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);
681 PL_lex_state = LEX_NORMAL;
683 Newx(parser->lex_brackstack, 120, char);
684 Newx(parser->lex_casestack, 12, char);
685 *parser->lex_casestack = '\0';
691 s = SvPV_const(line, len);
696 PL_linestr = newSVpvs("\n;");
697 } else if (SvREADONLY(line) || s[len-1] != ';') {
698 PL_linestr = newSVsv(line);
700 sv_catpvs(PL_linestr, "\n;");
703 SvREFCNT_inc_simple_void_NN(line);
706 /* PL_linestr needs to survive until end of scope, not just the next
707 FREETMPS. See changes 17505 and 17546 which fixed the symptoms only. */
708 SAVEFREESV(PL_linestr);
709 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = SvPVX(PL_linestr);
710 PL_bufend = PL_bufptr + SvCUR(PL_linestr);
711 PL_last_lop = PL_last_uni = NULL;
717 * Finalizer for lexing operations. Must be called when the parser is
718 * done with the lexer.
725 PL_doextract = FALSE;
730 * This subroutine has nothing to do with tilting, whether at windmills
731 * or pinball tables. Its name is short for "increment line". It
732 * increments the current line number in CopLINE(PL_curcop) and checks
733 * to see whether the line starts with a comment of the form
734 * # line 500 "foo.pm"
735 * If so, it sets the current line number and file to the values in the comment.
739 S_incline(pTHX_ const char *s)
746 CopLINE_inc(PL_curcop);
749 while (SPACE_OR_TAB(*s))
751 if (strnEQ(s, "line", 4))
755 if (SPACE_OR_TAB(*s))
759 while (SPACE_OR_TAB(*s))
767 while (SPACE_OR_TAB(*s))
769 if (*s == '"' && (t = strchr(s+1, '"'))) {
779 while (SPACE_OR_TAB(*e) || *e == '\r' || *e == '\f')
781 if (*e != '\n' && *e != '\0')
782 return; /* false alarm */
785 const STRLEN len = t - s;
787 const char * const cf = CopFILE(PL_curcop);
788 STRLEN tmplen = cf ? strlen(cf) : 0;
789 if (tmplen > 7 && strnEQ(cf, "(eval ", 6)) {
790 /* must copy *{"::_<(eval N)[oldfilename:L]"}
791 * to *{"::_<newfilename"} */
792 /* However, the long form of evals is only turned on by the
793 debugger - usually they're "(eval %lu)" */
797 STRLEN tmplen2 = len;
798 if (tmplen + 2 <= sizeof smallbuf)
801 Newx(tmpbuf, tmplen + 2, char);
804 memcpy(tmpbuf + 2, cf, tmplen);
806 gvp = (GV**)hv_fetch(PL_defstash, tmpbuf, tmplen, FALSE);
811 if (tmplen2 + 2 <= sizeof smallbuf)
814 Newx(tmpbuf2, tmplen2 + 2, char);
816 if (tmpbuf2 != smallbuf || tmpbuf != smallbuf) {
817 /* Either they malloc'd it, or we malloc'd it,
818 so no prefix is present in ours. */
823 memcpy(tmpbuf2 + 2, s, tmplen2);
826 gv2 = *(GV**)hv_fetch(PL_defstash, tmpbuf2, tmplen2, TRUE);
828 gv_init(gv2, PL_defstash, tmpbuf2, tmplen2, FALSE);
829 /* adjust ${"::_<newfilename"} to store the new file name */
830 GvSV(gv2) = newSVpvn(tmpbuf2 + 2, tmplen2 - 2);
831 GvHV(gv2) = (HV*)SvREFCNT_inc(GvHV(*gvp));
832 GvAV(gv2) = (AV*)SvREFCNT_inc(GvAV(*gvp));
835 if (tmpbuf2 != smallbuf) Safefree(tmpbuf2);
837 if (tmpbuf != smallbuf) Safefree(tmpbuf);
840 CopFILE_free(PL_curcop);
841 CopFILE_setn(PL_curcop, s, len);
843 CopLINE_set(PL_curcop, atoi(n)-1);
847 /* skip space before PL_thistoken */
850 S_skipspace0(pTHX_ register char *s)
857 PL_thiswhite = newSVpvs("");
858 sv_catsv(PL_thiswhite, PL_skipwhite);
859 sv_free(PL_skipwhite);
862 PL_realtokenstart = s - SvPVX(PL_linestr);
866 /* skip space after PL_thistoken */
869 S_skipspace1(pTHX_ register char *s)
871 const char *start = s;
872 I32 startoff = start - SvPVX(PL_linestr);
877 start = SvPVX(PL_linestr) + startoff;
878 if (!PL_thistoken && PL_realtokenstart >= 0) {
879 const char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
880 PL_thistoken = newSVpvn(tstart, start - tstart);
882 PL_realtokenstart = -1;
885 PL_nextwhite = newSVpvs("");
886 sv_catsv(PL_nextwhite, PL_skipwhite);
887 sv_free(PL_skipwhite);
894 S_skipspace2(pTHX_ register char *s, SV **svp)
897 const I32 bufptroff = PL_bufptr - SvPVX(PL_linestr);
898 const I32 startoff = s - SvPVX(PL_linestr);
901 PL_bufptr = SvPVX(PL_linestr) + bufptroff;
902 if (!PL_madskills || !svp)
904 start = SvPVX(PL_linestr) + startoff;
905 if (!PL_thistoken && PL_realtokenstart >= 0) {
906 char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
907 PL_thistoken = newSVpvn(tstart, start - tstart);
908 PL_realtokenstart = -1;
913 sv_setsv(*svp, PL_skipwhite);
914 sv_free(PL_skipwhite);
923 S_update_debugger_info(pTHX_ SV *orig_sv, const char *buf, STRLEN len)
925 AV *av = CopFILEAVx(PL_curcop);
927 SV * const sv = newSV_type(SVt_PVMG);
929 sv_setsv(sv, orig_sv);
931 sv_setpvn(sv, buf, len);
934 av_store(av, (I32)CopLINE(PL_curcop), sv);
940 * Called to gobble the appropriate amount and type of whitespace.
941 * Skips comments as well.
945 S_skipspace(pTHX_ register char *s)
950 int startoff = s - SvPVX(PL_linestr);
953 sv_free(PL_skipwhite);
958 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
959 while (s < PL_bufend && SPACE_OR_TAB(*s))
969 SSize_t oldprevlen, oldoldprevlen;
970 SSize_t oldloplen = 0, oldunilen = 0;
971 while (s < PL_bufend && isSPACE(*s)) {
972 if (*s++ == '\n' && PL_in_eval && !PL_rsfp)
977 if (s < PL_bufend && *s == '#') {
978 while (s < PL_bufend && *s != '\n')
982 if (PL_in_eval && !PL_rsfp) {
989 /* only continue to recharge the buffer if we're at the end
990 * of the buffer, we're not reading from a source filter, and
991 * we're in normal lexing mode
993 if (s < PL_bufend || !PL_rsfp || PL_sublex_info.sub_inwhat ||
994 PL_lex_state == LEX_FORMLINE)
1001 /* try to recharge the buffer */
1003 curoff = s - SvPVX(PL_linestr);
1006 if ((s = filter_gets(PL_linestr, PL_rsfp,
1007 (prevlen = SvCUR(PL_linestr)))) == NULL)
1010 if (PL_madskills && curoff != startoff) {
1012 PL_skipwhite = newSVpvs("");
1013 sv_catpvn(PL_skipwhite, SvPVX(PL_linestr) + startoff,
1017 /* mustn't throw out old stuff yet if madpropping */
1018 SvCUR(PL_linestr) = curoff;
1019 s = SvPVX(PL_linestr) + curoff;
1021 if (curoff && s[-1] == '\n')
1025 /* end of file. Add on the -p or -n magic */
1026 /* XXX these shouldn't really be added here, can't set PL_faketokens */
1029 sv_catpvs(PL_linestr,
1030 ";}continue{print or die qq(-p destination: $!\\n);}");
1032 sv_setpvs(PL_linestr,
1033 ";}continue{print or die qq(-p destination: $!\\n);}");
1035 PL_minus_n = PL_minus_p = 0;
1037 else if (PL_minus_n) {
1039 sv_catpvn(PL_linestr, ";}", 2);
1041 sv_setpvn(PL_linestr, ";}", 2);
1047 sv_catpvn(PL_linestr,";", 1);
1049 sv_setpvn(PL_linestr,";", 1);
1052 /* reset variables for next time we lex */
1053 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = s = PL_linestart
1059 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
1060 PL_last_lop = PL_last_uni = NULL;
1062 /* Close the filehandle. Could be from -P preprocessor,
1063 * STDIN, or a regular file. If we were reading code from
1064 * STDIN (because the commandline held no -e or filename)
1065 * then we don't close it, we reset it so the code can
1066 * read from STDIN too.
1069 if (PL_preprocess && !PL_in_eval)
1070 (void)PerlProc_pclose(PL_rsfp);
1071 else if ((PerlIO*)PL_rsfp == PerlIO_stdin())
1072 PerlIO_clearerr(PL_rsfp);
1074 (void)PerlIO_close(PL_rsfp);
1079 /* not at end of file, so we only read another line */
1080 /* make corresponding updates to old pointers, for yyerror() */
1081 oldprevlen = PL_oldbufptr - PL_bufend;
1082 oldoldprevlen = PL_oldoldbufptr - PL_bufend;
1084 oldunilen = PL_last_uni - PL_bufend;
1086 oldloplen = PL_last_lop - PL_bufend;
1087 PL_linestart = PL_bufptr = s + prevlen;
1088 PL_bufend = s + SvCUR(PL_linestr);
1090 PL_oldbufptr = s + oldprevlen;
1091 PL_oldoldbufptr = s + oldoldprevlen;
1093 PL_last_uni = s + oldunilen;
1095 PL_last_lop = s + oldloplen;
1098 /* debugger active and we're not compiling the debugger code,
1099 * so store the line into the debugger's array of lines
1101 if (PERLDB_LINE && PL_curstash != PL_debstash)
1102 update_debugger_info(NULL, PL_bufptr, PL_bufend - PL_bufptr);
1109 PL_skipwhite = newSVpvs("");
1110 curoff = s - SvPVX(PL_linestr);
1111 if (curoff - startoff)
1112 sv_catpvn(PL_skipwhite, SvPVX(PL_linestr) + startoff,
1121 * Check the unary operators to ensure there's no ambiguity in how they're
1122 * used. An ambiguous piece of code would be:
1124 * This doesn't mean rand() + 5. Because rand() is a unary operator,
1125 * the +5 is its argument.
1135 if (PL_oldoldbufptr != PL_last_uni)
1137 while (isSPACE(*PL_last_uni))
1140 while (isALNUM_lazy_if(s,UTF) || *s == '-')
1142 if ((t = strchr(s, '(')) && t < PL_bufptr)
1145 if (ckWARN_d(WARN_AMBIGUOUS)){
1146 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
1147 "Warning: Use of \"%.*s\" without parentheses is ambiguous",
1148 (int)(s - PL_last_uni), PL_last_uni);
1153 * LOP : macro to build a list operator. Its behaviour has been replaced
1154 * with a subroutine, S_lop() for which LOP is just another name.
1157 #define LOP(f,x) return lop(f,x,s)
1161 * Build a list operator (or something that might be one). The rules:
1162 * - if we have a next token, then it's a list operator [why?]
1163 * - if the next thing is an opening paren, then it's a function
1164 * - else it's a list operator
1168 S_lop(pTHX_ I32 f, int x, char *s)
1175 PL_last_lop = PL_oldbufptr;
1176 PL_last_lop_op = (OPCODE)f;
1179 return REPORT(LSTOP);
1182 return REPORT(LSTOP);
1185 return REPORT(FUNC);
1188 return REPORT(FUNC);
1190 return REPORT(LSTOP);
1196 * Sets up for an eventual force_next(). start_force(0) basically does
1197 * an unshift, while start_force(-1) does a push. yylex removes items
1202 S_start_force(pTHX_ int where)
1206 if (where < 0) /* so people can duplicate start_force(PL_curforce) */
1207 where = PL_lasttoke;
1208 assert(PL_curforce < 0 || PL_curforce == where);
1209 if (PL_curforce != where) {
1210 for (i = PL_lasttoke; i > where; --i) {
1211 PL_nexttoke[i] = PL_nexttoke[i-1];
1215 if (PL_curforce < 0) /* in case of duplicate start_force() */
1216 Zero(&PL_nexttoke[where], 1, NEXTTOKE);
1217 PL_curforce = where;
1220 curmad('^', newSVpvs(""));
1221 CURMAD('_', PL_nextwhite);
1226 S_curmad(pTHX_ char slot, SV *sv)
1232 if (PL_curforce < 0)
1233 where = &PL_thismad;
1235 where = &PL_nexttoke[PL_curforce].next_mad;
1238 sv_setpvn(sv, "", 0);
1241 if (UTF && is_utf8_string((U8*)SvPVX(sv), SvCUR(sv)))
1243 else if (PL_encoding) {
1244 sv_recode_to_utf8(sv, PL_encoding);
1249 /* keep a slot open for the head of the list? */
1250 if (slot != '_' && *where && (*where)->mad_key == '^') {
1251 (*where)->mad_key = slot;
1252 sv_free((*where)->mad_val);
1253 (*where)->mad_val = (void*)sv;
1256 addmad(newMADsv(slot, sv), where, 0);
1259 # define start_force(where) NOOP
1260 # define curmad(slot, sv) NOOP
1265 * When the lexer realizes it knows the next token (for instance,
1266 * it is reordering tokens for the parser) then it can call S_force_next
1267 * to know what token to return the next time the lexer is called. Caller
1268 * will need to set PL_nextval[] (or PL_nexttoke[].next_val with PERL_MAD),
1269 * and possibly PL_expect to ensure the lexer handles the token correctly.
1273 S_force_next(pTHX_ I32 type)
1277 if (PL_curforce < 0)
1278 start_force(PL_lasttoke);
1279 PL_nexttoke[PL_curforce].next_type = type;
1280 if (PL_lex_state != LEX_KNOWNEXT)
1281 PL_lex_defer = PL_lex_state;
1282 PL_lex_state = LEX_KNOWNEXT;
1283 PL_lex_expect = PL_expect;
1286 PL_nexttype[PL_nexttoke] = type;
1288 if (PL_lex_state != LEX_KNOWNEXT) {
1289 PL_lex_defer = PL_lex_state;
1290 PL_lex_expect = PL_expect;
1291 PL_lex_state = LEX_KNOWNEXT;
1297 S_newSV_maybe_utf8(pTHX_ const char *start, STRLEN len)
1300 SV * const sv = newSVpvn(start,len);
1301 if (UTF && !IN_BYTES && is_utf8_string((const U8*)start, len))
1308 * When the lexer knows the next thing is a word (for instance, it has
1309 * just seen -> and it knows that the next char is a word char, then
1310 * it calls S_force_word to stick the next word into the PL_nexttoke/val
1314 * char *start : buffer position (must be within PL_linestr)
1315 * int token : PL_next* will be this type of bare word (e.g., METHOD,WORD)
1316 * int check_keyword : if true, Perl checks to make sure the word isn't
1317 * a keyword (do this if the word is a label, e.g. goto FOO)
1318 * int allow_pack : if true, : characters will also be allowed (require,
1319 * use, etc. do this)
1320 * int allow_initial_tick : used by the "sub" lexer only.
1324 S_force_word(pTHX_ register char *start, int token, int check_keyword, int allow_pack, int allow_initial_tick)
1330 start = SKIPSPACE1(start);
1332 if (isIDFIRST_lazy_if(s,UTF) ||
1333 (allow_pack && *s == ':') ||
1334 (allow_initial_tick && *s == '\'') )
1336 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, allow_pack, &len);
1337 if (check_keyword && keyword(PL_tokenbuf, len, 0))
1339 start_force(PL_curforce);
1341 curmad('X', newSVpvn(start,s-start));
1342 if (token == METHOD) {
1347 PL_expect = XOPERATOR;
1350 NEXTVAL_NEXTTOKE.opval
1351 = (OP*)newSVOP(OP_CONST,0,
1352 S_newSV_maybe_utf8(aTHX_ PL_tokenbuf, len));
1353 NEXTVAL_NEXTTOKE.opval->op_private |= OPpCONST_BARE;
1361 * Called when the lexer wants $foo *foo &foo etc, but the program
1362 * text only contains the "foo" portion. The first argument is a pointer
1363 * to the "foo", and the second argument is the type symbol to prefix.
1364 * Forces the next token to be a "WORD".
1365 * Creates the symbol if it didn't already exist (via gv_fetchpv()).
1369 S_force_ident(pTHX_ register const char *s, int kind)
1373 const STRLEN len = strlen(s);
1374 OP* const o = (OP*)newSVOP(OP_CONST, 0, newSVpvn(s, len));
1375 start_force(PL_curforce);
1376 NEXTVAL_NEXTTOKE.opval = o;
1379 o->op_private = OPpCONST_ENTERED;
1380 /* XXX see note in pp_entereval() for why we forgo typo
1381 warnings if the symbol must be introduced in an eval.
1383 gv_fetchpvn_flags(s, len,
1384 PL_in_eval ? (GV_ADDMULTI | GV_ADDINEVAL)
1386 kind == '$' ? SVt_PV :
1387 kind == '@' ? SVt_PVAV :
1388 kind == '%' ? SVt_PVHV :
1396 Perl_str_to_version(pTHX_ SV *sv)
1401 const char *start = SvPV_const(sv,len);
1402 const char * const end = start + len;
1403 const bool utf = SvUTF8(sv) ? TRUE : FALSE;
1404 while (start < end) {
1408 n = utf8n_to_uvchr((U8*)start, len, &skip, 0);
1413 retval += ((NV)n)/nshift;
1422 * Forces the next token to be a version number.
1423 * If the next token appears to be an invalid version number, (e.g. "v2b"),
1424 * and if "guessing" is TRUE, then no new token is created (and the caller
1425 * must use an alternative parsing method).
1429 S_force_version(pTHX_ char *s, int guessing)
1435 I32 startoff = s - SvPVX(PL_linestr);
1444 while (isDIGIT(*d) || *d == '_' || *d == '.')
1448 start_force(PL_curforce);
1449 curmad('X', newSVpvn(s,d-s));
1452 if (*d == ';' || isSPACE(*d) || *d == '}' || !*d) {
1454 s = scan_num(s, &yylval);
1455 version = yylval.opval;
1456 ver = cSVOPx(version)->op_sv;
1457 if (SvPOK(ver) && !SvNIOK(ver)) {
1458 SvUPGRADE(ver, SVt_PVNV);
1459 SvNV_set(ver, str_to_version(ver));
1460 SvNOK_on(ver); /* hint that it is a version */
1463 else if (guessing) {
1466 sv_free(PL_nextwhite); /* let next token collect whitespace */
1468 s = SvPVX(PL_linestr) + startoff;
1476 if (PL_madskills && !version) {
1477 sv_free(PL_nextwhite); /* let next token collect whitespace */
1479 s = SvPVX(PL_linestr) + startoff;
1482 /* NOTE: The parser sees the package name and the VERSION swapped */
1483 start_force(PL_curforce);
1484 NEXTVAL_NEXTTOKE.opval = version;
1492 * Tokenize a quoted string passed in as an SV. It finds the next
1493 * chunk, up to end of string or a backslash. It may make a new
1494 * SV containing that chunk (if HINT_NEW_STRING is on). It also
1499 S_tokeq(pTHX_ SV *sv)
1503 register char *send;
1511 s = SvPV_force(sv, len);
1512 if (SvTYPE(sv) >= SVt_PVIV && SvIVX(sv) == -1)
1515 while (s < send && *s != '\\')
1520 if ( PL_hints & HINT_NEW_STRING ) {
1521 pv = sv_2mortal(newSVpvn(SvPVX_const(pv), len));
1527 if (s + 1 < send && (s[1] == '\\'))
1528 s++; /* all that, just for this */
1533 SvCUR_set(sv, d - SvPVX_const(sv));
1535 if ( PL_hints & HINT_NEW_STRING )
1536 return new_constant(NULL, 0, "q", sv, pv, "q");
1541 * Now come three functions related to double-quote context,
1542 * S_sublex_start, S_sublex_push, and S_sublex_done. They're used when
1543 * converting things like "\u\Lgnat" into ucfirst(lc("gnat")). They
1544 * interact with PL_lex_state, and create fake ( ... ) argument lists
1545 * to handle functions and concatenation.
1546 * They assume that whoever calls them will be setting up a fake
1547 * join call, because each subthing puts a ',' after it. This lets
1550 * join($, , 'lower ', lcfirst( 'uPpEr', ) ,)
1552 * (I'm not sure whether the spurious commas at the end of lcfirst's
1553 * arguments and join's arguments are created or not).
1558 * Assumes that yylval.ival is the op we're creating (e.g. OP_LCFIRST).
1560 * Pattern matching will set PL_lex_op to the pattern-matching op to
1561 * make (we return THING if yylval.ival is OP_NULL, PMFUNC otherwise).
1563 * OP_CONST and OP_READLINE are easy--just make the new op and return.
1565 * Everything else becomes a FUNC.
1567 * Sets PL_lex_state to LEX_INTERPPUSH unless (ival was OP_NULL or we
1568 * had an OP_CONST or OP_READLINE). This just sets us up for a
1569 * call to S_sublex_push().
1573 S_sublex_start(pTHX)
1576 register const I32 op_type = yylval.ival;
1578 if (op_type == OP_NULL) {
1579 yylval.opval = PL_lex_op;
1583 if (op_type == OP_CONST || op_type == OP_READLINE) {
1584 SV *sv = tokeq(PL_lex_stuff);
1586 if (SvTYPE(sv) == SVt_PVIV) {
1587 /* Overloaded constants, nothing fancy: Convert to SVt_PV: */
1589 const char * const p = SvPV_const(sv, len);
1590 SV * const nsv = newSVpvn(p, len);
1596 yylval.opval = (OP*)newSVOP(op_type, 0, sv);
1597 PL_lex_stuff = NULL;
1598 /* Allow <FH> // "foo" */
1599 if (op_type == OP_READLINE)
1600 PL_expect = XTERMORDORDOR;
1603 else if (op_type == OP_BACKTICK && PL_lex_op) {
1604 /* readpipe() vas overriden */
1605 cSVOPx(cLISTOPx(cUNOPx(PL_lex_op)->op_first)->op_first->op_sibling)->op_sv = tokeq(PL_lex_stuff);
1606 yylval.opval = PL_lex_op;
1608 PL_lex_stuff = NULL;
1612 PL_sublex_info.super_state = PL_lex_state;
1613 PL_sublex_info.sub_inwhat = op_type;
1614 PL_sublex_info.sub_op = PL_lex_op;
1615 PL_lex_state = LEX_INTERPPUSH;
1619 yylval.opval = PL_lex_op;
1629 * Create a new scope to save the lexing state. The scope will be
1630 * ended in S_sublex_done. Returns a '(', starting the function arguments
1631 * to the uc, lc, etc. found before.
1632 * Sets PL_lex_state to LEX_INTERPCONCAT.
1641 PL_lex_state = PL_sublex_info.super_state;
1642 SAVEI32(PL_lex_dojoin);
1643 SAVEI32(PL_lex_brackets);
1644 SAVEI32(PL_lex_casemods);
1645 SAVEI32(PL_lex_starts);
1646 SAVEI32(PL_lex_state);
1647 SAVEVPTR(PL_lex_inpat);
1648 SAVEI32(PL_lex_inwhat);
1649 SAVECOPLINE(PL_curcop);
1650 SAVEPPTR(PL_bufptr);
1651 SAVEPPTR(PL_bufend);
1652 SAVEPPTR(PL_oldbufptr);
1653 SAVEPPTR(PL_oldoldbufptr);
1654 SAVEPPTR(PL_last_lop);
1655 SAVEPPTR(PL_last_uni);
1656 SAVEPPTR(PL_linestart);
1657 SAVESPTR(PL_linestr);
1658 SAVEGENERICPV(PL_lex_brackstack);
1659 SAVEGENERICPV(PL_lex_casestack);
1661 PL_linestr = PL_lex_stuff;
1662 PL_lex_stuff = NULL;
1664 PL_bufend = PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart
1665 = SvPVX(PL_linestr);
1666 PL_bufend += SvCUR(PL_linestr);
1667 PL_last_lop = PL_last_uni = NULL;
1668 SAVEFREESV(PL_linestr);
1670 PL_lex_dojoin = FALSE;
1671 PL_lex_brackets = 0;
1672 Newx(PL_lex_brackstack, 120, char);
1673 Newx(PL_lex_casestack, 12, char);
1674 PL_lex_casemods = 0;
1675 *PL_lex_casestack = '\0';
1677 PL_lex_state = LEX_INTERPCONCAT;
1678 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
1680 PL_lex_inwhat = PL_sublex_info.sub_inwhat;
1681 if (PL_lex_inwhat == OP_MATCH || PL_lex_inwhat == OP_QR || PL_lex_inwhat == OP_SUBST)
1682 PL_lex_inpat = PL_sublex_info.sub_op;
1684 PL_lex_inpat = NULL;
1691 * Restores lexer state after a S_sublex_push.
1698 if (!PL_lex_starts++) {
1699 SV * const sv = newSVpvs("");
1700 if (SvUTF8(PL_linestr))
1702 PL_expect = XOPERATOR;
1703 yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
1707 if (PL_lex_casemods) { /* oops, we've got some unbalanced parens */
1708 PL_lex_state = LEX_INTERPCASEMOD;
1712 /* Is there a right-hand side to take care of? (s//RHS/ or tr//RHS/) */
1713 if (PL_lex_repl && (PL_lex_inwhat == OP_SUBST || PL_lex_inwhat == OP_TRANS)) {
1714 PL_linestr = PL_lex_repl;
1716 PL_bufend = PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart = SvPVX(PL_linestr);
1717 PL_bufend += SvCUR(PL_linestr);
1718 PL_last_lop = PL_last_uni = NULL;
1719 SAVEFREESV(PL_linestr);
1720 PL_lex_dojoin = FALSE;
1721 PL_lex_brackets = 0;
1722 PL_lex_casemods = 0;
1723 *PL_lex_casestack = '\0';
1725 if (SvEVALED(PL_lex_repl)) {
1726 PL_lex_state = LEX_INTERPNORMAL;
1728 /* we don't clear PL_lex_repl here, so that we can check later
1729 whether this is an evalled subst; that means we rely on the
1730 logic to ensure sublex_done() is called again only via the
1731 branch (in yylex()) that clears PL_lex_repl, else we'll loop */
1734 PL_lex_state = LEX_INTERPCONCAT;
1744 PL_endwhite = newSVpvs("");
1745 sv_catsv(PL_endwhite, PL_thiswhite);
1749 sv_setpvn(PL_thistoken,"",0);
1751 PL_realtokenstart = -1;
1755 PL_bufend = SvPVX(PL_linestr);
1756 PL_bufend += SvCUR(PL_linestr);
1757 PL_expect = XOPERATOR;
1758 PL_sublex_info.sub_inwhat = 0;
1766 Extracts a pattern, double-quoted string, or transliteration. This
1769 It looks at PL_lex_inwhat and PL_lex_inpat to find out whether it's
1770 processing a pattern (PL_lex_inpat is true), a transliteration
1771 (PL_lex_inwhat == OP_TRANS is true), or a double-quoted string.
1773 Returns a pointer to the character scanned up to. If this is
1774 advanced from the start pointer supplied (i.e. if anything was
1775 successfully parsed), will leave an OP for the substring scanned
1776 in yylval. Caller must intuit reason for not parsing further
1777 by looking at the next characters herself.
1781 double-quoted style: \r and \n
1782 regexp special ones: \D \s
1785 case and quoting: \U \Q \E
1786 stops on @ and $, but not for $ as tail anchor
1788 In transliterations:
1789 characters are VERY literal, except for - not at the start or end
1790 of the string, which indicates a range. If the range is in bytes,
1791 scan_const expands the range to the full set of intermediate
1792 characters. If the range is in utf8, the hyphen is replaced with
1793 a certain range mark which will be handled by pmtrans() in op.c.
1795 In double-quoted strings:
1797 double-quoted style: \r and \n
1799 deprecated backrefs: \1 (in substitution replacements)
1800 case and quoting: \U \Q \E
1803 scan_const does *not* construct ops to handle interpolated strings.
1804 It stops processing as soon as it finds an embedded $ or @ variable
1805 and leaves it to the caller to work out what's going on.
1807 embedded arrays (whether in pattern or not) could be:
1808 @foo, @::foo, @'foo, @{foo}, @$foo, @+, @-.
1810 $ in double-quoted strings must be the symbol of an embedded scalar.
1812 $ in pattern could be $foo or could be tail anchor. Assumption:
1813 it's a tail anchor if $ is the last thing in the string, or if it's
1814 followed by one of "()| \r\n\t"
1816 \1 (backreferences) are turned into $1
1818 The structure of the code is
1819 while (there's a character to process) {
1820 handle transliteration ranges
1821 skip regexp comments /(?#comment)/ and codes /(?{code})/
1822 skip #-initiated comments in //x patterns
1823 check for embedded arrays
1824 check for embedded scalars
1826 leave intact backslashes from leaveit (below)
1827 deprecate \1 in substitution replacements
1828 handle string-changing backslashes \l \U \Q \E, etc.
1829 switch (what was escaped) {
1830 handle \- in a transliteration (becomes a literal -)
1831 handle \132 (octal characters)
1832 handle \x15 and \x{1234} (hex characters)
1833 handle \N{name} (named characters)
1834 handle \cV (control characters)
1835 handle printf-style backslashes (\f, \r, \n, etc)
1837 } (end if backslash)
1838 } (end while character to read)
1843 S_scan_const(pTHX_ char *start)
1846 register char *send = PL_bufend; /* end of the constant */
1847 SV *sv = newSV(send - start); /* sv for the constant */
1848 register char *s = start; /* start of the constant */
1849 register char *d = SvPVX(sv); /* destination for copies */
1850 bool dorange = FALSE; /* are we in a translit range? */
1851 bool didrange = FALSE; /* did we just finish a range? */
1852 I32 has_utf8 = FALSE; /* Output constant is UTF8 */
1853 I32 this_utf8 = UTF; /* The source string is assumed to be UTF8 */
1856 UV literal_endpoint = 0;
1857 bool native_range = TRUE; /* turned to FALSE if the first endpoint is Unicode. */
1860 if (PL_lex_inwhat == OP_TRANS && PL_sublex_info.sub_op) {
1861 /* If we are doing a trans and we know we want UTF8 set expectation */
1862 has_utf8 = PL_sublex_info.sub_op->op_private & (OPpTRANS_FROM_UTF|OPpTRANS_TO_UTF);
1863 this_utf8 = PL_sublex_info.sub_op->op_private & (PL_lex_repl ? OPpTRANS_FROM_UTF : OPpTRANS_TO_UTF);
1867 while (s < send || dorange) {
1868 /* get transliterations out of the way (they're most literal) */
1869 if (PL_lex_inwhat == OP_TRANS) {
1870 /* expand a range A-Z to the full set of characters. AIE! */
1872 I32 i; /* current expanded character */
1873 I32 min; /* first character in range */
1874 I32 max; /* last character in range */
1885 char * const c = (char*)utf8_hop((U8*)d, -1);
1889 *c = (char)UTF_TO_NATIVE(0xff);
1890 /* mark the range as done, and continue */
1896 i = d - SvPVX_const(sv); /* remember current offset */
1899 SvLEN(sv) + (has_utf8 ?
1900 (512 - UTF_CONTINUATION_MARK +
1903 /* How many two-byte within 0..255: 128 in UTF-8,
1904 * 96 in UTF-8-mod. */
1906 SvGROW(sv, SvLEN(sv) + 256); /* never more than 256 chars in a range */
1908 d = SvPVX(sv) + i; /* refresh d after realloc */
1912 for (j = 0; j <= 1; j++) {
1913 char * const c = (char*)utf8_hop((U8*)d, -1);
1914 const UV uv = utf8n_to_uvchr((U8*)c, d - c, NULL, 0);
1920 max = (U8)0xff; /* only to \xff */
1921 uvmax = uv; /* \x{100} to uvmax */
1923 d = c; /* eat endpoint chars */
1928 d -= 2; /* eat the first char and the - */
1929 min = (U8)*d; /* first char in range */
1930 max = (U8)d[1]; /* last char in range */
1937 "Invalid range \"%c-%c\" in transliteration operator",
1938 (char)min, (char)max);
1942 if (literal_endpoint == 2 &&
1943 ((isLOWER(min) && isLOWER(max)) ||
1944 (isUPPER(min) && isUPPER(max)))) {
1946 for (i = min; i <= max; i++)
1948 *d++ = NATIVE_TO_NEED(has_utf8,i);
1950 for (i = min; i <= max; i++)
1952 *d++ = NATIVE_TO_NEED(has_utf8,i);
1957 for (i = min; i <= max; i++)
1960 const U8 ch = (U8)NATIVE_TO_UTF(i);
1961 if (UNI_IS_INVARIANT(ch))
1964 *d++ = (U8)UTF8_EIGHT_BIT_HI(ch);
1965 *d++ = (U8)UTF8_EIGHT_BIT_LO(ch);
1974 d = (char*)uvchr_to_utf8((U8*)d, 0x100);
1976 *d++ = (char)UTF_TO_NATIVE(0xff);
1978 d = (char*)uvchr_to_utf8((U8*)d, uvmax);
1982 /* mark the range as done, and continue */
1986 literal_endpoint = 0;
1991 /* range begins (ignore - as first or last char) */
1992 else if (*s == '-' && s+1 < send && s != start) {
1994 Perl_croak(aTHX_ "Ambiguous range in transliteration operator");
2001 *d++ = (char)UTF_TO_NATIVE(0xff); /* use illegal utf8 byte--see pmtrans */
2011 literal_endpoint = 0;
2012 native_range = TRUE;
2017 /* if we get here, we're not doing a transliteration */
2019 /* skip for regexp comments /(?#comment)/ and code /(?{code})/,
2020 except for the last char, which will be done separately. */
2021 else if (*s == '(' && PL_lex_inpat && s[1] == '?') {
2023 while (s+1 < send && *s != ')')
2024 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2026 else if (s[2] == '{' /* This should match regcomp.c */
2027 || (s[2] == '?' && s[3] == '{'))
2030 char *regparse = s + (s[2] == '{' ? 3 : 4);
2033 while (count && (c = *regparse)) {
2034 if (c == '\\' && regparse[1])
2042 if (*regparse != ')')
2043 regparse--; /* Leave one char for continuation. */
2044 while (s < regparse)
2045 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2049 /* likewise skip #-initiated comments in //x patterns */
2050 else if (*s == '#' && PL_lex_inpat &&
2051 ((PMOP*)PL_lex_inpat)->op_pmflags & PMf_EXTENDED) {
2052 while (s+1 < send && *s != '\n')
2053 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2056 /* check for embedded arrays
2057 (@foo, @::foo, @'foo, @{foo}, @$foo, @+, @-)
2059 else if (*s == '@' && s[1]) {
2060 if (isALNUM_lazy_if(s+1,UTF))
2062 if (strchr(":'{$", s[1]))
2064 if (!PL_lex_inpat && (s[1] == '+' || s[1] == '-'))
2065 break; /* in regexp, neither @+ nor @- are interpolated */
2068 /* check for embedded scalars. only stop if we're sure it's a
2071 else if (*s == '$') {
2072 if (!PL_lex_inpat) /* not a regexp, so $ must be var */
2074 if (s + 1 < send && !strchr("()| \r\n\t", s[1]))
2075 break; /* in regexp, $ might be tail anchor */
2078 /* End of else if chain - OP_TRANS rejoin rest */
2081 if (*s == '\\' && s+1 < send) {
2084 /* deprecate \1 in strings and substitution replacements */
2085 if (PL_lex_inwhat == OP_SUBST && !PL_lex_inpat &&
2086 isDIGIT(*s) && *s != '0' && !isDIGIT(s[1]))
2088 if (ckWARN(WARN_SYNTAX))
2089 Perl_warner(aTHX_ packWARN(WARN_SYNTAX), "\\%c better written as $%c", *s, *s);
2094 /* string-change backslash escapes */
2095 if (PL_lex_inwhat != OP_TRANS && *s && strchr("lLuUEQ", *s)) {
2099 /* skip any other backslash escapes in a pattern */
2100 else if (PL_lex_inpat) {
2101 *d++ = NATIVE_TO_NEED(has_utf8,'\\');
2102 goto default_action;
2105 /* if we get here, it's either a quoted -, or a digit */
2108 /* quoted - in transliterations */
2110 if (PL_lex_inwhat == OP_TRANS) {
2117 if ((isALPHA(*s) || isDIGIT(*s)) &&
2119 Perl_warner(aTHX_ packWARN(WARN_MISC),
2120 "Unrecognized escape \\%c passed through",
2122 /* default action is to copy the quoted character */
2123 goto default_action;
2126 /* \132 indicates an octal constant */
2127 case '0': case '1': case '2': case '3':
2128 case '4': case '5': case '6': case '7':
2132 uv = grok_oct(s, &len, &flags, NULL);
2135 goto NUM_ESCAPE_INSERT;
2137 /* \x24 indicates a hex constant */
2141 char* const e = strchr(s, '}');
2142 I32 flags = PERL_SCAN_ALLOW_UNDERSCORES |
2143 PERL_SCAN_DISALLOW_PREFIX;
2148 yyerror("Missing right brace on \\x{}");
2152 uv = grok_hex(s, &len, &flags, NULL);
2158 I32 flags = PERL_SCAN_DISALLOW_PREFIX;
2159 uv = grok_hex(s, &len, &flags, NULL);
2165 /* Insert oct or hex escaped character.
2166 * There will always enough room in sv since such
2167 * escapes will be longer than any UTF-8 sequence
2168 * they can end up as. */
2170 /* We need to map to chars to ASCII before doing the tests
2173 if (!UNI_IS_INVARIANT(NATIVE_TO_UNI(uv))) {
2174 if (!has_utf8 && uv > 255) {
2175 /* Might need to recode whatever we have
2176 * accumulated so far if it contains any
2179 * (Can't we keep track of that and avoid
2180 * this rescan? --jhi)
2184 for (c = (U8 *) SvPVX(sv); c < (U8 *)d; c++) {
2185 if (!NATIVE_IS_INVARIANT(*c)) {
2190 const STRLEN offset = d - SvPVX_const(sv);
2192 d = SvGROW(sv, SvLEN(sv) + hicount + 1) + offset;
2196 while (src >= (const U8 *)SvPVX_const(sv)) {
2197 if (!NATIVE_IS_INVARIANT(*src)) {
2198 const U8 ch = NATIVE_TO_ASCII(*src);
2199 *dst-- = (U8)UTF8_EIGHT_BIT_LO(ch);
2200 *dst-- = (U8)UTF8_EIGHT_BIT_HI(ch);
2210 if (has_utf8 || uv > 255) {
2211 d = (char*)uvchr_to_utf8((U8*)d, uv);
2213 if (PL_lex_inwhat == OP_TRANS &&
2214 PL_sublex_info.sub_op) {
2215 PL_sublex_info.sub_op->op_private |=
2216 (PL_lex_repl ? OPpTRANS_FROM_UTF
2220 if (uv > 255 && !dorange)
2221 native_range = FALSE;
2233 /* \N{LATIN SMALL LETTER A} is a named character */
2237 char* e = strchr(s, '}');
2244 yyerror("Missing right brace on \\N{}");
2248 if (e > s + 2 && s[1] == 'U' && s[2] == '+') {
2250 I32 flags = PERL_SCAN_ALLOW_UNDERSCORES |
2251 PERL_SCAN_DISALLOW_PREFIX;
2254 uv = grok_hex(s, &len, &flags, NULL);
2255 if ( e > s && len != (STRLEN)(e - s) ) {
2259 goto NUM_ESCAPE_INSERT;
2261 res = newSVpvn(s + 1, e - s - 1);
2262 type = newSVpvn(s - 2,e - s + 3);
2263 res = new_constant( NULL, 0, "charnames",
2264 res, NULL, SvPVX(type) );
2267 sv_utf8_upgrade(res);
2268 str = SvPV_const(res,len);
2269 #ifdef EBCDIC_NEVER_MIND
2270 /* charnames uses pack U and that has been
2271 * recently changed to do the below uni->native
2272 * mapping, so this would be redundant (and wrong,
2273 * the code point would be doubly converted).
2274 * But leave this in just in case the pack U change
2275 * gets revoked, but the semantics is still
2276 * desireable for charnames. --jhi */
2278 UV uv = utf8_to_uvchr((const U8*)str, 0);
2281 U8 tmpbuf[UTF8_MAXBYTES+1], *d;
2283 d = uvchr_to_utf8(tmpbuf, UNI_TO_NATIVE(uv));
2284 sv_setpvn(res, (char *)tmpbuf, d - tmpbuf);
2285 str = SvPV_const(res, len);
2289 if (!has_utf8 && SvUTF8(res)) {
2290 const char * const ostart = SvPVX_const(sv);
2291 SvCUR_set(sv, d - ostart);
2294 sv_utf8_upgrade(sv);
2295 /* this just broke our allocation above... */
2296 SvGROW(sv, (STRLEN)(send - start));
2297 d = SvPVX(sv) + SvCUR(sv);
2300 if (len > (STRLEN)(e - s + 4)) { /* I _guess_ 4 is \N{} --jhi */
2301 const char * const odest = SvPVX_const(sv);
2303 SvGROW(sv, (SvLEN(sv) + len - (e - s + 4)));
2304 d = SvPVX(sv) + (d - odest);
2308 native_range = FALSE; /* \N{} is guessed to be Unicode */
2310 Copy(str, d, len, char);
2317 yyerror("Missing braces on \\N{}");
2320 /* \c is a control character */
2329 *d++ = NATIVE_TO_NEED(has_utf8,toCTRL(c));
2332 yyerror("Missing control char name in \\c");
2336 /* printf-style backslashes, formfeeds, newlines, etc */
2338 *d++ = NATIVE_TO_NEED(has_utf8,'\b');
2341 *d++ = NATIVE_TO_NEED(has_utf8,'\n');
2344 *d++ = NATIVE_TO_NEED(has_utf8,'\r');
2347 *d++ = NATIVE_TO_NEED(has_utf8,'\f');
2350 *d++ = NATIVE_TO_NEED(has_utf8,'\t');
2353 *d++ = ASCII_TO_NEED(has_utf8,'\033');
2356 *d++ = ASCII_TO_NEED(has_utf8,'\007');
2362 } /* end if (backslash) */
2369 /* If we started with encoded form, or already know we want it
2370 and then encode the next character */
2371 if ((has_utf8 || this_utf8) && !NATIVE_IS_INVARIANT((U8)(*s))) {
2373 const UV nextuv = (this_utf8) ? utf8n_to_uvchr((U8*)s, send - s, &len, 0) : (UV) ((U8) *s);
2374 const STRLEN need = UNISKIP(NATIVE_TO_UNI(nextuv));
2377 /* encoded value larger than old, need extra space (NOTE: SvCUR() not set here) */
2378 const STRLEN off = d - SvPVX_const(sv);
2379 d = SvGROW(sv, SvLEN(sv) + (need-len)) + off;
2381 d = (char*)uvchr_to_utf8((U8*)d, nextuv);
2384 if (uv > 255 && !dorange)
2385 native_range = FALSE;
2389 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2391 } /* while loop to process each character */
2393 /* terminate the string and set up the sv */
2395 SvCUR_set(sv, d - SvPVX_const(sv));
2396 if (SvCUR(sv) >= SvLEN(sv))
2397 Perl_croak(aTHX_ "panic: constant overflowed allocated space");
2400 if (PL_encoding && !has_utf8) {
2401 sv_recode_to_utf8(sv, PL_encoding);
2407 if (PL_lex_inwhat == OP_TRANS && PL_sublex_info.sub_op) {
2408 PL_sublex_info.sub_op->op_private |=
2409 (PL_lex_repl ? OPpTRANS_FROM_UTF : OPpTRANS_TO_UTF);
2413 /* shrink the sv if we allocated more than we used */
2414 if (SvCUR(sv) + 5 < SvLEN(sv)) {
2415 SvPV_shrink_to_cur(sv);
2418 /* return the substring (via yylval) only if we parsed anything */
2419 if (s > PL_bufptr) {
2420 if ( PL_hints & ( PL_lex_inpat ? HINT_NEW_RE : HINT_NEW_STRING ) )
2421 sv = new_constant(start, s - start,
2422 (const char *)(PL_lex_inpat ? "qr" : "q"),
2425 (( PL_lex_inwhat == OP_TRANS
2427 : ( (PL_lex_inwhat == OP_SUBST && !PL_lex_inpat)
2430 yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
2437 * Returns TRUE if there's more to the expression (e.g., a subscript),
2440 * It deals with "$foo[3]" and /$foo[3]/ and /$foo[0123456789$]+/
2442 * ->[ and ->{ return TRUE
2443 * { and [ outside a pattern are always subscripts, so return TRUE
2444 * if we're outside a pattern and it's not { or [, then return FALSE
2445 * if we're in a pattern and the first char is a {
2446 * {4,5} (any digits around the comma) returns FALSE
2447 * if we're in a pattern and the first char is a [
2449 * [SOMETHING] has a funky algorithm to decide whether it's a
2450 * character class or not. It has to deal with things like
2451 * /$foo[-3]/ and /$foo[$bar]/ as well as /$foo[$\d]+/
2452 * anything else returns TRUE
2455 /* This is the one truly awful dwimmer necessary to conflate C and sed. */
2458 S_intuit_more(pTHX_ register char *s)
2461 if (PL_lex_brackets)
2463 if (*s == '-' && s[1] == '>' && (s[2] == '[' || s[2] == '{'))
2465 if (*s != '{' && *s != '[')
2470 /* In a pattern, so maybe we have {n,m}. */
2487 /* On the other hand, maybe we have a character class */
2490 if (*s == ']' || *s == '^')
2493 /* this is terrifying, and it works */
2494 int weight = 2; /* let's weigh the evidence */
2496 unsigned char un_char = 255, last_un_char;
2497 const char * const send = strchr(s,']');
2498 char tmpbuf[sizeof PL_tokenbuf * 4];
2500 if (!send) /* has to be an expression */
2503 Zero(seen,256,char);
2506 else if (isDIGIT(*s)) {
2508 if (isDIGIT(s[1]) && s[2] == ']')
2514 for (; s < send; s++) {
2515 last_un_char = un_char;
2516 un_char = (unsigned char)*s;
2521 weight -= seen[un_char] * 10;
2522 if (isALNUM_lazy_if(s+1,UTF)) {
2524 scan_ident(s, send, tmpbuf, sizeof tmpbuf, FALSE);
2525 len = (int)strlen(tmpbuf);
2526 if (len > 1 && gv_fetchpvn_flags(tmpbuf, len, 0, SVt_PV))
2531 else if (*s == '$' && s[1] &&
2532 strchr("[#!%*<>()-=",s[1])) {
2533 if (/*{*/ strchr("])} =",s[2]))
2542 if (strchr("wds]",s[1]))
2544 else if (seen[(U8)'\''] || seen[(U8)'"'])
2546 else if (strchr("rnftbxcav",s[1]))
2548 else if (isDIGIT(s[1])) {
2550 while (s[1] && isDIGIT(s[1]))
2560 if (strchr("aA01! ",last_un_char))
2562 if (strchr("zZ79~",s[1]))
2564 if (last_un_char == 255 && (isDIGIT(s[1]) || s[1] == '$'))
2565 weight -= 5; /* cope with negative subscript */
2568 if (!isALNUM(last_un_char)
2569 && !(last_un_char == '$' || last_un_char == '@'
2570 || last_un_char == '&')
2571 && isALPHA(*s) && s[1] && isALPHA(s[1])) {
2576 if (keyword(tmpbuf, d - tmpbuf, 0))
2579 if (un_char == last_un_char + 1)
2581 weight -= seen[un_char];
2586 if (weight >= 0) /* probably a character class */
2596 * Does all the checking to disambiguate
2598 * between foo(bar) and bar->foo. Returns 0 if not a method, otherwise
2599 * FUNCMETH (bar->foo(args)) or METHOD (bar->foo args).
2601 * First argument is the stuff after the first token, e.g. "bar".
2603 * Not a method if bar is a filehandle.
2604 * Not a method if foo is a subroutine prototyped to take a filehandle.
2605 * Not a method if it's really "Foo $bar"
2606 * Method if it's "foo $bar"
2607 * Not a method if it's really "print foo $bar"
2608 * Method if it's really "foo package::" (interpreted as package->foo)
2609 * Not a method if bar is known to be a subroutine ("sub bar; foo bar")
2610 * Not a method if bar is a filehandle or package, but is quoted with
2615 S_intuit_method(pTHX_ char *start, GV *gv, CV *cv)
2618 char *s = start + (*start == '$');
2619 char tmpbuf[sizeof PL_tokenbuf];
2627 if (SvTYPE(gv) == SVt_PVGV && GvIO(gv))
2631 const char *proto = SvPVX_const(cv);
2642 s = scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
2643 /* start is the beginning of the possible filehandle/object,
2644 * and s is the end of it
2645 * tmpbuf is a copy of it
2648 if (*start == '$') {
2649 if (gv || PL_last_lop_op == OP_PRINT || isUPPER(*PL_tokenbuf))
2652 len = start - SvPVX(PL_linestr);
2656 start = SvPVX(PL_linestr) + len;
2660 return *s == '(' ? FUNCMETH : METHOD;
2662 if (!keyword(tmpbuf, len, 0)) {
2663 if (len > 2 && tmpbuf[len - 2] == ':' && tmpbuf[len - 1] == ':') {
2667 soff = s - SvPVX(PL_linestr);
2671 indirgv = gv_fetchpvn_flags(tmpbuf, len, 0, SVt_PVCV);
2672 if (indirgv && GvCVu(indirgv))
2674 /* filehandle or package name makes it a method */
2675 if (!gv || GvIO(indirgv) || gv_stashpvn(tmpbuf, len, 0)) {
2677 soff = s - SvPVX(PL_linestr);
2680 if ((PL_bufend - s) >= 2 && *s == '=' && *(s+1) == '>')
2681 return 0; /* no assumptions -- "=>" quotes bearword */
2683 start_force(PL_curforce);
2684 NEXTVAL_NEXTTOKE.opval = (OP*)newSVOP(OP_CONST, 0,
2685 newSVpvn(tmpbuf,len));
2686 NEXTVAL_NEXTTOKE.opval->op_private = OPpCONST_BARE;
2688 curmad('X', newSVpvn(start,SvPVX(PL_linestr) + soff - start));
2693 PL_bufptr = SvPVX(PL_linestr) + soff; /* restart before space */
2695 return *s == '(' ? FUNCMETH : METHOD;
2703 * Return a string of Perl code to load the debugger. If PERL5DB
2704 * is set, it will return the contents of that, otherwise a
2705 * compile-time require of perl5db.pl.
2713 const char * const pdb = PerlEnv_getenv("PERL5DB");
2717 SETERRNO(0,SS_NORMAL);
2718 return "BEGIN { require 'perl5db.pl' }";
2724 /* Encoded script support. filter_add() effectively inserts a
2725 * 'pre-processing' function into the current source input stream.
2726 * Note that the filter function only applies to the current source file
2727 * (e.g., it will not affect files 'require'd or 'use'd by this one).
2729 * The datasv parameter (which may be NULL) can be used to pass
2730 * private data to this instance of the filter. The filter function
2731 * can recover the SV using the FILTER_DATA macro and use it to
2732 * store private buffers and state information.
2734 * The supplied datasv parameter is upgraded to a PVIO type
2735 * and the IoDIRP/IoANY field is used to store the function pointer,
2736 * and IOf_FAKE_DIRP is enabled on datasv to mark this as such.
2737 * Note that IoTOP_NAME, IoFMT_NAME, IoBOTTOM_NAME, if set for
2738 * private use must be set using malloc'd pointers.
2742 Perl_filter_add(pTHX_ filter_t funcp, SV *datasv)
2748 if (!PL_rsfp_filters)
2749 PL_rsfp_filters = newAV();
2752 SvUPGRADE(datasv, SVt_PVIO);
2753 IoANY(datasv) = FPTR2DPTR(void *, funcp); /* stash funcp into spare field */
2754 IoFLAGS(datasv) |= IOf_FAKE_DIRP;
2755 DEBUG_P(PerlIO_printf(Perl_debug_log, "filter_add func %p (%s)\n",
2756 FPTR2DPTR(void *, IoANY(datasv)),
2757 SvPV_nolen(datasv)));
2758 av_unshift(PL_rsfp_filters, 1);
2759 av_store(PL_rsfp_filters, 0, datasv) ;
2764 /* Delete most recently added instance of this filter function. */
2766 Perl_filter_del(pTHX_ filter_t funcp)
2772 DEBUG_P(PerlIO_printf(Perl_debug_log, "filter_del func %p",
2773 FPTR2DPTR(void*, funcp)));
2775 if (!PL_rsfp_filters || AvFILLp(PL_rsfp_filters)<0)
2777 /* if filter is on top of stack (usual case) just pop it off */
2778 datasv = FILTER_DATA(AvFILLp(PL_rsfp_filters));
2779 if (IoANY(datasv) == FPTR2DPTR(void *, funcp)) {
2780 IoFLAGS(datasv) &= ~IOf_FAKE_DIRP;
2781 IoANY(datasv) = (void *)NULL;
2782 sv_free(av_pop(PL_rsfp_filters));
2786 /* we need to search for the correct entry and clear it */
2787 Perl_die(aTHX_ "filter_del can only delete in reverse order (currently)");
2791 /* Invoke the idxth filter function for the current rsfp. */
2792 /* maxlen 0 = read one text line */
2794 Perl_filter_read(pTHX_ int idx, SV *buf_sv, int maxlen)
2799 /* This API is bad. It should have been using unsigned int for maxlen.
2800 Not sure if we want to change the API, but if not we should sanity
2801 check the value here. */
2802 const unsigned int correct_length
2811 if (!PL_rsfp_filters)
2813 if (idx > AvFILLp(PL_rsfp_filters)) { /* Any more filters? */
2814 /* Provide a default input filter to make life easy. */
2815 /* Note that we append to the line. This is handy. */
2816 DEBUG_P(PerlIO_printf(Perl_debug_log,
2817 "filter_read %d: from rsfp\n", idx));
2818 if (correct_length) {
2821 const int old_len = SvCUR(buf_sv);
2823 /* ensure buf_sv is large enough */
2824 SvGROW(buf_sv, (STRLEN)(old_len + correct_length)) ;
2825 if ((len = PerlIO_read(PL_rsfp, SvPVX(buf_sv) + old_len,
2826 correct_length)) <= 0) {
2827 if (PerlIO_error(PL_rsfp))
2828 return -1; /* error */
2830 return 0 ; /* end of file */
2832 SvCUR_set(buf_sv, old_len + len) ;
2835 if (sv_gets(buf_sv, PL_rsfp, SvCUR(buf_sv)) == NULL) {
2836 if (PerlIO_error(PL_rsfp))
2837 return -1; /* error */
2839 return 0 ; /* end of file */
2842 return SvCUR(buf_sv);
2844 /* Skip this filter slot if filter has been deleted */
2845 if ( (datasv = FILTER_DATA(idx)) == &PL_sv_undef) {
2846 DEBUG_P(PerlIO_printf(Perl_debug_log,
2847 "filter_read %d: skipped (filter deleted)\n",
2849 return FILTER_READ(idx+1, buf_sv, correct_length); /* recurse */
2851 /* Get function pointer hidden within datasv */
2852 funcp = DPTR2FPTR(filter_t, IoANY(datasv));
2853 DEBUG_P(PerlIO_printf(Perl_debug_log,
2854 "filter_read %d: via function %p (%s)\n",
2855 idx, (void*)datasv, SvPV_nolen_const(datasv)));
2856 /* Call function. The function is expected to */
2857 /* call "FILTER_READ(idx+1, buf_sv)" first. */
2858 /* Return: <0:error, =0:eof, >0:not eof */
2859 return (*funcp)(aTHX_ idx, buf_sv, correct_length);
2863 S_filter_gets(pTHX_ register SV *sv, register PerlIO *fp, STRLEN append)
2866 #ifdef PERL_CR_FILTER
2867 if (!PL_rsfp_filters) {
2868 filter_add(S_cr_textfilter,NULL);
2871 if (PL_rsfp_filters) {
2873 SvCUR_set(sv, 0); /* start with empty line */
2874 if (FILTER_READ(0, sv, 0) > 0)
2875 return ( SvPVX(sv) ) ;
2880 return (sv_gets(sv, fp, append));
2884 S_find_in_my_stash(pTHX_ const char *pkgname, I32 len)
2889 if (len == 11 && *pkgname == '_' && strEQ(pkgname, "__PACKAGE__"))
2893 (pkgname[len - 2] == ':' && pkgname[len - 1] == ':') &&
2894 (gv = gv_fetchpvn_flags(pkgname, len, 0, SVt_PVHV)))
2896 return GvHV(gv); /* Foo:: */
2899 /* use constant CLASS => 'MyClass' */
2900 gv = gv_fetchpvn_flags(pkgname, len, 0, SVt_PVCV);
2901 if (gv && GvCV(gv)) {
2902 SV * const sv = cv_const_sv(GvCV(gv));
2904 pkgname = SvPV_nolen_const(sv);
2907 return gv_stashpv(pkgname, 0);
2911 * S_readpipe_override
2912 * Check whether readpipe() is overriden, and generates the appropriate
2913 * optree, provided sublex_start() is called afterwards.
2916 S_readpipe_override(pTHX)
2919 GV *gv_readpipe = gv_fetchpvs("readpipe", GV_NOTQUAL, SVt_PVCV);
2920 yylval.ival = OP_BACKTICK;
2922 && GvCVu(gv_readpipe) && GvIMPORTED_CV(gv_readpipe))
2924 ((gvp = (GV**)hv_fetchs(PL_globalstash, "readpipe", FALSE))
2925 && (gv_readpipe = *gvp) && isGV_with_GP(gv_readpipe)
2926 && GvCVu(gv_readpipe) && GvIMPORTED_CV(gv_readpipe)))
2928 PL_lex_op = (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
2929 append_elem(OP_LIST,
2930 newSVOP(OP_CONST, 0, &PL_sv_undef), /* value will be read later */
2931 newCVREF(0, newGVOP(OP_GV, 0, gv_readpipe))));
2941 * The intent of this yylex wrapper is to minimize the changes to the
2942 * tokener when we aren't interested in collecting madprops. It remains
2943 * to be seen how successful this strategy will be...
2950 char *s = PL_bufptr;
2952 /* make sure PL_thiswhite is initialized */
2956 /* just do what yylex would do on pending identifier; leave PL_thiswhite alone */
2957 if (PL_pending_ident)
2958 return S_pending_ident(aTHX);
2960 /* previous token ate up our whitespace? */
2961 if (!PL_lasttoke && PL_nextwhite) {
2962 PL_thiswhite = PL_nextwhite;
2966 /* isolate the token, and figure out where it is without whitespace */
2967 PL_realtokenstart = -1;
2971 assert(PL_curforce < 0);
2973 if (!PL_thismad || PL_thismad->mad_key == '^') { /* not forced already? */
2974 if (!PL_thistoken) {
2975 if (PL_realtokenstart < 0 || !CopLINE(PL_curcop))
2976 PL_thistoken = newSVpvs("");
2978 char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
2979 PL_thistoken = newSVpvn(tstart, s - tstart);
2982 if (PL_thismad) /* install head */
2983 CURMAD('X', PL_thistoken);
2986 /* last whitespace of a sublex? */
2987 if (optype == ')' && PL_endwhite) {
2988 CURMAD('X', PL_endwhite);
2993 /* if no whitespace and we're at EOF, bail. Otherwise fake EOF below. */
2994 if (!PL_thiswhite && !PL_endwhite && !optype) {
2995 sv_free(PL_thistoken);
3000 /* put off final whitespace till peg */
3001 if (optype == ';' && !PL_rsfp) {
3002 PL_nextwhite = PL_thiswhite;
3005 else if (PL_thisopen) {
3006 CURMAD('q', PL_thisopen);
3008 sv_free(PL_thistoken);
3012 /* Store actual token text as madprop X */
3013 CURMAD('X', PL_thistoken);
3017 /* add preceding whitespace as madprop _ */
3018 CURMAD('_', PL_thiswhite);
3022 /* add quoted material as madprop = */
3023 CURMAD('=', PL_thisstuff);
3027 /* add terminating quote as madprop Q */
3028 CURMAD('Q', PL_thisclose);
3032 /* special processing based on optype */
3036 /* opval doesn't need a TOKEN since it can already store mp */
3047 append_madprops(PL_thismad, yylval.opval, 0);
3055 addmad(newMADsv('p', PL_endwhite), &PL_thismad, 0);
3064 /* remember any fake bracket that lexer is about to discard */
3065 if (PL_lex_brackets == 1 &&
3066 ((expectation)PL_lex_brackstack[0] & XFAKEBRACK))
3069 while (s < PL_bufend && (*s == ' ' || *s == '\t'))
3072 PL_thiswhite = newSVpvn(PL_bufptr, ++s - PL_bufptr);
3073 addmad(newMADsv('#', PL_thiswhite), &PL_thismad, 0);
3076 break; /* don't bother looking for trailing comment */
3085 /* attach a trailing comment to its statement instead of next token */
3089 if (PL_bufptr > PL_oldbufptr && PL_bufptr[-1] == optype) {
3091 while (s < PL_bufend && (*s == ' ' || *s == '\t'))
3093 if (*s == '\n' || *s == '#') {
3094 while (s < PL_bufend && *s != '\n')
3098 PL_thiswhite = newSVpvn(PL_bufptr, s - PL_bufptr);
3099 addmad(newMADsv('#', PL_thiswhite), &PL_thismad, 0);
3116 /* Create new token struct. Note: opvals return early above. */
3117 yylval.tkval = newTOKEN(optype, yylval, PL_thismad);
3124 S_tokenize_use(pTHX_ int is_use, char *s) {
3126 if (PL_expect != XSTATE)
3127 yyerror(Perl_form(aTHX_ "\"%s\" not allowed in expression",
3128 is_use ? "use" : "no"));
3130 if (isDIGIT(*s) || (*s == 'v' && isDIGIT(s[1]))) {
3131 s = force_version(s, TRUE);
3132 if (*s == ';' || (s = SKIPSPACE1(s), *s == ';')) {
3133 start_force(PL_curforce);
3134 NEXTVAL_NEXTTOKE.opval = NULL;
3137 else if (*s == 'v') {
3138 s = force_word(s,WORD,FALSE,TRUE,FALSE);
3139 s = force_version(s, FALSE);
3143 s = force_word(s,WORD,FALSE,TRUE,FALSE);
3144 s = force_version(s, FALSE);
3146 yylval.ival = is_use;
3150 static const char* const exp_name[] =
3151 { "OPERATOR", "TERM", "REF", "STATE", "BLOCK", "ATTRBLOCK",
3152 "ATTRTERM", "TERMBLOCK", "TERMORDORDOR"
3159 Works out what to call the token just pulled out of the input
3160 stream. The yacc parser takes care of taking the ops we return and
3161 stitching them into a tree.
3167 if read an identifier
3168 if we're in a my declaration
3169 croak if they tried to say my($foo::bar)
3170 build the ops for a my() declaration
3171 if it's an access to a my() variable
3172 are we in a sort block?
3173 croak if my($a); $a <=> $b
3174 build ops for access to a my() variable
3175 if in a dq string, and they've said @foo and we can't find @foo
3177 build ops for a bareword
3178 if we already built the token before, use it.
3183 #pragma segment Perl_yylex
3189 register char *s = PL_bufptr;
3194 /* orig_keyword, gvp, and gv are initialized here because
3195 * jump to the label just_a_word_zero can bypass their
3196 * initialization later. */
3197 I32 orig_keyword = 0;
3202 SV* tmp = newSVpvs("");
3203 PerlIO_printf(Perl_debug_log, "### %"IVdf":LEX_%s/X%s %s\n",
3204 (IV)CopLINE(PL_curcop),
3205 lex_state_names[PL_lex_state],
3206 exp_name[PL_expect],
3207 pv_display(tmp, s, strlen(s), 0, 60));
3210 /* check if there's an identifier for us to look at */
3211 if (PL_pending_ident)
3212 return REPORT(S_pending_ident(aTHX));
3214 /* no identifier pending identification */
3216 switch (PL_lex_state) {
3218 case LEX_NORMAL: /* Some compilers will produce faster */
3219 case LEX_INTERPNORMAL: /* code if we comment these out. */
3223 /* when we've already built the next token, just pull it out of the queue */
3227 yylval = PL_nexttoke[PL_lasttoke].next_val;
3229 PL_thismad = PL_nexttoke[PL_lasttoke].next_mad;
3230 PL_nexttoke[PL_lasttoke].next_mad = 0;
3231 if (PL_thismad && PL_thismad->mad_key == '_') {
3232 PL_thiswhite = (SV*)PL_thismad->mad_val;
3233 PL_thismad->mad_val = 0;
3234 mad_free(PL_thismad);
3239 PL_lex_state = PL_lex_defer;
3240 PL_expect = PL_lex_expect;
3241 PL_lex_defer = LEX_NORMAL;
3242 if (!PL_nexttoke[PL_lasttoke].next_type)
3247 yylval = PL_nextval[PL_nexttoke];
3249 PL_lex_state = PL_lex_defer;
3250 PL_expect = PL_lex_expect;
3251 PL_lex_defer = LEX_NORMAL;
3255 /* FIXME - can these be merged? */
3256 return(PL_nexttoke[PL_lasttoke].next_type);
3258 return REPORT(PL_nexttype[PL_nexttoke]);
3261 /* interpolated case modifiers like \L \U, including \Q and \E.
3262 when we get here, PL_bufptr is at the \
3264 case LEX_INTERPCASEMOD:
3266 if (PL_bufptr != PL_bufend && *PL_bufptr != '\\')
3267 Perl_croak(aTHX_ "panic: INTERPCASEMOD");
3269 /* handle \E or end of string */
3270 if (PL_bufptr == PL_bufend || PL_bufptr[1] == 'E') {
3272 if (PL_lex_casemods) {
3273 const char oldmod = PL_lex_casestack[--PL_lex_casemods];
3274 PL_lex_casestack[PL_lex_casemods] = '\0';
3276 if (PL_bufptr != PL_bufend
3277 && (oldmod == 'L' || oldmod == 'U' || oldmod == 'Q')) {
3279 PL_lex_state = LEX_INTERPCONCAT;
3282 PL_thistoken = newSVpvs("\\E");
3288 while (PL_bufptr != PL_bufend &&
3289 PL_bufptr[0] == '\\' && PL_bufptr[1] == 'E') {
3291 PL_thiswhite = newSVpvs("");
3292 sv_catpvn(PL_thiswhite, PL_bufptr, 2);
3296 if (PL_bufptr != PL_bufend)
3299 PL_lex_state = LEX_INTERPCONCAT;
3303 DEBUG_T({ PerlIO_printf(Perl_debug_log,
3304 "### Saw case modifier\n"); });
3306 if (s[1] == '\\' && s[2] == 'E') {
3309 PL_thiswhite = newSVpvs("");
3310 sv_catpvn(PL_thiswhite, PL_bufptr, 4);
3313 PL_lex_state = LEX_INTERPCONCAT;
3318 if (!PL_madskills) /* when just compiling don't need correct */
3319 if (strnEQ(s, "L\\u", 3) || strnEQ(s, "U\\l", 3))
3320 tmp = *s, *s = s[2], s[2] = (char)tmp; /* misordered... */
3321 if ((*s == 'L' || *s == 'U') &&
3322 (strchr(PL_lex_casestack, 'L') || strchr(PL_lex_casestack, 'U'))) {
3323 PL_lex_casestack[--PL_lex_casemods] = '\0';
3326 if (PL_lex_casemods > 10)
3327 Renew(PL_lex_casestack, PL_lex_casemods + 2, char);
3328 PL_lex_casestack[PL_lex_casemods++] = *s;
3329 PL_lex_casestack[PL_lex_casemods] = '\0';
3330 PL_lex_state = LEX_INTERPCONCAT;
3331 start_force(PL_curforce);
3332 NEXTVAL_NEXTTOKE.ival = 0;
3334 start_force(PL_curforce);
3336 NEXTVAL_NEXTTOKE.ival = OP_LCFIRST;
3338 NEXTVAL_NEXTTOKE.ival = OP_UCFIRST;
3340 NEXTVAL_NEXTTOKE.ival = OP_LC;
3342 NEXTVAL_NEXTTOKE.ival = OP_UC;
3344 NEXTVAL_NEXTTOKE.ival = OP_QUOTEMETA;
3346 Perl_croak(aTHX_ "panic: yylex");
3348 SV* const tmpsv = newSVpvs("");
3349 Perl_sv_catpvf(aTHX_ tmpsv, "\\%c", *s);
3355 if (PL_lex_starts) {
3361 sv_free(PL_thistoken);
3362 PL_thistoken = newSVpvs("");
3365 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
3366 if (PL_lex_casemods == 1 && PL_lex_inpat)
3375 case LEX_INTERPPUSH:
3376 return REPORT(sublex_push());
3378 case LEX_INTERPSTART:
3379 if (PL_bufptr == PL_bufend)
3380 return REPORT(sublex_done());
3381 DEBUG_T({ PerlIO_printf(Perl_debug_log,
3382 "### Interpolated variable\n"); });
3384 PL_lex_dojoin = (*PL_bufptr == '@');
3385 PL_lex_state = LEX_INTERPNORMAL;
3386 if (PL_lex_dojoin) {
3387 start_force(PL_curforce);
3388 NEXTVAL_NEXTTOKE.ival = 0;
3390 start_force(PL_curforce);
3391 force_ident("\"", '$');
3392 start_force(PL_curforce);
3393 NEXTVAL_NEXTTOKE.ival = 0;
3395 start_force(PL_curforce);
3396 NEXTVAL_NEXTTOKE.ival = 0;
3398 start_force(PL_curforce);
3399 NEXTVAL_NEXTTOKE.ival = OP_JOIN; /* emulate join($", ...) */
3402 if (PL_lex_starts++) {
3407 sv_free(PL_thistoken);
3408 PL_thistoken = newSVpvs("");
3411 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
3412 if (!PL_lex_casemods && PL_lex_inpat)
3419 case LEX_INTERPENDMAYBE:
3420 if (intuit_more(PL_bufptr)) {
3421 PL_lex_state = LEX_INTERPNORMAL; /* false alarm, more expr */
3427 if (PL_lex_dojoin) {
3428 PL_lex_dojoin = FALSE;
3429 PL_lex_state = LEX_INTERPCONCAT;
3433 sv_free(PL_thistoken);
3434 PL_thistoken = newSVpvs("");
3439 if (PL_lex_inwhat == OP_SUBST && PL_linestr == PL_lex_repl
3440 && SvEVALED(PL_lex_repl))
3442 if (PL_bufptr != PL_bufend)
3443 Perl_croak(aTHX_ "Bad evalled substitution pattern");
3447 case LEX_INTERPCONCAT:
3449 if (PL_lex_brackets)
3450 Perl_croak(aTHX_ "panic: INTERPCONCAT");
3452 if (PL_bufptr == PL_bufend)
3453 return REPORT(sublex_done());
3455 if (SvIVX(PL_linestr) == '\'') {
3456 SV *sv = newSVsv(PL_linestr);
3459 else if ( PL_hints & HINT_NEW_RE )
3460 sv = new_constant(NULL, 0, "qr", sv, sv, "q");
3461 yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
3465 s = scan_const(PL_bufptr);
3467 PL_lex_state = LEX_INTERPCASEMOD;
3469 PL_lex_state = LEX_INTERPSTART;
3472 if (s != PL_bufptr) {
3473 start_force(PL_curforce);
3475 curmad('X', newSVpvn(PL_bufptr,s-PL_bufptr));
3477 NEXTVAL_NEXTTOKE = yylval;
3480 if (PL_lex_starts++) {
3484 sv_free(PL_thistoken);
3485 PL_thistoken = newSVpvs("");
3488 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
3489 if (!PL_lex_casemods && PL_lex_inpat)
3502 PL_lex_state = LEX_NORMAL;
3503 s = scan_formline(PL_bufptr);
3504 if (!PL_lex_formbrack)
3510 PL_oldoldbufptr = PL_oldbufptr;
3516 sv_free(PL_thistoken);
3519 PL_realtokenstart = s - SvPVX(PL_linestr); /* assume but undo on ws */
3523 if (isIDFIRST_lazy_if(s,UTF))
3525 Perl_croak(aTHX_ "Unrecognized character \\x%02X", *s & 255);
3528 goto fake_eof; /* emulate EOF on ^D or ^Z */
3537 if (PL_lex_brackets) {
3538 yyerror((const char *)
3540 ? "Format not terminated"
3541 : "Missing right curly or square bracket"));
3543 DEBUG_T( { PerlIO_printf(Perl_debug_log,
3544 "### Tokener got EOF\n");
3548 if (s++ < PL_bufend)
3549 goto retry; /* ignore stray nulls */
3552 if (!PL_in_eval && !PL_preambled) {
3553 PL_preambled = TRUE;
3558 sv_setpv(PL_linestr,incl_perldb());
3559 if (SvCUR(PL_linestr))
3560 sv_catpvs(PL_linestr,";");
3562 while(AvFILLp(PL_preambleav) >= 0) {
3563 SV *tmpsv = av_shift(PL_preambleav);
3564 sv_catsv(PL_linestr, tmpsv);
3565 sv_catpvs(PL_linestr, ";");
3568 sv_free((SV*)PL_preambleav);
3569 PL_preambleav = NULL;
3571 if (PL_minus_n || PL_minus_p) {
3572 sv_catpvs(PL_linestr, "LINE: while (<>) {");
3574 sv_catpvs(PL_linestr,"chomp;");
3577 if ((*PL_splitstr == '/' || *PL_splitstr == '\''
3578 || *PL_splitstr == '"')
3579 && strchr(PL_splitstr + 1, *PL_splitstr))
3580 Perl_sv_catpvf(aTHX_ PL_linestr, "our @F=split(%s);", PL_splitstr);
3582 /* "q\0${splitstr}\0" is legal perl. Yes, even NUL
3583 bytes can be used as quoting characters. :-) */
3584 const char *splits = PL_splitstr;
3585 sv_catpvs(PL_linestr, "our @F=split(q\0");
3588 if (*splits == '\\')
3589 sv_catpvn(PL_linestr, splits, 1);
3590 sv_catpvn(PL_linestr, splits, 1);
3591 } while (*splits++);
3592 /* This loop will embed the trailing NUL of
3593 PL_linestr as the last thing it does before
3595 sv_catpvs(PL_linestr, ");");
3599 sv_catpvs(PL_linestr,"our @F=split(' ');");
3603 sv_catpvs(PL_linestr,"use feature ':5.10';");
3604 sv_catpvs(PL_linestr, "\n");
3605 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3606 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3607 PL_last_lop = PL_last_uni = NULL;
3608 if (PERLDB_LINE && PL_curstash != PL_debstash)
3609 update_debugger_info(PL_linestr, NULL, 0);
3613 bof = PL_rsfp ? TRUE : FALSE;
3614 if ((s = filter_gets(PL_linestr, PL_rsfp, 0)) == NULL) {
3617 PL_realtokenstart = -1;
3620 if (PL_preprocess && !PL_in_eval)
3621 (void)PerlProc_pclose(PL_rsfp);
3622 else if ((PerlIO *)PL_rsfp == PerlIO_stdin())
3623 PerlIO_clearerr(PL_rsfp);
3625 (void)PerlIO_close(PL_rsfp);
3627 PL_doextract = FALSE;
3629 if (!PL_in_eval && (PL_minus_n || PL_minus_p)) {
3634 sv_setpv(PL_linestr,
3637 ? ";}continue{print;}" : ";}"));
3638 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3639 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3640 PL_last_lop = PL_last_uni = NULL;
3641 PL_minus_n = PL_minus_p = 0;
3644 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3645 PL_last_lop = PL_last_uni = NULL;
3646 sv_setpvn(PL_linestr,"",0);
3647 TOKEN(';'); /* not infinite loop because rsfp is NULL now */
3649 /* If it looks like the start of a BOM or raw UTF-16,
3650 * check if it in fact is. */
3656 #ifdef PERLIO_IS_STDIO
3657 # ifdef __GNU_LIBRARY__
3658 # if __GNU_LIBRARY__ == 1 /* Linux glibc5 */
3659 # define FTELL_FOR_PIPE_IS_BROKEN
3663 # if __GLIBC__ == 1 /* maybe some glibc5 release had it like this? */
3664 # define FTELL_FOR_PIPE_IS_BROKEN
3669 #ifdef FTELL_FOR_PIPE_IS_BROKEN
3670 /* This loses the possibility to detect the bof
3671 * situation on perl -P when the libc5 is being used.
3672 * Workaround? Maybe attach some extra state to PL_rsfp?
3675 bof = PerlIO_tell(PL_rsfp) == SvCUR(PL_linestr);
3677 bof = PerlIO_tell(PL_rsfp) == (Off_t)SvCUR(PL_linestr);
3680 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3681 s = swallow_bom((U8*)s);
3685 /* Incest with pod. */
3688 sv_catsv(PL_thiswhite, PL_linestr);
3690 if (*s == '=' && strnEQ(s, "=cut", 4) && !isALPHA(s[4])) {
3691 sv_setpvn(PL_linestr, "", 0);
3692 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3693 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3694 PL_last_lop = PL_last_uni = NULL;
3695 PL_doextract = FALSE;
3699 } while (PL_doextract);
3700 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = s;
3701 if (PERLDB_LINE && PL_curstash != PL_debstash)
3702 update_debugger_info(PL_linestr, NULL, 0);
3703 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3704 PL_last_lop = PL_last_uni = NULL;
3705 if (CopLINE(PL_curcop) == 1) {
3706 while (s < PL_bufend && isSPACE(*s))
3708 if (*s == ':' && s[1] != ':') /* for csh execing sh scripts */
3712 PL_thiswhite = newSVpvn(PL_linestart, s - PL_linestart);
3716 if (*s == '#' && *(s+1) == '!')
3718 #ifdef ALTERNATE_SHEBANG
3720 static char const as[] = ALTERNATE_SHEBANG;
3721 if (*s == as[0] && strnEQ(s, as, sizeof(as) - 1))
3722 d = s + (sizeof(as) - 1);
3724 #endif /* ALTERNATE_SHEBANG */
3733 while (*d && !isSPACE(*d))
3737 #ifdef ARG_ZERO_IS_SCRIPT
3738 if (ipathend > ipath) {
3740 * HP-UX (at least) sets argv[0] to the script name,
3741 * which makes $^X incorrect. And Digital UNIX and Linux,
3742 * at least, set argv[0] to the basename of the Perl
3743 * interpreter. So, having found "#!", we'll set it right.
3745 SV * const x = GvSV(gv_fetchpvs("\030", GV_ADD|GV_NOTQUAL,
3747 assert(SvPOK(x) || SvGMAGICAL(x));
3748 if (sv_eq(x, CopFILESV(PL_curcop))) {
3749 sv_setpvn(x, ipath, ipathend - ipath);
3755 const char *bstart = SvPV_const(CopFILESV(PL_curcop),blen);
3756 const char * const lstart = SvPV_const(x,llen);
3758 bstart += blen - llen;
3759 if (strnEQ(bstart, lstart, llen) && bstart[-1] == '/') {
3760 sv_setpvn(x, ipath, ipathend - ipath);
3765 TAINT_NOT; /* $^X is always tainted, but that's OK */
3767 #endif /* ARG_ZERO_IS_SCRIPT */
3772 d = instr(s,"perl -");
3774 d = instr(s,"perl");
3776 /* avoid getting into infinite loops when shebang
3777 * line contains "Perl" rather than "perl" */
3779 for (d = ipathend-4; d >= ipath; --d) {
3780 if ((*d == 'p' || *d == 'P')
3781 && !ibcmp(d, "perl", 4))
3791 #ifdef ALTERNATE_SHEBANG
3793 * If the ALTERNATE_SHEBANG on this system starts with a
3794 * character that can be part of a Perl expression, then if
3795 * we see it but not "perl", we're probably looking at the
3796 * start of Perl code, not a request to hand off to some
3797 * other interpreter. Similarly, if "perl" is there, but
3798 * not in the first 'word' of the line, we assume the line
3799 * contains the start of the Perl program.
3801 if (d && *s != '#') {
3802 const char *c = ipath;
3803 while (*c && !strchr("; \t\r\n\f\v#", *c))
3806 d = NULL; /* "perl" not in first word; ignore */
3808 *s = '#'; /* Don't try to parse shebang line */
3810 #endif /* ALTERNATE_SHEBANG */
3811 #ifndef MACOS_TRADITIONAL
3816 !instr(s,"indir") &&
3817 instr(PL_origargv[0],"perl"))
3824 while (s < PL_bufend && isSPACE(*s))
3826 if (s < PL_bufend) {
3827 Newxz(newargv,PL_origargc+3,char*);
3829 while (s < PL_bufend && !isSPACE(*s))
3832 Copy(PL_origargv+1, newargv+2, PL_origargc+1, char*);
3835 newargv = PL_origargv;
3838 PerlProc_execv(ipath, EXEC_ARGV_CAST(newargv));
3840 Perl_croak(aTHX_ "Can't exec %s", ipath);
3844 while (*d && !isSPACE(*d))
3846 while (SPACE_OR_TAB(*d))
3850 const bool switches_done = PL_doswitches;
3851 const U32 oldpdb = PL_perldb;
3852 const bool oldn = PL_minus_n;
3853 const bool oldp = PL_minus_p;
3856 if (*d == 'M' || *d == 'm' || *d == 'C') {
3857 const char * const m = d;
3858 while (*d && !isSPACE(*d))
3860 Perl_croak(aTHX_ "Too late for \"-%.*s\" option",
3863 d = moreswitches(d);
3865 if (PL_doswitches && !switches_done) {
3866 int argc = PL_origargc;
3867 char **argv = PL_origargv;
3870 } while (argc && argv[0][0] == '-' && argv[0][1]);
3871 init_argv_symbols(argc,argv);
3873 if ((PERLDB_LINE && !oldpdb) ||
3874 ((PL_minus_n || PL_minus_p) && !(oldn || oldp)))
3875 /* if we have already added "LINE: while (<>) {",
3876 we must not do it again */
3878 sv_setpvn(PL_linestr, "", 0);
3879 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3880 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3881 PL_last_lop = PL_last_uni = NULL;
3882 PL_preambled = FALSE;
3884 (void)gv_fetchfile(PL_origfilename);
3891 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
3893 PL_lex_state = LEX_FORMLINE;
3898 #ifdef PERL_STRICT_CR
3899 Perl_warn(aTHX_ "Illegal character \\%03o (carriage return)", '\r');
3901 "\t(Maybe you didn't strip carriage returns after a network transfer?)\n");
3903 case ' ': case '\t': case '\f': case 013:
3904 #ifdef MACOS_TRADITIONAL
3908 PL_realtokenstart = -1;
3910 PL_thiswhite = newSVpvs("");
3911 sv_catpvn(PL_thiswhite, s, 1);
3918 PL_realtokenstart = -1;
3922 if (PL_lex_state != LEX_NORMAL || (PL_in_eval && !PL_rsfp)) {
3923 if (*s == '#' && s == PL_linestart && PL_in_eval && !PL_rsfp) {
3924 /* handle eval qq[#line 1 "foo"\n ...] */
3925 CopLINE_dec(PL_curcop);
3928 if (PL_madskills && !PL_lex_formbrack && !PL_in_eval) {
3930 if (!PL_in_eval || PL_rsfp)
3935 while (d < PL_bufend && *d != '\n')
3939 else if (d > PL_bufend) /* Found by Ilya: feed random input to Perl. */
3940 Perl_croak(aTHX_ "panic: input overflow");
3943 PL_thiswhite = newSVpvn(s, d - s);
3948 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
3950 PL_lex_state = LEX_FORMLINE;
3956 if (PL_madskills && CopLINE(PL_curcop) >= 1 && !PL_lex_formbrack) {
3957 if (CopLINE(PL_curcop) == 1 && s[0] == '#' && s[1] == '!') {
3960 TOKEN(PEG); /* make sure any #! line is accessible */
3965 /* if (PL_madskills && PL_lex_formbrack) { */
3967 while (d < PL_bufend && *d != '\n')
3971 else if (d > PL_bufend) /* Found by Ilya: feed random input to Perl. */
3972 Perl_croak(aTHX_ "panic: input overflow");
3973 if (PL_madskills && CopLINE(PL_curcop) >= 1) {
3975 PL_thiswhite = newSVpvs("");
3976 if (CopLINE(PL_curcop) == 1) {
3977 sv_setpvn(PL_thiswhite, "", 0);
3980 sv_catpvn(PL_thiswhite, s, d - s);
3994 if (s[1] && isALPHA(s[1]) && !isALNUM(s[2])) {
4002 while (s < PL_bufend && SPACE_OR_TAB(*s))
4005 if (strnEQ(s,"=>",2)) {
4006 s = force_word(PL_bufptr,WORD,FALSE,FALSE,FALSE);
4007 DEBUG_T( { printbuf("### Saw unary minus before =>, forcing word %s\n", s); } );
4008 OPERATOR('-'); /* unary minus */
4010 PL_last_uni = PL_oldbufptr;
4012 case 'r': ftst = OP_FTEREAD; break;
4013 case 'w': ftst = OP_FTEWRITE; break;
4014 case 'x': ftst = OP_FTEEXEC; break;
4015 case 'o': ftst = OP_FTEOWNED; break;
4016 case 'R': ftst = OP_FTRREAD; break;
4017 case 'W': ftst = OP_FTRWRITE; break;
4018 case 'X': ftst = OP_FTREXEC; break;
4019 case 'O': ftst = OP_FTROWNED; break;
4020 case 'e': ftst = OP_FTIS; break;
4021 case 'z': ftst = OP_FTZERO; break;
4022 case 's': ftst = OP_FTSIZE; break;
4023 case 'f': ftst = OP_FTFILE; break;
4024 case 'd': ftst = OP_FTDIR; break;
4025 case 'l': ftst = OP_FTLINK; break;
4026 case 'p': ftst = OP_FTPIPE; break;
4027 case 'S': ftst = OP_FTSOCK; break;
4028 case 'u': ftst = OP_FTSUID; break;
4029 case 'g': ftst = OP_FTSGID; break;
4030 case 'k': ftst = OP_FTSVTX; break;
4031 case 'b': ftst = OP_FTBLK; break;
4032 case 'c': ftst = OP_FTCHR; break;
4033 case 't': ftst = OP_FTTTY; break;
4034 case 'T': ftst = OP_FTTEXT; break;
4035 case 'B': ftst = OP_FTBINARY; break;
4036 case 'M': case 'A': case 'C':
4037 gv_fetchpvs("\024", GV_ADD|GV_NOTQUAL, SVt_PV);
4039 case 'M': ftst = OP_FTMTIME; break;
4040 case 'A': ftst = OP_FTATIME; break;
4041 case 'C': ftst = OP_FTCTIME; break;
4049 PL_last_lop_op = (OPCODE)ftst;
4050 DEBUG_T( { PerlIO_printf(Perl_debug_log,
4051 "### Saw file test %c\n", (int)tmp);
4056 /* Assume it was a minus followed by a one-letter named
4057 * subroutine call (or a -bareword), then. */
4058 DEBUG_T( { PerlIO_printf(Perl_debug_log,
4059 "### '-%c' looked like a file test but was not\n",
4066 const char tmp = *s++;
4069 if (PL_expect == XOPERATOR)
4074 else if (*s == '>') {
4077 if (isIDFIRST_lazy_if(s,UTF)) {
4078 s = force_word(s,METHOD,FALSE,TRUE,FALSE);
4086 if (PL_expect == XOPERATOR)
4089 if (isSPACE(*s) || !isSPACE(*PL_bufptr))
4091 OPERATOR('-'); /* unary minus */
4097 const char tmp = *s++;
4100 if (PL_expect == XOPERATOR)
4105 if (PL_expect == XOPERATOR)
4108 if (isSPACE(*s) || !isSPACE(*PL_bufptr))
4115 if (PL_expect != XOPERATOR) {
4116 s = scan_ident(s, PL_bufend, PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
4117 PL_expect = XOPERATOR;
4118 force_ident(PL_tokenbuf, '*');
4131 if (PL_expect == XOPERATOR) {
4135 PL_tokenbuf[0] = '%';
4136 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1,
4137 sizeof PL_tokenbuf - 1, FALSE);
4138 if (!PL_tokenbuf[1]) {
4141 PL_pending_ident = '%';
4152 && (PL_expect == XOPERATOR || PL_expect == XTERMORDORDOR))
4159 const char tmp = *s++;
4165 goto just_a_word_zero_gv;
4168 switch (PL_expect) {
4174 if (!PL_in_my || PL_lex_state != LEX_NORMAL)
4176 PL_bufptr = s; /* update in case we back off */
4182 PL_expect = XTERMBLOCK;
4185 stuffstart = s - SvPVX(PL_linestr) - 1;
4189 while (isIDFIRST_lazy_if(s,UTF)) {
4192 d = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
4193 if (isLOWER(*s) && (tmp = keyword(PL_tokenbuf, len, 0))) {
4194 if (tmp < 0) tmp = -tmp;
4209 sv = newSVpvn(s, len);
4211 d = scan_str(d,TRUE,TRUE);
4213 /* MUST advance bufptr here to avoid bogus
4214 "at end of line" context messages from yyerror().
4216 PL_bufptr = s + len;
4217 yyerror("Unterminated attribute parameter in attribute list");
4221 return REPORT(0); /* EOF indicator */
4225 sv_catsv(sv, PL_lex_stuff);
4226 attrs = append_elem(OP_LIST, attrs,
4227 newSVOP(OP_CONST, 0, sv));
4228 SvREFCNT_dec(PL_lex_stuff);
4229 PL_lex_stuff = NULL;
4232 if (len == 6 && strnEQ(SvPVX(sv), "unique", len)) {
4234 if (PL_in_my == KEY_our) {
4236 GvUNIQUE_on(cGVOPx_gv(yylval.opval));
4238 /* skip to avoid loading attributes.pm */
4240 deprecate(":unique");
4243 Perl_croak(aTHX_ "The 'unique' attribute may only be applied to 'our' variables");
4246 /* NOTE: any CV attrs applied here need to be part of
4247 the CVf_BUILTIN_ATTRS define in cv.h! */
4248 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "lvalue", len)) {
4250 CvLVALUE_on(PL_compcv);
4252 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "locked", len)) {
4254 CvLOCKED_on(PL_compcv);
4256 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "method", len)) {
4258 CvMETHOD_on(PL_compcv);
4260 else if (!PL_in_my && len == 9 && strnEQ(SvPVX(sv), "assertion", len)) {
4262 CvASSERTION_on(PL_compcv);
4264 /* After we've set the flags, it could be argued that
4265 we don't need to do the attributes.pm-based setting
4266 process, and shouldn't bother appending recognized
4267 flags. To experiment with that, uncomment the
4268 following "else". (Note that's already been
4269 uncommented. That keeps the above-applied built-in
4270 attributes from being intercepted (and possibly
4271 rejected) by a package's attribute routines, but is
4272 justified by the performance win for the common case
4273 of applying only built-in attributes.) */
4275 attrs = append_elem(OP_LIST, attrs,
4276 newSVOP(OP_CONST, 0,
4280 if (*s == ':' && s[1] != ':')
4283 break; /* require real whitespace or :'s */
4284 /* XXX losing whitespace on sequential attributes here */
4288 = (PL_expect == XOPERATOR ? '=' : '{'); /*'}(' for vi */
4289 if (*s != ';' && *s != '}' && *s != tmp
4290 && (tmp != '=' || *s != ')')) {
4291 const char q = ((*s == '\'') ? '"' : '\'');
4292 /* If here for an expression, and parsed no attrs, back
4294 if (tmp == '=' && !attrs) {
4298 /* MUST advance bufptr here to avoid bogus "at end of line"
4299 context messages from yyerror().
4302 yyerror( (const char *)
4304 ? Perl_form(aTHX_ "Invalid separator character "
4305 "%c%c%c in attribute list", q, *s, q)
4306 : "Unterminated attribute list" ) );
4314 start_force(PL_curforce);
4315 NEXTVAL_NEXTTOKE.opval = attrs;
4316 CURMAD('_', PL_nextwhite);
4321 PL_thistoken = newSVpvn(SvPVX(PL_linestr) + stuffstart,
4322 (s - SvPVX(PL_linestr)) - stuffstart);
4330 if (PL_last_lop == PL_oldoldbufptr || PL_last_uni == PL_oldoldbufptr)
4331 PL_oldbufptr = PL_oldoldbufptr; /* allow print(STDOUT 123) */
4339 const char tmp = *s++;
4344 const char tmp = *s++;
4352 if (PL_lex_brackets <= 0)
4353 yyerror("Unmatched right square bracket");
4356 if (PL_lex_state == LEX_INTERPNORMAL) {
4357 if (PL_lex_brackets == 0) {
4358 if (*s != '[' && *s != '{' && (*s != '-' || s[1] != '>'))
4359 PL_lex_state = LEX_INTERPEND;
4366 if (PL_lex_brackets > 100) {
4367 Renew(PL_lex_brackstack, PL_lex_brackets + 10, char);
4369 switch (PL_expect) {
4371 if (PL_lex_formbrack) {
4375 if (PL_oldoldbufptr == PL_last_lop)
4376 PL_lex_brackstack[PL_lex_brackets++] = XTERM;
4378 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
4379 OPERATOR(HASHBRACK);
4381 while (s < PL_bufend && SPACE_OR_TAB(*s))
4384 PL_tokenbuf[0] = '\0';
4385 if (d < PL_bufend && *d == '-') {
4386 PL_tokenbuf[0] = '-';
4388 while (d < PL_bufend && SPACE_OR_TAB(*d))
4391 if (d < PL_bufend && isIDFIRST_lazy_if(d,UTF)) {
4392 d = scan_word(d, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1,
4394 while (d < PL_bufend && SPACE_OR_TAB(*d))
4397 const char minus = (PL_tokenbuf[0] == '-');
4398 s = force_word(s + minus, WORD, FALSE, TRUE, FALSE);
4406 PL_lex_brackstack[PL_lex_brackets++] = XSTATE;
4411 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
4416 if (PL_oldoldbufptr == PL_last_lop)
4417 PL_lex_brackstack[PL_lex_brackets++] = XTERM;
4419 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
4422 if (PL_expect == XREF && PL_lex_state == LEX_INTERPNORMAL) {
4424 /* This hack is to get the ${} in the message. */
4426 yyerror("syntax error");
4429 OPERATOR(HASHBRACK);
4431 /* This hack serves to disambiguate a pair of curlies
4432 * as being a block or an anon hash. Normally, expectation
4433 * determines that, but in cases where we're not in a
4434 * position to expect anything in particular (like inside
4435 * eval"") we have to resolve the ambiguity. This code
4436 * covers the case where the first term in the curlies is a
4437 * quoted string. Most other cases need to be explicitly
4438 * disambiguated by prepending a "+" before the opening
4439 * curly in order to force resolution as an anon hash.
4441 * XXX should probably propagate the outer expectation
4442 * into eval"" to rely less on this hack, but that could
4443 * potentially break current behavior of eval"".
4447 if (*s == '\'' || *s == '"' || *s == '`') {
4448 /* common case: get past first string, handling escapes */
4449 for (t++; t < PL_bufend && *t != *s;)
4450 if (*t++ == '\\' && (*t == '\\' || *t == *s))
4454 else if (*s == 'q') {
4457 || ((*t == 'q' || *t == 'x') && ++t < PL_bufend
4460 /* skip q//-like construct */
4462 char open, close, term;
4465 while (t < PL_bufend && isSPACE(*t))
4467 /* check for q => */
4468 if (t+1 < PL_bufend && t[0] == '=' && t[1] == '>') {
4469 OPERATOR(HASHBRACK);
4473 if (term && (tmps = strchr("([{< )]}> )]}>",term)))
4477 for (t++; t < PL_bufend; t++) {
4478 if (*t == '\\' && t+1 < PL_bufend && open != '\\')
4480 else if (*t == open)
4484 for (t++; t < PL_bufend; t++) {
4485 if (*t == '\\' && t+1 < PL_bufend)
4487 else if (*t == close && --brackets <= 0)
4489 else if (*t == open)
4496 /* skip plain q word */
4497 while (t < PL_bufend && isALNUM_lazy_if(t,UTF))
4500 else if (isALNUM_lazy_if(t,UTF)) {
4502 while (t < PL_bufend && isALNUM_lazy_if(t,UTF))
4505 while (t < PL_bufend && isSPACE(*t))
4507 /* if comma follows first term, call it an anon hash */
4508 /* XXX it could be a comma expression with loop modifiers */
4509 if (t < PL_bufend && ((*t == ',' && (*s == 'q' || !isLOWER(*s)))
4510 || (*t == '=' && t[1] == '>')))
4511 OPERATOR(HASHBRACK);
4512 if (PL_expect == XREF)
4515 PL_lex_brackstack[PL_lex_brackets-1] = XSTATE;
4521 yylval.ival = CopLINE(PL_curcop);
4522 if (isSPACE(*s) || *s == '#')
4523 PL_copline = NOLINE; /* invalidate current command line number */
4528 if (PL_lex_brackets <= 0)
4529 yyerror("Unmatched right curly bracket");
4531 PL_expect = (expectation)PL_lex_brackstack[--PL_lex_brackets];
4532 if (PL_lex_brackets < PL_lex_formbrack && PL_lex_state != LEX_INTERPNORMAL)
4533 PL_lex_formbrack = 0;
4534 if (PL_lex_state == LEX_INTERPNORMAL) {
4535 if (PL_lex_brackets == 0) {
4536 if (PL_expect & XFAKEBRACK) {
4537 PL_expect &= XENUMMASK;
4538 PL_lex_state = LEX_INTERPEND;
4543 PL_thiswhite = newSVpvs("");
4544 sv_catpvn(PL_thiswhite,"}",1);
4547 return yylex(); /* ignore fake brackets */
4549 if (*s == '-' && s[1] == '>')
4550 PL_lex_state = LEX_INTERPENDMAYBE;
4551 else if (*s != '[' && *s != '{')
4552 PL_lex_state = LEX_INTERPEND;
4555 if (PL_expect & XFAKEBRACK) {
4556 PL_expect &= XENUMMASK;
4558 return yylex(); /* ignore fake brackets */
4560 start_force(PL_curforce);
4562 curmad('X', newSVpvn(s-1,1));
4563 CURMAD('_', PL_thiswhite);
4568 PL_thistoken = newSVpvs("");
4576 if (PL_expect == XOPERATOR) {
4577 if (PL_bufptr == PL_linestart && ckWARN(WARN_SEMICOLON)
4578 && isIDFIRST_lazy_if(s,UTF))
4580 CopLINE_dec(PL_curcop);
4581 Perl_warner(aTHX_ packWARN(WARN_SEMICOLON), PL_warn_nosemi);
4582 CopLINE_inc(PL_curcop);
4587 s = scan_ident(s - 1, PL_bufend, PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
4589 PL_expect = XOPERATOR;
4590 force_ident(PL_tokenbuf, '&');
4594 yylval.ival = (OPpENTERSUB_AMPER<<8);
4606 const char tmp = *s++;
4613 if (tmp && isSPACE(*s) && ckWARN(WARN_SYNTAX)
4614 && strchr("+-*/%.^&|<",tmp))
4615 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
4616 "Reversed %c= operator",(int)tmp);
4618 if (PL_expect == XSTATE && isALPHA(tmp) &&
4619 (s == PL_linestart+1 || s[-2] == '\n') )
4621 if (PL_in_eval && !PL_rsfp) {
4626 if (strnEQ(s,"=cut",4)) {
4642 PL_thiswhite = newSVpvs("");
4643 sv_catpvn(PL_thiswhite, PL_linestart,
4644 PL_bufend - PL_linestart);
4648 PL_doextract = TRUE;
4652 if (PL_lex_brackets < PL_lex_formbrack) {
4654 #ifdef PERL_STRICT_CR
4655 while (SPACE_OR_TAB(*t))
4657 while (SPACE_OR_TAB(*t) || *t == '\r')
4660 if (*t == '\n' || *t == '#') {
4671 const char tmp = *s++;
4673 /* was this !=~ where !~ was meant?
4674 * warn on m:!=~\s+([/?]|[msy]\W|tr\W): */
4676 if (*s == '~' && ckWARN(WARN_SYNTAX)) {
4677 const char *t = s+1;
4679 while (t < PL_bufend && isSPACE(*t))
4682 if (*t == '/' || *t == '?' ||
4683 ((*t == 'm' || *t == 's' || *t == 'y')
4684 && !isALNUM(t[1])) ||
4685 (*t == 't' && t[1] == 'r' && !isALNUM(t[2])))
4686 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
4687 "!=~ should be !~");
4697 if (PL_expect != XOPERATOR) {
4698 if (s[1] != '<' && !strchr(s,'>'))
4701 s = scan_heredoc(s);
4703 s = scan_inputsymbol(s);
4704 TERM(sublex_start());
4710 SHop(OP_LEFT_SHIFT);
4724 const char tmp = *s++;
4726 SHop(OP_RIGHT_SHIFT);
4727 else if (tmp == '=')
4736 if (PL_expect == XOPERATOR) {
4737 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
4739 deprecate_old(commaless_variable_list);
4740 return REPORT(','); /* grandfather non-comma-format format */
4744 if (s[1] == '#' && (isIDFIRST_lazy_if(s+2,UTF) || strchr("{$:+-", s[2]))) {
4745 PL_tokenbuf[0] = '@';
4746 s = scan_ident(s + 1, PL_bufend, PL_tokenbuf + 1,
4747 sizeof PL_tokenbuf - 1, FALSE);
4748 if (PL_expect == XOPERATOR)
4749 no_op("Array length", s);
4750 if (!PL_tokenbuf[1])
4752 PL_expect = XOPERATOR;
4753 PL_pending_ident = '#';
4757 PL_tokenbuf[0] = '$';
4758 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1,
4759 sizeof PL_tokenbuf - 1, FALSE);
4760 if (PL_expect == XOPERATOR)
4762 if (!PL_tokenbuf[1]) {
4764 yyerror("Final $ should be \\$ or $name");
4768 /* This kludge not intended to be bulletproof. */
4769 if (PL_tokenbuf[1] == '[' && !PL_tokenbuf[2]) {
4770 yylval.opval = newSVOP(OP_CONST, 0,
4771 newSViv(CopARYBASE_get(&PL_compiling)));
4772 yylval.opval->op_private = OPpCONST_ARYBASE;
4778 const char tmp = *s;
4779 if (PL_lex_state == LEX_NORMAL)
4782 if ((PL_expect != XREF || PL_oldoldbufptr == PL_last_lop)
4783 && intuit_more(s)) {
4785 PL_tokenbuf[0] = '@';
4786 if (ckWARN(WARN_SYNTAX)) {
4789 while (isSPACE(*t) || isALNUM_lazy_if(t,UTF) || *t == '$')
4792 PL_bufptr = PEEKSPACE(PL_bufptr); /* XXX can realloc */
4793 while (t < PL_bufend && *t != ']')
4795 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
4796 "Multidimensional syntax %.*s not supported",
4797 (int)((t - PL_bufptr) + 1), PL_bufptr);
4801 else if (*s == '{') {
4803 PL_tokenbuf[0] = '%';
4804 if (strEQ(PL_tokenbuf+1, "SIG") && ckWARN(WARN_SYNTAX)
4805 && (t = strchr(s, '}')) && (t = strchr(t, '=')))
4807 char tmpbuf[sizeof PL_tokenbuf];
4810 } while (isSPACE(*t));
4811 if (isIDFIRST_lazy_if(t,UTF)) {
4813 t = scan_word(t, tmpbuf, sizeof tmpbuf, TRUE,
4817 if (*t == ';' && get_cvn_flags(tmpbuf, len, 0))
4818 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
4819 "You need to quote \"%s\"",
4826 PL_expect = XOPERATOR;
4827 if (PL_lex_state == LEX_NORMAL && isSPACE((char)tmp)) {
4828 const bool islop = (PL_last_lop == PL_oldoldbufptr);
4829 if (!islop || PL_last_lop_op == OP_GREPSTART)
4830 PL_expect = XOPERATOR;
4831 else if (strchr("$@\"'`q", *s))
4832 PL_expect = XTERM; /* e.g. print $fh "foo" */
4833 else if (strchr("&*<%", *s) && isIDFIRST_lazy_if(s+1,UTF))
4834 PL_expect = XTERM; /* e.g. print $fh &sub */
4835 else if (isIDFIRST_lazy_if(s,UTF)) {
4836 char tmpbuf[sizeof PL_tokenbuf];
4838 scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
4839 if ((t2 = keyword(tmpbuf, len, 0))) {
4840 /* binary operators exclude handle interpretations */
4852 PL_expect = XTERM; /* e.g. print $fh length() */
4857 PL_expect = XTERM; /* e.g. print $fh subr() */
4860 else if (isDIGIT(*s))
4861 PL_expect = XTERM; /* e.g. print $fh 3 */
4862 else if (*s == '.' && isDIGIT(s[1]))
4863 PL_expect = XTERM; /* e.g. print $fh .3 */
4864 else if ((*s == '?' || *s == '-' || *s == '+')
4865 && !isSPACE(s[1]) && s[1] != '=')
4866 PL_expect = XTERM; /* e.g. print $fh -1 */
4867 else if (*s == '/' && !isSPACE(s[1]) && s[1] != '='
4869 PL_expect = XTERM; /* e.g. print $fh /.../
4870 XXX except DORDOR operator
4872 else if (*s == '<' && s[1] == '<' && !isSPACE(s[2])
4874 PL_expect = XTERM; /* print $fh <<"EOF" */
4877 PL_pending_ident = '$';
4881 if (PL_expect == XOPERATOR)
4883 PL_tokenbuf[0] = '@';
4884 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1, FALSE);
4885 if (!PL_tokenbuf[1]) {
4888 if (PL_lex_state == LEX_NORMAL)
4890 if ((PL_expect != XREF || PL_oldoldbufptr == PL_last_lop) && intuit_more(s)) {
4892 PL_tokenbuf[0] = '%';
4894 /* Warn about @ where they meant $. */
4895 if (*s == '[' || *s == '{') {
4896 if (ckWARN(WARN_SYNTAX)) {
4897 const char *t = s + 1;
4898 while (*t && (isALNUM_lazy_if(t,UTF) || strchr(" \t$#+-'\"", *t)))
4900 if (*t == '}' || *t == ']') {
4902 PL_bufptr = PEEKSPACE(PL_bufptr); /* XXX can realloc */
4903 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
4904 "Scalar value %.*s better written as $%.*s",
4905 (int)(t-PL_bufptr), PL_bufptr,
4906 (int)(t-PL_bufptr-1), PL_bufptr+1);
4911 PL_pending_ident = '@';
4914 case '/': /* may be division, defined-or, or pattern */
4915 if (PL_expect == XTERMORDORDOR && s[1] == '/') {
4919 case '?': /* may either be conditional or pattern */
4920 if(PL_expect == XOPERATOR) {
4928 /* A // operator. */
4938 /* Disable warning on "study /blah/" */
4939 if (PL_oldoldbufptr == PL_last_uni
4940 && (*PL_last_uni != 's' || s - PL_last_uni < 5
4941 || memNE(PL_last_uni, "study", 5)
4942 || isALNUM_lazy_if(PL_last_uni+5,UTF)
4945 s = scan_pat(s,OP_MATCH);
4946 TERM(sublex_start());
4950 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack
4951 #ifdef PERL_STRICT_CR
4954 && (s[1] == '\n' || (s[1] == '\r' && s[2] == '\n'))
4956 && (s == PL_linestart || s[-1] == '\n') )
4958 PL_lex_formbrack = 0;
4962 if (PL_expect == XOPERATOR || !isDIGIT(s[1])) {
4968 yylval.ival = OPf_SPECIAL;
4974 if (PL_expect != XOPERATOR)
4979 case '0': case '1': case '2': case '3': case '4':
4980 case '5': case '6': case '7': case '8': case '9':
4981 s = scan_num(s, &yylval);
4982 DEBUG_T( { printbuf("### Saw number in %s\n", s); } );
4983 if (PL_expect == XOPERATOR)
4988 s = scan_str(s,!!PL_madskills,FALSE);
4989 DEBUG_T( { printbuf("### Saw string before %s\n", s); } );
4990 if (PL_expect == XOPERATOR) {
4991 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
4993 deprecate_old(commaless_variable_list);
4994 return REPORT(','); /* grandfather non-comma-format format */
5001 yylval.ival = OP_CONST;
5002 TERM(sublex_start());
5005 s = scan_str(s,!!PL_madskills,FALSE);
5006 DEBUG_T( { printbuf("### Saw string before %s\n", s); } );
5007 if (PL_expect == XOPERATOR) {
5008 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
5010 deprecate_old(commaless_variable_list);
5011 return REPORT(','); /* grandfather non-comma-format format */
5018 yylval.ival = OP_CONST;
5019 /* FIXME. I think that this can be const if char *d is replaced by
5020 more localised variables. */
5021 for (d = SvPV(PL_lex_stuff, len); len; len--, d++) {
5022 if (*d == '$' || *d == '@' || *d == '\\' || !UTF8_IS_INVARIANT((U8)*d)) {
5023 yylval.ival = OP_STRINGIFY;
5027 TERM(sublex_start());
5030 s = scan_str(s,!!PL_madskills,FALSE);
5031 DEBUG_T( { printbuf("### Saw backtick string before %s\n", s); } );
5032 if (PL_expect == XOPERATOR)
5033 no_op("Backticks",s);
5036 readpipe_override();
5037 TERM(sublex_start());
5041 if (PL_lex_inwhat && isDIGIT(*s) && ckWARN(WARN_SYNTAX))
5042 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),"Can't use \\%c to mean $%c in expression",
5044 if (PL_expect == XOPERATOR)
5045 no_op("Backslash",s);
5049 if (isDIGIT(s[1]) && PL_expect != XOPERATOR) {
5050 char *start = s + 2;
5051 while (isDIGIT(*start) || *start == '_')
5053 if (*start == '.' && isDIGIT(start[1])) {
5054 s = scan_num(s, &yylval);
5057 /* avoid v123abc() or $h{v1}, allow C<print v10;> */
5058 else if (!isALPHA(*start) && (PL_expect == XTERM
5059 || PL_expect == XREF || PL_expect == XSTATE
5060 || PL_expect == XTERMORDORDOR)) {
5061 /* XXX Use gv_fetchpvn rather than stomping on a const string */
5062 const char c = *start;
5065 gv = gv_fetchpv(s, 0, SVt_PVCV);
5068 s = scan_num(s, &yylval);
5075 if (isDIGIT(s[1]) && PL_expect == XOPERATOR) {
5117 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
5119 /* Some keywords can be followed by any delimiter, including ':' */
5120 tmp = ((len == 1 && strchr("msyq", PL_tokenbuf[0])) ||
5121 (len == 2 && ((PL_tokenbuf[0] == 't' && PL_tokenbuf[1] == 'r') ||
5122 (PL_tokenbuf[0] == 'q' &&
5123 strchr("qwxr", PL_tokenbuf[1])))));
5125 /* x::* is just a word, unless x is "CORE" */
5126 if (!tmp && *s == ':' && s[1] == ':' && strNE(PL_tokenbuf, "CORE"))
5130 while (d < PL_bufend && isSPACE(*d))
5131 d++; /* no comments skipped here, or s### is misparsed */
5133 /* Is this a label? */
5134 if (!tmp && PL_expect == XSTATE
5135 && d < PL_bufend && *d == ':' && *(d + 1) != ':') {
5137 yylval.pval = CopLABEL_alloc(PL_tokenbuf);
5142 /* Check for keywords */
5143 tmp = keyword(PL_tokenbuf, len, 0);
5145 /* Is this a word before a => operator? */
5146 if (*d == '=' && d[1] == '>') {
5149 = (OP*)newSVOP(OP_CONST, 0,
5150 S_newSV_maybe_utf8(aTHX_ PL_tokenbuf, len));
5151 yylval.opval->op_private = OPpCONST_BARE;
5155 if (tmp < 0) { /* second-class keyword? */
5156 GV *ogv = NULL; /* override (winner) */
5157 GV *hgv = NULL; /* hidden (loser) */
5158 if (PL_expect != XOPERATOR && (*s != ':' || s[1] != ':')) {
5160 if ((gv = gv_fetchpvn_flags(PL_tokenbuf, len, 0, SVt_PVCV)) &&
5163 if (GvIMPORTED_CV(gv))
5165 else if (! CvMETHOD(cv))
5169 (gvp = (GV**)hv_fetch(PL_globalstash,PL_tokenbuf,len,FALSE)) &&
5170 (gv = *gvp) && isGV_with_GP(gv) &&
5171 GvCVu(gv) && GvIMPORTED_CV(gv))
5178 tmp = 0; /* overridden by import or by GLOBAL */
5181 && -tmp==KEY_lock /* XXX generalizable kludge */
5183 && !hv_fetchs(GvHVn(PL_incgv), "Thread.pm", FALSE))
5185 tmp = 0; /* any sub overrides "weak" keyword */
5187 else { /* no override */
5189 if (tmp == KEY_dump && ckWARN(WARN_MISC)) {
5190 Perl_warner(aTHX_ packWARN(WARN_MISC),
5191 "dump() better written as CORE::dump()");
5195 if (hgv && tmp != KEY_x && tmp != KEY_CORE
5196 && ckWARN(WARN_AMBIGUOUS)) /* never ambiguous */
5197 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
5198 "Ambiguous call resolved as CORE::%s(), %s",
5199 GvENAME(hgv), "qualify as such or use &");
5206 default: /* not a keyword */
5207 /* Trade off - by using this evil construction we can pull the
5208 variable gv into the block labelled keylookup. If not, then
5209 we have to give it function scope so that the goto from the
5210 earlier ':' case doesn't bypass the initialisation. */
5212 just_a_word_zero_gv:
5220 const char lastchar = (PL_bufptr == PL_oldoldbufptr ? 0 : PL_bufptr[-1]);
5223 SV *nextPL_nextwhite = 0;
5227 /* Get the rest if it looks like a package qualifier */
5229 if (*s == '\'' || (*s == ':' && s[1] == ':')) {
5231 s = scan_word(s, PL_tokenbuf + len, sizeof PL_tokenbuf - len,
5234 Perl_croak(aTHX_ "Bad name after %s%s", PL_tokenbuf,
5235 *s == '\'' ? "'" : "::");
5240 if (PL_expect == XOPERATOR) {
5241 if (PL_bufptr == PL_linestart) {
5242 CopLINE_dec(PL_curcop);
5243 Perl_warner(aTHX_ packWARN(WARN_SEMICOLON), PL_warn_nosemi);
5244 CopLINE_inc(PL_curcop);
5247 no_op("Bareword",s);
5250 /* Look for a subroutine with this name in current package,
5251 unless name is "Foo::", in which case Foo is a bearword
5252 (and a package name). */
5254 if (len > 2 && !PL_madskills &&
5255 PL_tokenbuf[len - 2] == ':' && PL_tokenbuf[len - 1] == ':')
5257 if (ckWARN(WARN_BAREWORD)
5258 && ! gv_fetchpvn_flags(PL_tokenbuf, len, 0, SVt_PVHV))
5259 Perl_warner(aTHX_ packWARN(WARN_BAREWORD),
5260 "Bareword \"%s\" refers to nonexistent package",
5263 PL_tokenbuf[len] = '\0';
5269 /* Mustn't actually add anything to a symbol table.
5270 But also don't want to "initialise" any placeholder
5271 constants that might already be there into full
5272 blown PVGVs with attached PVCV. */
5273 gv = gv_fetchpvn_flags(PL_tokenbuf, len,
5274 GV_NOADD_NOINIT, SVt_PVCV);
5279 /* if we saw a global override before, get the right name */
5282 sv = newSVpvs("CORE::GLOBAL::");
5283 sv_catpv(sv,PL_tokenbuf);
5286 /* If len is 0, newSVpv does strlen(), which is correct.
5287 If len is non-zero, then it will be the true length,
5288 and so the scalar will be created correctly. */
5289 sv = newSVpv(PL_tokenbuf,len);
5292 if (PL_madskills && !PL_thistoken) {
5293 char *start = SvPVX(PL_linestr) + PL_realtokenstart;
5294 PL_thistoken = newSVpv(start,s - start);
5295 PL_realtokenstart = s - SvPVX(PL_linestr);
5299 /* Presume this is going to be a bareword of some sort. */
5302 yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
5303 yylval.opval->op_private = OPpCONST_BARE;
5304 /* UTF-8 package name? */
5305 if (UTF && !IN_BYTES &&
5306 is_utf8_string((U8*)SvPVX_const(sv), SvCUR(sv)))
5309 /* And if "Foo::", then that's what it certainly is. */
5314 /* Do the explicit type check so that we don't need to force
5315 the initialisation of the symbol table to have a real GV.
5316 Beware - gv may not really be a PVGV, cv may not really be
5317 a PVCV, (because of the space optimisations that gv_init
5318 understands) But they're true if for this symbol there is
5319 respectively a typeglob and a subroutine.
5321 cv = gv ? ((SvTYPE(gv) == SVt_PVGV)
5322 /* Real typeglob, so get the real subroutine: */
5324 /* A proxy for a subroutine in this package? */
5325 : SvOK(gv) ? (CV *) gv : NULL)
5328 /* See if it's the indirect object for a list operator. */
5330 if (PL_oldoldbufptr &&
5331 PL_oldoldbufptr < PL_bufptr &&
5332 (PL_oldoldbufptr == PL_last_lop
5333 || PL_oldoldbufptr == PL_last_uni) &&
5334 /* NO SKIPSPACE BEFORE HERE! */
5335 (PL_expect == XREF ||
5336 ((PL_opargs[PL_last_lop_op] >> OASHIFT)& 7) == OA_FILEREF))
5338 bool immediate_paren = *s == '(';
5340 /* (Now we can afford to cross potential line boundary.) */
5341 s = SKIPSPACE2(s,nextPL_nextwhite);
5343 PL_nextwhite = nextPL_nextwhite; /* assume no & deception */
5346 /* Two barewords in a row may indicate method call. */
5348 if ((isIDFIRST_lazy_if(s,UTF) || *s == '$') &&
5349 (tmp = intuit_method(s, gv, cv)))
5352 /* If not a declared subroutine, it's an indirect object. */
5353 /* (But it's an indir obj regardless for sort.) */
5354 /* Also, if "_" follows a filetest operator, it's a bareword */
5357 ( !immediate_paren && (PL_last_lop_op == OP_SORT ||
5359 (PL_last_lop_op != OP_MAPSTART &&
5360 PL_last_lop_op != OP_GREPSTART))))
5361 || (PL_tokenbuf[0] == '_' && PL_tokenbuf[1] == '\0'
5362 && ((PL_opargs[PL_last_lop_op] & OA_CLASS_MASK) == OA_FILESTATOP))
5365 PL_expect = (PL_last_lop == PL_oldoldbufptr) ? XTERM : XOPERATOR;
5370 PL_expect = XOPERATOR;
5373 s = SKIPSPACE2(s,nextPL_nextwhite);
5374 PL_nextwhite = nextPL_nextwhite;
5379 /* Is this a word before a => operator? */
5380 if (*s == '=' && s[1] == '>' && !pkgname) {
5382 sv_setpv(((SVOP*)yylval.opval)->op_sv, PL_tokenbuf);
5383 if (UTF && !IN_BYTES && is_utf8_string((U8*)PL_tokenbuf, len))
5384 SvUTF8_on(((SVOP*)yylval.opval)->op_sv);
5388 /* If followed by a paren, it's certainly a subroutine. */
5393 while (SPACE_OR_TAB(*d))
5395 if (*d == ')' && (sv = gv_const_sv(gv))) {
5399 char *par = SvPVX(PL_linestr) + PL_realtokenstart;
5400 sv_catpvn(PL_thistoken, par, s - par);
5402 sv_free(PL_nextwhite);
5412 PL_nextwhite = PL_thiswhite;
5415 start_force(PL_curforce);
5417 NEXTVAL_NEXTTOKE.opval = yylval.opval;
5418 PL_expect = XOPERATOR;
5421 PL_nextwhite = nextPL_nextwhite;
5422 curmad('X', PL_thistoken);
5423 PL_thistoken = newSVpvs("");
5431 /* If followed by var or block, call it a method (unless sub) */
5433 if ((*s == '$' || *s == '{') && (!gv || !cv)) {
5434 PL_last_lop = PL_oldbufptr;
5435 PL_last_lop_op = OP_METHOD;
5439 /* If followed by a bareword, see if it looks like indir obj. */
5442 && (isIDFIRST_lazy_if(s,UTF) || *s == '$')
5443 && (tmp = intuit_method(s, gv, cv)))
5446 /* Not a method, so call it a subroutine (if defined) */
5449 if (lastchar == '-' && ckWARN_d(WARN_AMBIGUOUS))
5450 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
5451 "Ambiguous use of -%s resolved as -&%s()",
5452 PL_tokenbuf, PL_tokenbuf);
5453 /* Check for a constant sub */
5454 if ((sv = gv_const_sv(gv))) {
5456 SvREFCNT_dec(((SVOP*)yylval.opval)->op_sv);
5457 ((SVOP*)yylval.opval)->op_sv = SvREFCNT_inc_simple(sv);
5458 yylval.opval->op_private = 0;
5462 /* Resolve to GV now. */
5463 if (SvTYPE(gv) != SVt_PVGV) {
5464 gv = gv_fetchpv(PL_tokenbuf, 0, SVt_PVCV);
5465 assert (SvTYPE(gv) == SVt_PVGV);
5466 /* cv must have been some sort of placeholder, so
5467 now needs replacing with a real code reference. */
5471 op_free(yylval.opval);
5472 yylval.opval = newCVREF(0, newGVOP(OP_GV, 0, gv));
5473 yylval.opval->op_private |= OPpENTERSUB_NOPAREN;
5474 PL_last_lop = PL_oldbufptr;
5475 PL_last_lop_op = OP_ENTERSUB;
5476 /* Is there a prototype? */
5484 const char *proto = SvPV_const((SV*)cv, protolen);
5487 if ((*proto == '$' || *proto == '_') && proto[1] == '\0')
5489 while (*proto == ';')
5491 if (*proto == '&' && *s == '{') {
5492 sv_setpv(PL_subname,
5495 "__ANON__" : "__ANON__::__ANON__"));
5502 PL_nextwhite = PL_thiswhite;
5505 start_force(PL_curforce);
5506 NEXTVAL_NEXTTOKE.opval = yylval.opval;
5509 PL_nextwhite = nextPL_nextwhite;
5510 curmad('X', PL_thistoken);
5511 PL_thistoken = newSVpvs("");
5518 /* Guess harder when madskills require "best effort". */
5519 if (PL_madskills && (!gv || !GvCVu(gv))) {
5520 int probable_sub = 0;
5521 if (strchr("\"'`$@%0123456789!*+{[<", *s))
5523 else if (isALPHA(*s)) {
5527 d = scan_word(d, tmpbuf, sizeof tmpbuf, TRUE, &tmplen);
5528 if (!keyword(tmpbuf, tmplen, 0))
5531 while (d < PL_bufend && isSPACE(*d))
5533 if (*d == '=' && d[1] == '>')
5538 gv = gv_fetchpv(PL_tokenbuf, GV_ADD, SVt_PVCV);
5539 op_free(yylval.opval);
5540 yylval.opval = newCVREF(0, newGVOP(OP_GV, 0, gv));
5541 yylval.opval->op_private |= OPpENTERSUB_NOPAREN;
5542 PL_last_lop = PL_oldbufptr;
5543 PL_last_lop_op = OP_ENTERSUB;
5544 PL_nextwhite = PL_thiswhite;
5546 start_force(PL_curforce);
5547 NEXTVAL_NEXTTOKE.opval = yylval.opval;
5549 PL_nextwhite = nextPL_nextwhite;
5550 curmad('X', PL_thistoken);
5551 PL_thistoken = newSVpvs("");
5556 NEXTVAL_NEXTTOKE.opval = yylval.opval;
5563 /* Call it a bare word */
5565 if (PL_hints & HINT_STRICT_SUBS)
5566 yylval.opval->op_private |= OPpCONST_STRICT;
5569 if (lastchar != '-') {
5570 if (ckWARN(WARN_RESERVED)) {
5574 if (!*d && !gv_stashpv(PL_tokenbuf, 0))
5575 Perl_warner(aTHX_ packWARN(WARN_RESERVED), PL_warn_reserved,
5582 if ((lastchar == '*' || lastchar == '%' || lastchar == '&')
5583 && ckWARN_d(WARN_AMBIGUOUS)) {
5584 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
5585 "Operator or semicolon missing before %c%s",
5586 lastchar, PL_tokenbuf);
5587 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
5588 "Ambiguous use of %c resolved as operator %c",
5589 lastchar, lastchar);
5595 yylval.opval = (OP*)newSVOP(OP_CONST, 0,
5596 newSVpv(CopFILE(PL_curcop),0));
5600 yylval.opval = (OP*)newSVOP(OP_CONST, 0,
5601 Perl_newSVpvf(aTHX_ "%"IVdf, (IV)CopLINE(PL_curcop)));
5604 case KEY___PACKAGE__:
5605 yylval.opval = (OP*)newSVOP(OP_CONST, 0,
5607 ? newSVhek(HvNAME_HEK(PL_curstash))
5614 if (PL_rsfp && (!PL_in_eval || PL_tokenbuf[2] == 'D')) {
5615 const char *pname = "main";
5616 if (PL_tokenbuf[2] == 'D')
5617 pname = HvNAME_get(PL_curstash ? PL_curstash : PL_defstash);
5618 gv = gv_fetchpv(Perl_form(aTHX_ "%s::DATA", pname), GV_ADD,
5622 GvIOp(gv) = newIO();
5623 IoIFP(GvIOp(gv)) = PL_rsfp;
5624 #if defined(HAS_FCNTL) && defined(F_SETFD)
5626 const int fd = PerlIO_fileno(PL_rsfp);
5627 fcntl(fd,F_SETFD,fd >= 3);
5630 /* Mark this internal pseudo-handle as clean */
5631 IoFLAGS(GvIOp(gv)) |= IOf_UNTAINT;
5633 IoTYPE(GvIOp(gv)) = IoTYPE_PIPE;
5634 else if ((PerlIO*)PL_rsfp == PerlIO_stdin())
5635 IoTYPE(GvIOp(gv)) = IoTYPE_STD;
5637 IoTYPE(GvIOp(gv)) = IoTYPE_RDONLY;
5638 #if defined(WIN32) && !defined(PERL_TEXTMODE_SCRIPTS)
5639 /* if the script was opened in binmode, we need to revert
5640 * it to text mode for compatibility; but only iff it has CRs
5641 * XXX this is a questionable hack at best. */
5642 if (PL_bufend-PL_bufptr > 2
5643 && PL_bufend[-1] == '\n' && PL_bufend[-2] == '\r')
5646 if (IoTYPE(GvIOp(gv)) == IoTYPE_RDONLY) {
5647 loc = PerlIO_tell(PL_rsfp);
5648 (void)PerlIO_seek(PL_rsfp, 0L, 0);
5651 if (PerlLIO_setmode(PL_rsfp, O_TEXT) != -1) {
5653 if (PerlLIO_setmode(PerlIO_fileno(PL_rsfp), O_TEXT) != -1) {
5654 #endif /* NETWARE */
5655 #ifdef PERLIO_IS_STDIO /* really? */
5656 # if defined(__BORLANDC__)
5657 /* XXX see note in do_binmode() */
5658 ((FILE*)PL_rsfp)->flags &= ~_F_BIN;
5662 PerlIO_seek(PL_rsfp, loc, 0);
5666 #ifdef PERLIO_LAYERS
5669 PerlIO_apply_layers(aTHX_ PL_rsfp, NULL, ":utf8");
5670 else if (PL_encoding) {
5677 XPUSHs(PL_encoding);
5679 call_method("name", G_SCALAR);
5683 PerlIO_apply_layers(aTHX_ PL_rsfp, NULL,
5684 Perl_form(aTHX_ ":encoding(%"SVf")",
5693 if (PL_realtokenstart >= 0) {
5694 char *tstart = SvPVX(PL_linestr) + PL_realtokenstart;
5696 PL_endwhite = newSVpvs("");
5697 sv_catsv(PL_endwhite, PL_thiswhite);
5699 sv_catpvn(PL_endwhite, tstart, PL_bufend - tstart);
5700 PL_realtokenstart = -1;
5702 while ((s = filter_gets(PL_endwhite, PL_rsfp,
5703 SvCUR(PL_endwhite))) != Nullch) ;
5718 if (PL_expect == XSTATE) {
5725 if (*s == ':' && s[1] == ':') {
5728 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
5729 if (!(tmp = keyword(PL_tokenbuf, len, 0)))
5730 Perl_croak(aTHX_ "CORE::%s is not a keyword", PL_tokenbuf);
5733 else if (tmp == KEY_require || tmp == KEY_do)
5734 /* that's a way to remember we saw "CORE::" */
5747 LOP(OP_ACCEPT,XTERM);
5753 LOP(OP_ATAN2,XTERM);
5759 LOP(OP_BINMODE,XTERM);
5762 LOP(OP_BLESS,XTERM);
5771 /* When 'use switch' is in effect, continue has a dual
5772 life as a control operator. */
5774 if (!FEATURE_IS_ENABLED("switch"))
5777 /* We have to disambiguate the two senses of
5778 "continue". If the next token is a '{' then
5779 treat it as the start of a continue block;
5780 otherwise treat it as a control operator.
5792 (void)gv_fetchpvs("ENV", GV_ADD|GV_NOTQUAL, SVt_PVHV);
5809 if (!PL_cryptseen) {
5810 PL_cryptseen = TRUE;
5814 LOP(OP_CRYPT,XTERM);
5817 LOP(OP_CHMOD,XTERM);
5820 LOP(OP_CHOWN,XTERM);
5823 LOP(OP_CONNECT,XTERM);
5842 s = force_word(s,WORD,TRUE,TRUE,FALSE);
5843 if (orig_keyword == KEY_do) {
5852 PL_hints |= HINT_BLOCK_SCOPE;
5862 gv_fetchpvs("AnyDBM_File::ISA", GV_ADDMULTI, SVt_PVAV);
5863 LOP(OP_DBMOPEN,XTERM);
5869 s = force_word(s,WORD,TRUE,FALSE,FALSE);
5876 yylval.ival = CopLINE(PL_curcop);
5892 PL_expect = (*s == '{') ? XTERMBLOCK : XTERM;
5893 UNIBRACK(OP_ENTEREVAL);
5911 case KEY_endhostent:
5917 case KEY_endservent:
5920 case KEY_endprotoent:
5931 yylval.ival = CopLINE(PL_curcop);
5933 if (PL_expect == XSTATE && isIDFIRST_lazy_if(s,UTF)) {
5936 int soff = s - SvPVX(PL_linestr); /* for skipspace realloc */
5939 if ((PL_bufend - p) >= 3 &&
5940 strnEQ(p, "my", 2) && isSPACE(*(p + 2)))
5942 else if ((PL_bufend - p) >= 4 &&
5943 strnEQ(p, "our", 3) && isSPACE(*(p + 3)))
5946 if (isIDFIRST_lazy_if(p,UTF)) {
5947 p = scan_ident(p, PL_bufend,
5948 PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
5952 Perl_croak(aTHX_ "Missing $ on loop variable");
5954 s = SvPVX(PL_linestr) + soff;
5960 LOP(OP_FORMLINE,XTERM);
5966 LOP(OP_FCNTL,XTERM);
5972 LOP(OP_FLOCK,XTERM);
5981 LOP(OP_GREPSTART, XREF);
5984 s = force_word(s,WORD,TRUE,FALSE,FALSE);
5999 case KEY_getpriority:
6000 LOP(OP_GETPRIORITY,XTERM);
6002 case KEY_getprotobyname:
6005 case KEY_getprotobynumber:
6006 LOP(OP_GPBYNUMBER,XTERM);
6008 case KEY_getprotoent:
6020 case KEY_getpeername:
6021 UNI(OP_GETPEERNAME);
6023 case KEY_gethostbyname:
6026 case KEY_gethostbyaddr:
6027 LOP(OP_GHBYADDR,XTERM);
6029 case KEY_gethostent:
6032 case KEY_getnetbyname:
6035 case KEY_getnetbyaddr:
6036 LOP(OP_GNBYADDR,XTERM);
6041 case KEY_getservbyname:
6042 LOP(OP_GSBYNAME,XTERM);
6044 case KEY_getservbyport:
6045 LOP(OP_GSBYPORT,XTERM);
6047 case KEY_getservent:
6050 case KEY_getsockname:
6051 UNI(OP_GETSOCKNAME);
6053 case KEY_getsockopt:
6054 LOP(OP_GSOCKOPT,XTERM);
6069 yylval.ival = CopLINE(PL_curcop);
6080 yylval.ival = CopLINE(PL_curcop);
6084 LOP(OP_INDEX,XTERM);
6090 LOP(OP_IOCTL,XTERM);
6102 s = force_word(s,WORD,TRUE,FALSE,FALSE);
6134 LOP(OP_LISTEN,XTERM);
6143 s = scan_pat(s,OP_MATCH);
6144 TERM(sublex_start());
6147 LOP(OP_MAPSTART, XREF);
6150 LOP(OP_MKDIR,XTERM);
6153 LOP(OP_MSGCTL,XTERM);
6156 LOP(OP_MSGGET,XTERM);
6159 LOP(OP_MSGRCV,XTERM);
6162 LOP(OP_MSGSND,XTERM);
6169 if (isIDFIRST_lazy_if(s,UTF)) {
6173 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, TRUE, &len);
6174 if (len == 3 && strnEQ(PL_tokenbuf, "sub", 3))
6176 PL_in_my_stash = find_in_my_stash(PL_tokenbuf, len);
6177 if (!PL_in_my_stash) {
6180 my_snprintf(tmpbuf, sizeof(tmpbuf), "No such class %.1000s", PL_tokenbuf);
6184 if (PL_madskills) { /* just add type to declarator token */
6185 sv_catsv(PL_thistoken, PL_nextwhite);
6187 sv_catpvn(PL_thistoken, start, s - start);
6195 s = force_word(s,WORD,TRUE,FALSE,FALSE);
6202 s = tokenize_use(0, s);
6206 if (*s == '(' || (s = SKIPSPACE1(s), *s == '('))
6213 if (isIDFIRST_lazy_if(s,UTF)) {
6215 for (d = s; isALNUM_lazy_if(d,UTF);)
6217 for (t=d; isSPACE(*t);)
6219 if ( *t && strchr("|&*+-=!?:.", *t) && ckWARN_d(WARN_PRECEDENCE)
6221 && !(t[0] == '=' && t[1] == '>')
6223 int parms_len = (int)(d-s);
6224 Perl_warner(aTHX_ packWARN(WARN_PRECEDENCE),
6225 "Precedence problem: open %.*s should be open(%.*s)",
6226 parms_len, s, parms_len, s);
6232 yylval.ival = OP_OR;
6242 LOP(OP_OPEN_DIR,XTERM);
6245 checkcomma(s,PL_tokenbuf,"filehandle");
6249 checkcomma(s,PL_tokenbuf,"filehandle");
6268 s = force_word(s,WORD,FALSE,TRUE,FALSE);
6272 LOP(OP_PIPE_OP,XTERM);
6275 s = scan_str(s,!!PL_madskills,FALSE);
6278 yylval.ival = OP_CONST;
6279 TERM(sublex_start());
6285 s = scan_str(s,!!PL_madskills,FALSE);
6288 PL_expect = XOPERATOR;
6290 if (SvCUR(PL_lex_stuff)) {
6293 d = SvPV_force(PL_lex_stuff, len);
6295 for (; isSPACE(*d) && len; --len, ++d)
6300 if (!warned && ckWARN(WARN_QW)) {
6301 for (; !isSPACE(*d) && len; --len, ++d) {
6303 Perl_warner(aTHX_ packWARN(WARN_QW),
6304 "Possible attempt to separate words with commas");
6307 else if (*d == '#') {
6308 Perl_warner(aTHX_ packWARN(WARN_QW),
6309 "Possible attempt to put comments in qw() list");
6315 for (; !isSPACE(*d) && len; --len, ++d)
6318 sv = newSVpvn(b, d-b);
6319 if (DO_UTF8(PL_lex_stuff))
6321 words = append_elem(OP_LIST, words,
6322 newSVOP(OP_CONST, 0, tokeq(sv)));
6326 start_force(PL_curforce);
6327 NEXTVAL_NEXTTOKE.opval = words;
6332 SvREFCNT_dec(PL_lex_stuff);
6333 PL_lex_stuff = NULL;
6339 s = scan_str(s,!!PL_madskills,FALSE);
6342 yylval.ival = OP_STRINGIFY;
6343 if (SvIVX(PL_lex_stuff) == '\'')
6344 SvIV_set(PL_lex_stuff, 0); /* qq'$foo' should intepolate */
6345 TERM(sublex_start());
6348 s = scan_pat(s,OP_QR);
6349 TERM(sublex_start());
6352 s = scan_str(s,!!PL_madskills,FALSE);
6355 readpipe_override();
6356 TERM(sublex_start());
6364 s = force_version(s, FALSE);
6366 else if (*s != 'v' || !isDIGIT(s[1])
6367 || (s = force_version(s, TRUE), *s == 'v'))
6369 *PL_tokenbuf = '\0';
6370 s = force_word(s,WORD,TRUE,TRUE,FALSE);
6371 if (isIDFIRST_lazy_if(PL_tokenbuf,UTF))
6372 gv_stashpvn(PL_tokenbuf, strlen(PL_tokenbuf), GV_ADD);
6374 yyerror("<> should be quotes");
6376 if (orig_keyword == KEY_require) {
6384 PL_last_uni = PL_oldbufptr;
6385 PL_last_lop_op = OP_REQUIRE;
6387 return REPORT( (int)REQUIRE );
6393 s = force_word(s,WORD,TRUE,FALSE,FALSE);
6397 LOP(OP_RENAME,XTERM);
6406 LOP(OP_RINDEX,XTERM);
6416 UNIDOR(OP_READLINE);
6420 UNIDOR(OP_BACKTICK);
6429 LOP(OP_REVERSE,XTERM);
6432 UNIDOR(OP_READLINK);
6440 TERM(sublex_start());
6442 TOKEN(1); /* force error */
6445 checkcomma(s,PL_tokenbuf,"filehandle");
6455 LOP(OP_SELECT,XTERM);
6461 LOP(OP_SEMCTL,XTERM);
6464 LOP(OP_SEMGET,XTERM);
6467 LOP(OP_SEMOP,XTERM);
6473 LOP(OP_SETPGRP,XTERM);
6475 case KEY_setpriority:
6476 LOP(OP_SETPRIORITY,XTERM);
6478 case KEY_sethostent:
6484 case KEY_setservent:
6487 case KEY_setprotoent:
6497 LOP(OP_SEEKDIR,XTERM);
6499 case KEY_setsockopt:
6500 LOP(OP_SSOCKOPT,XTERM);
6506 LOP(OP_SHMCTL,XTERM);
6509 LOP(OP_SHMGET,XTERM);
6512 LOP(OP_SHMREAD,XTERM);
6515 LOP(OP_SHMWRITE,XTERM);
6518 LOP(OP_SHUTDOWN,XTERM);
6527 LOP(OP_SOCKET,XTERM);
6529 case KEY_socketpair:
6530 LOP(OP_SOCKPAIR,XTERM);
6533 checkcomma(s,PL_tokenbuf,"subroutine name");
6535 if (*s == ';' || *s == ')') /* probably a close */
6536 Perl_croak(aTHX_ "sort is now a reserved word");
6538 s = force_word(s,WORD,TRUE,TRUE,FALSE);
6542 LOP(OP_SPLIT,XTERM);
6545 LOP(OP_SPRINTF,XTERM);
6548 LOP(OP_SPLICE,XTERM);
6563 LOP(OP_SUBSTR,XTERM);
6569 char tmpbuf[sizeof PL_tokenbuf];
6570 SSize_t tboffset = 0;
6571 expectation attrful;
6572 bool have_name, have_proto;
6573 const int key = tmp;
6578 char *tstart = SvPVX(PL_linestr) + PL_realtokenstart;
6579 SV *subtoken = newSVpvn(tstart, s - tstart);
6583 s = SKIPSPACE2(s,tmpwhite);
6588 if (isIDFIRST_lazy_if(s,UTF) || *s == '\'' ||
6589 (*s == ':' && s[1] == ':'))
6596 attrful = XATTRBLOCK;
6597 /* remember buffer pos'n for later force_word */
6598 tboffset = s - PL_oldbufptr;
6599 d = scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
6602 nametoke = newSVpvn(s, d - s);
6604 if (memchr(tmpbuf, ':', len))
6605 sv_setpvn(PL_subname, tmpbuf, len);
6607 sv_setsv(PL_subname,PL_curstname);
6608 sv_catpvs(PL_subname,"::");
6609 sv_catpvn(PL_subname,tmpbuf,len);
6616 CURMAD('X', nametoke);
6617 CURMAD('_', tmpwhite);
6618 (void) force_word(PL_oldbufptr + tboffset, WORD,
6621 s = SKIPSPACE2(d,tmpwhite);
6628 Perl_croak(aTHX_ "Missing name in \"my sub\"");
6629 PL_expect = XTERMBLOCK;
6630 attrful = XATTRTERM;
6631 sv_setpvn(PL_subname,"?",1);
6635 if (key == KEY_format) {
6637 PL_lex_formbrack = PL_lex_brackets + 1;
6639 PL_thistoken = subtoken;
6643 (void) force_word(PL_oldbufptr + tboffset, WORD,
6649 /* Look for a prototype */
6652 bool bad_proto = FALSE;
6653 const bool warnsyntax = ckWARN(WARN_SYNTAX);
6655 s = scan_str(s,!!PL_madskills,FALSE);
6657 Perl_croak(aTHX_ "Prototype not terminated");
6658 /* strip spaces and check for bad characters */
6659 d = SvPVX(PL_lex_stuff);
6661 for (p = d; *p; ++p) {
6664 if (warnsyntax && !strchr("$@%*;[]&\\_", *p))
6670 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
6671 "Illegal character in prototype for %"SVf" : %s",
6672 SVfARG(PL_subname), d);
6673 SvCUR_set(PL_lex_stuff, tmp);
6678 CURMAD('q', PL_thisopen);
6679 CURMAD('_', tmpwhite);
6680 CURMAD('=', PL_thisstuff);
6681 CURMAD('Q', PL_thisclose);
6682 NEXTVAL_NEXTTOKE.opval =
6683 (OP*)newSVOP(OP_CONST, 0, PL_lex_stuff);
6684 PL_lex_stuff = Nullsv;
6687 s = SKIPSPACE2(s,tmpwhite);
6695 if (*s == ':' && s[1] != ':')
6696 PL_expect = attrful;
6697 else if (*s != '{' && key == KEY_sub) {
6699 Perl_croak(aTHX_ "Illegal declaration of anonymous subroutine");
6701 Perl_croak(aTHX_ "Illegal declaration of subroutine %"SVf, SVfARG(PL_subname));
6708 curmad('^', newSVpvs(""));
6709 CURMAD('_', tmpwhite);
6713 PL_thistoken = subtoken;
6716 NEXTVAL_NEXTTOKE.opval =
6717 (OP*)newSVOP(OP_CONST, 0, PL_lex_stuff);
6718 PL_lex_stuff = NULL;
6723 sv_setpv(PL_subname,
6725 (PL_curstash ? "__ANON__" : "__ANON__::__ANON__"));
6729 (void) force_word(PL_oldbufptr + tboffset, WORD,
6739 LOP(OP_SYSTEM,XREF);
6742 LOP(OP_SYMLINK,XTERM);
6745 LOP(OP_SYSCALL,XTERM);
6748 LOP(OP_SYSOPEN,XTERM);
6751 LOP(OP_SYSSEEK,XTERM);
6754 LOP(OP_SYSREAD,XTERM);
6757 LOP(OP_SYSWRITE,XTERM);
6761 TERM(sublex_start());
6782 LOP(OP_TRUNCATE,XTERM);
6794 yylval.ival = CopLINE(PL_curcop);
6798 yylval.ival = CopLINE(PL_curcop);
6802 LOP(OP_UNLINK,XTERM);
6808 LOP(OP_UNPACK,XTERM);
6811 LOP(OP_UTIME,XTERM);
6817 LOP(OP_UNSHIFT,XTERM);
6820 s = tokenize_use(1, s);
6830 yylval.ival = CopLINE(PL_curcop);
6834 yylval.ival = CopLINE(PL_curcop);
6838 PL_hints |= HINT_BLOCK_SCOPE;
6845 LOP(OP_WAITPID,XTERM);
6854 ctl_l[0] = toCTRL('L');
6856 gv_fetchpvn_flags(ctl_l, 1, GV_ADD|GV_NOTQUAL, SVt_PV);
6859 /* Make sure $^L is defined */
6860 gv_fetchpvs("\f", GV_ADD|GV_NOTQUAL, SVt_PV);
6865 if (PL_expect == XOPERATOR)
6871 yylval.ival = OP_XOR;
6876 TERM(sublex_start());
6881 #pragma segment Main
6885 S_pending_ident(pTHX)
6890 /* pit holds the identifier we read and pending_ident is reset */
6891 char pit = PL_pending_ident;
6892 PL_pending_ident = 0;
6894 /* PL_realtokenstart = realtokenend = PL_bufptr - SvPVX(PL_linestr); */
6895 DEBUG_T({ PerlIO_printf(Perl_debug_log,
6896 "### Pending identifier '%s'\n", PL_tokenbuf); });
6898 /* if we're in a my(), we can't allow dynamics here.
6899 $foo'bar has already been turned into $foo::bar, so
6900 just check for colons.
6902 if it's a legal name, the OP is a PADANY.
6905 if (PL_in_my == KEY_our) { /* "our" is merely analogous to "my" */
6906 if (strchr(PL_tokenbuf,':'))
6907 yyerror(Perl_form(aTHX_ "No package name allowed for "
6908 "variable %s in \"our\"",
6910 tmp = allocmy(PL_tokenbuf);
6913 if (strchr(PL_tokenbuf,':'))
6914 yyerror(Perl_form(aTHX_ PL_no_myglob,
6915 PL_in_my == KEY_my ? "my" : "state", PL_tokenbuf));
6917 yylval.opval = newOP(OP_PADANY, 0);
6918 yylval.opval->op_targ = allocmy(PL_tokenbuf);
6924 build the ops for accesses to a my() variable.
6926 Deny my($a) or my($b) in a sort block, *if* $a or $b is
6927 then used in a comparison. This catches most, but not
6928 all cases. For instance, it catches
6929 sort { my($a); $a <=> $b }
6931 sort { my($a); $a < $b ? -1 : $a == $b ? 0 : 1; }
6932 (although why you'd do that is anyone's guess).
6935 if (!strchr(PL_tokenbuf,':')) {
6937 tmp = pad_findmy(PL_tokenbuf);
6938 if (tmp != NOT_IN_PAD) {
6939 /* might be an "our" variable" */
6940 if (PAD_COMPNAME_FLAGS_isOUR(tmp)) {
6941 /* build ops for a bareword */
6942 HV * const stash = PAD_COMPNAME_OURSTASH(tmp);
6943 HEK * const stashname = HvNAME_HEK(stash);
6944 SV * const sym = newSVhek(stashname);
6945 sv_catpvs(sym, "::");
6946 sv_catpv(sym, PL_tokenbuf+1);
6947 yylval.opval = (OP*)newSVOP(OP_CONST, 0, sym);
6948 yylval.opval->op_private = OPpCONST_ENTERED;
6951 ? (GV_ADDMULTI | GV_ADDINEVAL)
6954 ((PL_tokenbuf[0] == '$') ? SVt_PV
6955 : (PL_tokenbuf[0] == '@') ? SVt_PVAV
6960 /* if it's a sort block and they're naming $a or $b */
6961 if (PL_last_lop_op == OP_SORT &&
6962 PL_tokenbuf[0] == '$' &&
6963 (PL_tokenbuf[1] == 'a' || PL_tokenbuf[1] == 'b')
6966 for (d = PL_in_eval ? PL_oldoldbufptr : PL_linestart;
6967 d < PL_bufend && *d != '\n';
6970 if (strnEQ(d,"<=>",3) || strnEQ(d,"cmp",3)) {
6971 Perl_croak(aTHX_ "Can't use \"my %s\" in sort comparison",
6977 yylval.opval = newOP(OP_PADANY, 0);
6978 yylval.opval->op_targ = tmp;
6984 Whine if they've said @foo in a doublequoted string,
6985 and @foo isn't a variable we can find in the symbol
6988 if (pit == '@' && PL_lex_state != LEX_NORMAL && !PL_lex_brackets) {
6989 GV *gv = gv_fetchpv(PL_tokenbuf+1, 0, SVt_PVAV);
6990 if ((!gv || ((PL_tokenbuf[0] == '@') ? !GvAV(gv) : !GvHV(gv)))
6991 && ckWARN(WARN_AMBIGUOUS)
6992 /* DO NOT warn for @- and @+ */
6993 && !( PL_tokenbuf[2] == '\0' &&
6994 ( PL_tokenbuf[1] == '-' || PL_tokenbuf[1] == '+' ))
6997 /* Downgraded from fatal to warning 20000522 mjd */
6998 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
6999 "Possible unintended interpolation of %s in string",
7004 /* build ops for a bareword */
7005 yylval.opval = (OP*)newSVOP(OP_CONST, 0, newSVpv(PL_tokenbuf+1, 0));
7006 yylval.opval->op_private = OPpCONST_ENTERED;
7009 /* If the identifier refers to a stash, don't autovivify it.
7010 * Change 24660 had the side effect of causing symbol table
7011 * hashes to always be defined, even if they were freshly
7012 * created and the only reference in the entire program was
7013 * the single statement with the defined %foo::bar:: test.
7014 * It appears that all code in the wild doing this actually
7015 * wants to know whether sub-packages have been loaded, so
7016 * by avoiding auto-vivifying symbol tables, we ensure that
7017 * defined %foo::bar:: continues to be false, and the existing
7018 * tests still give the expected answers, even though what
7019 * they're actually testing has now changed subtly.
7021 (*PL_tokenbuf == '%' && *(d = PL_tokenbuf + strlen(PL_tokenbuf) - 1) == ':' && d[-1] == ':'
7023 : PL_in_eval ? (GV_ADDMULTI | GV_ADDINEVAL) : GV_ADD),
7024 ((PL_tokenbuf[0] == '$') ? SVt_PV
7025 : (PL_tokenbuf[0] == '@') ? SVt_PVAV
7031 * The following code was generated by perl_keyword.pl.
7035 Perl_keyword (pTHX_ const char *name, I32 len, bool all_keywords)
7040 case 1: /* 5 tokens of length 1 */
7072 case 2: /* 18 tokens of length 2 */
7218 case 3: /* 29 tokens of length 3 */
7222 if (name[1] == 'N' &&
7285 if (name[1] == 'i' &&
7307 return (all_keywords || FEATURE_IS_ENABLED("err") ? -KEY_err : 0);
7325 if (name[1] == 'o' &&
7334 if (name[1] == 'e' &&
7343 if (name[1] == 'n' &&
7352 if (name[1] == 'o' &&
7361 if (name[1] == 'a' &&
7370 if (name[1] == 'o' &&
7432 if (name[1] == 'e' &&
7446 return (all_keywords || FEATURE_IS_ENABLED("say") ? KEY_say : 0);
7472 if (name[1] == 'i' &&
7481 if (name[1] == 's' &&
7490 if (name[1] == 'e' &&
7499 if (name[1] == 'o' &&
7511 case 4: /* 41 tokens of length 4 */
7515 if (name[1] == 'O' &&
7525 if (name[1] == 'N' &&
7535 if (name[1] == 'i' &&
7545 if (name[1] == 'h' &&
7555 if (name[1] == 'u' &&
7568 if (name[2] == 'c' &&
7577 if (name[2] == 's' &&
7586 if (name[2] == 'a' &&
7622 if (name[1] == 'o' &&
7635 if (name[2] == 't' &&
7644 if (name[2] == 'o' &&
7653 if (name[2] == 't' &&
7662 if (name[2] == 'e' &&
7675 if (name[1] == 'o' &&
7688 if (name[2] == 'y' &&
7697 if (name[2] == 'l' &&
7713 if (name[2] == 's' &&
7722 if (name[2] == 'n' &&
7731 if (name[2] == 'c' &&
7744 if (name[1] == 'e' &&
7754 if (name[1] == 'p' &&
7767 if (name[2] == 'c' &&
7776 if (name[2] == 'p' &&
7785 if (name[2] == 's' &&
7801 if (name[2] == 'n' &&
7871 if (name[2] == 'r' &&
7880 if (name[2] == 'r' &&
7889 if (name[2] == 'a' &&
7905 if (name[2] == 'l' &&
7967 if (name[2] == 'e' &&
7970 return (all_keywords || FEATURE_IS_ENABLED("switch") ? KEY_when : 0);
7983 case 5: /* 39 tokens of length 5 */
7987 if (name[1] == 'E' &&
7998 if (name[1] == 'H' &&
8012 if (name[2] == 'a' &&
8022 if (name[2] == 'a' &&
8039 if (name[2] == 'e' &&
8049 if (name[2] == 'e' &&
8053 return (all_keywords || FEATURE_IS_ENABLED("switch") ? -KEY_break : 0);
8069 if (name[3] == 'i' &&
8078 if (name[3] == 'o' &&
8114 if (name[2] == 'o' &&
8124 if (name[2] == 'y' &&
8138 if (name[1] == 'l' &&
8152 if (name[2] == 'n' &&
8162 if (name[2] == 'o' &&
8176 if (name[1] == 'i' &&
8181 return (all_keywords || FEATURE_IS_ENABLED("switch") ? KEY_given : 0);
8190 if (name[2] == 'd' &&
8200 if (name[2] == 'c' &&
8217 if (name[2] == 'c' &&
8227 if (name[2] == 't' &&
8241 if (name[1] == 'k' &&
8252 if (name[1] == 'r' &&
8266 if (name[2] == 's' &&
8276 if (name[2] == 'd' &&
8293 if (name[2] == 'm' &&
8303 if (name[2] == 'i' &&
8313 if (name[2] == 'e' &&
8323 if (name[2] == 'l' &&
8333 if (name[2] == 'a' &&
8346 if (name[3] == 't' &&
8349 return (all_keywords || FEATURE_IS_ENABLED("state") ? KEY_state : 0);
8355 if (name[3] == 'd' &&
8372 if (name[1] == 'i' &&
8386 if (name[2] == 'a' &&
8399 if (name[3] == 'e' &&
8434 if (name[2] == 'i' &&
8451 if (name[2] == 'i' &&
8461 if (name[2] == 'i' &&
8478 case 6: /* 33 tokens of length 6 */
8482 if (name[1] == 'c' &&
8497 if (name[2] == 'l' &&
8508 if (name[2] == 'r' &&
8523 if (name[1] == 'e' &&
8538 if (name[2] == 's' &&
8543 if(ckWARN_d(WARN_SYNTAX))
8544 Perl_warner(aTHX_ packWARN(WARN_SYNTAX), "elseif should be elsif");
8550 if (name[2] == 'i' &&
8568 if (name[2] == 'l' &&
8579 if (name[2] == 'r' &&
8594 if (name[1] == 'm' &&
8609 if (name[2] == 'n' &&
8620 if (name[2] == 's' &&
8635 if (name[1] == 's' &&
8641 if (name[4] == 't' &&
8650 if (name[4] == 'e' &&
8659 if (name[4] == 'c' &&
8668 if (name[4] == 'n' &&
8684 if (name[1] == 'r' &&
8702 if (name[3] == 'a' &&
8712 if (name[3] == 'u' &&
8726 if (name[2] == 'n' &&
8744 if (name[2] == 'a' &&
8758 if (name[3] == 'e' &&
8771 if (name[4] == 't' &&
8780 if (name[4] == 'e' &&
8802 if (name[4] == 't' &&
8811 if (name[4] == 'e' &&
8827 if (name[2] == 'c' &&
8838 if (name[2] == 'l' &&
8849 if (name[2] == 'b' &&
8860 if (name[2] == 's' &&
8883 if (name[4] == 's' &&
8892 if (name[4] == 'n' &&
8905 if (name[3] == 'a' &&
8922 if (name[1] == 'a' &&
8937 case 7: /* 29 tokens of length 7 */
8941 if (name[1] == 'E' &&
8954 if (name[1] == '_' &&
8967 if (name[1] == 'i' &&
8974 return -KEY_binmode;
8980 if (name[1] == 'o' &&
8987 return -KEY_connect;
8996 if (name[2] == 'm' &&
9002 return -KEY_dbmopen;
9013 if (name[4] == 'u' &&
9017 return (all_keywords || FEATURE_IS_ENABLED("switch") ? KEY_default : 0);
9023 if (name[4] == 'n' &&
9044 if (name[1] == 'o' &&
9057 if (name[1] == 'e' &&
9064 if (name[5] == 'r' &&
9067 return -KEY_getpgrp;
9073 if (name[5] == 'i' &&
9076 return -KEY_getppid;
9089 if (name[1] == 'c' &&
9096 return -KEY_lcfirst;
9102 if (name[1] == 'p' &&
9109 return -KEY_opendir;
9115 if (name[1] == 'a' &&
9133 if (name[3] == 'd' &&
9138 return -KEY_readdir;
9144 if (name[3] == 'u' &&
9155 if (name[3] == 'e' &&
9160 return -KEY_reverse;
9179 if (name[3] == 'k' &&
9184 return -KEY_seekdir;
9190 if (name[3] == 'p' &&
9195 return -KEY_setpgrp;
9205 if (name[2] == 'm' &&
9211 return -KEY_shmread;
9217 if (name[2] == 'r' &&
9223 return -KEY_sprintf;
9232 if (name[3] == 'l' &&
9237 return -KEY_symlink;
9246 if (name[4] == 'a' &&
9250 return -KEY_syscall;
9256 if (name[4] == 'p' &&
9260 return -KEY_sysopen;
9266 if (name[4] == 'e' &&
9270 return -KEY_sysread;
9276 if (name[4] == 'e' &&
9280 return -KEY_sysseek;
9298 if (name[1] == 'e' &&
9305 return -KEY_telldir;
9314 if (name[2] == 'f' &&
9320 return -KEY_ucfirst;
9326 if (name[2] == 's' &&
9332 return -KEY_unshift;
9342 if (name[1] == 'a' &&
9349 return -KEY_waitpid;
9358 case 8: /* 26 tokens of length 8 */
9362 if (name[1] == 'U' &&
9370 return KEY_AUTOLOAD;
9381 if (name[3] == 'A' &&
9387 return KEY___DATA__;
9393 if (name[3] == 'I' &&
9399 return -KEY___FILE__;
9405 if (name[3] == 'I' &&
9411 return -KEY___LINE__;
9427 if (name[2] == 'o' &&
9434 return -KEY_closedir;
9440 if (name[2] == 'n' &&
9447 return -KEY_continue;
9457 if (name[1] == 'b' &&
9465 return -KEY_dbmclose;
9471 if (name[1] == 'n' &&
9477 if (name[4] == 'r' &&
9482 return -KEY_endgrent;
9488 if (name[4] == 'w' &&
9493 return -KEY_endpwent;
9506 if (name[1] == 'o' &&
9514 return -KEY_formline;
9520 if (name[1] == 'e' &&
9531 if (name[6] == 'n' &&
9534 return -KEY_getgrent;
9540 if (name[6] == 'i' &&
9543 return -KEY_getgrgid;
9549 if (name[6] == 'a' &&
9552 return -KEY_getgrnam;
9565 if (name[4] == 'o' &&
9570 return -KEY_getlogin;
9581 if (name[6] == 'n' &&
9584 return -KEY_getpwent;
9590 if (name[6] == 'a' &&
9593 return -KEY_getpwnam;
9599 if (name[6] == 'i' &&
9602 return -KEY_getpwuid;
9622 if (name[1] == 'e' &&
9629 if (name[5] == 'i' &&
9636 return -KEY_readline;
9641 return -KEY_readlink;
9652 if (name[5] == 'i' &&
9656 return -KEY_readpipe;
9677 if (name[4] == 'r' &&
9682 return -KEY_setgrent;
9688 if (name[4] == 'w' &&
9693 return -KEY_setpwent;
9709 if (name[3] == 'w' &&
9715 return -KEY_shmwrite;
9721 if (name[3] == 't' &&
9727 return -KEY_shutdown;
9737 if (name[2] == 's' &&
9744 return -KEY_syswrite;
9754 if (name[1] == 'r' &&
9762 return -KEY_truncate;
9771 case 9: /* 9 tokens of length 9 */
9775 if (name[1] == 'N' &&
9784 return KEY_UNITCHECK;
9790 if (name[1] == 'n' &&
9799 return -KEY_endnetent;
9805 if (name[1] == 'e' &&
9814 return -KEY_getnetent;
9820 if (name[1] == 'o' &&
9829 return -KEY_localtime;
9835 if (name[1] == 'r' &&
9844 return KEY_prototype;
9850 if (name[1] == 'u' &&
9859 return -KEY_quotemeta;
9865 if (name[1] == 'e' &&
9874 return -KEY_rewinddir;
9880 if (name[1] == 'e' &&
9889 return -KEY_setnetent;
9895 if (name[1] == 'a' &&
9904 return -KEY_wantarray;
9913 case 10: /* 9 tokens of length 10 */
9917 if (name[1] == 'n' &&
9923 if (name[4] == 'o' &&
9930 return -KEY_endhostent;
9936 if (name[4] == 'e' &&
9943 return -KEY_endservent;
9956 if (name[1] == 'e' &&
9962 if (name[4] == 'o' &&
9969 return -KEY_gethostent;
9978 if (name[5] == 'r' &&
9984 return -KEY_getservent;
9990 if (name[5] == 'c' &&
9996 return -KEY_getsockopt;
10016 if (name[2] == 't')
10021 if (name[4] == 'o' &&
10028 return -KEY_sethostent;
10037 if (name[5] == 'r' &&
10043 return -KEY_setservent;
10049 if (name[5] == 'c' &&
10055 return -KEY_setsockopt;
10072 if (name[2] == 'c' &&
10081 return -KEY_socketpair;
10094 case 11: /* 8 tokens of length 11 */
10098 if (name[1] == '_' &&
10108 { /* __PACKAGE__ */
10109 return -KEY___PACKAGE__;
10115 if (name[1] == 'n' &&
10125 { /* endprotoent */
10126 return -KEY_endprotoent;
10132 if (name[1] == 'e' &&
10141 if (name[5] == 'e' &&
10147 { /* getpeername */
10148 return -KEY_getpeername;
10157 if (name[6] == 'o' &&
10162 { /* getpriority */
10163 return -KEY_getpriority;
10169 if (name[6] == 't' &&
10174 { /* getprotoent */
10175 return -KEY_getprotoent;
10189 if (name[4] == 'o' &&
10196 { /* getsockname */
10197 return -KEY_getsockname;
10210 if (name[1] == 'e' &&
10218 if (name[6] == 'o' &&
10223 { /* setpriority */
10224 return -KEY_setpriority;
10230 if (name[6] == 't' &&
10235 { /* setprotoent */
10236 return -KEY_setprotoent;
10252 case 12: /* 2 tokens of length 12 */
10253 if (name[0] == 'g' &&
10265 if (name[9] == 'd' &&
10268 { /* getnetbyaddr */
10269 return -KEY_getnetbyaddr;
10275 if (name[9] == 'a' &&
10278 { /* getnetbyname */
10279 return -KEY_getnetbyname;
10291 case 13: /* 4 tokens of length 13 */
10292 if (name[0] == 'g' &&
10299 if (name[4] == 'o' &&
10308 if (name[10] == 'd' &&
10311 { /* gethostbyaddr */
10312 return -KEY_gethostbyaddr;
10318 if (name[10] == 'a' &&
10321 { /* gethostbyname */
10322 return -KEY_gethostbyname;
10335 if (name[4] == 'e' &&
10344 if (name[10] == 'a' &&
10347 { /* getservbyname */
10348 return -KEY_getservbyname;
10354 if (name[10] == 'o' &&
10357 { /* getservbyport */
10358 return -KEY_getservbyport;
10377 case 14: /* 1 tokens of length 14 */
10378 if (name[0] == 'g' &&
10392 { /* getprotobyname */
10393 return -KEY_getprotobyname;
10398 case 16: /* 1 tokens of length 16 */
10399 if (name[0] == 'g' &&
10415 { /* getprotobynumber */
10416 return -KEY_getprotobynumber;
10430 S_checkcomma(pTHX_ const char *s, const char *name, const char *what)
10434 if (*s == ' ' && s[1] == '(') { /* XXX gotta be a better way */
10435 if (ckWARN(WARN_SYNTAX)) {
10438 for (w = s+2; *w && level; w++) {
10441 else if (*w == ')')
10444 while (isSPACE(*w))
10446 if (!*w || !strchr(";|})]oaiuw!=", *w)) /* an advisory hack only... */
10447 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
10448 "%s (...) interpreted as function",name);
10451 while (s < PL_bufend && isSPACE(*s))
10455 while (s < PL_bufend && isSPACE(*s))
10457 if (isIDFIRST_lazy_if(s,UTF)) {
10458 const char * const w = s++;
10459 while (isALNUM_lazy_if(s,UTF))
10461 while (s < PL_bufend && isSPACE(*s))
10465 if (keyword(w, s - w, 0))
10468 gv = gv_fetchpvn_flags(w, s - w, 0, SVt_PVCV);
10469 if (gv && GvCVu(gv))
10471 Perl_croak(aTHX_ "No comma allowed after %s", what);
10476 /* Either returns sv, or mortalizes sv and returns a new SV*.
10477 Best used as sv=new_constant(..., sv, ...).
10478 If s, pv are NULL, calls subroutine with one argument,
10479 and type is used with error messages only. */
10482 S_new_constant(pTHX_ const char *s, STRLEN len, const char *key, SV *sv, SV *pv,
10486 HV * const table = GvHV(PL_hintgv); /* ^H */
10490 const char *why1 = "", *why2 = "", *why3 = "";
10492 if (!table || !(PL_hints & HINT_LOCALIZE_HH)) {
10495 why2 = (const char *)
10496 (strEQ(key,"charnames")
10497 ? "(possibly a missing \"use charnames ...\")"
10499 msg = Perl_newSVpvf(aTHX_ "Constant(%s) unknown: %s",
10500 (type ? type: "undef"), why2);
10502 /* This is convoluted and evil ("goto considered harmful")
10503 * but I do not understand the intricacies of all the different
10504 * failure modes of %^H in here. The goal here is to make
10505 * the most probable error message user-friendly. --jhi */
10510 msg = Perl_newSVpvf(aTHX_ "Constant(%s): %s%s%s",
10511 (type ? type: "undef"), why1, why2, why3);
10513 yyerror(SvPVX_const(msg));
10517 cvp = hv_fetch(table, key, strlen(key), FALSE);
10518 if (!cvp || !SvOK(*cvp)) {
10521 why3 = "} is not defined";
10524 sv_2mortal(sv); /* Parent created it permanently */
10527 pv = sv_2mortal(newSVpvn(s, len));
10529 typesv = sv_2mortal(newSVpv(type, 0));
10531 typesv = &PL_sv_undef;
10533 PUSHSTACKi(PERLSI_OVERLOAD);
10545 call_sv(cv, G_SCALAR | ( PL_in_eval ? 0 : G_EVAL));
10549 /* Check the eval first */
10550 if (!PL_in_eval && SvTRUE(ERRSV)) {
10551 sv_catpvs(ERRSV, "Propagated");
10552 yyerror(SvPV_nolen_const(ERRSV)); /* Duplicates the message inside eval */
10554 res = SvREFCNT_inc_simple(sv);
10558 SvREFCNT_inc_simple_void(res);
10567 why1 = "Call to &{$^H{";
10569 why3 = "}} did not return a defined value";
10577 /* Returns a NUL terminated string, with the length of the string written to
10581 S_scan_word(pTHX_ register char *s, char *dest, STRLEN destlen, int allow_package, STRLEN *slp)
10584 register char *d = dest;
10585 register char * const e = d + destlen - 3; /* two-character token, ending NUL */
10588 Perl_croak(aTHX_ ident_too_long);
10589 if (isALNUM(*s)) /* UTF handled below */
10591 else if (allow_package && (*s == '\'') && isIDFIRST_lazy_if(s+1,UTF)) {
10596 else if (allow_package && (s[0] == ':') && (s[1] == ':') && (s[2] != '$')) {
10600 else if (UTF && UTF8_IS_START(*s) && isALNUM_utf8((U8*)s)) {
10601 char *t = s + UTF8SKIP(s);
10603 while (UTF8_IS_CONTINUED(*t) && is_utf8_mark((U8*)t))
10607 Perl_croak(aTHX_ ident_too_long);
10608 Copy(s, d, len, char);
10621 S_scan_ident(pTHX_ register char *s, register const char *send, char *dest, STRLEN destlen, I32 ck_uni)
10624 char *bracket = NULL;
10626 register char *d = dest;
10627 register char * const e = d + destlen + 3; /* two-character token, ending NUL */
10632 while (isDIGIT(*s)) {
10634 Perl_croak(aTHX_ ident_too_long);
10641 Perl_croak(aTHX_ ident_too_long);
10642 if (isALNUM(*s)) /* UTF handled below */
10644 else if (*s == '\'' && isIDFIRST_lazy_if(s+1,UTF)) {
10649 else if (*s == ':' && s[1] == ':') {
10653 else if (UTF && UTF8_IS_START(*s) && isALNUM_utf8((U8*)s)) {
10654 char *t = s + UTF8SKIP(s);
10655 while (UTF8_IS_CONTINUED(*t) && is_utf8_mark((U8*)t))
10657 if (d + (t - s) > e)
10658 Perl_croak(aTHX_ ident_too_long);
10659 Copy(s, d, t - s, char);
10670 if (PL_lex_state != LEX_NORMAL)
10671 PL_lex_state = LEX_INTERPENDMAYBE;
10674 if (*s == '$' && s[1] &&
10675 (isALNUM_lazy_if(s+1,UTF) || s[1] == '$' || s[1] == '{' || strnEQ(s+1,"::",2)) )
10688 if (*d == '^' && *s && isCONTROLVAR(*s)) {
10693 if (isSPACE(s[-1])) {
10695 const char ch = *s++;
10696 if (!SPACE_OR_TAB(ch)) {
10702 if (isIDFIRST_lazy_if(d,UTF)) {
10706 while ((end < send && isALNUM_lazy_if(end,UTF)) || *end == ':') {
10707 end += UTF8SKIP(end);
10708 while (end < send && UTF8_IS_CONTINUED(*end) && is_utf8_mark((U8*)end))
10709 end += UTF8SKIP(end);
10711 Copy(s, d, end - s, char);
10716 while ((isALNUM(*s) || *s == ':') && d < e)
10719 Perl_croak(aTHX_ ident_too_long);
10722 while (s < send && SPACE_OR_TAB(*s))
10724 if ((*s == '[' || (*s == '{' && strNE(dest, "sub")))) {
10725 if (ckWARN(WARN_AMBIGUOUS) && keyword(dest, d - dest, 0)) {
10726 const char * const brack =
10728 ((*s == '[') ? "[...]" : "{...}");
10729 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
10730 "Ambiguous use of %c{%s%s} resolved to %c%s%s",
10731 funny, dest, brack, funny, dest, brack);
10734 PL_lex_brackstack[PL_lex_brackets++] = (char)(XOPERATOR | XFAKEBRACK);
10738 /* Handle extended ${^Foo} variables
10739 * 1999-02-27 mjd-perl-patch@plover.com */
10740 else if (!isALNUM(*d) && !isPRINT(*d) /* isCTRL(d) */
10744 while (isALNUM(*s) && d < e) {
10748 Perl_croak(aTHX_ ident_too_long);
10753 if (PL_lex_state == LEX_INTERPNORMAL && !PL_lex_brackets) {
10754 PL_lex_state = LEX_INTERPEND;
10757 if (PL_lex_state == LEX_NORMAL) {
10758 if (ckWARN(WARN_AMBIGUOUS) &&
10759 (keyword(dest, d - dest, 0)
10760 || get_cvn_flags(dest, d - dest, 0)))
10764 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
10765 "Ambiguous use of %c{%s} resolved to %c%s",
10766 funny, dest, funny, dest);
10771 s = bracket; /* let the parser handle it */
10775 else if (PL_lex_state == LEX_INTERPNORMAL && !PL_lex_brackets && !intuit_more(s))
10776 PL_lex_state = LEX_INTERPEND;
10781 Perl_pmflag(pTHX_ U32* pmfl, int ch)
10783 PERL_UNUSED_CONTEXT;
10787 CASE_STD_PMMOD_FLAGS_PARSE_SET(pmfl);
10788 case GLOBAL_PAT_MOD: *pmfl |= PMf_GLOBAL; break;
10789 case CONTINUE_PAT_MOD: *pmfl |= PMf_CONTINUE; break;
10790 case ONCE_PAT_MOD: *pmfl |= PMf_KEEP; break;
10791 case KEEPCOPY_PAT_MOD: *pmfl |= PMf_KEEPCOPY; break;
10797 S_scan_pat(pTHX_ char *start, I32 type)
10801 char *s = scan_str(start,!!PL_madskills,FALSE);
10802 const char * const valid_flags =
10803 (const char *)((type == OP_QR) ? QR_PAT_MODS : M_PAT_MODS);
10810 const char * const delimiter = skipspace(start);
10814 ? "Search pattern not terminated or ternary operator parsed as search pattern"
10815 : "Search pattern not terminated" ));
10818 pm = (PMOP*)newPMOP(type, 0);
10819 if (PL_multi_open == '?')
10820 pm->op_pmflags |= PMf_ONCE;
10824 while (*s && strchr(valid_flags, *s))
10825 pmflag(&pm->op_pmflags,*s++);
10827 if (PL_madskills && modstart != s) {
10828 SV* tmptoken = newSVpvn(modstart, s - modstart);
10829 append_madprops(newMADPROP('m', MAD_SV, tmptoken, 0), (OP*)pm, 0);
10832 /* issue a warning if /c is specified,but /g is not */
10833 if ((pm->op_pmflags & PMf_CONTINUE) && !(pm->op_pmflags & PMf_GLOBAL)
10834 && ckWARN(WARN_REGEXP))
10836 Perl_warner(aTHX_ packWARN(WARN_REGEXP),
10837 "Use of /c modifier is meaningless without /g" );
10840 pm->op_pmpermflags = pm->op_pmflags;
10842 PL_lex_op = (OP*)pm;
10843 yylval.ival = OP_MATCH;
10848 S_scan_subst(pTHX_ char *start)
10859 yylval.ival = OP_NULL;
10861 s = scan_str(start,!!PL_madskills,FALSE);
10864 Perl_croak(aTHX_ "Substitution pattern not terminated");
10866 if (s[-1] == PL_multi_open)
10869 if (PL_madskills) {
10870 CURMAD('q', PL_thisopen);
10871 CURMAD('_', PL_thiswhite);
10872 CURMAD('E', PL_thisstuff);
10873 CURMAD('Q', PL_thisclose);
10874 PL_realtokenstart = s - SvPVX(PL_linestr);
10878 first_start = PL_multi_start;
10879 s = scan_str(s,!!PL_madskills,FALSE);
10881 if (PL_lex_stuff) {
10882 SvREFCNT_dec(PL_lex_stuff);
10883 PL_lex_stuff = NULL;
10885 Perl_croak(aTHX_ "Substitution replacement not terminated");
10887 PL_multi_start = first_start; /* so whole substitution is taken together */
10889 pm = (PMOP*)newPMOP(OP_SUBST, 0);
10892 if (PL_madskills) {
10893 CURMAD('z', PL_thisopen);
10894 CURMAD('R', PL_thisstuff);
10895 CURMAD('Z', PL_thisclose);
10901 if (*s == EXEC_PAT_MOD) {
10905 else if (strchr(S_PAT_MODS, *s))
10906 pmflag(&pm->op_pmflags,*s++);
10912 if (PL_madskills) {
10914 curmad('m', newSVpvn(modstart, s - modstart));
10915 append_madprops(PL_thismad, (OP*)pm, 0);
10919 if ((pm->op_pmflags & PMf_CONTINUE) && ckWARN(WARN_REGEXP)) {
10920 Perl_warner(aTHX_ packWARN(WARN_REGEXP), "Use of /c modifier is meaningless in s///" );
10924 SV * const repl = newSVpvs("");
10926 PL_sublex_info.super_bufptr = s;
10927 PL_sublex_info.super_bufend = PL_bufend;
10929 pm->op_pmflags |= PMf_EVAL;
10931 sv_catpv(repl, (const char *)(es ? "eval " : "do "));
10932 sv_catpvs(repl, "{");
10933 sv_catsv(repl, PL_lex_repl);
10934 if (strchr(SvPVX(PL_lex_repl), '#'))
10935 sv_catpvs(repl, "\n");
10936 sv_catpvs(repl, "}");
10938 SvREFCNT_dec(PL_lex_repl);
10939 PL_lex_repl = repl;
10942 pm->op_pmpermflags = pm->op_pmflags;
10943 PL_lex_op = (OP*)pm;
10944 yylval.ival = OP_SUBST;
10949 S_scan_trans(pTHX_ char *start)
10962 yylval.ival = OP_NULL;
10964 s = scan_str(start,!!PL_madskills,FALSE);
10966 Perl_croak(aTHX_ "Transliteration pattern not terminated");
10968 if (s[-1] == PL_multi_open)
10971 if (PL_madskills) {
10972 CURMAD('q', PL_thisopen);
10973 CURMAD('_', PL_thiswhite);
10974 CURMAD('E', PL_thisstuff);
10975 CURMAD('Q', PL_thisclose);
10976 PL_realtokenstart = s - SvPVX(PL_linestr);
10980 s = scan_str(s,!!PL_madskills,FALSE);
10982 if (PL_lex_stuff) {
10983 SvREFCNT_dec(PL_lex_stuff);
10984 PL_lex_stuff = NULL;
10986 Perl_croak(aTHX_ "Transliteration replacement not terminated");
10988 if (PL_madskills) {
10989 CURMAD('z', PL_thisopen);
10990 CURMAD('R', PL_thisstuff);
10991 CURMAD('Z', PL_thisclose);
10994 complement = del = squash = 0;
11001 complement = OPpTRANS_COMPLEMENT;
11004 del = OPpTRANS_DELETE;
11007 squash = OPpTRANS_SQUASH;
11016 tbl = (short *)PerlMemShared_calloc(complement&&!del?258:256, sizeof(short));
11017 o = newPVOP(OP_TRANS, 0, (char*)tbl);
11018 o->op_private &= ~OPpTRANS_ALL;
11019 o->op_private |= del|squash|complement|
11020 (DO_UTF8(PL_lex_stuff)? OPpTRANS_FROM_UTF : 0)|
11021 (DO_UTF8(PL_lex_repl) ? OPpTRANS_TO_UTF : 0);
11024 yylval.ival = OP_TRANS;
11027 if (PL_madskills) {
11029 curmad('m', newSVpvn(modstart, s - modstart));
11030 append_madprops(PL_thismad, o, 0);
11039 S_scan_heredoc(pTHX_ register char *s)
11043 I32 op_type = OP_SCALAR;
11047 const char *found_newline;
11051 const int outer = (PL_rsfp && !(PL_lex_inwhat == OP_SCALAR));
11053 I32 stuffstart = s - SvPVX(PL_linestr);
11056 PL_realtokenstart = -1;
11061 e = PL_tokenbuf + sizeof PL_tokenbuf - 1;
11065 while (SPACE_OR_TAB(*peek))
11067 if (*peek == '`' || *peek == '\'' || *peek =='"') {
11070 s = delimcpy(d, e, s, PL_bufend, term, &len);
11080 if (!isALNUM_lazy_if(s,UTF))
11081 deprecate_old("bare << to mean <<\"\"");
11082 for (; isALNUM_lazy_if(s,UTF); s++) {
11087 if (d >= PL_tokenbuf + sizeof PL_tokenbuf - 1)
11088 Perl_croak(aTHX_ "Delimiter for here document is too long");
11091 len = d - PL_tokenbuf;
11094 if (PL_madskills) {
11095 tstart = PL_tokenbuf + !outer;
11096 PL_thisclose = newSVpvn(tstart, len - !outer);
11097 tstart = SvPVX(PL_linestr) + stuffstart;
11098 PL_thisopen = newSVpvn(tstart, s - tstart);
11099 stuffstart = s - SvPVX(PL_linestr);
11102 #ifndef PERL_STRICT_CR
11103 d = strchr(s, '\r');
11105 char * const olds = s;
11107 while (s < PL_bufend) {
11113 else if (*s == '\n' && s[1] == '\r') { /* \015\013 on a mac? */
11122 SvCUR_set(PL_linestr, PL_bufend - SvPVX_const(PL_linestr));
11129 if ( outer || !(found_newline = (char*)memchr((void*)s, '\n', PL_bufend - s)) ) {
11130 herewas = newSVpvn(s,PL_bufend-s);
11134 herewas = newSVpvn(s-1,found_newline-s+1);
11137 herewas = newSVpvn(s,found_newline-s);
11141 if (PL_madskills) {
11142 tstart = SvPVX(PL_linestr) + stuffstart;
11144 sv_catpvn(PL_thisstuff, tstart, s - tstart);
11146 PL_thisstuff = newSVpvn(tstart, s - tstart);
11149 s += SvCUR(herewas);
11152 stuffstart = s - SvPVX(PL_linestr);
11158 tmpstr = newSV_type(SVt_PVIV);
11159 SvGROW(tmpstr, 80);
11160 if (term == '\'') {
11161 op_type = OP_CONST;
11162 SvIV_set(tmpstr, -1);
11164 else if (term == '`') {
11165 op_type = OP_BACKTICK;
11166 SvIV_set(tmpstr, '\\');
11170 PL_multi_start = CopLINE(PL_curcop);
11171 PL_multi_open = PL_multi_close = '<';
11172 term = *PL_tokenbuf;
11173 if (PL_lex_inwhat == OP_SUBST && PL_in_eval && !PL_rsfp) {
11174 char * const bufptr = PL_sublex_info.super_bufptr;
11175 char * const bufend = PL_sublex_info.super_bufend;
11176 char * const olds = s - SvCUR(herewas);
11177 s = strchr(bufptr, '\n');
11181 while (s < bufend &&
11182 (*s != term || memNE(s,PL_tokenbuf,len)) ) {
11184 CopLINE_inc(PL_curcop);
11187 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
11188 missingterm(PL_tokenbuf);
11190 sv_setpvn(herewas,bufptr,d-bufptr+1);
11191 sv_setpvn(tmpstr,d+1,s-d);
11193 sv_catpvn(herewas,s,bufend-s);
11194 Copy(SvPVX_const(herewas),bufptr,SvCUR(herewas) + 1,char);
11201 while (s < PL_bufend &&
11202 (*s != term || memNE(s,PL_tokenbuf,len)) ) {
11204 CopLINE_inc(PL_curcop);
11206 if (s >= PL_bufend) {
11207 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
11208 missingterm(PL_tokenbuf);
11210 sv_setpvn(tmpstr,d+1,s-d);
11212 if (PL_madskills) {
11214 sv_catpvn(PL_thisstuff, d + 1, s - d);
11216 PL_thisstuff = newSVpvn(d + 1, s - d);
11217 stuffstart = s - SvPVX(PL_linestr);
11221 CopLINE_inc(PL_curcop); /* the preceding stmt passes a newline */
11223 sv_catpvn(herewas,s,PL_bufend-s);
11224 sv_setsv(PL_linestr,herewas);
11225 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = s = PL_linestart = SvPVX(PL_linestr);
11226 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
11227 PL_last_lop = PL_last_uni = NULL;
11230 sv_setpvn(tmpstr,"",0); /* avoid "uninitialized" warning */
11231 while (s >= PL_bufend) { /* multiple line string? */
11233 if (PL_madskills) {
11234 tstart = SvPVX(PL_linestr) + stuffstart;
11236 sv_catpvn(PL_thisstuff, tstart, PL_bufend - tstart);
11238 PL_thisstuff = newSVpvn(tstart, PL_bufend - tstart);
11242 !(PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = filter_gets(PL_linestr, PL_rsfp, 0))) {
11243 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
11244 missingterm(PL_tokenbuf);
11247 stuffstart = s - SvPVX(PL_linestr);
11249 CopLINE_inc(PL_curcop);
11250 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
11251 PL_last_lop = PL_last_uni = NULL;
11252 #ifndef PERL_STRICT_CR
11253 if (PL_bufend - PL_linestart >= 2) {
11254 if ((PL_bufend[-2] == '\r' && PL_bufend[-1] == '\n') ||
11255 (PL_bufend[-2] == '\n' && PL_bufend[-1] == '\r'))
11257 PL_bufend[-2] = '\n';
11259 SvCUR_set(PL_linestr, PL_bufend - SvPVX_const(PL_linestr));
11261 else if (PL_bufend[-1] == '\r')
11262 PL_bufend[-1] = '\n';
11264 else if (PL_bufend - PL_linestart == 1 && PL_bufend[-1] == '\r')
11265 PL_bufend[-1] = '\n';
11267 if (PERLDB_LINE && PL_curstash != PL_debstash)
11268 update_debugger_info(PL_linestr, NULL, 0);
11269 if (*s == term && memEQ(s,PL_tokenbuf,len)) {
11270 STRLEN off = PL_bufend - 1 - SvPVX_const(PL_linestr);
11271 *(SvPVX(PL_linestr) + off ) = ' ';
11272 sv_catsv(PL_linestr,herewas);
11273 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
11274 s = SvPVX(PL_linestr) + off; /* In case PV of PL_linestr moved. */
11278 sv_catsv(tmpstr,PL_linestr);
11283 PL_multi_end = CopLINE(PL_curcop);
11284 if (SvCUR(tmpstr) + 5 < SvLEN(tmpstr)) {
11285 SvPV_shrink_to_cur(tmpstr);
11287 SvREFCNT_dec(herewas);
11289 if (UTF && is_utf8_string((U8*)SvPVX_const(tmpstr), SvCUR(tmpstr)))
11291 else if (PL_encoding)
11292 sv_recode_to_utf8(tmpstr, PL_encoding);
11294 PL_lex_stuff = tmpstr;
11295 yylval.ival = op_type;
11299 /* scan_inputsymbol
11300 takes: current position in input buffer
11301 returns: new position in input buffer
11302 side-effects: yylval and lex_op are set.
11307 <FH> read from filehandle
11308 <pkg::FH> read from package qualified filehandle
11309 <pkg'FH> read from package qualified filehandle
11310 <$fh> read from filehandle in $fh
11311 <*.h> filename glob
11316 S_scan_inputsymbol(pTHX_ char *start)
11319 register char *s = start; /* current position in buffer */
11323 char *d = PL_tokenbuf; /* start of temp holding space */
11324 const char * const e = PL_tokenbuf + sizeof PL_tokenbuf; /* end of temp holding space */
11326 end = strchr(s, '\n');
11329 s = delimcpy(d, e, s + 1, end, '>', &len); /* extract until > */
11331 /* die if we didn't have space for the contents of the <>,
11332 or if it didn't end, or if we see a newline
11335 if (len >= (I32)sizeof PL_tokenbuf)
11336 Perl_croak(aTHX_ "Excessively long <> operator");
11338 Perl_croak(aTHX_ "Unterminated <> operator");
11343 Remember, only scalar variables are interpreted as filehandles by
11344 this code. Anything more complex (e.g., <$fh{$num}>) will be
11345 treated as a glob() call.
11346 This code makes use of the fact that except for the $ at the front,
11347 a scalar variable and a filehandle look the same.
11349 if (*d == '$' && d[1]) d++;
11351 /* allow <Pkg'VALUE> or <Pkg::VALUE> */
11352 while (*d && (isALNUM_lazy_if(d,UTF) || *d == '\'' || *d == ':'))
11355 /* If we've tried to read what we allow filehandles to look like, and
11356 there's still text left, then it must be a glob() and not a getline.
11357 Use scan_str to pull out the stuff between the <> and treat it
11358 as nothing more than a string.
11361 if (d - PL_tokenbuf != len) {
11362 yylval.ival = OP_GLOB;
11364 s = scan_str(start,!!PL_madskills,FALSE);
11366 Perl_croak(aTHX_ "Glob not terminated");
11370 bool readline_overriden = FALSE;
11373 /* we're in a filehandle read situation */
11376 /* turn <> into <ARGV> */
11378 Copy("ARGV",d,5,char);
11380 /* Check whether readline() is overriden */
11381 gv_readline = gv_fetchpvs("readline", GV_NOTQUAL, SVt_PVCV);
11383 && GvCVu(gv_readline) && GvIMPORTED_CV(gv_readline))
11385 ((gvp = (GV**)hv_fetchs(PL_globalstash, "readline", FALSE))
11386 && (gv_readline = *gvp) && isGV_with_GP(gv_readline)
11387 && GvCVu(gv_readline) && GvIMPORTED_CV(gv_readline)))
11388 readline_overriden = TRUE;
11390 /* if <$fh>, create the ops to turn the variable into a
11394 /* try to find it in the pad for this block, otherwise find
11395 add symbol table ops
11397 const PADOFFSET tmp = pad_findmy(d);
11398 if (tmp != NOT_IN_PAD) {
11399 if (PAD_COMPNAME_FLAGS_isOUR(tmp)) {
11400 HV * const stash = PAD_COMPNAME_OURSTASH(tmp);
11401 HEK * const stashname = HvNAME_HEK(stash);
11402 SV * const sym = sv_2mortal(newSVhek(stashname));
11403 sv_catpvs(sym, "::");
11404 sv_catpv(sym, d+1);
11409 OP * const o = newOP(OP_PADSV, 0);
11411 PL_lex_op = readline_overriden
11412 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
11413 append_elem(OP_LIST, o,
11414 newCVREF(0, newGVOP(OP_GV,0,gv_readline))))
11415 : (OP*)newUNOP(OP_READLINE, 0, o);
11424 ? (GV_ADDMULTI | GV_ADDINEVAL)
11427 PL_lex_op = readline_overriden
11428 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
11429 append_elem(OP_LIST,
11430 newUNOP(OP_RV2SV, 0, newGVOP(OP_GV, 0, gv)),
11431 newCVREF(0, newGVOP(OP_GV, 0, gv_readline))))
11432 : (OP*)newUNOP(OP_READLINE, 0,
11433 newUNOP(OP_RV2SV, 0,
11434 newGVOP(OP_GV, 0, gv)));
11436 if (!readline_overriden)
11437 PL_lex_op->op_flags |= OPf_SPECIAL;
11438 /* we created the ops in PL_lex_op, so make yylval.ival a null op */
11439 yylval.ival = OP_NULL;
11442 /* If it's none of the above, it must be a literal filehandle
11443 (<Foo::BAR> or <FOO>) so build a simple readline OP */
11445 GV * const gv = gv_fetchpv(d, GV_ADD, SVt_PVIO);
11446 PL_lex_op = readline_overriden
11447 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
11448 append_elem(OP_LIST,
11449 newGVOP(OP_GV, 0, gv),
11450 newCVREF(0, newGVOP(OP_GV, 0, gv_readline))))
11451 : (OP*)newUNOP(OP_READLINE, 0, newGVOP(OP_GV, 0, gv));
11452 yylval.ival = OP_NULL;
11461 takes: start position in buffer
11462 keep_quoted preserve \ on the embedded delimiter(s)
11463 keep_delims preserve the delimiters around the string
11464 returns: position to continue reading from buffer
11465 side-effects: multi_start, multi_close, lex_repl or lex_stuff, and
11466 updates the read buffer.
11468 This subroutine pulls a string out of the input. It is called for:
11469 q single quotes q(literal text)
11470 ' single quotes 'literal text'
11471 qq double quotes qq(interpolate $here please)
11472 " double quotes "interpolate $here please"
11473 qx backticks qx(/bin/ls -l)
11474 ` backticks `/bin/ls -l`
11475 qw quote words @EXPORT_OK = qw( func() $spam )
11476 m// regexp match m/this/
11477 s/// regexp substitute s/this/that/
11478 tr/// string transliterate tr/this/that/
11479 y/// string transliterate y/this/that/
11480 ($*@) sub prototypes sub foo ($)
11481 (stuff) sub attr parameters sub foo : attr(stuff)
11482 <> readline or globs <FOO>, <>, <$fh>, or <*.c>
11484 In most of these cases (all but <>, patterns and transliterate)
11485 yylex() calls scan_str(). m// makes yylex() call scan_pat() which
11486 calls scan_str(). s/// makes yylex() call scan_subst() which calls
11487 scan_str(). tr/// and y/// make yylex() call scan_trans() which
11490 It skips whitespace before the string starts, and treats the first
11491 character as the delimiter. If the delimiter is one of ([{< then
11492 the corresponding "close" character )]}> is used as the closing
11493 delimiter. It allows quoting of delimiters, and if the string has
11494 balanced delimiters ([{<>}]) it allows nesting.
11496 On success, the SV with the resulting string is put into lex_stuff or,
11497 if that is already non-NULL, into lex_repl. The second case occurs only
11498 when parsing the RHS of the special constructs s/// and tr/// (y///).
11499 For convenience, the terminating delimiter character is stuffed into
11504 S_scan_str(pTHX_ char *start, int keep_quoted, int keep_delims)
11507 SV *sv; /* scalar value: string */
11508 const char *tmps; /* temp string, used for delimiter matching */
11509 register char *s = start; /* current position in the buffer */
11510 register char term; /* terminating character */
11511 register char *to; /* current position in the sv's data */
11512 I32 brackets = 1; /* bracket nesting level */
11513 bool has_utf8 = FALSE; /* is there any utf8 content? */
11514 I32 termcode; /* terminating char. code */
11515 U8 termstr[UTF8_MAXBYTES]; /* terminating string */
11516 STRLEN termlen; /* length of terminating string */
11517 int last_off = 0; /* last position for nesting bracket */
11523 /* skip space before the delimiter */
11529 if (PL_realtokenstart >= 0) {
11530 stuffstart = PL_realtokenstart;
11531 PL_realtokenstart = -1;
11534 stuffstart = start - SvPVX(PL_linestr);
11536 /* mark where we are, in case we need to report errors */
11539 /* after skipping whitespace, the next character is the terminator */
11542 termcode = termstr[0] = term;
11546 termcode = utf8_to_uvchr((U8*)s, &termlen);
11547 Copy(s, termstr, termlen, U8);
11548 if (!UTF8_IS_INVARIANT(term))
11552 /* mark where we are */
11553 PL_multi_start = CopLINE(PL_curcop);
11554 PL_multi_open = term;
11556 /* find corresponding closing delimiter */
11557 if (term && (tmps = strchr("([{< )]}> )]}>",term)))
11558 termcode = termstr[0] = term = tmps[5];
11560 PL_multi_close = term;
11562 /* create a new SV to hold the contents. 79 is the SV's initial length.
11563 What a random number. */
11564 sv = newSV_type(SVt_PVIV);
11566 SvIV_set(sv, termcode);
11567 (void)SvPOK_only(sv); /* validate pointer */
11569 /* move past delimiter and try to read a complete string */
11571 sv_catpvn(sv, s, termlen);
11574 tstart = SvPVX(PL_linestr) + stuffstart;
11575 if (!PL_thisopen && !keep_delims) {
11576 PL_thisopen = newSVpvn(tstart, s - tstart);
11577 stuffstart = s - SvPVX(PL_linestr);
11581 if (PL_encoding && !UTF) {
11585 int offset = s - SvPVX_const(PL_linestr);
11586 const bool found = sv_cat_decode(sv, PL_encoding, PL_linestr,
11587 &offset, (char*)termstr, termlen);
11588 const char * const ns = SvPVX_const(PL_linestr) + offset;
11589 char * const svlast = SvEND(sv) - 1;
11591 for (; s < ns; s++) {
11592 if (*s == '\n' && !PL_rsfp)
11593 CopLINE_inc(PL_curcop);
11596 goto read_more_line;
11598 /* handle quoted delimiters */
11599 if (SvCUR(sv) > 1 && *(svlast-1) == '\\') {
11601 for (t = svlast-2; t >= SvPVX_const(sv) && *t == '\\';)
11603 if ((svlast-1 - t) % 2) {
11604 if (!keep_quoted) {
11605 *(svlast-1) = term;
11607 SvCUR_set(sv, SvCUR(sv) - 1);
11612 if (PL_multi_open == PL_multi_close) {
11618 for (t = w = SvPVX(sv)+last_off; t < svlast; w++, t++) {
11619 /* At here, all closes are "was quoted" one,
11620 so we don't check PL_multi_close. */
11622 if (!keep_quoted && *(t+1) == PL_multi_open)
11627 else if (*t == PL_multi_open)
11635 SvCUR_set(sv, w - SvPVX_const(sv));
11637 last_off = w - SvPVX(sv);
11638 if (--brackets <= 0)
11643 if (!keep_delims) {
11644 SvCUR_set(sv, SvCUR(sv) - 1);
11650 /* extend sv if need be */
11651 SvGROW(sv, SvCUR(sv) + (PL_bufend - s) + 1);
11652 /* set 'to' to the next character in the sv's string */
11653 to = SvPVX(sv)+SvCUR(sv);
11655 /* if open delimiter is the close delimiter read unbridle */
11656 if (PL_multi_open == PL_multi_close) {
11657 for (; s < PL_bufend; s++,to++) {
11658 /* embedded newlines increment the current line number */
11659 if (*s == '\n' && !PL_rsfp)
11660 CopLINE_inc(PL_curcop);
11661 /* handle quoted delimiters */
11662 if (*s == '\\' && s+1 < PL_bufend && term != '\\') {
11663 if (!keep_quoted && s[1] == term)
11665 /* any other quotes are simply copied straight through */
11669 /* terminate when run out of buffer (the for() condition), or
11670 have found the terminator */
11671 else if (*s == term) {
11674 if (s+termlen <= PL_bufend && memEQ(s, (char*)termstr, termlen))
11677 else if (!has_utf8 && !UTF8_IS_INVARIANT((U8)*s) && UTF)
11683 /* if the terminator isn't the same as the start character (e.g.,
11684 matched brackets), we have to allow more in the quoting, and
11685 be prepared for nested brackets.
11688 /* read until we run out of string, or we find the terminator */
11689 for (; s < PL_bufend; s++,to++) {
11690 /* embedded newlines increment the line count */
11691 if (*s == '\n' && !PL_rsfp)
11692 CopLINE_inc(PL_curcop);
11693 /* backslashes can escape the open or closing characters */
11694 if (*s == '\\' && s+1 < PL_bufend) {
11695 if (!keep_quoted &&
11696 ((s[1] == PL_multi_open) || (s[1] == PL_multi_close)))
11701 /* allow nested opens and closes */
11702 else if (*s == PL_multi_close && --brackets <= 0)
11704 else if (*s == PL_multi_open)
11706 else if (!has_utf8 && !UTF8_IS_INVARIANT((U8)*s) && UTF)
11711 /* terminate the copied string and update the sv's end-of-string */
11713 SvCUR_set(sv, to - SvPVX_const(sv));
11716 * this next chunk reads more into the buffer if we're not done yet
11720 break; /* handle case where we are done yet :-) */
11722 #ifndef PERL_STRICT_CR
11723 if (to - SvPVX_const(sv) >= 2) {
11724 if ((to[-2] == '\r' && to[-1] == '\n') ||
11725 (to[-2] == '\n' && to[-1] == '\r'))
11729 SvCUR_set(sv, to - SvPVX_const(sv));
11731 else if (to[-1] == '\r')
11734 else if (to - SvPVX_const(sv) == 1 && to[-1] == '\r')
11739 /* if we're out of file, or a read fails, bail and reset the current
11740 line marker so we can report where the unterminated string began
11743 if (PL_madskills) {
11744 char * const tstart = SvPVX(PL_linestr) + stuffstart;
11746 sv_catpvn(PL_thisstuff, tstart, PL_bufend - tstart);
11748 PL_thisstuff = newSVpvn(tstart, PL_bufend - tstart);
11752 !(PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = filter_gets(PL_linestr, PL_rsfp, 0))) {
11754 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
11760 /* we read a line, so increment our line counter */
11761 CopLINE_inc(PL_curcop);
11763 /* update debugger info */
11764 if (PERLDB_LINE && PL_curstash != PL_debstash)
11765 update_debugger_info(PL_linestr, NULL, 0);
11767 /* having changed the buffer, we must update PL_bufend */
11768 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
11769 PL_last_lop = PL_last_uni = NULL;
11772 /* at this point, we have successfully read the delimited string */
11774 if (!PL_encoding || UTF) {
11776 if (PL_madskills) {
11777 char * const tstart = SvPVX(PL_linestr) + stuffstart;
11778 const int len = s - tstart;
11780 sv_catpvn(PL_thisstuff, tstart, len);
11782 PL_thisstuff = newSVpvn(tstart, len);
11783 if (!PL_thisclose && !keep_delims)
11784 PL_thisclose = newSVpvn(s,termlen);
11789 sv_catpvn(sv, s, termlen);
11794 if (PL_madskills) {
11795 char * const tstart = SvPVX(PL_linestr) + stuffstart;
11796 const int len = s - tstart - termlen;
11798 sv_catpvn(PL_thisstuff, tstart, len);
11800 PL_thisstuff = newSVpvn(tstart, len);
11801 if (!PL_thisclose && !keep_delims)
11802 PL_thisclose = newSVpvn(s - termlen,termlen);
11806 if (has_utf8 || PL_encoding)
11809 PL_multi_end = CopLINE(PL_curcop);
11811 /* if we allocated too much space, give some back */
11812 if (SvCUR(sv) + 5 < SvLEN(sv)) {
11813 SvLEN_set(sv, SvCUR(sv) + 1);
11814 SvPV_renew(sv, SvLEN(sv));
11817 /* decide whether this is the first or second quoted string we've read
11830 takes: pointer to position in buffer
11831 returns: pointer to new position in buffer
11832 side-effects: builds ops for the constant in yylval.op
11834 Read a number in any of the formats that Perl accepts:
11836 \d(_?\d)*(\.(\d(_?\d)*)?)?[Ee][\+\-]?(\d(_?\d)*) 12 12.34 12.
11837 \.\d(_?\d)*[Ee][\+\-]?(\d(_?\d)*) .34
11840 0x[0-9A-Fa-f](_?[0-9A-Fa-f])*
11842 Like most scan_ routines, it uses the PL_tokenbuf buffer to hold the
11845 If it reads a number without a decimal point or an exponent, it will
11846 try converting the number to an integer and see if it can do so
11847 without loss of precision.
11851 Perl_scan_num(pTHX_ const char *start, YYSTYPE* lvalp)
11854 register const char *s = start; /* current position in buffer */
11855 register char *d; /* destination in temp buffer */
11856 register char *e; /* end of temp buffer */
11857 NV nv; /* number read, as a double */
11858 SV *sv = NULL; /* place to put the converted number */
11859 bool floatit; /* boolean: int or float? */
11860 const char *lastub = NULL; /* position of last underbar */
11861 static char const number_too_long[] = "Number too long";
11863 /* We use the first character to decide what type of number this is */
11867 Perl_croak(aTHX_ "panic: scan_num");
11869 /* if it starts with a 0, it could be an octal number, a decimal in
11870 0.13 disguise, or a hexadecimal number, or a binary number. */
11874 u holds the "number so far"
11875 shift the power of 2 of the base
11876 (hex == 4, octal == 3, binary == 1)
11877 overflowed was the number more than we can hold?
11879 Shift is used when we add a digit. It also serves as an "are
11880 we in octal/hex/binary?" indicator to disallow hex characters
11881 when in octal mode.
11886 bool overflowed = FALSE;
11887 bool just_zero = TRUE; /* just plain 0 or binary number? */
11888 static const NV nvshift[5] = { 1.0, 2.0, 4.0, 8.0, 16.0 };
11889 static const char* const bases[5] =
11890 { "", "binary", "", "octal", "hexadecimal" };
11891 static const char* const Bases[5] =
11892 { "", "Binary", "", "Octal", "Hexadecimal" };
11893 static const char* const maxima[5] =
11895 "0b11111111111111111111111111111111",
11899 const char *base, *Base, *max;
11901 /* check for hex */
11906 } else if (s[1] == 'b') {
11911 /* check for a decimal in disguise */
11912 else if (s[1] == '.' || s[1] == 'e' || s[1] == 'E')
11914 /* so it must be octal */
11921 if (ckWARN(WARN_SYNTAX))
11922 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
11923 "Misplaced _ in number");
11927 base = bases[shift];
11928 Base = Bases[shift];
11929 max = maxima[shift];
11931 /* read the rest of the number */
11933 /* x is used in the overflow test,
11934 b is the digit we're adding on. */
11939 /* if we don't mention it, we're done */
11943 /* _ are ignored -- but warned about if consecutive */
11945 if (lastub && s == lastub + 1 && ckWARN(WARN_SYNTAX))
11946 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
11947 "Misplaced _ in number");
11951 /* 8 and 9 are not octal */
11952 case '8': case '9':
11954 yyerror(Perl_form(aTHX_ "Illegal octal digit '%c'", *s));
11958 case '2': case '3': case '4':
11959 case '5': case '6': case '7':
11961 yyerror(Perl_form(aTHX_ "Illegal binary digit '%c'", *s));
11964 case '0': case '1':
11965 b = *s++ & 15; /* ASCII digit -> value of digit */
11969 case 'a': case 'b': case 'c': case 'd': case 'e': case 'f':
11970 case 'A': case 'B': case 'C': case 'D': case 'E': case 'F':
11971 /* make sure they said 0x */
11974 b = (*s++ & 7) + 9;
11976 /* Prepare to put the digit we have onto the end
11977 of the number so far. We check for overflows.
11983 x = u << shift; /* make room for the digit */
11985 if ((x >> shift) != u
11986 && !(PL_hints & HINT_NEW_BINARY)) {
11989 if (ckWARN_d(WARN_OVERFLOW))
11990 Perl_warner(aTHX_ packWARN(WARN_OVERFLOW),
11991 "Integer overflow in %s number",
11994 u = x | b; /* add the digit to the end */
11997 n *= nvshift[shift];
11998 /* If an NV has not enough bits in its
11999 * mantissa to represent an UV this summing of
12000 * small low-order numbers is a waste of time
12001 * (because the NV cannot preserve the
12002 * low-order bits anyway): we could just
12003 * remember when did we overflow and in the
12004 * end just multiply n by the right
12012 /* if we get here, we had success: make a scalar value from
12017 /* final misplaced underbar check */
12018 if (s[-1] == '_') {
12019 if (ckWARN(WARN_SYNTAX))
12020 Perl_warner(aTHX_ packWARN(WARN_SYNTAX), "Misplaced _ in number");
12025 if (n > 4294967295.0 && ckWARN(WARN_PORTABLE))
12026 Perl_warner(aTHX_ packWARN(WARN_PORTABLE),
12027 "%s number > %s non-portable",
12033 if (u > 0xffffffff && ckWARN(WARN_PORTABLE))
12034 Perl_warner(aTHX_ packWARN(WARN_PORTABLE),
12035 "%s number > %s non-portable",
12040 if (just_zero && (PL_hints & HINT_NEW_INTEGER))
12041 sv = new_constant(start, s - start, "integer",
12043 else if (PL_hints & HINT_NEW_BINARY)
12044 sv = new_constant(start, s - start, "binary", sv, NULL, NULL);
12049 handle decimal numbers.
12050 we're also sent here when we read a 0 as the first digit
12052 case '1': case '2': case '3': case '4': case '5':
12053 case '6': case '7': case '8': case '9': case '.':
12056 e = PL_tokenbuf + sizeof PL_tokenbuf - 6; /* room for various punctuation */
12059 /* read next group of digits and _ and copy into d */
12060 while (isDIGIT(*s) || *s == '_') {
12061 /* skip underscores, checking for misplaced ones
12065 if (lastub && s == lastub + 1 && ckWARN(WARN_SYNTAX))
12066 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
12067 "Misplaced _ in number");
12071 /* check for end of fixed-length buffer */
12073 Perl_croak(aTHX_ number_too_long);
12074 /* if we're ok, copy the character */
12079 /* final misplaced underbar check */
12080 if (lastub && s == lastub + 1) {
12081 if (ckWARN(WARN_SYNTAX))
12082 Perl_warner(aTHX_ packWARN(WARN_SYNTAX), "Misplaced _ in number");
12085 /* read a decimal portion if there is one. avoid
12086 3..5 being interpreted as the number 3. followed
12089 if (*s == '.' && s[1] != '.') {
12094 if (ckWARN(WARN_SYNTAX))
12095 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
12096 "Misplaced _ in number");
12100 /* copy, ignoring underbars, until we run out of digits.
12102 for (; isDIGIT(*s) || *s == '_'; s++) {
12103 /* fixed length buffer check */
12105 Perl_croak(aTHX_ number_too_long);
12107 if (lastub && s == lastub + 1 && ckWARN(WARN_SYNTAX))
12108 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
12109 "Misplaced _ in number");
12115 /* fractional part ending in underbar? */
12116 if (s[-1] == '_') {
12117 if (ckWARN(WARN_SYNTAX))
12118 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
12119 "Misplaced _ in number");
12121 if (*s == '.' && isDIGIT(s[1])) {
12122 /* oops, it's really a v-string, but without the "v" */
12128 /* read exponent part, if present */
12129 if ((*s == 'e' || *s == 'E') && strchr("+-0123456789_", s[1])) {
12133 /* regardless of whether user said 3E5 or 3e5, use lower 'e' */
12134 *d++ = 'e'; /* At least some Mach atof()s don't grok 'E' */
12136 /* stray preinitial _ */
12138 if (ckWARN(WARN_SYNTAX))
12139 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
12140 "Misplaced _ in number");
12144 /* allow positive or negative exponent */
12145 if (*s == '+' || *s == '-')
12148 /* stray initial _ */
12150 if (ckWARN(WARN_SYNTAX))
12151 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
12152 "Misplaced _ in number");
12156 /* read digits of exponent */
12157 while (isDIGIT(*s) || *s == '_') {
12160 Perl_croak(aTHX_ number_too_long);
12164 if (((lastub && s == lastub + 1) ||
12165 (!isDIGIT(s[1]) && s[1] != '_'))
12166 && ckWARN(WARN_SYNTAX))
12167 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
12168 "Misplaced _ in number");
12175 /* make an sv from the string */
12179 We try to do an integer conversion first if no characters
12180 indicating "float" have been found.
12185 const int flags = grok_number (PL_tokenbuf, d - PL_tokenbuf, &uv);
12187 if (flags == IS_NUMBER_IN_UV) {
12189 sv_setiv(sv, uv); /* Prefer IVs over UVs. */
12192 } else if (flags == (IS_NUMBER_IN_UV | IS_NUMBER_NEG)) {
12193 if (uv <= (UV) IV_MIN)
12194 sv_setiv(sv, -(IV)uv);
12201 /* terminate the string */
12203 nv = Atof(PL_tokenbuf);
12207 if ( floatit ? (PL_hints & HINT_NEW_FLOAT) :
12208 (PL_hints & HINT_NEW_INTEGER) )
12209 sv = new_constant(PL_tokenbuf,
12212 (floatit ? "float" : "integer"),
12216 /* if it starts with a v, it could be a v-string */
12219 sv = newSV(5); /* preallocate storage space */
12220 s = scan_vstring(s,sv);
12224 /* make the op for the constant and return */
12227 lvalp->opval = newSVOP(OP_CONST, 0, sv);
12229 lvalp->opval = NULL;
12235 S_scan_formline(pTHX_ register char *s)
12238 register char *eol;
12240 SV * const stuff = newSVpvs("");
12241 bool needargs = FALSE;
12242 bool eofmt = FALSE;
12244 char *tokenstart = s;
12247 if (PL_madskills) {
12248 savewhite = PL_thiswhite;
12253 while (!needargs) {
12256 #ifdef PERL_STRICT_CR
12257 while (SPACE_OR_TAB(*t))
12260 while (SPACE_OR_TAB(*t) || *t == '\r')
12263 if (*t == '\n' || t == PL_bufend) {
12268 if (PL_in_eval && !PL_rsfp) {
12269 eol = (char *) memchr(s,'\n',PL_bufend-s);
12274 eol = PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
12276 for (t = s; t < eol; t++) {
12277 if (*t == '~' && t[1] == '~' && SvCUR(stuff)) {
12279 goto enough; /* ~~ must be first line in formline */
12281 if (*t == '@' || *t == '^')
12285 sv_catpvn(stuff, s, eol-s);
12286 #ifndef PERL_STRICT_CR
12287 if (eol-s > 1 && eol[-2] == '\r' && eol[-1] == '\n') {
12288 char *end = SvPVX(stuff) + SvCUR(stuff);
12291 SvCUR_set(stuff, SvCUR(stuff) - 1);
12301 if (PL_madskills) {
12303 sv_catpvn(PL_thistoken, tokenstart, PL_bufend - tokenstart);
12305 PL_thistoken = newSVpvn(tokenstart, PL_bufend - tokenstart);
12308 s = filter_gets(PL_linestr, PL_rsfp, 0);
12310 tokenstart = PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = SvPVX(PL_linestr);
12312 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = SvPVX(PL_linestr);
12314 PL_bufend = PL_bufptr + SvCUR(PL_linestr);
12315 PL_last_lop = PL_last_uni = NULL;
12324 if (SvCUR(stuff)) {
12327 PL_lex_state = LEX_NORMAL;
12328 start_force(PL_curforce);
12329 NEXTVAL_NEXTTOKE.ival = 0;
12333 PL_lex_state = LEX_FORMLINE;
12335 if (UTF && is_utf8_string((U8*)SvPVX_const(stuff), SvCUR(stuff)))
12337 else if (PL_encoding)
12338 sv_recode_to_utf8(stuff, PL_encoding);
12340 start_force(PL_curforce);
12341 NEXTVAL_NEXTTOKE.opval = (OP*)newSVOP(OP_CONST, 0, stuff);
12343 start_force(PL_curforce);
12344 NEXTVAL_NEXTTOKE.ival = OP_FORMLINE;
12348 SvREFCNT_dec(stuff);
12350 PL_lex_formbrack = 0;
12354 if (PL_madskills) {
12356 sv_catpvn(PL_thistoken, tokenstart, s - tokenstart);
12358 PL_thistoken = newSVpvn(tokenstart, s - tokenstart);
12359 PL_thiswhite = savewhite;
12371 PL_cshlen = strlen(PL_cshname);
12373 #if defined(USE_ITHREADS)
12374 PERL_UNUSED_CONTEXT;
12380 Perl_start_subparse(pTHX_ I32 is_format, U32 flags)
12383 const I32 oldsavestack_ix = PL_savestack_ix;
12384 CV* const outsidecv = PL_compcv;
12387 assert(SvTYPE(PL_compcv) == SVt_PVCV);
12389 SAVEI32(PL_subline);
12390 save_item(PL_subname);
12391 SAVESPTR(PL_compcv);
12393 PL_compcv = (CV*)newSV_type(is_format ? SVt_PVFM : SVt_PVCV);
12394 CvFLAGS(PL_compcv) |= flags;
12396 PL_subline = CopLINE(PL_curcop);
12397 CvPADLIST(PL_compcv) = pad_new(padnew_SAVE|padnew_SAVESUB);
12398 CvOUTSIDE(PL_compcv) = (CV*)SvREFCNT_inc_simple(outsidecv);
12399 CvOUTSIDE_SEQ(PL_compcv) = PL_cop_seqmax;
12401 return oldsavestack_ix;
12405 #pragma segment Perl_yylex
12408 Perl_yywarn(pTHX_ const char *s)
12411 PL_in_eval |= EVAL_WARNONLY;
12413 PL_in_eval &= ~EVAL_WARNONLY;
12418 Perl_yyerror(pTHX_ const char *s)
12421 const char *where = NULL;
12422 const char *context = NULL;
12425 int yychar = PL_parser->yychar;
12427 if (!yychar || (yychar == ';' && !PL_rsfp))
12429 else if (PL_oldoldbufptr && PL_bufptr > PL_oldoldbufptr &&
12430 PL_bufptr - PL_oldoldbufptr < 200 && PL_oldoldbufptr != PL_oldbufptr &&
12431 PL_oldbufptr != PL_bufptr) {
12434 The code below is removed for NetWare because it abends/crashes on NetWare
12435 when the script has error such as not having the closing quotes like:
12436 if ($var eq "value)
12437 Checking of white spaces is anyway done in NetWare code.
12440 while (isSPACE(*PL_oldoldbufptr))
12443 context = PL_oldoldbufptr;
12444 contlen = PL_bufptr - PL_oldoldbufptr;
12446 else if (PL_oldbufptr && PL_bufptr > PL_oldbufptr &&
12447 PL_bufptr - PL_oldbufptr < 200 && PL_oldbufptr != PL_bufptr) {
12450 The code below is removed for NetWare because it abends/crashes on NetWare
12451 when the script has error such as not having the closing quotes like:
12452 if ($var eq "value)
12453 Checking of white spaces is anyway done in NetWare code.
12456 while (isSPACE(*PL_oldbufptr))
12459 context = PL_oldbufptr;
12460 contlen = PL_bufptr - PL_oldbufptr;
12462 else if (yychar > 255)
12463 where = "next token ???";
12464 else if (yychar == -2) { /* YYEMPTY */
12465 if (PL_lex_state == LEX_NORMAL ||
12466 (PL_lex_state == LEX_KNOWNEXT && PL_lex_defer == LEX_NORMAL))
12467 where = "at end of line";
12468 else if (PL_lex_inpat)
12469 where = "within pattern";
12471 where = "within string";
12474 SV * const where_sv = sv_2mortal(newSVpvs("next char "));
12476 Perl_sv_catpvf(aTHX_ where_sv, "^%c", toCTRL(yychar));
12477 else if (isPRINT_LC(yychar))
12478 Perl_sv_catpvf(aTHX_ where_sv, "%c", yychar);
12480 Perl_sv_catpvf(aTHX_ where_sv, "\\%03o", yychar & 255);
12481 where = SvPVX_const(where_sv);
12483 msg = sv_2mortal(newSVpv(s, 0));
12484 Perl_sv_catpvf(aTHX_ msg, " at %s line %"IVdf", ",
12485 OutCopFILE(PL_curcop), (IV)CopLINE(PL_curcop));
12487 Perl_sv_catpvf(aTHX_ msg, "near \"%.*s\"\n", contlen, context);
12489 Perl_sv_catpvf(aTHX_ msg, "%s\n", where);
12490 if (PL_multi_start < PL_multi_end && (U32)(CopLINE(PL_curcop) - PL_multi_end) <= 1) {
12491 Perl_sv_catpvf(aTHX_ msg,
12492 " (Might be a runaway multi-line %c%c string starting on line %"IVdf")\n",
12493 (int)PL_multi_open,(int)PL_multi_close,(IV)PL_multi_start);
12496 if (PL_in_eval & EVAL_WARNONLY && ckWARN_d(WARN_SYNTAX))
12497 Perl_warner(aTHX_ packWARN(WARN_SYNTAX), "%"SVf, SVfARG(msg));
12500 if (PL_error_count >= 10) {
12501 if (PL_in_eval && SvCUR(ERRSV))
12502 Perl_croak(aTHX_ "%"SVf"%s has too many errors.\n",
12503 SVfARG(ERRSV), OutCopFILE(PL_curcop));
12505 Perl_croak(aTHX_ "%s has too many errors.\n",
12506 OutCopFILE(PL_curcop));
12509 PL_in_my_stash = NULL;
12513 #pragma segment Main
12517 S_swallow_bom(pTHX_ U8 *s)
12520 const STRLEN slen = SvCUR(PL_linestr);
12523 if (s[1] == 0xFE) {
12524 /* UTF-16 little-endian? (or UTF32-LE?) */
12525 if (s[2] == 0 && s[3] == 0) /* UTF-32 little-endian */
12526 Perl_croak(aTHX_ "Unsupported script encoding UTF32-LE");
12527 #ifndef PERL_NO_UTF16_FILTER
12528 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF16-LE script encoding (BOM)\n");
12531 if (PL_bufend > (char*)s) {
12535 filter_add(utf16rev_textfilter, NULL);
12536 Newx(news, (PL_bufend - (char*)s) * 3 / 2 + 1, U8);
12537 utf16_to_utf8_reversed(s, news,
12538 PL_bufend - (char*)s - 1,
12540 sv_setpvn(PL_linestr, (const char*)news, newlen);
12542 s = (U8*)SvPVX(PL_linestr);
12543 Copy(news, s, newlen, U8);
12547 SvUTF8_on(PL_linestr);
12548 s = (U8*)SvPVX(PL_linestr);
12550 /* FIXME - is this a general bug fix? */
12553 PL_bufend = SvPVX(PL_linestr) + newlen;
12556 Perl_croak(aTHX_ "Unsupported script encoding UTF16-LE");
12561 if (s[1] == 0xFF) { /* UTF-16 big-endian? */
12562 #ifndef PERL_NO_UTF16_FILTER
12563 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16BE script encoding (BOM)\n");
12566 if (PL_bufend > (char *)s) {
12570 filter_add(utf16_textfilter, NULL);
12571 Newx(news, (PL_bufend - (char*)s) * 3 / 2 + 1, U8);
12572 utf16_to_utf8(s, news,
12573 PL_bufend - (char*)s,
12575 sv_setpvn(PL_linestr, (const char*)news, newlen);
12577 SvUTF8_on(PL_linestr);
12578 s = (U8*)SvPVX(PL_linestr);
12579 PL_bufend = SvPVX(PL_linestr) + newlen;
12582 Perl_croak(aTHX_ "Unsupported script encoding UTF16-BE");
12587 if (slen > 2 && s[1] == 0xBB && s[2] == 0xBF) {
12588 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-8 script encoding (BOM)\n");
12589 s += 3; /* UTF-8 */
12595 if (s[2] == 0xFE && s[3] == 0xFF) {
12596 /* UTF-32 big-endian */
12597 Perl_croak(aTHX_ "Unsupported script encoding UTF32-BE");
12600 else if (s[2] == 0 && s[3] != 0) {
12603 * are a good indicator of UTF-16BE. */
12604 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16BE script encoding (no BOM)\n");
12610 if (slen > 3 && s[1] == 0x73 && s[2] == 0x66 && s[3] == 0x73) {
12611 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-8 script encoding (BOM)\n");
12612 s += 4; /* UTF-8 */
12618 if (slen > 3 && s[1] == 0 && s[2] != 0 && s[3] == 0) {
12621 * are a good indicator of UTF-16LE. */
12622 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16LE script encoding (no BOM)\n");
12631 * Restore a source filter.
12635 restore_rsfp(pTHX_ void *f)
12638 PerlIO * const fp = (PerlIO*)f;
12640 if (PL_rsfp == PerlIO_stdin())
12641 PerlIO_clearerr(PL_rsfp);
12642 else if (PL_rsfp && (PL_rsfp != fp))
12643 PerlIO_close(PL_rsfp);
12647 #ifndef PERL_NO_UTF16_FILTER
12649 utf16_textfilter(pTHX_ int idx, SV *sv, int maxlen)
12652 const STRLEN old = SvCUR(sv);
12653 const I32 count = FILTER_READ(idx+1, sv, maxlen);
12654 DEBUG_P(PerlIO_printf(Perl_debug_log,
12655 "utf16_textfilter(%p): %d %d (%d)\n",
12656 FPTR2DPTR(void *, utf16_textfilter),
12657 idx, maxlen, (int) count));
12661 Newx(tmps, SvCUR(sv) * 3 / 2 + 1, U8);
12662 Copy(SvPVX_const(sv), tmps, old, char);
12663 utf16_to_utf8((U8*)SvPVX_const(sv) + old, tmps + old,
12664 SvCUR(sv) - old, &newlen);
12665 sv_usepvn(sv, (char*)tmps, (STRLEN)newlen + old);
12667 DEBUG_P({sv_dump(sv);});
12672 utf16rev_textfilter(pTHX_ int idx, SV *sv, int maxlen)
12675 const STRLEN old = SvCUR(sv);
12676 const I32 count = FILTER_READ(idx+1, sv, maxlen);
12677 DEBUG_P(PerlIO_printf(Perl_debug_log,
12678 "utf16rev_textfilter(%p): %d %d (%d)\n",
12679 FPTR2DPTR(void *, utf16rev_textfilter),
12680 idx, maxlen, (int) count));
12684 Newx(tmps, SvCUR(sv) * 3 / 2 + 1, U8);
12685 Copy(SvPVX_const(sv), tmps, old, char);
12686 utf16_to_utf8((U8*)SvPVX_const(sv) + old, tmps + old,
12687 SvCUR(sv) - old, &newlen);
12688 sv_usepvn(sv, (char*)tmps, (STRLEN)newlen + old);
12690 DEBUG_P({ sv_dump(sv); });
12696 Returns a pointer to the next character after the parsed
12697 vstring, as well as updating the passed in sv.
12699 Function must be called like
12702 s = scan_vstring(s,sv);
12704 The sv should already be large enough to store the vstring
12705 passed in, for performance reasons.
12710 Perl_scan_vstring(pTHX_ const char *s, SV *sv)
12713 const char *pos = s;
12714 const char *start = s;
12715 if (*pos == 'v') pos++; /* get past 'v' */
12716 while (pos < PL_bufend && (isDIGIT(*pos) || *pos == '_'))
12718 if ( *pos != '.') {
12719 /* this may not be a v-string if followed by => */
12720 const char *next = pos;
12721 while (next < PL_bufend && isSPACE(*next))
12723 if ((PL_bufend - next) >= 2 && *next == '=' && next[1] == '>' ) {
12724 /* return string not v-string */
12725 sv_setpvn(sv,(char *)s,pos-s);
12726 return (char *)pos;
12730 if (!isALPHA(*pos)) {
12731 U8 tmpbuf[UTF8_MAXBYTES+1];
12734 s++; /* get past 'v' */
12736 sv_setpvn(sv, "", 0);
12739 /* this is atoi() that tolerates underscores */
12742 const char *end = pos;
12744 while (--end >= s) {
12746 const UV orev = rev;
12747 rev += (*end - '0') * mult;
12749 if (orev > rev && ckWARN_d(WARN_OVERFLOW))
12750 Perl_warner(aTHX_ packWARN(WARN_OVERFLOW),
12751 "Integer overflow in decimal number");
12755 if (rev > 0x7FFFFFFF)
12756 Perl_croak(aTHX_ "In EBCDIC the v-string components cannot exceed 2147483647");
12758 /* Append native character for the rev point */
12759 tmpend = uvchr_to_utf8(tmpbuf, rev);
12760 sv_catpvn(sv, (const char*)tmpbuf, tmpend - tmpbuf);
12761 if (!UNI_IS_INVARIANT(NATIVE_TO_UNI(rev)))
12763 if (pos + 1 < PL_bufend && *pos == '.' && isDIGIT(pos[1]))
12769 while (pos < PL_bufend && (isDIGIT(*pos) || *pos == '_'))
12773 sv_magic(sv,NULL,PERL_MAGIC_vstring,(const char*)start, pos-start);
12781 * c-indentation-style: bsd
12782 * c-basic-offset: 4
12783 * indent-tabs-mode: t
12786 * ex: set ts=8 sts=4 sw=4 noet: