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 char smallbuf[128], smallbuf2[128];
792 char *tmpbuf, *tmpbuf2;
794 STRLEN tmplen2 = len;
795 if (tmplen + 2 <= sizeof smallbuf)
798 Newx(tmpbuf, tmplen + 2, char);
799 if (tmplen2 + 2 <= sizeof smallbuf2)
802 Newx(tmpbuf2, tmplen2 + 2, char);
803 tmpbuf[0] = tmpbuf2[0] = '_';
804 tmpbuf[1] = tmpbuf2[1] = '<';
805 memcpy(tmpbuf + 2, cf, tmplen);
806 memcpy(tmpbuf2 + 2, s, tmplen2);
807 tmplen += 2; tmplen2 += 2;
808 gvp = (GV**)hv_fetch(PL_defstash, tmpbuf, tmplen, FALSE);
810 gv2 = *(GV**)hv_fetch(PL_defstash, tmpbuf2, tmplen2, TRUE);
812 gv_init(gv2, PL_defstash, tmpbuf2, tmplen2, FALSE);
813 /* adjust ${"::_<newfilename"} to store the new file name */
814 GvSV(gv2) = newSVpvn(tmpbuf2 + 2, tmplen2 - 2);
815 GvHV(gv2) = (HV*)SvREFCNT_inc(GvHV(*gvp));
816 GvAV(gv2) = (AV*)SvREFCNT_inc(GvAV(*gvp));
819 if (tmpbuf != smallbuf) Safefree(tmpbuf);
820 if (tmpbuf2 != smallbuf2) Safefree(tmpbuf2);
823 CopFILE_free(PL_curcop);
824 CopFILE_setn(PL_curcop, s, len);
826 CopLINE_set(PL_curcop, atoi(n)-1);
830 /* skip space before PL_thistoken */
833 S_skipspace0(pTHX_ register char *s)
840 PL_thiswhite = newSVpvs("");
841 sv_catsv(PL_thiswhite, PL_skipwhite);
842 sv_free(PL_skipwhite);
845 PL_realtokenstart = s - SvPVX(PL_linestr);
849 /* skip space after PL_thistoken */
852 S_skipspace1(pTHX_ register char *s)
854 const char *start = s;
855 I32 startoff = start - SvPVX(PL_linestr);
860 start = SvPVX(PL_linestr) + startoff;
861 if (!PL_thistoken && PL_realtokenstart >= 0) {
862 const char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
863 PL_thistoken = newSVpvn(tstart, start - tstart);
865 PL_realtokenstart = -1;
868 PL_nextwhite = newSVpvs("");
869 sv_catsv(PL_nextwhite, PL_skipwhite);
870 sv_free(PL_skipwhite);
877 S_skipspace2(pTHX_ register char *s, SV **svp)
880 const I32 bufptroff = PL_bufptr - SvPVX(PL_linestr);
881 const I32 startoff = s - SvPVX(PL_linestr);
884 PL_bufptr = SvPVX(PL_linestr) + bufptroff;
885 if (!PL_madskills || !svp)
887 start = SvPVX(PL_linestr) + startoff;
888 if (!PL_thistoken && PL_realtokenstart >= 0) {
889 char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
890 PL_thistoken = newSVpvn(tstart, start - tstart);
891 PL_realtokenstart = -1;
896 sv_setsv(*svp, PL_skipwhite);
897 sv_free(PL_skipwhite);
906 S_update_debugger_info(pTHX_ SV *orig_sv, const char *buf, STRLEN len)
908 AV *av = CopFILEAVx(PL_curcop);
910 SV * const sv = newSV(0);
911 sv_upgrade(sv, SVt_PVMG);
913 sv_setsv(sv, orig_sv);
915 sv_setpvn(sv, buf, len);
918 av_store(av, (I32)CopLINE(PL_curcop), sv);
924 * Called to gobble the appropriate amount and type of whitespace.
925 * Skips comments as well.
929 S_skipspace(pTHX_ register char *s)
934 int startoff = s - SvPVX(PL_linestr);
937 sv_free(PL_skipwhite);
942 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
943 while (s < PL_bufend && SPACE_OR_TAB(*s))
953 SSize_t oldprevlen, oldoldprevlen;
954 SSize_t oldloplen = 0, oldunilen = 0;
955 while (s < PL_bufend && isSPACE(*s)) {
956 if (*s++ == '\n' && PL_in_eval && !PL_rsfp)
961 if (s < PL_bufend && *s == '#') {
962 while (s < PL_bufend && *s != '\n')
966 if (PL_in_eval && !PL_rsfp) {
973 /* only continue to recharge the buffer if we're at the end
974 * of the buffer, we're not reading from a source filter, and
975 * we're in normal lexing mode
977 if (s < PL_bufend || !PL_rsfp || PL_sublex_info.sub_inwhat ||
978 PL_lex_state == LEX_FORMLINE)
985 /* try to recharge the buffer */
987 curoff = s - SvPVX(PL_linestr);
990 if ((s = filter_gets(PL_linestr, PL_rsfp,
991 (prevlen = SvCUR(PL_linestr)))) == NULL)
994 if (PL_madskills && curoff != startoff) {
996 PL_skipwhite = newSVpvs("");
997 sv_catpvn(PL_skipwhite, SvPVX(PL_linestr) + startoff,
1001 /* mustn't throw out old stuff yet if madpropping */
1002 SvCUR(PL_linestr) = curoff;
1003 s = SvPVX(PL_linestr) + curoff;
1005 if (curoff && s[-1] == '\n')
1009 /* end of file. Add on the -p or -n magic */
1010 /* XXX these shouldn't really be added here, can't set PL_faketokens */
1013 sv_catpv(PL_linestr,
1014 ";}continue{print or die qq(-p destination: $!\\n);}");
1016 sv_setpv(PL_linestr,
1017 ";}continue{print or die qq(-p destination: $!\\n);}");
1019 PL_minus_n = PL_minus_p = 0;
1021 else if (PL_minus_n) {
1023 sv_catpvn(PL_linestr, ";}", 2);
1025 sv_setpvn(PL_linestr, ";}", 2);
1031 sv_catpvn(PL_linestr,";", 1);
1033 sv_setpvn(PL_linestr,";", 1);
1036 /* reset variables for next time we lex */
1037 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = s = PL_linestart
1043 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
1044 PL_last_lop = PL_last_uni = NULL;
1046 /* Close the filehandle. Could be from -P preprocessor,
1047 * STDIN, or a regular file. If we were reading code from
1048 * STDIN (because the commandline held no -e or filename)
1049 * then we don't close it, we reset it so the code can
1050 * read from STDIN too.
1053 if (PL_preprocess && !PL_in_eval)
1054 (void)PerlProc_pclose(PL_rsfp);
1055 else if ((PerlIO*)PL_rsfp == PerlIO_stdin())
1056 PerlIO_clearerr(PL_rsfp);
1058 (void)PerlIO_close(PL_rsfp);
1063 /* not at end of file, so we only read another line */
1064 /* make corresponding updates to old pointers, for yyerror() */
1065 oldprevlen = PL_oldbufptr - PL_bufend;
1066 oldoldprevlen = PL_oldoldbufptr - PL_bufend;
1068 oldunilen = PL_last_uni - PL_bufend;
1070 oldloplen = PL_last_lop - PL_bufend;
1071 PL_linestart = PL_bufptr = s + prevlen;
1072 PL_bufend = s + SvCUR(PL_linestr);
1074 PL_oldbufptr = s + oldprevlen;
1075 PL_oldoldbufptr = s + oldoldprevlen;
1077 PL_last_uni = s + oldunilen;
1079 PL_last_lop = s + oldloplen;
1082 /* debugger active and we're not compiling the debugger code,
1083 * so store the line into the debugger's array of lines
1085 if (PERLDB_LINE && PL_curstash != PL_debstash)
1086 update_debugger_info(NULL, PL_bufptr, PL_bufend - PL_bufptr);
1093 PL_skipwhite = newSVpvs("");
1094 curoff = s - SvPVX(PL_linestr);
1095 if (curoff - startoff)
1096 sv_catpvn(PL_skipwhite, SvPVX(PL_linestr) + startoff,
1105 * Check the unary operators to ensure there's no ambiguity in how they're
1106 * used. An ambiguous piece of code would be:
1108 * This doesn't mean rand() + 5. Because rand() is a unary operator,
1109 * the +5 is its argument.
1119 if (PL_oldoldbufptr != PL_last_uni)
1121 while (isSPACE(*PL_last_uni))
1124 while (isALNUM_lazy_if(s,UTF) || *s == '-')
1126 if ((t = strchr(s, '(')) && t < PL_bufptr)
1129 if (ckWARN_d(WARN_AMBIGUOUS)){
1130 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
1131 "Warning: Use of \"%.*s\" without parentheses is ambiguous",
1132 (int)(s - PL_last_uni), PL_last_uni);
1137 * LOP : macro to build a list operator. Its behaviour has been replaced
1138 * with a subroutine, S_lop() for which LOP is just another name.
1141 #define LOP(f,x) return lop(f,x,s)
1145 * Build a list operator (or something that might be one). The rules:
1146 * - if we have a next token, then it's a list operator [why?]
1147 * - if the next thing is an opening paren, then it's a function
1148 * - else it's a list operator
1152 S_lop(pTHX_ I32 f, int x, char *s)
1159 PL_last_lop = PL_oldbufptr;
1160 PL_last_lop_op = (OPCODE)f;
1163 return REPORT(LSTOP);
1166 return REPORT(LSTOP);
1169 return REPORT(FUNC);
1172 return REPORT(FUNC);
1174 return REPORT(LSTOP);
1180 * Sets up for an eventual force_next(). start_force(0) basically does
1181 * an unshift, while start_force(-1) does a push. yylex removes items
1186 S_start_force(pTHX_ int where)
1190 if (where < 0) /* so people can duplicate start_force(PL_curforce) */
1191 where = PL_lasttoke;
1192 assert(PL_curforce < 0 || PL_curforce == where);
1193 if (PL_curforce != where) {
1194 for (i = PL_lasttoke; i > where; --i) {
1195 PL_nexttoke[i] = PL_nexttoke[i-1];
1199 if (PL_curforce < 0) /* in case of duplicate start_force() */
1200 Zero(&PL_nexttoke[where], 1, NEXTTOKE);
1201 PL_curforce = where;
1204 curmad('^', newSVpvs(""));
1205 CURMAD('_', PL_nextwhite);
1210 S_curmad(pTHX_ char slot, SV *sv)
1216 if (PL_curforce < 0)
1217 where = &PL_thismad;
1219 where = &PL_nexttoke[PL_curforce].next_mad;
1222 sv_setpvn(sv, "", 0);
1225 if (UTF && is_utf8_string((U8*)SvPVX(sv), SvCUR(sv)))
1227 else if (PL_encoding) {
1228 sv_recode_to_utf8(sv, PL_encoding);
1233 /* keep a slot open for the head of the list? */
1234 if (slot != '_' && *where && (*where)->mad_key == '^') {
1235 (*where)->mad_key = slot;
1236 sv_free((*where)->mad_val);
1237 (*where)->mad_val = (void*)sv;
1240 addmad(newMADsv(slot, sv), where, 0);
1243 # define start_force(where) NOOP
1244 # define curmad(slot, sv) NOOP
1249 * When the lexer realizes it knows the next token (for instance,
1250 * it is reordering tokens for the parser) then it can call S_force_next
1251 * to know what token to return the next time the lexer is called. Caller
1252 * will need to set PL_nextval[] (or PL_nexttoke[].next_val with PERL_MAD),
1253 * and possibly PL_expect to ensure the lexer handles the token correctly.
1257 S_force_next(pTHX_ I32 type)
1261 if (PL_curforce < 0)
1262 start_force(PL_lasttoke);
1263 PL_nexttoke[PL_curforce].next_type = type;
1264 if (PL_lex_state != LEX_KNOWNEXT)
1265 PL_lex_defer = PL_lex_state;
1266 PL_lex_state = LEX_KNOWNEXT;
1267 PL_lex_expect = PL_expect;
1270 PL_nexttype[PL_nexttoke] = type;
1272 if (PL_lex_state != LEX_KNOWNEXT) {
1273 PL_lex_defer = PL_lex_state;
1274 PL_lex_expect = PL_expect;
1275 PL_lex_state = LEX_KNOWNEXT;
1281 S_newSV_maybe_utf8(pTHX_ const char *start, STRLEN len)
1284 SV * const sv = newSVpvn(start,len);
1285 if (UTF && !IN_BYTES && is_utf8_string((const U8*)start, len))
1292 * When the lexer knows the next thing is a word (for instance, it has
1293 * just seen -> and it knows that the next char is a word char, then
1294 * it calls S_force_word to stick the next word into the PL_nexttoke/val
1298 * char *start : buffer position (must be within PL_linestr)
1299 * int token : PL_next* will be this type of bare word (e.g., METHOD,WORD)
1300 * int check_keyword : if true, Perl checks to make sure the word isn't
1301 * a keyword (do this if the word is a label, e.g. goto FOO)
1302 * int allow_pack : if true, : characters will also be allowed (require,
1303 * use, etc. do this)
1304 * int allow_initial_tick : used by the "sub" lexer only.
1308 S_force_word(pTHX_ register char *start, int token, int check_keyword, int allow_pack, int allow_initial_tick)
1314 start = SKIPSPACE1(start);
1316 if (isIDFIRST_lazy_if(s,UTF) ||
1317 (allow_pack && *s == ':') ||
1318 (allow_initial_tick && *s == '\'') )
1320 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, allow_pack, &len);
1321 if (check_keyword && keyword(PL_tokenbuf, len, 0))
1323 start_force(PL_curforce);
1325 curmad('X', newSVpvn(start,s-start));
1326 if (token == METHOD) {
1331 PL_expect = XOPERATOR;
1334 NEXTVAL_NEXTTOKE.opval
1335 = (OP*)newSVOP(OP_CONST,0,
1336 S_newSV_maybe_utf8(aTHX_ PL_tokenbuf, len));
1337 NEXTVAL_NEXTTOKE.opval->op_private |= OPpCONST_BARE;
1345 * Called when the lexer wants $foo *foo &foo etc, but the program
1346 * text only contains the "foo" portion. The first argument is a pointer
1347 * to the "foo", and the second argument is the type symbol to prefix.
1348 * Forces the next token to be a "WORD".
1349 * Creates the symbol if it didn't already exist (via gv_fetchpv()).
1353 S_force_ident(pTHX_ register const char *s, int kind)
1357 const STRLEN len = strlen(s);
1358 OP* const o = (OP*)newSVOP(OP_CONST, 0, newSVpvn(s, len));
1359 start_force(PL_curforce);
1360 NEXTVAL_NEXTTOKE.opval = o;
1363 o->op_private = OPpCONST_ENTERED;
1364 /* XXX see note in pp_entereval() for why we forgo typo
1365 warnings if the symbol must be introduced in an eval.
1367 gv_fetchpvn_flags(s, len,
1368 PL_in_eval ? (GV_ADDMULTI | GV_ADDINEVAL)
1370 kind == '$' ? SVt_PV :
1371 kind == '@' ? SVt_PVAV :
1372 kind == '%' ? SVt_PVHV :
1380 Perl_str_to_version(pTHX_ SV *sv)
1385 const char *start = SvPV_const(sv,len);
1386 const char * const end = start + len;
1387 const bool utf = SvUTF8(sv) ? TRUE : FALSE;
1388 while (start < end) {
1392 n = utf8n_to_uvchr((U8*)start, len, &skip, 0);
1397 retval += ((NV)n)/nshift;
1406 * Forces the next token to be a version number.
1407 * If the next token appears to be an invalid version number, (e.g. "v2b"),
1408 * and if "guessing" is TRUE, then no new token is created (and the caller
1409 * must use an alternative parsing method).
1413 S_force_version(pTHX_ char *s, int guessing)
1419 I32 startoff = s - SvPVX(PL_linestr);
1428 while (isDIGIT(*d) || *d == '_' || *d == '.')
1432 start_force(PL_curforce);
1433 curmad('X', newSVpvn(s,d-s));
1436 if (*d == ';' || isSPACE(*d) || *d == '}' || !*d) {
1438 s = scan_num(s, &yylval);
1439 version = yylval.opval;
1440 ver = cSVOPx(version)->op_sv;
1441 if (SvPOK(ver) && !SvNIOK(ver)) {
1442 SvUPGRADE(ver, SVt_PVNV);
1443 SvNV_set(ver, str_to_version(ver));
1444 SvNOK_on(ver); /* hint that it is a version */
1447 else if (guessing) {
1450 sv_free(PL_nextwhite); /* let next token collect whitespace */
1452 s = SvPVX(PL_linestr) + startoff;
1460 if (PL_madskills && !version) {
1461 sv_free(PL_nextwhite); /* let next token collect whitespace */
1463 s = SvPVX(PL_linestr) + startoff;
1466 /* NOTE: The parser sees the package name and the VERSION swapped */
1467 start_force(PL_curforce);
1468 NEXTVAL_NEXTTOKE.opval = version;
1476 * Tokenize a quoted string passed in as an SV. It finds the next
1477 * chunk, up to end of string or a backslash. It may make a new
1478 * SV containing that chunk (if HINT_NEW_STRING is on). It also
1483 S_tokeq(pTHX_ SV *sv)
1487 register char *send;
1495 s = SvPV_force(sv, len);
1496 if (SvTYPE(sv) >= SVt_PVIV && SvIVX(sv) == -1)
1499 while (s < send && *s != '\\')
1504 if ( PL_hints & HINT_NEW_STRING ) {
1505 pv = sv_2mortal(newSVpvn(SvPVX_const(pv), len));
1511 if (s + 1 < send && (s[1] == '\\'))
1512 s++; /* all that, just for this */
1517 SvCUR_set(sv, d - SvPVX_const(sv));
1519 if ( PL_hints & HINT_NEW_STRING )
1520 return new_constant(NULL, 0, "q", sv, pv, "q");
1525 * Now come three functions related to double-quote context,
1526 * S_sublex_start, S_sublex_push, and S_sublex_done. They're used when
1527 * converting things like "\u\Lgnat" into ucfirst(lc("gnat")). They
1528 * interact with PL_lex_state, and create fake ( ... ) argument lists
1529 * to handle functions and concatenation.
1530 * They assume that whoever calls them will be setting up a fake
1531 * join call, because each subthing puts a ',' after it. This lets
1534 * join($, , 'lower ', lcfirst( 'uPpEr', ) ,)
1536 * (I'm not sure whether the spurious commas at the end of lcfirst's
1537 * arguments and join's arguments are created or not).
1542 * Assumes that yylval.ival is the op we're creating (e.g. OP_LCFIRST).
1544 * Pattern matching will set PL_lex_op to the pattern-matching op to
1545 * make (we return THING if yylval.ival is OP_NULL, PMFUNC otherwise).
1547 * OP_CONST and OP_READLINE are easy--just make the new op and return.
1549 * Everything else becomes a FUNC.
1551 * Sets PL_lex_state to LEX_INTERPPUSH unless (ival was OP_NULL or we
1552 * had an OP_CONST or OP_READLINE). This just sets us up for a
1553 * call to S_sublex_push().
1557 S_sublex_start(pTHX)
1560 register const I32 op_type = yylval.ival;
1562 if (op_type == OP_NULL) {
1563 yylval.opval = PL_lex_op;
1567 if (op_type == OP_CONST || op_type == OP_READLINE) {
1568 SV *sv = tokeq(PL_lex_stuff);
1570 if (SvTYPE(sv) == SVt_PVIV) {
1571 /* Overloaded constants, nothing fancy: Convert to SVt_PV: */
1573 const char * const p = SvPV_const(sv, len);
1574 SV * const nsv = newSVpvn(p, len);
1580 yylval.opval = (OP*)newSVOP(op_type, 0, sv);
1581 PL_lex_stuff = NULL;
1582 /* Allow <FH> // "foo" */
1583 if (op_type == OP_READLINE)
1584 PL_expect = XTERMORDORDOR;
1587 else if (op_type == OP_BACKTICK && PL_lex_op) {
1588 /* readpipe() vas overriden */
1589 cSVOPx(cLISTOPx(cUNOPx(PL_lex_op)->op_first)->op_first->op_sibling)->op_sv = tokeq(PL_lex_stuff);
1590 yylval.opval = PL_lex_op;
1592 PL_lex_stuff = NULL;
1596 PL_sublex_info.super_state = PL_lex_state;
1597 PL_sublex_info.sub_inwhat = op_type;
1598 PL_sublex_info.sub_op = PL_lex_op;
1599 PL_lex_state = LEX_INTERPPUSH;
1603 yylval.opval = PL_lex_op;
1613 * Create a new scope to save the lexing state. The scope will be
1614 * ended in S_sublex_done. Returns a '(', starting the function arguments
1615 * to the uc, lc, etc. found before.
1616 * Sets PL_lex_state to LEX_INTERPCONCAT.
1625 PL_lex_state = PL_sublex_info.super_state;
1626 SAVEI32(PL_lex_dojoin);
1627 SAVEI32(PL_lex_brackets);
1628 SAVEI32(PL_lex_casemods);
1629 SAVEI32(PL_lex_starts);
1630 SAVEI32(PL_lex_state);
1631 SAVEVPTR(PL_lex_inpat);
1632 SAVEI32(PL_lex_inwhat);
1633 SAVECOPLINE(PL_curcop);
1634 SAVEPPTR(PL_bufptr);
1635 SAVEPPTR(PL_bufend);
1636 SAVEPPTR(PL_oldbufptr);
1637 SAVEPPTR(PL_oldoldbufptr);
1638 SAVEPPTR(PL_last_lop);
1639 SAVEPPTR(PL_last_uni);
1640 SAVEPPTR(PL_linestart);
1641 SAVESPTR(PL_linestr);
1642 SAVEGENERICPV(PL_lex_brackstack);
1643 SAVEGENERICPV(PL_lex_casestack);
1645 PL_linestr = PL_lex_stuff;
1646 PL_lex_stuff = NULL;
1648 PL_bufend = PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart
1649 = SvPVX(PL_linestr);
1650 PL_bufend += SvCUR(PL_linestr);
1651 PL_last_lop = PL_last_uni = NULL;
1652 SAVEFREESV(PL_linestr);
1654 PL_lex_dojoin = FALSE;
1655 PL_lex_brackets = 0;
1656 Newx(PL_lex_brackstack, 120, char);
1657 Newx(PL_lex_casestack, 12, char);
1658 PL_lex_casemods = 0;
1659 *PL_lex_casestack = '\0';
1661 PL_lex_state = LEX_INTERPCONCAT;
1662 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
1664 PL_lex_inwhat = PL_sublex_info.sub_inwhat;
1665 if (PL_lex_inwhat == OP_MATCH || PL_lex_inwhat == OP_QR || PL_lex_inwhat == OP_SUBST)
1666 PL_lex_inpat = PL_sublex_info.sub_op;
1668 PL_lex_inpat = NULL;
1675 * Restores lexer state after a S_sublex_push.
1682 if (!PL_lex_starts++) {
1683 SV * const sv = newSVpvs("");
1684 if (SvUTF8(PL_linestr))
1686 PL_expect = XOPERATOR;
1687 yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
1691 if (PL_lex_casemods) { /* oops, we've got some unbalanced parens */
1692 PL_lex_state = LEX_INTERPCASEMOD;
1696 /* Is there a right-hand side to take care of? (s//RHS/ or tr//RHS/) */
1697 if (PL_lex_repl && (PL_lex_inwhat == OP_SUBST || PL_lex_inwhat == OP_TRANS)) {
1698 PL_linestr = PL_lex_repl;
1700 PL_bufend = PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart = SvPVX(PL_linestr);
1701 PL_bufend += SvCUR(PL_linestr);
1702 PL_last_lop = PL_last_uni = NULL;
1703 SAVEFREESV(PL_linestr);
1704 PL_lex_dojoin = FALSE;
1705 PL_lex_brackets = 0;
1706 PL_lex_casemods = 0;
1707 *PL_lex_casestack = '\0';
1709 if (SvEVALED(PL_lex_repl)) {
1710 PL_lex_state = LEX_INTERPNORMAL;
1712 /* we don't clear PL_lex_repl here, so that we can check later
1713 whether this is an evalled subst; that means we rely on the
1714 logic to ensure sublex_done() is called again only via the
1715 branch (in yylex()) that clears PL_lex_repl, else we'll loop */
1718 PL_lex_state = LEX_INTERPCONCAT;
1728 PL_endwhite = newSVpvs("");
1729 sv_catsv(PL_endwhite, PL_thiswhite);
1733 sv_setpvn(PL_thistoken,"",0);
1735 PL_realtokenstart = -1;
1739 PL_bufend = SvPVX(PL_linestr);
1740 PL_bufend += SvCUR(PL_linestr);
1741 PL_expect = XOPERATOR;
1742 PL_sublex_info.sub_inwhat = 0;
1750 Extracts a pattern, double-quoted string, or transliteration. This
1753 It looks at PL_lex_inwhat and PL_lex_inpat to find out whether it's
1754 processing a pattern (PL_lex_inpat is true), a transliteration
1755 (PL_lex_inwhat == OP_TRANS is true), or a double-quoted string.
1757 Returns a pointer to the character scanned up to. If this is
1758 advanced from the start pointer supplied (i.e. if anything was
1759 successfully parsed), will leave an OP for the substring scanned
1760 in yylval. Caller must intuit reason for not parsing further
1761 by looking at the next characters herself.
1765 double-quoted style: \r and \n
1766 regexp special ones: \D \s
1769 case and quoting: \U \Q \E
1770 stops on @ and $, but not for $ as tail anchor
1772 In transliterations:
1773 characters are VERY literal, except for - not at the start or end
1774 of the string, which indicates a range. If the range is in bytes,
1775 scan_const expands the range to the full set of intermediate
1776 characters. If the range is in utf8, the hyphen is replaced with
1777 a certain range mark which will be handled by pmtrans() in op.c.
1779 In double-quoted strings:
1781 double-quoted style: \r and \n
1783 deprecated backrefs: \1 (in substitution replacements)
1784 case and quoting: \U \Q \E
1787 scan_const does *not* construct ops to handle interpolated strings.
1788 It stops processing as soon as it finds an embedded $ or @ variable
1789 and leaves it to the caller to work out what's going on.
1791 embedded arrays (whether in pattern or not) could be:
1792 @foo, @::foo, @'foo, @{foo}, @$foo, @+, @-.
1794 $ in double-quoted strings must be the symbol of an embedded scalar.
1796 $ in pattern could be $foo or could be tail anchor. Assumption:
1797 it's a tail anchor if $ is the last thing in the string, or if it's
1798 followed by one of "()| \r\n\t"
1800 \1 (backreferences) are turned into $1
1802 The structure of the code is
1803 while (there's a character to process) {
1804 handle transliteration ranges
1805 skip regexp comments /(?#comment)/ and codes /(?{code})/
1806 skip #-initiated comments in //x patterns
1807 check for embedded arrays
1808 check for embedded scalars
1810 leave intact backslashes from leaveit (below)
1811 deprecate \1 in substitution replacements
1812 handle string-changing backslashes \l \U \Q \E, etc.
1813 switch (what was escaped) {
1814 handle \- in a transliteration (becomes a literal -)
1815 handle \132 (octal characters)
1816 handle \x15 and \x{1234} (hex characters)
1817 handle \N{name} (named characters)
1818 handle \cV (control characters)
1819 handle printf-style backslashes (\f, \r, \n, etc)
1821 } (end if backslash)
1822 } (end while character to read)
1827 S_scan_const(pTHX_ char *start)
1830 register char *send = PL_bufend; /* end of the constant */
1831 SV *sv = newSV(send - start); /* sv for the constant */
1832 register char *s = start; /* start of the constant */
1833 register char *d = SvPVX(sv); /* destination for copies */
1834 bool dorange = FALSE; /* are we in a translit range? */
1835 bool didrange = FALSE; /* did we just finish a range? */
1836 I32 has_utf8 = FALSE; /* Output constant is UTF8 */
1837 I32 this_utf8 = UTF; /* The source string is assumed to be UTF8 */
1840 UV literal_endpoint = 0;
1841 bool native_range = TRUE; /* turned to FALSE if the first endpoint is Unicode. */
1844 if (PL_lex_inwhat == OP_TRANS && PL_sublex_info.sub_op) {
1845 /* If we are doing a trans and we know we want UTF8 set expectation */
1846 has_utf8 = PL_sublex_info.sub_op->op_private & (OPpTRANS_FROM_UTF|OPpTRANS_TO_UTF);
1847 this_utf8 = PL_sublex_info.sub_op->op_private & (PL_lex_repl ? OPpTRANS_FROM_UTF : OPpTRANS_TO_UTF);
1851 while (s < send || dorange) {
1852 /* get transliterations out of the way (they're most literal) */
1853 if (PL_lex_inwhat == OP_TRANS) {
1854 /* expand a range A-Z to the full set of characters. AIE! */
1856 I32 i; /* current expanded character */
1857 I32 min; /* first character in range */
1858 I32 max; /* last character in range */
1869 char * const c = (char*)utf8_hop((U8*)d, -1);
1873 *c = (char)UTF_TO_NATIVE(0xff);
1874 /* mark the range as done, and continue */
1880 i = d - SvPVX_const(sv); /* remember current offset */
1883 SvLEN(sv) + (has_utf8 ?
1884 (512 - UTF_CONTINUATION_MARK +
1887 /* How many two-byte within 0..255: 128 in UTF-8,
1888 * 96 in UTF-8-mod. */
1890 SvGROW(sv, SvLEN(sv) + 256); /* never more than 256 chars in a range */
1892 d = SvPVX(sv) + i; /* refresh d after realloc */
1896 for (j = 0; j <= 1; j++) {
1897 char * const c = (char*)utf8_hop((U8*)d, -1);
1898 const UV uv = utf8n_to_uvchr((U8*)c, d - c, NULL, 0);
1904 max = (U8)0xff; /* only to \xff */
1905 uvmax = uv; /* \x{100} to uvmax */
1907 d = c; /* eat endpoint chars */
1912 d -= 2; /* eat the first char and the - */
1913 min = (U8)*d; /* first char in range */
1914 max = (U8)d[1]; /* last char in range */
1921 "Invalid range \"%c-%c\" in transliteration operator",
1922 (char)min, (char)max);
1926 if (literal_endpoint == 2 &&
1927 ((isLOWER(min) && isLOWER(max)) ||
1928 (isUPPER(min) && isUPPER(max)))) {
1930 for (i = min; i <= max; i++)
1932 *d++ = NATIVE_TO_NEED(has_utf8,i);
1934 for (i = min; i <= max; i++)
1936 *d++ = NATIVE_TO_NEED(has_utf8,i);
1941 for (i = min; i <= max; i++)
1944 const U8 ch = (U8)NATIVE_TO_UTF(i);
1945 if (UNI_IS_INVARIANT(ch))
1948 *d++ = (U8)UTF8_EIGHT_BIT_HI(ch);
1949 *d++ = (U8)UTF8_EIGHT_BIT_LO(ch);
1958 d = (char*)uvchr_to_utf8((U8*)d, 0x100);
1960 *d++ = (char)UTF_TO_NATIVE(0xff);
1962 d = (char*)uvchr_to_utf8((U8*)d, uvmax);
1966 /* mark the range as done, and continue */
1970 literal_endpoint = 0;
1975 /* range begins (ignore - as first or last char) */
1976 else if (*s == '-' && s+1 < send && s != start) {
1978 Perl_croak(aTHX_ "Ambiguous range in transliteration operator");
1985 *d++ = (char)UTF_TO_NATIVE(0xff); /* use illegal utf8 byte--see pmtrans */
1995 literal_endpoint = 0;
1996 native_range = TRUE;
2001 /* if we get here, we're not doing a transliteration */
2003 /* skip for regexp comments /(?#comment)/ and code /(?{code})/,
2004 except for the last char, which will be done separately. */
2005 else if (*s == '(' && PL_lex_inpat && s[1] == '?') {
2007 while (s+1 < send && *s != ')')
2008 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2010 else if (s[2] == '{' /* This should match regcomp.c */
2011 || ((s[2] == 'p' || s[2] == '?') && s[3] == '{'))
2014 char *regparse = s + (s[2] == '{' ? 3 : 4);
2017 while (count && (c = *regparse)) {
2018 if (c == '\\' && regparse[1])
2026 if (*regparse != ')')
2027 regparse--; /* Leave one char for continuation. */
2028 while (s < regparse)
2029 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2033 /* likewise skip #-initiated comments in //x patterns */
2034 else if (*s == '#' && PL_lex_inpat &&
2035 ((PMOP*)PL_lex_inpat)->op_pmflags & PMf_EXTENDED) {
2036 while (s+1 < send && *s != '\n')
2037 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2040 /* check for embedded arrays
2041 (@foo, @::foo, @'foo, @{foo}, @$foo, @+, @-)
2043 else if (*s == '@' && s[1]) {
2044 if (isALNUM_lazy_if(s+1,UTF))
2046 if (strchr(":'{$", s[1]))
2048 if (!PL_lex_inpat && (s[1] == '+' || s[1] == '-'))
2049 break; /* in regexp, neither @+ nor @- are interpolated */
2052 /* check for embedded scalars. only stop if we're sure it's a
2055 else if (*s == '$') {
2056 if (!PL_lex_inpat) /* not a regexp, so $ must be var */
2058 if (s + 1 < send && !strchr("()| \r\n\t", s[1]))
2059 break; /* in regexp, $ might be tail anchor */
2062 /* End of else if chain - OP_TRANS rejoin rest */
2065 if (*s == '\\' && s+1 < send) {
2068 /* deprecate \1 in strings and substitution replacements */
2069 if (PL_lex_inwhat == OP_SUBST && !PL_lex_inpat &&
2070 isDIGIT(*s) && *s != '0' && !isDIGIT(s[1]))
2072 if (ckWARN(WARN_SYNTAX))
2073 Perl_warner(aTHX_ packWARN(WARN_SYNTAX), "\\%c better written as $%c", *s, *s);
2078 /* string-change backslash escapes */
2079 if (PL_lex_inwhat != OP_TRANS && *s && strchr("lLuUEQ", *s)) {
2083 /* skip any other backslash escapes in a pattern */
2084 else if (PL_lex_inpat) {
2085 *d++ = NATIVE_TO_NEED(has_utf8,'\\');
2086 goto default_action;
2089 /* if we get here, it's either a quoted -, or a digit */
2092 /* quoted - in transliterations */
2094 if (PL_lex_inwhat == OP_TRANS) {
2101 if ((isALPHA(*s) || isDIGIT(*s)) &&
2103 Perl_warner(aTHX_ packWARN(WARN_MISC),
2104 "Unrecognized escape \\%c passed through",
2106 /* default action is to copy the quoted character */
2107 goto default_action;
2110 /* \132 indicates an octal constant */
2111 case '0': case '1': case '2': case '3':
2112 case '4': case '5': case '6': case '7':
2116 uv = grok_oct(s, &len, &flags, NULL);
2119 goto NUM_ESCAPE_INSERT;
2121 /* \x24 indicates a hex constant */
2125 char* const e = strchr(s, '}');
2126 I32 flags = PERL_SCAN_ALLOW_UNDERSCORES |
2127 PERL_SCAN_DISALLOW_PREFIX;
2132 yyerror("Missing right brace on \\x{}");
2136 uv = grok_hex(s, &len, &flags, NULL);
2142 I32 flags = PERL_SCAN_DISALLOW_PREFIX;
2143 uv = grok_hex(s, &len, &flags, NULL);
2149 /* Insert oct or hex escaped character.
2150 * There will always enough room in sv since such
2151 * escapes will be longer than any UTF-8 sequence
2152 * they can end up as. */
2154 /* We need to map to chars to ASCII before doing the tests
2157 if (!UNI_IS_INVARIANT(NATIVE_TO_UNI(uv))) {
2158 if (!has_utf8 && uv > 255) {
2159 /* Might need to recode whatever we have
2160 * accumulated so far if it contains any
2163 * (Can't we keep track of that and avoid
2164 * this rescan? --jhi)
2168 for (c = (U8 *) SvPVX(sv); c < (U8 *)d; c++) {
2169 if (!NATIVE_IS_INVARIANT(*c)) {
2174 const STRLEN offset = d - SvPVX_const(sv);
2176 d = SvGROW(sv, SvLEN(sv) + hicount + 1) + offset;
2180 while (src >= (const U8 *)SvPVX_const(sv)) {
2181 if (!NATIVE_IS_INVARIANT(*src)) {
2182 const U8 ch = NATIVE_TO_ASCII(*src);
2183 *dst-- = (U8)UTF8_EIGHT_BIT_LO(ch);
2184 *dst-- = (U8)UTF8_EIGHT_BIT_HI(ch);
2194 if (has_utf8 || uv > 255) {
2195 d = (char*)uvchr_to_utf8((U8*)d, uv);
2197 if (PL_lex_inwhat == OP_TRANS &&
2198 PL_sublex_info.sub_op) {
2199 PL_sublex_info.sub_op->op_private |=
2200 (PL_lex_repl ? OPpTRANS_FROM_UTF
2204 if (uv > 255 && !dorange)
2205 native_range = FALSE;
2217 /* \N{LATIN SMALL LETTER A} is a named character */
2221 char* e = strchr(s, '}');
2228 yyerror("Missing right brace on \\N{}");
2232 if (e > s + 2 && s[1] == 'U' && s[2] == '+') {
2234 I32 flags = PERL_SCAN_ALLOW_UNDERSCORES |
2235 PERL_SCAN_DISALLOW_PREFIX;
2238 uv = grok_hex(s, &len, &flags, NULL);
2239 if ( e > s && len != (STRLEN)(e - s) ) {
2243 goto NUM_ESCAPE_INSERT;
2245 res = newSVpvn(s + 1, e - s - 1);
2246 type = newSVpvn(s - 2,e - s + 3);
2247 res = new_constant( NULL, 0, "charnames",
2248 res, NULL, SvPVX(type) );
2251 sv_utf8_upgrade(res);
2252 str = SvPV_const(res,len);
2253 #ifdef EBCDIC_NEVER_MIND
2254 /* charnames uses pack U and that has been
2255 * recently changed to do the below uni->native
2256 * mapping, so this would be redundant (and wrong,
2257 * the code point would be doubly converted).
2258 * But leave this in just in case the pack U change
2259 * gets revoked, but the semantics is still
2260 * desireable for charnames. --jhi */
2262 UV uv = utf8_to_uvchr((const U8*)str, 0);
2265 U8 tmpbuf[UTF8_MAXBYTES+1], *d;
2267 d = uvchr_to_utf8(tmpbuf, UNI_TO_NATIVE(uv));
2268 sv_setpvn(res, (char *)tmpbuf, d - tmpbuf);
2269 str = SvPV_const(res, len);
2273 if (!has_utf8 && SvUTF8(res)) {
2274 const char * const ostart = SvPVX_const(sv);
2275 SvCUR_set(sv, d - ostart);
2278 sv_utf8_upgrade(sv);
2279 /* this just broke our allocation above... */
2280 SvGROW(sv, (STRLEN)(send - start));
2281 d = SvPVX(sv) + SvCUR(sv);
2284 if (len > (STRLEN)(e - s + 4)) { /* I _guess_ 4 is \N{} --jhi */
2285 const char * const odest = SvPVX_const(sv);
2287 SvGROW(sv, (SvLEN(sv) + len - (e - s + 4)));
2288 d = SvPVX(sv) + (d - odest);
2292 native_range = FALSE; /* \N{} is guessed to be Unicode */
2294 Copy(str, d, len, char);
2301 yyerror("Missing braces on \\N{}");
2304 /* \c is a control character */
2313 *d++ = NATIVE_TO_NEED(has_utf8,toCTRL(c));
2316 yyerror("Missing control char name in \\c");
2320 /* printf-style backslashes, formfeeds, newlines, etc */
2322 *d++ = NATIVE_TO_NEED(has_utf8,'\b');
2325 *d++ = NATIVE_TO_NEED(has_utf8,'\n');
2328 *d++ = NATIVE_TO_NEED(has_utf8,'\r');
2331 *d++ = NATIVE_TO_NEED(has_utf8,'\f');
2334 *d++ = NATIVE_TO_NEED(has_utf8,'\t');
2337 *d++ = ASCII_TO_NEED(has_utf8,'\033');
2340 *d++ = ASCII_TO_NEED(has_utf8,'\007');
2346 } /* end if (backslash) */
2353 /* If we started with encoded form, or already know we want it
2354 and then encode the next character */
2355 if ((has_utf8 || this_utf8) && !NATIVE_IS_INVARIANT((U8)(*s))) {
2357 const UV nextuv = (this_utf8) ? utf8n_to_uvchr((U8*)s, send - s, &len, 0) : (UV) ((U8) *s);
2358 const STRLEN need = UNISKIP(NATIVE_TO_UNI(nextuv));
2361 /* encoded value larger than old, need extra space (NOTE: SvCUR() not set here) */
2362 const STRLEN off = d - SvPVX_const(sv);
2363 d = SvGROW(sv, SvLEN(sv) + (need-len)) + off;
2365 d = (char*)uvchr_to_utf8((U8*)d, nextuv);
2368 if (uv > 255 && !dorange)
2369 native_range = FALSE;
2373 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2375 } /* while loop to process each character */
2377 /* terminate the string and set up the sv */
2379 SvCUR_set(sv, d - SvPVX_const(sv));
2380 if (SvCUR(sv) >= SvLEN(sv))
2381 Perl_croak(aTHX_ "panic: constant overflowed allocated space");
2384 if (PL_encoding && !has_utf8) {
2385 sv_recode_to_utf8(sv, PL_encoding);
2391 if (PL_lex_inwhat == OP_TRANS && PL_sublex_info.sub_op) {
2392 PL_sublex_info.sub_op->op_private |=
2393 (PL_lex_repl ? OPpTRANS_FROM_UTF : OPpTRANS_TO_UTF);
2397 /* shrink the sv if we allocated more than we used */
2398 if (SvCUR(sv) + 5 < SvLEN(sv)) {
2399 SvPV_shrink_to_cur(sv);
2402 /* return the substring (via yylval) only if we parsed anything */
2403 if (s > PL_bufptr) {
2404 if ( PL_hints & ( PL_lex_inpat ? HINT_NEW_RE : HINT_NEW_STRING ) )
2405 sv = new_constant(start, s - start,
2406 (const char *)(PL_lex_inpat ? "qr" : "q"),
2409 (( PL_lex_inwhat == OP_TRANS
2411 : ( (PL_lex_inwhat == OP_SUBST && !PL_lex_inpat)
2414 yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
2421 * Returns TRUE if there's more to the expression (e.g., a subscript),
2424 * It deals with "$foo[3]" and /$foo[3]/ and /$foo[0123456789$]+/
2426 * ->[ and ->{ return TRUE
2427 * { and [ outside a pattern are always subscripts, so return TRUE
2428 * if we're outside a pattern and it's not { or [, then return FALSE
2429 * if we're in a pattern and the first char is a {
2430 * {4,5} (any digits around the comma) returns FALSE
2431 * if we're in a pattern and the first char is a [
2433 * [SOMETHING] has a funky algorithm to decide whether it's a
2434 * character class or not. It has to deal with things like
2435 * /$foo[-3]/ and /$foo[$bar]/ as well as /$foo[$\d]+/
2436 * anything else returns TRUE
2439 /* This is the one truly awful dwimmer necessary to conflate C and sed. */
2442 S_intuit_more(pTHX_ register char *s)
2445 if (PL_lex_brackets)
2447 if (*s == '-' && s[1] == '>' && (s[2] == '[' || s[2] == '{'))
2449 if (*s != '{' && *s != '[')
2454 /* In a pattern, so maybe we have {n,m}. */
2471 /* On the other hand, maybe we have a character class */
2474 if (*s == ']' || *s == '^')
2477 /* this is terrifying, and it works */
2478 int weight = 2; /* let's weigh the evidence */
2480 unsigned char un_char = 255, last_un_char;
2481 const char * const send = strchr(s,']');
2482 char tmpbuf[sizeof PL_tokenbuf * 4];
2484 if (!send) /* has to be an expression */
2487 Zero(seen,256,char);
2490 else if (isDIGIT(*s)) {
2492 if (isDIGIT(s[1]) && s[2] == ']')
2498 for (; s < send; s++) {
2499 last_un_char = un_char;
2500 un_char = (unsigned char)*s;
2505 weight -= seen[un_char] * 10;
2506 if (isALNUM_lazy_if(s+1,UTF)) {
2508 scan_ident(s, send, tmpbuf, sizeof tmpbuf, FALSE);
2509 len = (int)strlen(tmpbuf);
2510 if (len > 1 && gv_fetchpvn_flags(tmpbuf, len, 0, SVt_PV))
2515 else if (*s == '$' && s[1] &&
2516 strchr("[#!%*<>()-=",s[1])) {
2517 if (/*{*/ strchr("])} =",s[2]))
2526 if (strchr("wds]",s[1]))
2528 else if (seen[(U8)'\''] || seen[(U8)'"'])
2530 else if (strchr("rnftbxcav",s[1]))
2532 else if (isDIGIT(s[1])) {
2534 while (s[1] && isDIGIT(s[1]))
2544 if (strchr("aA01! ",last_un_char))
2546 if (strchr("zZ79~",s[1]))
2548 if (last_un_char == 255 && (isDIGIT(s[1]) || s[1] == '$'))
2549 weight -= 5; /* cope with negative subscript */
2552 if (!isALNUM(last_un_char)
2553 && !(last_un_char == '$' || last_un_char == '@'
2554 || last_un_char == '&')
2555 && isALPHA(*s) && s[1] && isALPHA(s[1])) {
2560 if (keyword(tmpbuf, d - tmpbuf, 0))
2563 if (un_char == last_un_char + 1)
2565 weight -= seen[un_char];
2570 if (weight >= 0) /* probably a character class */
2580 * Does all the checking to disambiguate
2582 * between foo(bar) and bar->foo. Returns 0 if not a method, otherwise
2583 * FUNCMETH (bar->foo(args)) or METHOD (bar->foo args).
2585 * First argument is the stuff after the first token, e.g. "bar".
2587 * Not a method if bar is a filehandle.
2588 * Not a method if foo is a subroutine prototyped to take a filehandle.
2589 * Not a method if it's really "Foo $bar"
2590 * Method if it's "foo $bar"
2591 * Not a method if it's really "print foo $bar"
2592 * Method if it's really "foo package::" (interpreted as package->foo)
2593 * Not a method if bar is known to be a subroutine ("sub bar; foo bar")
2594 * Not a method if bar is a filehandle or package, but is quoted with
2599 S_intuit_method(pTHX_ char *start, GV *gv, CV *cv)
2602 char *s = start + (*start == '$');
2603 char tmpbuf[sizeof PL_tokenbuf];
2611 if (SvTYPE(gv) == SVt_PVGV && GvIO(gv))
2615 const char *proto = SvPVX_const(cv);
2626 s = scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
2627 /* start is the beginning of the possible filehandle/object,
2628 * and s is the end of it
2629 * tmpbuf is a copy of it
2632 if (*start == '$') {
2633 if (gv || PL_last_lop_op == OP_PRINT || isUPPER(*PL_tokenbuf))
2636 len = start - SvPVX(PL_linestr);
2640 start = SvPVX(PL_linestr) + len;
2644 return *s == '(' ? FUNCMETH : METHOD;
2646 if (!keyword(tmpbuf, len, 0)) {
2647 if (len > 2 && tmpbuf[len - 2] == ':' && tmpbuf[len - 1] == ':') {
2651 soff = s - SvPVX(PL_linestr);
2655 indirgv = gv_fetchpvn_flags(tmpbuf, len, 0, SVt_PVCV);
2656 if (indirgv && GvCVu(indirgv))
2658 /* filehandle or package name makes it a method */
2659 if (!gv || GvIO(indirgv) || gv_stashpvn(tmpbuf, len, 0)) {
2661 soff = s - SvPVX(PL_linestr);
2664 if ((PL_bufend - s) >= 2 && *s == '=' && *(s+1) == '>')
2665 return 0; /* no assumptions -- "=>" quotes bearword */
2667 start_force(PL_curforce);
2668 NEXTVAL_NEXTTOKE.opval = (OP*)newSVOP(OP_CONST, 0,
2669 newSVpvn(tmpbuf,len));
2670 NEXTVAL_NEXTTOKE.opval->op_private = OPpCONST_BARE;
2672 curmad('X', newSVpvn(start,SvPVX(PL_linestr) + soff - start));
2677 PL_bufptr = SvPVX(PL_linestr) + soff; /* restart before space */
2679 return *s == '(' ? FUNCMETH : METHOD;
2687 * Return a string of Perl code to load the debugger. If PERL5DB
2688 * is set, it will return the contents of that, otherwise a
2689 * compile-time require of perl5db.pl.
2697 const char * const pdb = PerlEnv_getenv("PERL5DB");
2701 SETERRNO(0,SS_NORMAL);
2702 return "BEGIN { require 'perl5db.pl' }";
2708 /* Encoded script support. filter_add() effectively inserts a
2709 * 'pre-processing' function into the current source input stream.
2710 * Note that the filter function only applies to the current source file
2711 * (e.g., it will not affect files 'require'd or 'use'd by this one).
2713 * The datasv parameter (which may be NULL) can be used to pass
2714 * private data to this instance of the filter. The filter function
2715 * can recover the SV using the FILTER_DATA macro and use it to
2716 * store private buffers and state information.
2718 * The supplied datasv parameter is upgraded to a PVIO type
2719 * and the IoDIRP/IoANY field is used to store the function pointer,
2720 * and IOf_FAKE_DIRP is enabled on datasv to mark this as such.
2721 * Note that IoTOP_NAME, IoFMT_NAME, IoBOTTOM_NAME, if set for
2722 * private use must be set using malloc'd pointers.
2726 Perl_filter_add(pTHX_ filter_t funcp, SV *datasv)
2732 if (!PL_rsfp_filters)
2733 PL_rsfp_filters = newAV();
2736 SvUPGRADE(datasv, SVt_PVIO);
2737 IoANY(datasv) = FPTR2DPTR(void *, funcp); /* stash funcp into spare field */
2738 IoFLAGS(datasv) |= IOf_FAKE_DIRP;
2739 DEBUG_P(PerlIO_printf(Perl_debug_log, "filter_add func %p (%s)\n",
2740 FPTR2DPTR(void *, IoANY(datasv)),
2741 SvPV_nolen(datasv)));
2742 av_unshift(PL_rsfp_filters, 1);
2743 av_store(PL_rsfp_filters, 0, datasv) ;
2748 /* Delete most recently added instance of this filter function. */
2750 Perl_filter_del(pTHX_ filter_t funcp)
2756 DEBUG_P(PerlIO_printf(Perl_debug_log, "filter_del func %p",
2757 FPTR2DPTR(void*, funcp)));
2759 if (!PL_rsfp_filters || AvFILLp(PL_rsfp_filters)<0)
2761 /* if filter is on top of stack (usual case) just pop it off */
2762 datasv = FILTER_DATA(AvFILLp(PL_rsfp_filters));
2763 if (IoANY(datasv) == FPTR2DPTR(void *, funcp)) {
2764 IoFLAGS(datasv) &= ~IOf_FAKE_DIRP;
2765 IoANY(datasv) = (void *)NULL;
2766 sv_free(av_pop(PL_rsfp_filters));
2770 /* we need to search for the correct entry and clear it */
2771 Perl_die(aTHX_ "filter_del can only delete in reverse order (currently)");
2775 /* Invoke the idxth filter function for the current rsfp. */
2776 /* maxlen 0 = read one text line */
2778 Perl_filter_read(pTHX_ int idx, SV *buf_sv, int maxlen)
2783 /* This API is bad. It should have been using unsigned int for maxlen.
2784 Not sure if we want to change the API, but if not we should sanity
2785 check the value here. */
2786 const unsigned int correct_length
2795 if (!PL_rsfp_filters)
2797 if (idx > AvFILLp(PL_rsfp_filters)) { /* Any more filters? */
2798 /* Provide a default input filter to make life easy. */
2799 /* Note that we append to the line. This is handy. */
2800 DEBUG_P(PerlIO_printf(Perl_debug_log,
2801 "filter_read %d: from rsfp\n", idx));
2802 if (correct_length) {
2805 const int old_len = SvCUR(buf_sv);
2807 /* ensure buf_sv is large enough */
2808 SvGROW(buf_sv, (STRLEN)(old_len + correct_length)) ;
2809 if ((len = PerlIO_read(PL_rsfp, SvPVX(buf_sv) + old_len,
2810 correct_length)) <= 0) {
2811 if (PerlIO_error(PL_rsfp))
2812 return -1; /* error */
2814 return 0 ; /* end of file */
2816 SvCUR_set(buf_sv, old_len + len) ;
2819 if (sv_gets(buf_sv, PL_rsfp, SvCUR(buf_sv)) == NULL) {
2820 if (PerlIO_error(PL_rsfp))
2821 return -1; /* error */
2823 return 0 ; /* end of file */
2826 return SvCUR(buf_sv);
2828 /* Skip this filter slot if filter has been deleted */
2829 if ( (datasv = FILTER_DATA(idx)) == &PL_sv_undef) {
2830 DEBUG_P(PerlIO_printf(Perl_debug_log,
2831 "filter_read %d: skipped (filter deleted)\n",
2833 return FILTER_READ(idx+1, buf_sv, correct_length); /* recurse */
2835 /* Get function pointer hidden within datasv */
2836 funcp = DPTR2FPTR(filter_t, IoANY(datasv));
2837 DEBUG_P(PerlIO_printf(Perl_debug_log,
2838 "filter_read %d: via function %p (%s)\n",
2839 idx, (void*)datasv, SvPV_nolen_const(datasv)));
2840 /* Call function. The function is expected to */
2841 /* call "FILTER_READ(idx+1, buf_sv)" first. */
2842 /* Return: <0:error, =0:eof, >0:not eof */
2843 return (*funcp)(aTHX_ idx, buf_sv, correct_length);
2847 S_filter_gets(pTHX_ register SV *sv, register PerlIO *fp, STRLEN append)
2850 #ifdef PERL_CR_FILTER
2851 if (!PL_rsfp_filters) {
2852 filter_add(S_cr_textfilter,NULL);
2855 if (PL_rsfp_filters) {
2857 SvCUR_set(sv, 0); /* start with empty line */
2858 if (FILTER_READ(0, sv, 0) > 0)
2859 return ( SvPVX(sv) ) ;
2864 return (sv_gets(sv, fp, append));
2868 S_find_in_my_stash(pTHX_ const char *pkgname, I32 len)
2873 if (len == 11 && *pkgname == '_' && strEQ(pkgname, "__PACKAGE__"))
2877 (pkgname[len - 2] == ':' && pkgname[len - 1] == ':') &&
2878 (gv = gv_fetchpvn_flags(pkgname, len, 0, SVt_PVHV)))
2880 return GvHV(gv); /* Foo:: */
2883 /* use constant CLASS => 'MyClass' */
2884 gv = gv_fetchpvn_flags(pkgname, len, 0, SVt_PVCV);
2885 if (gv && GvCV(gv)) {
2886 SV * const sv = cv_const_sv(GvCV(gv));
2888 pkgname = SvPV_nolen_const(sv);
2891 return gv_stashpv(pkgname, 0);
2895 * S_readpipe_override
2896 * Check whether readpipe() is overriden, and generates the appropriate
2897 * optree, provided sublex_start() is called afterwards.
2900 S_readpipe_override(pTHX)
2903 GV *gv_readpipe = gv_fetchpvs("readpipe", GV_NOTQUAL, SVt_PVCV);
2904 yylval.ival = OP_BACKTICK;
2906 && GvCVu(gv_readpipe) && GvIMPORTED_CV(gv_readpipe))
2908 ((gvp = (GV**)hv_fetchs(PL_globalstash, "readpipe", FALSE))
2909 && (gv_readpipe = *gvp) != (GV*)&PL_sv_undef
2910 && GvCVu(gv_readpipe) && GvIMPORTED_CV(gv_readpipe)))
2912 PL_lex_op = (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
2913 append_elem(OP_LIST,
2914 newSVOP(OP_CONST, 0, &PL_sv_undef), /* value will be read later */
2915 newCVREF(0, newGVOP(OP_GV, 0, gv_readpipe))));
2925 * The intent of this yylex wrapper is to minimize the changes to the
2926 * tokener when we aren't interested in collecting madprops. It remains
2927 * to be seen how successful this strategy will be...
2934 char *s = PL_bufptr;
2936 /* make sure PL_thiswhite is initialized */
2940 /* just do what yylex would do on pending identifier; leave PL_thiswhite alone */
2941 if (PL_pending_ident)
2942 return S_pending_ident(aTHX);
2944 /* previous token ate up our whitespace? */
2945 if (!PL_lasttoke && PL_nextwhite) {
2946 PL_thiswhite = PL_nextwhite;
2950 /* isolate the token, and figure out where it is without whitespace */
2951 PL_realtokenstart = -1;
2955 assert(PL_curforce < 0);
2957 if (!PL_thismad || PL_thismad->mad_key == '^') { /* not forced already? */
2958 if (!PL_thistoken) {
2959 if (PL_realtokenstart < 0 || !CopLINE(PL_curcop))
2960 PL_thistoken = newSVpvs("");
2962 char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
2963 PL_thistoken = newSVpvn(tstart, s - tstart);
2966 if (PL_thismad) /* install head */
2967 CURMAD('X', PL_thistoken);
2970 /* last whitespace of a sublex? */
2971 if (optype == ')' && PL_endwhite) {
2972 CURMAD('X', PL_endwhite);
2977 /* if no whitespace and we're at EOF, bail. Otherwise fake EOF below. */
2978 if (!PL_thiswhite && !PL_endwhite && !optype) {
2979 sv_free(PL_thistoken);
2984 /* put off final whitespace till peg */
2985 if (optype == ';' && !PL_rsfp) {
2986 PL_nextwhite = PL_thiswhite;
2989 else if (PL_thisopen) {
2990 CURMAD('q', PL_thisopen);
2992 sv_free(PL_thistoken);
2996 /* Store actual token text as madprop X */
2997 CURMAD('X', PL_thistoken);
3001 /* add preceding whitespace as madprop _ */
3002 CURMAD('_', PL_thiswhite);
3006 /* add quoted material as madprop = */
3007 CURMAD('=', PL_thisstuff);
3011 /* add terminating quote as madprop Q */
3012 CURMAD('Q', PL_thisclose);
3016 /* special processing based on optype */
3020 /* opval doesn't need a TOKEN since it can already store mp */
3031 append_madprops(PL_thismad, yylval.opval, 0);
3039 addmad(newMADsv('p', PL_endwhite), &PL_thismad, 0);
3048 /* remember any fake bracket that lexer is about to discard */
3049 if (PL_lex_brackets == 1 &&
3050 ((expectation)PL_lex_brackstack[0] & XFAKEBRACK))
3053 while (s < PL_bufend && (*s == ' ' || *s == '\t'))
3056 PL_thiswhite = newSVpvn(PL_bufptr, ++s - PL_bufptr);
3057 addmad(newMADsv('#', PL_thiswhite), &PL_thismad, 0);
3060 break; /* don't bother looking for trailing comment */
3069 /* attach a trailing comment to its statement instead of next token */
3073 if (PL_bufptr > PL_oldbufptr && PL_bufptr[-1] == optype) {
3075 while (s < PL_bufend && (*s == ' ' || *s == '\t'))
3077 if (*s == '\n' || *s == '#') {
3078 while (s < PL_bufend && *s != '\n')
3082 PL_thiswhite = newSVpvn(PL_bufptr, s - PL_bufptr);
3083 addmad(newMADsv('#', PL_thiswhite), &PL_thismad, 0);
3100 /* Create new token struct. Note: opvals return early above. */
3101 yylval.tkval = newTOKEN(optype, yylval, PL_thismad);
3108 S_tokenize_use(pTHX_ int is_use, char *s) {
3110 if (PL_expect != XSTATE)
3111 yyerror(Perl_form(aTHX_ "\"%s\" not allowed in expression",
3112 is_use ? "use" : "no"));
3114 if (isDIGIT(*s) || (*s == 'v' && isDIGIT(s[1]))) {
3115 s = force_version(s, TRUE);
3116 if (*s == ';' || (s = SKIPSPACE1(s), *s == ';')) {
3117 start_force(PL_curforce);
3118 NEXTVAL_NEXTTOKE.opval = NULL;
3121 else if (*s == 'v') {
3122 s = force_word(s,WORD,FALSE,TRUE,FALSE);
3123 s = force_version(s, FALSE);
3127 s = force_word(s,WORD,FALSE,TRUE,FALSE);
3128 s = force_version(s, FALSE);
3130 yylval.ival = is_use;
3134 static const char* const exp_name[] =
3135 { "OPERATOR", "TERM", "REF", "STATE", "BLOCK", "ATTRBLOCK",
3136 "ATTRTERM", "TERMBLOCK", "TERMORDORDOR"
3143 Works out what to call the token just pulled out of the input
3144 stream. The yacc parser takes care of taking the ops we return and
3145 stitching them into a tree.
3151 if read an identifier
3152 if we're in a my declaration
3153 croak if they tried to say my($foo::bar)
3154 build the ops for a my() declaration
3155 if it's an access to a my() variable
3156 are we in a sort block?
3157 croak if my($a); $a <=> $b
3158 build ops for access to a my() variable
3159 if in a dq string, and they've said @foo and we can't find @foo
3161 build ops for a bareword
3162 if we already built the token before, use it.
3167 #pragma segment Perl_yylex
3173 register char *s = PL_bufptr;
3178 /* orig_keyword, gvp, and gv are initialized here because
3179 * jump to the label just_a_word_zero can bypass their
3180 * initialization later. */
3181 I32 orig_keyword = 0;
3186 SV* tmp = newSVpvs("");
3187 PerlIO_printf(Perl_debug_log, "### %"IVdf":LEX_%s/X%s %s\n",
3188 (IV)CopLINE(PL_curcop),
3189 lex_state_names[PL_lex_state],
3190 exp_name[PL_expect],
3191 pv_display(tmp, s, strlen(s), 0, 60));
3194 /* check if there's an identifier for us to look at */
3195 if (PL_pending_ident)
3196 return REPORT(S_pending_ident(aTHX));
3198 /* no identifier pending identification */
3200 switch (PL_lex_state) {
3202 case LEX_NORMAL: /* Some compilers will produce faster */
3203 case LEX_INTERPNORMAL: /* code if we comment these out. */
3207 /* when we've already built the next token, just pull it out of the queue */
3211 yylval = PL_nexttoke[PL_lasttoke].next_val;
3213 PL_thismad = PL_nexttoke[PL_lasttoke].next_mad;
3214 PL_nexttoke[PL_lasttoke].next_mad = 0;
3215 if (PL_thismad && PL_thismad->mad_key == '_') {
3216 PL_thiswhite = (SV*)PL_thismad->mad_val;
3217 PL_thismad->mad_val = 0;
3218 mad_free(PL_thismad);
3223 PL_lex_state = PL_lex_defer;
3224 PL_expect = PL_lex_expect;
3225 PL_lex_defer = LEX_NORMAL;
3226 if (!PL_nexttoke[PL_lasttoke].next_type)
3231 yylval = PL_nextval[PL_nexttoke];
3233 PL_lex_state = PL_lex_defer;
3234 PL_expect = PL_lex_expect;
3235 PL_lex_defer = LEX_NORMAL;
3239 /* FIXME - can these be merged? */
3240 return(PL_nexttoke[PL_lasttoke].next_type);
3242 return REPORT(PL_nexttype[PL_nexttoke]);
3245 /* interpolated case modifiers like \L \U, including \Q and \E.
3246 when we get here, PL_bufptr is at the \
3248 case LEX_INTERPCASEMOD:
3250 if (PL_bufptr != PL_bufend && *PL_bufptr != '\\')
3251 Perl_croak(aTHX_ "panic: INTERPCASEMOD");
3253 /* handle \E or end of string */
3254 if (PL_bufptr == PL_bufend || PL_bufptr[1] == 'E') {
3256 if (PL_lex_casemods) {
3257 const char oldmod = PL_lex_casestack[--PL_lex_casemods];
3258 PL_lex_casestack[PL_lex_casemods] = '\0';
3260 if (PL_bufptr != PL_bufend
3261 && (oldmod == 'L' || oldmod == 'U' || oldmod == 'Q')) {
3263 PL_lex_state = LEX_INTERPCONCAT;
3266 PL_thistoken = newSVpvs("\\E");
3272 while (PL_bufptr != PL_bufend &&
3273 PL_bufptr[0] == '\\' && PL_bufptr[1] == 'E') {
3275 PL_thiswhite = newSVpvs("");
3276 sv_catpvn(PL_thiswhite, PL_bufptr, 2);
3280 if (PL_bufptr != PL_bufend)
3283 PL_lex_state = LEX_INTERPCONCAT;
3287 DEBUG_T({ PerlIO_printf(Perl_debug_log,
3288 "### Saw case modifier\n"); });
3290 if (s[1] == '\\' && s[2] == 'E') {
3293 PL_thiswhite = newSVpvs("");
3294 sv_catpvn(PL_thiswhite, PL_bufptr, 4);
3297 PL_lex_state = LEX_INTERPCONCAT;
3302 if (!PL_madskills) /* when just compiling don't need correct */
3303 if (strnEQ(s, "L\\u", 3) || strnEQ(s, "U\\l", 3))
3304 tmp = *s, *s = s[2], s[2] = (char)tmp; /* misordered... */
3305 if ((*s == 'L' || *s == 'U') &&
3306 (strchr(PL_lex_casestack, 'L') || strchr(PL_lex_casestack, 'U'))) {
3307 PL_lex_casestack[--PL_lex_casemods] = '\0';
3310 if (PL_lex_casemods > 10)
3311 Renew(PL_lex_casestack, PL_lex_casemods + 2, char);
3312 PL_lex_casestack[PL_lex_casemods++] = *s;
3313 PL_lex_casestack[PL_lex_casemods] = '\0';
3314 PL_lex_state = LEX_INTERPCONCAT;
3315 start_force(PL_curforce);
3316 NEXTVAL_NEXTTOKE.ival = 0;
3318 start_force(PL_curforce);
3320 NEXTVAL_NEXTTOKE.ival = OP_LCFIRST;
3322 NEXTVAL_NEXTTOKE.ival = OP_UCFIRST;
3324 NEXTVAL_NEXTTOKE.ival = OP_LC;
3326 NEXTVAL_NEXTTOKE.ival = OP_UC;
3328 NEXTVAL_NEXTTOKE.ival = OP_QUOTEMETA;
3330 Perl_croak(aTHX_ "panic: yylex");
3332 SV* const tmpsv = newSVpvs("");
3333 Perl_sv_catpvf(aTHX_ tmpsv, "\\%c", *s);
3339 if (PL_lex_starts) {
3345 sv_free(PL_thistoken);
3346 PL_thistoken = newSVpvs("");
3349 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
3350 if (PL_lex_casemods == 1 && PL_lex_inpat)
3359 case LEX_INTERPPUSH:
3360 return REPORT(sublex_push());
3362 case LEX_INTERPSTART:
3363 if (PL_bufptr == PL_bufend)
3364 return REPORT(sublex_done());
3365 DEBUG_T({ PerlIO_printf(Perl_debug_log,
3366 "### Interpolated variable\n"); });
3368 PL_lex_dojoin = (*PL_bufptr == '@');
3369 PL_lex_state = LEX_INTERPNORMAL;
3370 if (PL_lex_dojoin) {
3371 start_force(PL_curforce);
3372 NEXTVAL_NEXTTOKE.ival = 0;
3374 start_force(PL_curforce);
3375 force_ident("\"", '$');
3376 start_force(PL_curforce);
3377 NEXTVAL_NEXTTOKE.ival = 0;
3379 start_force(PL_curforce);
3380 NEXTVAL_NEXTTOKE.ival = 0;
3382 start_force(PL_curforce);
3383 NEXTVAL_NEXTTOKE.ival = OP_JOIN; /* emulate join($", ...) */
3386 if (PL_lex_starts++) {
3391 sv_free(PL_thistoken);
3392 PL_thistoken = newSVpvs("");
3395 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
3396 if (!PL_lex_casemods && PL_lex_inpat)
3403 case LEX_INTERPENDMAYBE:
3404 if (intuit_more(PL_bufptr)) {
3405 PL_lex_state = LEX_INTERPNORMAL; /* false alarm, more expr */
3411 if (PL_lex_dojoin) {
3412 PL_lex_dojoin = FALSE;
3413 PL_lex_state = LEX_INTERPCONCAT;
3417 sv_free(PL_thistoken);
3418 PL_thistoken = newSVpvs("");
3423 if (PL_lex_inwhat == OP_SUBST && PL_linestr == PL_lex_repl
3424 && SvEVALED(PL_lex_repl))
3426 if (PL_bufptr != PL_bufend)
3427 Perl_croak(aTHX_ "Bad evalled substitution pattern");
3431 case LEX_INTERPCONCAT:
3433 if (PL_lex_brackets)
3434 Perl_croak(aTHX_ "panic: INTERPCONCAT");
3436 if (PL_bufptr == PL_bufend)
3437 return REPORT(sublex_done());
3439 if (SvIVX(PL_linestr) == '\'') {
3440 SV *sv = newSVsv(PL_linestr);
3443 else if ( PL_hints & HINT_NEW_RE )
3444 sv = new_constant(NULL, 0, "qr", sv, sv, "q");
3445 yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
3449 s = scan_const(PL_bufptr);
3451 PL_lex_state = LEX_INTERPCASEMOD;
3453 PL_lex_state = LEX_INTERPSTART;
3456 if (s != PL_bufptr) {
3457 start_force(PL_curforce);
3459 curmad('X', newSVpvn(PL_bufptr,s-PL_bufptr));
3461 NEXTVAL_NEXTTOKE = yylval;
3464 if (PL_lex_starts++) {
3468 sv_free(PL_thistoken);
3469 PL_thistoken = newSVpvs("");
3472 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
3473 if (!PL_lex_casemods && PL_lex_inpat)
3486 PL_lex_state = LEX_NORMAL;
3487 s = scan_formline(PL_bufptr);
3488 if (!PL_lex_formbrack)
3494 PL_oldoldbufptr = PL_oldbufptr;
3500 sv_free(PL_thistoken);
3503 PL_realtokenstart = s - SvPVX(PL_linestr); /* assume but undo on ws */
3507 if (isIDFIRST_lazy_if(s,UTF))
3509 Perl_croak(aTHX_ "Unrecognized character \\x%02X", *s & 255);
3512 goto fake_eof; /* emulate EOF on ^D or ^Z */
3521 if (PL_lex_brackets) {
3522 yyerror((const char *)
3524 ? "Format not terminated"
3525 : "Missing right curly or square bracket"));
3527 DEBUG_T( { PerlIO_printf(Perl_debug_log,
3528 "### Tokener got EOF\n");
3532 if (s++ < PL_bufend)
3533 goto retry; /* ignore stray nulls */
3536 if (!PL_in_eval && !PL_preambled) {
3537 PL_preambled = TRUE;
3542 sv_setpv(PL_linestr,incl_perldb());
3543 if (SvCUR(PL_linestr))
3544 sv_catpvs(PL_linestr,";");
3546 while(AvFILLp(PL_preambleav) >= 0) {
3547 SV *tmpsv = av_shift(PL_preambleav);
3548 sv_catsv(PL_linestr, tmpsv);
3549 sv_catpvs(PL_linestr, ";");
3552 sv_free((SV*)PL_preambleav);
3553 PL_preambleav = NULL;
3555 if (PL_minus_n || PL_minus_p) {
3556 sv_catpvs(PL_linestr, "LINE: while (<>) {");
3558 sv_catpvs(PL_linestr,"chomp;");
3561 if ((*PL_splitstr == '/' || *PL_splitstr == '\''
3562 || *PL_splitstr == '"')
3563 && strchr(PL_splitstr + 1, *PL_splitstr))
3564 Perl_sv_catpvf(aTHX_ PL_linestr, "our @F=split(%s);", PL_splitstr);
3566 /* "q\0${splitstr}\0" is legal perl. Yes, even NUL
3567 bytes can be used as quoting characters. :-) */
3568 const char *splits = PL_splitstr;
3569 sv_catpvs(PL_linestr, "our @F=split(q\0");
3572 if (*splits == '\\')
3573 sv_catpvn(PL_linestr, splits, 1);
3574 sv_catpvn(PL_linestr, splits, 1);
3575 } while (*splits++);
3576 /* This loop will embed the trailing NUL of
3577 PL_linestr as the last thing it does before
3579 sv_catpvs(PL_linestr, ");");
3583 sv_catpvs(PL_linestr,"our @F=split(' ');");
3587 sv_catpvs(PL_linestr,"use feature ':5.10';");
3588 sv_catpvs(PL_linestr, "\n");
3589 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3590 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3591 PL_last_lop = PL_last_uni = NULL;
3592 if (PERLDB_LINE && PL_curstash != PL_debstash)
3593 update_debugger_info(PL_linestr, NULL, 0);
3597 bof = PL_rsfp ? TRUE : FALSE;
3598 if ((s = filter_gets(PL_linestr, PL_rsfp, 0)) == NULL) {
3601 PL_realtokenstart = -1;
3604 if (PL_preprocess && !PL_in_eval)
3605 (void)PerlProc_pclose(PL_rsfp);
3606 else if ((PerlIO *)PL_rsfp == PerlIO_stdin())
3607 PerlIO_clearerr(PL_rsfp);
3609 (void)PerlIO_close(PL_rsfp);
3611 PL_doextract = FALSE;
3613 if (!PL_in_eval && (PL_minus_n || PL_minus_p)) {
3618 sv_setpv(PL_linestr,
3621 ? ";}continue{print;}" : ";}"));
3622 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3623 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3624 PL_last_lop = PL_last_uni = NULL;
3625 PL_minus_n = PL_minus_p = 0;
3628 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3629 PL_last_lop = PL_last_uni = NULL;
3630 sv_setpvn(PL_linestr,"",0);
3631 TOKEN(';'); /* not infinite loop because rsfp is NULL now */
3633 /* If it looks like the start of a BOM or raw UTF-16,
3634 * check if it in fact is. */
3640 #ifdef PERLIO_IS_STDIO
3641 # ifdef __GNU_LIBRARY__
3642 # if __GNU_LIBRARY__ == 1 /* Linux glibc5 */
3643 # define FTELL_FOR_PIPE_IS_BROKEN
3647 # if __GLIBC__ == 1 /* maybe some glibc5 release had it like this? */
3648 # define FTELL_FOR_PIPE_IS_BROKEN
3653 #ifdef FTELL_FOR_PIPE_IS_BROKEN
3654 /* This loses the possibility to detect the bof
3655 * situation on perl -P when the libc5 is being used.
3656 * Workaround? Maybe attach some extra state to PL_rsfp?
3659 bof = PerlIO_tell(PL_rsfp) == SvCUR(PL_linestr);
3661 bof = PerlIO_tell(PL_rsfp) == (Off_t)SvCUR(PL_linestr);
3664 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3665 s = swallow_bom((U8*)s);
3669 /* Incest with pod. */
3672 sv_catsv(PL_thiswhite, PL_linestr);
3674 if (*s == '=' && strnEQ(s, "=cut", 4) && !isALPHA(s[4])) {
3675 sv_setpvn(PL_linestr, "", 0);
3676 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3677 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3678 PL_last_lop = PL_last_uni = NULL;
3679 PL_doextract = FALSE;
3683 } while (PL_doextract);
3684 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = s;
3685 if (PERLDB_LINE && PL_curstash != PL_debstash)
3686 update_debugger_info(PL_linestr, NULL, 0);
3687 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3688 PL_last_lop = PL_last_uni = NULL;
3689 if (CopLINE(PL_curcop) == 1) {
3690 while (s < PL_bufend && isSPACE(*s))
3692 if (*s == ':' && s[1] != ':') /* for csh execing sh scripts */
3696 PL_thiswhite = newSVpvn(PL_linestart, s - PL_linestart);
3700 if (*s == '#' && *(s+1) == '!')
3702 #ifdef ALTERNATE_SHEBANG
3704 static char const as[] = ALTERNATE_SHEBANG;
3705 if (*s == as[0] && strnEQ(s, as, sizeof(as) - 1))
3706 d = s + (sizeof(as) - 1);
3708 #endif /* ALTERNATE_SHEBANG */
3717 while (*d && !isSPACE(*d))
3721 #ifdef ARG_ZERO_IS_SCRIPT
3722 if (ipathend > ipath) {
3724 * HP-UX (at least) sets argv[0] to the script name,
3725 * which makes $^X incorrect. And Digital UNIX and Linux,
3726 * at least, set argv[0] to the basename of the Perl
3727 * interpreter. So, having found "#!", we'll set it right.
3729 SV * const x = GvSV(gv_fetchpvs("\030", GV_ADD|GV_NOTQUAL,
3731 assert(SvPOK(x) || SvGMAGICAL(x));
3732 if (sv_eq(x, CopFILESV(PL_curcop))) {
3733 sv_setpvn(x, ipath, ipathend - ipath);
3739 const char *bstart = SvPV_const(CopFILESV(PL_curcop),blen);
3740 const char * const lstart = SvPV_const(x,llen);
3742 bstart += blen - llen;
3743 if (strnEQ(bstart, lstart, llen) && bstart[-1] == '/') {
3744 sv_setpvn(x, ipath, ipathend - ipath);
3749 TAINT_NOT; /* $^X is always tainted, but that's OK */
3751 #endif /* ARG_ZERO_IS_SCRIPT */
3756 d = instr(s,"perl -");
3758 d = instr(s,"perl");
3760 /* avoid getting into infinite loops when shebang
3761 * line contains "Perl" rather than "perl" */
3763 for (d = ipathend-4; d >= ipath; --d) {
3764 if ((*d == 'p' || *d == 'P')
3765 && !ibcmp(d, "perl", 4))
3775 #ifdef ALTERNATE_SHEBANG
3777 * If the ALTERNATE_SHEBANG on this system starts with a
3778 * character that can be part of a Perl expression, then if
3779 * we see it but not "perl", we're probably looking at the
3780 * start of Perl code, not a request to hand off to some
3781 * other interpreter. Similarly, if "perl" is there, but
3782 * not in the first 'word' of the line, we assume the line
3783 * contains the start of the Perl program.
3785 if (d && *s != '#') {
3786 const char *c = ipath;
3787 while (*c && !strchr("; \t\r\n\f\v#", *c))
3790 d = NULL; /* "perl" not in first word; ignore */
3792 *s = '#'; /* Don't try to parse shebang line */
3794 #endif /* ALTERNATE_SHEBANG */
3795 #ifndef MACOS_TRADITIONAL
3800 !instr(s,"indir") &&
3801 instr(PL_origargv[0],"perl"))
3808 while (s < PL_bufend && isSPACE(*s))
3810 if (s < PL_bufend) {
3811 Newxz(newargv,PL_origargc+3,char*);
3813 while (s < PL_bufend && !isSPACE(*s))
3816 Copy(PL_origargv+1, newargv+2, PL_origargc+1, char*);
3819 newargv = PL_origargv;
3822 PerlProc_execv(ipath, EXEC_ARGV_CAST(newargv));
3824 Perl_croak(aTHX_ "Can't exec %s", ipath);
3828 while (*d && !isSPACE(*d))
3830 while (SPACE_OR_TAB(*d))
3834 const bool switches_done = PL_doswitches;
3835 const U32 oldpdb = PL_perldb;
3836 const bool oldn = PL_minus_n;
3837 const bool oldp = PL_minus_p;
3840 if (*d == 'M' || *d == 'm' || *d == 'C') {
3841 const char * const m = d;
3842 while (*d && !isSPACE(*d))
3844 Perl_croak(aTHX_ "Too late for \"-%.*s\" option",
3847 d = moreswitches(d);
3849 if (PL_doswitches && !switches_done) {
3850 int argc = PL_origargc;
3851 char **argv = PL_origargv;
3854 } while (argc && argv[0][0] == '-' && argv[0][1]);
3855 init_argv_symbols(argc,argv);
3857 if ((PERLDB_LINE && !oldpdb) ||
3858 ((PL_minus_n || PL_minus_p) && !(oldn || oldp)))
3859 /* if we have already added "LINE: while (<>) {",
3860 we must not do it again */
3862 sv_setpvn(PL_linestr, "", 0);
3863 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3864 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3865 PL_last_lop = PL_last_uni = NULL;
3866 PL_preambled = FALSE;
3868 (void)gv_fetchfile(PL_origfilename);
3875 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
3877 PL_lex_state = LEX_FORMLINE;
3882 #ifdef PERL_STRICT_CR
3883 Perl_warn(aTHX_ "Illegal character \\%03o (carriage return)", '\r');
3885 "\t(Maybe you didn't strip carriage returns after a network transfer?)\n");
3887 case ' ': case '\t': case '\f': case 013:
3888 #ifdef MACOS_TRADITIONAL
3892 PL_realtokenstart = -1;
3901 PL_realtokenstart = -1;
3905 if (PL_lex_state != LEX_NORMAL || (PL_in_eval && !PL_rsfp)) {
3906 if (*s == '#' && s == PL_linestart && PL_in_eval && !PL_rsfp) {
3907 /* handle eval qq[#line 1 "foo"\n ...] */
3908 CopLINE_dec(PL_curcop);
3911 if (PL_madskills && !PL_lex_formbrack && !PL_in_eval) {
3913 if (!PL_in_eval || PL_rsfp)
3918 while (d < PL_bufend && *d != '\n')
3922 else if (d > PL_bufend) /* Found by Ilya: feed random input to Perl. */
3923 Perl_croak(aTHX_ "panic: input overflow");
3926 PL_thiswhite = newSVpvn(s, d - s);
3931 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
3933 PL_lex_state = LEX_FORMLINE;
3939 if (PL_madskills && CopLINE(PL_curcop) >= 1 && !PL_lex_formbrack) {
3940 if (CopLINE(PL_curcop) == 1 && s[0] == '#' && s[1] == '!') {
3943 TOKEN(PEG); /* make sure any #! line is accessible */
3948 /* if (PL_madskills && PL_lex_formbrack) { */
3950 while (d < PL_bufend && *d != '\n')
3954 else if (d > PL_bufend) /* Found by Ilya: feed random input to Perl. */
3955 Perl_croak(aTHX_ "panic: input overflow");
3956 if (PL_madskills && CopLINE(PL_curcop) >= 1) {
3958 PL_thiswhite = newSVpvs("");
3959 if (CopLINE(PL_curcop) == 1) {
3960 sv_setpvn(PL_thiswhite, "", 0);
3963 sv_catpvn(PL_thiswhite, s, d - s);
3977 if (s[1] && isALPHA(s[1]) && !isALNUM(s[2])) {
3985 while (s < PL_bufend && SPACE_OR_TAB(*s))
3988 if (strnEQ(s,"=>",2)) {
3989 s = force_word(PL_bufptr,WORD,FALSE,FALSE,FALSE);
3990 DEBUG_T( { printbuf("### Saw unary minus before =>, forcing word %s\n", s); } );
3991 OPERATOR('-'); /* unary minus */
3993 PL_last_uni = PL_oldbufptr;
3995 case 'r': ftst = OP_FTEREAD; break;
3996 case 'w': ftst = OP_FTEWRITE; break;
3997 case 'x': ftst = OP_FTEEXEC; break;
3998 case 'o': ftst = OP_FTEOWNED; break;
3999 case 'R': ftst = OP_FTRREAD; break;
4000 case 'W': ftst = OP_FTRWRITE; break;
4001 case 'X': ftst = OP_FTREXEC; break;
4002 case 'O': ftst = OP_FTROWNED; break;
4003 case 'e': ftst = OP_FTIS; break;
4004 case 'z': ftst = OP_FTZERO; break;
4005 case 's': ftst = OP_FTSIZE; break;
4006 case 'f': ftst = OP_FTFILE; break;
4007 case 'd': ftst = OP_FTDIR; break;
4008 case 'l': ftst = OP_FTLINK; break;
4009 case 'p': ftst = OP_FTPIPE; break;
4010 case 'S': ftst = OP_FTSOCK; break;
4011 case 'u': ftst = OP_FTSUID; break;
4012 case 'g': ftst = OP_FTSGID; break;
4013 case 'k': ftst = OP_FTSVTX; break;
4014 case 'b': ftst = OP_FTBLK; break;
4015 case 'c': ftst = OP_FTCHR; break;
4016 case 't': ftst = OP_FTTTY; break;
4017 case 'T': ftst = OP_FTTEXT; break;
4018 case 'B': ftst = OP_FTBINARY; break;
4019 case 'M': case 'A': case 'C':
4020 gv_fetchpvs("\024", GV_ADD|GV_NOTQUAL, SVt_PV);
4022 case 'M': ftst = OP_FTMTIME; break;
4023 case 'A': ftst = OP_FTATIME; break;
4024 case 'C': ftst = OP_FTCTIME; break;
4032 PL_last_lop_op = (OPCODE)ftst;
4033 DEBUG_T( { PerlIO_printf(Perl_debug_log,
4034 "### Saw file test %c\n", (int)tmp);
4039 /* Assume it was a minus followed by a one-letter named
4040 * subroutine call (or a -bareword), then. */
4041 DEBUG_T( { PerlIO_printf(Perl_debug_log,
4042 "### '-%c' looked like a file test but was not\n",
4049 const char tmp = *s++;
4052 if (PL_expect == XOPERATOR)
4057 else if (*s == '>') {
4060 if (isIDFIRST_lazy_if(s,UTF)) {
4061 s = force_word(s,METHOD,FALSE,TRUE,FALSE);
4069 if (PL_expect == XOPERATOR)
4072 if (isSPACE(*s) || !isSPACE(*PL_bufptr))
4074 OPERATOR('-'); /* unary minus */
4080 const char tmp = *s++;
4083 if (PL_expect == XOPERATOR)
4088 if (PL_expect == XOPERATOR)
4091 if (isSPACE(*s) || !isSPACE(*PL_bufptr))
4098 if (PL_expect != XOPERATOR) {
4099 s = scan_ident(s, PL_bufend, PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
4100 PL_expect = XOPERATOR;
4101 force_ident(PL_tokenbuf, '*');
4114 if (PL_expect == XOPERATOR) {
4118 PL_tokenbuf[0] = '%';
4119 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1, TRUE);
4120 if (!PL_tokenbuf[1]) {
4123 PL_pending_ident = '%';
4134 && (PL_expect == XOPERATOR || PL_expect == XTERMORDORDOR))
4141 const char tmp = *s++;
4147 goto just_a_word_zero_gv;
4150 switch (PL_expect) {
4156 if (!PL_in_my || PL_lex_state != LEX_NORMAL)
4158 PL_bufptr = s; /* update in case we back off */
4164 PL_expect = XTERMBLOCK;
4167 stuffstart = s - SvPVX(PL_linestr) - 1;
4171 while (isIDFIRST_lazy_if(s,UTF)) {
4174 d = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
4175 if (isLOWER(*s) && (tmp = keyword(PL_tokenbuf, len, 0))) {
4176 if (tmp < 0) tmp = -tmp;
4191 sv = newSVpvn(s, len);
4193 d = scan_str(d,TRUE,TRUE);
4195 /* MUST advance bufptr here to avoid bogus
4196 "at end of line" context messages from yyerror().
4198 PL_bufptr = s + len;
4199 yyerror("Unterminated attribute parameter in attribute list");
4203 return REPORT(0); /* EOF indicator */
4207 sv_catsv(sv, PL_lex_stuff);
4208 attrs = append_elem(OP_LIST, attrs,
4209 newSVOP(OP_CONST, 0, sv));
4210 SvREFCNT_dec(PL_lex_stuff);
4211 PL_lex_stuff = NULL;
4214 if (len == 6 && strnEQ(SvPVX(sv), "unique", len)) {
4216 if (PL_in_my == KEY_our) {
4218 GvUNIQUE_on(cGVOPx_gv(yylval.opval));
4220 /* skip to avoid loading attributes.pm */
4222 deprecate(":unique");
4225 Perl_croak(aTHX_ "The 'unique' attribute may only be applied to 'our' variables");
4228 /* NOTE: any CV attrs applied here need to be part of
4229 the CVf_BUILTIN_ATTRS define in cv.h! */
4230 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "lvalue", len)) {
4232 CvLVALUE_on(PL_compcv);
4234 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "locked", len)) {
4236 CvLOCKED_on(PL_compcv);
4238 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "method", len)) {
4240 CvMETHOD_on(PL_compcv);
4242 else if (!PL_in_my && len == 9 && strnEQ(SvPVX(sv), "assertion", len)) {
4244 CvASSERTION_on(PL_compcv);
4246 /* After we've set the flags, it could be argued that
4247 we don't need to do the attributes.pm-based setting
4248 process, and shouldn't bother appending recognized
4249 flags. To experiment with that, uncomment the
4250 following "else". (Note that's already been
4251 uncommented. That keeps the above-applied built-in
4252 attributes from being intercepted (and possibly
4253 rejected) by a package's attribute routines, but is
4254 justified by the performance win for the common case
4255 of applying only built-in attributes.) */
4257 attrs = append_elem(OP_LIST, attrs,
4258 newSVOP(OP_CONST, 0,
4262 if (*s == ':' && s[1] != ':')
4265 break; /* require real whitespace or :'s */
4266 /* XXX losing whitespace on sequential attributes here */
4270 = (PL_expect == XOPERATOR ? '=' : '{'); /*'}(' for vi */
4271 if (*s != ';' && *s != '}' && *s != tmp
4272 && (tmp != '=' || *s != ')')) {
4273 const char q = ((*s == '\'') ? '"' : '\'');
4274 /* If here for an expression, and parsed no attrs, back
4276 if (tmp == '=' && !attrs) {
4280 /* MUST advance bufptr here to avoid bogus "at end of line"
4281 context messages from yyerror().
4284 yyerror( (const char *)
4286 ? Perl_form(aTHX_ "Invalid separator character "
4287 "%c%c%c in attribute list", q, *s, q)
4288 : "Unterminated attribute list" ) );
4296 start_force(PL_curforce);
4297 NEXTVAL_NEXTTOKE.opval = attrs;
4298 CURMAD('_', PL_nextwhite);
4303 PL_thistoken = newSVpvn(SvPVX(PL_linestr) + stuffstart,
4304 (s - SvPVX(PL_linestr)) - stuffstart);
4312 if (PL_last_lop == PL_oldoldbufptr || PL_last_uni == PL_oldoldbufptr)
4313 PL_oldbufptr = PL_oldoldbufptr; /* allow print(STDOUT 123) */
4321 const char tmp = *s++;
4326 const char tmp = *s++;
4334 if (PL_lex_brackets <= 0)
4335 yyerror("Unmatched right square bracket");
4338 if (PL_lex_state == LEX_INTERPNORMAL) {
4339 if (PL_lex_brackets == 0) {
4340 if (*s != '[' && *s != '{' && (*s != '-' || s[1] != '>'))
4341 PL_lex_state = LEX_INTERPEND;
4348 if (PL_lex_brackets > 100) {
4349 Renew(PL_lex_brackstack, PL_lex_brackets + 10, char);
4351 switch (PL_expect) {
4353 if (PL_lex_formbrack) {
4357 if (PL_oldoldbufptr == PL_last_lop)
4358 PL_lex_brackstack[PL_lex_brackets++] = XTERM;
4360 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
4361 OPERATOR(HASHBRACK);
4363 while (s < PL_bufend && SPACE_OR_TAB(*s))
4366 PL_tokenbuf[0] = '\0';
4367 if (d < PL_bufend && *d == '-') {
4368 PL_tokenbuf[0] = '-';
4370 while (d < PL_bufend && SPACE_OR_TAB(*d))
4373 if (d < PL_bufend && isIDFIRST_lazy_if(d,UTF)) {
4374 d = scan_word(d, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1,
4376 while (d < PL_bufend && SPACE_OR_TAB(*d))
4379 const char minus = (PL_tokenbuf[0] == '-');
4380 s = force_word(s + minus, WORD, FALSE, TRUE, FALSE);
4388 PL_lex_brackstack[PL_lex_brackets++] = XSTATE;
4393 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
4398 if (PL_oldoldbufptr == PL_last_lop)
4399 PL_lex_brackstack[PL_lex_brackets++] = XTERM;
4401 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
4404 if (PL_expect == XREF && PL_lex_state == LEX_INTERPNORMAL) {
4406 /* This hack is to get the ${} in the message. */
4408 yyerror("syntax error");
4411 OPERATOR(HASHBRACK);
4413 /* This hack serves to disambiguate a pair of curlies
4414 * as being a block or an anon hash. Normally, expectation
4415 * determines that, but in cases where we're not in a
4416 * position to expect anything in particular (like inside
4417 * eval"") we have to resolve the ambiguity. This code
4418 * covers the case where the first term in the curlies is a
4419 * quoted string. Most other cases need to be explicitly
4420 * disambiguated by prepending a "+" before the opening
4421 * curly in order to force resolution as an anon hash.
4423 * XXX should probably propagate the outer expectation
4424 * into eval"" to rely less on this hack, but that could
4425 * potentially break current behavior of eval"".
4429 if (*s == '\'' || *s == '"' || *s == '`') {
4430 /* common case: get past first string, handling escapes */
4431 for (t++; t < PL_bufend && *t != *s;)
4432 if (*t++ == '\\' && (*t == '\\' || *t == *s))
4436 else if (*s == 'q') {
4439 || ((*t == 'q' || *t == 'x') && ++t < PL_bufend
4442 /* skip q//-like construct */
4444 char open, close, term;
4447 while (t < PL_bufend && isSPACE(*t))
4449 /* check for q => */
4450 if (t+1 < PL_bufend && t[0] == '=' && t[1] == '>') {
4451 OPERATOR(HASHBRACK);
4455 if (term && (tmps = strchr("([{< )]}> )]}>",term)))
4459 for (t++; t < PL_bufend; t++) {
4460 if (*t == '\\' && t+1 < PL_bufend && open != '\\')
4462 else if (*t == open)
4466 for (t++; t < PL_bufend; t++) {
4467 if (*t == '\\' && t+1 < PL_bufend)
4469 else if (*t == close && --brackets <= 0)
4471 else if (*t == open)
4478 /* skip plain q word */
4479 while (t < PL_bufend && isALNUM_lazy_if(t,UTF))
4482 else if (isALNUM_lazy_if(t,UTF)) {
4484 while (t < PL_bufend && isALNUM_lazy_if(t,UTF))
4487 while (t < PL_bufend && isSPACE(*t))
4489 /* if comma follows first term, call it an anon hash */
4490 /* XXX it could be a comma expression with loop modifiers */
4491 if (t < PL_bufend && ((*t == ',' && (*s == 'q' || !isLOWER(*s)))
4492 || (*t == '=' && t[1] == '>')))
4493 OPERATOR(HASHBRACK);
4494 if (PL_expect == XREF)
4497 PL_lex_brackstack[PL_lex_brackets-1] = XSTATE;
4503 yylval.ival = CopLINE(PL_curcop);
4504 if (isSPACE(*s) || *s == '#')
4505 PL_copline = NOLINE; /* invalidate current command line number */
4510 if (PL_lex_brackets <= 0)
4511 yyerror("Unmatched right curly bracket");
4513 PL_expect = (expectation)PL_lex_brackstack[--PL_lex_brackets];
4514 if (PL_lex_brackets < PL_lex_formbrack && PL_lex_state != LEX_INTERPNORMAL)
4515 PL_lex_formbrack = 0;
4516 if (PL_lex_state == LEX_INTERPNORMAL) {
4517 if (PL_lex_brackets == 0) {
4518 if (PL_expect & XFAKEBRACK) {
4519 PL_expect &= XENUMMASK;
4520 PL_lex_state = LEX_INTERPEND;
4525 PL_thiswhite = newSVpvs("");
4526 sv_catpvn(PL_thiswhite,"}",1);
4529 return yylex(); /* ignore fake brackets */
4531 if (*s == '-' && s[1] == '>')
4532 PL_lex_state = LEX_INTERPENDMAYBE;
4533 else if (*s != '[' && *s != '{')
4534 PL_lex_state = LEX_INTERPEND;
4537 if (PL_expect & XFAKEBRACK) {
4538 PL_expect &= XENUMMASK;
4540 return yylex(); /* ignore fake brackets */
4542 start_force(PL_curforce);
4544 curmad('X', newSVpvn(s-1,1));
4545 CURMAD('_', PL_thiswhite);
4550 PL_thistoken = newSVpvs("");
4558 if (PL_expect == XOPERATOR) {
4559 if (PL_bufptr == PL_linestart && ckWARN(WARN_SEMICOLON)
4560 && isIDFIRST_lazy_if(s,UTF))
4562 CopLINE_dec(PL_curcop);
4563 Perl_warner(aTHX_ packWARN(WARN_SEMICOLON), PL_warn_nosemi);
4564 CopLINE_inc(PL_curcop);
4569 s = scan_ident(s - 1, PL_bufend, PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
4571 PL_expect = XOPERATOR;
4572 force_ident(PL_tokenbuf, '&');
4576 yylval.ival = (OPpENTERSUB_AMPER<<8);
4588 const char tmp = *s++;
4595 if (tmp && isSPACE(*s) && ckWARN(WARN_SYNTAX)
4596 && strchr("+-*/%.^&|<",tmp))
4597 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
4598 "Reversed %c= operator",(int)tmp);
4600 if (PL_expect == XSTATE && isALPHA(tmp) &&
4601 (s == PL_linestart+1 || s[-2] == '\n') )
4603 if (PL_in_eval && !PL_rsfp) {
4608 if (strnEQ(s,"=cut",4)) {
4624 PL_thiswhite = newSVpvs("");
4625 sv_catpvn(PL_thiswhite, PL_linestart,
4626 PL_bufend - PL_linestart);
4630 PL_doextract = TRUE;
4634 if (PL_lex_brackets < PL_lex_formbrack) {
4636 #ifdef PERL_STRICT_CR
4637 while (SPACE_OR_TAB(*t))
4639 while (SPACE_OR_TAB(*t) || *t == '\r')
4642 if (*t == '\n' || *t == '#') {
4653 const char tmp = *s++;
4655 /* was this !=~ where !~ was meant?
4656 * warn on m:!=~\s+([/?]|[msy]\W|tr\W): */
4658 if (*s == '~' && ckWARN(WARN_SYNTAX)) {
4659 const char *t = s+1;
4661 while (t < PL_bufend && isSPACE(*t))
4664 if (*t == '/' || *t == '?' ||
4665 ((*t == 'm' || *t == 's' || *t == 'y')
4666 && !isALNUM(t[1])) ||
4667 (*t == 't' && t[1] == 'r' && !isALNUM(t[2])))
4668 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
4669 "!=~ should be !~");
4679 if (PL_expect != XOPERATOR) {
4680 if (s[1] != '<' && !strchr(s,'>'))
4683 s = scan_heredoc(s);
4685 s = scan_inputsymbol(s);
4686 TERM(sublex_start());
4692 SHop(OP_LEFT_SHIFT);
4706 const char tmp = *s++;
4708 SHop(OP_RIGHT_SHIFT);
4709 else if (tmp == '=')
4718 if (PL_expect == XOPERATOR) {
4719 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
4721 deprecate_old(commaless_variable_list);
4722 return REPORT(','); /* grandfather non-comma-format format */
4726 if (s[1] == '#' && (isIDFIRST_lazy_if(s+2,UTF) || strchr("{$:+-", s[2]))) {
4727 PL_tokenbuf[0] = '@';
4728 s = scan_ident(s + 1, PL_bufend, PL_tokenbuf + 1,
4729 sizeof PL_tokenbuf - 1, FALSE);
4730 if (PL_expect == XOPERATOR)
4731 no_op("Array length", s);
4732 if (!PL_tokenbuf[1])
4734 PL_expect = XOPERATOR;
4735 PL_pending_ident = '#';
4739 PL_tokenbuf[0] = '$';
4740 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1,
4741 sizeof PL_tokenbuf - 1, FALSE);
4742 if (PL_expect == XOPERATOR)
4744 if (!PL_tokenbuf[1]) {
4746 yyerror("Final $ should be \\$ or $name");
4750 /* This kludge not intended to be bulletproof. */
4751 if (PL_tokenbuf[1] == '[' && !PL_tokenbuf[2]) {
4752 yylval.opval = newSVOP(OP_CONST, 0,
4753 newSViv(CopARYBASE_get(&PL_compiling)));
4754 yylval.opval->op_private = OPpCONST_ARYBASE;
4760 const char tmp = *s;
4761 if (PL_lex_state == LEX_NORMAL)
4764 if ((PL_expect != XREF || PL_oldoldbufptr == PL_last_lop)
4765 && intuit_more(s)) {
4767 PL_tokenbuf[0] = '@';
4768 if (ckWARN(WARN_SYNTAX)) {
4771 while (isSPACE(*t) || isALNUM_lazy_if(t,UTF) || *t == '$')
4774 PL_bufptr = PEEKSPACE(PL_bufptr); /* XXX can realloc */
4775 while (t < PL_bufend && *t != ']')
4777 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
4778 "Multidimensional syntax %.*s not supported",
4779 (int)((t - PL_bufptr) + 1), PL_bufptr);
4783 else if (*s == '{') {
4785 PL_tokenbuf[0] = '%';
4786 if (strEQ(PL_tokenbuf+1, "SIG") && ckWARN(WARN_SYNTAX)
4787 && (t = strchr(s, '}')) && (t = strchr(t, '=')))
4789 char tmpbuf[sizeof PL_tokenbuf];
4792 } while (isSPACE(*t));
4793 if (isIDFIRST_lazy_if(t,UTF)) {
4795 t = scan_word(t, tmpbuf, sizeof tmpbuf, TRUE,
4799 if (*t == ';' && get_cvn_flags(tmpbuf, len, 0))
4800 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
4801 "You need to quote \"%s\"",
4808 PL_expect = XOPERATOR;
4809 if (PL_lex_state == LEX_NORMAL && isSPACE((char)tmp)) {
4810 const bool islop = (PL_last_lop == PL_oldoldbufptr);
4811 if (!islop || PL_last_lop_op == OP_GREPSTART)
4812 PL_expect = XOPERATOR;
4813 else if (strchr("$@\"'`q", *s))
4814 PL_expect = XTERM; /* e.g. print $fh "foo" */
4815 else if (strchr("&*<%", *s) && isIDFIRST_lazy_if(s+1,UTF))
4816 PL_expect = XTERM; /* e.g. print $fh &sub */
4817 else if (isIDFIRST_lazy_if(s,UTF)) {
4818 char tmpbuf[sizeof PL_tokenbuf];
4820 scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
4821 if ((t2 = keyword(tmpbuf, len, 0))) {
4822 /* binary operators exclude handle interpretations */
4834 PL_expect = XTERM; /* e.g. print $fh length() */
4839 PL_expect = XTERM; /* e.g. print $fh subr() */
4842 else if (isDIGIT(*s))
4843 PL_expect = XTERM; /* e.g. print $fh 3 */
4844 else if (*s == '.' && isDIGIT(s[1]))
4845 PL_expect = XTERM; /* e.g. print $fh .3 */
4846 else if ((*s == '?' || *s == '-' || *s == '+')
4847 && !isSPACE(s[1]) && s[1] != '=')
4848 PL_expect = XTERM; /* e.g. print $fh -1 */
4849 else if (*s == '/' && !isSPACE(s[1]) && s[1] != '='
4851 PL_expect = XTERM; /* e.g. print $fh /.../
4852 XXX except DORDOR operator
4854 else if (*s == '<' && s[1] == '<' && !isSPACE(s[2])
4856 PL_expect = XTERM; /* print $fh <<"EOF" */
4859 PL_pending_ident = '$';
4863 if (PL_expect == XOPERATOR)
4865 PL_tokenbuf[0] = '@';
4866 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1, FALSE);
4867 if (!PL_tokenbuf[1]) {
4870 if (PL_lex_state == LEX_NORMAL)
4872 if ((PL_expect != XREF || PL_oldoldbufptr == PL_last_lop) && intuit_more(s)) {
4874 PL_tokenbuf[0] = '%';
4876 /* Warn about @ where they meant $. */
4877 if (*s == '[' || *s == '{') {
4878 if (ckWARN(WARN_SYNTAX)) {
4879 const char *t = s + 1;
4880 while (*t && (isALNUM_lazy_if(t,UTF) || strchr(" \t$#+-'\"", *t)))
4882 if (*t == '}' || *t == ']') {
4884 PL_bufptr = PEEKSPACE(PL_bufptr); /* XXX can realloc */
4885 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
4886 "Scalar value %.*s better written as $%.*s",
4887 (int)(t-PL_bufptr), PL_bufptr,
4888 (int)(t-PL_bufptr-1), PL_bufptr+1);
4893 PL_pending_ident = '@';
4896 case '/': /* may be division, defined-or, or pattern */
4897 if (PL_expect == XTERMORDORDOR && s[1] == '/') {
4901 case '?': /* may either be conditional or pattern */
4902 if(PL_expect == XOPERATOR) {
4910 /* A // operator. */
4920 /* Disable warning on "study /blah/" */
4921 if (PL_oldoldbufptr == PL_last_uni
4922 && (*PL_last_uni != 's' || s - PL_last_uni < 5
4923 || memNE(PL_last_uni, "study", 5)
4924 || isALNUM_lazy_if(PL_last_uni+5,UTF)
4927 s = scan_pat(s,OP_MATCH);
4928 TERM(sublex_start());
4932 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack
4933 #ifdef PERL_STRICT_CR
4936 && (s[1] == '\n' || (s[1] == '\r' && s[2] == '\n'))
4938 && (s == PL_linestart || s[-1] == '\n') )
4940 PL_lex_formbrack = 0;
4944 if (PL_expect == XOPERATOR || !isDIGIT(s[1])) {
4950 yylval.ival = OPf_SPECIAL;
4956 if (PL_expect != XOPERATOR)
4961 case '0': case '1': case '2': case '3': case '4':
4962 case '5': case '6': case '7': case '8': case '9':
4963 s = scan_num(s, &yylval);
4964 DEBUG_T( { printbuf("### Saw number in %s\n", s); } );
4965 if (PL_expect == XOPERATOR)
4970 s = scan_str(s,!!PL_madskills,FALSE);
4971 DEBUG_T( { printbuf("### Saw string before %s\n", s); } );
4972 if (PL_expect == XOPERATOR) {
4973 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
4975 deprecate_old(commaless_variable_list);
4976 return REPORT(','); /* grandfather non-comma-format format */
4983 yylval.ival = OP_CONST;
4984 TERM(sublex_start());
4987 s = scan_str(s,!!PL_madskills,FALSE);
4988 DEBUG_T( { printbuf("### Saw string before %s\n", s); } );
4989 if (PL_expect == XOPERATOR) {
4990 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
4992 deprecate_old(commaless_variable_list);
4993 return REPORT(','); /* grandfather non-comma-format format */
5000 yylval.ival = OP_CONST;
5001 /* FIXME. I think that this can be const if char *d is replaced by
5002 more localised variables. */
5003 for (d = SvPV(PL_lex_stuff, len); len; len--, d++) {
5004 if (*d == '$' || *d == '@' || *d == '\\' || !UTF8_IS_INVARIANT((U8)*d)) {
5005 yylval.ival = OP_STRINGIFY;
5009 TERM(sublex_start());
5012 s = scan_str(s,!!PL_madskills,FALSE);
5013 DEBUG_T( { printbuf("### Saw backtick string before %s\n", s); } );
5014 if (PL_expect == XOPERATOR)
5015 no_op("Backticks",s);
5018 readpipe_override();
5019 TERM(sublex_start());
5023 if (PL_lex_inwhat && isDIGIT(*s) && ckWARN(WARN_SYNTAX))
5024 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),"Can't use \\%c to mean $%c in expression",
5026 if (PL_expect == XOPERATOR)
5027 no_op("Backslash",s);
5031 if (isDIGIT(s[1]) && PL_expect != XOPERATOR) {
5032 char *start = s + 2;
5033 while (isDIGIT(*start) || *start == '_')
5035 if (*start == '.' && isDIGIT(start[1])) {
5036 s = scan_num(s, &yylval);
5039 /* avoid v123abc() or $h{v1}, allow C<print v10;> */
5040 else if (!isALPHA(*start) && (PL_expect == XTERM
5041 || PL_expect == XREF || PL_expect == XSTATE
5042 || PL_expect == XTERMORDORDOR)) {
5043 /* XXX Use gv_fetchpvn rather than stomping on a const string */
5044 const char c = *start;
5047 gv = gv_fetchpv(s, 0, SVt_PVCV);
5050 s = scan_num(s, &yylval);
5057 if (isDIGIT(s[1]) && PL_expect == XOPERATOR) {
5099 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
5101 /* Some keywords can be followed by any delimiter, including ':' */
5102 tmp = ((len == 1 && strchr("msyq", PL_tokenbuf[0])) ||
5103 (len == 2 && ((PL_tokenbuf[0] == 't' && PL_tokenbuf[1] == 'r') ||
5104 (PL_tokenbuf[0] == 'q' &&
5105 strchr("qwxr", PL_tokenbuf[1])))));
5107 /* x::* is just a word, unless x is "CORE" */
5108 if (!tmp && *s == ':' && s[1] == ':' && strNE(PL_tokenbuf, "CORE"))
5112 while (d < PL_bufend && isSPACE(*d))
5113 d++; /* no comments skipped here, or s### is misparsed */
5115 /* Is this a label? */
5116 if (!tmp && PL_expect == XSTATE
5117 && d < PL_bufend && *d == ':' && *(d + 1) != ':') {
5119 yylval.pval = CopLABEL_alloc(PL_tokenbuf);
5124 /* Check for keywords */
5125 tmp = keyword(PL_tokenbuf, len, 0);
5127 /* Is this a word before a => operator? */
5128 if (*d == '=' && d[1] == '>') {
5131 = (OP*)newSVOP(OP_CONST, 0,
5132 S_newSV_maybe_utf8(aTHX_ PL_tokenbuf, len));
5133 yylval.opval->op_private = OPpCONST_BARE;
5137 if (tmp < 0) { /* second-class keyword? */
5138 GV *ogv = NULL; /* override (winner) */
5139 GV *hgv = NULL; /* hidden (loser) */
5140 if (PL_expect != XOPERATOR && (*s != ':' || s[1] != ':')) {
5142 if ((gv = gv_fetchpvn_flags(PL_tokenbuf, len, 0, SVt_PVCV)) &&
5145 if (GvIMPORTED_CV(gv))
5147 else if (! CvMETHOD(cv))
5151 (gvp = (GV**)hv_fetch(PL_globalstash,PL_tokenbuf,len,FALSE)) &&
5152 (gv = *gvp) != (GV*)&PL_sv_undef &&
5153 GvCVu(gv) && GvIMPORTED_CV(gv))
5160 tmp = 0; /* overridden by import or by GLOBAL */
5163 && -tmp==KEY_lock /* XXX generalizable kludge */
5165 && !hv_fetchs(GvHVn(PL_incgv), "Thread.pm", FALSE))
5167 tmp = 0; /* any sub overrides "weak" keyword */
5169 else { /* no override */
5171 if (tmp == KEY_dump && ckWARN(WARN_MISC)) {
5172 Perl_warner(aTHX_ packWARN(WARN_MISC),
5173 "dump() better written as CORE::dump()");
5177 if (hgv && tmp != KEY_x && tmp != KEY_CORE
5178 && ckWARN(WARN_AMBIGUOUS)) /* never ambiguous */
5179 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
5180 "Ambiguous call resolved as CORE::%s(), %s",
5181 GvENAME(hgv), "qualify as such or use &");
5188 default: /* not a keyword */
5189 /* Trade off - by using this evil construction we can pull the
5190 variable gv into the block labelled keylookup. If not, then
5191 we have to give it function scope so that the goto from the
5192 earlier ':' case doesn't bypass the initialisation. */
5194 just_a_word_zero_gv:
5202 const char lastchar = (PL_bufptr == PL_oldoldbufptr ? 0 : PL_bufptr[-1]);
5205 SV *nextPL_nextwhite = 0;
5209 /* Get the rest if it looks like a package qualifier */
5211 if (*s == '\'' || (*s == ':' && s[1] == ':')) {
5213 s = scan_word(s, PL_tokenbuf + len, sizeof PL_tokenbuf - len,
5216 Perl_croak(aTHX_ "Bad name after %s%s", PL_tokenbuf,
5217 *s == '\'' ? "'" : "::");
5222 if (PL_expect == XOPERATOR) {
5223 if (PL_bufptr == PL_linestart) {
5224 CopLINE_dec(PL_curcop);
5225 Perl_warner(aTHX_ packWARN(WARN_SEMICOLON), PL_warn_nosemi);
5226 CopLINE_inc(PL_curcop);
5229 no_op("Bareword",s);
5232 /* Look for a subroutine with this name in current package,
5233 unless name is "Foo::", in which case Foo is a bearword
5234 (and a package name). */
5236 if (len > 2 && !PL_madskills &&
5237 PL_tokenbuf[len - 2] == ':' && PL_tokenbuf[len - 1] == ':')
5239 if (ckWARN(WARN_BAREWORD)
5240 && ! gv_fetchpvn_flags(PL_tokenbuf, len, 0, SVt_PVHV))
5241 Perl_warner(aTHX_ packWARN(WARN_BAREWORD),
5242 "Bareword \"%s\" refers to nonexistent package",
5245 PL_tokenbuf[len] = '\0';
5251 /* Mustn't actually add anything to a symbol table.
5252 But also don't want to "initialise" any placeholder
5253 constants that might already be there into full
5254 blown PVGVs with attached PVCV. */
5255 gv = gv_fetchpvn_flags(PL_tokenbuf, len,
5256 GV_NOADD_NOINIT, SVt_PVCV);
5261 /* if we saw a global override before, get the right name */
5264 sv = newSVpvs("CORE::GLOBAL::");
5265 sv_catpv(sv,PL_tokenbuf);
5268 /* If len is 0, newSVpv does strlen(), which is correct.
5269 If len is non-zero, then it will be the true length,
5270 and so the scalar will be created correctly. */
5271 sv = newSVpv(PL_tokenbuf,len);
5274 if (PL_madskills && !PL_thistoken) {
5275 char *start = SvPVX(PL_linestr) + PL_realtokenstart;
5276 PL_thistoken = newSVpv(start,s - start);
5277 PL_realtokenstart = s - SvPVX(PL_linestr);
5281 /* Presume this is going to be a bareword of some sort. */
5284 yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
5285 yylval.opval->op_private = OPpCONST_BARE;
5286 /* UTF-8 package name? */
5287 if (UTF && !IN_BYTES &&
5288 is_utf8_string((U8*)SvPVX_const(sv), SvCUR(sv)))
5291 /* And if "Foo::", then that's what it certainly is. */
5296 /* Do the explicit type check so that we don't need to force
5297 the initialisation of the symbol table to have a real GV.
5298 Beware - gv may not really be a PVGV, cv may not really be
5299 a PVCV, (because of the space optimisations that gv_init
5300 understands) But they're true if for this symbol there is
5301 respectively a typeglob and a subroutine.
5303 cv = gv ? ((SvTYPE(gv) == SVt_PVGV)
5304 /* Real typeglob, so get the real subroutine: */
5306 /* A proxy for a subroutine in this package? */
5307 : SvOK(gv) ? (CV *) gv : NULL)
5310 /* See if it's the indirect object for a list operator. */
5312 if (PL_oldoldbufptr &&
5313 PL_oldoldbufptr < PL_bufptr &&
5314 (PL_oldoldbufptr == PL_last_lop
5315 || PL_oldoldbufptr == PL_last_uni) &&
5316 /* NO SKIPSPACE BEFORE HERE! */
5317 (PL_expect == XREF ||
5318 ((PL_opargs[PL_last_lop_op] >> OASHIFT)& 7) == OA_FILEREF))
5320 bool immediate_paren = *s == '(';
5322 /* (Now we can afford to cross potential line boundary.) */
5323 s = SKIPSPACE2(s,nextPL_nextwhite);
5325 PL_nextwhite = nextPL_nextwhite; /* assume no & deception */
5328 /* Two barewords in a row may indicate method call. */
5330 if ((isIDFIRST_lazy_if(s,UTF) || *s == '$') &&
5331 (tmp = intuit_method(s, gv, cv)))
5334 /* If not a declared subroutine, it's an indirect object. */
5335 /* (But it's an indir obj regardless for sort.) */
5336 /* Also, if "_" follows a filetest operator, it's a bareword */
5339 ( !immediate_paren && (PL_last_lop_op == OP_SORT ||
5341 (PL_last_lop_op != OP_MAPSTART &&
5342 PL_last_lop_op != OP_GREPSTART))))
5343 || (PL_tokenbuf[0] == '_' && PL_tokenbuf[1] == '\0'
5344 && ((PL_opargs[PL_last_lop_op] & OA_CLASS_MASK) == OA_FILESTATOP))
5347 PL_expect = (PL_last_lop == PL_oldoldbufptr) ? XTERM : XOPERATOR;
5352 PL_expect = XOPERATOR;
5355 s = SKIPSPACE2(s,nextPL_nextwhite);
5356 PL_nextwhite = nextPL_nextwhite;
5361 /* Is this a word before a => operator? */
5362 if (*s == '=' && s[1] == '>' && !pkgname) {
5364 sv_setpv(((SVOP*)yylval.opval)->op_sv, PL_tokenbuf);
5365 if (UTF && !IN_BYTES && is_utf8_string((U8*)PL_tokenbuf, len))
5366 SvUTF8_on(((SVOP*)yylval.opval)->op_sv);
5370 /* If followed by a paren, it's certainly a subroutine. */
5375 while (SPACE_OR_TAB(*d))
5377 if (*d == ')' && (sv = gv_const_sv(gv))) {
5381 char *par = SvPVX(PL_linestr) + PL_realtokenstart;
5382 sv_catpvn(PL_thistoken, par, s - par);
5384 sv_free(PL_nextwhite);
5394 PL_nextwhite = PL_thiswhite;
5397 start_force(PL_curforce);
5399 NEXTVAL_NEXTTOKE.opval = yylval.opval;
5400 PL_expect = XOPERATOR;
5403 PL_nextwhite = nextPL_nextwhite;
5404 curmad('X', PL_thistoken);
5405 PL_thistoken = newSVpvs("");
5413 /* If followed by var or block, call it a method (unless sub) */
5415 if ((*s == '$' || *s == '{') && (!gv || !cv)) {
5416 PL_last_lop = PL_oldbufptr;
5417 PL_last_lop_op = OP_METHOD;
5421 /* If followed by a bareword, see if it looks like indir obj. */
5424 && (isIDFIRST_lazy_if(s,UTF) || *s == '$')
5425 && (tmp = intuit_method(s, gv, cv)))
5428 /* Not a method, so call it a subroutine (if defined) */
5431 if (lastchar == '-' && ckWARN_d(WARN_AMBIGUOUS))
5432 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
5433 "Ambiguous use of -%s resolved as -&%s()",
5434 PL_tokenbuf, PL_tokenbuf);
5435 /* Check for a constant sub */
5436 if ((sv = gv_const_sv(gv))) {
5438 SvREFCNT_dec(((SVOP*)yylval.opval)->op_sv);
5439 ((SVOP*)yylval.opval)->op_sv = SvREFCNT_inc_simple(sv);
5440 yylval.opval->op_private = 0;
5444 /* Resolve to GV now. */
5445 if (SvTYPE(gv) != SVt_PVGV) {
5446 gv = gv_fetchpv(PL_tokenbuf, 0, SVt_PVCV);
5447 assert (SvTYPE(gv) == SVt_PVGV);
5448 /* cv must have been some sort of placeholder, so
5449 now needs replacing with a real code reference. */
5453 op_free(yylval.opval);
5454 yylval.opval = newCVREF(0, newGVOP(OP_GV, 0, gv));
5455 yylval.opval->op_private |= OPpENTERSUB_NOPAREN;
5456 PL_last_lop = PL_oldbufptr;
5457 PL_last_lop_op = OP_ENTERSUB;
5458 /* Is there a prototype? */
5466 const char *proto = SvPV_const((SV*)cv, protolen);
5469 if ((*proto == '$' || *proto == '_') && proto[1] == '\0')
5471 while (*proto == ';')
5473 if (*proto == '&' && *s == '{') {
5474 sv_setpv(PL_subname,
5477 "__ANON__" : "__ANON__::__ANON__"));
5484 PL_nextwhite = PL_thiswhite;
5487 start_force(PL_curforce);
5488 NEXTVAL_NEXTTOKE.opval = yylval.opval;
5491 PL_nextwhite = nextPL_nextwhite;
5492 curmad('X', PL_thistoken);
5493 PL_thistoken = newSVpvs("");
5500 /* Guess harder when madskills require "best effort". */
5501 if (PL_madskills && (!gv || !GvCVu(gv))) {
5502 int probable_sub = 0;
5503 if (strchr("\"'`$@%0123456789!*+{[<", *s))
5505 else if (isALPHA(*s)) {
5509 d = scan_word(d, tmpbuf, sizeof tmpbuf, TRUE, &tmplen);
5510 if (!keyword(tmpbuf, tmplen, 0))
5513 while (d < PL_bufend && isSPACE(*d))
5515 if (*d == '=' && d[1] == '>')
5520 gv = gv_fetchpv(PL_tokenbuf, TRUE, SVt_PVCV);
5521 op_free(yylval.opval);
5522 yylval.opval = newCVREF(0, newGVOP(OP_GV, 0, gv));
5523 yylval.opval->op_private |= OPpENTERSUB_NOPAREN;
5524 PL_last_lop = PL_oldbufptr;
5525 PL_last_lop_op = OP_ENTERSUB;
5526 PL_nextwhite = PL_thiswhite;
5528 start_force(PL_curforce);
5529 NEXTVAL_NEXTTOKE.opval = yylval.opval;
5531 PL_nextwhite = nextPL_nextwhite;
5532 curmad('X', PL_thistoken);
5533 PL_thistoken = newSVpvs("");
5538 NEXTVAL_NEXTTOKE.opval = yylval.opval;
5545 /* Call it a bare word */
5547 if (PL_hints & HINT_STRICT_SUBS)
5548 yylval.opval->op_private |= OPpCONST_STRICT;
5551 if (lastchar != '-') {
5552 if (ckWARN(WARN_RESERVED)) {
5556 if (!*d && !gv_stashpv(PL_tokenbuf, 0))
5557 Perl_warner(aTHX_ packWARN(WARN_RESERVED), PL_warn_reserved,
5564 if ((lastchar == '*' || lastchar == '%' || lastchar == '&')
5565 && ckWARN_d(WARN_AMBIGUOUS)) {
5566 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
5567 "Operator or semicolon missing before %c%s",
5568 lastchar, PL_tokenbuf);
5569 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
5570 "Ambiguous use of %c resolved as operator %c",
5571 lastchar, lastchar);
5577 yylval.opval = (OP*)newSVOP(OP_CONST, 0,
5578 newSVpv(CopFILE(PL_curcop),0));
5582 yylval.opval = (OP*)newSVOP(OP_CONST, 0,
5583 Perl_newSVpvf(aTHX_ "%"IVdf, (IV)CopLINE(PL_curcop)));
5586 case KEY___PACKAGE__:
5587 yylval.opval = (OP*)newSVOP(OP_CONST, 0,
5589 ? newSVhek(HvNAME_HEK(PL_curstash))
5596 if (PL_rsfp && (!PL_in_eval || PL_tokenbuf[2] == 'D')) {
5597 const char *pname = "main";
5598 if (PL_tokenbuf[2] == 'D')
5599 pname = HvNAME_get(PL_curstash ? PL_curstash : PL_defstash);
5600 gv = gv_fetchpv(Perl_form(aTHX_ "%s::DATA", pname), GV_ADD,
5604 GvIOp(gv) = newIO();
5605 IoIFP(GvIOp(gv)) = PL_rsfp;
5606 #if defined(HAS_FCNTL) && defined(F_SETFD)
5608 const int fd = PerlIO_fileno(PL_rsfp);
5609 fcntl(fd,F_SETFD,fd >= 3);
5612 /* Mark this internal pseudo-handle as clean */
5613 IoFLAGS(GvIOp(gv)) |= IOf_UNTAINT;
5615 IoTYPE(GvIOp(gv)) = IoTYPE_PIPE;
5616 else if ((PerlIO*)PL_rsfp == PerlIO_stdin())
5617 IoTYPE(GvIOp(gv)) = IoTYPE_STD;
5619 IoTYPE(GvIOp(gv)) = IoTYPE_RDONLY;
5620 #if defined(WIN32) && !defined(PERL_TEXTMODE_SCRIPTS)
5621 /* if the script was opened in binmode, we need to revert
5622 * it to text mode for compatibility; but only iff it has CRs
5623 * XXX this is a questionable hack at best. */
5624 if (PL_bufend-PL_bufptr > 2
5625 && PL_bufend[-1] == '\n' && PL_bufend[-2] == '\r')
5628 if (IoTYPE(GvIOp(gv)) == IoTYPE_RDONLY) {
5629 loc = PerlIO_tell(PL_rsfp);
5630 (void)PerlIO_seek(PL_rsfp, 0L, 0);
5633 if (PerlLIO_setmode(PL_rsfp, O_TEXT) != -1) {
5635 if (PerlLIO_setmode(PerlIO_fileno(PL_rsfp), O_TEXT) != -1) {
5636 #endif /* NETWARE */
5637 #ifdef PERLIO_IS_STDIO /* really? */
5638 # if defined(__BORLANDC__)
5639 /* XXX see note in do_binmode() */
5640 ((FILE*)PL_rsfp)->flags &= ~_F_BIN;
5644 PerlIO_seek(PL_rsfp, loc, 0);
5648 #ifdef PERLIO_LAYERS
5651 PerlIO_apply_layers(aTHX_ PL_rsfp, NULL, ":utf8");
5652 else if (PL_encoding) {
5659 XPUSHs(PL_encoding);
5661 call_method("name", G_SCALAR);
5665 PerlIO_apply_layers(aTHX_ PL_rsfp, NULL,
5666 Perl_form(aTHX_ ":encoding(%"SVf")",
5675 if (PL_realtokenstart >= 0) {
5676 char *tstart = SvPVX(PL_linestr) + PL_realtokenstart;
5678 PL_endwhite = newSVpvs("");
5679 sv_catsv(PL_endwhite, PL_thiswhite);
5681 sv_catpvn(PL_endwhite, tstart, PL_bufend - tstart);
5682 PL_realtokenstart = -1;
5684 while ((s = filter_gets(PL_endwhite, PL_rsfp,
5685 SvCUR(PL_endwhite))) != Nullch) ;
5700 if (PL_expect == XSTATE) {
5707 if (*s == ':' && s[1] == ':') {
5710 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
5711 if (!(tmp = keyword(PL_tokenbuf, len, 0)))
5712 Perl_croak(aTHX_ "CORE::%s is not a keyword", PL_tokenbuf);
5715 else if (tmp == KEY_require || tmp == KEY_do)
5716 /* that's a way to remember we saw "CORE::" */
5729 LOP(OP_ACCEPT,XTERM);
5735 LOP(OP_ATAN2,XTERM);
5741 LOP(OP_BINMODE,XTERM);
5744 LOP(OP_BLESS,XTERM);
5753 /* When 'use switch' is in effect, continue has a dual
5754 life as a control operator. */
5756 if (!FEATURE_IS_ENABLED("switch"))
5759 /* We have to disambiguate the two senses of
5760 "continue". If the next token is a '{' then
5761 treat it as the start of a continue block;
5762 otherwise treat it as a control operator.
5774 (void)gv_fetchpvs("ENV", GV_ADD|GV_NOTQUAL, SVt_PVHV);
5791 if (!PL_cryptseen) {
5792 PL_cryptseen = TRUE;
5796 LOP(OP_CRYPT,XTERM);
5799 LOP(OP_CHMOD,XTERM);
5802 LOP(OP_CHOWN,XTERM);
5805 LOP(OP_CONNECT,XTERM);
5824 s = force_word(s,WORD,TRUE,TRUE,FALSE);
5825 if (orig_keyword == KEY_do) {
5834 PL_hints |= HINT_BLOCK_SCOPE;
5844 gv_fetchpvs("AnyDBM_File::ISA", GV_ADDMULTI, SVt_PVAV);
5845 LOP(OP_DBMOPEN,XTERM);
5851 s = force_word(s,WORD,TRUE,FALSE,FALSE);
5858 yylval.ival = CopLINE(PL_curcop);
5874 PL_expect = (*s == '{') ? XTERMBLOCK : XTERM;
5875 UNIBRACK(OP_ENTEREVAL);
5893 case KEY_endhostent:
5899 case KEY_endservent:
5902 case KEY_endprotoent:
5913 yylval.ival = CopLINE(PL_curcop);
5915 if (PL_expect == XSTATE && isIDFIRST_lazy_if(s,UTF)) {
5918 int soff = s - SvPVX(PL_linestr); /* for skipspace realloc */
5921 if ((PL_bufend - p) >= 3 &&
5922 strnEQ(p, "my", 2) && isSPACE(*(p + 2)))
5924 else if ((PL_bufend - p) >= 4 &&
5925 strnEQ(p, "our", 3) && isSPACE(*(p + 3)))
5928 if (isIDFIRST_lazy_if(p,UTF)) {
5929 p = scan_ident(p, PL_bufend,
5930 PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
5934 Perl_croak(aTHX_ "Missing $ on loop variable");
5936 s = SvPVX(PL_linestr) + soff;
5942 LOP(OP_FORMLINE,XTERM);
5948 LOP(OP_FCNTL,XTERM);
5954 LOP(OP_FLOCK,XTERM);
5963 LOP(OP_GREPSTART, XREF);
5966 s = force_word(s,WORD,TRUE,FALSE,FALSE);
5981 case KEY_getpriority:
5982 LOP(OP_GETPRIORITY,XTERM);
5984 case KEY_getprotobyname:
5987 case KEY_getprotobynumber:
5988 LOP(OP_GPBYNUMBER,XTERM);
5990 case KEY_getprotoent:
6002 case KEY_getpeername:
6003 UNI(OP_GETPEERNAME);
6005 case KEY_gethostbyname:
6008 case KEY_gethostbyaddr:
6009 LOP(OP_GHBYADDR,XTERM);
6011 case KEY_gethostent:
6014 case KEY_getnetbyname:
6017 case KEY_getnetbyaddr:
6018 LOP(OP_GNBYADDR,XTERM);
6023 case KEY_getservbyname:
6024 LOP(OP_GSBYNAME,XTERM);
6026 case KEY_getservbyport:
6027 LOP(OP_GSBYPORT,XTERM);
6029 case KEY_getservent:
6032 case KEY_getsockname:
6033 UNI(OP_GETSOCKNAME);
6035 case KEY_getsockopt:
6036 LOP(OP_GSOCKOPT,XTERM);
6051 yylval.ival = CopLINE(PL_curcop);
6062 yylval.ival = CopLINE(PL_curcop);
6066 LOP(OP_INDEX,XTERM);
6072 LOP(OP_IOCTL,XTERM);
6084 s = force_word(s,WORD,TRUE,FALSE,FALSE);
6116 LOP(OP_LISTEN,XTERM);
6125 s = scan_pat(s,OP_MATCH);
6126 TERM(sublex_start());
6129 LOP(OP_MAPSTART, XREF);
6132 LOP(OP_MKDIR,XTERM);
6135 LOP(OP_MSGCTL,XTERM);
6138 LOP(OP_MSGGET,XTERM);
6141 LOP(OP_MSGRCV,XTERM);
6144 LOP(OP_MSGSND,XTERM);
6151 if (isIDFIRST_lazy_if(s,UTF)) {
6155 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, TRUE, &len);
6156 if (len == 3 && strnEQ(PL_tokenbuf, "sub", 3))
6158 PL_in_my_stash = find_in_my_stash(PL_tokenbuf, len);
6159 if (!PL_in_my_stash) {
6162 my_snprintf(tmpbuf, sizeof(tmpbuf), "No such class %.1000s", PL_tokenbuf);
6166 if (PL_madskills) { /* just add type to declarator token */
6167 sv_catsv(PL_thistoken, PL_nextwhite);
6169 sv_catpvn(PL_thistoken, start, s - start);
6177 s = force_word(s,WORD,TRUE,FALSE,FALSE);
6184 s = tokenize_use(0, s);
6188 if (*s == '(' || (s = SKIPSPACE1(s), *s == '('))
6195 if (isIDFIRST_lazy_if(s,UTF)) {
6197 for (d = s; isALNUM_lazy_if(d,UTF);)
6199 for (t=d; isSPACE(*t);)
6201 if ( *t && strchr("|&*+-=!?:.", *t) && ckWARN_d(WARN_PRECEDENCE)
6203 && !(t[0] == '=' && t[1] == '>')
6205 int parms_len = (int)(d-s);
6206 Perl_warner(aTHX_ packWARN(WARN_PRECEDENCE),
6207 "Precedence problem: open %.*s should be open(%.*s)",
6208 parms_len, s, parms_len, s);
6214 yylval.ival = OP_OR;
6224 LOP(OP_OPEN_DIR,XTERM);
6227 checkcomma(s,PL_tokenbuf,"filehandle");
6231 checkcomma(s,PL_tokenbuf,"filehandle");
6250 s = force_word(s,WORD,FALSE,TRUE,FALSE);
6254 LOP(OP_PIPE_OP,XTERM);
6257 s = scan_str(s,!!PL_madskills,FALSE);
6260 yylval.ival = OP_CONST;
6261 TERM(sublex_start());
6267 s = scan_str(s,!!PL_madskills,FALSE);
6270 PL_expect = XOPERATOR;
6272 if (SvCUR(PL_lex_stuff)) {
6275 d = SvPV_force(PL_lex_stuff, len);
6277 for (; isSPACE(*d) && len; --len, ++d)
6282 if (!warned && ckWARN(WARN_QW)) {
6283 for (; !isSPACE(*d) && len; --len, ++d) {
6285 Perl_warner(aTHX_ packWARN(WARN_QW),
6286 "Possible attempt to separate words with commas");
6289 else if (*d == '#') {
6290 Perl_warner(aTHX_ packWARN(WARN_QW),
6291 "Possible attempt to put comments in qw() list");
6297 for (; !isSPACE(*d) && len; --len, ++d)
6300 sv = newSVpvn(b, d-b);
6301 if (DO_UTF8(PL_lex_stuff))
6303 words = append_elem(OP_LIST, words,
6304 newSVOP(OP_CONST, 0, tokeq(sv)));
6308 start_force(PL_curforce);
6309 NEXTVAL_NEXTTOKE.opval = words;
6314 SvREFCNT_dec(PL_lex_stuff);
6315 PL_lex_stuff = NULL;
6321 s = scan_str(s,!!PL_madskills,FALSE);
6324 yylval.ival = OP_STRINGIFY;
6325 if (SvIVX(PL_lex_stuff) == '\'')
6326 SvIV_set(PL_lex_stuff, 0); /* qq'$foo' should intepolate */
6327 TERM(sublex_start());
6330 s = scan_pat(s,OP_QR);
6331 TERM(sublex_start());
6334 s = scan_str(s,!!PL_madskills,FALSE);
6337 readpipe_override();
6338 TERM(sublex_start());
6346 s = force_version(s, FALSE);
6348 else if (*s != 'v' || !isDIGIT(s[1])
6349 || (s = force_version(s, TRUE), *s == 'v'))
6351 *PL_tokenbuf = '\0';
6352 s = force_word(s,WORD,TRUE,TRUE,FALSE);
6353 if (isIDFIRST_lazy_if(PL_tokenbuf,UTF))
6354 gv_stashpvn(PL_tokenbuf, strlen(PL_tokenbuf), GV_ADD);
6356 yyerror("<> should be quotes");
6358 if (orig_keyword == KEY_require) {
6366 PL_last_uni = PL_oldbufptr;
6367 PL_last_lop_op = OP_REQUIRE;
6369 return REPORT( (int)REQUIRE );
6375 s = force_word(s,WORD,TRUE,FALSE,FALSE);
6379 LOP(OP_RENAME,XTERM);
6388 LOP(OP_RINDEX,XTERM);
6398 UNIDOR(OP_READLINE);
6411 LOP(OP_REVERSE,XTERM);
6414 UNIDOR(OP_READLINK);
6422 TERM(sublex_start());
6424 TOKEN(1); /* force error */
6427 checkcomma(s,PL_tokenbuf,"filehandle");
6437 LOP(OP_SELECT,XTERM);
6443 LOP(OP_SEMCTL,XTERM);
6446 LOP(OP_SEMGET,XTERM);
6449 LOP(OP_SEMOP,XTERM);
6455 LOP(OP_SETPGRP,XTERM);
6457 case KEY_setpriority:
6458 LOP(OP_SETPRIORITY,XTERM);
6460 case KEY_sethostent:
6466 case KEY_setservent:
6469 case KEY_setprotoent:
6479 LOP(OP_SEEKDIR,XTERM);
6481 case KEY_setsockopt:
6482 LOP(OP_SSOCKOPT,XTERM);
6488 LOP(OP_SHMCTL,XTERM);
6491 LOP(OP_SHMGET,XTERM);
6494 LOP(OP_SHMREAD,XTERM);
6497 LOP(OP_SHMWRITE,XTERM);
6500 LOP(OP_SHUTDOWN,XTERM);
6509 LOP(OP_SOCKET,XTERM);
6511 case KEY_socketpair:
6512 LOP(OP_SOCKPAIR,XTERM);
6515 checkcomma(s,PL_tokenbuf,"subroutine name");
6517 if (*s == ';' || *s == ')') /* probably a close */
6518 Perl_croak(aTHX_ "sort is now a reserved word");
6520 s = force_word(s,WORD,TRUE,TRUE,FALSE);
6524 LOP(OP_SPLIT,XTERM);
6527 LOP(OP_SPRINTF,XTERM);
6530 LOP(OP_SPLICE,XTERM);
6545 LOP(OP_SUBSTR,XTERM);
6551 char tmpbuf[sizeof PL_tokenbuf];
6552 SSize_t tboffset = 0;
6553 expectation attrful;
6554 bool have_name, have_proto;
6555 const int key = tmp;
6560 char *tstart = SvPVX(PL_linestr) + PL_realtokenstart;
6561 SV *subtoken = newSVpvn(tstart, s - tstart);
6565 s = SKIPSPACE2(s,tmpwhite);
6570 if (isIDFIRST_lazy_if(s,UTF) || *s == '\'' ||
6571 (*s == ':' && s[1] == ':'))
6578 attrful = XATTRBLOCK;
6579 /* remember buffer pos'n for later force_word */
6580 tboffset = s - PL_oldbufptr;
6581 d = scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
6584 nametoke = newSVpvn(s, d - s);
6586 if (strchr(tmpbuf, ':'))
6587 sv_setpv(PL_subname, tmpbuf);
6589 sv_setsv(PL_subname,PL_curstname);
6590 sv_catpvs(PL_subname,"::");
6591 sv_catpvn(PL_subname,tmpbuf,len);
6598 CURMAD('X', nametoke);
6599 CURMAD('_', tmpwhite);
6600 (void) force_word(PL_oldbufptr + tboffset, WORD,
6603 s = SKIPSPACE2(d,tmpwhite);
6610 Perl_croak(aTHX_ "Missing name in \"my sub\"");
6611 PL_expect = XTERMBLOCK;
6612 attrful = XATTRTERM;
6613 sv_setpvn(PL_subname,"?",1);
6617 if (key == KEY_format) {
6619 PL_lex_formbrack = PL_lex_brackets + 1;
6621 PL_thistoken = subtoken;
6625 (void) force_word(PL_oldbufptr + tboffset, WORD,
6631 /* Look for a prototype */
6634 bool bad_proto = FALSE;
6635 const bool warnsyntax = ckWARN(WARN_SYNTAX);
6637 s = scan_str(s,!!PL_madskills,FALSE);
6639 Perl_croak(aTHX_ "Prototype not terminated");
6640 /* strip spaces and check for bad characters */
6641 d = SvPVX(PL_lex_stuff);
6643 for (p = d; *p; ++p) {
6646 if (warnsyntax && !strchr("$@%*;[]&\\_", *p))
6652 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
6653 "Illegal character in prototype for %"SVf" : %s",
6654 SVfARG(PL_subname), d);
6655 SvCUR_set(PL_lex_stuff, tmp);
6660 CURMAD('q', PL_thisopen);
6661 CURMAD('_', tmpwhite);
6662 CURMAD('=', PL_thisstuff);
6663 CURMAD('Q', PL_thisclose);
6664 NEXTVAL_NEXTTOKE.opval =
6665 (OP*)newSVOP(OP_CONST, 0, PL_lex_stuff);
6666 PL_lex_stuff = Nullsv;
6669 s = SKIPSPACE2(s,tmpwhite);
6677 if (*s == ':' && s[1] != ':')
6678 PL_expect = attrful;
6679 else if (*s != '{' && key == KEY_sub) {
6681 Perl_croak(aTHX_ "Illegal declaration of anonymous subroutine");
6683 Perl_croak(aTHX_ "Illegal declaration of subroutine %"SVf, SVfARG(PL_subname));
6690 curmad('^', newSVpvs(""));
6691 CURMAD('_', tmpwhite);
6695 PL_thistoken = subtoken;
6698 NEXTVAL_NEXTTOKE.opval =
6699 (OP*)newSVOP(OP_CONST, 0, PL_lex_stuff);
6700 PL_lex_stuff = NULL;
6705 sv_setpv(PL_subname,
6707 (PL_curstash ? "__ANON__" : "__ANON__::__ANON__"));
6711 (void) force_word(PL_oldbufptr + tboffset, WORD,
6721 LOP(OP_SYSTEM,XREF);
6724 LOP(OP_SYMLINK,XTERM);
6727 LOP(OP_SYSCALL,XTERM);
6730 LOP(OP_SYSOPEN,XTERM);
6733 LOP(OP_SYSSEEK,XTERM);
6736 LOP(OP_SYSREAD,XTERM);
6739 LOP(OP_SYSWRITE,XTERM);
6743 TERM(sublex_start());
6764 LOP(OP_TRUNCATE,XTERM);
6776 yylval.ival = CopLINE(PL_curcop);
6780 yylval.ival = CopLINE(PL_curcop);
6784 LOP(OP_UNLINK,XTERM);
6790 LOP(OP_UNPACK,XTERM);
6793 LOP(OP_UTIME,XTERM);
6799 LOP(OP_UNSHIFT,XTERM);
6802 s = tokenize_use(1, s);
6812 yylval.ival = CopLINE(PL_curcop);
6816 yylval.ival = CopLINE(PL_curcop);
6820 PL_hints |= HINT_BLOCK_SCOPE;
6827 LOP(OP_WAITPID,XTERM);
6836 ctl_l[0] = toCTRL('L');
6838 gv_fetchpvn_flags(ctl_l, 1, GV_ADD|GV_NOTQUAL, SVt_PV);
6841 /* Make sure $^L is defined */
6842 gv_fetchpvs("\f", GV_ADD|GV_NOTQUAL, SVt_PV);
6847 if (PL_expect == XOPERATOR)
6853 yylval.ival = OP_XOR;
6858 TERM(sublex_start());
6863 #pragma segment Main
6867 S_pending_ident(pTHX)
6872 /* pit holds the identifier we read and pending_ident is reset */
6873 char pit = PL_pending_ident;
6874 PL_pending_ident = 0;
6876 /* PL_realtokenstart = realtokenend = PL_bufptr - SvPVX(PL_linestr); */
6877 DEBUG_T({ PerlIO_printf(Perl_debug_log,
6878 "### Pending identifier '%s'\n", PL_tokenbuf); });
6880 /* if we're in a my(), we can't allow dynamics here.
6881 $foo'bar has already been turned into $foo::bar, so
6882 just check for colons.
6884 if it's a legal name, the OP is a PADANY.
6887 if (PL_in_my == KEY_our) { /* "our" is merely analogous to "my" */
6888 if (strchr(PL_tokenbuf,':'))
6889 yyerror(Perl_form(aTHX_ "No package name allowed for "
6890 "variable %s in \"our\"",
6892 tmp = allocmy(PL_tokenbuf);
6895 if (strchr(PL_tokenbuf,':'))
6896 yyerror(Perl_form(aTHX_ PL_no_myglob,
6897 PL_in_my == KEY_my ? "my" : "state", PL_tokenbuf));
6899 yylval.opval = newOP(OP_PADANY, 0);
6900 yylval.opval->op_targ = allocmy(PL_tokenbuf);
6906 build the ops for accesses to a my() variable.
6908 Deny my($a) or my($b) in a sort block, *if* $a or $b is
6909 then used in a comparison. This catches most, but not
6910 all cases. For instance, it catches
6911 sort { my($a); $a <=> $b }
6913 sort { my($a); $a < $b ? -1 : $a == $b ? 0 : 1; }
6914 (although why you'd do that is anyone's guess).
6917 if (!strchr(PL_tokenbuf,':')) {
6919 tmp = pad_findmy(PL_tokenbuf);
6920 if (tmp != NOT_IN_PAD) {
6921 /* might be an "our" variable" */
6922 if (PAD_COMPNAME_FLAGS_isOUR(tmp)) {
6923 /* build ops for a bareword */
6924 HV * const stash = PAD_COMPNAME_OURSTASH(tmp);
6925 HEK * const stashname = HvNAME_HEK(stash);
6926 SV * const sym = newSVhek(stashname);
6927 sv_catpvs(sym, "::");
6928 sv_catpv(sym, PL_tokenbuf+1);
6929 yylval.opval = (OP*)newSVOP(OP_CONST, 0, sym);
6930 yylval.opval->op_private = OPpCONST_ENTERED;
6933 ? (GV_ADDMULTI | GV_ADDINEVAL)
6936 ((PL_tokenbuf[0] == '$') ? SVt_PV
6937 : (PL_tokenbuf[0] == '@') ? SVt_PVAV
6942 /* if it's a sort block and they're naming $a or $b */
6943 if (PL_last_lop_op == OP_SORT &&
6944 PL_tokenbuf[0] == '$' &&
6945 (PL_tokenbuf[1] == 'a' || PL_tokenbuf[1] == 'b')
6948 for (d = PL_in_eval ? PL_oldoldbufptr : PL_linestart;
6949 d < PL_bufend && *d != '\n';
6952 if (strnEQ(d,"<=>",3) || strnEQ(d,"cmp",3)) {
6953 Perl_croak(aTHX_ "Can't use \"my %s\" in sort comparison",
6959 yylval.opval = newOP(OP_PADANY, 0);
6960 yylval.opval->op_targ = tmp;
6966 Whine if they've said @foo in a doublequoted string,
6967 and @foo isn't a variable we can find in the symbol
6970 if (pit == '@' && PL_lex_state != LEX_NORMAL && !PL_lex_brackets) {
6971 GV *gv = gv_fetchpv(PL_tokenbuf+1, 0, SVt_PVAV);
6972 if ((!gv || ((PL_tokenbuf[0] == '@') ? !GvAV(gv) : !GvHV(gv)))
6973 && ckWARN(WARN_AMBIGUOUS))
6975 /* Downgraded from fatal to warning 20000522 mjd */
6976 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
6977 "Possible unintended interpolation of %s in string",
6982 /* build ops for a bareword */
6983 yylval.opval = (OP*)newSVOP(OP_CONST, 0, newSVpv(PL_tokenbuf+1, 0));
6984 yylval.opval->op_private = OPpCONST_ENTERED;
6987 /* If the identifier refers to a stash, don't autovivify it.
6988 * Change 24660 had the side effect of causing symbol table
6989 * hashes to always be defined, even if they were freshly
6990 * created and the only reference in the entire program was
6991 * the single statement with the defined %foo::bar:: test.
6992 * It appears that all code in the wild doing this actually
6993 * wants to know whether sub-packages have been loaded, so
6994 * by avoiding auto-vivifying symbol tables, we ensure that
6995 * defined %foo::bar:: continues to be false, and the existing
6996 * tests still give the expected answers, even though what
6997 * they're actually testing has now changed subtly.
6999 (*PL_tokenbuf == '%' && *(d = PL_tokenbuf + strlen(PL_tokenbuf) - 1) == ':' && d[-1] == ':'
7001 : PL_in_eval ? (GV_ADDMULTI | GV_ADDINEVAL) : GV_ADD),
7002 ((PL_tokenbuf[0] == '$') ? SVt_PV
7003 : (PL_tokenbuf[0] == '@') ? SVt_PVAV
7009 * The following code was generated by perl_keyword.pl.
7013 Perl_keyword (pTHX_ const char *name, I32 len, bool all_keywords)
7018 case 1: /* 5 tokens of length 1 */
7050 case 2: /* 18 tokens of length 2 */
7196 case 3: /* 29 tokens of length 3 */
7200 if (name[1] == 'N' &&
7263 if (name[1] == 'i' &&
7285 return (all_keywords || FEATURE_IS_ENABLED("err") ? -KEY_err : 0);
7303 if (name[1] == 'o' &&
7312 if (name[1] == 'e' &&
7321 if (name[1] == 'n' &&
7330 if (name[1] == 'o' &&
7339 if (name[1] == 'a' &&
7348 if (name[1] == 'o' &&
7410 if (name[1] == 'e' &&
7424 return (all_keywords || FEATURE_IS_ENABLED("say") ? KEY_say : 0);
7450 if (name[1] == 'i' &&
7459 if (name[1] == 's' &&
7468 if (name[1] == 'e' &&
7477 if (name[1] == 'o' &&
7489 case 4: /* 41 tokens of length 4 */
7493 if (name[1] == 'O' &&
7503 if (name[1] == 'N' &&
7513 if (name[1] == 'i' &&
7523 if (name[1] == 'h' &&
7533 if (name[1] == 'u' &&
7546 if (name[2] == 'c' &&
7555 if (name[2] == 's' &&
7564 if (name[2] == 'a' &&
7600 if (name[1] == 'o' &&
7613 if (name[2] == 't' &&
7622 if (name[2] == 'o' &&
7631 if (name[2] == 't' &&
7640 if (name[2] == 'e' &&
7653 if (name[1] == 'o' &&
7666 if (name[2] == 'y' &&
7675 if (name[2] == 'l' &&
7691 if (name[2] == 's' &&
7700 if (name[2] == 'n' &&
7709 if (name[2] == 'c' &&
7722 if (name[1] == 'e' &&
7732 if (name[1] == 'p' &&
7745 if (name[2] == 'c' &&
7754 if (name[2] == 'p' &&
7763 if (name[2] == 's' &&
7779 if (name[2] == 'n' &&
7849 if (name[2] == 'r' &&
7858 if (name[2] == 'r' &&
7867 if (name[2] == 'a' &&
7883 if (name[2] == 'l' &&
7945 if (name[2] == 'e' &&
7948 return (all_keywords || FEATURE_IS_ENABLED("switch") ? KEY_when : 0);
7961 case 5: /* 39 tokens of length 5 */
7965 if (name[1] == 'E' &&
7976 if (name[1] == 'H' &&
7990 if (name[2] == 'a' &&
8000 if (name[2] == 'a' &&
8017 if (name[2] == 'e' &&
8027 if (name[2] == 'e' &&
8031 return (all_keywords || FEATURE_IS_ENABLED("switch") ? -KEY_break : 0);
8047 if (name[3] == 'i' &&
8056 if (name[3] == 'o' &&
8092 if (name[2] == 'o' &&
8102 if (name[2] == 'y' &&
8116 if (name[1] == 'l' &&
8130 if (name[2] == 'n' &&
8140 if (name[2] == 'o' &&
8154 if (name[1] == 'i' &&
8159 return (all_keywords || FEATURE_IS_ENABLED("switch") ? KEY_given : 0);
8168 if (name[2] == 'd' &&
8178 if (name[2] == 'c' &&
8195 if (name[2] == 'c' &&
8205 if (name[2] == 't' &&
8219 if (name[1] == 'k' &&
8230 if (name[1] == 'r' &&
8244 if (name[2] == 's' &&
8254 if (name[2] == 'd' &&
8271 if (name[2] == 'm' &&
8281 if (name[2] == 'i' &&
8291 if (name[2] == 'e' &&
8301 if (name[2] == 'l' &&
8311 if (name[2] == 'a' &&
8324 if (name[3] == 't' &&
8327 return (all_keywords || FEATURE_IS_ENABLED("state") ? KEY_state : 0);
8333 if (name[3] == 'd' &&
8350 if (name[1] == 'i' &&
8364 if (name[2] == 'a' &&
8377 if (name[3] == 'e' &&
8412 if (name[2] == 'i' &&
8429 if (name[2] == 'i' &&
8439 if (name[2] == 'i' &&
8456 case 6: /* 33 tokens of length 6 */
8460 if (name[1] == 'c' &&
8475 if (name[2] == 'l' &&
8486 if (name[2] == 'r' &&
8501 if (name[1] == 'e' &&
8516 if (name[2] == 's' &&
8521 if(ckWARN_d(WARN_SYNTAX))
8522 Perl_warner(aTHX_ packWARN(WARN_SYNTAX), "elseif should be elsif");
8528 if (name[2] == 'i' &&
8546 if (name[2] == 'l' &&
8557 if (name[2] == 'r' &&
8572 if (name[1] == 'm' &&
8587 if (name[2] == 'n' &&
8598 if (name[2] == 's' &&
8613 if (name[1] == 's' &&
8619 if (name[4] == 't' &&
8628 if (name[4] == 'e' &&
8637 if (name[4] == 'c' &&
8646 if (name[4] == 'n' &&
8662 if (name[1] == 'r' &&
8680 if (name[3] == 'a' &&
8690 if (name[3] == 'u' &&
8704 if (name[2] == 'n' &&
8722 if (name[2] == 'a' &&
8736 if (name[3] == 'e' &&
8749 if (name[4] == 't' &&
8758 if (name[4] == 'e' &&
8780 if (name[4] == 't' &&
8789 if (name[4] == 'e' &&
8805 if (name[2] == 'c' &&
8816 if (name[2] == 'l' &&
8827 if (name[2] == 'b' &&
8838 if (name[2] == 's' &&
8861 if (name[4] == 's' &&
8870 if (name[4] == 'n' &&
8883 if (name[3] == 'a' &&
8900 if (name[1] == 'a' &&
8915 case 7: /* 29 tokens of length 7 */
8919 if (name[1] == 'E' &&
8932 if (name[1] == '_' &&
8945 if (name[1] == 'i' &&
8952 return -KEY_binmode;
8958 if (name[1] == 'o' &&
8965 return -KEY_connect;
8974 if (name[2] == 'm' &&
8980 return -KEY_dbmopen;
8991 if (name[4] == 'u' &&
8995 return (all_keywords || FEATURE_IS_ENABLED("switch") ? KEY_default : 0);
9001 if (name[4] == 'n' &&
9022 if (name[1] == 'o' &&
9035 if (name[1] == 'e' &&
9042 if (name[5] == 'r' &&
9045 return -KEY_getpgrp;
9051 if (name[5] == 'i' &&
9054 return -KEY_getppid;
9067 if (name[1] == 'c' &&
9074 return -KEY_lcfirst;
9080 if (name[1] == 'p' &&
9087 return -KEY_opendir;
9093 if (name[1] == 'a' &&
9111 if (name[3] == 'd' &&
9116 return -KEY_readdir;
9122 if (name[3] == 'u' &&
9133 if (name[3] == 'e' &&
9138 return -KEY_reverse;
9157 if (name[3] == 'k' &&
9162 return -KEY_seekdir;
9168 if (name[3] == 'p' &&
9173 return -KEY_setpgrp;
9183 if (name[2] == 'm' &&
9189 return -KEY_shmread;
9195 if (name[2] == 'r' &&
9201 return -KEY_sprintf;
9210 if (name[3] == 'l' &&
9215 return -KEY_symlink;
9224 if (name[4] == 'a' &&
9228 return -KEY_syscall;
9234 if (name[4] == 'p' &&
9238 return -KEY_sysopen;
9244 if (name[4] == 'e' &&
9248 return -KEY_sysread;
9254 if (name[4] == 'e' &&
9258 return -KEY_sysseek;
9276 if (name[1] == 'e' &&
9283 return -KEY_telldir;
9292 if (name[2] == 'f' &&
9298 return -KEY_ucfirst;
9304 if (name[2] == 's' &&
9310 return -KEY_unshift;
9320 if (name[1] == 'a' &&
9327 return -KEY_waitpid;
9336 case 8: /* 26 tokens of length 8 */
9340 if (name[1] == 'U' &&
9348 return KEY_AUTOLOAD;
9359 if (name[3] == 'A' &&
9365 return KEY___DATA__;
9371 if (name[3] == 'I' &&
9377 return -KEY___FILE__;
9383 if (name[3] == 'I' &&
9389 return -KEY___LINE__;
9405 if (name[2] == 'o' &&
9412 return -KEY_closedir;
9418 if (name[2] == 'n' &&
9425 return -KEY_continue;
9435 if (name[1] == 'b' &&
9443 return -KEY_dbmclose;
9449 if (name[1] == 'n' &&
9455 if (name[4] == 'r' &&
9460 return -KEY_endgrent;
9466 if (name[4] == 'w' &&
9471 return -KEY_endpwent;
9484 if (name[1] == 'o' &&
9492 return -KEY_formline;
9498 if (name[1] == 'e' &&
9509 if (name[6] == 'n' &&
9512 return -KEY_getgrent;
9518 if (name[6] == 'i' &&
9521 return -KEY_getgrgid;
9527 if (name[6] == 'a' &&
9530 return -KEY_getgrnam;
9543 if (name[4] == 'o' &&
9548 return -KEY_getlogin;
9559 if (name[6] == 'n' &&
9562 return -KEY_getpwent;
9568 if (name[6] == 'a' &&
9571 return -KEY_getpwnam;
9577 if (name[6] == 'i' &&
9580 return -KEY_getpwuid;
9600 if (name[1] == 'e' &&
9607 if (name[5] == 'i' &&
9614 return -KEY_readline;
9619 return -KEY_readlink;
9630 if (name[5] == 'i' &&
9634 return -KEY_readpipe;
9655 if (name[4] == 'r' &&
9660 return -KEY_setgrent;
9666 if (name[4] == 'w' &&
9671 return -KEY_setpwent;
9687 if (name[3] == 'w' &&
9693 return -KEY_shmwrite;
9699 if (name[3] == 't' &&
9705 return -KEY_shutdown;
9715 if (name[2] == 's' &&
9722 return -KEY_syswrite;
9732 if (name[1] == 'r' &&
9740 return -KEY_truncate;
9749 case 9: /* 9 tokens of length 9 */
9753 if (name[1] == 'N' &&
9762 return KEY_UNITCHECK;
9768 if (name[1] == 'n' &&
9777 return -KEY_endnetent;
9783 if (name[1] == 'e' &&
9792 return -KEY_getnetent;
9798 if (name[1] == 'o' &&
9807 return -KEY_localtime;
9813 if (name[1] == 'r' &&
9822 return KEY_prototype;
9828 if (name[1] == 'u' &&
9837 return -KEY_quotemeta;
9843 if (name[1] == 'e' &&
9852 return -KEY_rewinddir;
9858 if (name[1] == 'e' &&
9867 return -KEY_setnetent;
9873 if (name[1] == 'a' &&
9882 return -KEY_wantarray;
9891 case 10: /* 9 tokens of length 10 */
9895 if (name[1] == 'n' &&
9901 if (name[4] == 'o' &&
9908 return -KEY_endhostent;
9914 if (name[4] == 'e' &&
9921 return -KEY_endservent;
9934 if (name[1] == 'e' &&
9940 if (name[4] == 'o' &&
9947 return -KEY_gethostent;
9956 if (name[5] == 'r' &&
9962 return -KEY_getservent;
9968 if (name[5] == 'c' &&
9974 return -KEY_getsockopt;
9999 if (name[4] == 'o' &&
10006 return -KEY_sethostent;
10015 if (name[5] == 'r' &&
10021 return -KEY_setservent;
10027 if (name[5] == 'c' &&
10033 return -KEY_setsockopt;
10050 if (name[2] == 'c' &&
10059 return -KEY_socketpair;
10072 case 11: /* 8 tokens of length 11 */
10076 if (name[1] == '_' &&
10086 { /* __PACKAGE__ */
10087 return -KEY___PACKAGE__;
10093 if (name[1] == 'n' &&
10103 { /* endprotoent */
10104 return -KEY_endprotoent;
10110 if (name[1] == 'e' &&
10119 if (name[5] == 'e' &&
10125 { /* getpeername */
10126 return -KEY_getpeername;
10135 if (name[6] == 'o' &&
10140 { /* getpriority */
10141 return -KEY_getpriority;
10147 if (name[6] == 't' &&
10152 { /* getprotoent */
10153 return -KEY_getprotoent;
10167 if (name[4] == 'o' &&
10174 { /* getsockname */
10175 return -KEY_getsockname;
10188 if (name[1] == 'e' &&
10196 if (name[6] == 'o' &&
10201 { /* setpriority */
10202 return -KEY_setpriority;
10208 if (name[6] == 't' &&
10213 { /* setprotoent */
10214 return -KEY_setprotoent;
10230 case 12: /* 2 tokens of length 12 */
10231 if (name[0] == 'g' &&
10243 if (name[9] == 'd' &&
10246 { /* getnetbyaddr */
10247 return -KEY_getnetbyaddr;
10253 if (name[9] == 'a' &&
10256 { /* getnetbyname */
10257 return -KEY_getnetbyname;
10269 case 13: /* 4 tokens of length 13 */
10270 if (name[0] == 'g' &&
10277 if (name[4] == 'o' &&
10286 if (name[10] == 'd' &&
10289 { /* gethostbyaddr */
10290 return -KEY_gethostbyaddr;
10296 if (name[10] == 'a' &&
10299 { /* gethostbyname */
10300 return -KEY_gethostbyname;
10313 if (name[4] == 'e' &&
10322 if (name[10] == 'a' &&
10325 { /* getservbyname */
10326 return -KEY_getservbyname;
10332 if (name[10] == 'o' &&
10335 { /* getservbyport */
10336 return -KEY_getservbyport;
10355 case 14: /* 1 tokens of length 14 */
10356 if (name[0] == 'g' &&
10370 { /* getprotobyname */
10371 return -KEY_getprotobyname;
10376 case 16: /* 1 tokens of length 16 */
10377 if (name[0] == 'g' &&
10393 { /* getprotobynumber */
10394 return -KEY_getprotobynumber;
10408 S_checkcomma(pTHX_ const char *s, const char *name, const char *what)
10412 if (*s == ' ' && s[1] == '(') { /* XXX gotta be a better way */
10413 if (ckWARN(WARN_SYNTAX)) {
10416 for (w = s+2; *w && level; w++) {
10419 else if (*w == ')')
10422 while (isSPACE(*w))
10424 if (!*w || !strchr(";|})]oaiuw!=", *w)) /* an advisory hack only... */
10425 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
10426 "%s (...) interpreted as function",name);
10429 while (s < PL_bufend && isSPACE(*s))
10433 while (s < PL_bufend && isSPACE(*s))
10435 if (isIDFIRST_lazy_if(s,UTF)) {
10436 const char * const w = s++;
10437 while (isALNUM_lazy_if(s,UTF))
10439 while (s < PL_bufend && isSPACE(*s))
10443 if (keyword(w, s - w, 0))
10446 gv = gv_fetchpvn_flags(w, s - w, 0, SVt_PVCV);
10447 if (gv && GvCVu(gv))
10449 Perl_croak(aTHX_ "No comma allowed after %s", what);
10454 /* Either returns sv, or mortalizes sv and returns a new SV*.
10455 Best used as sv=new_constant(..., sv, ...).
10456 If s, pv are NULL, calls subroutine with one argument,
10457 and type is used with error messages only. */
10460 S_new_constant(pTHX_ const char *s, STRLEN len, const char *key, SV *sv, SV *pv,
10464 HV * const table = GvHV(PL_hintgv); /* ^H */
10468 const char *why1 = "", *why2 = "", *why3 = "";
10470 if (!table || !(PL_hints & HINT_LOCALIZE_HH)) {
10473 why2 = (const char *)
10474 (strEQ(key,"charnames")
10475 ? "(possibly a missing \"use charnames ...\")"
10477 msg = Perl_newSVpvf(aTHX_ "Constant(%s) unknown: %s",
10478 (type ? type: "undef"), why2);
10480 /* This is convoluted and evil ("goto considered harmful")
10481 * but I do not understand the intricacies of all the different
10482 * failure modes of %^H in here. The goal here is to make
10483 * the most probable error message user-friendly. --jhi */
10488 msg = Perl_newSVpvf(aTHX_ "Constant(%s): %s%s%s",
10489 (type ? type: "undef"), why1, why2, why3);
10491 yyerror(SvPVX_const(msg));
10495 cvp = hv_fetch(table, key, strlen(key), FALSE);
10496 if (!cvp || !SvOK(*cvp)) {
10499 why3 = "} is not defined";
10502 sv_2mortal(sv); /* Parent created it permanently */
10505 pv = sv_2mortal(newSVpvn(s, len));
10507 typesv = sv_2mortal(newSVpv(type, 0));
10509 typesv = &PL_sv_undef;
10511 PUSHSTACKi(PERLSI_OVERLOAD);
10523 call_sv(cv, G_SCALAR | ( PL_in_eval ? 0 : G_EVAL));
10527 /* Check the eval first */
10528 if (!PL_in_eval && SvTRUE(ERRSV)) {
10529 sv_catpvs(ERRSV, "Propagated");
10530 yyerror(SvPV_nolen_const(ERRSV)); /* Duplicates the message inside eval */
10532 res = SvREFCNT_inc_simple(sv);
10536 SvREFCNT_inc_simple_void(res);
10545 why1 = "Call to &{$^H{";
10547 why3 = "}} did not return a defined value";
10555 /* Returns a NUL terminated string, with the length of the string written to
10559 S_scan_word(pTHX_ register char *s, char *dest, STRLEN destlen, int allow_package, STRLEN *slp)
10562 register char *d = dest;
10563 register char * const e = d + destlen - 3; /* two-character token, ending NUL */
10566 Perl_croak(aTHX_ ident_too_long);
10567 if (isALNUM(*s)) /* UTF handled below */
10569 else if (allow_package && (*s == '\'') && isIDFIRST_lazy_if(s+1,UTF)) {
10574 else if (allow_package && (s[0] == ':') && (s[1] == ':') && (s[2] != '$')) {
10578 else if (UTF && UTF8_IS_START(*s) && isALNUM_utf8((U8*)s)) {
10579 char *t = s + UTF8SKIP(s);
10581 while (UTF8_IS_CONTINUED(*t) && is_utf8_mark((U8*)t))
10585 Perl_croak(aTHX_ ident_too_long);
10586 Copy(s, d, len, char);
10599 S_scan_ident(pTHX_ register char *s, register const char *send, char *dest, STRLEN destlen, I32 ck_uni)
10602 char *bracket = NULL;
10604 register char *d = dest;
10605 register char * const e = d + destlen + 3; /* two-character token, ending NUL */
10610 while (isDIGIT(*s)) {
10612 Perl_croak(aTHX_ ident_too_long);
10619 Perl_croak(aTHX_ ident_too_long);
10620 if (isALNUM(*s)) /* UTF handled below */
10622 else if (*s == '\'' && isIDFIRST_lazy_if(s+1,UTF)) {
10627 else if (*s == ':' && s[1] == ':') {
10631 else if (UTF && UTF8_IS_START(*s) && isALNUM_utf8((U8*)s)) {
10632 char *t = s + UTF8SKIP(s);
10633 while (UTF8_IS_CONTINUED(*t) && is_utf8_mark((U8*)t))
10635 if (d + (t - s) > e)
10636 Perl_croak(aTHX_ ident_too_long);
10637 Copy(s, d, t - s, char);
10648 if (PL_lex_state != LEX_NORMAL)
10649 PL_lex_state = LEX_INTERPENDMAYBE;
10652 if (*s == '$' && s[1] &&
10653 (isALNUM_lazy_if(s+1,UTF) || s[1] == '$' || s[1] == '{' || strnEQ(s+1,"::",2)) )
10666 if (*d == '^' && *s && isCONTROLVAR(*s)) {
10671 if (isSPACE(s[-1])) {
10673 const char ch = *s++;
10674 if (!SPACE_OR_TAB(ch)) {
10680 if (isIDFIRST_lazy_if(d,UTF)) {
10684 while ((end < send && isALNUM_lazy_if(end,UTF)) || *end == ':') {
10685 end += UTF8SKIP(end);
10686 while (end < send && UTF8_IS_CONTINUED(*end) && is_utf8_mark((U8*)end))
10687 end += UTF8SKIP(end);
10689 Copy(s, d, end - s, char);
10694 while ((isALNUM(*s) || *s == ':') && d < e)
10697 Perl_croak(aTHX_ ident_too_long);
10700 while (s < send && SPACE_OR_TAB(*s))
10702 if ((*s == '[' || (*s == '{' && strNE(dest, "sub")))) {
10703 if (ckWARN(WARN_AMBIGUOUS) && keyword(dest, d - dest, 0)) {
10704 const char * const brack =
10706 ((*s == '[') ? "[...]" : "{...}");
10707 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
10708 "Ambiguous use of %c{%s%s} resolved to %c%s%s",
10709 funny, dest, brack, funny, dest, brack);
10712 PL_lex_brackstack[PL_lex_brackets++] = (char)(XOPERATOR | XFAKEBRACK);
10716 /* Handle extended ${^Foo} variables
10717 * 1999-02-27 mjd-perl-patch@plover.com */
10718 else if (!isALNUM(*d) && !isPRINT(*d) /* isCTRL(d) */
10722 while (isALNUM(*s) && d < e) {
10726 Perl_croak(aTHX_ ident_too_long);
10731 if (PL_lex_state == LEX_INTERPNORMAL && !PL_lex_brackets) {
10732 PL_lex_state = LEX_INTERPEND;
10735 if (PL_lex_state == LEX_NORMAL) {
10736 if (ckWARN(WARN_AMBIGUOUS) &&
10737 (keyword(dest, d - dest, 0)
10738 || get_cvn_flags(dest, d - dest, 0)))
10742 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
10743 "Ambiguous use of %c{%s} resolved to %c%s",
10744 funny, dest, funny, dest);
10749 s = bracket; /* let the parser handle it */
10753 else if (PL_lex_state == LEX_INTERPNORMAL && !PL_lex_brackets && !intuit_more(s))
10754 PL_lex_state = LEX_INTERPEND;
10759 Perl_pmflag(pTHX_ U32* pmfl, int ch)
10761 PERL_UNUSED_CONTEXT;
10765 CASE_STD_PMMOD_FLAGS_PARSE_SET(pmfl);
10766 case GLOBAL_PAT_MOD: *pmfl |= PMf_GLOBAL; break;
10767 case CONTINUE_PAT_MOD: *pmfl |= PMf_CONTINUE; break;
10768 case ONCE_PAT_MOD: *pmfl |= PMf_KEEP; break;
10769 case KEEPCOPY_PAT_MOD: *pmfl |= PMf_KEEPCOPY; break;
10775 S_scan_pat(pTHX_ char *start, I32 type)
10779 char *s = scan_str(start,!!PL_madskills,FALSE);
10780 const char * const valid_flags =
10781 (const char *)((type == OP_QR) ? QR_PAT_MODS : M_PAT_MODS);
10788 const char * const delimiter = skipspace(start);
10792 ? "Search pattern not terminated or ternary operator parsed as search pattern"
10793 : "Search pattern not terminated" ));
10796 pm = (PMOP*)newPMOP(type, 0);
10797 if (PL_multi_open == '?')
10798 pm->op_pmflags |= PMf_ONCE;
10802 while (*s && strchr(valid_flags, *s))
10803 pmflag(&pm->op_pmflags,*s++);
10805 if (PL_madskills && modstart != s) {
10806 SV* tmptoken = newSVpvn(modstart, s - modstart);
10807 append_madprops(newMADPROP('m', MAD_SV, tmptoken, 0), (OP*)pm, 0);
10810 /* issue a warning if /c is specified,but /g is not */
10811 if ((pm->op_pmflags & PMf_CONTINUE) && !(pm->op_pmflags & PMf_GLOBAL)
10812 && ckWARN(WARN_REGEXP))
10814 Perl_warner(aTHX_ packWARN(WARN_REGEXP),
10815 "Use of /c modifier is meaningless without /g" );
10818 pm->op_pmpermflags = pm->op_pmflags;
10820 PL_lex_op = (OP*)pm;
10821 yylval.ival = OP_MATCH;
10826 S_scan_subst(pTHX_ char *start)
10837 yylval.ival = OP_NULL;
10839 s = scan_str(start,!!PL_madskills,FALSE);
10842 Perl_croak(aTHX_ "Substitution pattern not terminated");
10844 if (s[-1] == PL_multi_open)
10847 if (PL_madskills) {
10848 CURMAD('q', PL_thisopen);
10849 CURMAD('_', PL_thiswhite);
10850 CURMAD('E', PL_thisstuff);
10851 CURMAD('Q', PL_thisclose);
10852 PL_realtokenstart = s - SvPVX(PL_linestr);
10856 first_start = PL_multi_start;
10857 s = scan_str(s,!!PL_madskills,FALSE);
10859 if (PL_lex_stuff) {
10860 SvREFCNT_dec(PL_lex_stuff);
10861 PL_lex_stuff = NULL;
10863 Perl_croak(aTHX_ "Substitution replacement not terminated");
10865 PL_multi_start = first_start; /* so whole substitution is taken together */
10867 pm = (PMOP*)newPMOP(OP_SUBST, 0);
10870 if (PL_madskills) {
10871 CURMAD('z', PL_thisopen);
10872 CURMAD('R', PL_thisstuff);
10873 CURMAD('Z', PL_thisclose);
10879 if (*s == EXEC_PAT_MOD) {
10883 else if (strchr(S_PAT_MODS, *s))
10884 pmflag(&pm->op_pmflags,*s++);
10890 if (PL_madskills) {
10892 curmad('m', newSVpvn(modstart, s - modstart));
10893 append_madprops(PL_thismad, (OP*)pm, 0);
10897 if ((pm->op_pmflags & PMf_CONTINUE) && ckWARN(WARN_REGEXP)) {
10898 Perl_warner(aTHX_ packWARN(WARN_REGEXP), "Use of /c modifier is meaningless in s///" );
10902 SV * const repl = newSVpvs("");
10904 PL_sublex_info.super_bufptr = s;
10905 PL_sublex_info.super_bufend = PL_bufend;
10907 pm->op_pmflags |= PMf_EVAL;
10909 sv_catpv(repl, (const char *)(es ? "eval " : "do "));
10910 sv_catpvs(repl, "{");
10911 sv_catsv(repl, PL_lex_repl);
10912 if (strchr(SvPVX(PL_lex_repl), '#'))
10913 sv_catpvs(repl, "\n");
10914 sv_catpvs(repl, "}");
10916 SvREFCNT_dec(PL_lex_repl);
10917 PL_lex_repl = repl;
10920 pm->op_pmpermflags = pm->op_pmflags;
10921 PL_lex_op = (OP*)pm;
10922 yylval.ival = OP_SUBST;
10927 S_scan_trans(pTHX_ char *start)
10940 yylval.ival = OP_NULL;
10942 s = scan_str(start,!!PL_madskills,FALSE);
10944 Perl_croak(aTHX_ "Transliteration pattern not terminated");
10946 if (s[-1] == PL_multi_open)
10949 if (PL_madskills) {
10950 CURMAD('q', PL_thisopen);
10951 CURMAD('_', PL_thiswhite);
10952 CURMAD('E', PL_thisstuff);
10953 CURMAD('Q', PL_thisclose);
10954 PL_realtokenstart = s - SvPVX(PL_linestr);
10958 s = scan_str(s,!!PL_madskills,FALSE);
10960 if (PL_lex_stuff) {
10961 SvREFCNT_dec(PL_lex_stuff);
10962 PL_lex_stuff = NULL;
10964 Perl_croak(aTHX_ "Transliteration replacement not terminated");
10966 if (PL_madskills) {
10967 CURMAD('z', PL_thisopen);
10968 CURMAD('R', PL_thisstuff);
10969 CURMAD('Z', PL_thisclose);
10972 complement = del = squash = 0;
10979 complement = OPpTRANS_COMPLEMENT;
10982 del = OPpTRANS_DELETE;
10985 squash = OPpTRANS_SQUASH;
10994 tbl = (short *)PerlMemShared_calloc(complement&&!del?258:256, sizeof(short));
10995 o = newPVOP(OP_TRANS, 0, (char*)tbl);
10996 o->op_private &= ~OPpTRANS_ALL;
10997 o->op_private |= del|squash|complement|
10998 (DO_UTF8(PL_lex_stuff)? OPpTRANS_FROM_UTF : 0)|
10999 (DO_UTF8(PL_lex_repl) ? OPpTRANS_TO_UTF : 0);
11002 yylval.ival = OP_TRANS;
11005 if (PL_madskills) {
11007 curmad('m', newSVpvn(modstart, s - modstart));
11008 append_madprops(PL_thismad, o, 0);
11017 S_scan_heredoc(pTHX_ register char *s)
11021 I32 op_type = OP_SCALAR;
11025 const char *found_newline;
11029 const int outer = (PL_rsfp && !(PL_lex_inwhat == OP_SCALAR));
11031 I32 stuffstart = s - SvPVX(PL_linestr);
11034 PL_realtokenstart = -1;
11039 e = PL_tokenbuf + sizeof PL_tokenbuf - 1;
11043 while (SPACE_OR_TAB(*peek))
11045 if (*peek == '`' || *peek == '\'' || *peek =='"') {
11048 s = delimcpy(d, e, s, PL_bufend, term, &len);
11058 if (!isALNUM_lazy_if(s,UTF))
11059 deprecate_old("bare << to mean <<\"\"");
11060 for (; isALNUM_lazy_if(s,UTF); s++) {
11065 if (d >= PL_tokenbuf + sizeof PL_tokenbuf - 1)
11066 Perl_croak(aTHX_ "Delimiter for here document is too long");
11069 len = d - PL_tokenbuf;
11072 if (PL_madskills) {
11073 tstart = PL_tokenbuf + !outer;
11074 PL_thisclose = newSVpvn(tstart, len - !outer);
11075 tstart = SvPVX(PL_linestr) + stuffstart;
11076 PL_thisopen = newSVpvn(tstart, s - tstart);
11077 stuffstart = s - SvPVX(PL_linestr);
11080 #ifndef PERL_STRICT_CR
11081 d = strchr(s, '\r');
11083 char * const olds = s;
11085 while (s < PL_bufend) {
11091 else if (*s == '\n' && s[1] == '\r') { /* \015\013 on a mac? */
11100 SvCUR_set(PL_linestr, PL_bufend - SvPVX_const(PL_linestr));
11107 if ( outer || !(found_newline = (char*)memchr((void*)s, '\n', PL_bufend - s)) ) {
11108 herewas = newSVpvn(s,PL_bufend-s);
11112 herewas = newSVpvn(s-1,found_newline-s+1);
11115 herewas = newSVpvn(s,found_newline-s);
11119 if (PL_madskills) {
11120 tstart = SvPVX(PL_linestr) + stuffstart;
11122 sv_catpvn(PL_thisstuff, tstart, s - tstart);
11124 PL_thisstuff = newSVpvn(tstart, s - tstart);
11127 s += SvCUR(herewas);
11130 stuffstart = s - SvPVX(PL_linestr);
11136 tmpstr = newSV(79);
11137 sv_upgrade(tmpstr, SVt_PVIV);
11138 if (term == '\'') {
11139 op_type = OP_CONST;
11140 SvIV_set(tmpstr, -1);
11142 else if (term == '`') {
11143 op_type = OP_BACKTICK;
11144 SvIV_set(tmpstr, '\\');
11148 PL_multi_start = CopLINE(PL_curcop);
11149 PL_multi_open = PL_multi_close = '<';
11150 term = *PL_tokenbuf;
11151 if (PL_lex_inwhat == OP_SUBST && PL_in_eval && !PL_rsfp) {
11152 char * const bufptr = PL_sublex_info.super_bufptr;
11153 char * const bufend = PL_sublex_info.super_bufend;
11154 char * const olds = s - SvCUR(herewas);
11155 s = strchr(bufptr, '\n');
11159 while (s < bufend &&
11160 (*s != term || memNE(s,PL_tokenbuf,len)) ) {
11162 CopLINE_inc(PL_curcop);
11165 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
11166 missingterm(PL_tokenbuf);
11168 sv_setpvn(herewas,bufptr,d-bufptr+1);
11169 sv_setpvn(tmpstr,d+1,s-d);
11171 sv_catpvn(herewas,s,bufend-s);
11172 Copy(SvPVX_const(herewas),bufptr,SvCUR(herewas) + 1,char);
11179 while (s < PL_bufend &&
11180 (*s != term || memNE(s,PL_tokenbuf,len)) ) {
11182 CopLINE_inc(PL_curcop);
11184 if (s >= PL_bufend) {
11185 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
11186 missingterm(PL_tokenbuf);
11188 sv_setpvn(tmpstr,d+1,s-d);
11190 if (PL_madskills) {
11192 sv_catpvn(PL_thisstuff, d + 1, s - d);
11194 PL_thisstuff = newSVpvn(d + 1, s - d);
11195 stuffstart = s - SvPVX(PL_linestr);
11199 CopLINE_inc(PL_curcop); /* the preceding stmt passes a newline */
11201 sv_catpvn(herewas,s,PL_bufend-s);
11202 sv_setsv(PL_linestr,herewas);
11203 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = s = PL_linestart = SvPVX(PL_linestr);
11204 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
11205 PL_last_lop = PL_last_uni = NULL;
11208 sv_setpvn(tmpstr,"",0); /* avoid "uninitialized" warning */
11209 while (s >= PL_bufend) { /* multiple line string? */
11211 if (PL_madskills) {
11212 tstart = SvPVX(PL_linestr) + stuffstart;
11214 sv_catpvn(PL_thisstuff, tstart, PL_bufend - tstart);
11216 PL_thisstuff = newSVpvn(tstart, PL_bufend - tstart);
11220 !(PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = filter_gets(PL_linestr, PL_rsfp, 0))) {
11221 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
11222 missingterm(PL_tokenbuf);
11225 stuffstart = s - SvPVX(PL_linestr);
11227 CopLINE_inc(PL_curcop);
11228 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
11229 PL_last_lop = PL_last_uni = NULL;
11230 #ifndef PERL_STRICT_CR
11231 if (PL_bufend - PL_linestart >= 2) {
11232 if ((PL_bufend[-2] == '\r' && PL_bufend[-1] == '\n') ||
11233 (PL_bufend[-2] == '\n' && PL_bufend[-1] == '\r'))
11235 PL_bufend[-2] = '\n';
11237 SvCUR_set(PL_linestr, PL_bufend - SvPVX_const(PL_linestr));
11239 else if (PL_bufend[-1] == '\r')
11240 PL_bufend[-1] = '\n';
11242 else if (PL_bufend - PL_linestart == 1 && PL_bufend[-1] == '\r')
11243 PL_bufend[-1] = '\n';
11245 if (PERLDB_LINE && PL_curstash != PL_debstash)
11246 update_debugger_info(PL_linestr, NULL, 0);
11247 if (*s == term && memEQ(s,PL_tokenbuf,len)) {
11248 STRLEN off = PL_bufend - 1 - SvPVX_const(PL_linestr);
11249 *(SvPVX(PL_linestr) + off ) = ' ';
11250 sv_catsv(PL_linestr,herewas);
11251 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
11252 s = SvPVX(PL_linestr) + off; /* In case PV of PL_linestr moved. */
11256 sv_catsv(tmpstr,PL_linestr);
11261 PL_multi_end = CopLINE(PL_curcop);
11262 if (SvCUR(tmpstr) + 5 < SvLEN(tmpstr)) {
11263 SvPV_shrink_to_cur(tmpstr);
11265 SvREFCNT_dec(herewas);
11267 if (UTF && is_utf8_string((U8*)SvPVX_const(tmpstr), SvCUR(tmpstr)))
11269 else if (PL_encoding)
11270 sv_recode_to_utf8(tmpstr, PL_encoding);
11272 PL_lex_stuff = tmpstr;
11273 yylval.ival = op_type;
11277 /* scan_inputsymbol
11278 takes: current position in input buffer
11279 returns: new position in input buffer
11280 side-effects: yylval and lex_op are set.
11285 <FH> read from filehandle
11286 <pkg::FH> read from package qualified filehandle
11287 <pkg'FH> read from package qualified filehandle
11288 <$fh> read from filehandle in $fh
11289 <*.h> filename glob
11294 S_scan_inputsymbol(pTHX_ char *start)
11297 register char *s = start; /* current position in buffer */
11301 char *d = PL_tokenbuf; /* start of temp holding space */
11302 const char * const e = PL_tokenbuf + sizeof PL_tokenbuf; /* end of temp holding space */
11304 end = strchr(s, '\n');
11307 s = delimcpy(d, e, s + 1, end, '>', &len); /* extract until > */
11309 /* die if we didn't have space for the contents of the <>,
11310 or if it didn't end, or if we see a newline
11313 if (len >= (I32)sizeof PL_tokenbuf)
11314 Perl_croak(aTHX_ "Excessively long <> operator");
11316 Perl_croak(aTHX_ "Unterminated <> operator");
11321 Remember, only scalar variables are interpreted as filehandles by
11322 this code. Anything more complex (e.g., <$fh{$num}>) will be
11323 treated as a glob() call.
11324 This code makes use of the fact that except for the $ at the front,
11325 a scalar variable and a filehandle look the same.
11327 if (*d == '$' && d[1]) d++;
11329 /* allow <Pkg'VALUE> or <Pkg::VALUE> */
11330 while (*d && (isALNUM_lazy_if(d,UTF) || *d == '\'' || *d == ':'))
11333 /* If we've tried to read what we allow filehandles to look like, and
11334 there's still text left, then it must be a glob() and not a getline.
11335 Use scan_str to pull out the stuff between the <> and treat it
11336 as nothing more than a string.
11339 if (d - PL_tokenbuf != len) {
11340 yylval.ival = OP_GLOB;
11342 s = scan_str(start,!!PL_madskills,FALSE);
11344 Perl_croak(aTHX_ "Glob not terminated");
11348 bool readline_overriden = FALSE;
11351 /* we're in a filehandle read situation */
11354 /* turn <> into <ARGV> */
11356 Copy("ARGV",d,5,char);
11358 /* Check whether readline() is overriden */
11359 gv_readline = gv_fetchpvs("readline", GV_NOTQUAL, SVt_PVCV);
11361 && GvCVu(gv_readline) && GvIMPORTED_CV(gv_readline))
11363 ((gvp = (GV**)hv_fetchs(PL_globalstash, "readline", FALSE))
11364 && (gv_readline = *gvp) != (GV*)&PL_sv_undef
11365 && GvCVu(gv_readline) && GvIMPORTED_CV(gv_readline)))
11366 readline_overriden = TRUE;
11368 /* if <$fh>, create the ops to turn the variable into a
11372 /* try to find it in the pad for this block, otherwise find
11373 add symbol table ops
11375 const PADOFFSET tmp = pad_findmy(d);
11376 if (tmp != NOT_IN_PAD) {
11377 if (PAD_COMPNAME_FLAGS_isOUR(tmp)) {
11378 HV * const stash = PAD_COMPNAME_OURSTASH(tmp);
11379 HEK * const stashname = HvNAME_HEK(stash);
11380 SV * const sym = sv_2mortal(newSVhek(stashname));
11381 sv_catpvs(sym, "::");
11382 sv_catpv(sym, d+1);
11387 OP * const o = newOP(OP_PADSV, 0);
11389 PL_lex_op = readline_overriden
11390 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
11391 append_elem(OP_LIST, o,
11392 newCVREF(0, newGVOP(OP_GV,0,gv_readline))))
11393 : (OP*)newUNOP(OP_READLINE, 0, o);
11402 ? (GV_ADDMULTI | GV_ADDINEVAL)
11405 PL_lex_op = readline_overriden
11406 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
11407 append_elem(OP_LIST,
11408 newUNOP(OP_RV2SV, 0, newGVOP(OP_GV, 0, gv)),
11409 newCVREF(0, newGVOP(OP_GV, 0, gv_readline))))
11410 : (OP*)newUNOP(OP_READLINE, 0,
11411 newUNOP(OP_RV2SV, 0,
11412 newGVOP(OP_GV, 0, gv)));
11414 if (!readline_overriden)
11415 PL_lex_op->op_flags |= OPf_SPECIAL;
11416 /* we created the ops in PL_lex_op, so make yylval.ival a null op */
11417 yylval.ival = OP_NULL;
11420 /* If it's none of the above, it must be a literal filehandle
11421 (<Foo::BAR> or <FOO>) so build a simple readline OP */
11423 GV * const gv = gv_fetchpv(d, GV_ADD, SVt_PVIO);
11424 PL_lex_op = readline_overriden
11425 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
11426 append_elem(OP_LIST,
11427 newGVOP(OP_GV, 0, gv),
11428 newCVREF(0, newGVOP(OP_GV, 0, gv_readline))))
11429 : (OP*)newUNOP(OP_READLINE, 0, newGVOP(OP_GV, 0, gv));
11430 yylval.ival = OP_NULL;
11439 takes: start position in buffer
11440 keep_quoted preserve \ on the embedded delimiter(s)
11441 keep_delims preserve the delimiters around the string
11442 returns: position to continue reading from buffer
11443 side-effects: multi_start, multi_close, lex_repl or lex_stuff, and
11444 updates the read buffer.
11446 This subroutine pulls a string out of the input. It is called for:
11447 q single quotes q(literal text)
11448 ' single quotes 'literal text'
11449 qq double quotes qq(interpolate $here please)
11450 " double quotes "interpolate $here please"
11451 qx backticks qx(/bin/ls -l)
11452 ` backticks `/bin/ls -l`
11453 qw quote words @EXPORT_OK = qw( func() $spam )
11454 m// regexp match m/this/
11455 s/// regexp substitute s/this/that/
11456 tr/// string transliterate tr/this/that/
11457 y/// string transliterate y/this/that/
11458 ($*@) sub prototypes sub foo ($)
11459 (stuff) sub attr parameters sub foo : attr(stuff)
11460 <> readline or globs <FOO>, <>, <$fh>, or <*.c>
11462 In most of these cases (all but <>, patterns and transliterate)
11463 yylex() calls scan_str(). m// makes yylex() call scan_pat() which
11464 calls scan_str(). s/// makes yylex() call scan_subst() which calls
11465 scan_str(). tr/// and y/// make yylex() call scan_trans() which
11468 It skips whitespace before the string starts, and treats the first
11469 character as the delimiter. If the delimiter is one of ([{< then
11470 the corresponding "close" character )]}> is used as the closing
11471 delimiter. It allows quoting of delimiters, and if the string has
11472 balanced delimiters ([{<>}]) it allows nesting.
11474 On success, the SV with the resulting string is put into lex_stuff or,
11475 if that is already non-NULL, into lex_repl. The second case occurs only
11476 when parsing the RHS of the special constructs s/// and tr/// (y///).
11477 For convenience, the terminating delimiter character is stuffed into
11482 S_scan_str(pTHX_ char *start, int keep_quoted, int keep_delims)
11485 SV *sv; /* scalar value: string */
11486 const char *tmps; /* temp string, used for delimiter matching */
11487 register char *s = start; /* current position in the buffer */
11488 register char term; /* terminating character */
11489 register char *to; /* current position in the sv's data */
11490 I32 brackets = 1; /* bracket nesting level */
11491 bool has_utf8 = FALSE; /* is there any utf8 content? */
11492 I32 termcode; /* terminating char. code */
11493 U8 termstr[UTF8_MAXBYTES]; /* terminating string */
11494 STRLEN termlen; /* length of terminating string */
11495 int last_off = 0; /* last position for nesting bracket */
11501 /* skip space before the delimiter */
11507 if (PL_realtokenstart >= 0) {
11508 stuffstart = PL_realtokenstart;
11509 PL_realtokenstart = -1;
11512 stuffstart = start - SvPVX(PL_linestr);
11514 /* mark where we are, in case we need to report errors */
11517 /* after skipping whitespace, the next character is the terminator */
11520 termcode = termstr[0] = term;
11524 termcode = utf8_to_uvchr((U8*)s, &termlen);
11525 Copy(s, termstr, termlen, U8);
11526 if (!UTF8_IS_INVARIANT(term))
11530 /* mark where we are */
11531 PL_multi_start = CopLINE(PL_curcop);
11532 PL_multi_open = term;
11534 /* find corresponding closing delimiter */
11535 if (term && (tmps = strchr("([{< )]}> )]}>",term)))
11536 termcode = termstr[0] = term = tmps[5];
11538 PL_multi_close = term;
11540 /* create a new SV to hold the contents. 79 is the SV's initial length.
11541 What a random number. */
11543 sv_upgrade(sv, SVt_PVIV);
11544 SvIV_set(sv, termcode);
11545 (void)SvPOK_only(sv); /* validate pointer */
11547 /* move past delimiter and try to read a complete string */
11549 sv_catpvn(sv, s, termlen);
11552 tstart = SvPVX(PL_linestr) + stuffstart;
11553 if (!PL_thisopen && !keep_delims) {
11554 PL_thisopen = newSVpvn(tstart, s - tstart);
11555 stuffstart = s - SvPVX(PL_linestr);
11559 if (PL_encoding && !UTF) {
11563 int offset = s - SvPVX_const(PL_linestr);
11564 const bool found = sv_cat_decode(sv, PL_encoding, PL_linestr,
11565 &offset, (char*)termstr, termlen);
11566 const char * const ns = SvPVX_const(PL_linestr) + offset;
11567 char * const svlast = SvEND(sv) - 1;
11569 for (; s < ns; s++) {
11570 if (*s == '\n' && !PL_rsfp)
11571 CopLINE_inc(PL_curcop);
11574 goto read_more_line;
11576 /* handle quoted delimiters */
11577 if (SvCUR(sv) > 1 && *(svlast-1) == '\\') {
11579 for (t = svlast-2; t >= SvPVX_const(sv) && *t == '\\';)
11581 if ((svlast-1 - t) % 2) {
11582 if (!keep_quoted) {
11583 *(svlast-1) = term;
11585 SvCUR_set(sv, SvCUR(sv) - 1);
11590 if (PL_multi_open == PL_multi_close) {
11596 for (t = w = SvPVX(sv)+last_off; t < svlast; w++, t++) {
11597 /* At here, all closes are "was quoted" one,
11598 so we don't check PL_multi_close. */
11600 if (!keep_quoted && *(t+1) == PL_multi_open)
11605 else if (*t == PL_multi_open)
11613 SvCUR_set(sv, w - SvPVX_const(sv));
11615 last_off = w - SvPVX(sv);
11616 if (--brackets <= 0)
11621 if (!keep_delims) {
11622 SvCUR_set(sv, SvCUR(sv) - 1);
11628 /* extend sv if need be */
11629 SvGROW(sv, SvCUR(sv) + (PL_bufend - s) + 1);
11630 /* set 'to' to the next character in the sv's string */
11631 to = SvPVX(sv)+SvCUR(sv);
11633 /* if open delimiter is the close delimiter read unbridle */
11634 if (PL_multi_open == PL_multi_close) {
11635 for (; s < PL_bufend; s++,to++) {
11636 /* embedded newlines increment the current line number */
11637 if (*s == '\n' && !PL_rsfp)
11638 CopLINE_inc(PL_curcop);
11639 /* handle quoted delimiters */
11640 if (*s == '\\' && s+1 < PL_bufend && term != '\\') {
11641 if (!keep_quoted && s[1] == term)
11643 /* any other quotes are simply copied straight through */
11647 /* terminate when run out of buffer (the for() condition), or
11648 have found the terminator */
11649 else if (*s == term) {
11652 if (s+termlen <= PL_bufend && memEQ(s, (char*)termstr, termlen))
11655 else if (!has_utf8 && !UTF8_IS_INVARIANT((U8)*s) && UTF)
11661 /* if the terminator isn't the same as the start character (e.g.,
11662 matched brackets), we have to allow more in the quoting, and
11663 be prepared for nested brackets.
11666 /* read until we run out of string, or we find the terminator */
11667 for (; s < PL_bufend; s++,to++) {
11668 /* embedded newlines increment the line count */
11669 if (*s == '\n' && !PL_rsfp)
11670 CopLINE_inc(PL_curcop);
11671 /* backslashes can escape the open or closing characters */
11672 if (*s == '\\' && s+1 < PL_bufend) {
11673 if (!keep_quoted &&
11674 ((s[1] == PL_multi_open) || (s[1] == PL_multi_close)))
11679 /* allow nested opens and closes */
11680 else if (*s == PL_multi_close && --brackets <= 0)
11682 else if (*s == PL_multi_open)
11684 else if (!has_utf8 && !UTF8_IS_INVARIANT((U8)*s) && UTF)
11689 /* terminate the copied string and update the sv's end-of-string */
11691 SvCUR_set(sv, to - SvPVX_const(sv));
11694 * this next chunk reads more into the buffer if we're not done yet
11698 break; /* handle case where we are done yet :-) */
11700 #ifndef PERL_STRICT_CR
11701 if (to - SvPVX_const(sv) >= 2) {
11702 if ((to[-2] == '\r' && to[-1] == '\n') ||
11703 (to[-2] == '\n' && to[-1] == '\r'))
11707 SvCUR_set(sv, to - SvPVX_const(sv));
11709 else if (to[-1] == '\r')
11712 else if (to - SvPVX_const(sv) == 1 && to[-1] == '\r')
11717 /* if we're out of file, or a read fails, bail and reset the current
11718 line marker so we can report where the unterminated string began
11721 if (PL_madskills) {
11722 char * const tstart = SvPVX(PL_linestr) + stuffstart;
11724 sv_catpvn(PL_thisstuff, tstart, PL_bufend - tstart);
11726 PL_thisstuff = newSVpvn(tstart, PL_bufend - tstart);
11730 !(PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = filter_gets(PL_linestr, PL_rsfp, 0))) {
11732 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
11738 /* we read a line, so increment our line counter */
11739 CopLINE_inc(PL_curcop);
11741 /* update debugger info */
11742 if (PERLDB_LINE && PL_curstash != PL_debstash)
11743 update_debugger_info(PL_linestr, NULL, 0);
11745 /* having changed the buffer, we must update PL_bufend */
11746 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
11747 PL_last_lop = PL_last_uni = NULL;
11750 /* at this point, we have successfully read the delimited string */
11752 if (!PL_encoding || UTF) {
11754 if (PL_madskills) {
11755 char * const tstart = SvPVX(PL_linestr) + stuffstart;
11756 const int len = s - tstart;
11758 sv_catpvn(PL_thisstuff, tstart, len);
11760 PL_thisstuff = newSVpvn(tstart, len);
11761 if (!PL_thisclose && !keep_delims)
11762 PL_thisclose = newSVpvn(s,termlen);
11767 sv_catpvn(sv, s, termlen);
11772 if (PL_madskills) {
11773 char * const tstart = SvPVX(PL_linestr) + stuffstart;
11774 const int len = s - tstart - termlen;
11776 sv_catpvn(PL_thisstuff, tstart, len);
11778 PL_thisstuff = newSVpvn(tstart, len);
11779 if (!PL_thisclose && !keep_delims)
11780 PL_thisclose = newSVpvn(s - termlen,termlen);
11784 if (has_utf8 || PL_encoding)
11787 PL_multi_end = CopLINE(PL_curcop);
11789 /* if we allocated too much space, give some back */
11790 if (SvCUR(sv) + 5 < SvLEN(sv)) {
11791 SvLEN_set(sv, SvCUR(sv) + 1);
11792 SvPV_renew(sv, SvLEN(sv));
11795 /* decide whether this is the first or second quoted string we've read
11808 takes: pointer to position in buffer
11809 returns: pointer to new position in buffer
11810 side-effects: builds ops for the constant in yylval.op
11812 Read a number in any of the formats that Perl accepts:
11814 \d(_?\d)*(\.(\d(_?\d)*)?)?[Ee][\+\-]?(\d(_?\d)*) 12 12.34 12.
11815 \.\d(_?\d)*[Ee][\+\-]?(\d(_?\d)*) .34
11818 0x[0-9A-Fa-f](_?[0-9A-Fa-f])*
11820 Like most scan_ routines, it uses the PL_tokenbuf buffer to hold the
11823 If it reads a number without a decimal point or an exponent, it will
11824 try converting the number to an integer and see if it can do so
11825 without loss of precision.
11829 Perl_scan_num(pTHX_ const char *start, YYSTYPE* lvalp)
11832 register const char *s = start; /* current position in buffer */
11833 register char *d; /* destination in temp buffer */
11834 register char *e; /* end of temp buffer */
11835 NV nv; /* number read, as a double */
11836 SV *sv = NULL; /* place to put the converted number */
11837 bool floatit; /* boolean: int or float? */
11838 const char *lastub = NULL; /* position of last underbar */
11839 static char const number_too_long[] = "Number too long";
11841 /* We use the first character to decide what type of number this is */
11845 Perl_croak(aTHX_ "panic: scan_num");
11847 /* if it starts with a 0, it could be an octal number, a decimal in
11848 0.13 disguise, or a hexadecimal number, or a binary number. */
11852 u holds the "number so far"
11853 shift the power of 2 of the base
11854 (hex == 4, octal == 3, binary == 1)
11855 overflowed was the number more than we can hold?
11857 Shift is used when we add a digit. It also serves as an "are
11858 we in octal/hex/binary?" indicator to disallow hex characters
11859 when in octal mode.
11864 bool overflowed = FALSE;
11865 bool just_zero = TRUE; /* just plain 0 or binary number? */
11866 static const NV nvshift[5] = { 1.0, 2.0, 4.0, 8.0, 16.0 };
11867 static const char* const bases[5] =
11868 { "", "binary", "", "octal", "hexadecimal" };
11869 static const char* const Bases[5] =
11870 { "", "Binary", "", "Octal", "Hexadecimal" };
11871 static const char* const maxima[5] =
11873 "0b11111111111111111111111111111111",
11877 const char *base, *Base, *max;
11879 /* check for hex */
11884 } else if (s[1] == 'b') {
11889 /* check for a decimal in disguise */
11890 else if (s[1] == '.' || s[1] == 'e' || s[1] == 'E')
11892 /* so it must be octal */
11899 if (ckWARN(WARN_SYNTAX))
11900 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
11901 "Misplaced _ in number");
11905 base = bases[shift];
11906 Base = Bases[shift];
11907 max = maxima[shift];
11909 /* read the rest of the number */
11911 /* x is used in the overflow test,
11912 b is the digit we're adding on. */
11917 /* if we don't mention it, we're done */
11921 /* _ are ignored -- but warned about if consecutive */
11923 if (lastub && s == lastub + 1 && ckWARN(WARN_SYNTAX))
11924 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
11925 "Misplaced _ in number");
11929 /* 8 and 9 are not octal */
11930 case '8': case '9':
11932 yyerror(Perl_form(aTHX_ "Illegal octal digit '%c'", *s));
11936 case '2': case '3': case '4':
11937 case '5': case '6': case '7':
11939 yyerror(Perl_form(aTHX_ "Illegal binary digit '%c'", *s));
11942 case '0': case '1':
11943 b = *s++ & 15; /* ASCII digit -> value of digit */
11947 case 'a': case 'b': case 'c': case 'd': case 'e': case 'f':
11948 case 'A': case 'B': case 'C': case 'D': case 'E': case 'F':
11949 /* make sure they said 0x */
11952 b = (*s++ & 7) + 9;
11954 /* Prepare to put the digit we have onto the end
11955 of the number so far. We check for overflows.
11961 x = u << shift; /* make room for the digit */
11963 if ((x >> shift) != u
11964 && !(PL_hints & HINT_NEW_BINARY)) {
11967 if (ckWARN_d(WARN_OVERFLOW))
11968 Perl_warner(aTHX_ packWARN(WARN_OVERFLOW),
11969 "Integer overflow in %s number",
11972 u = x | b; /* add the digit to the end */
11975 n *= nvshift[shift];
11976 /* If an NV has not enough bits in its
11977 * mantissa to represent an UV this summing of
11978 * small low-order numbers is a waste of time
11979 * (because the NV cannot preserve the
11980 * low-order bits anyway): we could just
11981 * remember when did we overflow and in the
11982 * end just multiply n by the right
11990 /* if we get here, we had success: make a scalar value from
11995 /* final misplaced underbar check */
11996 if (s[-1] == '_') {
11997 if (ckWARN(WARN_SYNTAX))
11998 Perl_warner(aTHX_ packWARN(WARN_SYNTAX), "Misplaced _ in number");
12003 if (n > 4294967295.0 && ckWARN(WARN_PORTABLE))
12004 Perl_warner(aTHX_ packWARN(WARN_PORTABLE),
12005 "%s number > %s non-portable",
12011 if (u > 0xffffffff && ckWARN(WARN_PORTABLE))
12012 Perl_warner(aTHX_ packWARN(WARN_PORTABLE),
12013 "%s number > %s non-portable",
12018 if (just_zero && (PL_hints & HINT_NEW_INTEGER))
12019 sv = new_constant(start, s - start, "integer",
12021 else if (PL_hints & HINT_NEW_BINARY)
12022 sv = new_constant(start, s - start, "binary", sv, NULL, NULL);
12027 handle decimal numbers.
12028 we're also sent here when we read a 0 as the first digit
12030 case '1': case '2': case '3': case '4': case '5':
12031 case '6': case '7': case '8': case '9': case '.':
12034 e = PL_tokenbuf + sizeof PL_tokenbuf - 6; /* room for various punctuation */
12037 /* read next group of digits and _ and copy into d */
12038 while (isDIGIT(*s) || *s == '_') {
12039 /* skip underscores, checking for misplaced ones
12043 if (lastub && s == lastub + 1 && ckWARN(WARN_SYNTAX))
12044 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
12045 "Misplaced _ in number");
12049 /* check for end of fixed-length buffer */
12051 Perl_croak(aTHX_ number_too_long);
12052 /* if we're ok, copy the character */
12057 /* final misplaced underbar check */
12058 if (lastub && s == lastub + 1) {
12059 if (ckWARN(WARN_SYNTAX))
12060 Perl_warner(aTHX_ packWARN(WARN_SYNTAX), "Misplaced _ in number");
12063 /* read a decimal portion if there is one. avoid
12064 3..5 being interpreted as the number 3. followed
12067 if (*s == '.' && s[1] != '.') {
12072 if (ckWARN(WARN_SYNTAX))
12073 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
12074 "Misplaced _ in number");
12078 /* copy, ignoring underbars, until we run out of digits.
12080 for (; isDIGIT(*s) || *s == '_'; s++) {
12081 /* fixed length buffer check */
12083 Perl_croak(aTHX_ number_too_long);
12085 if (lastub && s == lastub + 1 && ckWARN(WARN_SYNTAX))
12086 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
12087 "Misplaced _ in number");
12093 /* fractional part ending in underbar? */
12094 if (s[-1] == '_') {
12095 if (ckWARN(WARN_SYNTAX))
12096 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
12097 "Misplaced _ in number");
12099 if (*s == '.' && isDIGIT(s[1])) {
12100 /* oops, it's really a v-string, but without the "v" */
12106 /* read exponent part, if present */
12107 if ((*s == 'e' || *s == 'E') && strchr("+-0123456789_", s[1])) {
12111 /* regardless of whether user said 3E5 or 3e5, use lower 'e' */
12112 *d++ = 'e'; /* At least some Mach atof()s don't grok 'E' */
12114 /* stray preinitial _ */
12116 if (ckWARN(WARN_SYNTAX))
12117 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
12118 "Misplaced _ in number");
12122 /* allow positive or negative exponent */
12123 if (*s == '+' || *s == '-')
12126 /* stray initial _ */
12128 if (ckWARN(WARN_SYNTAX))
12129 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
12130 "Misplaced _ in number");
12134 /* read digits of exponent */
12135 while (isDIGIT(*s) || *s == '_') {
12138 Perl_croak(aTHX_ number_too_long);
12142 if (((lastub && s == lastub + 1) ||
12143 (!isDIGIT(s[1]) && s[1] != '_'))
12144 && ckWARN(WARN_SYNTAX))
12145 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
12146 "Misplaced _ in number");
12153 /* make an sv from the string */
12157 We try to do an integer conversion first if no characters
12158 indicating "float" have been found.
12163 const int flags = grok_number (PL_tokenbuf, d - PL_tokenbuf, &uv);
12165 if (flags == IS_NUMBER_IN_UV) {
12167 sv_setiv(sv, uv); /* Prefer IVs over UVs. */
12170 } else if (flags == (IS_NUMBER_IN_UV | IS_NUMBER_NEG)) {
12171 if (uv <= (UV) IV_MIN)
12172 sv_setiv(sv, -(IV)uv);
12179 /* terminate the string */
12181 nv = Atof(PL_tokenbuf);
12185 if ( floatit ? (PL_hints & HINT_NEW_FLOAT) :
12186 (PL_hints & HINT_NEW_INTEGER) )
12187 sv = new_constant(PL_tokenbuf,
12190 (floatit ? "float" : "integer"),
12194 /* if it starts with a v, it could be a v-string */
12197 sv = newSV(5); /* preallocate storage space */
12198 s = scan_vstring(s,sv);
12202 /* make the op for the constant and return */
12205 lvalp->opval = newSVOP(OP_CONST, 0, sv);
12207 lvalp->opval = NULL;
12213 S_scan_formline(pTHX_ register char *s)
12216 register char *eol;
12218 SV * const stuff = newSVpvs("");
12219 bool needargs = FALSE;
12220 bool eofmt = FALSE;
12222 char *tokenstart = s;
12225 if (PL_madskills) {
12226 savewhite = PL_thiswhite;
12231 while (!needargs) {
12234 #ifdef PERL_STRICT_CR
12235 while (SPACE_OR_TAB(*t))
12238 while (SPACE_OR_TAB(*t) || *t == '\r')
12241 if (*t == '\n' || t == PL_bufend) {
12246 if (PL_in_eval && !PL_rsfp) {
12247 eol = (char *) memchr(s,'\n',PL_bufend-s);
12252 eol = PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
12254 for (t = s; t < eol; t++) {
12255 if (*t == '~' && t[1] == '~' && SvCUR(stuff)) {
12257 goto enough; /* ~~ must be first line in formline */
12259 if (*t == '@' || *t == '^')
12263 sv_catpvn(stuff, s, eol-s);
12264 #ifndef PERL_STRICT_CR
12265 if (eol-s > 1 && eol[-2] == '\r' && eol[-1] == '\n') {
12266 char *end = SvPVX(stuff) + SvCUR(stuff);
12269 SvCUR_set(stuff, SvCUR(stuff) - 1);
12279 if (PL_madskills) {
12281 sv_catpvn(PL_thistoken, tokenstart, PL_bufend - tokenstart);
12283 PL_thistoken = newSVpvn(tokenstart, PL_bufend - tokenstart);
12286 s = filter_gets(PL_linestr, PL_rsfp, 0);
12288 tokenstart = PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = SvPVX(PL_linestr);
12290 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = SvPVX(PL_linestr);
12292 PL_bufend = PL_bufptr + SvCUR(PL_linestr);
12293 PL_last_lop = PL_last_uni = NULL;
12302 if (SvCUR(stuff)) {
12305 PL_lex_state = LEX_NORMAL;
12306 start_force(PL_curforce);
12307 NEXTVAL_NEXTTOKE.ival = 0;
12311 PL_lex_state = LEX_FORMLINE;
12313 if (UTF && is_utf8_string((U8*)SvPVX_const(stuff), SvCUR(stuff)))
12315 else if (PL_encoding)
12316 sv_recode_to_utf8(stuff, PL_encoding);
12318 start_force(PL_curforce);
12319 NEXTVAL_NEXTTOKE.opval = (OP*)newSVOP(OP_CONST, 0, stuff);
12321 start_force(PL_curforce);
12322 NEXTVAL_NEXTTOKE.ival = OP_FORMLINE;
12326 SvREFCNT_dec(stuff);
12328 PL_lex_formbrack = 0;
12332 if (PL_madskills) {
12334 sv_catpvn(PL_thistoken, tokenstart, s - tokenstart);
12336 PL_thistoken = newSVpvn(tokenstart, s - tokenstart);
12337 PL_thiswhite = savewhite;
12349 PL_cshlen = strlen(PL_cshname);
12351 #if defined(USE_ITHREADS)
12352 PERL_UNUSED_CONTEXT;
12358 Perl_start_subparse(pTHX_ I32 is_format, U32 flags)
12361 const I32 oldsavestack_ix = PL_savestack_ix;
12362 CV* const outsidecv = PL_compcv;
12365 assert(SvTYPE(PL_compcv) == SVt_PVCV);
12367 SAVEI32(PL_subline);
12368 save_item(PL_subname);
12369 SAVESPTR(PL_compcv);
12371 PL_compcv = (CV*)newSV(0);
12372 sv_upgrade((SV *)PL_compcv, is_format ? SVt_PVFM : SVt_PVCV);
12373 CvFLAGS(PL_compcv) |= flags;
12375 PL_subline = CopLINE(PL_curcop);
12376 CvPADLIST(PL_compcv) = pad_new(padnew_SAVE|padnew_SAVESUB);
12377 CvOUTSIDE(PL_compcv) = (CV*)SvREFCNT_inc_simple(outsidecv);
12378 CvOUTSIDE_SEQ(PL_compcv) = PL_cop_seqmax;
12380 return oldsavestack_ix;
12384 #pragma segment Perl_yylex
12387 Perl_yywarn(pTHX_ const char *s)
12390 PL_in_eval |= EVAL_WARNONLY;
12392 PL_in_eval &= ~EVAL_WARNONLY;
12397 Perl_yyerror(pTHX_ const char *s)
12400 const char *where = NULL;
12401 const char *context = NULL;
12404 int yychar = PL_parser->yychar;
12406 if (!yychar || (yychar == ';' && !PL_rsfp))
12408 else if (PL_oldoldbufptr && PL_bufptr > PL_oldoldbufptr &&
12409 PL_bufptr - PL_oldoldbufptr < 200 && PL_oldoldbufptr != PL_oldbufptr &&
12410 PL_oldbufptr != PL_bufptr) {
12413 The code below is removed for NetWare because it abends/crashes on NetWare
12414 when the script has error such as not having the closing quotes like:
12415 if ($var eq "value)
12416 Checking of white spaces is anyway done in NetWare code.
12419 while (isSPACE(*PL_oldoldbufptr))
12422 context = PL_oldoldbufptr;
12423 contlen = PL_bufptr - PL_oldoldbufptr;
12425 else if (PL_oldbufptr && PL_bufptr > PL_oldbufptr &&
12426 PL_bufptr - PL_oldbufptr < 200 && PL_oldbufptr != PL_bufptr) {
12429 The code below is removed for NetWare because it abends/crashes on NetWare
12430 when the script has error such as not having the closing quotes like:
12431 if ($var eq "value)
12432 Checking of white spaces is anyway done in NetWare code.
12435 while (isSPACE(*PL_oldbufptr))
12438 context = PL_oldbufptr;
12439 contlen = PL_bufptr - PL_oldbufptr;
12441 else if (yychar > 255)
12442 where = "next token ???";
12443 else if (yychar == -2) { /* YYEMPTY */
12444 if (PL_lex_state == LEX_NORMAL ||
12445 (PL_lex_state == LEX_KNOWNEXT && PL_lex_defer == LEX_NORMAL))
12446 where = "at end of line";
12447 else if (PL_lex_inpat)
12448 where = "within pattern";
12450 where = "within string";
12453 SV * const where_sv = sv_2mortal(newSVpvs("next char "));
12455 Perl_sv_catpvf(aTHX_ where_sv, "^%c", toCTRL(yychar));
12456 else if (isPRINT_LC(yychar))
12457 Perl_sv_catpvf(aTHX_ where_sv, "%c", yychar);
12459 Perl_sv_catpvf(aTHX_ where_sv, "\\%03o", yychar & 255);
12460 where = SvPVX_const(where_sv);
12462 msg = sv_2mortal(newSVpv(s, 0));
12463 Perl_sv_catpvf(aTHX_ msg, " at %s line %"IVdf", ",
12464 OutCopFILE(PL_curcop), (IV)CopLINE(PL_curcop));
12466 Perl_sv_catpvf(aTHX_ msg, "near \"%.*s\"\n", contlen, context);
12468 Perl_sv_catpvf(aTHX_ msg, "%s\n", where);
12469 if (PL_multi_start < PL_multi_end && (U32)(CopLINE(PL_curcop) - PL_multi_end) <= 1) {
12470 Perl_sv_catpvf(aTHX_ msg,
12471 " (Might be a runaway multi-line %c%c string starting on line %"IVdf")\n",
12472 (int)PL_multi_open,(int)PL_multi_close,(IV)PL_multi_start);
12475 if (PL_in_eval & EVAL_WARNONLY && ckWARN_d(WARN_SYNTAX))
12476 Perl_warner(aTHX_ packWARN(WARN_SYNTAX), "%"SVf, SVfARG(msg));
12479 if (PL_error_count >= 10) {
12480 if (PL_in_eval && SvCUR(ERRSV))
12481 Perl_croak(aTHX_ "%"SVf"%s has too many errors.\n",
12482 SVfARG(ERRSV), OutCopFILE(PL_curcop));
12484 Perl_croak(aTHX_ "%s has too many errors.\n",
12485 OutCopFILE(PL_curcop));
12488 PL_in_my_stash = NULL;
12492 #pragma segment Main
12496 S_swallow_bom(pTHX_ U8 *s)
12499 const STRLEN slen = SvCUR(PL_linestr);
12502 if (s[1] == 0xFE) {
12503 /* UTF-16 little-endian? (or UTF32-LE?) */
12504 if (s[2] == 0 && s[3] == 0) /* UTF-32 little-endian */
12505 Perl_croak(aTHX_ "Unsupported script encoding UTF32-LE");
12506 #ifndef PERL_NO_UTF16_FILTER
12507 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF16-LE script encoding (BOM)\n");
12510 if (PL_bufend > (char*)s) {
12514 filter_add(utf16rev_textfilter, NULL);
12515 Newx(news, (PL_bufend - (char*)s) * 3 / 2 + 1, U8);
12516 utf16_to_utf8_reversed(s, news,
12517 PL_bufend - (char*)s - 1,
12519 sv_setpvn(PL_linestr, (const char*)news, newlen);
12521 s = (U8*)SvPVX(PL_linestr);
12522 Copy(news, s, newlen, U8);
12526 SvUTF8_on(PL_linestr);
12527 s = (U8*)SvPVX(PL_linestr);
12529 /* FIXME - is this a general bug fix? */
12532 PL_bufend = SvPVX(PL_linestr) + newlen;
12535 Perl_croak(aTHX_ "Unsupported script encoding UTF16-LE");
12540 if (s[1] == 0xFF) { /* UTF-16 big-endian? */
12541 #ifndef PERL_NO_UTF16_FILTER
12542 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16BE script encoding (BOM)\n");
12545 if (PL_bufend > (char *)s) {
12549 filter_add(utf16_textfilter, NULL);
12550 Newx(news, (PL_bufend - (char*)s) * 3 / 2 + 1, U8);
12551 utf16_to_utf8(s, news,
12552 PL_bufend - (char*)s,
12554 sv_setpvn(PL_linestr, (const char*)news, newlen);
12556 SvUTF8_on(PL_linestr);
12557 s = (U8*)SvPVX(PL_linestr);
12558 PL_bufend = SvPVX(PL_linestr) + newlen;
12561 Perl_croak(aTHX_ "Unsupported script encoding UTF16-BE");
12566 if (slen > 2 && s[1] == 0xBB && s[2] == 0xBF) {
12567 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-8 script encoding (BOM)\n");
12568 s += 3; /* UTF-8 */
12574 if (s[2] == 0xFE && s[3] == 0xFF) {
12575 /* UTF-32 big-endian */
12576 Perl_croak(aTHX_ "Unsupported script encoding UTF32-BE");
12579 else if (s[2] == 0 && s[3] != 0) {
12582 * are a good indicator of UTF-16BE. */
12583 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16BE script encoding (no BOM)\n");
12589 if (slen > 3 && s[1] == 0x73 && s[2] == 0x66 && s[3] == 0x73) {
12590 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-8 script encoding (BOM)\n");
12591 s += 4; /* UTF-8 */
12597 if (slen > 3 && s[1] == 0 && s[2] != 0 && s[3] == 0) {
12600 * are a good indicator of UTF-16LE. */
12601 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16LE script encoding (no BOM)\n");
12610 * Restore a source filter.
12614 restore_rsfp(pTHX_ void *f)
12617 PerlIO * const fp = (PerlIO*)f;
12619 if (PL_rsfp == PerlIO_stdin())
12620 PerlIO_clearerr(PL_rsfp);
12621 else if (PL_rsfp && (PL_rsfp != fp))
12622 PerlIO_close(PL_rsfp);
12626 #ifndef PERL_NO_UTF16_FILTER
12628 utf16_textfilter(pTHX_ int idx, SV *sv, int maxlen)
12631 const STRLEN old = SvCUR(sv);
12632 const I32 count = FILTER_READ(idx+1, sv, maxlen);
12633 DEBUG_P(PerlIO_printf(Perl_debug_log,
12634 "utf16_textfilter(%p): %d %d (%d)\n",
12635 FPTR2DPTR(void *, utf16_textfilter),
12636 idx, maxlen, (int) count));
12640 Newx(tmps, SvCUR(sv) * 3 / 2 + 1, U8);
12641 Copy(SvPVX_const(sv), tmps, old, char);
12642 utf16_to_utf8((U8*)SvPVX_const(sv) + old, tmps + old,
12643 SvCUR(sv) - old, &newlen);
12644 sv_usepvn(sv, (char*)tmps, (STRLEN)newlen + old);
12646 DEBUG_P({sv_dump(sv);});
12651 utf16rev_textfilter(pTHX_ int idx, SV *sv, int maxlen)
12654 const STRLEN old = SvCUR(sv);
12655 const I32 count = FILTER_READ(idx+1, sv, maxlen);
12656 DEBUG_P(PerlIO_printf(Perl_debug_log,
12657 "utf16rev_textfilter(%p): %d %d (%d)\n",
12658 FPTR2DPTR(void *, utf16rev_textfilter),
12659 idx, maxlen, (int) count));
12663 Newx(tmps, SvCUR(sv) * 3 / 2 + 1, U8);
12664 Copy(SvPVX_const(sv), tmps, old, char);
12665 utf16_to_utf8((U8*)SvPVX_const(sv) + old, tmps + old,
12666 SvCUR(sv) - old, &newlen);
12667 sv_usepvn(sv, (char*)tmps, (STRLEN)newlen + old);
12669 DEBUG_P({ sv_dump(sv); });
12675 Returns a pointer to the next character after the parsed
12676 vstring, as well as updating the passed in sv.
12678 Function must be called like
12681 s = scan_vstring(s,sv);
12683 The sv should already be large enough to store the vstring
12684 passed in, for performance reasons.
12689 Perl_scan_vstring(pTHX_ const char *s, SV *sv)
12692 const char *pos = s;
12693 const char *start = s;
12694 if (*pos == 'v') pos++; /* get past 'v' */
12695 while (pos < PL_bufend && (isDIGIT(*pos) || *pos == '_'))
12697 if ( *pos != '.') {
12698 /* this may not be a v-string if followed by => */
12699 const char *next = pos;
12700 while (next < PL_bufend && isSPACE(*next))
12702 if ((PL_bufend - next) >= 2 && *next == '=' && next[1] == '>' ) {
12703 /* return string not v-string */
12704 sv_setpvn(sv,(char *)s,pos-s);
12705 return (char *)pos;
12709 if (!isALPHA(*pos)) {
12710 U8 tmpbuf[UTF8_MAXBYTES+1];
12713 s++; /* get past 'v' */
12715 sv_setpvn(sv, "", 0);
12718 /* this is atoi() that tolerates underscores */
12721 const char *end = pos;
12723 while (--end >= s) {
12725 const UV orev = rev;
12726 rev += (*end - '0') * mult;
12728 if (orev > rev && ckWARN_d(WARN_OVERFLOW))
12729 Perl_warner(aTHX_ packWARN(WARN_OVERFLOW),
12730 "Integer overflow in decimal number");
12734 if (rev > 0x7FFFFFFF)
12735 Perl_croak(aTHX_ "In EBCDIC the v-string components cannot exceed 2147483647");
12737 /* Append native character for the rev point */
12738 tmpend = uvchr_to_utf8(tmpbuf, rev);
12739 sv_catpvn(sv, (const char*)tmpbuf, tmpend - tmpbuf);
12740 if (!UNI_IS_INVARIANT(NATIVE_TO_UNI(rev)))
12742 if (pos + 1 < PL_bufend && *pos == '.' && isDIGIT(pos[1]))
12748 while (pos < PL_bufend && (isDIGIT(*pos) || *pos == '_'))
12752 sv_magic(sv,NULL,PERL_MAGIC_vstring,(const char*)start, pos-start);
12760 * c-indentation-style: bsd
12761 * c-basic-offset: 4
12762 * indent-tabs-mode: t
12765 * ex: set ts=8 sts=4 sw=4 noet: