVOP*)yylval.opval)->op_sv);
5386 /* If followed by a paren, it's certainly a subroutine. */
5391 while (SPACE_OR_TAB(*d))
5393 if (*d == ')' && (sv = gv_const_sv(gv))) {
5397 char *par = SvPVX(PL_linestr) + PL_realtokenstart;
5398 sv_catpvn(PL_thistoken, par, s - par);
5400 sv_free(PL_nextwhite);
5410 PL_nextwhite = PL_thiswhite;
5413 start_force(PL_curforce);
5415 NEXTVAL_NEXTTOKE.opval = yylval.opval;
5416 PL_expect = XOPERATOR;
5419 PL_nextwhite = nextPL_nextwhite;
5420 curmad('X', PL_thistoken);
5421 PL_thistoken = newSVpvs("");
5429 /* If followed by var or block, call it a method (unless sub) */
5431 if ((*s == '$' || *s == '{') && (!gv || !cv)) {
5432 PL_last_lop = PL_oldbufptr;
5433 PL_last_lop_op = OP_METHOD;
5437 /* If followed by a bareword, see if it looks like indir obj. */
5440 && (isIDFIRST_lazy_if(s,UTF) || *s == '$')
5441 && (tmp = intuit_method(s, gv, cv)))
5444 /* Not a method, so call it a subroutine (if defined) */
5447 if (lastchar == '-' && ckWARN_d(WARN_AMBIGUOUS))
5448 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
5449 "Ambiguous use of -%s resolved as -&%s()",
5450 PL_tokenbuf, PL_tokenbuf);
5451 /* Check for a constant sub */
5452 if ((sv = gv_const_sv(gv))) {
5454 SvREFCNT_dec(((SVOP*)yylval.opval)->op_sv);
5455 ((SVOP*)yylval.opval)->op_sv = SvREFCNT_inc_simple(sv);
5456 yylval.opval->op_private = 0;
5460 /* Resolve to GV now. */
5461 if (SvTYPE(gv) != SVt_PVGV) {
5462 gv = gv_fetchpv(PL_tokenbuf, 0, SVt_PVCV);
5463 assert (SvTYPE(gv) == SVt_PVGV);
5464 /* cv must have been some sort of placeholder, so
5465 now needs replacing with a real code reference. */
5469 op_free(yylval.opval);
5470 yylval.opval = newCVREF(0, newGVOP(OP_GV, 0, gv));
5471 yylval.opval->op_private |= OPpENTERSUB_NOPAREN;
5472 PL_last_lop = PL_oldbufptr;
5473 PL_last_lop_op = OP_ENTERSUB;
5474 /* Is there a prototype? */
5482 const char *proto = SvPV_const((SV*)cv, protolen);
5485 if ((*proto == '$' || *proto == '_') && proto[1] == '\0')
5487 while (*proto == ';')
5489 if (*proto == '&' && *s == '{') {
5490 sv_setpv(PL_subname,
5493 "__ANON__" : "__ANON__::__ANON__"));
5500 PL_nextwhite = PL_thiswhite;
5503 start_force(PL_curforce);
5504 NEXTVAL_NEXTTOKE.opval = yylval.opval;
5507 PL_nextwhite = nextPL_nextwhite;
5508 curmad('X', PL_thistoken);
5509 PL_thistoken = newSVpvs("");
5516 /* Guess harder when madskills require "best effort". */
5517 if (PL_madskills && (!gv || !GvCVu(gv))) {
5518 int probable_sub = 0;
5519 if (strchr("\"'`$@%0123456789!*+{[<", *s))
5521 else if (isALPHA(*s)) {
5525 d = scan_word(d, tmpbuf, sizeof tmpbuf, TRUE, &tmplen);
5526 if (!keyword(tmpbuf, tmplen, 0))
5529 while (d < PL_bufend && isSPACE(*d))
5531 if (*d == '=' && d[1] == '>')
5536 gv = gv_fetchpv(PL_tokenbuf, GV_ADD, SVt_PVCV);
5537 op_free(yylval.opval);
5538 yylval.opval = newCVREF(0, newGVOP(OP_GV, 0, gv));
5539 yylval.opval->op_private |= OPpENTERSUB_NOPAREN;
5540 PL_last_lop = PL_oldbufptr;
5541 PL_last_lop_op = OP_ENTERSUB;
5542 PL_nextwhite = PL_thiswhite;
5544 start_force(PL_curforce);
5545 NEXTVAL_NEXTTOKE.opval = yylval.opval;
5547 PL_nextwhite = nextPL_nextwhite;
5548 curmad('X', PL_thistoken);
5549 PL_thistoken = newSVpvs("");
5554 NEXTVAL_NEXTTOKE.opval = yylval.opval;
5561 /* Call it a bare word */
5563 if (PL_hints & HINT_STRICT_SUBS)
5564 yylval.opval->op_private |= OPpCONST_STRICT;
5567 if (lastchar != '-') {
5568 if (ckWARN(WARN_RESERVED)) {
5572 if (!*d && !gv_stashpv(PL_tokenbuf, 0))
5573 Perl_warner(aTHX_ packWARN(WARN_RESERVED), PL_warn_reserved,
5580 if ((lastchar == '*' || lastchar == '%' || lastchar == '&')
5581 && ckWARN_d(WARN_AMBIGUOUS)) {
5582 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
5583 "Operator or semicolon missing before %c%s",
5584 lastchar, PL_tokenbuf);
5585 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
5586 "Ambiguous use of %c resolved as operator %c",
5587 lastchar, lastchar);
5593 yylval.opval = (OP*)newSVOP(OP_CONST, 0,
5594 newSVpv(CopFILE(PL_curcop),0));
5598 yylval.opval = (OP*)newSVOP(OP_CONST, 0,
5599 Perl_newSVpvf(aTHX_ "%"IVdf, (IV)CopLINE(PL_curcop)));
5602 case KEY___PACKAGE__:
5603 yylval.opval = (OP*)newSVOP(OP_CONST, 0,
5605 ? newSVhek(HvNAME_HEK(PL_curstash))
5612 if (PL_rsfp && (!PL_in_eval || PL_tokenbuf[2] == 'D')) {
5613 const char *pname = "main";
5614 if (PL_tokenbuf[2] == 'D')
5615 pname = HvNAME_get(PL_curstash ? PL_curstash : PL_defstash);
5616 gv = gv_fetchpv(Perl_form(aTHX_ "%s::DATA", pname), GV_ADD,
5620 GvIOp(gv) = newIO();
5621 IoIFP(GvIOp(gv)) = PL_rsfp;
5622 #if defined(HAS_FCNTL) && defined(F_SETFD)
5624 const int fd = PerlIO_fileno(PL_rsfp);
5625 fcntl(fd,F_SETFD,fd >= 3);
5628 /* Mark this internal pseudo-handle as clean */
5629 IoFLAGS(GvIOp(gv)) |= IOf_UNTAINT;
5631 IoTYPE(GvIOp(gv)) = IoTYPE_PIPE;
5632 else if ((PerlIO*)PL_rsfp == PerlIO_stdin())
5633 IoTYPE(GvIOp(gv)) = IoTYPE_STD;
5635 IoTYPE(GvIOp(gv)) = IoTYPE_RDONLY;
5636 #if defined(WIN32) && !defined(PERL_TEXTMODE_SCRIPTS)
5637 /* if the script was opened in binmode, we need to revert
5638 * it to text mode for compatibility; but only iff it has CRs
5639 * XXX this is a questionable hack at best. */
5640 if (PL_bufend-PL_bufptr > 2
5641 && PL_bufend[-1] == '\n' && PL_bufend[-2] == '\r')
5644 if (IoTYPE(GvIOp(gv)) == IoTYPE_RDONLY) {
5645 loc = PerlIO_tell(PL_rsfp);
5646 (void)PerlIO_seek(PL_rsfp, 0L, 0);
5649 if (PerlLIO_setmode(PL_rsfp, O_TEXT) != -1) {
5651 if (PerlLIO_setmode(PerlIO_fileno(PL_rsfp), O_TEXT) != -1) {
5652 #endif /* NETWARE */
5653 #ifdef PERLIO_IS_STDIO /* really? */
5654 # if defined(__BORLANDC__)
5655 /* XXX see note in do_binmode() */
5656 ((FILE*)PL_rsfp)->flags &= ~_F_BIN;
5660 PerlIO_seek(PL_rsfp, loc, 0);
5664 #ifdef PERLIO_LAYERS
5667 PerlIO_apply_layers(aTHX_ PL_rsfp, NULL, ":utf8");
5668 else if (PL_encoding) {
5675 XPUSHs(PL_encoding);
5677 call_method("name", G_SCALAR);
5681 PerlIO_apply_layers(aTHX_ PL_rsfp, NULL,
5682 Perl_form(aTHX_ ":encoding(%"SVf")",
5691 if (PL_realtokenstart >= 0) {
5692 char *tstart = SvPVX(PL_linestr) + PL_realtokenstart;
5694 PL_endwhite = newSVpvs("");
5695 sv_catsv(PL_endwhite, PL_thiswhite);
5697 sv_catpvn(PL_endwhite, tstart, PL_bufend - tstart);
5698 PL_realtokenstart = -1;
5700 while ((s = filter_gets(PL_endwhite, PL_rsfp,
5701 SvCUR(PL_endwhite))) != Nullch) ;
5716 if (PL_expect == XSTATE) {
5723 if (*s == ':' && s[1] == ':') {
5726 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
5727 if (!(tmp = keyword(PL_tokenbuf, len, 0)))
5728 Perl_croak(aTHX_ "CORE::%s is not a keyword", PL_tokenbuf);
5731 else if (tmp == KEY_require || tmp == KEY_do)
5732 /* that's a way to remember we saw "CORE::" */
5745 LOP(OP_ACCEPT,XTERM);
5751 LOP(OP_ATAN2,XTERM);
5757 LOP(OP_BINMODE,XTERM);
5760 LOP(OP_BLESS,XTERM);
5769 /* When 'use switch' is in effect, continue has a dual
5770 life as a control operator. */
5772 if (!FEATURE_IS_ENABLED("switch"))
5775 /* We have to disambiguate the two senses of
5776 "continue". If the next token is a '{' then
5777 treat it as the start of a continue block;
5778 otherwise treat it as a control operator.
5790 (void)gv_fetchpvs("ENV", GV_ADD|GV_NOTQUAL, SVt_PVHV);
5807 if (!PL_cryptseen) {
5808 PL_cryptseen = TRUE;
5812 LOP(OP_CRYPT,XTERM);
5815 LOP(OP_CHMOD,XTERM);
5818 LOP(OP_CHOWN,XTERM);
5821 LOP(OP_CONNECT,XTERM);
5840 s = force_word(s,WORD,TRUE,TRUE,FALSE);
5841 if (orig_keyword == KEY_do) {
5850 PL_hints |= HINT_BLOCK_SCOPE;
5860 gv_fetchpvs("AnyDBM_File::ISA", GV_ADDMULTI, SVt_PVAV);
5861 LOP(OP_DBMOPEN,XTERM);
5867 s = force_word(s,WORD,TRUE,FALSE,FALSE);
5874 yylval.ival = CopLINE(PL_curcop);
5890 PL_expect = (*s == '{') ? XTERMBLOCK : XTERM;
5891 UNIBRACK(OP_ENTEREVAL);
5909 case KEY_endhostent:
5915 case KEY_endservent:
5918 case KEY_endprotoent:
5929 yylval.ival = CopLINE(PL_curcop);
5931 if (PL_expect == XSTATE && isIDFIRST_lazy_if(s,UTF)) {
5934 int soff = s - SvPVX(PL_linestr); /* for skipspace realloc */
5937 if ((PL_bufend - p) >= 3 &&
5938 strnEQ(p, "my", 2) && isSPACE(*(p + 2)))
5940 else if ((PL_bufend - p) >= 4 &&
5941 strnEQ(p, "our", 3) && isSPACE(*(p + 3)))
5944 if (isIDFIRST_lazy_if(p,UTF)) {
5945 p = scan_ident(p, PL_bufend,
5946 PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
5950 Perl_croak(aTHX_ "Missing $ on loop variable");
5952 s = SvPVX(PL_linestr) + soff;
5958 LOP(OP_FORMLINE,XTERM);
5964 LOP(OP_FCNTL,XTERM);
5970 LOP(OP_FLOCK,XTERM);
5979 LOP(OP_GREPSTART, XREF);
5982 s = force_word(s,WORD,TRUE,FALSE,FALSE);
5997 case KEY_getpriority:
5998 LOP(OP_GETPRIORITY,XTERM);
6000 case KEY_getprotobyname:
6003 case KEY_getprotobynumber:
6004 LOP(OP_GPBYNUMBER,XTERM);
6006 case KEY_getprotoent:
6018 case KEY_getpeername:
6019 UNI(OP_GETPEERNAME);
6021 case KEY_gethostbyname:
6024 case KEY_gethostbyaddr:
6025 LOP(OP_GHBYADDR,XTERM);
6027 case KEY_gethostent:
6030 case KEY_getnetbyname:
6033 case KEY_getnetbyaddr:
6034 LOP(OP_GNBYADDR,XTERM);
6039 case KEY_getservbyname:
6040 LOP(OP_GSBYNAME,XTERM);
6042 case KEY_getservbyport:
6043 LOP(OP_GSBYPORT,XTERM);
6045 case KEY_getservent:
6048 case KEY_getsockname:
6049 UNI(OP_GETSOCKNAME);
6051 case KEY_getsockopt:
6052 LOP(OP_GSOCKOPT,XTERM);
6067 yylval.ival = CopLINE(PL_curcop);
6078 yylval.ival = CopLINE(PL_curcop);
6082 LOP(OP_INDEX,XTERM);
6088 LOP(OP_IOCTL,XTERM);
6100 s = force_word(s,WORD,TRUE,FALSE,FALSE);
6132 LOP(OP_LISTEN,XTERM);
6141 s = scan_pat(s,OP_MATCH);
6142 TERM(sublex_start());
6145 LOP(OP_MAPSTART, XREF);
6148 LOP(OP_MKDIR,XTERM);
6151 LOP(OP_MSGCTL,XTERM);
6154 LOP(OP_MSGGET,XTERM);
6157 LOP(OP_MSGRCV,XTERM);
6160 LOP(OP_MSGSND,XTERM);
6167 if (isIDFIRST_lazy_if(s,UTF)) {
6171 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, TRUE, &len);
6172 if (len == 3 && strnEQ(PL_tokenbuf, "sub", 3))
6174 PL_in_my_stash = find_in_my_stash(PL_tokenbuf, len);
6175 if (!PL_in_my_stash) {
6178 my_snprintf(tmpbuf, sizeof(tmpbuf), "No such class %.1000s", PL_tokenbuf);
6182 if (PL_madskills) { /* just add type to declarator token */
6183 sv_catsv(PL_thistoken, PL_nextwhite);
6185 sv_catpvn(PL_thistoken, start, s - start);
6193 s = force_word(s,WORD,TRUE,FALSE,FALSE);
6200 s = tokenize_use(0, s);
6204 if (*s == '(' || (s = SKIPSPACE1(s), *s == '('))
6211 if (isIDFIRST_lazy_if(s,UTF)) {
6213 for (d = s; isALNUM_lazy_if(d,UTF);)
6215 for (t=d; isSPACE(*t);)
6217 if ( *t && strchr("|&*+-=!?:.", *t) && ckWARN_d(WARN_PRECEDENCE)
6219 && !(t[0] == '=' && t[1] == '>')
6221 int parms_len = (int)(d-s);
6222 Perl_warner(aTHX_ packWARN(WARN_PRECEDENCE),
6223 "Precedence problem: open %.*s should be open(%.*s)",
6224 parms_len, s, parms_len, s);
6230 yylval.ival = OP_OR;
6240 LOP(OP_OPEN_DIR,XTERM);
6243 checkcomma(s,PL_tokenbuf,"filehandle");
6247 checkcomma(s,PL_tokenbuf,"filehandle");
6266 s = force_word(s,WORD,FALSE,TRUE,FALSE);
6270 LOP(OP_PIPE_OP,XTERM);
6273 s = scan_str(s,!!PL_madskills,FALSE);
6276 yylval.ival = OP_CONST;
6277 TERM(sublex_start());
6283 s = scan_str(s,!!PL_madskills,FALSE);
6286 PL_expect = XOPERATOR;
6288 if (SvCUR(PL_lex_stuff)) {
6291 d = SvPV_force(PL_lex_stuff, len);
6293 for (; isSPACE(*d) && len; --len, ++d)
6298 if (!warned && ckWARN(WARN_QW)) {
6299 for (; !isSPACE(*d) && len; --len, ++d) {
6301 Perl_warner(aTHX_ packWARN(WARN_QW),
6302 "Possible attempt to separate words with commas");
6305 else if (*d == '#') {
6306 Perl_warner(aTHX_ packWARN(WARN_QW),
6307 "Possible attempt to put comments in qw() list");
6313 for (; !isSPACE(*d) && len; --len, ++d)
6316 sv = newSVpvn(b, d-b);
6317 if (DO_UTF8(PL_lex_stuff))
6319 words = append_elem(OP_LIST, words,
6320 newSVOP(OP_CONST, 0, tokeq(sv)));
6324 start_force(PL_curforce);
6325 NEXTVAL_NEXTTOKE.opval = words;
6330 SvREFCNT_dec(PL_lex_stuff);
6331 PL_lex_stuff = NULL;
6337 s = scan_str(s,!!PL_madskills,FALSE);
6340 yylval.ival = OP_STRINGIFY;
6341 if (SvIVX(PL_lex_stuff) == '\'')
6342 SvIV_set(PL_lex_stuff, 0); /* qq'$foo' should intepolate */
6343 TERM(sublex_start());
6346 s = scan_pat(s,OP_QR);
6347 TERM(sublex_start());
6350 s = scan_str(s,!!PL_madskills,FALSE);
6353 readpipe_override();
6354 TERM(sublex_start());
6362 s = force_version(s, FALSE);
6364 else if (*s != 'v' || !isDIGIT(s[1])
6365 || (s = force_version(s, TRUE), *s == 'v'))
6367 *PL_tokenbuf = '\0';
6368 s = force_word(s,WORD,TRUE,TRUE,FALSE);
6369 if (isIDFIRST_lazy_if(PL_tokenbuf,UTF))
6370 gv_stashpvn(PL_tokenbuf, strlen(PL_tokenbuf), GV_ADD);
6372 yyerror("<> should be quotes");
6374 if (orig_keyword == KEY_require) {
6382 PL_last_uni = PL_oldbufptr;
6383 PL_last_lop_op = OP_REQUIRE;
6385 return REPORT( (int)REQUIRE );
6391 s = force_word(s,WORD,TRUE,FALSE,FALSE);
6395 LOP(OP_RENAME,XTERM);
6404 LOP(OP_RINDEX,XTERM);
6414 UNIDOR(OP_READLINE);
6427 LOP(OP_REVERSE,XTERM);
6430 UNIDOR(OP_READLINK);
6438 TERM(sublex_start());
6440 TOKEN(1); /* force error */
6443 checkcomma(s,PL_tokenbuf,"filehandle");
6453 LOP(OP_SELECT,XTERM);
6459 LOP(OP_SEMCTL,XTERM);
6462 LOP(OP_SEMGET,XTERM);
6465 LOP(OP_SEMOP,XTERM);
6471 LOP(OP_SETPGRP,XTERM);
6473 case KEY_setpriority:
6474 LOP(OP_SETPRIORITY,XTERM);
6476 case KEY_sethostent:
6482 case KEY_setservent:
6485 case KEY_setprotoent:
6495 LOP(OP_SEEKDIR,XTERM);
6497 case KEY_setsockopt:
6498 LOP(OP_SSOCKOPT,XTERM);
6504 LOP(OP_SHMCTL,XTERM);
6507 LOP(OP_SHMGET,XTERM);
6510 LOP(OP_SHMREAD,XTERM);
6513 LOP(OP_SHMWRITE,XTERM);
6516 LOP(OP_SHUTDOWN,XTERM);
6525 LOP(OP_SOCKET,XTERM);
6527 case KEY_socketpair:
6528 LOP(OP_SOCKPAIR,XTERM);
6531 checkcomma(s,PL_tokenbuf,"subroutine name");
6533 if (*s == ';' || *s == ')') /* probably a close */
6534 Perl_croak(aTHX_ "sort is now a reserved word");
6536 s = force_word(s,WORD,TRUE,TRUE,FALSE);
6540 LOP(OP_SPLIT,XTERM);
6543 LOP(OP_SPRINTF,XTERM);
6546 LOP(OP_SPLICE,XTERM);
6561 LOP(OP_SUBSTR,XTERM);
6567 char tmpbuf[sizeof PL_tokenbuf];
6568 SSize_t tboffset = 0;
6569 expectation attrful;
6570 bool have_name, have_proto;
6571 const int key = tmp;
6576 char *tstart = SvPVX(PL_linestr) + PL_realtokenstart;
6577 SV *subtoken = newSVpvn(tstart, s - tstart);
6581 s = SKIPSPACE2(s,tmpwhite);
6586 if (isIDFIRST_lazy_if(s,UTF) || *s == '\'' ||
6587 (*s == ':' && s[1] == ':'))
6594 attrful = XATTRBLOCK;
6595 /* remember buffer pos'n for later force_word */
6596 tboffset = s - PL_oldbufptr;
6597 d = scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
6600 nametoke = newSVpvn(s, d - s);
6602 if (strchr(tmpbuf, ':'))
6603 sv_setpv(PL_subname, tmpbuf);
6605 sv_setsv(PL_subname,PL_curstname);
6606 sv_catpvs(PL_subname,"::");
6607 sv_catpvn(PL_subname,tmpbuf,len);
6614 CURMAD('X', nametoke);
6615 CURMAD('_', tmpwhite);
6616 (void) force_word(PL_oldbufptr + tboffset, WORD,
6619 s = SKIPSPACE2(d,tmpwhite);
6626 Perl_croak(aTHX_ "Missing name in \"my sub\"");
6627 PL_expect = XTERMBLOCK;
6628 attrful = XATTRTERM;
6629 sv_setpvn(PL_subname,"?",1);
6633 if (key == KEY_format) {
6635 PL_lex_formbrack = PL_lex_brackets + 1;
6637 PL_thistoken = subtoken;
6641 (void) force_word(PL_oldbufptr + tboffset, WORD,
6647 /* Look for a prototype */
6650 bool bad_proto = FALSE;
6651 const bool warnsyntax = ckWARN(WARN_SYNTAX);
6653 s = scan_str(s,!!PL_madskills,FALSE);
6655 Perl_croak(aTHX_ "Prototype not terminated");
6656 /* strip spaces and check for bad characters */
6657 d = SvPVX(PL_lex_stuff);
6659 for (p = d; *p; ++p) {
6662 if (warnsyntax && !strchr("$@%*;[]&\\_", *p))
6668 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
6669 "Illegal character in prototype for %"SVf" : %s",
6670 SVfARG(PL_subname), d);
6671 SvCUR_set(PL_lex_stuff, tmp);
6676 CURMAD('q', PL_thisopen);
6677 CURMAD('_', tmpwhite);
6678 CURMAD('=', PL_thisstuff);
6679 CURMAD('Q', PL_thisclose);
6680 NEXTVAL_NEXTTOKE.opval =
6681 (OP*)newSVOP(OP_CONST, 0, PL_lex_stuff);
6682 PL_lex_stuff = Nullsv;
6685 s = SKIPSPACE2(s,tmpwhite);
6693 if (*s == ':' && s[1] != ':')
6694 PL_expect = attrful;
6695 else if (*s != '{' && key == KEY_sub) {
6697 Perl_croak(aTHX_ "Illegal declaration of anonymous subroutine");
6699 Perl_croak(aTHX_ "Illegal declaration of subroutine %"SVf, SVfARG(PL_subname));
6706 curmad('^', newSVpvs(""));
6707 CURMAD('_', tmpwhite);
6711 PL_thistoken = subtoken;
6714 NEXTVAL_NEXTTOKE.opval =
6715 (OP*)newSVOP(OP_CONST, 0, PL_lex_stuff);
6716 PL_lex_stuff = NULL;
6721 sv_setpv(PL_subname,
6723 (PL_curstash ? "__ANON__" : "__ANON__::__ANON__"));
6727 (void) force_word(PL_oldbufptr + tboffset, WORD,
6737 LOP(OP_SYSTEM,XREF);
6740 LOP(OP_SYMLINK,XTERM);
6743 LOP(OP_SYSCALL,XTERM);
6746 LOP(OP_SYSOPEN,XTERM);
6749 LOP(OP_SYSSEEK,XTERM);
6752 LOP(OP_SYSREAD,XTERM);
6755 LOP(OP_SYSWRITE,XTERM);
6759 TERM(sublex_start());
6780 LOP(OP_TRUNCATE,XTERM);
6792 yylval.ival = CopLINE(PL_curcop);
6796 yylval.ival = CopLINE(PL_curcop);
6800 LOP(OP_UNLINK,XTERM);
6806 LOP(OP_UNPACK,XTERM);
6809 LOP(OP_UTIME,XTERM);
6815 LOP(OP_UNSHIFT,XTERM);
6818 s = tokenize_use(1, s);
6828 yylval.ival = CopLINE(PL_curcop);
6832 yylval.ival = CopLINE(PL_curcop);
6836 PL_hints |= HINT_BLOCK_SCOPE;
6843 LOP(OP_WAITPID,XTERM);
6852 ctl_l[0] = toCTRL('L');
6854 gv_fetchpvn_flags(ctl_l, 1, GV_ADD|GV_NOTQUAL, SVt_PV);
6857 /* Make sure $^L is defined */
6858 gv_fetchpvs("\f", GV_ADD|GV_NOTQUAL, SVt_PV);
6863 if (PL_expect == XOPERATOR)
6869 yylval.ival = OP_XOR;
6874 TERM(sublex_start());
6879 #pragma segment Main
6883 S_pending_ident(pTHX)
6888 /* pit holds the identifier we read and pending_ident is reset */
6889 char pit = PL_pending_ident;
6890 PL_pending_ident = 0;
6892 /* PL_realtokenstart = realtokenend = PL_bufptr - SvPVX(PL_linestr); */
6893 DEBUG_T({ PerlIO_printf(Perl_debug_log,
6894 "### Pending identifier '%s'\n", PL_tokenbuf); });
6896 /* if we're in a my(), we can't allow dynamics here.
6897 $foo'bar has already been turned into $foo::bar, so
6898 just check for colons.
6900 if it's a legal name, the OP is a PADANY.
6903 if (PL_in_my == KEY_our) { /* "our" is merely analogous to "my" */
6904 if (strchr(PL_tokenbuf,':'))
6905 yyerror(Perl_form(aTHX_ "No package name allowed for "
6906 "variable %s in \"our\"",
6908 tmp = allocmy(PL_tokenbuf);
6911 if (strchr(PL_tokenbuf,':'))
6912 yyerror(Perl_form(aTHX_ PL_no_myglob,
6913 PL_in_my == KEY_my ? "my" : "state", PL_tokenbuf));
6915 yylval.opval = newOP(OP_PADANY, 0);
6916 yylval.opval->op_targ = allocmy(PL_tokenbuf);
6922 build the ops for accesses to a my() variable.
6924 Deny my($a) or my($b) in a sort block, *if* $a or $b is
6925 then used in a comparison. This catches most, but not
6926 all cases. For instance, it catches
6927 sort { my($a); $a <=> $b }
6929 sort { my($a); $a < $b ? -1 : $a == $b ? 0 : 1; }
6930 (although why you'd do that is anyone's guess).
6933 if (!strchr(PL_tokenbuf,':')) {
6935 tmp = pad_findmy(PL_tokenbuf);
6936 if (tmp != NOT_IN_PAD) {
6937 /* might be an "our" variable" */
6938 if (PAD_COMPNAME_FLAGS_isOUR(tmp)) {
6939 /* build ops for a bareword */
6940 HV * const stash = PAD_COMPNAME_OURSTASH(tmp);
6941 HEK * const stashname = HvNAME_HEK(stash);
6942 SV * const sym = newSVhek(stashname);
6943 sv_catpvs(sym, "::");
6944 sv_catpv(sym, PL_tokenbuf+1);
6945 yylval.opval = (OP*)newSVOP(OP_CONST, 0, sym);
6946 yylval.opval->op_private = OPpCONST_ENTERED;
6949 ? (GV_ADDMULTI | GV_ADDINEVAL)
6952 ((PL_tokenbuf[0] == '$') ? SVt_PV
6953 : (PL_tokenbuf[0] == '@') ? SVt_PVAV
6958 /* if it's a sort block and they're naming $a or $b */
6959 if (PL_last_lop_op == OP_SORT &&
6960 PL_tokenbuf[0] == '$' &&
6961 (PL_tokenbuf[1] == 'a' || PL_tokenbuf[1] == 'b')
6964 for (d = PL_in_eval ? PL_oldoldbufptr : PL_linestart;
6965 d < PL_bufend && *d != '\n';
6968 if (strnEQ(d,"<=>",3) || strnEQ(d,"cmp",3)) {
6969 Perl_croak(aTHX_ "Can't use \"my %s\" in sort comparison",
6975 yylval.opval = newOP(OP_PADANY, 0);
6976 yylval.opval->op_targ = tmp;
6982 Whine if they've said @foo in a doublequoted string,
6983 and @foo isn't a variable we can find in the symbol
6986 if (pit == '@' && PL_lex_state != LEX_NORMAL && !PL_lex_brackets) {
6987 GV *gv = gv_fetchpv(PL_tokenbuf+1, 0, SVt_PVAV);
6988 if ((!gv || ((PL_tokenbuf[0] == '@') ? !GvAV(gv) : !GvHV(gv)))
6989 && ckWARN(WARN_AMBIGUOUS))
6991 /* Downgraded from fatal to warning 20000522 mjd */
6992 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
6993 "Possible unintended interpolation of %s in string",
6998 /* build ops for a bareword */
6999 yylval.opval = (OP*)newSVOP(OP_CONST, 0, newSVpv(PL_tokenbuf+1, 0));
7000 yylval.opval->op_private = OPpCONST_ENTERED;
7003 /* If the identifier refers to a stash, don't autovivify it.
7004 * Change 24660 had the side effect of causing symbol table
7005 * hashes to always be defined, even if they were freshly
7006 * created and the only reference in the entire program was
7007 * the single statement with the defined %foo::bar:: test.
7008 * It appears that all code in the wild doing this actually
7009 * wants to know whether sub-packages have been loaded, so
7010 * by avoiding auto-vivifying symbol tables, we ensure that
7011 * defined %foo::bar:: continues to be false, and the existing
7012 * tests still give the expected answers, even though what
7013 * they're actually testing has now changed subtly.
7015 (*PL_tokenbuf == '%' && *(d = PL_tokenbuf + strlen(PL_tokenbuf) - 1) == ':' && d[-1] == ':'
7017 : PL_in_eval ? (GV_ADDMULTI | GV_ADDINEVAL) : GV_ADD),
7018 ((PL_tokenbuf[0] == '$') ? SVt_PV
7019 : (PL_tokenbuf[0] == '@') ? SVt_PVAV
7025 * The following code was generated by perl_keyword.pl.
7029 Perl_keyword (pTHX_ const char *name, I32 len, bool all_keywords)
7034 case 1: /* 5 tokens of length 1 */
7066 case 2: /* 18 tokens of length 2 */
7212 case 3: /* 29 tokens of length 3 */
7216 if (name[1] == 'N' &&
7279 if (name[1] == 'i' &&
7301 return (all_keywords || FEATURE_IS_ENABLED("err") ? -KEY_err : 0);
7319 if (name[1] == 'o' &&
7328 if (name[1] == 'e' &&
7337 if (name[1] == 'n' &&
7346 if (name[1] == 'o' &&
7355 if (name[1] == 'a' &&
7364 if (name[1] == 'o' &&
7426 if (name[1] == 'e' &&
7440 return (all_keywords || FEATURE_IS_ENABLED("say") ? KEY_say : 0);
7466 if (name[1] == 'i' &&
7475 if (name[1] == 's' &&
7484 if (name[1] == 'e' &&
7493 if (name[1] == 'o' &&
7505 case 4: /* 41 tokens of length 4 */
7509 if (name[1] == 'O' &&
7519 if (name[1] == 'N' &&
7529 if (name[1] == 'i' &&
7539 if (name[1] == 'h' &&
7549 if (name[1] == 'u' &&
7562 if (name[2] == 'c' &&
7571 if (name[2] == 's' &&
7580 if (name[2] == 'a' &&
7616 if (name[1] == 'o' &&
7629 if (name[2] == 't' &&
7638 if (name[2] == 'o' &&
7647 if (name[2] == 't' &&
7656 if (name[2] == 'e' &&
7669 if (name[1] == 'o' &&
7682 if (name[2] == 'y' &&
7691 if (name[2] == 'l' &&
7707 if (name[2] == 's' &&
7716 if (name[2] == 'n' &&
7725 if (name[2] == 'c' &&
7738 if (name[1] == 'e' &&
7748 if (name[1] == 'p' &&
7761 if (name[2] == 'c' &&
7770 if (name[2] == 'p' &&
7779 if (name[2] == 's' &&
7795 if (name[2] == 'n' &&
7865 if (name[2] == 'r' &&
7874 if (name[2] == 'r' &&
7883 if (name[2] == 'a' &&
7899 if (name[2] == 'l' &&
7961 if (name[2] == 'e' &&
7964 return (all_keywords || FEATURE_IS_ENABLED("switch") ? KEY_when : 0);
7977 case 5: /* 39 tokens of length 5 */
7981 if (name[1] == 'E' &&
7992 if (name[1] == 'H' &&
8006 if (name[2] == 'a' &&
8016 if (name[2] == 'a' &&
8033 if (name[2] == 'e' &&
8043 if (name[2] == 'e' &&
8047 return (all_keywords || FEATURE_IS_ENABLED("switch") ? -KEY_break : 0);
8063 if (name[3] == 'i' &&
8072 if (name[3] == 'o' &&
8108 if (name[2] == 'o' &&
8118 if (name[2] == 'y' &&
8132 if (name[1] == 'l' &&
8146 if (name[2] == 'n' &&
8156 if (name[2] == 'o' &&
8170 if (name[1] == 'i' &&
8175 return (all_keywords || FEATURE_IS_ENABLED("switch") ? KEY_given : 0);
8184 if (name[2] == 'd' &&
8194 if (name[2] == 'c' &&
8211 if (name[2] == 'c' &&
8221 if (name[2] == 't' &&
8235 if (name[1] == 'k' &&
8246 if (name[1] == 'r' &&
8260 if (name[2] == 's' &&
8270 if (name[2] == 'd' &&
8287 if (name[2] == 'm' &&
8297 if (name[2] == 'i' &&
8307 if (name[2] == 'e' &&
8317 if (name[2] == 'l' &&
8327 if (name[2] == 'a' &&
8340 if (name[3] == 't' &&
8343 return (all_keywords || FEATURE_IS_ENABLED("state") ? KEY_state : 0);
8349 if (name[3] == 'd' &&
8366 if (name[1] == 'i' &&
8380 if (name[2] == 'a' &&
8393 if (name[3] == 'e' &&
8428 if (name[2] == 'i' &&
8445 if (name[2] == 'i' &&
8455 if (name[2] == 'i' &&
8472 case 6: /* 33 tokens of length 6 */
8476 if (name[1] == 'c' &&
8491 if (name[2] == 'l' &&
8502 if (name[2] == 'r' &&
8517 if (name[1] == 'e' &&
8532 if (name[2] == 's' &&
8537 if(ckWARN_d(WARN_SYNTAX))
8538 Perl_warner(aTHX_ packWARN(WARN_SYNTAX), "elseif should be elsif");
8544 if (name[2] == 'i' &&
8562 if (name[2] == 'l' &&
8573 if (name[2] == 'r' &&
8588 if (name[1] == 'm' &&
8603 if (name[2] == 'n' &&
8614 if (name[2] == 's' &&
8629 if (name[1] == 's' &&
8635 if (name[4] == 't' &&
8644 if (name[4] == 'e' &&
8653 if (name[4] == 'c' &&
8662 if (name[4] == 'n' &&
8678 if (name[1] == 'r' &&
8696 if (name[3] == 'a' &&
8706 if (name[3] == 'u' &&
8720 if (name[2] == 'n' &&
8738 if (name[2] == 'a' &&
8752 if (name[3] == 'e' &&
8765 if (name[4] == 't' &&
8774 if (name[4] == 'e' &&
8796 if (name[4] == 't' &&
8805 if (name[4] == 'e' &&
8821 if (name[2] == 'c' &&
8832 if (name[2] == 'l' &&
8843 if (name[2] == 'b' &&
8854 if (name[2] == 's' &&
8877 if (name[4] == 's' &&
8886 if (name[4] == 'n' &&
8899 if (name[3] == 'a' &&
8916 if (name[1] == 'a' &&
8931 case 7: /* 29 tokens of length 7 */
8935 if (name[1] == 'E' &&
8948 if (name[1] == '_' &&
8961 if (name[1] == 'i' &&
8968 return -KEY_binmode;
8974 if (name[1] == 'o' &&
8981 return -KEY_connect;
8990 if (name[2] == 'm' &&
8996 return -KEY_dbmopen;
9007 if (name[4] == 'u' &&
9011 return (all_keywords || FEATURE_IS_ENABLED("switch") ? KEY_default : 0);
9017 if (name[4] == 'n' &&
9038 if (name[1] == 'o' &&
9051 if (name[1] == 'e' &&
9058 if (name[5] == 'r' &&
9061 return -KEY_getpgrp;
9067 if (name[5] == 'i' &&
9070 return -KEY_getppid;
9083 if (name[1] == 'c' &&
9090 return -KEY_lcfirst;
9096 if (name[1] == 'p' &&
9103 return -KEY_opendir;
9109 if (name[1] == 'a' &&
9127 if (name[3] == 'd' &&
9132 return -KEY_readdir;
9138 if (name[3] == 'u' &&
9149 if (name[3] == 'e' &&
9154 return -KEY_reverse;
9173 if (name[3] == 'k' &&
9178 return -KEY_seekdir;
9184 if (name[3] == 'p' &&
9189 return -KEY_setpgrp;
9199 if (name[2] == 'm' &&
9205 return -KEY_shmread;
9211 if (name[2] == 'r' &&
9217 return -KEY_sprintf;
9226 if (name[3] == 'l' &&
9231 return -KEY_symlink;
9240 if (name[4] == 'a' &&
9244 return -KEY_syscall;
9250 if (name[4] == 'p' &&
9254 return -KEY_sysopen;
9260 if (name[4] == 'e' &&
9264 return -KEY_sysread;
9270 if (name[4] == 'e' &&
9274 return -KEY_sysseek;
9292 if (name[1] == 'e' &&
9299 return -KEY_telldir;
9308 if (name[2] == 'f' &&
9314 return -KEY_ucfirst;
9320 if (name[2] == 's' &&
9326 return -KEY_unshift;
9336 if (name[1] == 'a' &&
9343 return -KEY_waitpid;
9352 case 8: /* 26 tokens of length 8 */
9356 if (name[1] == 'U' &&
9364 return KEY_AUTOLOAD;
9375 if (name[3] == 'A' &&
9381 return KEY___DATA__;
9387 if (name[3] == 'I' &&
9393 return -KEY___FILE__;
9399 if (name[3] == 'I' &&
9405 return -KEY___LINE__;
9421 if (name[2] == 'o' &&
9428 return -KEY_closedir;
9434 if (name[2] == 'n' &&
9441 return -KEY_continue;
9451 if (name[1] == 'b' &&
9459 return -KEY_dbmclose;
9465 if (name[1] == 'n' &&
9471 if (name[4] == 'r' &&
9476 return -KEY_endgrent;
9482 if (name[4] == 'w' &&
9487 return -KEY_endpwent;
9500 if (name[1] == 'o' &&
9508 return -KEY_formline;
9514 if (name[1] == 'e' &&
9525 if (name[6] == 'n' &&
9528 return -KEY_getgrent;
9534 if (name[6] == 'i' &&
9537 return -KEY_getgrgid;
9543 if (name[6] == 'a' &&
9546 return -KEY_getgrnam;
9559 if (name[4] == 'o' &&
9564 return -KEY_getlogin;
9575 if (name[6] == 'n' &&
9578 return -KEY_getpwent;
9584 if (name[6] == 'a' &&
9587 return -KEY_getpwnam;
9593 if (name[6] == 'i' &&
9596 return -KEY_getpwuid;
9616 if (name[1] == 'e' &&
9623 if (name[5] == 'i' &&
9630 return -KEY_readline;
9635 return -KEY_readlink;
9646 if (name[5] == 'i' &&
9650 return -KEY_readpipe;
9671 if (name[4] == 'r' &&
9676 return -KEY_setgrent;
9682 if (name[4] == 'w' &&
9687 return -KEY_setpwent;
9703 if (name[3] == 'w' &&
9709 return -KEY_shmwrite;
9715 if (name[3] == 't' &&
9721 return -KEY_shutdown;
9731 if (name[2] == 's' &&
9738 return -KEY_syswrite;
9748 if (name[1] == 'r' &&
9756 return -KEY_truncate;
9765 case 9: /* 9 tokens of length 9 */
9769 if (name[1] == 'N' &&
9778 return KEY_UNITCHECK;
9784 if (name[1] == 'n' &&
9793 return -KEY_endnetent;
9799 if (name[1] == 'e' &&
9808 return -KEY_getnetent;
9814 if (name[1] == 'o' &&
9823 return -KEY_localtime;
9829 if (name[1] == 'r' &&
9838 return KEY_prototype;
9844 if (name[1] == 'u' &&
9853 return -KEY_quotemeta;
9859 if (name[1] == 'e' &&
9868 return -KEY_rewinddir;
9874 if (name[1] == 'e' &&
9883 return -KEY_setnetent;
9889 if (name[1] == 'a' &&
9898 return -KEY_wantarray;
9907 case 10: /* 9 tokens of length 10 */
9911 if (name[1] == 'n' &&
9917 if (name[4] == 'o' &&
9924 return -KEY_endhostent;
9930 if (name[4] == 'e' &&
9937 return -KEY_endservent;
9950 if (name[1] == 'e' &&
9956 if (name[4] == 'o' &&
9963 return -KEY_gethostent;
9972 if (name[5] == 'r' &&
9978 return -KEY_getservent;
9984 if (name[5] == 'c' &&
9990 return -KEY_getsockopt;
10010 if (name[2] == 't')
10015 if (name[4] == 'o' &&
10022 return -KEY_sethostent;
10031 if (name[5] == 'r' &&
10037 return -KEY_setservent;
10043 if (name[5] == 'c' &&
10049 return -KEY_setsockopt;
10066 if (name[2] == 'c' &&
10075 return -KEY_socketpair;
10088 case 11: /* 8 tokens of length 11 */
10092 if (name[1] == '_' &&
10102 { /* __PACKAGE__ */
10103 return -KEY___PACKAGE__;
10109 if (name[1] == 'n' &&
10119 { /* endprotoent */
10120 return -KEY_endprotoent;
10126 if (name[1] == 'e' &&
10135 if (name[5] == 'e' &&
10141 { /* getpeername */
10142 return -KEY_getpeername;
10151 if (name[6] == 'o' &&
10156 { /* getpriority */
10157 return -KEY_getpriority;
10163 if (name[6] == 't' &&
10168 { /* getprotoent */
10169 return -KEY_getprotoent;
10183 if (name[4] == 'o' &&
10190 { /* getsockname */
10191 return -KEY_getsockname;
10204 if (name[1] == 'e' &&
10212 if (name[6] == 'o' &&
10217 { /* setpriority */
10218 return -KEY_setpriority;
10224 if (name[6] == 't' &&
10229 { /* setprotoent */
10230 return -KEY_setprotoent;
10246 case 12: /* 2 tokens of length 12 */
10247 if (name[0] == 'g' &&
10259 if (name[9] == 'd' &&
10262 { /* getnetbyaddr */
10263 return -KEY_getnetbyaddr;
10269 if (name[9] == 'a' &&
10272 { /* getnetbyname */
10273 return -KEY_getnetbyname;
10285 case 13: /* 4 tokens of length 13 */
10286 if (name[0] == 'g' &&
10293 if (name[4] == 'o' &&
10302 if (name[10] == 'd' &&
10305 { /* gethostbyaddr */
10306 return -KEY_gethostbyaddr;
10312 if (name[10] == 'a' &&
10315 { /* gethostbyname */
10316 return -KEY_gethostbyname;
10329 if (name[4] == 'e' &&
10338 if (name[10] == 'a' &&
10341 { /* getservbyname */
10342 return -KEY_getservbyname;
10348 if (name[10] == 'o' &&
10351 { /* getservbyport */
10352 return -KEY_getservbyport;
10371 case 14: /* 1 tokens of length 14 */
10372 if (name[0] == 'g' &&
10386 { /* getprotobyname */
10387 return -KEY_getprotobyname;
10392 case 16: /* 1 tokens of length 16 */
10393 if (name[0] == 'g' &&
10409 { /* getprotobynumber */
10410 return -KEY_getprotobynumber;
10424 S_checkcomma(pTHX_ const char *s, const char *name, const char *what)
10428 if (*s == ' ' && s[1] == '(') { /* XXX gotta be a better way */
10429 if (ckWARN(WARN_SYNTAX)) {
10432 for (w = s+2; *w && level; w++) {
10435 else if (*w == ')')
10438 while (isSPACE(*w))
10440 if (!*w || !strchr(";|})]oaiuw!=", *w)) /* an advisory hack only... */
10441 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
10442 "%s (...) interpreted as function",name);
10445 while (s < PL_bufend && isSPACE(*s))
10449 while (s < PL_bufend && isSPACE(*s))
10451 if (isIDFIRST_lazy_if(s,UTF)) {
10452 const char * const w = s++;
10453 while (isALNUM_lazy_if(s,UTF))
10455 while (s < PL_bufend && isSPACE(*s))
10459 if (keyword(w, s - w, 0))
10462 gv = gv_fetchpvn_flags(w, s - w, 0, SVt_PVCV);
10463 if (gv && GvCVu(gv))
10465 Perl_croak(aTHX_ "No comma allowed after %s", what);
10470 /* Either returns sv, or mortalizes sv and returns a new SV*.
10471 Best used as sv=new_constant(..., sv, ...).
10472 If s, pv are NULL, calls subroutine with one argument,
10473 and type is used with error messages only. */
10476 S_new_constant(pTHX_ const char *s, STRLEN len, const char *key, SV *sv, SV *pv,
10480 HV * const table = GvHV(PL_hintgv); /* ^H */
10484 const char *why1 = "", *why2 = "", *why3 = "";
10486 if (!table || !(PL_hints & HINT_LOCALIZE_HH)) {
10489 why2 = (const char *)
10490 (strEQ(key,"charnames")
10491 ? "(possibly a missing \"use charnames ...\")"
10493 msg = Perl_newSVpvf(aTHX_ "Constant(%s) unknown: %s",
10494 (type ? type: "undef"), why2);
10496 /* This is convoluted and evil ("goto considered harmful")
10497 * but I do not understand the intricacies of all the different
10498 * failure modes of %^H in here. The goal here is to make
10499 * the most probable error message user-friendly. --jhi */
10504 msg = Perl_newSVpvf(aTHX_ "Constant(%s): %s%s%s",
10505 (type ? type: "undef"), why1, why2, why3);
10507 yyerror(SvPVX_const(msg));
10511 cvp = hv_fetch(table, key, strlen(key), FALSE);
10512 if (!cvp || !SvOK(*cvp)) {
10515 why3 = "} is not defined";
10518 sv_2mortal(sv); /* Parent created it permanently */
10521 pv = sv_2mortal(newSVpvn(s, len));
10523 typesv = sv_2mortal(newSVpv(type, 0));
10525 typesv = &PL_sv_undef;
10527 PUSHSTACKi(PERLSI_OVERLOAD);
10539 call_sv(cv, G_SCALAR | ( PL_in_eval ? 0 : G_EVAL));
10543 /* Check the eval first */
10544 if (!PL_in_eval && SvTRUE(ERRSV)) {
10545 sv_catpvs(ERRSV, "Propagated");
10546 yyerror(SvPV_nolen_const(ERRSV)); /* Duplicates the message inside eval */
10548 res = SvREFCNT_inc_simple(sv);
10552 SvREFCNT_inc_simple_void(res);
10561 why1 = "Call to &{$^H{";
10563 why3 = "}} did not return a defined value";
10571 /* Returns a NUL terminated string, with the length of the string written to
10575 S_scan_word(pTHX_ register char *s, char *dest, STRLEN destlen, int allow_package, STRLEN *slp)
10578 register char *d = dest;
10579 register char * const e = d + destlen - 3; /* two-character token, ending NUL */
10582 Perl_croak(aTHX_ ident_too_long);
10583 if (isALNUM(*s)) /* UTF handled below */
10585 else if (allow_package && (*s == '\'') && isIDFIRST_lazy_if(s+1,UTF)) {
10590 else if (allow_package && (s[0] == ':') && (s[1] == ':') && (s[2] != '$')) {
10594 else if (UTF && UTF8_IS_START(*s) && isALNUM_utf8((U8*)s)) {
10595 char *t = s + UTF8SKIP(s);
10597 while (UTF8_IS_CONTINUED(*t) && is_utf8_mark((U8*)t))
10601 Perl_croak(aTHX_ ident_too_long);
10602 Copy(s, d, len, char);
10615 S_scan_ident(pTHX_ register char *s, register const char *send, char *dest, STRLEN destlen, I32 ck_uni)
10618 char *bracket = NULL;
10620 register char *d = dest;
10621 register char * const e = d + destlen + 3; /* two-character token, ending NUL */
10626 while (isDIGIT(*s)) {
10628 Perl_croak(aTHX_ ident_too_long);
10635 Perl_croak(aTHX_ ident_too_long);
10636 if (isALNUM(*s)) /* UTF handled below */
10638 else if (*s == '\'' && isIDFIRST_lazy_if(s+1,UTF)) {
10643 else if (*s == ':' && s[1] == ':') {
10647 else if (UTF && UTF8_IS_START(*s) && isALNUM_utf8((U8*)s)) {
10648 char *t = s + UTF8SKIP(s);
10649 while (UTF8_IS_CONTINUED(*t) && is_utf8_mark((U8*)t))
10651 if (d + (t - s) > e)
10652 Perl_croak(aTHX_ ident_too_long);
10653 Copy(s, d, t - s, char);
10664 if (PL_lex_state != LEX_NORMAL)
10665 PL_lex_state = LEX_INTERPENDMAYBE;
10668 if (*s == '$' && s[1] &&
10669 (isALNUM_lazy_if(s+1,UTF) || s[1] == '$' || s[1] == '{' || strnEQ(s+1,"::",2)) )
10682 if (*d == '^' && *s && isCONTROLVAR(*s)) {
10687 if (isSPACE(s[-1])) {
10689 const char ch = *s++;
10690 if (!SPACE_OR_TAB(ch)) {
10696 if (isIDFIRST_lazy_if(d,UTF)) {
10700 while ((end < send && isALNUM_lazy_if(end,UTF)) || *end == ':') {
10701 end += UTF8SKIP(end);
10702 while (end < send && UTF8_IS_CONTINUED(*end) && is_utf8_mark((U8*)end))
10703 end += UTF8SKIP(end);
10705 Copy(s, d, end - s, char);
10710 while ((isALNUM(*s) || *s == ':') && d < e)
10713 Perl_croak(aTHX_ ident_too_long);
10716 while (s < send && SPACE_OR_TAB(*s))
10718 if ((*s == '[' || (*s == '{' && strNE(dest, "sub")))) {
10719 if (ckWARN(WARN_AMBIGUOUS) && keyword(dest, d - dest, 0)) {
10720 const char * const brack =
10722 ((*s == '[') ? "[...]" : "{...}");
10723 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
10724 "Ambiguous use of %c{%s%s} resolved to %c%s%s",
10725 funny, dest, brack, funny, dest, brack);
10728 PL_lex_brackstack[PL_lex_brackets++] = (char)(XOPERATOR | XFAKEBRACK);
10732 /* Handle extended ${^Foo} variables
10733 * 1999-02-27 mjd-perl-patch@plover.com */
10734 else if (!isALNUM(*d) && !isPRINT(*d) /* isCTRL(d) */
10738 while (isALNUM(*s) && d < e) {
10742 Perl_croak(aTHX_ ident_too_long);
10747 if (PL_lex_state == LEX_INTERPNORMAL && !PL_lex_brackets) {
10748 PL_lex_state = LEX_INTERPEND;
10751 if (PL_lex_state == LEX_NORMAL) {
10752 if (ckWARN(WARN_AMBIGUOUS) &&
10753 (keyword(dest, d - dest, 0)
10754 || get_cvn_flags(dest, d - dest, 0)))
10758 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
10759 "Ambiguous use of %c{%s} resolved to %c%s",
10760 funny, dest, funny, dest);
10765 s = bracket; /* let the parser handle it */
10769 else if (PL_lex_state == LEX_INTERPNORMAL && !PL_lex_brackets && !intuit_more(s))
10770 PL_lex_state = LEX_INTERPEND;
10775 Perl_pmflag(pTHX_ U32* pmfl, int ch)
10777 PERL_UNUSED_CONTEXT;
10781 CASE_STD_PMMOD_FLAGS_PARSE_SET(pmfl);
10782 case GLOBAL_PAT_MOD: *pmfl |= PMf_GLOBAL; break;
10783 case CONTINUE_PAT_MOD: *pmfl |= PMf_CONTINUE; break;
10784 case ONCE_PAT_MOD: *pmfl |= PMf_KEEP; break;
10785 case KEEPCOPY_PAT_MOD: *pmfl |= PMf_KEEPCOPY; break;
10791 S_scan_pat(pTHX_ char *start, I32 type)
10795 char *s = scan_str(start,!!PL_madskills,FALSE);
10796 const char * const valid_flags =
10797 (const char *)((type == OP_QR) ? QR_PAT_MODS : M_PAT_MODS);
10804 const char * const delimiter = skipspace(start);
10808 ? "Search pattern not terminated or ternary operator parsed as search pattern"
10809 : "Search pattern not terminated" ));
10812 pm = (PMOP*)newPMOP(type, 0);
10813 if (PL_multi_open == '?')
10814 pm->op_pmflags |= PMf_ONCE;
10818 while (*s && strchr(valid_flags, *s))
10819 pmflag(&pm->op_pmflags,*s++);
10821 if (PL_madskills && modstart != s) {
10822 SV* tmptoken = newSVpvn(modstart, s - modstart);
10823 append_madprops(newMADPROP('m', MAD_SV, tmptoken, 0), (OP*)pm, 0);
10826 /* issue a warning if /c is specified,but /g is not */
10827 if ((pm->op_pmflags & PMf_CONTINUE) && !(pm->op_pmflags & PMf_GLOBAL)
10828 && ckWARN(WARN_REGEXP))
10830 Perl_warner(aTHX_ packWARN(WARN_REGEXP),
10831 "Use of /c modifier is meaningless without /g" );
10834 pm->op_pmpermflags = pm->op_pmflags;
10836 PL_lex_op = (OP*)pm;
10837 yylval.ival = OP_MATCH;
10842 S_scan_subst(pTHX_ char *start)
10853 yylval.ival = OP_NULL;
10855 s = scan_str(start,!!PL_madskills,FALSE);
10858 Perl_croak(aTHX_ "Substitution pattern not terminated");
10860 if (s[-1] == PL_multi_open)
10863 if (PL_madskills) {
10864 CURMAD('q', PL_thisopen);
10865 CURMAD('_', PL_thiswhite);
10866 CURMAD('E', PL_thisstuff);
10867 CURMAD('Q', PL_thisclose);
10868 PL_realtokenstart = s - SvPVX(PL_linestr);
10872 first_start = PL_multi_start;
10873 s = scan_str(s,!!PL_madskills,FALSE);
10875 if (PL_lex_stuff) {
10876 SvREFCNT_dec(PL_lex_stuff);
10877 PL_lex_stuff = NULL;
10879 Perl_croak(aTHX_ "Substitution replacement not terminated");
10881 PL_multi_start = first_start; /* so whole substitution is taken together */
10883 pm = (PMOP*)newPMOP(OP_SUBST, 0);
10886 if (PL_madskills) {
10887 CURMAD('z', PL_thisopen);
10888 CURMAD('R', PL_thisstuff);
10889 CURMAD('Z', PL_thisclose);
10895 if (*s == EXEC_PAT_MOD) {
10899 else if (strchr(S_PAT_MODS, *s))
10900 pmflag(&pm->op_pmflags,*s++);
10906 if (PL_madskills) {
10908 curmad('m', newSVpvn(modstart, s - modstart));
10909 append_madprops(PL_thismad, (OP*)pm, 0);
10913 if ((pm->op_pmflags & PMf_CONTINUE) && ckWARN(WARN_REGEXP)) {
10914 Perl_warner(aTHX_ packWARN(WARN_REGEXP), "Use of /c modifier is meaningless in s///" );
10918 SV * const repl = newSVpvs("");
10920 PL_sublex_info.super_bufptr = s;
10921 PL_sublex_info.super_bufend = PL_bufend;
10923 pm->op_pmflags |= PMf_EVAL;
10925 sv_catpv(repl, (const char *)(es ? "eval " : "do "));
10926 sv_catpvs(repl, "{");
10927 sv_catsv(repl, PL_lex_repl);
10928 if (strchr(SvPVX(PL_lex_repl), '#'))
10929 sv_catpvs(repl, "\n");
10930 sv_catpvs(repl, "}");
10932 SvREFCNT_dec(PL_lex_repl);
10933 PL_lex_repl = repl;
10936 pm->op_pmpermflags = pm->op_pmflags;
10937 PL_lex_op = (OP*)pm;
10938 yylval.ival = OP_SUBST;
10943 S_scan_trans(pTHX_ char *start)
10956 yylval.ival = OP_NULL;
10958 s = scan_str(start,!!PL_madskills,FALSE);
10960 Perl_croak(aTHX_ "Transliteration pattern not terminated");
10962 if (s[-1] == PL_multi_open)
10965 if (PL_madskills) {
10966 CURMAD('q', PL_thisopen);
10967 CURMAD('_', PL_thiswhite);
10968 CURMAD('E', PL_thisstuff);
10969 CURMAD('Q', PL_thisclose);
10970 PL_realtokenstart = s - SvPVX(PL_linestr);
10974 s = scan_str(s,!!PL_madskills,FALSE);
10976 if (PL_lex_stuff) {
10977 SvREFCNT_dec(PL_lex_stuff);
10978 PL_lex_stuff = NULL;
10980 Perl_croak(aTHX_ "Transliteration replacement not terminated");
10982 if (PL_madskills) {
10983 CURMAD('z', PL_thisopen);
10984 CURMAD('R', PL_thisstuff);
10985 CURMAD('Z', PL_thisclose);
10988 complement = del = squash = 0;
10995 complement = OPpTRANS_COMPLEMENT;
10998 del = OPpTRANS_DELETE;
11001 squash = OPpTRANS_SQUASH;
11010 tbl = (short *)PerlMemShared_calloc(complement&&!del?258:256, sizeof(short));
11011 o = newPVOP(OP_TRANS, 0, (char*)tbl);
11012 o->op_private &= ~OPpTRANS_ALL;
11013 o->op_private |= del|squash|complement|
11014 (DO_UTF8(PL_lex_stuff)? OPpTRANS_FROM_UTF : 0)|
11015 (DO_UTF8(PL_lex_repl) ? OPpTRANS_TO_UTF : 0);
11018 yylval.ival = OP_TRANS;
11021 if (PL_madskills) {
11023 curmad('m', newSVpvn(modstart, s - modstart));
11024 append_madprops(PL_thismad, o, 0);
11033 S_scan_heredoc(pTHX_ register char *s)
11037 I32 op_type = OP_SCALAR;
11041 const char *found_newline;
11045 const int outer = (PL_rsfp && !(PL_lex_inwhat == OP_SCALAR));
11047 I32 stuffstart = s - SvPVX(PL_linestr);
11050 PL_realtokenstart = -1;
11055 e = PL_tokenbuf + sizeof PL_tokenbuf - 1;
11059 while (SPACE_OR_TAB(*peek))
11061 if (*peek == '`' || *peek == '\'' || *peek =='"') {
11064 s = delimcpy(d, e, s, PL_bufend, term, &len);
11074 if (!isALNUM_lazy_if(s,UTF))
11075 deprecate_old("bare << to mean <<\"\"");
11076 for (; isALNUM_lazy_if(s,UTF); s++) {
11081 if (d >= PL_tokenbuf + sizeof PL_tokenbuf - 1)
11082 Perl_croak(aTHX_ "Delimiter for here document is too long");
11085 len = d - PL_tokenbuf;
11088 if (PL_madskills) {
11089 tstart = PL_tokenbuf + !outer;
11090 PL_thisclose = newSVpvn(tstart, len - !outer);
11091 tstart = SvPVX(PL_linestr) + stuffstart;
11092 PL_thisopen = newSVpvn(tstart, s - tstart);
11093 stuffstart = s - SvPVX(PL_linestr);
11096 #ifndef PERL_STRICT_CR
11097 d = strchr(s, '\r');
11099 char * const olds = s;
11101 while (s < PL_bufend) {
11107 else if (*s == '\n' && s[1] == '\r') { /* \015\013 on a mac? */
11116 SvCUR_set(PL_linestr, PL_bufend - SvPVX_const(PL_linestr));
11123 if ( outer || !(found_newline = (char*)memchr((void*)s, '\n', PL_bufend - s)) ) {
11124 herewas = newSVpvn(s,PL_bufend-s);
11128 herewas = newSVpvn(s-1,found_newline-s+1);
11131 herewas = newSVpvn(s,found_newline-s);
11135 if (PL_madskills) {
11136 tstart = SvPVX(PL_linestr) + stuffstart;
11138 sv_catpvn(PL_thisstuff, tstart, s - tstart);
11140 PL_thisstuff = newSVpvn(tstart, s - tstart);
11143 s += SvCUR(herewas);
11146 stuffstart = s - SvPVX(PL_linestr);
11152 tmpstr = newSV(79);
11153 sv_upgrade(tmpstr, SVt_PVIV);
11154 if (term == '\'') {
11155 op_type = OP_CONST;
11156 SvIV_set(tmpstr, -1);
11158 else if (term == '`') {
11159 op_type = OP_BACKTICK;
11160 SvIV_set(tmpstr, '\\');
11164 PL_multi_start = CopLINE(PL_curcop);
11165 PL_multi_open = PL_multi_close = '<';
11166 term = *PL_tokenbuf;
11167 if (PL_lex_inwhat == OP_SUBST && PL_in_eval && !PL_rsfp) {
11168 char * const bufptr = PL_sublex_info.super_bufptr;
11169 char * const bufend = PL_sublex_info.super_bufend;
11170 char * const olds = s - SvCUR(herewas);
11171 s = strchr(bufptr, '\n');
11175 while (s < bufend &&
11176 (*s != term || memNE(s,PL_tokenbuf,len)) ) {
11178 CopLINE_inc(PL_curcop);
11181 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
11182 missingterm(PL_tokenbuf);
11184 sv_setpvn(herewas,bufptr,d-bufptr+1);
11185 sv_setpvn(tmpstr,d+1,s-d);
11187 sv_catpvn(herewas,s,bufend-s);
11188 Copy(SvPVX_const(herewas),bufptr,SvCUR(herewas) + 1,char);
11195 while (s < PL_bufend &&
11196 (*s != term || memNE(s,PL_tokenbuf,len)) ) {
11198 CopLINE_inc(PL_curcop);
11200 if (s >= PL_bufend) {
11201 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
11202 missingterm(PL_tokenbuf);
11204 sv_setpvn(tmpstr,d+1,s-d);
11206 if (PL_madskills) {
11208 sv_catpvn(PL_thisstuff, d + 1, s - d);
11210 PL_thisstuff = newSVpvn(d + 1, s - d);
11211 stuffstart = s - SvPVX(PL_linestr);
11215 CopLINE_inc(PL_curcop); /* the preceding stmt passes a newline */
11217 sv_catpvn(herewas,s,PL_bufend-s);
11218 sv_setsv(PL_linestr,herewas);
11219 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = s = PL_linestart = SvPVX(PL_linestr);
11220 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
11221 PL_last_lop = PL_last_uni = NULL;
11224 sv_setpvn(tmpstr,"",0); /* avoid "uninitialized" warning */
11225 while (s >= PL_bufend) { /* multiple line string? */
11227 if (PL_madskills) {
11228 tstart = SvPVX(PL_linestr) + stuffstart;
11230 sv_catpvn(PL_thisstuff, tstart, PL_bufend - tstart);
11232 PL_thisstuff = newSVpvn(tstart, PL_bufend - tstart);
11236 !(PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = filter_gets(PL_linestr, PL_rsfp, 0))) {
11237 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
11238 missingterm(PL_tokenbuf);
11241 stuffstart = s - SvPVX(PL_linestr);
11243 CopLINE_inc(PL_curcop);
11244 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
11245 PL_last_lop = PL_last_uni = NULL;
11246 #ifndef PERL_STRICT_CR
11247 if (PL_bufend - PL_linestart >= 2) {
11248 if ((PL_bufend[-2] == '\r' && PL_bufend[-1] == '\n') ||
11249 (PL_bufend[-2] == '\n' && PL_bufend[-1] == '\r'))
11251 PL_bufend[-2] = '\n';
11253 SvCUR_set(PL_linestr, PL_bufend - SvPVX_const(PL_linestr));
11255 else if (PL_bufend[-1] == '\r')
11256 PL_bufend[-1] = '\n';
11258 else if (PL_bufend - PL_linestart == 1 && PL_bufend[-1] == '\r')
11259 PL_bufend[-1] = '\n';
11261 if (PERLDB_LINE && PL_curstash != PL_debstash)
11262 update_debugger_info(PL_linestr, NULL, 0);
11263 if (*s == term && memEQ(s,PL_tokenbuf,len)) {
11264 STRLEN off = PL_bufend - 1 - SvPVX_const(PL_linestr);
11265 *(SvPVX(PL_linestr) + off ) = ' ';
11266 sv_catsv(PL_linestr,herewas);
11267 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
11268 s = SvPVX(PL_linestr) + off; /* In case PV of PL_linestr moved. */
11272 sv_catsv(tmpstr,PL_linestr);
11277 PL_multi_end = CopLINE(PL_curcop);
11278 if (SvCUR(tmpstr) + 5 < SvLEN(tmpstr)) {
11279 SvPV_shrink_to_cur(tmpstr);
11281 SvREFCNT_dec(herewas);
11283 if (UTF && is_utf8_string((U8*)SvPVX_const(tmpstr), SvCUR(tmpstr)))
11285 else if (PL_encoding)
11286 sv_recode_to_utf8(tmpstr, PL_encoding);
11288 PL_lex_stuff = tmpstr;
11289 yylval.ival = op_type;
11293 /* scan_inputsymbol
11294 takes: current position in input buffer
11295 returns: new position in input buffer
11296 side-effects: yylval and lex_op are set.
11301 <FH> read from filehandle
11302 <pkg::FH> read from package qualified filehandle
11303 <pkg'FH> read from package qualified filehandle
11304 <$fh> read from filehandle in $fh
11305 <*.h> filename glob
11310 S_scan_inputsymbol(pTHX_ char *start)
11313 register char *s = start; /* current position in buffer */
11317 char *d = PL_tokenbuf; /* start of temp holding space */
11318 const char * const e = PL_tokenbuf + sizeof PL_tokenbuf; /* end of temp holding space */
11320 end = strchr(s, '\n');
11323 s = delimcpy(d, e, s + 1, end, '>', &len); /* extract until > */
11325 /* die if we didn't have space for the contents of the <>,
11326 or if it didn't end, or if we see a newline
11329 if (len >= (I32)sizeof PL_tokenbuf)
11330 Perl_croak(aTHX_ "Excessively long <> operator");
11332 Perl_croak(aTHX_ "Unterminated <> operator");
11337 Remember, only scalar variables are interpreted as filehandles by
11338 this code. Anything more complex (e.g., <$fh{$num}>) will be
11339 treated as a glob() call.
11340 This code makes use of the fact that except for the $ at the front,
11341 a scalar variable and a filehandle look the same.
11343 if (*d == '$' && d[1]) d++;
11345 /* allow <Pkg'VALUE> or <Pkg::VALUE> */
11346 while (*d && (isALNUM_lazy_if(d,UTF) || *d == '\'' || *d == ':'))
11349 /* If we've tried to read what we allow filehandles to look like, and
11350 there's still text left, then it must be a glob() and not a getline.
11351 Use scan_str to pull out the stuff between the <> and treat it
11352 as nothing more than a string.
11355 if (d - PL_tokenbuf != len) {
11356 yylval.ival = OP_GLOB;
11358 s = scan_str(start,!!PL_madskills,FALSE);
11360 Perl_croak(aTHX_ "Glob not terminated");
11364 bool readline_overriden = FALSE;
11367 /* we're in a filehandle read situation */
11370 /* turn <> into <ARGV> */
11372 Copy("ARGV",d,5,char);
11374 /* Check whether readline() is overriden */
11375 gv_readline = gv_fetchpvs("readline", GV_NOTQUAL, SVt_PVCV);
11377 && GvCVu(gv_readline) && GvIMPORTED_CV(gv_readline))
11379 ((gvp = (GV**)hv_fetchs(PL_globalstash, "readline", FALSE))
11380 && (gv_readline = *gvp) != (GV*)&PL_sv_undef
11381 && GvCVu(gv_readline) && GvIMPORTED_CV(gv_readline)))
11382 readline_overriden = TRUE;
11384 /* if <$fh>, create the ops to turn the variable into a
11388 /* try to find it in the pad for this block, otherwise find
11389 add symbol table ops
11391 const PADOFFSET tmp = pad_findmy(d);
11392 if (tmp != NOT_IN_PAD) {
11393 if (PAD_COMPNAME_FLAGS_isOUR(tmp)) {
11394 HV * const stash = PAD_COMPNAME_OURSTASH(tmp);
11395 HEK * const stashname = HvNAME_HEK(stash);
11396 SV * const sym = sv_2mortal(newSVhek(stashname));
11397 sv_catpvs(sym, "::");
11398 sv_catpv(sym, d+1);
11403 OP * const o = newOP(OP_PADSV, 0);
11405 PL_lex_op = readline_overriden
11406 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
11407 append_elem(OP_LIST, o,
11408 newCVREF(0, newGVOP(OP_GV,0,gv_readline))))
11409 : (OP*)newUNOP(OP_READLINE, 0, o);
11418 ? (GV_ADDMULTI | GV_ADDINEVAL)
11421 PL_lex_op = readline_overriden
11422 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
11423 append_elem(OP_LIST,
11424 newUNOP(OP_RV2SV, 0, newGVOP(OP_GV, 0, gv)),
11425 newCVREF(0, newGVOP(OP_GV, 0, gv_readline))))
11426 : (OP*)newUNOP(OP_READLINE, 0,
11427 newUNOP(OP_RV2SV, 0,
11428 newGVOP(OP_GV, 0, gv)));
11430 if (!readline_overriden)
11431 PL_lex_op->op_flags |= OPf_SPECIAL;
11432 /* we created the ops in PL_lex_op, so make yylval.ival a null op */
11433 yylval.ival = OP_NULL;
11436 /* If it's none of the above, it must be a literal filehandle
11437 (<Foo::BAR> or <FOO>) so build a simple readline OP */
11439 GV * const gv = gv_fetchpv(d, GV_ADD, SVt_PVIO);
11440 PL_lex_op = readline_overriden
11441 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
11442 append_elem(OP_LIST,
11443 newGVOP(OP_GV, 0, gv),
11444 newCVREF(0, newGVOP(OP_GV, 0, gv_readline))))
11445 : (OP*)newUNOP(OP_READLINE, 0, newGVOP(OP_GV, 0, gv));
11446 yylval.ival = OP_NULL;
11455 takes: start position in buffer
11456 keep_quoted preserve \ on the embedded delimiter(s)
11457 keep_delims preserve the delimiters around the string
11458 returns: position to continue reading from buffer
11459 side-effects: multi_start, multi_close, lex_repl or lex_stuff, and
11460 updates the read buffer.
11462 This subroutine pulls a string out of the input. It is called for:
11463 q single quotes q(literal text)
11464 ' single quotes 'literal text'
11465 qq double quotes qq(interpolate $here please)
11466 " double quotes "interpolate $here please"
11467 qx backticks qx(/bin/ls -l)
11468 ` backticks `/bin/ls -l`
11469 qw quote words @EXPORT_OK = qw( func() $spam )
11470 m// regexp match m/this/
11471 s/// regexp substitute s/this/that/
11472 tr/// string transliterate tr/this/that/
11473 y/// string transliterate y/this/that/
11474 ($*@) sub prototypes sub foo ($)
11475 (stuff) sub attr parameters sub foo : attr(stuff)
11476 <> readline or globs <FOO>, <>, <$fh>, or <*.c>
11478 In most of these cases (all but <>, patterns and transliterate)
11479 yylex() calls scan_str(). m// makes yylex() call scan_pat() which
11480 calls scan_str(). s/// makes yylex() call scan_subst() which calls
11481 scan_str(). tr/// and y/// make yylex() call scan_trans() which
11484 It skips whitespace before the string starts, and treats the first
11485 character as the delimiter. If the delimiter is one of ([{< then
11486 the corresponding "close" character )]}> is used as the closing
11487 delimiter. It allows quoting of delimiters, and if the string has
11488 balanced delimiters ([{<>}]) it allows nesting.
11490 On success, the SV with the resulting string is put into lex_stuff or,
11491 if that is already non-NULL, into lex_repl. The second case occurs only
11492 when parsing the RHS of the special constructs s/// and tr/// (y///).
11493 For convenience, the terminating delimiter character is stuffed into
11498 S_scan_str(pTHX_ char *start, int keep_quoted, int keep_delims)
11501 SV *sv; /* scalar value: string */
11502 const char *tmps; /* temp string, used for delimiter matching */
11503 register char *s = start; /* current position in the buffer */
11504 register char term; /* terminating character */
11505 register char *to; /* current position in the sv's data */
11506 I32 brackets = 1; /* bracket nesting level */
11507 bool has_utf8 = FALSE; /* is there any utf8 content? */
11508 I32 termcode; /* terminating char. code */
11509 U8 termstr[UTF8_MAXBYTES]; /* terminating string */
11510 STRLEN termlen; /* length of terminating string */
11511 int last_off = 0; /* last position for nesting bracket */
11517 /* skip space before the delimiter */
11523 if (PL_realtokenstart >= 0) {
11524 stuffstart = PL_realtokenstart;
11525 PL_realtokenstart = -1;
11528 stuffstart = start - SvPVX(PL_linestr);
11530 /* mark where we are, in case we need to report errors */
11533 /* after skipping whitespace, the next character is the terminator */
11536 termcode = termstr[0] = term;
11540 termcode = utf8_to_uvchr((U8*)s, &termlen);
11541 Copy(s, termstr, termlen, U8);
11542 if (!UTF8_IS_INVARIANT(term))
11546 /* mark where we are */
11547 PL_multi_start = CopLINE(PL_curcop);
11548 PL_multi_open = term;
11550 /* find corresponding closing delimiter */
11551 if (term && (tmps = strchr("([{< )]}> )]}>",term)))
11552 termcode = termstr[0] = term = tmps[5];
11554 PL_multi_close = term;
11556 /* create a new SV to hold the contents. 79 is the SV's initial length.
11557 What a random number. */
11559 sv_upgrade(sv, SVt_PVIV);
11560 SvIV_set(sv, termcode);
11561 (void)SvPOK_only(sv); /* validate pointer */
11563 /* move past delimiter and try to read a complete string */
11565 sv_catpvn(sv, s, termlen);
11568 tstart = SvPVX(PL_linestr) + stuffstart;
11569 if (!PL_thisopen && !keep_delims) {
11570 PL_thisopen = newSVpvn(tstart, s - tstart);
11571 stuffstart = s - SvPVX(PL_linestr);
11575 if (PL_encoding && !UTF) {
11579 int offset = s - SvPVX_const(PL_linestr);
11580 const bool found = sv_cat_decode(sv, PL_encoding, PL_linestr,
11581 &offset, (char*)termstr, termlen);
11582 const char * const ns = SvPVX_const(PL_linestr) + offset;
11583 char * const svlast = SvEND(sv) - 1;
11585 for (; s < ns; s++) {
11586 if (*s == '\n' && !PL_rsfp)
11587 CopLINE_inc(PL_curcop);
11590 goto read_more_line;
11592 /* handle quoted delimiters */
11593 if (SvCUR(sv) > 1 && *(svlast-1) == '\\') {
11595 for (t = svlast-2; t >= SvPVX_const(sv) && *t == '\\';)
11597 if ((svlast-1 - t) % 2) {
11598 if (!keep_quoted) {
11599 *(svlast-1) = term;
11601 SvCUR_set(sv, SvCUR(sv) - 1);
11606 if (PL_multi_open == PL_multi_close) {
11612 for (t = w = SvPVX(sv)+last_off; t < svlast; w++, t++) {
11613 /* At here, all closes are "was quoted" one,
11614 so we don't check PL_multi_close. */
11616 if (!keep_quoted && *(t+1) == PL_multi_open)
11621 else if (*t == PL_multi_open)
11629 SvCUR_set(sv, w - SvPVX_const(sv));
11631 last_off = w - SvPVX(sv);
11632 if (--brackets <= 0)
11637 if (!keep_delims) {
11638 SvCUR_set(sv, SvCUR(sv) - 1);
11644 /* extend sv if need be */
11645 SvGROW(sv, SvCUR(sv) + (PL_bufend - s) + 1);
11646 /* set 'to' to the next character in the sv's string */
11647 to = SvPVX(sv)+SvCUR(sv);
11649 /* if open delimiter is the close delimiter read unbridle */
11650 if (PL_multi_open == PL_multi_close) {
11651 for (; s < PL_bufend; s++,to++) {
11652 /* embedded newlines increment the current line number */
11653 if (*s == '\n' && !PL_rsfp)
11654 CopLINE_inc(PL_curcop);
11655 /* handle quoted delimiters */
11656 if (*s == '\\' && s+1 < PL_bufend && term != '\\') {
11657 if (!keep_quoted && s[1] == term)
11659 /* any other quotes are simply copied straight through */
11663 /* terminate when run out of buffer (the for() condition), or
11664 have found the terminator */
11665 else if (*s == term) {
11668 if (s+termlen <= PL_bufend && memEQ(s, (char*)termstr, termlen))
11671 else if (!has_utf8 && !UTF8_IS_INVARIANT((U8)*s) && UTF)
11677 /* if the terminator isn't the same as the start character (e.g.,
11678 matched brackets), we have to allow more in the quoting, and
11679 be prepared for nested brackets.
11682 /* read until we run out of string, or we find the terminator */
11683 for (; s < PL_bufend; s++,to++) {
11684 /* embedded newlines increment the line count */
11685 if (*s == '\n' && !PL_rsfp)
11686 CopLINE_inc(PL_curcop);
11687 /* backslashes can escape the open or closing characters */
11688 if (*s == '\\' && s+1 < PL_bufend) {
11689 if (!keep_quoted &&
11690 ((s[1] == PL_multi_open) || (s[1] == PL_multi_close)))
11695 /* allow nested opens and closes */
11696 else if (*s == PL_multi_close && --brackets <= 0)
11698 else if (*s == PL_multi_open)
11700 else if (!has_utf8 && !UTF8_IS_INVARIANT((U8)*s) && UTF)
11705 /* terminate the copied string and update the sv's end-of-string */
11707 SvCUR_set(sv, to - SvPVX_const(sv));
11710 * this next chunk reads more into the buffer if we're not done yet
11714 break; /* handle case where we are done yet :-) */
11716 #ifndef PERL_STRICT_CR
11717 if (to - SvPVX_const(sv) >= 2) {
11718 if ((to[-2] == '\r' && to[-1] == '\n') ||
11719 (to[-2] == '\n' && to[-1] == '\r'))
11723 SvCUR_set(sv, to - SvPVX_const(sv));
11725 else if (to[-1] == '\r')
11728 else if (to - SvPVX_const(sv) == 1 && to[-1] == '\r')
11733 /* if we're out of file, or a read fails, bail and reset the current
11734 line marker so we can report where the unterminated string began
11737 if (PL_madskills) {
11738 char * const tstart = SvPVX(PL_linestr) + stuffstart;
11740 sv_catpvn(PL_thisstuff, tstart, PL_bufend - tstart);
11742 PL_thisstuff = newSVpvn(tstart, PL_bufend - tstart);
11746 !(PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = filter_gets(PL_linestr, PL_rsfp, 0))) {
11748 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
11754 /* we read a line, so increment our line counter */
11755 CopLINE_inc(PL_curcop);
11757 /* update debugger info */
11758 if (PERLDB_LINE && PL_curstash != PL_debstash)
11759 update_debugger_info(PL_linestr, NULL, 0);
11761 /* having changed the buffer, we must update PL_bufend */
11762 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
11763 PL_last_lop = PL_last_uni = NULL;
11766 /* at this point, we have successfully read the delimited string */
11768 if (!PL_encoding || UTF) {
11770 if (PL_madskills) {
11771 char * const tstart = SvPVX(PL_linestr) + stuffstart;
11772 const int len = s - tstart;
11774 sv_catpvn(PL_thisstuff, tstart, len);
11776 PL_thisstuff = newSVpvn(tstart, len);
11777 if (!PL_thisclose && !keep_delims)
11778 PL_thisclose = newSVpvn(s,termlen);
11783 sv_catpvn(sv, s, termlen);
11788 if (PL_madskills) {
11789 char * const tstart = SvPVX(PL_linestr) + stuffstart;
11790 const int len = s - tstart - termlen;
11792 sv_catpvn(PL_thisstuff, tstart, len);
11794 PL_thisstuff = newSVpvn(tstart, len);
11795 if (!PL_thisclose && !keep_delims)
11796 PL_thisclose = newSVpvn(s - termlen,termlen);
11800 if (has_utf8 || PL_encoding)
11803 PL_multi_end = CopLINE(PL_curcop);
11805 /* if we allocated too much space, give some back */
11806 if (SvCUR(sv) + 5 < SvLEN(sv)) {
11807 SvLEN_set(sv, SvCUR(sv) + 1);
11808 SvPV_renew(sv, SvLEN(sv));
11811 /* decide whether this is the first or second quoted string we've read
11824 takes: pointer to position in buffer
11825 returns: pointer to new position in buffer
11826 side-effects: builds ops for the constant in yylval.op
11828 Read a number in any of the formats that Perl accepts:
11830 \d(_?\d)*(\.(\d(_?\d)*)?)?[Ee][\+\-]?(\d(_?\d)*) 12 12.34 12.
11831 \.\d(_?\d)*[Ee][\+\-]?(\d(_?\d)*) .34
11834 0x[0-9A-Fa-f](_?[0-9A-Fa-f])*
11836 Like most scan_ routines, it uses the PL_tokenbuf buffer to hold the
11839 If it reads a number without a decimal point or an exponent, it will
11840 try converting the number to an integer and see if it can do so
11841 without loss of precision.
11845 Perl_scan_num(pTHX_ const char *start, YYSTYPE* lvalp)
11848 register const char *s = start; /* current position in buffer */
11849 register char *d; /* destination in temp buffer */
11850 register char *e; /* end of temp buffer */
11851 NV nv; /* number read, as a double */
11852 SV *sv = NULL; /* place to put the converted number */
11853 bool floatit; /* boolean: int or float? */
11854 const char *lastub = NULL; /* position of last underbar */
11855 static char const number_too_long[] = "Number too long";
11857 /* We use the first character to decide what type of number this is */
11861 Perl_croak(aTHX_ "panic: scan_num");
11863 /* if it starts with a 0, it could be an octal number, a decimal in
11864 0.13 disguise, or a hexadecimal number, or a binary number. */
11868 u holds the "number so far"
11869 shift the power of 2 of the base
11870 (hex == 4, octal == 3, binary == 1)
11871 overflowed was the number more than we can hold?
11873 Shift is used when we add a digit. It also serves as an "are
11874 we in octal/hex/binary?" indicator to disallow hex characters
11875 when in octal mode.
11880 bool overflowed = FALSE;
11881 bool just_zero = TRUE; /* just plain 0 or binary number? */
11882 static const NV nvshift[5] = { 1.0, 2.0, 4.0, 8.0, 16.0 };
11883 static const char* const bases[5] =
11884 { "", "binary", "", "octal", "hexadecimal" };
11885 static const char* const Bases[5] =
11886 { "", "Binary", "", "Octal", "Hexadecimal" };
11887 static const char* const maxima[5] =
11889 "0b11111111111111111111111111111111",
11893 const char *base, *Base, *max;
11895 /* check for hex */
11900 } else if (s[1] == 'b') {
11905 /* check for a decimal in disguise */
11906 else if (s[1] == '.' || s[1] == 'e' || s[1] == 'E')
11908 /* so it must be octal */
11915 if (ckWARN(WARN_SYNTAX))
11916 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
11917 "Misplaced _ in number");
11921 base = bases[shift];
11922 Base = Bases[shift];
11923 max = maxima[shift];
11925 /* read the rest of the number */
11927 /* x is used in the overflow test,
11928 b is the digit we're adding on. */
11933 /* if we don't mention it, we're done */
11937 /* _ are ignored -- but warned about if consecutive */
11939 if (lastub && s == lastub + 1 && ckWARN(WARN_SYNTAX))
11940 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
11941 "Misplaced _ in number");
11945 /* 8 and 9 are not octal */
11946 case '8': case '9':
11948 yyerror(Perl_form(aTHX_ "Illegal octal digit '%c'", *s));
11952 case '2': case '3': case '4':
11953 case '5': case '6': case '7':
11955 yyerror(Perl_form(aTHX_ "Illegal binary digit '%c'", *s));
11958 case '0': case '1':
11959 b = *s++ & 15; /* ASCII digit -> value of digit */
11963 case 'a': case 'b': case 'c': case 'd': case 'e': case 'f':
11964 case 'A': case 'B': case 'C': case 'D': case 'E': case 'F':
11965 /* make sure they said 0x */
11968 b = (*s++ & 7) + 9;
11970 /* Prepare to put the digit we have onto the end
11971 of the number so far. We check for overflows.
11977 x = u << shift; /* make room for the digit */
11979 if ((x >> shift) != u
11980 && !(PL_hints & HINT_NEW_BINARY)) {
11983 if (ckWARN_d(WARN_OVERFLOW))
11984 Perl_warner(aTHX_ packWARN(WARN_OVERFLOW),
11985 "Integer overflow in %s number",
11988 u = x | b; /* add the digit to the end */
11991 n *= nvshift[shift];
11992 /* If an NV has not enough bits in its
11993 * mantissa to represent an UV this summing of
11994 * small low-order numbers is a waste of time
11995 * (because the NV cannot preserve the
11996 * low-order bits anyway): we could just
11997 * remember when did we overflow and in the
11998 * end just multiply n by the right
12006 /* if we get here, we had success: make a scalar value from
12011 /* final misplaced underbar check */
12012 if (s[-1] == '_') {
12013 if (ckWARN(WARN_SYNTAX))
12014 Perl_warner(aTHX_ packWARN(WARN_SYNTAX), "Misplaced _ in number");
12019 if (n > 4294967295.0 && ckWARN(WARN_PORTABLE))
12020 Perl_warner(aTHX_ packWARN(WARN_PORTABLE),
12021 "%s number > %s non-portable",
12027 if (u > 0xffffffff && ckWARN(WARN_PORTABLE))
12028 Perl_warner(aTHX_ packWARN(WARN_PORTABLE),
12029 "%s number > %s non-portable",
12034 if (just_zero && (PL_hints & HINT_NEW_INTEGER))
12035 sv = new_constant(start, s - start, "integer",
12037 else if (PL_hints & HINT_NEW_BINARY)
12038 sv = new_constant(start, s - start, "binary", sv, NULL, NULL);
12043 handle decimal numbers.
12044 we're also sent here when we read a 0 as the first digit
12046 case '1': case '2': case '3': case '4': case '5':
12047 case '6': case '7': case '8': case '9': case '.':
12050 e = PL_tokenbuf + sizeof PL_tokenbuf - 6; /* room for various punctuation */
12053 /* read next group of digits and _ and copy into d */
12054 while (isDIGIT(*s) || *s == '_') {
12055 /* skip underscores, checking for misplaced ones
12059 if (lastub && s == lastub + 1 && ckWARN(WARN_SYNTAX))
12060 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
12061 "Misplaced _ in number");
12065 /* check for end of fixed-length buffer */
12067 Perl_croak(aTHX_ number_too_long);
12068 /* if we're ok, copy the character */
12073 /* final misplaced underbar check */
12074 if (lastub && s == lastub + 1) {
12075 if (ckWARN(WARN_SYNTAX))
12076 Perl_warner(aTHX_ packWARN(WARN_SYNTAX), "Misplaced _ in number");
12079 /* read a decimal portion if there is one. avoid
12080 3..5 being interpreted as the number 3. followed
12083 if (*s == '.' && s[1] != '.') {
12088 if (ckWARN(WARN_SYNTAX))
12089 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
12090 "Misplaced _ in number");
12094 /* copy, ignoring underbars, until we run out of digits.
12096 for (; isDIGIT(*s) || *s == '_'; s++) {
12097 /* fixed length buffer check */
12099 Perl_croak(aTHX_ number_too_long);
12101 if (lastub && s == lastub + 1 && ckWARN(WARN_SYNTAX))
12102 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
12103 "Misplaced _ in number");
12109 /* fractional part ending in underbar? */
12110 if (s[-1] == '_') {
12111 if (ckWARN(WARN_SYNTAX))
12112 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
12113 "Misplaced _ in number");
12115 if (*s == '.' && isDIGIT(s[1])) {
12116 /* oops, it's really a v-string, but without the "v" */
12122 /* read exponent part, if present */
12123 if ((*s == 'e' || *s == 'E') && strchr("+-0123456789_", s[1])) {
12127 /* regardless of whether user said 3E5 or 3e5, use lower 'e' */
12128 *d++ = 'e'; /* At least some Mach atof()s don't grok 'E' */
12130 /* stray preinitial _ */
12132 if (ckWARN(WARN_SYNTAX))
12133 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
12134 "Misplaced _ in number");
12138 /* allow positive or negative exponent */
12139 if (*s == '+' || *s == '-')
12142 /* stray initial _ */
12144 if (ckWARN(WARN_SYNTAX))
12145 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
12146 "Misplaced _ in number");
12150 /* read digits of exponent */
12151 while (isDIGIT(*s) || *s == '_') {
12154 Perl_croak(aTHX_ number_too_long);
12158 if (((lastub && s == lastub + 1) ||
12159 (!isDIGIT(s[1]) && s[1] != '_'))
12160 && ckWARN(WARN_SYNTAX))
12161 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
12162 "Misplaced _ in number");
12169 /* make an sv from the string */
12173 We try to do an integer conversion first if no characters
12174 indicating "float" have been found.
12179 const int flags = grok_number (PL_tokenbuf, d - PL_tokenbuf, &uv);
12181 if (flags == IS_NUMBER_IN_UV) {
12183 sv_setiv(sv, uv); /* Prefer IVs over UVs. */
12186 } else if (flags == (IS_NUMBER_IN_UV | IS_NUMBER_NEG)) {
12187 if (uv <= (UV) IV_MIN)
12188 sv_setiv(sv, -(IV)uv);
12195 /* terminate the string */
12197 nv = Atof(PL_tokenbuf);
12201 if ( floatit ? (PL_hints & HINT_NEW_FLOAT) :
12202 (PL_hints & HINT_NEW_INTEGER) )
12203 sv = new_constant(PL_tokenbuf,
12206 (floatit ? "float" : "integer"),
12210 /* if it starts with a v, it could be a v-string */
12213 sv = newSV(5); /* preallocate storage space */
12214 s = scan_vstring(s,sv);
12218 /* make the op for the constant and return */
12221 lvalp->opval = newSVOP(OP_CONST, 0, sv);
12223 lvalp->opval = NULL;
12229 S_scan_formline(pTHX_ register char *s)
12232 register char *eol;
12234 SV * const stuff = newSVpvs("");
12235 bool needargs = FALSE;
12236 bool eofmt = FALSE;
12238 char *tokenstart = s;
12241 if (PL_madskills) {
12242 savewhite = PL_thiswhite;
12247 while (!needargs) {
12250 #ifdef PERL_STRICT_CR
12251 while (SPACE_OR_TAB(*t))
12254 while (SPACE_OR_TAB(*t) || *t == '\r')
12257 if (*t == '\n' || t == PL_bufend) {
12262 if (PL_in_eval && !PL_rsfp) {
12263 eol = (char *) memchr(s,'\n',PL_bufend-s);
12268 eol = PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
12270 for (t = s; t < eol; t++) {
12271 if (*t == '~' && t[1] == '~' && SvCUR(stuff)) {
12273 goto enough; /* ~~ must be first line in formline */
12275 if (*t == '@' || *t == '^')
12279 sv_catpvn(stuff, s, eol-s);
12280 #ifndef PERL_STRICT_CR
12281 if (eol-s > 1 && eol[-2] == '\r' && eol[-1] == '\n') {
12282 char *end = SvPVX(stuff) + SvCUR(stuff);
12285 SvCUR_set(stuff, SvCUR(stuff) - 1);
12295 if (PL_madskills) {
12297 sv_catpvn(PL_thistoken, tokenstart, PL_bufend - tokenstart);
12299 PL_thistoken = newSVpvn(tokenstart, PL_bufend - tokenstart);
12302 s = filter_gets(PL_linestr, PL_rsfp, 0);
12304 tokenstart = PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = SvPVX(PL_linestr);
12306 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = SvPVX(PL_linestr);
12308 PL_bufend = PL_bufptr + SvCUR(PL_linestr);
12309 PL_last_lop = PL_last_uni = NULL;
12318 if (SvCUR(stuff)) {
12321 PL_lex_state = LEX_NORMAL;
12322 start_force(PL_curforce);
12323 NEXTVAL_NEXTTOKE.ival = 0;
12327 PL_lex_state = LEX_FORMLINE;
12329 if (UTF && is_utf8_string((U8*)SvPVX_const(stuff), SvCUR(stuff)))
12331 else if (PL_encoding)
12332 sv_recode_to_utf8(stuff, PL_encoding);
12334 start_force(PL_curforce);
12335 NEXTVAL_NEXTTOKE.opval = (OP*)newSVOP(OP_CONST, 0, stuff);
12337 start_force(PL_curforce);
12338 NEXTVAL_NEXTTOKE.ival = OP_FORMLINE;
12342 SvREFCNT_dec(stuff);
12344 PL_lex_formbrack = 0;
12348 if (PL_madskills) {
12350 sv_catpvn(PL_thistoken, tokenstart, s - tokenstart);
12352 PL_thistoken = newSVpvn(tokenstart, s - tokenstart);
12353 PL_thiswhite = savewhite;
12365 PL_cshlen = strlen(PL_cshname);
12367 #if defined(USE_ITHREADS)
12368 PERL_UNUSED_CONTEXT;
12374 Perl_start_subparse(pTHX_ I32 is_format, U32 flags)
12377 const I32 oldsavestack_ix = PL_savestack_ix;
12378 CV* const outsidecv = PL_compcv;
12381 assert(SvTYPE(PL_compcv) == SVt_PVCV);
12383 SAVEI32(PL_subline);
12384 save_item(PL_subname);
12385 SAVESPTR(PL_compcv);
12387 PL_compcv = (CV*)newSV(0);
12388 sv_upgrade((SV *)PL_compcv, is_format ? SVt_PVFM : SVt_PVCV);
12389 CvFLAGS(PL_compcv) |= flags;
12391 PL_subline = CopLINE(PL_curcop);
12392 CvPADLIST(PL_compcv) = pad_new(padnew_SAVE|padnew_SAVESUB);
12393 CvOUTSIDE(PL_compcv) = (CV*)SvREFCNT_inc_simple(outsidecv);
12394 CvOUTSIDE_SEQ(PL_compcv) = PL_cop_seqmax;
12396 return oldsavestack_ix;
12400 #pragma segment Perl_yylex
12403 Perl_yywarn(pTHX_ const char *s)
12406 PL_in_eval |= EVAL_WARNONLY;
12408 PL_in_eval &= ~EVAL_WARNONLY;
12413 Perl_yyerror(pTHX_ const char *s)
12416 const char *where = NULL;
12417 const char *context = NULL;
12420 int yychar = PL_parser->yychar;
12422 if (!yychar || (yychar == ';' && !PL_rsfp))
12424 else if (PL_oldoldbufptr && PL_bufptr > PL_oldoldbufptr &&
12425 PL_bufptr - PL_oldoldbufptr < 200 && PL_oldoldbufptr != PL_oldbufptr &&
12426 PL_oldbufptr != PL_bufptr) {
12429 The code below is removed for NetWare because it abends/crashes on NetWare
12430 when the script has error such as not having the closing quotes like:
12431 if ($var eq "value)
12432 Checking of white spaces is anyway done in NetWare code.
12435 while (isSPACE(*PL_oldoldbufptr))
12438 context = PL_oldoldbufptr;
12439 contlen = PL_bufptr - PL_oldoldbufptr;
12441 else if (PL_oldbufptr && PL_bufptr > PL_oldbufptr &&
12442 PL_bufptr - PL_oldbufptr < 200 && PL_oldbufptr != PL_bufptr) {
12445 The code below is removed for NetWare because it abends/crashes on NetWare
12446 when the script has error such as not having the closing quotes like:
12447 if ($var eq "value)
12448 Checking of white spaces is anyway done in NetWare code.
12451 while (isSPACE(*PL_oldbufptr))
12454 context = PL_oldbufptr;
12455 contlen = PL_bufptr - PL_oldbufptr;
12457 else if (yychar > 255)
12458 where = "next token ???";
12459 else if (yychar == -2) { /* YYEMPTY */
12460 if (PL_lex_state == LEX_NORMAL ||
12461 (PL_lex_state == LEX_KNOWNEXT && PL_lex_defer == LEX_NORMAL))
12462 where = "at end of line";
12463 else if (PL_lex_inpat)
12464 where = "within pattern";
12466 where = "within string";
12469 SV * const where_sv = sv_2mortal(newSVpvs("next char "));
12471 Perl_sv_catpvf(aTHX_ where_sv, "^%c", toCTRL(yychar));
12472 else if (isPRINT_LC(yychar))
12473 Perl_sv_catpvf(aTHX_ where_sv, "%c", yychar);
12475 Perl_sv_catpvf(aTHX_ where_sv, "\\%03o", yychar & 255);
12476 where = SvPVX_const(where_sv);
12478 msg = sv_2mortal(newSVpv(s, 0));
12479 Perl_sv_catpvf(aTHX_ msg, " at %s line %"IVdf", ",
12480 OutCopFILE(PL_curcop), (IV)CopLINE(PL_curcop));
12482 Perl_sv_catpvf(aTHX_ msg, "near \"%.*s\"\n", contlen, context);
12484 Perl_sv_catpvf(aTHX_ msg, "%s\n", where);
12485 if (PL_multi_start < PL_multi_end && (U32)(CopLINE(PL_curcop) - PL_multi_end) <= 1) {
12486 Perl_sv_catpvf(aTHX_ msg,
12487 " (Might be a runaway multi-line %c%c string starting on line %"IVdf")\n",
12488 (int)PL_multi_open,(int)PL_multi_close,(IV)PL_multi_start);
12491 if (PL_in_eval & EVAL_WARNONLY && ckWARN_d(WARN_SYNTAX))
12492 Perl_warner(aTHX_ packWARN(WARN_SYNTAX), "%"SVf, SVfARG(msg));
12495 if (PL_error_count >= 10) {
12496 if (PL_in_eval && SvCUR(ERRSV))
12497 Perl_croak(aTHX_ "%"SVf"%s has too many errors.\n",
12498 SVfARG(ERRSV), OutCopFILE(PL_curcop));
12500 Perl_croak(aTHX_ "%s has too many errors.\n",
12501 OutCopFILE(PL_curcop));
12504 PL_in_my_stash = NULL;
12508 #pragma segment Main
12512 S_swallow_bom(pTHX_ U8 *s)
12515 const STRLEN slen = SvCUR(PL_linestr);
12518 if (s[1] == 0xFE) {
12519 /* UTF-16 little-endian? (or UTF32-LE?) */
12520 if (s[2] == 0 && s[3] == 0) /* UTF-32 little-endian */
12521 Perl_croak(aTHX_ "Unsupported script encoding UTF32-LE");
12522 #ifndef PERL_NO_UTF16_FILTER
12523 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF16-LE script encoding (BOM)\n");
12526 if (PL_bufend > (char*)s) {
12530 filter_add(utf16rev_textfilter, NULL);
12531 Newx(news, (PL_bufend - (char*)s) * 3 / 2 + 1, U8);
12532 utf16_to_utf8_reversed(s, news,
12533 PL_bufend - (char*)s - 1,
12535 sv_setpvn(PL_linestr, (const char*)news, newlen);
12537 s = (U8*)SvPVX(PL_linestr);
12538 Copy(news, s, newlen, U8);
12542 SvUTF8_on(PL_linestr);
12543 s = (U8*)SvPVX(PL_linestr);
12545 /* FIXME - is this a general bug fix? */
12548 PL_bufend = SvPVX(PL_linestr) + newlen;
12551 Perl_croak(aTHX_ "Unsupported script encoding UTF16-LE");
12556 if (s[1] == 0xFF) { /* UTF-16 big-endian? */
12557 #ifndef PERL_NO_UTF16_FILTER
12558 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16BE script encoding (BOM)\n");
12561 if (PL_bufend > (char *)s) {
12565 filter_add(utf16_textfilter, NULL);
12566 Newx(news, (PL_bufend - (char*)s) * 3 / 2 + 1, U8);
12567 utf16_to_utf8(s, news,
12568 PL_bufend - (char*)s,
12570 sv_setpvn(PL_linestr, (const char*)news, newlen);
12572 SvUTF8_on(PL_linestr);
12573 s = (U8*)SvPVX(PL_linestr);
12574 PL_bufend = SvPVX(PL_linestr) + newlen;
12577 Perl_croak(aTHX_ "Unsupported script encoding UTF16-BE");
12582 if (slen > 2 && s[1] == 0xBB && s[2] == 0xBF) {
12583 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-8 script encoding (BOM)\n");
12584 s += 3; /* UTF-8 */
12590 if (s[2] == 0xFE && s[3] == 0xFF) {
12591 /* UTF-32 big-endian */
12592 Perl_croak(aTHX_ "Unsupported script encoding UTF32-BE");
12595 else if (s[2] == 0 && s[3] != 0) {
12598 * are a good indicator of UTF-16BE. */
12599 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16BE script encoding (no BOM)\n");
12605 if (slen > 3 && s[1] == 0x73 && s[2] == 0x66 && s[3] == 0x73) {
12606 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-8 script encoding (BOM)\n");
12607 s += 4; /* UTF-8 */
12613 if (slen > 3 && s[1] == 0 && s[2] != 0 && s[3] == 0) {
12616 * are a good indicator of UTF-16LE. */
12617 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16LE script encoding (no BOM)\n");
12626 * Restore a source filter.
12630 restore_rsfp(pTHX_ void *f)
12633 PerlIO * const fp = (PerlIO*)f;
12635 if (PL_rsfp == PerlIO_stdin())
12636 PerlIO_clearerr(PL_rsfp);
12637 else if (PL_rsfp && (PL_rsfp != fp))
12638 PerlIO_close(PL_rsfp);
12642 #ifndef PERL_NO_UTF16_FILTER
12644 utf16_textfilter(pTHX_ int idx, SV *sv, int maxlen)
12647 const STRLEN old = SvCUR(sv);
12648 const I32 count = FILTER_READ(idx+1, sv, maxlen);
12649 DEBUG_P(PerlIO_printf(Perl_debug_log,
12650 "utf16_textfilter(%p): %d %d (%d)\n",
12651 FPTR2DPTR(void *, utf16_textfilter),
12652 idx, maxlen, (int) count));
12656 Newx(tmps, SvCUR(sv) * 3 / 2 + 1, U8);
12657 Copy(SvPVX_const(sv), tmps, old, char);
12658 utf16_to_utf8((U8*)SvPVX_const(sv) + old, tmps + old,
12659 SvCUR(sv) - old, &newlen);
12660 sv_usepvn(sv, (char*)tmps, (STRLEN)newlen + old);
12662 DEBUG_P({sv_dump(sv);});
12667 utf16rev_textfilter(pTHX_ int idx, SV *sv, int maxlen)
12670 const STRLEN old = SvCUR(sv);
12671 const I32 count = FILTER_READ(idx+1, sv, maxlen);
12672 DEBUG_P(PerlIO_printf(Perl_debug_log,
12673 "utf16rev_textfilter(%p): %d %d (%d)\n",
12674 FPTR2DPTR(void *, utf16rev_textfilter),
12675 idx, maxlen, (int) count));
12679 Newx(tmps, SvCUR(sv) * 3 / 2 + 1, U8);
12680 Copy(SvPVX_const(sv), tmps, old, char);
12681 utf16_to_utf8((U8*)SvPVX_const(sv) + old, tmps + old,
12682 SvCUR(sv) - old, &newlen);
12683 sv_usepvn(sv, (char*)tmps, (STRLEN)newlen + old);
12685 DEBUG_P({ sv_dump(sv); });
12691 Returns a pointer to the next character after the parsed
12692 vstring, as well as updating the passed in sv.
12694 Function must be called like
12697 s = scan_vstring(s,sv);
12699 The sv should already be large enough to store the vstring
12700 passed in, for performance reasons.
12705 Perl_scan_vstring(pTHX_ const char *s, SV *sv)
12708 const char *pos = s;
12709 const char *start = s;
12710 if (*pos == 'v') pos++; /* get past 'v' */
12711 while (pos < PL_bufend && (isDIGIT(*pos) || *pos == '_'))
12713 if ( *pos != '.') {
12714 /* this may not be a v-string if followed by => */
12715 const char *next = pos;
12716 while (next < PL_bufend && isSPACE(*next))
12718 if ((PL_bufend - next) >= 2 && *next == '=' && next[1] == '>' ) {
12719 /* return string not v-string */
12720 sv_setpvn(sv,(char *)s,pos-s);
12721 return (char *)pos;
12725 if (!isALPHA(*pos)) {
12726 U8 tmpbuf[UTF8_MAXBYTES+1];
12729 s++; /* get past 'v' */
12731 sv_setpvn(sv, "", 0);
12734 /* this is atoi() that tolerates underscores */
12737 const char *end = pos;
12739 while (--end >= s) {
12741 const UV orev = rev;
12742 rev += (*end - '0') * mult;
12744 if (orev > rev && ckWARN_d(WARN_OVERFLOW))
12745 Perl_warner(aTHX_ packWARN(WARN_OVERFLOW),
12746 "Integer overflow in decimal number");
12750 if (rev > 0x7FFFFFFF)
12751 Perl_croak(aTHX_ "In EBCDIC the v-string components cannot exceed 2147483647");
12753 /* Append native character for the rev point */
12754 tmpend = uvchr_to_utf8(tmpbuf, rev);
12755 sv_catpvn(sv, (const char*)tmpbuf, tmpend - tmpbuf);
12756 if (!UNI_IS_INVARIANT(NATIVE_TO_UNI(rev)))
12758 if (pos + 1 < PL_bufend && *pos == '.' && isDIGIT(pos[1]))
12764 while (pos < PL_bufend && (isDIGIT(*pos) || *pos == '_'))
12768 sv_magic(sv,NULL,PERL_MAGIC_vstring,(const char*)start, pos-start);
12776 * c-indentation-style: bsd
12777 * c-basic-offset: 4
12778 * indent-tabs-mode: t
12781 * ex: set ts=8 sts=4 sw=4 noet: