3 * Copyright (C) 1991, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999,
4 * 2000, 2001, 2002, 2003, 2004, 2005, 2006, 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 static const char ident_too_long[] = "Identifier too long";
29 static const char commaless_variable_list[] = "comma-less variable list";
31 static void restore_rsfp(pTHX_ void *f);
32 #ifndef PERL_NO_UTF16_FILTER
33 static I32 utf16_textfilter(pTHX_ int idx, SV *sv, int maxlen);
34 static I32 utf16rev_textfilter(pTHX_ int idx, SV *sv, int maxlen);
38 # define CURMAD(slot,sv) if (PL_madskills) { curmad(slot,sv); sv = 0; }
39 # define NEXTVAL_NEXTTOKE PL_nexttoke[PL_curforce].next_val
41 # define CURMAD(slot,sv)
42 # define NEXTVAL_NEXTTOKE PL_nextval[PL_nexttoke]
45 #define XFAKEBRACK 128
48 #ifdef USE_UTF8_SCRIPTS
49 # define UTF (!IN_BYTES)
51 # define UTF ((PL_linestr && DO_UTF8(PL_linestr)) || (PL_hints & HINT_UTF8))
54 /* In variables named $^X, these are the legal values for X.
55 * 1999-02-27 mjd-perl-patch@plover.com */
56 #define isCONTROLVAR(x) (isUPPER(x) || strchr("[\\]^_?", (x)))
58 /* On MacOS, respect nonbreaking spaces */
59 #ifdef MACOS_TRADITIONAL
60 #define SPACE_OR_TAB(c) ((c)==' '||(c)=='\312'||(c)=='\t')
62 #define SPACE_OR_TAB(c) ((c)==' '||(c)=='\t')
65 /* LEX_* are values for PL_lex_state, the state of the lexer.
66 * They are arranged oddly so that the guard on the switch statement
67 * can get by with a single comparison (if the compiler is smart enough).
70 /* #define LEX_NOTPARSING 11 is done in perl.h. */
72 #define LEX_NORMAL 10 /* normal code (ie not within "...") */
73 #define LEX_INTERPNORMAL 9 /* code within a string, eg "$foo[$x+1]" */
74 #define LEX_INTERPCASEMOD 8 /* expecting a \U, \Q or \E etc */
75 #define LEX_INTERPPUSH 7 /* starting a new sublex parse level */
76 #define LEX_INTERPSTART 6 /* expecting the start of a $var */
78 /* at end of code, eg "$x" followed by: */
79 #define LEX_INTERPEND 5 /* ... eg not one of [, { or -> */
80 #define LEX_INTERPENDMAYBE 4 /* ... eg one of [, { or -> */
82 #define LEX_INTERPCONCAT 3 /* expecting anything, eg at start of
83 string or after \E, $foo, etc */
84 #define LEX_INTERPCONST 2 /* NOT USED */
85 #define LEX_FORMLINE 1 /* expecting a format line */
86 #define LEX_KNOWNEXT 0 /* next token known; just return it */
90 static const char* const lex_state_names[] = {
109 #include "keywords.h"
111 /* CLINE is a macro that ensures PL_copline has a sane value */
116 #define CLINE (PL_copline = (CopLINE(PL_curcop) < PL_copline ? CopLINE(PL_curcop) : PL_copline))
119 # define SKIPSPACE0(s) skipspace0(s)
120 # define SKIPSPACE1(s) skipspace1(s)
121 # define SKIPSPACE2(s,tsv) skipspace2(s,&tsv)
122 # define PEEKSPACE(s) skipspace2(s,0)
124 # define SKIPSPACE0(s) skipspace(s)
125 # define SKIPSPACE1(s) skipspace(s)
126 # define SKIPSPACE2(s,tsv) skipspace(s)
127 # define PEEKSPACE(s) skipspace(s)
131 * Convenience functions to return different tokens and prime the
132 * lexer for the next token. They all take an argument.
134 * TOKEN : generic token (used for '(', DOLSHARP, etc)
135 * OPERATOR : generic operator
136 * AOPERATOR : assignment operator
137 * PREBLOCK : beginning the block after an if, while, foreach, ...
138 * PRETERMBLOCK : beginning a non-code-defining {} block (eg, hash ref)
139 * PREREF : *EXPR where EXPR is not a simple identifier
140 * TERM : expression term
141 * LOOPX : loop exiting command (goto, last, dump, etc)
142 * FTST : file test operator
143 * FUN0 : zero-argument function
144 * FUN1 : not used, except for not, which isn't a UNIOP
145 * BOop : bitwise or or xor
147 * SHop : shift operator
148 * PWop : power operator
149 * PMop : pattern-matching operator
150 * Aop : addition-level operator
151 * Mop : multiplication-level operator
152 * Eop : equality-testing operator
153 * Rop : relational operator <= != gt
155 * Also see LOP and lop() below.
158 #ifdef DEBUGGING /* Serve -DT. */
159 # define REPORT(retval) tokereport((I32)retval)
161 # define REPORT(retval) (retval)
164 #define TOKEN(retval) return ( PL_bufptr = s, REPORT(retval))
165 #define OPERATOR(retval) return (PL_expect = XTERM, PL_bufptr = s, REPORT(retval))
166 #define AOPERATOR(retval) return ao((PL_expect = XTERM, PL_bufptr = s, REPORT(retval)))
167 #define PREBLOCK(retval) return (PL_expect = XBLOCK,PL_bufptr = s, REPORT(retval))
168 #define PRETERMBLOCK(retval) return (PL_expect = XTERMBLOCK,PL_bufptr = s, REPORT(retval))
169 #define PREREF(retval) return (PL_expect = XREF,PL_bufptr = s, REPORT(retval))
170 #define TERM(retval) return (CLINE, PL_expect = XOPERATOR, PL_bufptr = s, REPORT(retval))
171 #define LOOPX(f) return (yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)LOOPEX))
172 #define FTST(f) return (yylval.ival=f, PL_expect=XTERMORDORDOR, PL_bufptr=s, REPORT((int)UNIOP))
173 #define FUN0(f) return (yylval.ival=f, PL_expect=XOPERATOR, PL_bufptr=s, REPORT((int)FUNC0))
174 #define FUN1(f) return (yylval.ival=f, PL_expect=XOPERATOR, PL_bufptr=s, REPORT((int)FUNC1))
175 #define BOop(f) return ao((yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)BITOROP)))
176 #define BAop(f) return ao((yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)BITANDOP)))
177 #define SHop(f) return ao((yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)SHIFTOP)))
178 #define PWop(f) return ao((yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)POWOP)))
179 #define PMop(f) return(yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)MATCHOP))
180 #define Aop(f) return ao((yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)ADDOP)))
181 #define Mop(f) return ao((yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)MULOP)))
182 #define Eop(f) return (yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)EQOP))
183 #define Rop(f) return (yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)RELOP))
185 /* This bit of chicanery makes a unary function followed by
186 * a parenthesis into a function with one argument, highest precedence.
187 * The UNIDOR macro is for unary functions that can be followed by the //
188 * operator (such as C<shift // 0>).
190 #define UNI2(f,x) { \
194 PL_last_uni = PL_oldbufptr; \
195 PL_last_lop_op = f; \
197 return REPORT( (int)FUNC1 ); \
199 return REPORT( *s=='(' ? (int)FUNC1 : (int)UNIOP ); \
201 #define UNI(f) UNI2(f,XTERM)
202 #define UNIDOR(f) UNI2(f,XTERMORDORDOR)
204 #define UNIBRACK(f) { \
207 PL_last_uni = PL_oldbufptr; \
209 return REPORT( (int)FUNC1 ); \
211 return REPORT( (*s == '(') ? (int)FUNC1 : (int)UNIOP ); \
214 /* grandfather return to old style */
215 #define OLDLOP(f) return(yylval.ival=f,PL_expect = XTERM,PL_bufptr = s,(int)LSTOP)
219 /* how to interpret the yylval associated with the token */
223 TOKENTYPE_OPNUM, /* yylval.ival contains an opcode number */
229 static struct debug_tokens {
231 enum token_type type;
233 } const debug_tokens[] =
235 { ADDOP, TOKENTYPE_OPNUM, "ADDOP" },
236 { ANDAND, TOKENTYPE_NONE, "ANDAND" },
237 { ANDOP, TOKENTYPE_NONE, "ANDOP" },
238 { ANONSUB, TOKENTYPE_IVAL, "ANONSUB" },
239 { ARROW, TOKENTYPE_NONE, "ARROW" },
240 { ASSIGNOP, TOKENTYPE_OPNUM, "ASSIGNOP" },
241 { BITANDOP, TOKENTYPE_OPNUM, "BITANDOP" },
242 { BITOROP, TOKENTYPE_OPNUM, "BITOROP" },
243 { COLONATTR, TOKENTYPE_NONE, "COLONATTR" },
244 { CONTINUE, TOKENTYPE_NONE, "CONTINUE" },
245 { DEFAULT, TOKENTYPE_NONE, "DEFAULT" },
246 { DO, TOKENTYPE_NONE, "DO" },
247 { DOLSHARP, TOKENTYPE_NONE, "DOLSHARP" },
248 { DORDOR, TOKENTYPE_NONE, "DORDOR" },
249 { DOROP, TOKENTYPE_OPNUM, "DOROP" },
250 { DOTDOT, TOKENTYPE_IVAL, "DOTDOT" },
251 { ELSE, TOKENTYPE_NONE, "ELSE" },
252 { ELSIF, TOKENTYPE_IVAL, "ELSIF" },
253 { EQOP, TOKENTYPE_OPNUM, "EQOP" },
254 { FOR, TOKENTYPE_IVAL, "FOR" },
255 { FORMAT, TOKENTYPE_NONE, "FORMAT" },
256 { FUNC, TOKENTYPE_OPNUM, "FUNC" },
257 { FUNC0, TOKENTYPE_OPNUM, "FUNC0" },
258 { FUNC0SUB, TOKENTYPE_OPVAL, "FUNC0SUB" },
259 { FUNC1, TOKENTYPE_OPNUM, "FUNC1" },
260 { FUNCMETH, TOKENTYPE_OPVAL, "FUNCMETH" },
261 { GIVEN, TOKENTYPE_IVAL, "GIVEN" },
262 { HASHBRACK, TOKENTYPE_NONE, "HASHBRACK" },
263 { IF, TOKENTYPE_IVAL, "IF" },
264 { LABEL, TOKENTYPE_PVAL, "LABEL" },
265 { LOCAL, TOKENTYPE_IVAL, "LOCAL" },
266 { LOOPEX, TOKENTYPE_OPNUM, "LOOPEX" },
267 { LSTOP, TOKENTYPE_OPNUM, "LSTOP" },
268 { LSTOPSUB, TOKENTYPE_OPVAL, "LSTOPSUB" },
269 { MATCHOP, TOKENTYPE_OPNUM, "MATCHOP" },
270 { METHOD, TOKENTYPE_OPVAL, "METHOD" },
271 { MULOP, TOKENTYPE_OPNUM, "MULOP" },
272 { MY, TOKENTYPE_IVAL, "MY" },
273 { MYSUB, TOKENTYPE_NONE, "MYSUB" },
274 { NOAMP, TOKENTYPE_NONE, "NOAMP" },
275 { NOTOP, TOKENTYPE_NONE, "NOTOP" },
276 { OROP, TOKENTYPE_IVAL, "OROP" },
277 { OROR, TOKENTYPE_NONE, "OROR" },
278 { PACKAGE, TOKENTYPE_NONE, "PACKAGE" },
279 { PMFUNC, TOKENTYPE_OPVAL, "PMFUNC" },
280 { POSTDEC, TOKENTYPE_NONE, "POSTDEC" },
281 { POSTINC, TOKENTYPE_NONE, "POSTINC" },
282 { POWOP, TOKENTYPE_OPNUM, "POWOP" },
283 { PREDEC, TOKENTYPE_NONE, "PREDEC" },
284 { PREINC, TOKENTYPE_NONE, "PREINC" },
285 { PRIVATEREF, TOKENTYPE_OPVAL, "PRIVATEREF" },
286 { REFGEN, TOKENTYPE_NONE, "REFGEN" },
287 { RELOP, TOKENTYPE_OPNUM, "RELOP" },
288 { SHIFTOP, TOKENTYPE_OPNUM, "SHIFTOP" },
289 { SUB, TOKENTYPE_NONE, "SUB" },
290 { THING, TOKENTYPE_OPVAL, "THING" },
291 { UMINUS, TOKENTYPE_NONE, "UMINUS" },
292 { UNIOP, TOKENTYPE_OPNUM, "UNIOP" },
293 { UNIOPSUB, TOKENTYPE_OPVAL, "UNIOPSUB" },
294 { UNLESS, TOKENTYPE_IVAL, "UNLESS" },
295 { UNTIL, TOKENTYPE_IVAL, "UNTIL" },
296 { USE, TOKENTYPE_IVAL, "USE" },
297 { WHEN, TOKENTYPE_IVAL, "WHEN" },
298 { WHILE, TOKENTYPE_IVAL, "WHILE" },
299 { WORD, TOKENTYPE_OPVAL, "WORD" },
300 { 0, TOKENTYPE_NONE, NULL }
303 /* dump the returned token in rv, plus any optional arg in yylval */
306 S_tokereport(pTHX_ I32 rv)
310 const char *name = NULL;
311 enum token_type type = TOKENTYPE_NONE;
312 const struct debug_tokens *p;
313 SV* const report = newSVpvs("<== ");
315 for (p = debug_tokens; p->token; p++) {
316 if (p->token == (int)rv) {
323 Perl_sv_catpv(aTHX_ report, name);
324 else if ((char)rv > ' ' && (char)rv < '~')
325 Perl_sv_catpvf(aTHX_ report, "'%c'", (char)rv);
327 sv_catpvs(report, "EOF");
329 Perl_sv_catpvf(aTHX_ report, "?? %"IVdf, (IV)rv);
332 case TOKENTYPE_GVVAL: /* doesn't appear to be used */
335 Perl_sv_catpvf(aTHX_ report, "(ival=%"IVdf")", (IV)yylval.ival);
337 case TOKENTYPE_OPNUM:
338 Perl_sv_catpvf(aTHX_ report, "(ival=op_%s)",
339 PL_op_name[yylval.ival]);
342 Perl_sv_catpvf(aTHX_ report, "(pval=\"%s\")", yylval.pval);
344 case TOKENTYPE_OPVAL:
346 Perl_sv_catpvf(aTHX_ report, "(opval=op_%s)",
347 PL_op_name[yylval.opval->op_type]);
348 if (yylval.opval->op_type == OP_CONST) {
349 Perl_sv_catpvf(aTHX_ report, " %s",
350 SvPEEK(cSVOPx_sv(yylval.opval)));
355 sv_catpvs(report, "(opval=null)");
358 PerlIO_printf(Perl_debug_log, "### %s\n\n", SvPV_nolen_const(report));
364 /* print the buffer with suitable escapes */
367 S_printbuf(pTHX_ const char* fmt, const char* s)
369 SV* const tmp = newSVpvs("");
370 PerlIO_printf(Perl_debug_log, fmt, pv_display(tmp, s, strlen(s), 0, 60));
379 * This subroutine detects &&=, ||=, and //= and turns an ANDAND, OROR or DORDOR
380 * into an OP_ANDASSIGN, OP_ORASSIGN, or OP_DORASSIGN
384 S_ao(pTHX_ int toketype)
387 if (*PL_bufptr == '=') {
389 if (toketype == ANDAND)
390 yylval.ival = OP_ANDASSIGN;
391 else if (toketype == OROR)
392 yylval.ival = OP_ORASSIGN;
393 else if (toketype == DORDOR)
394 yylval.ival = OP_DORASSIGN;
402 * When Perl expects an operator and finds something else, no_op
403 * prints the warning. It always prints "<something> found where
404 * operator expected. It prints "Missing semicolon on previous line?"
405 * if the surprise occurs at the start of the line. "do you need to
406 * predeclare ..." is printed out for code like "sub bar; foo bar $x"
407 * where the compiler doesn't know if foo is a method call or a function.
408 * It prints "Missing operator before end of line" if there's nothing
409 * after the missing operator, or "... before <...>" if there is something
410 * after the missing operator.
414 S_no_op(pTHX_ const char *what, char *s)
417 char * const oldbp = PL_bufptr;
418 const bool is_first = (PL_oldbufptr == PL_linestart);
424 yywarn(Perl_form(aTHX_ "%s found where operator expected", what));
425 if (ckWARN_d(WARN_SYNTAX)) {
427 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
428 "\t(Missing semicolon on previous line?)\n");
429 else if (PL_oldoldbufptr && isIDFIRST_lazy_if(PL_oldoldbufptr,UTF)) {
431 for (t = PL_oldoldbufptr; (isALNUM_lazy_if(t,UTF) || *t == ':'); t++)
433 if (t < PL_bufptr && isSPACE(*t))
434 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
435 "\t(Do you need to predeclare %.*s?)\n",
436 (int)(t - PL_oldoldbufptr), PL_oldoldbufptr);
440 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
441 "\t(Missing operator before %.*s?)\n", (int)(s - oldbp), oldbp);
449 * Complain about missing quote/regexp/heredoc terminator.
450 * If it's called with NULL then it cauterizes the line buffer.
451 * If we're in a delimited string and the delimiter is a control
452 * character, it's reformatted into a two-char sequence like ^C.
457 S_missingterm(pTHX_ char *s)
463 char * const nl = strrchr(s,'\n');
469 iscntrl(PL_multi_close)
471 PL_multi_close < 32 || PL_multi_close == 127
475 tmpbuf[1] = (char)toCTRL(PL_multi_close);
480 *tmpbuf = (char)PL_multi_close;
484 q = strchr(s,'"') ? '\'' : '"';
485 Perl_croak(aTHX_ "Can't find string terminator %c%s%c anywhere before EOF",q,s,q);
488 #define FEATURE_IS_ENABLED(name) \
489 ((0 != (PL_hints & HINT_LOCALIZE_HH)) \
490 && S_feature_is_enabled(aTHX_ STR_WITH_LEN(name)))
492 * S_feature_is_enabled
493 * Check whether the named feature is enabled.
496 S_feature_is_enabled(pTHX_ const char *name, STRLEN namelen)
499 HV * const hinthv = GvHV(PL_hintgv);
500 char he_name[32] = "feature_";
501 (void) my_strlcpy(&he_name[8], name, 24);
503 return (hinthv && hv_exists(hinthv, he_name, 8 + namelen));
511 Perl_deprecate(pTHX_ const char *s)
513 if (ckWARN(WARN_DEPRECATED))
514 Perl_warner(aTHX_ packWARN(WARN_DEPRECATED), "Use of %s is deprecated", s);
518 Perl_deprecate_old(pTHX_ const char *s)
520 /* This function should NOT be called for any new deprecated warnings */
521 /* Use Perl_deprecate instead */
523 /* It is here to maintain backward compatibility with the pre-5.8 */
524 /* warnings category hierarchy. The "deprecated" category used to */
525 /* live under the "syntax" category. It is now a top-level category */
526 /* in its own right. */
528 if (ckWARN2(WARN_DEPRECATED, WARN_SYNTAX))
529 Perl_warner(aTHX_ packWARN2(WARN_DEPRECATED, WARN_SYNTAX),
530 "Use of %s is deprecated", s);
534 * experimental text filters for win32 carriage-returns, utf16-to-utf8 and
535 * utf16-to-utf8-reversed.
538 #ifdef PERL_CR_FILTER
542 register const char *s = SvPVX_const(sv);
543 register const char * const e = s + SvCUR(sv);
544 /* outer loop optimized to do nothing if there are no CR-LFs */
546 if (*s++ == '\r' && *s == '\n') {
547 /* hit a CR-LF, need to copy the rest */
548 register char *d = s - 1;
551 if (*s == '\r' && s[1] == '\n')
562 S_cr_textfilter(pTHX_ int idx, SV *sv, int maxlen)
564 const I32 count = FILTER_READ(idx+1, sv, maxlen);
565 if (count > 0 && !maxlen)
573 * Initialize variables. Uses the Perl save_stack to save its state (for
574 * recursive calls to the parser).
578 Perl_lex_start(pTHX_ SV *line)
584 SAVEI32(PL_lex_dojoin);
585 SAVEI32(PL_lex_brackets);
586 SAVEI32(PL_lex_casemods);
587 SAVEI32(PL_lex_starts);
588 SAVEI32(PL_lex_state);
589 SAVEVPTR(PL_lex_inpat);
590 SAVEI32(PL_lex_inwhat);
592 if (PL_lex_state == LEX_KNOWNEXT) {
593 I32 toke = PL_lasttoke;
594 while (--toke >= 0) {
595 SAVEI32(PL_nexttoke[toke].next_type);
596 SAVEVPTR(PL_nexttoke[toke].next_val);
598 SAVEVPTR(PL_nexttoke[toke].next_mad);
600 SAVEI32(PL_lasttoke);
602 SAVESPTR(PL_endwhite);
603 SAVESPTR(PL_thistoken);
604 SAVESPTR(PL_thiswhite);
605 SAVESPTR(PL_nextwhite);
606 SAVESPTR(PL_thisopen);
607 SAVESPTR(PL_thisclose);
608 SAVESPTR(PL_thisstuff);
609 SAVEVPTR(PL_thismad);
610 SAVEI32(PL_realtokenstart);
611 SAVEI32(PL_faketokens);
612 SAVESPTR(PL_skipwhite);
613 SAVEI32(PL_curforce);
615 if (PL_lex_state == LEX_KNOWNEXT) {
616 I32 toke = PL_nexttoke;
617 while (--toke >= 0) {
618 SAVEI32(PL_nexttype[toke]);
619 SAVEVPTR(PL_nextval[toke]);
621 SAVEI32(PL_nexttoke);
624 SAVECOPLINE(PL_curcop);
627 SAVEPPTR(PL_oldbufptr);
628 SAVEPPTR(PL_oldoldbufptr);
629 SAVEPPTR(PL_last_lop);
630 SAVEPPTR(PL_last_uni);
631 SAVEPPTR(PL_linestart);
632 SAVESPTR(PL_linestr);
633 SAVEGENERICPV(PL_lex_brackstack);
634 SAVEGENERICPV(PL_lex_casestack);
635 SAVEDESTRUCTOR_X(restore_rsfp, PL_rsfp);
636 SAVESPTR(PL_lex_stuff);
637 SAVEI32(PL_lex_defer);
638 SAVEI32(PL_sublex_info.sub_inwhat);
639 SAVEI32(PL_sublex_info.super_state);
640 SAVEVPTR(PL_sublex_info.sub_op);
641 SAVEPPTR(PL_sublex_info.super_bufptr);
642 SAVEPPTR(PL_sublex_info.super_bufend);
643 SAVESPTR(PL_lex_repl);
645 SAVEINT(PL_lex_expect);
646 SAVEI32(PL_lex_formbrack);
648 SAVEI32(PL_multi_close);
649 SAVEI32(PL_multi_open);
650 SAVEI32(PL_multi_start);
651 SAVEI8(PL_pending_ident);
652 SAVEBOOL(PL_preambled);
654 PL_lex_state = LEX_NORMAL;
658 Newx(PL_lex_brackstack, 120, char);
659 Newx(PL_lex_casestack, 12, char);
661 *PL_lex_casestack = '\0';
672 PL_realtokenstart = 0;
684 PL_sublex_info.sub_inwhat = 0;
685 PL_sublex_info.super_state = 0;
686 PL_sublex_info.sub_op = NULL;
687 PL_sublex_info.super_bufptr = NULL;
688 PL_sublex_info.super_bufend = NULL;
690 PL_lex_formbrack = 0;
695 PL_pending_ident = '\0';
696 PL_preambled = FALSE;
699 s = SvPV_const(line, len);
704 PL_linestr = newSVpvs("\n;");
705 } else if (SvREADONLY(line) || s[len-1] != ';') {
706 PL_linestr = newSVsv(line);
708 sv_catpvs(PL_linestr, "\n;");
711 SvREFCNT_inc_simple_void_NN(line);
714 /* PL_linestr needs to survive until end of scope, not just the next
715 FREETMPS. See changes 17505 and 17546 which fixed the symptoms only. */
716 SAVEFREESV(PL_linestr);
717 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = SvPVX(PL_linestr);
718 PL_bufend = PL_bufptr + SvCUR(PL_linestr);
719 PL_last_lop = PL_last_uni = NULL;
725 * Finalizer for lexing operations. Must be called when the parser is
726 * done with the lexer.
733 PL_doextract = FALSE;
738 * This subroutine has nothing to do with tilting, whether at windmills
739 * or pinball tables. Its name is short for "increment line". It
740 * increments the current line number in CopLINE(PL_curcop) and checks
741 * to see whether the line starts with a comment of the form
742 * # line 500 "foo.pm"
743 * If so, it sets the current line number and file to the values in the comment.
747 S_incline(pTHX_ char *s)
755 CopLINE_inc(PL_curcop);
758 while (SPACE_OR_TAB(*s))
760 if (strnEQ(s, "line", 4))
764 if (SPACE_OR_TAB(*s))
768 while (SPACE_OR_TAB(*s))
776 while (SPACE_OR_TAB(*s))
778 if (*s == '"' && (t = strchr(s+1, '"'))) {
788 while (SPACE_OR_TAB(*e) || *e == '\r' || *e == '\f')
790 if (*e != '\n' && *e != '\0')
791 return; /* false alarm */
797 const char * const cf = CopFILE(PL_curcop);
798 STRLEN tmplen = cf ? strlen(cf) : 0;
799 if (tmplen > 7 && strnEQ(cf, "(eval ", 6)) {
800 /* must copy *{"::_<(eval N)[oldfilename:L]"}
801 * to *{"::_<newfilename"} */
802 char smallbuf[256], smallbuf2[256];
803 char *tmpbuf, *tmpbuf2;
805 STRLEN tmplen2 = strlen(s);
806 if (tmplen + 3 < sizeof smallbuf)
809 Newx(tmpbuf, tmplen + 3, char);
810 if (tmplen2 + 3 < sizeof smallbuf2)
813 Newx(tmpbuf2, tmplen2 + 3, char);
814 tmpbuf[0] = tmpbuf2[0] = '_';
815 tmpbuf[1] = tmpbuf2[1] = '<';
816 memcpy(tmpbuf + 2, cf, ++tmplen);
817 memcpy(tmpbuf2 + 2, s, ++tmplen2);
819 gvp = (GV**)hv_fetch(PL_defstash, tmpbuf, tmplen, FALSE);
821 gv2 = *(GV**)hv_fetch(PL_defstash, tmpbuf2, tmplen2, TRUE);
823 gv_init(gv2, PL_defstash, tmpbuf2, tmplen2, FALSE);
824 /* adjust ${"::_<newfilename"} to store the new file name */
825 GvSV(gv2) = newSVpvn(tmpbuf2 + 2, tmplen2 - 2);
826 GvHV(gv2) = (HV*)SvREFCNT_inc(GvHV(*gvp));
827 GvAV(gv2) = (AV*)SvREFCNT_inc(GvAV(*gvp));
830 if (tmpbuf != smallbuf) Safefree(tmpbuf);
831 if (tmpbuf2 != smallbuf2) Safefree(tmpbuf2);
834 CopFILE_free(PL_curcop);
835 CopFILE_set(PL_curcop, s);
838 CopLINE_set(PL_curcop, atoi(n)-1);
842 /* skip space before PL_thistoken */
845 S_skipspace0(pTHX_ register char *s)
852 PL_thiswhite = newSVpvs("");
853 sv_catsv(PL_thiswhite, PL_skipwhite);
854 sv_free(PL_skipwhite);
857 PL_realtokenstart = s - SvPVX(PL_linestr);
861 /* skip space after PL_thistoken */
864 S_skipspace1(pTHX_ register char *s)
866 const char *start = s;
867 I32 startoff = start - SvPVX(PL_linestr);
872 start = SvPVX(PL_linestr) + startoff;
873 if (!PL_thistoken && PL_realtokenstart >= 0) {
874 const char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
875 PL_thistoken = newSVpvn(tstart, start - tstart);
877 PL_realtokenstart = -1;
880 PL_nextwhite = newSVpvs("");
881 sv_catsv(PL_nextwhite, PL_skipwhite);
882 sv_free(PL_skipwhite);
889 S_skipspace2(pTHX_ register char *s, SV **svp)
892 const I32 bufptroff = PL_bufptr - SvPVX(PL_linestr);
893 const I32 startoff = s - SvPVX(PL_linestr);
896 PL_bufptr = SvPVX(PL_linestr) + bufptroff;
897 if (!PL_madskills || !svp)
899 start = SvPVX(PL_linestr) + startoff;
900 if (!PL_thistoken && PL_realtokenstart >= 0) {
901 char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
902 PL_thistoken = newSVpvn(tstart, start - tstart);
903 PL_realtokenstart = -1;
908 sv_setsv(*svp, PL_skipwhite);
909 sv_free(PL_skipwhite);
918 S_update_debugger_info_pv(pTHX_ const char *buf, STRLEN len)
920 AV *av = CopFILEAVx(PL_curcop);
922 SV * const sv = newSV(0);
923 sv_upgrade(sv, SVt_PVMG);
924 sv_setpvn(sv, buf, len);
927 av_store(av, (I32)CopLINE(PL_curcop), sv);
932 S_update_debugger_info_sv(pTHX_ SV *orig_sv)
934 AV *av = CopFILEAVx(PL_curcop);
936 SV * const sv = newSV(0);
937 sv_upgrade(sv, SVt_PVMG);
938 sv_setsv(sv, orig_sv);
941 av_store(av, (I32)CopLINE(PL_curcop), sv);
947 * Called to gobble the appropriate amount and type of whitespace.
948 * Skips comments as well.
952 S_skipspace(pTHX_ register char *s)
957 int startoff = s - SvPVX(PL_linestr);
960 sv_free(PL_skipwhite);
965 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
966 while (s < PL_bufend && SPACE_OR_TAB(*s))
976 SSize_t oldprevlen, oldoldprevlen;
977 SSize_t oldloplen = 0, oldunilen = 0;
978 while (s < PL_bufend && isSPACE(*s)) {
979 if (*s++ == '\n' && PL_in_eval && !PL_rsfp)
984 if (s < PL_bufend && *s == '#') {
985 while (s < PL_bufend && *s != '\n')
989 if (PL_in_eval && !PL_rsfp) {
996 /* only continue to recharge the buffer if we're at the end
997 * of the buffer, we're not reading from a source filter, and
998 * we're in normal lexing mode
1000 if (s < PL_bufend || !PL_rsfp || PL_sublex_info.sub_inwhat ||
1001 PL_lex_state == LEX_FORMLINE)
1008 /* try to recharge the buffer */
1010 curoff = s - SvPVX(PL_linestr);
1013 if ((s = filter_gets(PL_linestr, PL_rsfp,
1014 (prevlen = SvCUR(PL_linestr)))) == NULL)
1017 if (PL_madskills && curoff != startoff) {
1019 PL_skipwhite = newSVpvs("");
1020 sv_catpvn(PL_skipwhite, SvPVX(PL_linestr) + startoff,
1024 /* mustn't throw out old stuff yet if madpropping */
1025 SvCUR(PL_linestr) = curoff;
1026 s = SvPVX(PL_linestr) + curoff;
1028 if (curoff && s[-1] == '\n')
1032 /* end of file. Add on the -p or -n magic */
1033 /* XXX these shouldn't really be added here, can't set PL_faketokens */
1036 sv_catpv(PL_linestr,
1037 ";}continue{print or die qq(-p destination: $!\\n);}");
1039 sv_setpv(PL_linestr,
1040 ";}continue{print or die qq(-p destination: $!\\n);}");
1042 PL_minus_n = PL_minus_p = 0;
1044 else if (PL_minus_n) {
1046 sv_catpvn(PL_linestr, ";}", 2);
1048 sv_setpvn(PL_linestr, ";}", 2);
1054 sv_catpvn(PL_linestr,";", 1);
1056 sv_setpvn(PL_linestr,";", 1);
1059 /* reset variables for next time we lex */
1060 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = s = PL_linestart
1066 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
1067 PL_last_lop = PL_last_uni = NULL;
1069 /* Close the filehandle. Could be from -P preprocessor,
1070 * STDIN, or a regular file. If we were reading code from
1071 * STDIN (because the commandline held no -e or filename)
1072 * then we don't close it, we reset it so the code can
1073 * read from STDIN too.
1076 if (PL_preprocess && !PL_in_eval)
1077 (void)PerlProc_pclose(PL_rsfp);
1078 else if ((PerlIO*)PL_rsfp == PerlIO_stdin())
1079 PerlIO_clearerr(PL_rsfp);
1081 (void)PerlIO_close(PL_rsfp);
1086 /* not at end of file, so we only read another line */
1087 /* make corresponding updates to old pointers, for yyerror() */
1088 oldprevlen = PL_oldbufptr - PL_bufend;
1089 oldoldprevlen = PL_oldoldbufptr - PL_bufend;
1091 oldunilen = PL_last_uni - PL_bufend;
1093 oldloplen = PL_last_lop - PL_bufend;
1094 PL_linestart = PL_bufptr = s + prevlen;
1095 PL_bufend = s + SvCUR(PL_linestr);
1097 PL_oldbufptr = s + oldprevlen;
1098 PL_oldoldbufptr = s + oldoldprevlen;
1100 PL_last_uni = s + oldunilen;
1102 PL_last_lop = s + oldloplen;
1105 /* debugger active and we're not compiling the debugger code,
1106 * so store the line into the debugger's array of lines
1108 if (PERLDB_LINE && PL_curstash != PL_debstash)
1109 update_debugger_info_pv(PL_bufptr, PL_bufend - PL_bufptr);
1116 PL_skipwhite = newSVpvs("");
1117 curoff = s - SvPVX(PL_linestr);
1118 if (curoff - startoff)
1119 sv_catpvn(PL_skipwhite, SvPVX(PL_linestr) + startoff,
1128 * Check the unary operators to ensure there's no ambiguity in how they're
1129 * used. An ambiguous piece of code would be:
1131 * This doesn't mean rand() + 5. Because rand() is a unary operator,
1132 * the +5 is its argument.
1142 if (PL_oldoldbufptr != PL_last_uni)
1144 while (isSPACE(*PL_last_uni))
1147 while (isALNUM_lazy_if(s,UTF) || *s == '-')
1149 if ((t = strchr(s, '(')) && t < PL_bufptr)
1152 if (ckWARN_d(WARN_AMBIGUOUS)){
1153 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
1154 "Warning: Use of \"%.*s\" without parentheses is ambiguous",
1155 (int)(s - PL_last_uni), PL_last_uni);
1160 * LOP : macro to build a list operator. Its behaviour has been replaced
1161 * with a subroutine, S_lop() for which LOP is just another name.
1164 #define LOP(f,x) return lop(f,x,s)
1168 * Build a list operator (or something that might be one). The rules:
1169 * - if we have a next token, then it's a list operator [why?]
1170 * - if the next thing is an opening paren, then it's a function
1171 * - else it's a list operator
1175 S_lop(pTHX_ I32 f, int x, char *s)
1182 PL_last_lop = PL_oldbufptr;
1183 PL_last_lop_op = (OPCODE)f;
1186 return REPORT(LSTOP);
1189 return REPORT(LSTOP);
1192 return REPORT(FUNC);
1195 return REPORT(FUNC);
1197 return REPORT(LSTOP);
1203 * Sets up for an eventual force_next(). start_force(0) basically does
1204 * an unshift, while start_force(-1) does a push. yylex removes items
1209 S_start_force(pTHX_ int where)
1213 if (where < 0) /* so people can duplicate start_force(PL_curforce) */
1214 where = PL_lasttoke;
1215 assert(PL_curforce < 0 || PL_curforce == where);
1216 if (PL_curforce != where) {
1217 for (i = PL_lasttoke; i > where; --i) {
1218 PL_nexttoke[i] = PL_nexttoke[i-1];
1222 if (PL_curforce < 0) /* in case of duplicate start_force() */
1223 Zero(&PL_nexttoke[where], 1, NEXTTOKE);
1224 PL_curforce = where;
1227 curmad('^', newSVpvs(""));
1228 CURMAD('_', PL_nextwhite);
1233 S_curmad(pTHX_ char slot, SV *sv)
1239 if (PL_curforce < 0)
1240 where = &PL_thismad;
1242 where = &PL_nexttoke[PL_curforce].next_mad;
1245 sv_setpvn(sv, "", 0);
1248 if (UTF && is_utf8_string((U8*)SvPVX(sv), SvCUR(sv)))
1250 else if (PL_encoding) {
1251 sv_recode_to_utf8(sv, PL_encoding);
1256 /* keep a slot open for the head of the list? */
1257 if (slot != '_' && *where && (*where)->mad_key == '^') {
1258 (*where)->mad_key = slot;
1259 sv_free((*where)->mad_val);
1260 (*where)->mad_val = (void*)sv;
1263 addmad(newMADsv(slot, sv), where, 0);
1266 # define start_force(where) NOOP
1267 # define curmad(slot, sv) NOOP
1272 * When the lexer realizes it knows the next token (for instance,
1273 * it is reordering tokens for the parser) then it can call S_force_next
1274 * to know what token to return the next time the lexer is called. Caller
1275 * will need to set PL_nextval[] (or PL_nexttoke[].next_val with PERL_MAD),
1276 * and possibly PL_expect to ensure the lexer handles the token correctly.
1280 S_force_next(pTHX_ I32 type)
1284 if (PL_curforce < 0)
1285 start_force(PL_lasttoke);
1286 PL_nexttoke[PL_curforce].next_type = type;
1287 if (PL_lex_state != LEX_KNOWNEXT)
1288 PL_lex_defer = PL_lex_state;
1289 PL_lex_state = LEX_KNOWNEXT;
1290 PL_lex_expect = PL_expect;
1293 PL_nexttype[PL_nexttoke] = type;
1295 if (PL_lex_state != LEX_KNOWNEXT) {
1296 PL_lex_defer = PL_lex_state;
1297 PL_lex_expect = PL_expect;
1298 PL_lex_state = LEX_KNOWNEXT;
1304 S_newSV_maybe_utf8(pTHX_ const char *start, STRLEN len)
1307 SV * const sv = newSVpvn(start,len);
1308 if (UTF && !IN_BYTES && is_utf8_string((const U8*)start, len))
1315 * When the lexer knows the next thing is a word (for instance, it has
1316 * just seen -> and it knows that the next char is a word char, then
1317 * it calls S_force_word to stick the next word into the PL_nexttoke/val
1321 * char *start : buffer position (must be within PL_linestr)
1322 * int token : PL_next* will be this type of bare word (e.g., METHOD,WORD)
1323 * int check_keyword : if true, Perl checks to make sure the word isn't
1324 * a keyword (do this if the word is a label, e.g. goto FOO)
1325 * int allow_pack : if true, : characters will also be allowed (require,
1326 * use, etc. do this)
1327 * int allow_initial_tick : used by the "sub" lexer only.
1331 S_force_word(pTHX_ register char *start, int token, int check_keyword, int allow_pack, int allow_initial_tick)
1337 start = SKIPSPACE1(start);
1339 if (isIDFIRST_lazy_if(s,UTF) ||
1340 (allow_pack && *s == ':') ||
1341 (allow_initial_tick && *s == '\'') )
1343 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, allow_pack, &len);
1344 if (check_keyword && keyword(PL_tokenbuf, len, 0))
1346 start_force(PL_curforce);
1348 curmad('X', newSVpvn(start,s-start));
1349 if (token == METHOD) {
1354 PL_expect = XOPERATOR;
1357 NEXTVAL_NEXTTOKE.opval
1358 = (OP*)newSVOP(OP_CONST,0,
1359 S_newSV_maybe_utf8(aTHX_ PL_tokenbuf, len));
1360 NEXTVAL_NEXTTOKE.opval->op_private |= OPpCONST_BARE;
1368 * Called when the lexer wants $foo *foo &foo etc, but the program
1369 * text only contains the "foo" portion. The first argument is a pointer
1370 * to the "foo", and the second argument is the type symbol to prefix.
1371 * Forces the next token to be a "WORD".
1372 * Creates the symbol if it didn't already exist (via gv_fetchpv()).
1376 S_force_ident(pTHX_ register const char *s, int kind)
1380 const STRLEN len = strlen(s);
1381 OP* const o = (OP*)newSVOP(OP_CONST, 0, newSVpvn(s, len));
1382 start_force(PL_curforce);
1383 NEXTVAL_NEXTTOKE.opval = o;
1386 o->op_private = OPpCONST_ENTERED;
1387 /* XXX see note in pp_entereval() for why we forgo typo
1388 warnings if the symbol must be introduced in an eval.
1390 gv_fetchpvn_flags(s, len,
1391 PL_in_eval ? (GV_ADDMULTI | GV_ADDINEVAL)
1393 kind == '$' ? SVt_PV :
1394 kind == '@' ? SVt_PVAV :
1395 kind == '%' ? SVt_PVHV :
1403 Perl_str_to_version(pTHX_ SV *sv)
1408 const char *start = SvPV_const(sv,len);
1409 const char * const end = start + len;
1410 const bool utf = SvUTF8(sv) ? TRUE : FALSE;
1411 while (start < end) {
1415 n = utf8n_to_uvchr((U8*)start, len, &skip, 0);
1420 retval += ((NV)n)/nshift;
1429 * Forces the next token to be a version number.
1430 * If the next token appears to be an invalid version number, (e.g. "v2b"),
1431 * and if "guessing" is TRUE, then no new token is created (and the caller
1432 * must use an alternative parsing method).
1436 S_force_version(pTHX_ char *s, int guessing)
1442 I32 startoff = s - SvPVX(PL_linestr);
1451 while (isDIGIT(*d) || *d == '_' || *d == '.')
1455 start_force(PL_curforce);
1456 curmad('X', newSVpvn(s,d-s));
1459 if (*d == ';' || isSPACE(*d) || *d == '}' || !*d) {
1461 s = scan_num(s, &yylval);
1462 version = yylval.opval;
1463 ver = cSVOPx(version)->op_sv;
1464 if (SvPOK(ver) && !SvNIOK(ver)) {
1465 SvUPGRADE(ver, SVt_PVNV);
1466 SvNV_set(ver, str_to_version(ver));
1467 SvNOK_on(ver); /* hint that it is a version */
1470 else if (guessing) {
1473 sv_free(PL_nextwhite); /* let next token collect whitespace */
1475 s = SvPVX(PL_linestr) + startoff;
1483 if (PL_madskills && !version) {
1484 sv_free(PL_nextwhite); /* let next token collect whitespace */
1486 s = SvPVX(PL_linestr) + startoff;
1489 /* NOTE: The parser sees the package name and the VERSION swapped */
1490 start_force(PL_curforce);
1491 NEXTVAL_NEXTTOKE.opval = version;
1499 * Tokenize a quoted string passed in as an SV. It finds the next
1500 * chunk, up to end of string or a backslash. It may make a new
1501 * SV containing that chunk (if HINT_NEW_STRING is on). It also
1506 S_tokeq(pTHX_ SV *sv)
1510 register char *send;
1518 s = SvPV_force(sv, len);
1519 if (SvTYPE(sv) >= SVt_PVIV && SvIVX(sv) == -1)
1522 while (s < send && *s != '\\')
1527 if ( PL_hints & HINT_NEW_STRING ) {
1528 pv = sv_2mortal(newSVpvn(SvPVX_const(pv), len));
1534 if (s + 1 < send && (s[1] == '\\'))
1535 s++; /* all that, just for this */
1540 SvCUR_set(sv, d - SvPVX_const(sv));
1542 if ( PL_hints & HINT_NEW_STRING )
1543 return new_constant(NULL, 0, "q", sv, pv, "q");
1548 * Now come three functions related to double-quote context,
1549 * S_sublex_start, S_sublex_push, and S_sublex_done. They're used when
1550 * converting things like "\u\Lgnat" into ucfirst(lc("gnat")). They
1551 * interact with PL_lex_state, and create fake ( ... ) argument lists
1552 * to handle functions and concatenation.
1553 * They assume that whoever calls them will be setting up a fake
1554 * join call, because each subthing puts a ',' after it. This lets
1557 * join($, , 'lower ', lcfirst( 'uPpEr', ) ,)
1559 * (I'm not sure whether the spurious commas at the end of lcfirst's
1560 * arguments and join's arguments are created or not).
1565 * Assumes that yylval.ival is the op we're creating (e.g. OP_LCFIRST).
1567 * Pattern matching will set PL_lex_op to the pattern-matching op to
1568 * make (we return THING if yylval.ival is OP_NULL, PMFUNC otherwise).
1570 * OP_CONST and OP_READLINE are easy--just make the new op and return.
1572 * Everything else becomes a FUNC.
1574 * Sets PL_lex_state to LEX_INTERPPUSH unless (ival was OP_NULL or we
1575 * had an OP_CONST or OP_READLINE). This just sets us up for a
1576 * call to S_sublex_push().
1580 S_sublex_start(pTHX)
1583 register const I32 op_type = yylval.ival;
1585 if (op_type == OP_NULL) {
1586 yylval.opval = PL_lex_op;
1590 if (op_type == OP_CONST || op_type == OP_READLINE) {
1591 SV *sv = tokeq(PL_lex_stuff);
1593 if (SvTYPE(sv) == SVt_PVIV) {
1594 /* Overloaded constants, nothing fancy: Convert to SVt_PV: */
1596 const char * const p = SvPV_const(sv, len);
1597 SV * const nsv = newSVpvn(p, len);
1603 yylval.opval = (OP*)newSVOP(op_type, 0, sv);
1604 PL_lex_stuff = NULL;
1605 /* Allow <FH> // "foo" */
1606 if (op_type == OP_READLINE)
1607 PL_expect = XTERMORDORDOR;
1610 else if (op_type == OP_BACKTICK && PL_lex_op) {
1611 /* readpipe() vas overriden */
1612 cSVOPx(cLISTOPx(cUNOPx(PL_lex_op)->op_first)->op_first->op_sibling)->op_sv = tokeq(PL_lex_stuff);
1613 yylval.opval = PL_lex_op;
1615 PL_lex_stuff = NULL;
1619 PL_sublex_info.super_state = PL_lex_state;
1620 PL_sublex_info.sub_inwhat = op_type;
1621 PL_sublex_info.sub_op = PL_lex_op;
1622 PL_lex_state = LEX_INTERPPUSH;
1626 yylval.opval = PL_lex_op;
1636 * Create a new scope to save the lexing state. The scope will be
1637 * ended in S_sublex_done. Returns a '(', starting the function arguments
1638 * to the uc, lc, etc. found before.
1639 * Sets PL_lex_state to LEX_INTERPCONCAT.
1648 PL_lex_state = PL_sublex_info.super_state;
1649 SAVEI32(PL_lex_dojoin);
1650 SAVEI32(PL_lex_brackets);
1651 SAVEI32(PL_lex_casemods);
1652 SAVEI32(PL_lex_starts);
1653 SAVEI32(PL_lex_state);
1654 SAVEVPTR(PL_lex_inpat);
1655 SAVEI32(PL_lex_inwhat);
1656 SAVECOPLINE(PL_curcop);
1657 SAVEPPTR(PL_bufptr);
1658 SAVEPPTR(PL_bufend);
1659 SAVEPPTR(PL_oldbufptr);
1660 SAVEPPTR(PL_oldoldbufptr);
1661 SAVEPPTR(PL_last_lop);
1662 SAVEPPTR(PL_last_uni);
1663 SAVEPPTR(PL_linestart);
1664 SAVESPTR(PL_linestr);
1665 SAVEGENERICPV(PL_lex_brackstack);
1666 SAVEGENERICPV(PL_lex_casestack);
1668 PL_linestr = PL_lex_stuff;
1669 PL_lex_stuff = NULL;
1671 PL_bufend = PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart
1672 = SvPVX(PL_linestr);
1673 PL_bufend += SvCUR(PL_linestr);
1674 PL_last_lop = PL_last_uni = NULL;
1675 SAVEFREESV(PL_linestr);
1677 PL_lex_dojoin = FALSE;
1678 PL_lex_brackets = 0;
1679 Newx(PL_lex_brackstack, 120, char);
1680 Newx(PL_lex_casestack, 12, char);
1681 PL_lex_casemods = 0;
1682 *PL_lex_casestack = '\0';
1684 PL_lex_state = LEX_INTERPCONCAT;
1685 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
1687 PL_lex_inwhat = PL_sublex_info.sub_inwhat;
1688 if (PL_lex_inwhat == OP_MATCH || PL_lex_inwhat == OP_QR || PL_lex_inwhat == OP_SUBST)
1689 PL_lex_inpat = PL_sublex_info.sub_op;
1691 PL_lex_inpat = NULL;
1698 * Restores lexer state after a S_sublex_push.
1705 if (!PL_lex_starts++) {
1706 SV * const sv = newSVpvs("");
1707 if (SvUTF8(PL_linestr))
1709 PL_expect = XOPERATOR;
1710 yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
1714 if (PL_lex_casemods) { /* oops, we've got some unbalanced parens */
1715 PL_lex_state = LEX_INTERPCASEMOD;
1719 /* Is there a right-hand side to take care of? (s//RHS/ or tr//RHS/) */
1720 if (PL_lex_repl && (PL_lex_inwhat == OP_SUBST || PL_lex_inwhat == OP_TRANS)) {
1721 PL_linestr = PL_lex_repl;
1723 PL_bufend = PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart = SvPVX(PL_linestr);
1724 PL_bufend += SvCUR(PL_linestr);
1725 PL_last_lop = PL_last_uni = NULL;
1726 SAVEFREESV(PL_linestr);
1727 PL_lex_dojoin = FALSE;
1728 PL_lex_brackets = 0;
1729 PL_lex_casemods = 0;
1730 *PL_lex_casestack = '\0';
1732 if (SvEVALED(PL_lex_repl)) {
1733 PL_lex_state = LEX_INTERPNORMAL;
1735 /* we don't clear PL_lex_repl here, so that we can check later
1736 whether this is an evalled subst; that means we rely on the
1737 logic to ensure sublex_done() is called again only via the
1738 branch (in yylex()) that clears PL_lex_repl, else we'll loop */
1741 PL_lex_state = LEX_INTERPCONCAT;
1751 PL_endwhite = newSVpvs("");
1752 sv_catsv(PL_endwhite, PL_thiswhite);
1756 sv_setpvn(PL_thistoken,"",0);
1758 PL_realtokenstart = -1;
1762 PL_bufend = SvPVX(PL_linestr);
1763 PL_bufend += SvCUR(PL_linestr);
1764 PL_expect = XOPERATOR;
1765 PL_sublex_info.sub_inwhat = 0;
1773 Extracts a pattern, double-quoted string, or transliteration. This
1776 It looks at PL_lex_inwhat and PL_lex_inpat to find out whether it's
1777 processing a pattern (PL_lex_inpat is true), a transliteration
1778 (PL_lex_inwhat == OP_TRANS is true), or a double-quoted string.
1780 Returns a pointer to the character scanned up to. If this is
1781 advanced from the start pointer supplied (i.e. if anything was
1782 successfully parsed), will leave an OP for the substring scanned
1783 in yylval. Caller must intuit reason for not parsing further
1784 by looking at the next characters herself.
1788 double-quoted style: \r and \n
1789 regexp special ones: \D \s
1792 case and quoting: \U \Q \E
1793 stops on @ and $, but not for $ as tail anchor
1795 In transliterations:
1796 characters are VERY literal, except for - not at the start or end
1797 of the string, which indicates a range. If the range is in bytes,
1798 scan_const expands the range to the full set of intermediate
1799 characters. If the range is in utf8, the hyphen is replaced with
1800 a certain range mark which will be handled by pmtrans() in op.c.
1802 In double-quoted strings:
1804 double-quoted style: \r and \n
1806 deprecated backrefs: \1 (in substitution replacements)
1807 case and quoting: \U \Q \E
1810 scan_const does *not* construct ops to handle interpolated strings.
1811 It stops processing as soon as it finds an embedded $ or @ variable
1812 and leaves it to the caller to work out what's going on.
1814 embedded arrays (whether in pattern or not) could be:
1815 @foo, @::foo, @'foo, @{foo}, @$foo, @+, @-.
1817 $ in double-quoted strings must be the symbol of an embedded scalar.
1819 $ in pattern could be $foo or could be tail anchor. Assumption:
1820 it's a tail anchor if $ is the last thing in the string, or if it's
1821 followed by one of "()| \r\n\t"
1823 \1 (backreferences) are turned into $1
1825 The structure of the code is
1826 while (there's a character to process) {
1827 handle transliteration ranges
1828 skip regexp comments /(?#comment)/ and codes /(?{code})/
1829 skip #-initiated comments in //x patterns
1830 check for embedded arrays
1831 check for embedded scalars
1833 leave intact backslashes from leaveit (below)
1834 deprecate \1 in substitution replacements
1835 handle string-changing backslashes \l \U \Q \E, etc.
1836 switch (what was escaped) {
1837 handle \- in a transliteration (becomes a literal -)
1838 handle \132 (octal characters)
1839 handle \x15 and \x{1234} (hex characters)
1840 handle \N{name} (named characters)
1841 handle \cV (control characters)
1842 handle printf-style backslashes (\f, \r, \n, etc)
1844 } (end if backslash)
1845 } (end while character to read)
1850 S_scan_const(pTHX_ char *start)
1853 register char *send = PL_bufend; /* end of the constant */
1854 SV *sv = newSV(send - start); /* sv for the constant */
1855 register char *s = start; /* start of the constant */
1856 register char *d = SvPVX(sv); /* destination for copies */
1857 bool dorange = FALSE; /* are we in a translit range? */
1858 bool didrange = FALSE; /* did we just finish a range? */
1859 I32 has_utf8 = FALSE; /* Output constant is UTF8 */
1860 I32 this_utf8 = UTF; /* The source string is assumed to be UTF8 */
1863 UV literal_endpoint = 0;
1864 bool native_range = TRUE; /* turned to FALSE if the first endpoint is Unicode. */
1867 if (PL_lex_inwhat == OP_TRANS && PL_sublex_info.sub_op) {
1868 /* If we are doing a trans and we know we want UTF8 set expectation */
1869 has_utf8 = PL_sublex_info.sub_op->op_private & (OPpTRANS_FROM_UTF|OPpTRANS_TO_UTF);
1870 this_utf8 = PL_sublex_info.sub_op->op_private & (PL_lex_repl ? OPpTRANS_FROM_UTF : OPpTRANS_TO_UTF);
1874 while (s < send || dorange) {
1875 /* get transliterations out of the way (they're most literal) */
1876 if (PL_lex_inwhat == OP_TRANS) {
1877 /* expand a range A-Z to the full set of characters. AIE! */
1879 I32 i; /* current expanded character */
1880 I32 min; /* first character in range */
1881 I32 max; /* last character in range */
1892 char * const c = (char*)utf8_hop((U8*)d, -1);
1896 *c = (char)UTF_TO_NATIVE(0xff);
1897 /* mark the range as done, and continue */
1903 i = d - SvPVX_const(sv); /* remember current offset */
1906 SvLEN(sv) + (has_utf8 ?
1907 (512 - UTF_CONTINUATION_MARK +
1910 /* How many two-byte within 0..255: 128 in UTF-8,
1911 * 96 in UTF-8-mod. */
1913 SvGROW(sv, SvLEN(sv) + 256); /* never more than 256 chars in a range */
1915 d = SvPVX(sv) + i; /* refresh d after realloc */
1919 for (j = 0; j <= 1; j++) {
1920 char * const c = (char*)utf8_hop((U8*)d, -1);
1921 const UV uv = utf8n_to_uvchr((U8*)c, d - c, NULL, 0);
1927 max = (U8)0xff; /* only to \xff */
1928 uvmax = uv; /* \x{100} to uvmax */
1930 d = c; /* eat endpoint chars */
1935 d -= 2; /* eat the first char and the - */
1936 min = (U8)*d; /* first char in range */
1937 max = (U8)d[1]; /* last char in range */
1944 "Invalid range \"%c-%c\" in transliteration operator",
1945 (char)min, (char)max);
1949 if (literal_endpoint == 2 &&
1950 ((isLOWER(min) && isLOWER(max)) ||
1951 (isUPPER(min) && isUPPER(max)))) {
1953 for (i = min; i <= max; i++)
1955 *d++ = NATIVE_TO_NEED(has_utf8,i);
1957 for (i = min; i <= max; i++)
1959 *d++ = NATIVE_TO_NEED(has_utf8,i);
1964 for (i = min; i <= max; i++)
1967 const U8 ch = (U8)NATIVE_TO_UTF(i);
1968 if (UNI_IS_INVARIANT(ch))
1971 *d++ = (U8)UTF8_EIGHT_BIT_HI(ch);
1972 *d++ = (U8)UTF8_EIGHT_BIT_LO(ch);
1981 d = (char*)uvchr_to_utf8((U8*)d, 0x100);
1983 *d++ = (char)UTF_TO_NATIVE(0xff);
1985 d = (char*)uvchr_to_utf8((U8*)d, uvmax);
1989 /* mark the range as done, and continue */
1993 literal_endpoint = 0;
1998 /* range begins (ignore - as first or last char) */
1999 else if (*s == '-' && s+1 < send && s != start) {
2001 Perl_croak(aTHX_ "Ambiguous range in transliteration operator");
2008 *d++ = (char)UTF_TO_NATIVE(0xff); /* use illegal utf8 byte--see pmtrans */
2018 literal_endpoint = 0;
2019 native_range = TRUE;
2024 /* if we get here, we're not doing a transliteration */
2026 /* skip for regexp comments /(?#comment)/ and code /(?{code})/,
2027 except for the last char, which will be done separately. */
2028 else if (*s == '(' && PL_lex_inpat && s[1] == '?') {
2030 while (s+1 < send && *s != ')')
2031 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2033 else if (s[2] == '{' /* This should match regcomp.c */
2034 || ((s[2] == 'p' || s[2] == '?') && s[3] == '{'))
2037 char *regparse = s + (s[2] == '{' ? 3 : 4);
2040 while (count && (c = *regparse)) {
2041 if (c == '\\' && regparse[1])
2049 if (*regparse != ')')
2050 regparse--; /* Leave one char for continuation. */
2051 while (s < regparse)
2052 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2056 /* likewise skip #-initiated comments in //x patterns */
2057 else if (*s == '#' && PL_lex_inpat &&
2058 ((PMOP*)PL_lex_inpat)->op_pmflags & PMf_EXTENDED) {
2059 while (s+1 < send && *s != '\n')
2060 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2063 /* check for embedded arrays
2064 (@foo, @::foo, @'foo, @{foo}, @$foo, @+, @-)
2066 else if (*s == '@' && s[1]) {
2067 if (isALNUM_lazy_if(s+1,UTF))
2069 if (strchr(":'{$", s[1]))
2071 if (!PL_lex_inpat && (s[1] == '+' || s[1] == '-'))
2072 break; /* in regexp, neither @+ nor @- are interpolated */
2075 /* check for embedded scalars. only stop if we're sure it's a
2078 else if (*s == '$') {
2079 if (!PL_lex_inpat) /* not a regexp, so $ must be var */
2081 if (s + 1 < send && !strchr("()| \r\n\t", s[1]))
2082 break; /* in regexp, $ might be tail anchor */
2085 /* End of else if chain - OP_TRANS rejoin rest */
2088 if (*s == '\\' && s+1 < send) {
2091 /* deprecate \1 in strings and substitution replacements */
2092 if (PL_lex_inwhat == OP_SUBST && !PL_lex_inpat &&
2093 isDIGIT(*s) && *s != '0' && !isDIGIT(s[1]))
2095 if (ckWARN(WARN_SYNTAX))
2096 Perl_warner(aTHX_ packWARN(WARN_SYNTAX), "\\%c better written as $%c", *s, *s);
2101 /* string-change backslash escapes */
2102 if (PL_lex_inwhat != OP_TRANS && *s && strchr("lLuUEQ", *s)) {
2106 /* skip any other backslash escapes in a pattern */
2107 else if (PL_lex_inpat) {
2108 *d++ = NATIVE_TO_NEED(has_utf8,'\\');
2109 goto default_action;
2112 /* if we get here, it's either a quoted -, or a digit */
2115 /* quoted - in transliterations */
2117 if (PL_lex_inwhat == OP_TRANS) {
2124 if ((isALPHA(*s) || isDIGIT(*s)) &&
2126 Perl_warner(aTHX_ packWARN(WARN_MISC),
2127 "Unrecognized escape \\%c passed through",
2129 /* default action is to copy the quoted character */
2130 goto default_action;
2133 /* \132 indicates an octal constant */
2134 case '0': case '1': case '2': case '3':
2135 case '4': case '5': case '6': case '7':
2139 uv = grok_oct(s, &len, &flags, NULL);
2142 goto NUM_ESCAPE_INSERT;
2144 /* \x24 indicates a hex constant */
2148 char* const e = strchr(s, '}');
2149 I32 flags = PERL_SCAN_ALLOW_UNDERSCORES |
2150 PERL_SCAN_DISALLOW_PREFIX;
2155 yyerror("Missing right brace on \\x{}");
2159 uv = grok_hex(s, &len, &flags, NULL);
2165 I32 flags = PERL_SCAN_DISALLOW_PREFIX;
2166 uv = grok_hex(s, &len, &flags, NULL);
2172 /* Insert oct or hex escaped character.
2173 * There will always enough room in sv since such
2174 * escapes will be longer than any UTF-8 sequence
2175 * they can end up as. */
2177 /* We need to map to chars to ASCII before doing the tests
2180 if (!UNI_IS_INVARIANT(NATIVE_TO_UNI(uv))) {
2181 if (!has_utf8 && uv > 255) {
2182 /* Might need to recode whatever we have
2183 * accumulated so far if it contains any
2186 * (Can't we keep track of that and avoid
2187 * this rescan? --jhi)
2191 for (c = (U8 *) SvPVX(sv); c < (U8 *)d; c++) {
2192 if (!NATIVE_IS_INVARIANT(*c)) {
2197 const STRLEN offset = d - SvPVX_const(sv);
2199 d = SvGROW(sv, SvLEN(sv) + hicount + 1) + offset;
2203 while (src >= (const U8 *)SvPVX_const(sv)) {
2204 if (!NATIVE_IS_INVARIANT(*src)) {
2205 const U8 ch = NATIVE_TO_ASCII(*src);
2206 *dst-- = (U8)UTF8_EIGHT_BIT_LO(ch);
2207 *dst-- = (U8)UTF8_EIGHT_BIT_HI(ch);
2217 if (has_utf8 || uv > 255) {
2218 d = (char*)uvchr_to_utf8((U8*)d, uv);
2220 if (PL_lex_inwhat == OP_TRANS &&
2221 PL_sublex_info.sub_op) {
2222 PL_sublex_info.sub_op->op_private |=
2223 (PL_lex_repl ? OPpTRANS_FROM_UTF
2227 if (uv > 255 && !dorange)
2228 native_range = FALSE;
2240 /* \N{LATIN SMALL LETTER A} is a named character */
2244 char* e = strchr(s, '}');
2251 yyerror("Missing right brace on \\N{}");
2255 if (e > s + 2 && s[1] == 'U' && s[2] == '+') {
2257 I32 flags = PERL_SCAN_ALLOW_UNDERSCORES |
2258 PERL_SCAN_DISALLOW_PREFIX;
2261 uv = grok_hex(s, &len, &flags, NULL);
2262 if ( e > s && len != (STRLEN)(e - s) ) {
2266 goto NUM_ESCAPE_INSERT;
2268 res = newSVpvn(s + 1, e - s - 1);
2269 type = newSVpvn(s - 2,e - s + 3);
2270 res = new_constant( NULL, 0, "charnames",
2271 res, NULL, SvPVX(type) );
2274 sv_utf8_upgrade(res);
2275 str = SvPV_const(res,len);
2276 #ifdef EBCDIC_NEVER_MIND
2277 /* charnames uses pack U and that has been
2278 * recently changed to do the below uni->native
2279 * mapping, so this would be redundant (and wrong,
2280 * the code point would be doubly converted).
2281 * But leave this in just in case the pack U change
2282 * gets revoked, but the semantics is still
2283 * desireable for charnames. --jhi */
2285 UV uv = utf8_to_uvchr((const U8*)str, 0);
2288 U8 tmpbuf[UTF8_MAXBYTES+1], *d;
2290 d = uvchr_to_utf8(tmpbuf, UNI_TO_NATIVE(uv));
2291 sv_setpvn(res, (char *)tmpbuf, d - tmpbuf);
2292 str = SvPV_const(res, len);
2296 if (!has_utf8 && SvUTF8(res)) {
2297 const char * const ostart = SvPVX_const(sv);
2298 SvCUR_set(sv, d - ostart);
2301 sv_utf8_upgrade(sv);
2302 /* this just broke our allocation above... */
2303 SvGROW(sv, (STRLEN)(send - start));
2304 d = SvPVX(sv) + SvCUR(sv);
2307 if (len > (STRLEN)(e - s + 4)) { /* I _guess_ 4 is \N{} --jhi */
2308 const char * const odest = SvPVX_const(sv);
2310 SvGROW(sv, (SvLEN(sv) + len - (e - s + 4)));
2311 d = SvPVX(sv) + (d - odest);
2315 native_range = FALSE; /* \N{} is guessed to be Unicode */
2317 Copy(str, d, len, char);
2324 yyerror("Missing braces on \\N{}");
2327 /* \c is a control character */
2336 *d++ = NATIVE_TO_NEED(has_utf8,toCTRL(c));
2339 yyerror("Missing control char name in \\c");
2343 /* printf-style backslashes, formfeeds, newlines, etc */
2345 *d++ = NATIVE_TO_NEED(has_utf8,'\b');
2348 *d++ = NATIVE_TO_NEED(has_utf8,'\n');
2351 *d++ = NATIVE_TO_NEED(has_utf8,'\r');
2354 *d++ = NATIVE_TO_NEED(has_utf8,'\f');
2357 *d++ = NATIVE_TO_NEED(has_utf8,'\t');
2360 *d++ = ASCII_TO_NEED(has_utf8,'\033');
2363 *d++ = ASCII_TO_NEED(has_utf8,'\007');
2369 } /* end if (backslash) */
2376 /* If we started with encoded form, or already know we want it
2377 and then encode the next character */
2378 if ((has_utf8 || this_utf8) && !NATIVE_IS_INVARIANT((U8)(*s))) {
2380 const UV nextuv = (this_utf8) ? utf8n_to_uvchr((U8*)s, send - s, &len, 0) : (UV) ((U8) *s);
2381 const STRLEN need = UNISKIP(NATIVE_TO_UNI(nextuv));
2384 /* encoded value larger than old, need extra space (NOTE: SvCUR() not set here) */
2385 const STRLEN off = d - SvPVX_const(sv);
2386 d = SvGROW(sv, SvLEN(sv) + (need-len)) + off;
2388 d = (char*)uvchr_to_utf8((U8*)d, nextuv);
2391 if (uv > 255 && !dorange)
2392 native_range = FALSE;
2396 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2398 } /* while loop to process each character */
2400 /* terminate the string and set up the sv */
2402 SvCUR_set(sv, d - SvPVX_const(sv));
2403 if (SvCUR(sv) >= SvLEN(sv))
2404 Perl_croak(aTHX_ "panic: constant overflowed allocated space");
2407 if (PL_encoding && !has_utf8) {
2408 sv_recode_to_utf8(sv, PL_encoding);
2414 if (PL_lex_inwhat == OP_TRANS && PL_sublex_info.sub_op) {
2415 PL_sublex_info.sub_op->op_private |=
2416 (PL_lex_repl ? OPpTRANS_FROM_UTF : OPpTRANS_TO_UTF);
2420 /* shrink the sv if we allocated more than we used */
2421 if (SvCUR(sv) + 5 < SvLEN(sv)) {
2422 SvPV_shrink_to_cur(sv);
2425 /* return the substring (via yylval) only if we parsed anything */
2426 if (s > PL_bufptr) {
2427 if ( PL_hints & ( PL_lex_inpat ? HINT_NEW_RE : HINT_NEW_STRING ) )
2428 sv = new_constant(start, s - start,
2429 (const char *)(PL_lex_inpat ? "qr" : "q"),
2432 (( PL_lex_inwhat == OP_TRANS
2434 : ( (PL_lex_inwhat == OP_SUBST && !PL_lex_inpat)
2437 yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
2444 * Returns TRUE if there's more to the expression (e.g., a subscript),
2447 * It deals with "$foo[3]" and /$foo[3]/ and /$foo[0123456789$]+/
2449 * ->[ and ->{ return TRUE
2450 * { and [ outside a pattern are always subscripts, so return TRUE
2451 * if we're outside a pattern and it's not { or [, then return FALSE
2452 * if we're in a pattern and the first char is a {
2453 * {4,5} (any digits around the comma) returns FALSE
2454 * if we're in a pattern and the first char is a [
2456 * [SOMETHING] has a funky algorithm to decide whether it's a
2457 * character class or not. It has to deal with things like
2458 * /$foo[-3]/ and /$foo[$bar]/ as well as /$foo[$\d]+/
2459 * anything else returns TRUE
2462 /* This is the one truly awful dwimmer necessary to conflate C and sed. */
2465 S_intuit_more(pTHX_ register char *s)
2468 if (PL_lex_brackets)
2470 if (*s == '-' && s[1] == '>' && (s[2] == '[' || s[2] == '{'))
2472 if (*s != '{' && *s != '[')
2477 /* In a pattern, so maybe we have {n,m}. */
2494 /* On the other hand, maybe we have a character class */
2497 if (*s == ']' || *s == '^')
2500 /* this is terrifying, and it works */
2501 int weight = 2; /* let's weigh the evidence */
2503 unsigned char un_char = 255, last_un_char;
2504 const char * const send = strchr(s,']');
2505 char tmpbuf[sizeof PL_tokenbuf * 4];
2507 if (!send) /* has to be an expression */
2510 Zero(seen,256,char);
2513 else if (isDIGIT(*s)) {
2515 if (isDIGIT(s[1]) && s[2] == ']')
2521 for (; s < send; s++) {
2522 last_un_char = un_char;
2523 un_char = (unsigned char)*s;
2528 weight -= seen[un_char] * 10;
2529 if (isALNUM_lazy_if(s+1,UTF)) {
2531 scan_ident(s, send, tmpbuf, sizeof tmpbuf, FALSE);
2532 len = (int)strlen(tmpbuf);
2533 if (len > 1 && gv_fetchpvn_flags(tmpbuf, len, 0, SVt_PV))
2538 else if (*s == '$' && s[1] &&
2539 strchr("[#!%*<>()-=",s[1])) {
2540 if (/*{*/ strchr("])} =",s[2]))
2549 if (strchr("wds]",s[1]))
2551 else if (seen[(U8)'\''] || seen[(U8)'"'])
2553 else if (strchr("rnftbxcav",s[1]))
2555 else if (isDIGIT(s[1])) {
2557 while (s[1] && isDIGIT(s[1]))
2567 if (strchr("aA01! ",last_un_char))
2569 if (strchr("zZ79~",s[1]))
2571 if (last_un_char == 255 && (isDIGIT(s[1]) || s[1] == '$'))
2572 weight -= 5; /* cope with negative subscript */
2575 if (!isALNUM(last_un_char)
2576 && !(last_un_char == '$' || last_un_char == '@'
2577 || last_un_char == '&')
2578 && isALPHA(*s) && s[1] && isALPHA(s[1])) {
2583 if (keyword(tmpbuf, d - tmpbuf, 0))
2586 if (un_char == last_un_char + 1)
2588 weight -= seen[un_char];
2593 if (weight >= 0) /* probably a character class */
2603 * Does all the checking to disambiguate
2605 * between foo(bar) and bar->foo. Returns 0 if not a method, otherwise
2606 * FUNCMETH (bar->foo(args)) or METHOD (bar->foo args).
2608 * First argument is the stuff after the first token, e.g. "bar".
2610 * Not a method if bar is a filehandle.
2611 * Not a method if foo is a subroutine prototyped to take a filehandle.
2612 * Not a method if it's really "Foo $bar"
2613 * Method if it's "foo $bar"
2614 * Not a method if it's really "print foo $bar"
2615 * Method if it's really "foo package::" (interpreted as package->foo)
2616 * Not a method if bar is known to be a subroutine ("sub bar; foo bar")
2617 * Not a method if bar is a filehandle or package, but is quoted with
2622 S_intuit_method(pTHX_ char *start, GV *gv, CV *cv)
2625 char *s = start + (*start == '$');
2626 char tmpbuf[sizeof PL_tokenbuf];
2634 if (SvTYPE(gv) == SVt_PVGV && GvIO(gv))
2638 const char *proto = SvPVX_const(cv);
2649 s = scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
2650 /* start is the beginning of the possible filehandle/object,
2651 * and s is the end of it
2652 * tmpbuf is a copy of it
2655 if (*start == '$') {
2656 if (gv || PL_last_lop_op == OP_PRINT || isUPPER(*PL_tokenbuf))
2659 len = start - SvPVX(PL_linestr);
2663 start = SvPVX(PL_linestr) + len;
2667 return *s == '(' ? FUNCMETH : METHOD;
2669 if (!keyword(tmpbuf, len, 0)) {
2670 if (len > 2 && tmpbuf[len - 2] == ':' && tmpbuf[len - 1] == ':') {
2674 soff = s - SvPVX(PL_linestr);
2678 indirgv = gv_fetchpvn_flags(tmpbuf, len, 0, SVt_PVCV);
2679 if (indirgv && GvCVu(indirgv))
2681 /* filehandle or package name makes it a method */
2682 if (!gv || GvIO(indirgv) || gv_stashpvn(tmpbuf, len, FALSE)) {
2684 soff = s - SvPVX(PL_linestr);
2687 if ((PL_bufend - s) >= 2 && *s == '=' && *(s+1) == '>')
2688 return 0; /* no assumptions -- "=>" quotes bearword */
2690 start_force(PL_curforce);
2691 NEXTVAL_NEXTTOKE.opval = (OP*)newSVOP(OP_CONST, 0,
2692 newSVpvn(tmpbuf,len));
2693 NEXTVAL_NEXTTOKE.opval->op_private = OPpCONST_BARE;
2695 curmad('X', newSVpvn(start,SvPVX(PL_linestr) + soff - start));
2700 PL_bufptr = SvPVX(PL_linestr) + soff; /* restart before space */
2702 return *s == '(' ? FUNCMETH : METHOD;
2710 * Return a string of Perl code to load the debugger. If PERL5DB
2711 * is set, it will return the contents of that, otherwise a
2712 * compile-time require of perl5db.pl.
2720 const char * const pdb = PerlEnv_getenv("PERL5DB");
2724 SETERRNO(0,SS_NORMAL);
2725 return "BEGIN { require 'perl5db.pl' }";
2731 /* Encoded script support. filter_add() effectively inserts a
2732 * 'pre-processing' function into the current source input stream.
2733 * Note that the filter function only applies to the current source file
2734 * (e.g., it will not affect files 'require'd or 'use'd by this one).
2736 * The datasv parameter (which may be NULL) can be used to pass
2737 * private data to this instance of the filter. The filter function
2738 * can recover the SV using the FILTER_DATA macro and use it to
2739 * store private buffers and state information.
2741 * The supplied datasv parameter is upgraded to a PVIO type
2742 * and the IoDIRP/IoANY field is used to store the function pointer,
2743 * and IOf_FAKE_DIRP is enabled on datasv to mark this as such.
2744 * Note that IoTOP_NAME, IoFMT_NAME, IoBOTTOM_NAME, if set for
2745 * private use must be set using malloc'd pointers.
2749 Perl_filter_add(pTHX_ filter_t funcp, SV *datasv)
2755 if (!PL_rsfp_filters)
2756 PL_rsfp_filters = newAV();
2759 SvUPGRADE(datasv, SVt_PVIO);
2760 IoANY(datasv) = FPTR2DPTR(void *, funcp); /* stash funcp into spare field */
2761 IoFLAGS(datasv) |= IOf_FAKE_DIRP;
2762 DEBUG_P(PerlIO_printf(Perl_debug_log, "filter_add func %p (%s)\n",
2763 FPTR2DPTR(void *, IoANY(datasv)),
2764 SvPV_nolen(datasv)));
2765 av_unshift(PL_rsfp_filters, 1);
2766 av_store(PL_rsfp_filters, 0, datasv) ;
2771 /* Delete most recently added instance of this filter function. */
2773 Perl_filter_del(pTHX_ filter_t funcp)
2779 DEBUG_P(PerlIO_printf(Perl_debug_log, "filter_del func %p",
2780 FPTR2DPTR(void*, funcp)));
2782 if (!PL_rsfp_filters || AvFILLp(PL_rsfp_filters)<0)
2784 /* if filter is on top of stack (usual case) just pop it off */
2785 datasv = FILTER_DATA(AvFILLp(PL_rsfp_filters));
2786 if (IoANY(datasv) == FPTR2DPTR(void *, funcp)) {
2787 IoFLAGS(datasv) &= ~IOf_FAKE_DIRP;
2788 IoANY(datasv) = (void *)NULL;
2789 sv_free(av_pop(PL_rsfp_filters));
2793 /* we need to search for the correct entry and clear it */
2794 Perl_die(aTHX_ "filter_del can only delete in reverse order (currently)");
2798 /* Invoke the idxth filter function for the current rsfp. */
2799 /* maxlen 0 = read one text line */
2801 Perl_filter_read(pTHX_ int idx, SV *buf_sv, int maxlen)
2806 /* This API is bad. It should have been using unsigned int for maxlen.
2807 Not sure if we want to change the API, but if not we should sanity
2808 check the value here. */
2809 const unsigned int correct_length
2818 if (!PL_rsfp_filters)
2820 if (idx > AvFILLp(PL_rsfp_filters)) { /* Any more filters? */
2821 /* Provide a default input filter to make life easy. */
2822 /* Note that we append to the line. This is handy. */
2823 DEBUG_P(PerlIO_printf(Perl_debug_log,
2824 "filter_read %d: from rsfp\n", idx));
2825 if (correct_length) {
2828 const int old_len = SvCUR(buf_sv);
2830 /* ensure buf_sv is large enough */
2831 SvGROW(buf_sv, (STRLEN)(old_len + correct_length)) ;
2832 if ((len = PerlIO_read(PL_rsfp, SvPVX(buf_sv) + old_len,
2833 correct_length)) <= 0) {
2834 if (PerlIO_error(PL_rsfp))
2835 return -1; /* error */
2837 return 0 ; /* end of file */
2839 SvCUR_set(buf_sv, old_len + len) ;
2842 if (sv_gets(buf_sv, PL_rsfp, SvCUR(buf_sv)) == NULL) {
2843 if (PerlIO_error(PL_rsfp))
2844 return -1; /* error */
2846 return 0 ; /* end of file */
2849 return SvCUR(buf_sv);
2851 /* Skip this filter slot if filter has been deleted */
2852 if ( (datasv = FILTER_DATA(idx)) == &PL_sv_undef) {
2853 DEBUG_P(PerlIO_printf(Perl_debug_log,
2854 "filter_read %d: skipped (filter deleted)\n",
2856 return FILTER_READ(idx+1, buf_sv, correct_length); /* recurse */
2858 /* Get function pointer hidden within datasv */
2859 funcp = DPTR2FPTR(filter_t, IoANY(datasv));
2860 DEBUG_P(PerlIO_printf(Perl_debug_log,
2861 "filter_read %d: via function %p (%s)\n",
2862 idx, (void*)datasv, SvPV_nolen_const(datasv)));
2863 /* Call function. The function is expected to */
2864 /* call "FILTER_READ(idx+1, buf_sv)" first. */
2865 /* Return: <0:error, =0:eof, >0:not eof */
2866 return (*funcp)(aTHX_ idx, buf_sv, correct_length);
2870 S_filter_gets(pTHX_ register SV *sv, register PerlIO *fp, STRLEN append)
2873 #ifdef PERL_CR_FILTER
2874 if (!PL_rsfp_filters) {
2875 filter_add(S_cr_textfilter,NULL);
2878 if (PL_rsfp_filters) {
2880 SvCUR_set(sv, 0); /* start with empty line */
2881 if (FILTER_READ(0, sv, 0) > 0)
2882 return ( SvPVX(sv) ) ;
2887 return (sv_gets(sv, fp, append));
2891 S_find_in_my_stash(pTHX_ const char *pkgname, I32 len)
2896 if (len == 11 && *pkgname == '_' && strEQ(pkgname, "__PACKAGE__"))
2900 (pkgname[len - 2] == ':' && pkgname[len - 1] == ':') &&
2901 (gv = gv_fetchpvn_flags(pkgname, len, 0, SVt_PVHV)))
2903 return GvHV(gv); /* Foo:: */
2906 /* use constant CLASS => 'MyClass' */
2907 gv = gv_fetchpvn_flags(pkgname, len, 0, SVt_PVCV);
2908 if (gv && GvCV(gv)) {
2909 SV * const sv = cv_const_sv(GvCV(gv));
2911 pkgname = SvPV_nolen_const(sv);
2914 return gv_stashpv(pkgname, FALSE);
2918 * S_readpipe_override
2919 * Check whether readpipe() is overriden, and generates the appropriate
2920 * optree, provided sublex_start() is called afterwards.
2923 S_readpipe_override(pTHX)
2926 GV *gv_readpipe = gv_fetchpvs("readpipe", GV_NOTQUAL, SVt_PVCV);
2927 yylval.ival = OP_BACKTICK;
2929 && GvCVu(gv_readpipe) && GvIMPORTED_CV(gv_readpipe))
2931 ((gvp = (GV**)hv_fetchs(PL_globalstash, "readpipe", FALSE))
2932 && (gv_readpipe = *gvp) != (GV*)&PL_sv_undef
2933 && GvCVu(gv_readpipe) && GvIMPORTED_CV(gv_readpipe)))
2935 PL_lex_op = (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
2936 append_elem(OP_LIST,
2937 newSVOP(OP_CONST, 0, &PL_sv_undef), /* value will be read later */
2938 newCVREF(0, newGVOP(OP_GV, 0, gv_readpipe))));
2948 * The intent of this yylex wrapper is to minimize the changes to the
2949 * tokener when we aren't interested in collecting madprops. It remains
2950 * to be seen how successful this strategy will be...
2957 char *s = PL_bufptr;
2959 /* make sure PL_thiswhite is initialized */
2963 /* just do what yylex would do on pending identifier; leave PL_thiswhite alone */
2964 if (PL_pending_ident)
2965 return S_pending_ident(aTHX);
2967 /* previous token ate up our whitespace? */
2968 if (!PL_lasttoke && PL_nextwhite) {
2969 PL_thiswhite = PL_nextwhite;
2973 /* isolate the token, and figure out where it is without whitespace */
2974 PL_realtokenstart = -1;
2978 assert(PL_curforce < 0);
2980 if (!PL_thismad || PL_thismad->mad_key == '^') { /* not forced already? */
2981 if (!PL_thistoken) {
2982 if (PL_realtokenstart < 0 || !CopLINE(PL_curcop))
2983 PL_thistoken = newSVpvs("");
2985 char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
2986 PL_thistoken = newSVpvn(tstart, s - tstart);
2989 if (PL_thismad) /* install head */
2990 CURMAD('X', PL_thistoken);
2993 /* last whitespace of a sublex? */
2994 if (optype == ')' && PL_endwhite) {
2995 CURMAD('X', PL_endwhite);
3000 /* if no whitespace and we're at EOF, bail. Otherwise fake EOF below. */
3001 if (!PL_thiswhite && !PL_endwhite && !optype) {
3002 sv_free(PL_thistoken);
3007 /* put off final whitespace till peg */
3008 if (optype == ';' && !PL_rsfp) {
3009 PL_nextwhite = PL_thiswhite;
3012 else if (PL_thisopen) {
3013 CURMAD('q', PL_thisopen);
3015 sv_free(PL_thistoken);
3019 /* Store actual token text as madprop X */
3020 CURMAD('X', PL_thistoken);
3024 /* add preceding whitespace as madprop _ */
3025 CURMAD('_', PL_thiswhite);
3029 /* add quoted material as madprop = */
3030 CURMAD('=', PL_thisstuff);
3034 /* add terminating quote as madprop Q */
3035 CURMAD('Q', PL_thisclose);
3039 /* special processing based on optype */
3043 /* opval doesn't need a TOKEN since it can already store mp */
3054 append_madprops(PL_thismad, yylval.opval, 0);
3062 addmad(newMADsv('p', PL_endwhite), &PL_thismad, 0);
3071 /* remember any fake bracket that lexer is about to discard */
3072 if (PL_lex_brackets == 1 &&
3073 ((expectation)PL_lex_brackstack[0] & XFAKEBRACK))
3076 while (s < PL_bufend && (*s == ' ' || *s == '\t'))
3079 PL_thiswhite = newSVpvn(PL_bufptr, ++s - PL_bufptr);
3080 addmad(newMADsv('#', PL_thiswhite), &PL_thismad, 0);
3083 break; /* don't bother looking for trailing comment */
3092 /* attach a trailing comment to its statement instead of next token */
3096 if (PL_bufptr > PL_oldbufptr && PL_bufptr[-1] == optype) {
3098 while (s < PL_bufend && (*s == ' ' || *s == '\t'))
3100 if (*s == '\n' || *s == '#') {
3101 while (s < PL_bufend && *s != '\n')
3105 PL_thiswhite = newSVpvn(PL_bufptr, s - PL_bufptr);
3106 addmad(newMADsv('#', PL_thiswhite), &PL_thismad, 0);
3123 /* Create new token struct. Note: opvals return early above. */
3124 yylval.tkval = newTOKEN(optype, yylval, PL_thismad);
3131 S_tokenize_use(pTHX_ int is_use, char *s) {
3133 if (PL_expect != XSTATE)
3134 yyerror(Perl_form(aTHX_ "\"%s\" not allowed in expression",
3135 is_use ? "use" : "no"));
3137 if (isDIGIT(*s) || (*s == 'v' && isDIGIT(s[1]))) {
3138 s = force_version(s, TRUE);
3139 if (*s == ';' || (s = SKIPSPACE1(s), *s == ';')) {
3140 start_force(PL_curforce);
3141 NEXTVAL_NEXTTOKE.opval = NULL;
3144 else if (*s == 'v') {
3145 s = force_word(s,WORD,FALSE,TRUE,FALSE);
3146 s = force_version(s, FALSE);
3150 s = force_word(s,WORD,FALSE,TRUE,FALSE);
3151 s = force_version(s, FALSE);
3153 yylval.ival = is_use;
3157 static const char* const exp_name[] =
3158 { "OPERATOR", "TERM", "REF", "STATE", "BLOCK", "ATTRBLOCK",
3159 "ATTRTERM", "TERMBLOCK", "TERMORDORDOR"
3166 Works out what to call the token just pulled out of the input
3167 stream. The yacc parser takes care of taking the ops we return and
3168 stitching them into a tree.
3174 if read an identifier
3175 if we're in a my declaration
3176 croak if they tried to say my($foo::bar)
3177 build the ops for a my() declaration
3178 if it's an access to a my() variable
3179 are we in a sort block?
3180 croak if my($a); $a <=> $b
3181 build ops for access to a my() variable
3182 if in a dq string, and they've said @foo and we can't find @foo
3184 build ops for a bareword
3185 if we already built the token before, use it.
3190 #pragma segment Perl_yylex
3196 register char *s = PL_bufptr;
3201 /* orig_keyword, gvp, and gv are initialized here because
3202 * jump to the label just_a_word_zero can bypass their
3203 * initialization later. */
3204 I32 orig_keyword = 0;
3209 SV* tmp = newSVpvs("");
3210 PerlIO_printf(Perl_debug_log, "### %"IVdf":LEX_%s/X%s %s\n",
3211 (IV)CopLINE(PL_curcop),
3212 lex_state_names[PL_lex_state],
3213 exp_name[PL_expect],
3214 pv_display(tmp, s, strlen(s), 0, 60));
3217 /* check if there's an identifier for us to look at */
3218 if (PL_pending_ident)
3219 return REPORT(S_pending_ident(aTHX));
3221 /* no identifier pending identification */
3223 switch (PL_lex_state) {
3225 case LEX_NORMAL: /* Some compilers will produce faster */
3226 case LEX_INTERPNORMAL: /* code if we comment these out. */
3230 /* when we've already built the next token, just pull it out of the queue */
3234 yylval = PL_nexttoke[PL_lasttoke].next_val;
3236 PL_thismad = PL_nexttoke[PL_lasttoke].next_mad;
3237 PL_nexttoke[PL_lasttoke].next_mad = 0;
3238 if (PL_thismad && PL_thismad->mad_key == '_') {
3239 PL_thiswhite = (SV*)PL_thismad->mad_val;
3240 PL_thismad->mad_val = 0;
3241 mad_free(PL_thismad);
3246 PL_lex_state = PL_lex_defer;
3247 PL_expect = PL_lex_expect;
3248 PL_lex_defer = LEX_NORMAL;
3249 if (!PL_nexttoke[PL_lasttoke].next_type)
3254 yylval = PL_nextval[PL_nexttoke];
3256 PL_lex_state = PL_lex_defer;
3257 PL_expect = PL_lex_expect;
3258 PL_lex_defer = LEX_NORMAL;
3262 /* FIXME - can these be merged? */
3263 return(PL_nexttoke[PL_lasttoke].next_type);
3265 return REPORT(PL_nexttype[PL_nexttoke]);
3268 /* interpolated case modifiers like \L \U, including \Q and \E.
3269 when we get here, PL_bufptr is at the \
3271 case LEX_INTERPCASEMOD:
3273 if (PL_bufptr != PL_bufend && *PL_bufptr != '\\')
3274 Perl_croak(aTHX_ "panic: INTERPCASEMOD");
3276 /* handle \E or end of string */
3277 if (PL_bufptr == PL_bufend || PL_bufptr[1] == 'E') {
3279 if (PL_lex_casemods) {
3280 const char oldmod = PL_lex_casestack[--PL_lex_casemods];
3281 PL_lex_casestack[PL_lex_casemods] = '\0';
3283 if (PL_bufptr != PL_bufend
3284 && (oldmod == 'L' || oldmod == 'U' || oldmod == 'Q')) {
3286 PL_lex_state = LEX_INTERPCONCAT;
3289 PL_thistoken = newSVpvs("\\E");
3295 while (PL_bufptr != PL_bufend &&
3296 PL_bufptr[0] == '\\' && PL_bufptr[1] == 'E') {
3298 PL_thiswhite = newSVpvs("");
3299 sv_catpvn(PL_thiswhite, PL_bufptr, 2);
3303 if (PL_bufptr != PL_bufend)
3306 PL_lex_state = LEX_INTERPCONCAT;
3310 DEBUG_T({ PerlIO_printf(Perl_debug_log,
3311 "### Saw case modifier\n"); });
3313 if (s[1] == '\\' && s[2] == 'E') {
3316 PL_thiswhite = newSVpvs("");
3317 sv_catpvn(PL_thiswhite, PL_bufptr, 4);
3320 PL_lex_state = LEX_INTERPCONCAT;
3325 if (!PL_madskills) /* when just compiling don't need correct */
3326 if (strnEQ(s, "L\\u", 3) || strnEQ(s, "U\\l", 3))
3327 tmp = *s, *s = s[2], s[2] = (char)tmp; /* misordered... */
3328 if ((*s == 'L' || *s == 'U') &&
3329 (strchr(PL_lex_casestack, 'L') || strchr(PL_lex_casestack, 'U'))) {
3330 PL_lex_casestack[--PL_lex_casemods] = '\0';
3333 if (PL_lex_casemods > 10)
3334 Renew(PL_lex_casestack, PL_lex_casemods + 2, char);
3335 PL_lex_casestack[PL_lex_casemods++] = *s;
3336 PL_lex_casestack[PL_lex_casemods] = '\0';
3337 PL_lex_state = LEX_INTERPCONCAT;
3338 start_force(PL_curforce);
3339 NEXTVAL_NEXTTOKE.ival = 0;
3341 start_force(PL_curforce);
3343 NEXTVAL_NEXTTOKE.ival = OP_LCFIRST;
3345 NEXTVAL_NEXTTOKE.ival = OP_UCFIRST;
3347 NEXTVAL_NEXTTOKE.ival = OP_LC;
3349 NEXTVAL_NEXTTOKE.ival = OP_UC;
3351 NEXTVAL_NEXTTOKE.ival = OP_QUOTEMETA;
3353 Perl_croak(aTHX_ "panic: yylex");
3355 SV* const tmpsv = newSVpvs("");
3356 Perl_sv_catpvf(aTHX_ tmpsv, "\\%c", *s);
3362 if (PL_lex_starts) {
3368 sv_free(PL_thistoken);
3369 PL_thistoken = newSVpvs("");
3372 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
3373 if (PL_lex_casemods == 1 && PL_lex_inpat)
3382 case LEX_INTERPPUSH:
3383 return REPORT(sublex_push());
3385 case LEX_INTERPSTART:
3386 if (PL_bufptr == PL_bufend)
3387 return REPORT(sublex_done());
3388 DEBUG_T({ PerlIO_printf(Perl_debug_log,
3389 "### Interpolated variable\n"); });
3391 PL_lex_dojoin = (*PL_bufptr == '@');
3392 PL_lex_state = LEX_INTERPNORMAL;
3393 if (PL_lex_dojoin) {
3394 start_force(PL_curforce);
3395 NEXTVAL_NEXTTOKE.ival = 0;
3397 start_force(PL_curforce);
3398 force_ident("\"", '$');
3399 start_force(PL_curforce);
3400 NEXTVAL_NEXTTOKE.ival = 0;
3402 start_force(PL_curforce);
3403 NEXTVAL_NEXTTOKE.ival = 0;
3405 start_force(PL_curforce);
3406 NEXTVAL_NEXTTOKE.ival = OP_JOIN; /* emulate join($", ...) */
3409 if (PL_lex_starts++) {
3414 sv_free(PL_thistoken);
3415 PL_thistoken = newSVpvs("");
3418 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
3419 if (!PL_lex_casemods && PL_lex_inpat)
3426 case LEX_INTERPENDMAYBE:
3427 if (intuit_more(PL_bufptr)) {
3428 PL_lex_state = LEX_INTERPNORMAL; /* false alarm, more expr */
3434 if (PL_lex_dojoin) {
3435 PL_lex_dojoin = FALSE;
3436 PL_lex_state = LEX_INTERPCONCAT;
3440 sv_free(PL_thistoken);
3441 PL_thistoken = newSVpvs("");
3446 if (PL_lex_inwhat == OP_SUBST && PL_linestr == PL_lex_repl
3447 && SvEVALED(PL_lex_repl))
3449 if (PL_bufptr != PL_bufend)
3450 Perl_croak(aTHX_ "Bad evalled substitution pattern");
3454 case LEX_INTERPCONCAT:
3456 if (PL_lex_brackets)
3457 Perl_croak(aTHX_ "panic: INTERPCONCAT");
3459 if (PL_bufptr == PL_bufend)
3460 return REPORT(sublex_done());
3462 if (SvIVX(PL_linestr) == '\'') {
3463 SV *sv = newSVsv(PL_linestr);
3466 else if ( PL_hints & HINT_NEW_RE )
3467 sv = new_constant(NULL, 0, "qr", sv, sv, "q");
3468 yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
3472 s = scan_const(PL_bufptr);
3474 PL_lex_state = LEX_INTERPCASEMOD;
3476 PL_lex_state = LEX_INTERPSTART;
3479 if (s != PL_bufptr) {
3480 start_force(PL_curforce);
3482 curmad('X', newSVpvn(PL_bufptr,s-PL_bufptr));
3484 NEXTVAL_NEXTTOKE = yylval;
3487 if (PL_lex_starts++) {
3491 sv_free(PL_thistoken);
3492 PL_thistoken = newSVpvs("");
3495 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
3496 if (!PL_lex_casemods && PL_lex_inpat)
3509 PL_lex_state = LEX_NORMAL;
3510 s = scan_formline(PL_bufptr);
3511 if (!PL_lex_formbrack)
3517 PL_oldoldbufptr = PL_oldbufptr;
3523 sv_free(PL_thistoken);
3526 PL_realtokenstart = s - SvPVX(PL_linestr); /* assume but undo on ws */
3530 if (isIDFIRST_lazy_if(s,UTF))
3532 Perl_croak(aTHX_ "Unrecognized character \\x%02X", *s & 255);
3535 goto fake_eof; /* emulate EOF on ^D or ^Z */
3544 if (PL_lex_brackets) {
3545 yyerror((const char *)
3547 ? "Format not terminated"
3548 : "Missing right curly or square bracket"));
3550 DEBUG_T( { PerlIO_printf(Perl_debug_log,
3551 "### Tokener got EOF\n");
3555 if (s++ < PL_bufend)
3556 goto retry; /* ignore stray nulls */
3559 if (!PL_in_eval && !PL_preambled) {
3560 PL_preambled = TRUE;
3565 sv_setpv(PL_linestr,incl_perldb());
3566 if (SvCUR(PL_linestr))
3567 sv_catpvs(PL_linestr,";");
3569 while(AvFILLp(PL_preambleav) >= 0) {
3570 SV *tmpsv = av_shift(PL_preambleav);
3571 sv_catsv(PL_linestr, tmpsv);
3572 sv_catpvs(PL_linestr, ";");
3575 sv_free((SV*)PL_preambleav);
3576 PL_preambleav = NULL;
3578 if (PL_minus_n || PL_minus_p) {
3579 sv_catpvs(PL_linestr, "LINE: while (<>) {");
3581 sv_catpvs(PL_linestr,"chomp;");
3584 if ((*PL_splitstr == '/' || *PL_splitstr == '\''
3585 || *PL_splitstr == '"')
3586 && strchr(PL_splitstr + 1, *PL_splitstr))
3587 Perl_sv_catpvf(aTHX_ PL_linestr, "our @F=split(%s);", PL_splitstr);
3589 /* "q\0${splitstr}\0" is legal perl. Yes, even NUL
3590 bytes can be used as quoting characters. :-) */
3591 const char *splits = PL_splitstr;
3592 sv_catpvs(PL_linestr, "our @F=split(q\0");
3595 if (*splits == '\\')
3596 sv_catpvn(PL_linestr, splits, 1);
3597 sv_catpvn(PL_linestr, splits, 1);
3598 } while (*splits++);
3599 /* This loop will embed the trailing NUL of
3600 PL_linestr as the last thing it does before
3602 sv_catpvs(PL_linestr, ");");
3606 sv_catpvs(PL_linestr,"our @F=split(' ');");
3610 sv_catpvs(PL_linestr,"use feature ':5.10';");
3611 sv_catpvs(PL_linestr, "\n");
3612 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3613 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3614 PL_last_lop = PL_last_uni = NULL;
3615 if (PERLDB_LINE && PL_curstash != PL_debstash)
3616 update_debugger_info_sv(PL_linestr);
3620 bof = PL_rsfp ? TRUE : FALSE;
3621 if ((s = filter_gets(PL_linestr, PL_rsfp, 0)) == NULL) {
3624 PL_realtokenstart = -1;
3627 if (PL_preprocess && !PL_in_eval)
3628 (void)PerlProc_pclose(PL_rsfp);
3629 else if ((PerlIO *)PL_rsfp == PerlIO_stdin())
3630 PerlIO_clearerr(PL_rsfp);
3632 (void)PerlIO_close(PL_rsfp);
3634 PL_doextract = FALSE;
3636 if (!PL_in_eval && (PL_minus_n || PL_minus_p)) {
3641 sv_setpv(PL_linestr,
3644 ? ";}continue{print;}" : ";}"));
3645 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3646 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3647 PL_last_lop = PL_last_uni = NULL;
3648 PL_minus_n = PL_minus_p = 0;
3651 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3652 PL_last_lop = PL_last_uni = NULL;
3653 sv_setpvn(PL_linestr,"",0);
3654 TOKEN(';'); /* not infinite loop because rsfp is NULL now */
3656 /* If it looks like the start of a BOM or raw UTF-16,
3657 * check if it in fact is. */
3663 #ifdef PERLIO_IS_STDIO
3664 # ifdef __GNU_LIBRARY__
3665 # if __GNU_LIBRARY__ == 1 /* Linux glibc5 */
3666 # define FTELL_FOR_PIPE_IS_BROKEN
3670 # if __GLIBC__ == 1 /* maybe some glibc5 release had it like this? */
3671 # define FTELL_FOR_PIPE_IS_BROKEN
3676 #ifdef FTELL_FOR_PIPE_IS_BROKEN
3677 /* This loses the possibility to detect the bof
3678 * situation on perl -P when the libc5 is being used.
3679 * Workaround? Maybe attach some extra state to PL_rsfp?
3682 bof = PerlIO_tell(PL_rsfp) == SvCUR(PL_linestr);
3684 bof = PerlIO_tell(PL_rsfp) == (Off_t)SvCUR(PL_linestr);
3687 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3688 s = swallow_bom((U8*)s);
3692 /* Incest with pod. */
3695 sv_catsv(PL_thiswhite, PL_linestr);
3697 if (*s == '=' && strnEQ(s, "=cut", 4)) {
3698 sv_setpvn(PL_linestr, "", 0);
3699 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3700 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3701 PL_last_lop = PL_last_uni = NULL;
3702 PL_doextract = FALSE;
3706 } while (PL_doextract);
3707 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = s;
3708 if (PERLDB_LINE && PL_curstash != PL_debstash)
3709 update_debugger_info_sv(PL_linestr);
3710 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3711 PL_last_lop = PL_last_uni = NULL;
3712 if (CopLINE(PL_curcop) == 1) {
3713 while (s < PL_bufend && isSPACE(*s))
3715 if (*s == ':' && s[1] != ':') /* for csh execing sh scripts */
3719 PL_thiswhite = newSVpvn(PL_linestart, s - PL_linestart);
3723 if (*s == '#' && *(s+1) == '!')
3725 #ifdef ALTERNATE_SHEBANG
3727 static char const as[] = ALTERNATE_SHEBANG;
3728 if (*s == as[0] && strnEQ(s, as, sizeof(as) - 1))
3729 d = s + (sizeof(as) - 1);
3731 #endif /* ALTERNATE_SHEBANG */
3740 while (*d && !isSPACE(*d))
3744 #ifdef ARG_ZERO_IS_SCRIPT
3745 if (ipathend > ipath) {
3747 * HP-UX (at least) sets argv[0] to the script name,
3748 * which makes $^X incorrect. And Digital UNIX and Linux,
3749 * at least, set argv[0] to the basename of the Perl
3750 * interpreter. So, having found "#!", we'll set it right.
3752 SV * const x = GvSV(gv_fetchpvs("\030", GV_ADD|GV_NOTQUAL,
3754 assert(SvPOK(x) || SvGMAGICAL(x));
3755 if (sv_eq(x, CopFILESV(PL_curcop))) {
3756 sv_setpvn(x, ipath, ipathend - ipath);
3762 const char *bstart = SvPV_const(CopFILESV(PL_curcop),blen);
3763 const char * const lstart = SvPV_const(x,llen);
3765 bstart += blen - llen;
3766 if (strnEQ(bstart, lstart, llen) && bstart[-1] == '/') {
3767 sv_setpvn(x, ipath, ipathend - ipath);
3772 TAINT_NOT; /* $^X is always tainted, but that's OK */
3774 #endif /* ARG_ZERO_IS_SCRIPT */
3779 d = instr(s,"perl -");
3781 d = instr(s,"perl");
3783 /* avoid getting into infinite loops when shebang
3784 * line contains "Perl" rather than "perl" */
3786 for (d = ipathend-4; d >= ipath; --d) {
3787 if ((*d == 'p' || *d == 'P')
3788 && !ibcmp(d, "perl", 4))
3798 #ifdef ALTERNATE_SHEBANG
3800 * If the ALTERNATE_SHEBANG on this system starts with a
3801 * character that can be part of a Perl expression, then if
3802 * we see it but not "perl", we're probably looking at the
3803 * start of Perl code, not a request to hand off to some
3804 * other interpreter. Similarly, if "perl" is there, but
3805 * not in the first 'word' of the line, we assume the line
3806 * contains the start of the Perl program.
3808 if (d && *s != '#') {
3809 const char *c = ipath;
3810 while (*c && !strchr("; \t\r\n\f\v#", *c))
3813 d = NULL; /* "perl" not in first word; ignore */
3815 *s = '#'; /* Don't try to parse shebang line */
3817 #endif /* ALTERNATE_SHEBANG */
3818 #ifndef MACOS_TRADITIONAL
3823 !instr(s,"indir") &&
3824 instr(PL_origargv[0],"perl"))
3831 while (s < PL_bufend && isSPACE(*s))
3833 if (s < PL_bufend) {
3834 Newxz(newargv,PL_origargc+3,char*);
3836 while (s < PL_bufend && !isSPACE(*s))
3839 Copy(PL_origargv+1, newargv+2, PL_origargc+1, char*);
3842 newargv = PL_origargv;
3845 PerlProc_execv(ipath, EXEC_ARGV_CAST(newargv));
3847 Perl_croak(aTHX_ "Can't exec %s", ipath);
3851 while (*d && !isSPACE(*d))
3853 while (SPACE_OR_TAB(*d))
3857 const bool switches_done = PL_doswitches;
3858 const U32 oldpdb = PL_perldb;
3859 const bool oldn = PL_minus_n;
3860 const bool oldp = PL_minus_p;
3863 if (*d == 'M' || *d == 'm' || *d == 'C') {
3864 const char * const m = d;
3865 while (*d && !isSPACE(*d))
3867 Perl_croak(aTHX_ "Too late for \"-%.*s\" option",
3870 d = moreswitches(d);
3872 if (PL_doswitches && !switches_done) {
3873 int argc = PL_origargc;
3874 char **argv = PL_origargv;
3877 } while (argc && argv[0][0] == '-' && argv[0][1]);
3878 init_argv_symbols(argc,argv);
3880 if ((PERLDB_LINE && !oldpdb) ||
3881 ((PL_minus_n || PL_minus_p) && !(oldn || oldp)))
3882 /* if we have already added "LINE: while (<>) {",
3883 we must not do it again */
3885 sv_setpvn(PL_linestr, "", 0);
3886 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3887 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3888 PL_last_lop = PL_last_uni = NULL;
3889 PL_preambled = FALSE;
3891 (void)gv_fetchfile(PL_origfilename);
3898 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
3900 PL_lex_state = LEX_FORMLINE;
3905 #ifdef PERL_STRICT_CR
3906 Perl_warn(aTHX_ "Illegal character \\%03o (carriage return)", '\r');
3908 "\t(Maybe you didn't strip carriage returns after a network transfer?)\n");
3910 case ' ': case '\t': case '\f': case 013:
3911 #ifdef MACOS_TRADITIONAL
3915 PL_realtokenstart = -1;
3924 PL_realtokenstart = -1;
3928 if (PL_lex_state != LEX_NORMAL || (PL_in_eval && !PL_rsfp)) {
3929 if (*s == '#' && s == PL_linestart && PL_in_eval && !PL_rsfp) {
3930 /* handle eval qq[#line 1 "foo"\n ...] */
3931 CopLINE_dec(PL_curcop);
3934 if (PL_madskills && !PL_lex_formbrack && !PL_in_eval) {
3936 if (!PL_in_eval || PL_rsfp)
3941 while (d < PL_bufend && *d != '\n')
3945 else if (d > PL_bufend) /* Found by Ilya: feed random input to Perl. */
3946 Perl_croak(aTHX_ "panic: input overflow");
3949 PL_thiswhite = newSVpvn(s, d - s);
3954 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
3956 PL_lex_state = LEX_FORMLINE;
3962 if (PL_madskills && CopLINE(PL_curcop) >= 1 && !PL_lex_formbrack) {
3963 if (CopLINE(PL_curcop) == 1 && s[0] == '#' && s[1] == '!') {
3966 TOKEN(PEG); /* make sure any #! line is accessible */
3971 /* if (PL_madskills && PL_lex_formbrack) { */
3973 while (d < PL_bufend && *d != '\n')
3977 else if (d > PL_bufend) /* Found by Ilya: feed random input to Perl. */
3978 Perl_croak(aTHX_ "panic: input overflow");
3979 if (PL_madskills && CopLINE(PL_curcop) >= 1) {
3981 PL_thiswhite = newSVpvs("");
3982 if (CopLINE(PL_curcop) == 1) {
3983 sv_setpvn(PL_thiswhite, "", 0);
3986 sv_catpvn(PL_thiswhite, s, d - s);
4000 if (s[1] && isALPHA(s[1]) && !isALNUM(s[2])) {
4008 while (s < PL_bufend && SPACE_OR_TAB(*s))
4011 if (strnEQ(s,"=>",2)) {
4012 s = force_word(PL_bufptr,WORD,FALSE,FALSE,FALSE);
4013 DEBUG_T( { printbuf("### Saw unary minus before =>, forcing word %s\n", s); } );
4014 OPERATOR('-'); /* unary minus */
4016 PL_last_uni = PL_oldbufptr;
4018 case 'r': ftst = OP_FTEREAD; break;
4019 case 'w': ftst = OP_FTEWRITE; break;
4020 case 'x': ftst = OP_FTEEXEC; break;
4021 case 'o': ftst = OP_FTEOWNED; break;
4022 case 'R': ftst = OP_FTRREAD; break;
4023 case 'W': ftst = OP_FTRWRITE; break;
4024 case 'X': ftst = OP_FTREXEC; break;
4025 case 'O': ftst = OP_FTROWNED; break;
4026 case 'e': ftst = OP_FTIS; break;
4027 case 'z': ftst = OP_FTZERO; break;
4028 case 's': ftst = OP_FTSIZE; break;
4029 case 'f': ftst = OP_FTFILE; break;
4030 case 'd': ftst = OP_FTDIR; break;
4031 case 'l': ftst = OP_FTLINK; break;
4032 case 'p': ftst = OP_FTPIPE; break;
4033 case 'S': ftst = OP_FTSOCK; break;
4034 case 'u': ftst = OP_FTSUID; break;
4035 case 'g': ftst = OP_FTSGID; break;
4036 case 'k': ftst = OP_FTSVTX; break;
4037 case 'b': ftst = OP_FTBLK; break;
4038 case 'c': ftst = OP_FTCHR; break;
4039 case 't': ftst = OP_FTTTY; break;
4040 case 'T': ftst = OP_FTTEXT; break;
4041 case 'B': ftst = OP_FTBINARY; break;
4042 case 'M': case 'A': case 'C':
4043 gv_fetchpvs("\024", GV_ADD|GV_NOTQUAL, SVt_PV);
4045 case 'M': ftst = OP_FTMTIME; break;
4046 case 'A': ftst = OP_FTATIME; break;
4047 case 'C': ftst = OP_FTCTIME; break;
4055 PL_last_lop_op = (OPCODE)ftst;
4056 DEBUG_T( { PerlIO_printf(Perl_debug_log,
4057 "### Saw file test %c\n", (int)tmp);
4062 /* Assume it was a minus followed by a one-letter named
4063 * subroutine call (or a -bareword), then. */
4064 DEBUG_T( { PerlIO_printf(Perl_debug_log,
4065 "### '-%c' looked like a file test but was not\n",
4072 const char tmp = *s++;
4075 if (PL_expect == XOPERATOR)
4080 else if (*s == '>') {
4083 if (isIDFIRST_lazy_if(s,UTF)) {
4084 s = force_word(s,METHOD,FALSE,TRUE,FALSE);
4092 if (PL_expect == XOPERATOR)
4095 if (isSPACE(*s) || !isSPACE(*PL_bufptr))
4097 OPERATOR('-'); /* unary minus */
4103 const char tmp = *s++;
4106 if (PL_expect == XOPERATOR)
4111 if (PL_expect == XOPERATOR)
4114 if (isSPACE(*s) || !isSPACE(*PL_bufptr))
4121 if (PL_expect != XOPERATOR) {
4122 s = scan_ident(s, PL_bufend, PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
4123 PL_expect = XOPERATOR;
4124 force_ident(PL_tokenbuf, '*');
4137 if (PL_expect == XOPERATOR) {
4141 PL_tokenbuf[0] = '%';
4142 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1, TRUE);
4143 if (!PL_tokenbuf[1]) {
4146 PL_pending_ident = '%';
4157 && (PL_expect == XOPERATOR || PL_expect == XTERMORDORDOR)
4158 && FEATURE_IS_ENABLED("~~"))
4165 const char tmp = *s++;
4171 goto just_a_word_zero_gv;
4174 switch (PL_expect) {
4180 if (!PL_in_my || PL_lex_state != LEX_NORMAL)
4182 PL_bufptr = s; /* update in case we back off */
4188 PL_expect = XTERMBLOCK;
4191 stuffstart = s - SvPVX(PL_linestr) - 1;
4195 while (isIDFIRST_lazy_if(s,UTF)) {
4198 d = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
4199 if (isLOWER(*s) && (tmp = keyword(PL_tokenbuf, len, 0))) {
4200 if (tmp < 0) tmp = -tmp;
4215 sv = newSVpvn(s, len);
4217 d = scan_str(d,TRUE,TRUE);
4219 /* MUST advance bufptr here to avoid bogus
4220 "at end of line" context messages from yyerror().
4222 PL_bufptr = s + len;
4223 yyerror("Unterminated attribute parameter in attribute list");
4227 return REPORT(0); /* EOF indicator */
4231 sv_catsv(sv, PL_lex_stuff);
4232 attrs = append_elem(OP_LIST, attrs,
4233 newSVOP(OP_CONST, 0, sv));
4234 SvREFCNT_dec(PL_lex_stuff);
4235 PL_lex_stuff = NULL;
4238 if (len == 6 && strnEQ(SvPVX(sv), "unique", len)) {
4240 if (PL_in_my == KEY_our) {
4242 GvUNIQUE_on(cGVOPx_gv(yylval.opval));
4244 /* skip to avoid loading attributes.pm */
4246 deprecate(":unique");
4249 Perl_croak(aTHX_ "The 'unique' attribute may only be applied to 'our' variables");
4252 /* NOTE: any CV attrs applied here need to be part of
4253 the CVf_BUILTIN_ATTRS define in cv.h! */
4254 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "lvalue", len)) {
4256 CvLVALUE_on(PL_compcv);
4258 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "locked", len)) {
4260 CvLOCKED_on(PL_compcv);
4262 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "method", len)) {
4264 CvMETHOD_on(PL_compcv);
4266 else if (!PL_in_my && len == 9 && strnEQ(SvPVX(sv), "assertion", len)) {
4268 CvASSERTION_on(PL_compcv);
4270 /* After we've set the flags, it could be argued that
4271 we don't need to do the attributes.pm-based setting
4272 process, and shouldn't bother appending recognized
4273 flags. To experiment with that, uncomment the
4274 following "else". (Note that's already been
4275 uncommented. That keeps the above-applied built-in
4276 attributes from being intercepted (and possibly
4277 rejected) by a package's attribute routines, but is
4278 justified by the performance win for the common case
4279 of applying only built-in attributes.) */
4281 attrs = append_elem(OP_LIST, attrs,
4282 newSVOP(OP_CONST, 0,
4286 if (*s == ':' && s[1] != ':')
4289 break; /* require real whitespace or :'s */
4290 /* XXX losing whitespace on sequential attributes here */
4294 = (PL_expect == XOPERATOR ? '=' : '{'); /*'}(' for vi */
4295 if (*s != ';' && *s != '}' && *s != tmp
4296 && (tmp != '=' || *s != ')')) {
4297 const char q = ((*s == '\'') ? '"' : '\'');
4298 /* If here for an expression, and parsed no attrs, back
4300 if (tmp == '=' && !attrs) {
4304 /* MUST advance bufptr here to avoid bogus "at end of line"
4305 context messages from yyerror().
4308 yyerror( (const char *)
4310 ? Perl_form(aTHX_ "Invalid separator character "
4311 "%c%c%c in attribute list", q, *s, q)
4312 : "Unterminated attribute list" ) );
4320 start_force(PL_curforce);
4321 NEXTVAL_NEXTTOKE.opval = attrs;
4322 CURMAD('_', PL_nextwhite);
4327 PL_thistoken = newSVpvn(SvPVX(PL_linestr) + stuffstart,
4328 (s - SvPVX(PL_linestr)) - stuffstart);
4336 if (PL_last_lop == PL_oldoldbufptr || PL_last_uni == PL_oldoldbufptr)
4337 PL_oldbufptr = PL_oldoldbufptr; /* allow print(STDOUT 123) */
4345 const char tmp = *s++;
4350 const char tmp = *s++;
4358 if (PL_lex_brackets <= 0)
4359 yyerror("Unmatched right square bracket");
4362 if (PL_lex_state == LEX_INTERPNORMAL) {
4363 if (PL_lex_brackets == 0) {
4364 if (*s != '[' && *s != '{' && (*s != '-' || s[1] != '>'))
4365 PL_lex_state = LEX_INTERPEND;
4372 if (PL_lex_brackets > 100) {
4373 Renew(PL_lex_brackstack, PL_lex_brackets + 10, char);
4375 switch (PL_expect) {
4377 if (PL_lex_formbrack) {
4381 if (PL_oldoldbufptr == PL_last_lop)
4382 PL_lex_brackstack[PL_lex_brackets++] = XTERM;
4384 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
4385 OPERATOR(HASHBRACK);
4387 while (s < PL_bufend && SPACE_OR_TAB(*s))
4390 PL_tokenbuf[0] = '\0';
4391 if (d < PL_bufend && *d == '-') {
4392 PL_tokenbuf[0] = '-';
4394 while (d < PL_bufend && SPACE_OR_TAB(*d))
4397 if (d < PL_bufend && isIDFIRST_lazy_if(d,UTF)) {
4398 d = scan_word(d, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1,
4400 while (d < PL_bufend && SPACE_OR_TAB(*d))
4403 const char minus = (PL_tokenbuf[0] == '-');
4404 s = force_word(s + minus, WORD, FALSE, TRUE, FALSE);
4412 PL_lex_brackstack[PL_lex_brackets++] = XSTATE;
4417 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
4422 if (PL_oldoldbufptr == PL_last_lop)
4423 PL_lex_brackstack[PL_lex_brackets++] = XTERM;
4425 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
4428 if (PL_expect == XREF && PL_lex_state == LEX_INTERPNORMAL) {
4430 /* This hack is to get the ${} in the message. */
4432 yyerror("syntax error");
4435 OPERATOR(HASHBRACK);
4437 /* This hack serves to disambiguate a pair of curlies
4438 * as being a block or an anon hash. Normally, expectation
4439 * determines that, but in cases where we're not in a
4440 * position to expect anything in particular (like inside
4441 * eval"") we have to resolve the ambiguity. This code
4442 * covers the case where the first term in the curlies is a
4443 * quoted string. Most other cases need to be explicitly
4444 * disambiguated by prepending a "+" before the opening
4445 * curly in order to force resolution as an anon hash.
4447 * XXX should probably propagate the outer expectation
4448 * into eval"" to rely less on this hack, but that could
4449 * potentially break current behavior of eval"".
4453 if (*s == '\'' || *s == '"' || *s == '`') {
4454 /* common case: get past first string, handling escapes */
4455 for (t++; t < PL_bufend && *t != *s;)
4456 if (*t++ == '\\' && (*t == '\\' || *t == *s))
4460 else if (*s == 'q') {
4463 || ((*t == 'q' || *t == 'x') && ++t < PL_bufend
4466 /* skip q//-like construct */
4468 char open, close, term;
4471 while (t < PL_bufend && isSPACE(*t))
4473 /* check for q => */
4474 if (t+1 < PL_bufend && t[0] == '=' && t[1] == '>') {
4475 OPERATOR(HASHBRACK);
4479 if (term && (tmps = strchr("([{< )]}> )]}>",term)))
4483 for (t++; t < PL_bufend; t++) {
4484 if (*t == '\\' && t+1 < PL_bufend && open != '\\')
4486 else if (*t == open)
4490 for (t++; t < PL_bufend; t++) {
4491 if (*t == '\\' && t+1 < PL_bufend)
4493 else if (*t == close && --brackets <= 0)
4495 else if (*t == open)
4502 /* skip plain q word */
4503 while (t < PL_bufend && isALNUM_lazy_if(t,UTF))
4506 else if (isALNUM_lazy_if(t,UTF)) {
4508 while (t < PL_bufend && isALNUM_lazy_if(t,UTF))
4511 while (t < PL_bufend && isSPACE(*t))
4513 /* if comma follows first term, call it an anon hash */
4514 /* XXX it could be a comma expression with loop modifiers */
4515 if (t < PL_bufend && ((*t == ',' && (*s == 'q' || !isLOWER(*s)))
4516 || (*t == '=' && t[1] == '>')))
4517 OPERATOR(HASHBRACK);
4518 if (PL_expect == XREF)
4521 PL_lex_brackstack[PL_lex_brackets-1] = XSTATE;
4527 yylval.ival = CopLINE(PL_curcop);
4528 if (isSPACE(*s) || *s == '#')
4529 PL_copline = NOLINE; /* invalidate current command line number */
4534 if (PL_lex_brackets <= 0)
4535 yyerror("Unmatched right curly bracket");
4537 PL_expect = (expectation)PL_lex_brackstack[--PL_lex_brackets];
4538 if (PL_lex_brackets < PL_lex_formbrack && PL_lex_state != LEX_INTERPNORMAL)
4539 PL_lex_formbrack = 0;
4540 if (PL_lex_state == LEX_INTERPNORMAL) {
4541 if (PL_lex_brackets == 0) {
4542 if (PL_expect & XFAKEBRACK) {
4543 PL_expect &= XENUMMASK;
4544 PL_lex_state = LEX_INTERPEND;
4549 PL_thiswhite = newSVpvs("");
4550 sv_catpvn(PL_thiswhite,"}",1);
4553 return yylex(); /* ignore fake brackets */
4555 if (*s == '-' && s[1] == '>')
4556 PL_lex_state = LEX_INTERPENDMAYBE;
4557 else if (*s != '[' && *s != '{')
4558 PL_lex_state = LEX_INTERPEND;
4561 if (PL_expect & XFAKEBRACK) {
4562 PL_expect &= XENUMMASK;
4564 return yylex(); /* ignore fake brackets */
4566 start_force(PL_curforce);
4568 curmad('X', newSVpvn(s-1,1));
4569 CURMAD('_', PL_thiswhite);
4574 PL_thistoken = newSVpvs("");
4582 if (PL_expect == XOPERATOR) {
4583 if (PL_bufptr == PL_linestart && ckWARN(WARN_SEMICOLON)
4584 && isIDFIRST_lazy_if(s,UTF))
4586 CopLINE_dec(PL_curcop);
4587 Perl_warner(aTHX_ packWARN(WARN_SEMICOLON), PL_warn_nosemi);
4588 CopLINE_inc(PL_curcop);
4593 s = scan_ident(s - 1, PL_bufend, PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
4595 PL_expect = XOPERATOR;
4596 force_ident(PL_tokenbuf, '&');
4600 yylval.ival = (OPpENTERSUB_AMPER<<8);
4612 const char tmp = *s++;
4619 if (tmp && isSPACE(*s) && ckWARN(WARN_SYNTAX)
4620 && strchr("+-*/%.^&|<",tmp))
4621 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
4622 "Reversed %c= operator",(int)tmp);
4624 if (PL_expect == XSTATE && isALPHA(tmp) &&
4625 (s == PL_linestart+1 || s[-2] == '\n') )
4627 if (PL_in_eval && !PL_rsfp) {
4632 if (strnEQ(s,"=cut",4)) {
4648 PL_thiswhite = newSVpvs("");
4649 sv_catpvn(PL_thiswhite, PL_linestart,
4650 PL_bufend - PL_linestart);
4654 PL_doextract = TRUE;
4658 if (PL_lex_brackets < PL_lex_formbrack) {
4660 #ifdef PERL_STRICT_CR
4661 while (SPACE_OR_TAB(*t))
4663 while (SPACE_OR_TAB(*t) || *t == '\r')
4666 if (*t == '\n' || *t == '#') {
4677 const char tmp = *s++;
4679 /* was this !=~ where !~ was meant?
4680 * warn on m:!=~\s+([/?]|[msy]\W|tr\W): */
4682 if (*s == '~' && ckWARN(WARN_SYNTAX)) {
4683 const char *t = s+1;
4685 while (t < PL_bufend && isSPACE(*t))
4688 if (*t == '/' || *t == '?' ||
4689 ((*t == 'm' || *t == 's' || *t == 'y')
4690 && !isALNUM(t[1])) ||
4691 (*t == 't' && t[1] == 'r' && !isALNUM(t[2])))
4692 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
4693 "!=~ should be !~");
4703 if (PL_expect != XOPERATOR) {
4704 if (s[1] != '<' && !strchr(s,'>'))
4707 s = scan_heredoc(s);
4709 s = scan_inputsymbol(s);
4710 TERM(sublex_start());
4716 SHop(OP_LEFT_SHIFT);
4730 const char tmp = *s++;
4732 SHop(OP_RIGHT_SHIFT);
4733 else if (tmp == '=')
4742 if (PL_expect == XOPERATOR) {
4743 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
4745 deprecate_old(commaless_variable_list);
4746 return REPORT(','); /* grandfather non-comma-format format */
4750 if (s[1] == '#' && (isIDFIRST_lazy_if(s+2,UTF) || strchr("{$:+-", s[2]))) {
4751 PL_tokenbuf[0] = '@';
4752 s = scan_ident(s + 1, PL_bufend, PL_tokenbuf + 1,
4753 sizeof PL_tokenbuf - 1, FALSE);
4754 if (PL_expect == XOPERATOR)
4755 no_op("Array length", s);
4756 if (!PL_tokenbuf[1])
4758 PL_expect = XOPERATOR;
4759 PL_pending_ident = '#';
4763 PL_tokenbuf[0] = '$';
4764 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1,
4765 sizeof PL_tokenbuf - 1, FALSE);
4766 if (PL_expect == XOPERATOR)
4768 if (!PL_tokenbuf[1]) {
4770 yyerror("Final $ should be \\$ or $name");
4774 /* This kludge not intended to be bulletproof. */
4775 if (PL_tokenbuf[1] == '[' && !PL_tokenbuf[2]) {
4776 yylval.opval = newSVOP(OP_CONST, 0,
4777 newSViv(CopARYBASE_get(&PL_compiling)));
4778 yylval.opval->op_private = OPpCONST_ARYBASE;
4784 const char tmp = *s;
4785 if (PL_lex_state == LEX_NORMAL)
4788 if ((PL_expect != XREF || PL_oldoldbufptr == PL_last_lop)
4789 && intuit_more(s)) {
4791 PL_tokenbuf[0] = '@';
4792 if (ckWARN(WARN_SYNTAX)) {
4795 while (isSPACE(*t) || isALNUM_lazy_if(t,UTF) || *t == '$')
4798 PL_bufptr = PEEKSPACE(PL_bufptr); /* XXX can realloc */
4799 while (t < PL_bufend && *t != ']')
4801 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
4802 "Multidimensional syntax %.*s not supported",
4803 (int)((t - PL_bufptr) + 1), PL_bufptr);
4807 else if (*s == '{') {
4809 PL_tokenbuf[0] = '%';
4810 if (strEQ(PL_tokenbuf+1, "SIG") && ckWARN(WARN_SYNTAX)
4811 && (t = strchr(s, '}')) && (t = strchr(t, '=')))
4813 char tmpbuf[sizeof PL_tokenbuf];
4816 } while (isSPACE(*t));
4817 if (isIDFIRST_lazy_if(t,UTF)) {
4819 t = scan_word(t, tmpbuf, sizeof tmpbuf, TRUE,
4823 if (*t == ';' && get_cv(tmpbuf, FALSE))
4824 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
4825 "You need to quote \"%s\"",
4832 PL_expect = XOPERATOR;
4833 if (PL_lex_state == LEX_NORMAL && isSPACE((char)tmp)) {
4834 const bool islop = (PL_last_lop == PL_oldoldbufptr);
4835 if (!islop || PL_last_lop_op == OP_GREPSTART)
4836 PL_expect = XOPERATOR;
4837 else if (strchr("$@\"'`q", *s))
4838 PL_expect = XTERM; /* e.g. print $fh "foo" */
4839 else if (strchr("&*<%", *s) && isIDFIRST_lazy_if(s+1,UTF))
4840 PL_expect = XTERM; /* e.g. print $fh &sub */
4841 else if (isIDFIRST_lazy_if(s,UTF)) {
4842 char tmpbuf[sizeof PL_tokenbuf];
4844 scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
4845 if ((t2 = keyword(tmpbuf, len, 0))) {
4846 /* binary operators exclude handle interpretations */
4858 PL_expect = XTERM; /* e.g. print $fh length() */
4863 PL_expect = XTERM; /* e.g. print $fh subr() */
4866 else if (isDIGIT(*s))
4867 PL_expect = XTERM; /* e.g. print $fh 3 */
4868 else if (*s == '.' && isDIGIT(s[1]))
4869 PL_expect = XTERM; /* e.g. print $fh .3 */
4870 else if ((*s == '?' || *s == '-' || *s == '+')
4871 && !isSPACE(s[1]) && s[1] != '=')
4872 PL_expect = XTERM; /* e.g. print $fh -1 */
4873 else if (*s == '/' && !isSPACE(s[1]) && s[1] != '='
4875 PL_expect = XTERM; /* e.g. print $fh /.../
4876 XXX except DORDOR operator
4878 else if (*s == '<' && s[1] == '<' && !isSPACE(s[2])
4880 PL_expect = XTERM; /* print $fh <<"EOF" */
4883 PL_pending_ident = '$';
4887 if (PL_expect == XOPERATOR)
4889 PL_tokenbuf[0] = '@';
4890 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1, FALSE);
4891 if (!PL_tokenbuf[1]) {
4894 if (PL_lex_state == LEX_NORMAL)
4896 if ((PL_expect != XREF || PL_oldoldbufptr == PL_last_lop) && intuit_more(s)) {
4898 PL_tokenbuf[0] = '%';
4900 /* Warn about @ where they meant $. */
4901 if (*s == '[' || *s == '{') {
4902 if (ckWARN(WARN_SYNTAX)) {
4903 const char *t = s + 1;
4904 while (*t && (isALNUM_lazy_if(t,UTF) || strchr(" \t$#+-'\"", *t)))
4906 if (*t == '}' || *t == ']') {
4908 PL_bufptr = PEEKSPACE(PL_bufptr); /* XXX can realloc */
4909 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
4910 "Scalar value %.*s better written as $%.*s",
4911 (int)(t-PL_bufptr), PL_bufptr,
4912 (int)(t-PL_bufptr-1), PL_bufptr+1);
4917 PL_pending_ident = '@';
4920 case '/': /* may be division, defined-or, or pattern */
4921 if (PL_expect == XTERMORDORDOR && s[1] == '/') {
4925 case '?': /* may either be conditional or pattern */
4926 if(PL_expect == XOPERATOR) {
4934 /* A // operator. */
4944 /* Disable warning on "study /blah/" */
4945 if (PL_oldoldbufptr == PL_last_uni
4946 && (*PL_last_uni != 's' || s - PL_last_uni < 5
4947 || memNE(PL_last_uni, "study", 5)
4948 || isALNUM_lazy_if(PL_last_uni+5,UTF)
4951 s = scan_pat(s,OP_MATCH);
4952 TERM(sublex_start());
4956 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack
4957 #ifdef PERL_STRICT_CR
4960 && (s[1] == '\n' || (s[1] == '\r' && s[2] == '\n'))
4962 && (s == PL_linestart || s[-1] == '\n') )
4964 PL_lex_formbrack = 0;
4968 if (PL_expect == XOPERATOR || !isDIGIT(s[1])) {
4974 yylval.ival = OPf_SPECIAL;
4980 if (PL_expect != XOPERATOR)
4985 case '0': case '1': case '2': case '3': case '4':
4986 case '5': case '6': case '7': case '8': case '9':
4987 s = scan_num(s, &yylval);
4988 DEBUG_T( { printbuf("### Saw number in %s\n", s); } );
4989 if (PL_expect == XOPERATOR)
4994 s = scan_str(s,!!PL_madskills,FALSE);
4995 DEBUG_T( { printbuf("### Saw string before %s\n", s); } );
4996 if (PL_expect == XOPERATOR) {
4997 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
4999 deprecate_old(commaless_variable_list);
5000 return REPORT(','); /* grandfather non-comma-format format */
5007 yylval.ival = OP_CONST;
5008 TERM(sublex_start());
5011 s = scan_str(s,!!PL_madskills,FALSE);
5012 DEBUG_T( { printbuf("### Saw string before %s\n", s); } );
5013 if (PL_expect == XOPERATOR) {
5014 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
5016 deprecate_old(commaless_variable_list);
5017 return REPORT(','); /* grandfather non-comma-format format */
5024 yylval.ival = OP_CONST;
5025 /* FIXME. I think that this can be const if char *d is replaced by
5026 more localised variables. */
5027 for (d = SvPV(PL_lex_stuff, len); len; len--, d++) {
5028 if (*d == '$' || *d == '@' || *d == '\\' || !UTF8_IS_INVARIANT((U8)*d)) {
5029 yylval.ival = OP_STRINGIFY;
5033 TERM(sublex_start());
5036 s = scan_str(s,!!PL_madskills,FALSE);
5037 DEBUG_T( { printbuf("### Saw backtick string before %s\n", s); } );
5038 if (PL_expect == XOPERATOR)
5039 no_op("Backticks",s);
5042 readpipe_override();
5043 TERM(sublex_start());
5047 if (PL_lex_inwhat && isDIGIT(*s) && ckWARN(WARN_SYNTAX))
5048 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),"Can't use \\%c to mean $%c in expression",
5050 if (PL_expect == XOPERATOR)
5051 no_op("Backslash",s);
5055 if (isDIGIT(s[1]) && PL_expect != XOPERATOR) {
5056 char *start = s + 2;
5057 while (isDIGIT(*start) || *start == '_')
5059 if (*start == '.' && isDIGIT(start[1])) {
5060 s = scan_num(s, &yylval);
5063 /* avoid v123abc() or $h{v1}, allow C<print v10;> */
5064 else if (!isALPHA(*start) && (PL_expect == XTERM
5065 || PL_expect == XREF || PL_expect == XSTATE
5066 || PL_expect == XTERMORDORDOR)) {
5067 /* XXX Use gv_fetchpvn rather than stomping on a const string */
5068 const char c = *start;
5071 gv = gv_fetchpv(s, 0, SVt_PVCV);
5074 s = scan_num(s, &yylval);
5081 if (isDIGIT(s[1]) && PL_expect == XOPERATOR) {
5123 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
5125 /* Some keywords can be followed by any delimiter, including ':' */
5126 tmp = ((len == 1 && strchr("msyq", PL_tokenbuf[0])) ||
5127 (len == 2 && ((PL_tokenbuf[0] == 't' && PL_tokenbuf[1] == 'r') ||
5128 (PL_tokenbuf[0] == 'q' &&
5129 strchr("qwxr", PL_tokenbuf[1])))));
5131 /* x::* is just a word, unless x is "CORE" */
5132 if (!tmp && *s == ':' && s[1] == ':' && strNE(PL_tokenbuf, "CORE"))
5136 while (d < PL_bufend && isSPACE(*d))
5137 d++; /* no comments skipped here, or s### is misparsed */
5139 /* Is this a label? */
5140 if (!tmp && PL_expect == XSTATE
5141 && d < PL_bufend && *d == ':' && *(d + 1) != ':') {
5143 yylval.pval = CopLABEL_alloc(PL_tokenbuf);
5148 /* Check for keywords */
5149 tmp = keyword(PL_tokenbuf, len, 0);
5151 /* Is this a word before a => operator? */
5152 if (*d == '=' && d[1] == '>') {
5155 = (OP*)newSVOP(OP_CONST, 0,
5156 S_newSV_maybe_utf8(aTHX_ PL_tokenbuf, len));
5157 yylval.opval->op_private = OPpCONST_BARE;
5161 if (tmp < 0) { /* second-class keyword? */
5162 GV *ogv = NULL; /* override (winner) */
5163 GV *hgv = NULL; /* hidden (loser) */
5164 if (PL_expect != XOPERATOR && (*s != ':' || s[1] != ':')) {
5166 if ((gv = gv_fetchpvn_flags(PL_tokenbuf, len, 0, SVt_PVCV)) &&
5169 if (GvIMPORTED_CV(gv))
5171 else if (! CvMETHOD(cv))
5175 (gvp = (GV**)hv_fetch(PL_globalstash,PL_tokenbuf,len,FALSE)) &&
5176 (gv = *gvp) != (GV*)&PL_sv_undef &&
5177 GvCVu(gv) && GvIMPORTED_CV(gv))
5184 tmp = 0; /* overridden by import or by GLOBAL */
5187 && -tmp==KEY_lock /* XXX generalizable kludge */
5189 && !hv_fetchs(GvHVn(PL_incgv), "Thread.pm", FALSE))
5191 tmp = 0; /* any sub overrides "weak" keyword */
5193 else { /* no override */
5195 if (tmp == KEY_dump && ckWARN(WARN_MISC)) {
5196 Perl_warner(aTHX_ packWARN(WARN_MISC),
5197 "dump() better written as CORE::dump()");
5201 if (hgv && tmp != KEY_x && tmp != KEY_CORE
5202 && ckWARN(WARN_AMBIGUOUS)) /* never ambiguous */
5203 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
5204 "Ambiguous call resolved as CORE::%s(), %s",
5205 GvENAME(hgv), "qualify as such or use &");
5212 default: /* not a keyword */
5213 /* Trade off - by using this evil construction we can pull the
5214 variable gv into the block labelled keylookup. If not, then
5215 we have to give it function scope so that the goto from the
5216 earlier ':' case doesn't bypass the initialisation. */
5218 just_a_word_zero_gv:
5226 const char lastchar = (PL_bufptr == PL_oldoldbufptr ? 0 : PL_bufptr[-1]);
5229 SV *nextPL_nextwhite = 0;
5233 /* Get the rest if it looks like a package qualifier */
5235 if (*s == '\'' || (*s == ':' && s[1] == ':')) {
5237 s = scan_word(s, PL_tokenbuf + len, sizeof PL_tokenbuf - len,
5240 Perl_croak(aTHX_ "Bad name after %s%s", PL_tokenbuf,
5241 *s == '\'' ? "'" : "::");
5246 if (PL_expect == XOPERATOR) {
5247 if (PL_bufptr == PL_linestart) {
5248 CopLINE_dec(PL_curcop);
5249 Perl_warner(aTHX_ packWARN(WARN_SEMICOLON), PL_warn_nosemi);
5250 CopLINE_inc(PL_curcop);
5253 no_op("Bareword",s);
5256 /* Look for a subroutine with this name in current package,
5257 unless name is "Foo::", in which case Foo is a bearword
5258 (and a package name). */
5260 if (len > 2 && !PL_madskills &&
5261 PL_tokenbuf[len - 2] == ':' && PL_tokenbuf[len - 1] == ':')
5263 if (ckWARN(WARN_BAREWORD)
5264 && ! gv_fetchpvn_flags(PL_tokenbuf, len, 0, SVt_PVHV))
5265 Perl_warner(aTHX_ packWARN(WARN_BAREWORD),
5266 "Bareword \"%s\" refers to nonexistent package",
5269 PL_tokenbuf[len] = '\0';
5275 /* Mustn't actually add anything to a symbol table.
5276 But also don't want to "initialise" any placeholder
5277 constants that might already be there into full
5278 blown PVGVs with attached PVCV. */
5279 gv = gv_fetchpvn_flags(PL_tokenbuf, len,
5280 GV_NOADD_NOINIT, SVt_PVCV);
5285 /* if we saw a global override before, get the right name */
5288 sv = newSVpvs("CORE::GLOBAL::");
5289 sv_catpv(sv,PL_tokenbuf);
5292 /* If len is 0, newSVpv does strlen(), which is correct.
5293 If len is non-zero, then it will be the true length,
5294 and so the scalar will be created correctly. */
5295 sv = newSVpv(PL_tokenbuf,len);
5298 if (PL_madskills && !PL_thistoken) {
5299 char *start = SvPVX(PL_linestr) + PL_realtokenstart;
5300 PL_thistoken = newSVpv(start,s - start);
5301 PL_realtokenstart = s - SvPVX(PL_linestr);
5305 /* Presume this is going to be a bareword of some sort. */
5308 yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
5309 yylval.opval->op_private = OPpCONST_BARE;
5310 /* UTF-8 package name? */
5311 if (UTF && !IN_BYTES &&
5312 is_utf8_string((U8*)SvPVX_const(sv), SvCUR(sv)))
5315 /* And if "Foo::", then that's what it certainly is. */
5320 /* Do the explicit type check so that we don't need to force
5321 the initialisation of the symbol table to have a real GV.
5322 Beware - gv may not really be a PVGV, cv may not really be
5323 a PVCV, (because of the space optimisations that gv_init
5324 understands) But they're true if for this symbol there is
5325 respectively a typeglob and a subroutine.
5327 cv = gv ? ((SvTYPE(gv) == SVt_PVGV)
5328 /* Real typeglob, so get the real subroutine: */
5330 /* A proxy for a subroutine in this package? */
5331 : SvOK(gv) ? (CV *) gv : NULL)
5334 /* See if it's the indirect object for a list operator. */
5336 if (PL_oldoldbufptr &&
5337 PL_oldoldbufptr < PL_bufptr &&
5338 (PL_oldoldbufptr == PL_last_lop
5339 || PL_oldoldbufptr == PL_last_uni) &&
5340 /* NO SKIPSPACE BEFORE HERE! */
5341 (PL_expect == XREF ||
5342 ((PL_opargs[PL_last_lop_op] >> OASHIFT)& 7) == OA_FILEREF))
5344 bool immediate_paren = *s == '(';
5346 /* (Now we can afford to cross potential line boundary.) */
5347 s = SKIPSPACE2(s,nextPL_nextwhite);
5349 PL_nextwhite = nextPL_nextwhite; /* assume no & deception */
5352 /* Two barewords in a row may indicate method call. */
5354 if ((isIDFIRST_lazy_if(s,UTF) || *s == '$') &&
5355 (tmp = intuit_method(s, gv, cv)))
5358 /* If not a declared subroutine, it's an indirect object. */
5359 /* (But it's an indir obj regardless for sort.) */
5360 /* Also, if "_" follows a filetest operator, it's a bareword */
5363 ( !immediate_paren && (PL_last_lop_op == OP_SORT ||
5365 (PL_last_lop_op != OP_MAPSTART &&
5366 PL_last_lop_op != OP_GREPSTART))))
5367 || (PL_tokenbuf[0] == '_' && PL_tokenbuf[1] == '\0'
5368 && ((PL_opargs[PL_last_lop_op] & OA_CLASS_MASK) == OA_FILESTATOP))
5371 PL_expect = (PL_last_lop == PL_oldoldbufptr) ? XTERM : XOPERATOR;
5376 PL_expect = XOPERATOR;
5379 s = SKIPSPACE2(s,nextPL_nextwhite);
5380 PL_nextwhite = nextPL_nextwhite;
5385 /* Is this a word before a => operator? */
5386 if (*s == '=' && s[1] == '>' && !pkgname) {
5388 sv_setpv(((SVOP*)yylval.opval)->op_sv, PL_tokenbuf);
5389 if (UTF && !IN_BYTES && is_utf8_string((U8*)PL_tokenbuf, len))
5390 SvUTF8_on(((SVOP*)yylval.opval)->op_sv);
5394 /* If followed by a paren, it's certainly a subroutine. */
5399 while (SPACE_OR_TAB(*d))
5401 if (*d == ')' && (sv = gv_const_sv(gv))) {
5405 char *par = SvPVX(PL_linestr) + PL_realtokenstart;
5406 sv_catpvn(PL_thistoken, par, s - par);
5408 sv_free(PL_nextwhite);
5418 PL_nextwhite = PL_thiswhite;
5421 start_force(PL_curforce);
5423 NEXTVAL_NEXTTOKE.opval = yylval.opval;
5424 PL_expect = XOPERATOR;
5427 PL_nextwhite = nextPL_nextwhite;
5428 curmad('X', PL_thistoken);
5429 PL_thistoken = newSVpvs("");
5437 /* If followed by var or block, call it a method (unless sub) */
5439 if ((*s == '$' || *s == '{') && (!gv || !cv)) {
5440 PL_last_lop = PL_oldbufptr;
5441 PL_last_lop_op = OP_METHOD;
5445 /* If followed by a bareword, see if it looks like indir obj. */
5448 && (isIDFIRST_lazy_if(s,UTF) || *s == '$')
5449 && (tmp = intuit_method(s, gv, cv)))
5452 /* Not a method, so call it a subroutine (if defined) */
5455 if (lastchar == '-' && ckWARN_d(WARN_AMBIGUOUS))
5456 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
5457 "Ambiguous use of -%s resolved as -&%s()",
5458 PL_tokenbuf, PL_tokenbuf);
5459 /* Check for a constant sub */
5460 if ((sv = gv_const_sv(gv))) {
5462 SvREFCNT_dec(((SVOP*)yylval.opval)->op_sv);
5463 ((SVOP*)yylval.opval)->op_sv = SvREFCNT_inc_simple(sv);
5464 yylval.opval->op_private = 0;
5468 /* Resolve to GV now. */
5469 if (SvTYPE(gv) != SVt_PVGV) {
5470 gv = gv_fetchpv(PL_tokenbuf, 0, SVt_PVCV);
5471 assert (SvTYPE(gv) == SVt_PVGV);
5472 /* cv must have been some sort of placeholder, so
5473 now needs replacing with a real code reference. */
5477 op_free(yylval.opval);
5478 yylval.opval = newCVREF(0, newGVOP(OP_GV, 0, gv));
5479 yylval.opval->op_private |= OPpENTERSUB_NOPAREN;
5480 PL_last_lop = PL_oldbufptr;
5481 PL_last_lop_op = OP_ENTERSUB;
5482 /* Is there a prototype? */
5490 const char *proto = SvPV_const((SV*)cv, protolen);
5493 if ((*proto == '$' || *proto == '_') && proto[1] == '\0')
5495 while (*proto == ';')
5497 if (*proto == '&' && *s == '{') {
5498 sv_setpv(PL_subname,
5501 "__ANON__" : "__ANON__::__ANON__"));
5508 PL_nextwhite = PL_thiswhite;
5511 start_force(PL_curforce);
5512 NEXTVAL_NEXTTOKE.opval = yylval.opval;
5515 PL_nextwhite = nextPL_nextwhite;
5516 curmad('X', PL_thistoken);
5517 PL_thistoken = newSVpvs("");
5524 /* Guess harder when madskills require "best effort". */
5525 if (PL_madskills && (!gv || !GvCVu(gv))) {
5526 int probable_sub = 0;
5527 if (strchr("\"'`$@%0123456789!*+{[<", *s))
5529 else if (isALPHA(*s)) {
5533 d = scan_word(d, tmpbuf, sizeof tmpbuf, TRUE, &tmplen);
5534 if (!keyword(tmpbuf, tmplen, 0))
5537 while (d < PL_bufend && isSPACE(*d))
5539 if (*d == '=' && d[1] == '>')
5544 gv = gv_fetchpv(PL_tokenbuf, TRUE, SVt_PVCV);
5545 op_free(yylval.opval);
5546 yylval.opval = newCVREF(0, newGVOP(OP_GV, 0, gv));
5547 yylval.opval->op_private |= OPpENTERSUB_NOPAREN;
5548 PL_last_lop = PL_oldbufptr;
5549 PL_last_lop_op = OP_ENTERSUB;
5550 PL_nextwhite = PL_thiswhite;
5552 start_force(PL_curforce);
5553 NEXTVAL_NEXTTOKE.opval = yylval.opval;
5555 PL_nextwhite = nextPL_nextwhite;
5556 curmad('X', PL_thistoken);
5557 PL_thistoken = newSVpvs("");
5562 NEXTVAL_NEXTTOKE.opval = yylval.opval;
5569 /* Call it a bare word */
5571 if (PL_hints & HINT_STRICT_SUBS)
5572 yylval.opval->op_private |= OPpCONST_STRICT;
5575 if (lastchar != '-') {
5576 if (ckWARN(WARN_RESERVED)) {
5580 if (!*d && !gv_stashpv(PL_tokenbuf,FALSE))
5581 Perl_warner(aTHX_ packWARN(WARN_RESERVED), PL_warn_reserved,
5588 if ((lastchar == '*' || lastchar == '%' || lastchar == '&')
5589 && ckWARN_d(WARN_AMBIGUOUS)) {
5590 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
5591 "Operator or semicolon missing before %c%s",
5592 lastchar, PL_tokenbuf);
5593 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
5594 "Ambiguous use of %c resolved as operator %c",
5595 lastchar, lastchar);
5601 yylval.opval = (OP*)newSVOP(OP_CONST, 0,
5602 newSVpv(CopFILE(PL_curcop),0));
5606 yylval.opval = (OP*)newSVOP(OP_CONST, 0,
5607 Perl_newSVpvf(aTHX_ "%"IVdf, (IV)CopLINE(PL_curcop)));
5610 case KEY___PACKAGE__:
5611 yylval.opval = (OP*)newSVOP(OP_CONST, 0,
5613 ? newSVhek(HvNAME_HEK(PL_curstash))
5620 if (PL_rsfp && (!PL_in_eval || PL_tokenbuf[2] == 'D')) {
5621 const char *pname = "main";
5622 if (PL_tokenbuf[2] == 'D')
5623 pname = HvNAME_get(PL_curstash ? PL_curstash : PL_defstash);
5624 gv = gv_fetchpv(Perl_form(aTHX_ "%s::DATA", pname), GV_ADD,
5628 GvIOp(gv) = newIO();
5629 IoIFP(GvIOp(gv)) = PL_rsfp;
5630 #if defined(HAS_FCNTL) && defined(F_SETFD)
5632 const int fd = PerlIO_fileno(PL_rsfp);
5633 fcntl(fd,F_SETFD,fd >= 3);
5636 /* Mark this internal pseudo-handle as clean */
5637 IoFLAGS(GvIOp(gv)) |= IOf_UNTAINT;
5639 IoTYPE(GvIOp(gv)) = IoTYPE_PIPE;
5640 else if ((PerlIO*)PL_rsfp == PerlIO_stdin())
5641 IoTYPE(GvIOp(gv)) = IoTYPE_STD;
5643 IoTYPE(GvIOp(gv)) = IoTYPE_RDONLY;
5644 #if defined(WIN32) && !defined(PERL_TEXTMODE_SCRIPTS)
5645 /* if the script was opened in binmode, we need to revert
5646 * it to text mode for compatibility; but only iff it has CRs
5647 * XXX this is a questionable hack at best. */
5648 if (PL_bufend-PL_bufptr > 2
5649 && PL_bufend[-1] == '\n' && PL_bufend[-2] == '\r')
5652 if (IoTYPE(GvIOp(gv)) == IoTYPE_RDONLY) {
5653 loc = PerlIO_tell(PL_rsfp);
5654 (void)PerlIO_seek(PL_rsfp, 0L, 0);
5657 if (PerlLIO_setmode(PL_rsfp, O_TEXT) != -1) {
5659 if (PerlLIO_setmode(PerlIO_fileno(PL_rsfp), O_TEXT) != -1) {
5660 #endif /* NETWARE */
5661 #ifdef PERLIO_IS_STDIO /* really? */
5662 # if defined(__BORLANDC__)
5663 /* XXX see note in do_binmode() */
5664 ((FILE*)PL_rsfp)->flags &= ~_F_BIN;
5668 PerlIO_seek(PL_rsfp, loc, 0);
5672 #ifdef PERLIO_LAYERS
5675 PerlIO_apply_layers(aTHX_ PL_rsfp, NULL, ":utf8");
5676 else if (PL_encoding) {
5683 XPUSHs(PL_encoding);
5685 call_method("name", G_SCALAR);
5689 PerlIO_apply_layers(aTHX_ PL_rsfp, NULL,
5690 Perl_form(aTHX_ ":encoding(%"SVf")",
5699 if (PL_realtokenstart >= 0) {
5700 char *tstart = SvPVX(PL_linestr) + PL_realtokenstart;
5702 PL_endwhite = newSVpvs("");
5703 sv_catsv(PL_endwhite, PL_thiswhite);
5705 sv_catpvn(PL_endwhite, tstart, PL_bufend - tstart);
5706 PL_realtokenstart = -1;
5708 while ((s = filter_gets(PL_endwhite, PL_rsfp,
5709 SvCUR(PL_endwhite))) != Nullch) ;
5724 if (PL_expect == XSTATE) {
5731 if (*s == ':' && s[1] == ':') {
5734 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
5735 if (!(tmp = keyword(PL_tokenbuf, len, 0)))
5736 Perl_croak(aTHX_ "CORE::%s is not a keyword", PL_tokenbuf);
5739 else if (tmp == KEY_require || tmp == KEY_do)
5740 /* that's a way to remember we saw "CORE::" */
5753 LOP(OP_ACCEPT,XTERM);
5759 LOP(OP_ATAN2,XTERM);
5765 LOP(OP_BINMODE,XTERM);
5768 LOP(OP_BLESS,XTERM);
5777 /* When 'use switch' is in effect, continue has a dual
5778 life as a control operator. */
5780 if (!FEATURE_IS_ENABLED("switch"))
5783 /* We have to disambiguate the two senses of
5784 "continue". If the next token is a '{' then
5785 treat it as the start of a continue block;
5786 otherwise treat it as a control operator.
5798 (void)gv_fetchpvs("ENV", GV_ADD|GV_NOTQUAL, SVt_PVHV);
5815 if (!PL_cryptseen) {
5816 PL_cryptseen = TRUE;
5820 LOP(OP_CRYPT,XTERM);
5823 LOP(OP_CHMOD,XTERM);
5826 LOP(OP_CHOWN,XTERM);
5829 LOP(OP_CONNECT,XTERM);
5848 s = force_word(s,WORD,TRUE,TRUE,FALSE);
5849 if (orig_keyword == KEY_do) {
5858 PL_hints |= HINT_BLOCK_SCOPE;
5868 gv_fetchpvs("AnyDBM_File::ISA", GV_ADDMULTI, SVt_PVAV);
5869 LOP(OP_DBMOPEN,XTERM);
5875 s = force_word(s,WORD,TRUE,FALSE,FALSE);
5882 yylval.ival = CopLINE(PL_curcop);
5898 PL_expect = (*s == '{') ? XTERMBLOCK : XTERM;
5899 UNIBRACK(OP_ENTEREVAL);
5917 case KEY_endhostent:
5923 case KEY_endservent:
5926 case KEY_endprotoent:
5937 yylval.ival = CopLINE(PL_curcop);
5939 if (PL_expect == XSTATE && isIDFIRST_lazy_if(s,UTF)) {
5942 int soff = s - SvPVX(PL_linestr); /* for skipspace realloc */
5945 if ((PL_bufend - p) >= 3 &&
5946 strnEQ(p, "my", 2) && isSPACE(*(p + 2)))
5948 else if ((PL_bufend - p) >= 4 &&
5949 strnEQ(p, "our", 3) && isSPACE(*(p + 3)))
5952 if (isIDFIRST_lazy_if(p,UTF)) {
5953 p = scan_ident(p, PL_bufend,
5954 PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
5958 Perl_croak(aTHX_ "Missing $ on loop variable");
5960 s = SvPVX(PL_linestr) + soff;
5966 LOP(OP_FORMLINE,XTERM);
5972 LOP(OP_FCNTL,XTERM);
5978 LOP(OP_FLOCK,XTERM);
5987 LOP(OP_GREPSTART, XREF);
5990 s = force_word(s,WORD,TRUE,FALSE,FALSE);
6005 case KEY_getpriority:
6006 LOP(OP_GETPRIORITY,XTERM);
6008 case KEY_getprotobyname:
6011 case KEY_getprotobynumber:
6012 LOP(OP_GPBYNUMBER,XTERM);
6014 case KEY_getprotoent:
6026 case KEY_getpeername:
6027 UNI(OP_GETPEERNAME);
6029 case KEY_gethostbyname:
6032 case KEY_gethostbyaddr:
6033 LOP(OP_GHBYADDR,XTERM);
6035 case KEY_gethostent:
6038 case KEY_getnetbyname:
6041 case KEY_getnetbyaddr:
6042 LOP(OP_GNBYADDR,XTERM);
6047 case KEY_getservbyname:
6048 LOP(OP_GSBYNAME,XTERM);
6050 case KEY_getservbyport:
6051 LOP(OP_GSBYPORT,XTERM);
6053 case KEY_getservent:
6056 case KEY_getsockname:
6057 UNI(OP_GETSOCKNAME);
6059 case KEY_getsockopt:
6060 LOP(OP_GSOCKOPT,XTERM);
6075 yylval.ival = CopLINE(PL_curcop);
6086 yylval.ival = CopLINE(PL_curcop);
6090 LOP(OP_INDEX,XTERM);
6096 LOP(OP_IOCTL,XTERM);
6108 s = force_word(s,WORD,TRUE,FALSE,FALSE);
6140 LOP(OP_LISTEN,XTERM);
6149 s = scan_pat(s,OP_MATCH);
6150 TERM(sublex_start());
6153 LOP(OP_MAPSTART, XREF);
6156 LOP(OP_MKDIR,XTERM);
6159 LOP(OP_MSGCTL,XTERM);
6162 LOP(OP_MSGGET,XTERM);
6165 LOP(OP_MSGRCV,XTERM);
6168 LOP(OP_MSGSND,XTERM);
6175 if (isIDFIRST_lazy_if(s,UTF)) {
6179 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, TRUE, &len);
6180 if (len == 3 && strnEQ(PL_tokenbuf, "sub", 3))
6182 PL_in_my_stash = find_in_my_stash(PL_tokenbuf, len);
6183 if (!PL_in_my_stash) {
6186 my_snprintf(tmpbuf, sizeof(tmpbuf), "No such class %.1000s", PL_tokenbuf);
6190 if (PL_madskills) { /* just add type to declarator token */
6191 sv_catsv(PL_thistoken, PL_nextwhite);
6193 sv_catpvn(PL_thistoken, start, s - start);
6201 s = force_word(s,WORD,TRUE,FALSE,FALSE);
6208 s = tokenize_use(0, s);
6212 if (*s == '(' || (s = SKIPSPACE1(s), *s == '('))
6219 if (isIDFIRST_lazy_if(s,UTF)) {
6221 for (d = s; isALNUM_lazy_if(d,UTF);)
6223 for (t=d; isSPACE(*t);)
6225 if ( *t && strchr("|&*+-=!?:.", *t) && ckWARN_d(WARN_PRECEDENCE)
6227 && !(t[0] == '=' && t[1] == '>')
6229 int parms_len = (int)(d-s);
6230 Perl_warner(aTHX_ packWARN(WARN_PRECEDENCE),
6231 "Precedence problem: open %.*s should be open(%.*s)",
6232 parms_len, s, parms_len, s);
6238 yylval.ival = OP_OR;
6248 LOP(OP_OPEN_DIR,XTERM);
6251 checkcomma(s,PL_tokenbuf,"filehandle");
6255 checkcomma(s,PL_tokenbuf,"filehandle");
6274 s = force_word(s,WORD,FALSE,TRUE,FALSE);
6278 LOP(OP_PIPE_OP,XTERM);
6281 s = scan_str(s,!!PL_madskills,FALSE);
6284 yylval.ival = OP_CONST;
6285 TERM(sublex_start());
6291 s = scan_str(s,!!PL_madskills,FALSE);
6294 PL_expect = XOPERATOR;
6296 if (SvCUR(PL_lex_stuff)) {
6299 d = SvPV_force(PL_lex_stuff, len);
6301 for (; isSPACE(*d) && len; --len, ++d)
6306 if (!warned && ckWARN(WARN_QW)) {
6307 for (; !isSPACE(*d) && len; --len, ++d) {
6309 Perl_warner(aTHX_ packWARN(WARN_QW),
6310 "Possible attempt to separate words with commas");
6313 else if (*d == '#') {
6314 Perl_warner(aTHX_ packWARN(WARN_QW),
6315 "Possible attempt to put comments in qw() list");
6321 for (; !isSPACE(*d) && len; --len, ++d)
6324 sv = newSVpvn(b, d-b);
6325 if (DO_UTF8(PL_lex_stuff))
6327 words = append_elem(OP_LIST, words,
6328 newSVOP(OP_CONST, 0, tokeq(sv)));
6332 start_force(PL_curforce);
6333 NEXTVAL_NEXTTOKE.opval = words;
6338 SvREFCNT_dec(PL_lex_stuff);
6339 PL_lex_stuff = NULL;
6345 s = scan_str(s,!!PL_madskills,FALSE);
6348 yylval.ival = OP_STRINGIFY;
6349 if (SvIVX(PL_lex_stuff) == '\'')
6350 SvIV_set(PL_lex_stuff, 0); /* qq'$foo' should intepolate */
6351 TERM(sublex_start());
6354 s = scan_pat(s,OP_QR);
6355 TERM(sublex_start());
6358 s = scan_str(s,!!PL_madskills,FALSE);
6361 readpipe_override();
6362 TERM(sublex_start());
6370 s = force_version(s, FALSE);
6372 else if (*s != 'v' || !isDIGIT(s[1])
6373 || (s = force_version(s, TRUE), *s == 'v'))
6375 *PL_tokenbuf = '\0';
6376 s = force_word(s,WORD,TRUE,TRUE,FALSE);
6377 if (isIDFIRST_lazy_if(PL_tokenbuf,UTF))
6378 gv_stashpvn(PL_tokenbuf, strlen(PL_tokenbuf), TRUE);
6380 yyerror("<> should be quotes");
6382 if (orig_keyword == KEY_require) {
6390 PL_last_uni = PL_oldbufptr;
6391 PL_last_lop_op = OP_REQUIRE;
6393 return REPORT( (int)REQUIRE );
6399 s = force_word(s,WORD,TRUE,FALSE,FALSE);
6403 LOP(OP_RENAME,XTERM);
6412 LOP(OP_RINDEX,XTERM);
6422 UNIDOR(OP_READLINE);
6435 LOP(OP_REVERSE,XTERM);
6438 UNIDOR(OP_READLINK);
6446 TERM(sublex_start());
6448 TOKEN(1); /* force error */
6451 checkcomma(s,PL_tokenbuf,"filehandle");
6461 LOP(OP_SELECT,XTERM);
6467 LOP(OP_SEMCTL,XTERM);
6470 LOP(OP_SEMGET,XTERM);
6473 LOP(OP_SEMOP,XTERM);
6479 LOP(OP_SETPGRP,XTERM);
6481 case KEY_setpriority:
6482 LOP(OP_SETPRIORITY,XTERM);
6484 case KEY_sethostent:
6490 case KEY_setservent:
6493 case KEY_setprotoent:
6503 LOP(OP_SEEKDIR,XTERM);
6505 case KEY_setsockopt:
6506 LOP(OP_SSOCKOPT,XTERM);
6512 LOP(OP_SHMCTL,XTERM);
6515 LOP(OP_SHMGET,XTERM);
6518 LOP(OP_SHMREAD,XTERM);
6521 LOP(OP_SHMWRITE,XTERM);
6524 LOP(OP_SHUTDOWN,XTERM);
6533 LOP(OP_SOCKET,XTERM);
6535 case KEY_socketpair:
6536 LOP(OP_SOCKPAIR,XTERM);
6539 checkcomma(s,PL_tokenbuf,"subroutine name");
6541 if (*s == ';' || *s == ')') /* probably a close */
6542 Perl_croak(aTHX_ "sort is now a reserved word");
6544 s = force_word(s,WORD,TRUE,TRUE,FALSE);
6548 LOP(OP_SPLIT,XTERM);
6551 LOP(OP_SPRINTF,XTERM);
6554 LOP(OP_SPLICE,XTERM);
6569 LOP(OP_SUBSTR,XTERM);
6575 char tmpbuf[sizeof PL_tokenbuf];
6576 SSize_t tboffset = 0;
6577 expectation attrful;
6578 bool have_name, have_proto;
6579 const int key = tmp;
6584 char *tstart = SvPVX(PL_linestr) + PL_realtokenstart;
6585 SV *subtoken = newSVpvn(tstart, s - tstart);
6589 s = SKIPSPACE2(s,tmpwhite);
6594 if (isIDFIRST_lazy_if(s,UTF) || *s == '\'' ||
6595 (*s == ':' && s[1] == ':'))
6602 attrful = XATTRBLOCK;
6603 /* remember buffer pos'n for later force_word */
6604 tboffset = s - PL_oldbufptr;
6605 d = scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
6608 nametoke = newSVpvn(s, d - s);
6610 if (strchr(tmpbuf, ':'))
6611 sv_setpv(PL_subname, tmpbuf);
6613 sv_setsv(PL_subname,PL_curstname);
6614 sv_catpvs(PL_subname,"::");
6615 sv_catpvn(PL_subname,tmpbuf,len);
6622 CURMAD('X', nametoke);
6623 CURMAD('_', tmpwhite);
6624 (void) force_word(PL_oldbufptr + tboffset, WORD,
6627 s = SKIPSPACE2(d,tmpwhite);
6634 Perl_croak(aTHX_ "Missing name in \"my sub\"");
6635 PL_expect = XTERMBLOCK;
6636 attrful = XATTRTERM;
6637 sv_setpvn(PL_subname,"?",1);
6641 if (key == KEY_format) {
6643 PL_lex_formbrack = PL_lex_brackets + 1;
6645 PL_thistoken = subtoken;
6649 (void) force_word(PL_oldbufptr + tboffset, WORD,
6655 /* Look for a prototype */
6658 bool bad_proto = FALSE;
6659 const bool warnsyntax = ckWARN(WARN_SYNTAX);
6661 s = scan_str(s,!!PL_madskills,FALSE);
6663 Perl_croak(aTHX_ "Prototype not terminated");
6664 /* strip spaces and check for bad characters */
6665 d = SvPVX(PL_lex_stuff);
6667 for (p = d; *p; ++p) {
6670 if (warnsyntax && !strchr("$@%*;[]&\\_", *p))
6676 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
6677 "Illegal character in prototype for %"SVf" : %s",
6678 (void*)PL_subname, d);
6679 SvCUR_set(PL_lex_stuff, tmp);
6684 CURMAD('q', PL_thisopen);
6685 CURMAD('_', tmpwhite);
6686 CURMAD('=', PL_thisstuff);
6687 CURMAD('Q', PL_thisclose);
6688 NEXTVAL_NEXTTOKE.opval =
6689 (OP*)newSVOP(OP_CONST, 0, PL_lex_stuff);
6690 PL_lex_stuff = Nullsv;
6693 s = SKIPSPACE2(s,tmpwhite);
6701 if (*s == ':' && s[1] != ':')
6702 PL_expect = attrful;
6703 else if (*s != '{' && key == KEY_sub) {
6705 Perl_croak(aTHX_ "Illegal declaration of anonymous subroutine");
6707 Perl_croak(aTHX_ "Illegal declaration of subroutine %"SVf, (void*)PL_subname);
6714 curmad('^', newSVpvs(""));
6715 CURMAD('_', tmpwhite);
6719 PL_thistoken = subtoken;
6722 NEXTVAL_NEXTTOKE.opval =
6723 (OP*)newSVOP(OP_CONST, 0, PL_lex_stuff);
6724 PL_lex_stuff = NULL;
6729 sv_setpv(PL_subname,
6731 (PL_curstash ? "__ANON__" : "__ANON__::__ANON__"));
6735 (void) force_word(PL_oldbufptr + tboffset, WORD,
6745 LOP(OP_SYSTEM,XREF);
6748 LOP(OP_SYMLINK,XTERM);
6751 LOP(OP_SYSCALL,XTERM);
6754 LOP(OP_SYSOPEN,XTERM);
6757 LOP(OP_SYSSEEK,XTERM);
6760 LOP(OP_SYSREAD,XTERM);
6763 LOP(OP_SYSWRITE,XTERM);
6767 TERM(sublex_start());
6788 LOP(OP_TRUNCATE,XTERM);
6800 yylval.ival = CopLINE(PL_curcop);
6804 yylval.ival = CopLINE(PL_curcop);
6808 LOP(OP_UNLINK,XTERM);
6814 LOP(OP_UNPACK,XTERM);
6817 LOP(OP_UTIME,XTERM);
6823 LOP(OP_UNSHIFT,XTERM);
6826 s = tokenize_use(1, s);
6836 yylval.ival = CopLINE(PL_curcop);
6840 yylval.ival = CopLINE(PL_curcop);
6844 PL_hints |= HINT_BLOCK_SCOPE;
6851 LOP(OP_WAITPID,XTERM);
6860 ctl_l[0] = toCTRL('L');
6862 gv_fetchpvn_flags(ctl_l, 1, GV_ADD|GV_NOTQUAL, SVt_PV);
6865 /* Make sure $^L is defined */
6866 gv_fetchpvs("\f", GV_ADD|GV_NOTQUAL, SVt_PV);
6871 if (PL_expect == XOPERATOR)
6877 yylval.ival = OP_XOR;
6882 TERM(sublex_start());
6887 #pragma segment Main
6891 S_pending_ident(pTHX)
6896 /* pit holds the identifier we read and pending_ident is reset */
6897 char pit = PL_pending_ident;
6898 PL_pending_ident = 0;
6900 /* PL_realtokenstart = realtokenend = PL_bufptr - SvPVX(PL_linestr); */
6901 DEBUG_T({ PerlIO_printf(Perl_debug_log,
6902 "### Pending identifier '%s'\n", PL_tokenbuf); });
6904 /* if we're in a my(), we can't allow dynamics here.
6905 $foo'bar has already been turned into $foo::bar, so
6906 just check for colons.
6908 if it's a legal name, the OP is a PADANY.
6911 if (PL_in_my == KEY_our) { /* "our" is merely analogous to "my" */
6912 if (strchr(PL_tokenbuf,':'))
6913 yyerror(Perl_form(aTHX_ "No package name allowed for "
6914 "variable %s in \"our\"",
6916 tmp = allocmy(PL_tokenbuf);
6919 if (strchr(PL_tokenbuf,':'))
6920 yyerror(Perl_form(aTHX_ PL_no_myglob,
6921 PL_in_my == KEY_my ? "my" : "state", PL_tokenbuf));
6923 yylval.opval = newOP(OP_PADANY, 0);
6924 yylval.opval->op_targ = allocmy(PL_tokenbuf);
6930 build the ops for accesses to a my() variable.
6932 Deny my($a) or my($b) in a sort block, *if* $a or $b is
6933 then used in a comparison. This catches most, but not
6934 all cases. For instance, it catches
6935 sort { my($a); $a <=> $b }
6937 sort { my($a); $a < $b ? -1 : $a == $b ? 0 : 1; }
6938 (although why you'd do that is anyone's guess).
6941 if (!strchr(PL_tokenbuf,':')) {
6943 tmp = pad_findmy(PL_tokenbuf);
6944 if (tmp != NOT_IN_PAD) {
6945 /* might be an "our" variable" */
6946 if (PAD_COMPNAME_FLAGS_isOUR(tmp)) {
6947 /* build ops for a bareword */
6948 HV * const stash = PAD_COMPNAME_OURSTASH(tmp);
6949 HEK * const stashname = HvNAME_HEK(stash);
6950 SV * const sym = newSVhek(stashname);
6951 sv_catpvs(sym, "::");
6952 sv_catpv(sym, PL_tokenbuf+1);
6953 yylval.opval = (OP*)newSVOP(OP_CONST, 0, sym);
6954 yylval.opval->op_private = OPpCONST_ENTERED;
6957 ? (GV_ADDMULTI | GV_ADDINEVAL)
6960 ((PL_tokenbuf[0] == '$') ? SVt_PV
6961 : (PL_tokenbuf[0] == '@') ? SVt_PVAV
6966 /* if it's a sort block and they're naming $a or $b */
6967 if (PL_last_lop_op == OP_SORT &&
6968 PL_tokenbuf[0] == '$' &&
6969 (PL_tokenbuf[1] == 'a' || PL_tokenbuf[1] == 'b')
6972 for (d = PL_in_eval ? PL_oldoldbufptr : PL_linestart;
6973 d < PL_bufend && *d != '\n';
6976 if (strnEQ(d,"<=>",3) || strnEQ(d,"cmp",3)) {
6977 Perl_croak(aTHX_ "Can't use \"my %s\" in sort comparison",
6983 yylval.opval = newOP(OP_PADANY, 0);
6984 yylval.opval->op_targ = tmp;
6990 Whine if they've said @foo in a doublequoted string,
6991 and @foo isn't a variable we can find in the symbol
6994 if (pit == '@' && PL_lex_state != LEX_NORMAL && !PL_lex_brackets) {
6995 GV *gv = gv_fetchpv(PL_tokenbuf+1, 0, SVt_PVAV);
6996 if ((!gv || ((PL_tokenbuf[0] == '@') ? !GvAV(gv) : !GvHV(gv)))
6997 && ckWARN(WARN_AMBIGUOUS))
6999 /* Downgraded from fatal to warning 20000522 mjd */
7000 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
7001 "Possible unintended interpolation of %s in string",
7006 /* build ops for a bareword */
7007 yylval.opval = (OP*)newSVOP(OP_CONST, 0, newSVpv(PL_tokenbuf+1, 0));
7008 yylval.opval->op_private = OPpCONST_ENTERED;
7011 /* If the identifier refers to a stash, don't autovivify it.
7012 * Change 24660 had the side effect of causing symbol table
7013 * hashes to always be defined, even if they were freshly
7014 * created and the only reference in the entire program was
7015 * the single statement with the defined %foo::bar:: test.
7016 * It appears that all code in the wild doing this actually
7017 * wants to know whether sub-packages have been loaded, so
7018 * by avoiding auto-vivifying symbol tables, we ensure that
7019 * defined %foo::bar:: continues to be false, and the existing
7020 * tests still give the expected answers, even though what
7021 * they're actually testing has now changed subtly.
7023 (*PL_tokenbuf == '%' && *(d = PL_tokenbuf + strlen(PL_tokenbuf) - 1) == ':' && d[-1] == ':'
7025 : PL_in_eval ? (GV_ADDMULTI | GV_ADDINEVAL) : GV_ADD),
7026 ((PL_tokenbuf[0] == '$') ? SVt_PV
7027 : (PL_tokenbuf[0] == '@') ? SVt_PVAV
7033 * The following code was generated by perl_keyword.pl.
7037 Perl_keyword (pTHX_ const char *name, I32 len, bool all_keywords)
7042 case 1: /* 5 tokens of length 1 */
7074 case 2: /* 18 tokens of length 2 */
7220 case 3: /* 29 tokens of length 3 */
7224 if (name[1] == 'N' &&
7287 if (name[1] == 'i' &&
7309 return (all_keywords || FEATURE_IS_ENABLED("err") ? -KEY_err : 0);
7327 if (name[1] == 'o' &&
7336 if (name[1] == 'e' &&
7345 if (name[1] == 'n' &&
7354 if (name[1] == 'o' &&
7363 if (name[1] == 'a' &&
7372 if (name[1] == 'o' &&
7434 if (name[1] == 'e' &&
7448 return (all_keywords || FEATURE_IS_ENABLED("say") ? KEY_say : 0);
7474 if (name[1] == 'i' &&
7483 if (name[1] == 's' &&
7492 if (name[1] == 'e' &&
7501 if (name[1] == 'o' &&
7513 case 4: /* 41 tokens of length 4 */
7517 if (name[1] == 'O' &&
7527 if (name[1] == 'N' &&
7537 if (name[1] == 'i' &&
7547 if (name[1] == 'h' &&
7557 if (name[1] == 'u' &&
7570 if (name[2] == 'c' &&
7579 if (name[2] == 's' &&
7588 if (name[2] == 'a' &&
7624 if (name[1] == 'o' &&
7637 if (name[2] == 't' &&
7646 if (name[2] == 'o' &&
7655 if (name[2] == 't' &&
7664 if (name[2] == 'e' &&
7677 if (name[1] == 'o' &&
7690 if (name[2] == 'y' &&
7699 if (name[2] == 'l' &&
7715 if (name[2] == 's' &&
7724 if (name[2] == 'n' &&
7733 if (name[2] == 'c' &&
7746 if (name[1] == 'e' &&
7756 if (name[1] == 'p' &&
7769 if (name[2] == 'c' &&
7778 if (name[2] == 'p' &&
7787 if (name[2] == 's' &&
7803 if (name[2] == 'n' &&
7873 if (name[2] == 'r' &&
7882 if (name[2] == 'r' &&
7891 if (name[2] == 'a' &&
7907 if (name[2] == 'l' &&
7969 if (name[2] == 'e' &&
7972 return (all_keywords || FEATURE_IS_ENABLED("switch") ? KEY_when : 0);
7985 case 5: /* 39 tokens of length 5 */
7989 if (name[1] == 'E' &&
8000 if (name[1] == 'H' &&
8014 if (name[2] == 'a' &&
8024 if (name[2] == 'a' &&
8041 if (name[2] == 'e' &&
8051 if (name[2] == 'e' &&
8055 return (all_keywords || FEATURE_IS_ENABLED("switch") ? -KEY_break : 0);
8071 if (name[3] == 'i' &&
8080 if (name[3] == 'o' &&
8116 if (name[2] == 'o' &&
8126 if (name[2] == 'y' &&
8140 if (name[1] == 'l' &&
8154 if (name[2] == 'n' &&
8164 if (name[2] == 'o' &&
8178 if (name[1] == 'i' &&
8183 return (all_keywords || FEATURE_IS_ENABLED("switch") ? KEY_given : 0);
8192 if (name[2] == 'd' &&
8202 if (name[2] == 'c' &&
8219 if (name[2] == 'c' &&
8229 if (name[2] == 't' &&
8243 if (name[1] == 'k' &&
8254 if (name[1] == 'r' &&
8268 if (name[2] == 's' &&
8278 if (name[2] == 'd' &&
8295 if (name[2] == 'm' &&
8305 if (name[2] == 'i' &&
8315 if (name[2] == 'e' &&
8325 if (name[2] == 'l' &&
8335 if (name[2] == 'a' &&
8348 if (name[3] == 't' &&
8351 return (all_keywords || FEATURE_IS_ENABLED("state") ? KEY_state : 0);
8357 if (name[3] == 'd' &&
8374 if (name[1] == 'i' &&
8388 if (name[2] == 'a' &&
8401 if (name[3] == 'e' &&
8436 if (name[2] == 'i' &&
8453 if (name[2] == 'i' &&
8463 if (name[2] == 'i' &&
8480 case 6: /* 33 tokens of length 6 */
8484 if (name[1] == 'c' &&
8499 if (name[2] == 'l' &&
8510 if (name[2] == 'r' &&
8525 if (name[1] == 'e' &&
8540 if (name[2] == 's' &&
8545 if(ckWARN_d(WARN_SYNTAX))
8546 Perl_warner(aTHX_ packWARN(WARN_SYNTAX), "elseif should be elsif");
8552 if (name[2] == 'i' &&
8570 if (name[2] == 'l' &&
8581 if (name[2] == 'r' &&
8596 if (name[1] == 'm' &&
8611 if (name[2] == 'n' &&
8622 if (name[2] == 's' &&
8637 if (name[1] == 's' &&
8643 if (name[4] == 't' &&
8652 if (name[4] == 'e' &&
8661 if (name[4] == 'c' &&
8670 if (name[4] == 'n' &&
8686 if (name[1] == 'r' &&
8704 if (name[3] == 'a' &&
8714 if (name[3] == 'u' &&
8728 if (name[2] == 'n' &&
8746 if (name[2] == 'a' &&
8760 if (name[3] == 'e' &&
8773 if (name[4] == 't' &&
8782 if (name[4] == 'e' &&
8804 if (name[4] == 't' &&
8813 if (name[4] == 'e' &&
8829 if (name[2] == 'c' &&
8840 if (name[2] == 'l' &&
8851 if (name[2] == 'b' &&
8862 if (name[2] == 's' &&
8885 if (name[4] == 's' &&
8894 if (name[4] == 'n' &&
8907 if (name[3] == 'a' &&
8924 if (name[1] == 'a' &&
8939 case 7: /* 29 tokens of length 7 */
8943 if (name[1] == 'E' &&
8956 if (name[1] == '_' &&
8969 if (name[1] == 'i' &&
8976 return -KEY_binmode;
8982 if (name[1] == 'o' &&
8989 return -KEY_connect;
8998 if (name[2] == 'm' &&
9004 return -KEY_dbmopen;
9015 if (name[4] == 'u' &&
9019 return (all_keywords || FEATURE_IS_ENABLED("switch") ? KEY_default : 0);
9025 if (name[4] == 'n' &&
9046 if (name[1] == 'o' &&
9059 if (name[1] == 'e' &&
9066 if (name[5] == 'r' &&
9069 return -KEY_getpgrp;
9075 if (name[5] == 'i' &&
9078 return -KEY_getppid;
9091 if (name[1] == 'c' &&
9098 return -KEY_lcfirst;
9104 if (name[1] == 'p' &&
9111 return -KEY_opendir;
9117 if (name[1] == 'a' &&
9135 if (name[3] == 'd' &&
9140 return -KEY_readdir;
9146 if (name[3] == 'u' &&
9157 if (name[3] == 'e' &&
9162 return -KEY_reverse;
9181 if (name[3] == 'k' &&
9186 return -KEY_seekdir;
9192 if (name[3] == 'p' &&
9197 return -KEY_setpgrp;
9207 if (name[2] == 'm' &&
9213 return -KEY_shmread;
9219 if (name[2] == 'r' &&
9225 return -KEY_sprintf;
9234 if (name[3] == 'l' &&
9239 return -KEY_symlink;
9248 if (name[4] == 'a' &&
9252 return -KEY_syscall;
9258 if (name[4] == 'p' &&
9262 return -KEY_sysopen;
9268 if (name[4] == 'e' &&
9272 return -KEY_sysread;
9278 if (name[4] == 'e' &&
9282 return -KEY_sysseek;
9300 if (name[1] == 'e' &&
9307 return -KEY_telldir;
9316 if (name[2] == 'f' &&
9322 return -KEY_ucfirst;
9328 if (name[2] == 's' &&
9334 return -KEY_unshift;
9344 if (name[1] == 'a' &&
9351 return -KEY_waitpid;
9360 case 8: /* 26 tokens of length 8 */
9364 if (name[1] == 'U' &&
9372 return KEY_AUTOLOAD;
9383 if (name[3] == 'A' &&
9389 return KEY___DATA__;
9395 if (name[3] == 'I' &&
9401 return -KEY___FILE__;
9407 if (name[3] == 'I' &&
9413 return -KEY___LINE__;
9429 if (name[2] == 'o' &&
9436 return -KEY_closedir;
9442 if (name[2] == 'n' &&
9449 return -KEY_continue;
9459 if (name[1] == 'b' &&
9467 return -KEY_dbmclose;
9473 if (name[1] == 'n' &&
9479 if (name[4] == 'r' &&
9484 return -KEY_endgrent;
9490 if (name[4] == 'w' &&
9495 return -KEY_endpwent;
9508 if (name[1] == 'o' &&
9516 return -KEY_formline;
9522 if (name[1] == 'e' &&
9533 if (name[6] == 'n' &&
9536 return -KEY_getgrent;
9542 if (name[6] == 'i' &&
9545 return -KEY_getgrgid;
9551 if (name[6] == 'a' &&
9554 return -KEY_getgrnam;
9567 if (name[4] == 'o' &&
9572 return -KEY_getlogin;
9583 if (name[6] == 'n' &&
9586 return -KEY_getpwent;
9592 if (name[6] == 'a' &&
9595 return -KEY_getpwnam;
9601 if (name[6] == 'i' &&
9604 return -KEY_getpwuid;
9624 if (name[1] == 'e' &&
9631 if (name[5] == 'i' &&
9638 return -KEY_readline;
9643 return -KEY_readlink;
9654 if (name[5] == 'i' &&
9658 return -KEY_readpipe;
9679 if (name[4] == 'r' &&
9684 return -KEY_setgrent;
9690 if (name[4] == 'w' &&
9695 return -KEY_setpwent;
9711 if (name[3] == 'w' &&
9717 return -KEY_shmwrite;
9723 if (name[3] == 't' &&
9729 return -KEY_shutdown;
9739 if (name[2] == 's' &&
9746 return -KEY_syswrite;
9756 if (name[1] == 'r' &&
9764 return -KEY_truncate;
9773 case 9: /* 9 tokens of length 9 */
9777 if (name[1] == 'N' &&
9786 return KEY_UNITCHECK;
9792 if (name[1] == 'n' &&
9801 return -KEY_endnetent;
9807 if (name[1] == 'e' &&
9816 return -KEY_getnetent;
9822 if (name[1] == 'o' &&
9831 return -KEY_localtime;
9837 if (name[1] == 'r' &&
9846 return KEY_prototype;
9852 if (name[1] == 'u' &&
9861 return -KEY_quotemeta;
9867 if (name[1] == 'e' &&
9876 return -KEY_rewinddir;
9882 if (name[1] == 'e' &&
9891 return -KEY_setnetent;
9897 if (name[1] == 'a' &&
9906 return -KEY_wantarray;
9915 case 10: /* 9 tokens of length 10 */
9919 if (name[1] == 'n' &&
9925 if (name[4] == 'o' &&
9932 return -KEY_endhostent;
9938 if (name[4] == 'e' &&
9945 return -KEY_endservent;
9958 if (name[1] == 'e' &&
9964 if (name[4] == 'o' &&
9971 return -KEY_gethostent;
9980 if (name[5] == 'r' &&
9986 return -KEY_getservent;
9992 if (name[5] == 'c' &&
9998 return -KEY_getsockopt;
10018 if (name[2] == 't')
10023 if (name[4] == 'o' &&
10030 return -KEY_sethostent;
10039 if (name[5] == 'r' &&
10045 return -KEY_setservent;
10051 if (name[5] == 'c' &&
10057 return -KEY_setsockopt;
10074 if (name[2] == 'c' &&
10083 return -KEY_socketpair;
10096 case 11: /* 8 tokens of length 11 */
10100 if (name[1] == '_' &&
10110 { /* __PACKAGE__ */
10111 return -KEY___PACKAGE__;
10117 if (name[1] == 'n' &&
10127 { /* endprotoent */
10128 return -KEY_endprotoent;
10134 if (name[1] == 'e' &&
10143 if (name[5] == 'e' &&
10149 { /* getpeername */
10150 return -KEY_getpeername;
10159 if (name[6] == 'o' &&
10164 { /* getpriority */
10165 return -KEY_getpriority;
10171 if (name[6] == 't' &&
10176 { /* getprotoent */
10177 return -KEY_getprotoent;
10191 if (name[4] == 'o' &&
10198 { /* getsockname */
10199 return -KEY_getsockname;
10212 if (name[1] == 'e' &&
10220 if (name[6] == 'o' &&
10225 { /* setpriority */
10226 return -KEY_setpriority;
10232 if (name[6] == 't' &&
10237 { /* setprotoent */
10238 return -KEY_setprotoent;
10254 case 12: /* 2 tokens of length 12 */
10255 if (name[0] == 'g' &&
10267 if (name[9] == 'd' &&
10270 { /* getnetbyaddr */
10271 return -KEY_getnetbyaddr;
10277 if (name[9] == 'a' &&
10280 { /* getnetbyname */
10281 return -KEY_getnetbyname;
10293 case 13: /* 4 tokens of length 13 */
10294 if (name[0] == 'g' &&
10301 if (name[4] == 'o' &&
10310 if (name[10] == 'd' &&
10313 { /* gethostbyaddr */
10314 return -KEY_gethostbyaddr;
10320 if (name[10] == 'a' &&
10323 { /* gethostbyname */
10324 return -KEY_gethostbyname;
10337 if (name[4] == 'e' &&
10346 if (name[10] == 'a' &&
10349 { /* getservbyname */
10350 return -KEY_getservbyname;
10356 if (name[10] == 'o' &&
10359 { /* getservbyport */
10360 return -KEY_getservbyport;
10379 case 14: /* 1 tokens of length 14 */
10380 if (name[0] == 'g' &&
10394 { /* getprotobyname */
10395 return -KEY_getprotobyname;
10400 case 16: /* 1 tokens of length 16 */
10401 if (name[0] == 'g' &&
10417 { /* getprotobynumber */
10418 return -KEY_getprotobynumber;
10432 S_checkcomma(pTHX_ const char *s, const char *name, const char *what)
10436 if (*s == ' ' && s[1] == '(') { /* XXX gotta be a better way */
10437 if (ckWARN(WARN_SYNTAX)) {
10440 for (w = s+2; *w && level; w++) {
10443 else if (*w == ')')
10446 while (isSPACE(*w))
10448 if (!*w || !strchr(";|})]oaiuw!=", *w)) /* an advisory hack only... */
10449 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
10450 "%s (...) interpreted as function",name);
10453 while (s < PL_bufend && isSPACE(*s))
10457 while (s < PL_bufend && isSPACE(*s))
10459 if (isIDFIRST_lazy_if(s,UTF)) {
10460 const char * const w = s++;
10461 while (isALNUM_lazy_if(s,UTF))
10463 while (s < PL_bufend && isSPACE(*s))
10467 if (keyword(w, s - w, 0))
10470 gv = gv_fetchpvn_flags(w, s - w, 0, SVt_PVCV);
10471 if (gv && GvCVu(gv))
10473 Perl_croak(aTHX_ "No comma allowed after %s", what);
10478 /* Either returns sv, or mortalizes sv and returns a new SV*.
10479 Best used as sv=new_constant(..., sv, ...).
10480 If s, pv are NULL, calls subroutine with one argument,
10481 and type is used with error messages only. */
10484 S_new_constant(pTHX_ const char *s, STRLEN len, const char *key, SV *sv, SV *pv,
10488 HV * const table = GvHV(PL_hintgv); /* ^H */
10492 const char *why1 = "", *why2 = "", *why3 = "";
10494 if (!table || !(PL_hints & HINT_LOCALIZE_HH)) {
10497 why2 = (const char *)
10498 (strEQ(key,"charnames")
10499 ? "(possibly a missing \"use charnames ...\")"
10501 msg = Perl_newSVpvf(aTHX_ "Constant(%s) unknown: %s",
10502 (type ? type: "undef"), why2);
10504 /* This is convoluted and evil ("goto considered harmful")
10505 * but I do not understand the intricacies of all the different
10506 * failure modes of %^H in here. The goal here is to make
10507 * the most probable error message user-friendly. --jhi */
10512 msg = Perl_newSVpvf(aTHX_ "Constant(%s): %s%s%s",
10513 (type ? type: "undef"), why1, why2, why3);
10515 yyerror(SvPVX_const(msg));
10519 cvp = hv_fetch(table, key, strlen(key), FALSE);
10520 if (!cvp || !SvOK(*cvp)) {
10523 why3 = "} is not defined";
10526 sv_2mortal(sv); /* Parent created it permanently */
10529 pv = sv_2mortal(newSVpvn(s, len));
10531 typesv = sv_2mortal(newSVpv(type, 0));
10533 typesv = &PL_sv_undef;
10535 PUSHSTACKi(PERLSI_OVERLOAD);
10547 call_sv(cv, G_SCALAR | ( PL_in_eval ? 0 : G_EVAL));
10551 /* Check the eval first */
10552 if (!PL_in_eval && SvTRUE(ERRSV)) {
10553 sv_catpvs(ERRSV, "Propagated");
10554 yyerror(SvPV_nolen_const(ERRSV)); /* Duplicates the message inside eval */
10556 res = SvREFCNT_inc_simple(sv);
10560 SvREFCNT_inc_simple_void(res);
10569 why1 = "Call to &{$^H{";
10571 why3 = "}} did not return a defined value";
10579 /* Returns a NUL terminated string, with the length of the string written to
10583 S_scan_word(pTHX_ register char *s, char *dest, STRLEN destlen, int allow_package, STRLEN *slp)
10586 register char *d = dest;
10587 register char * const e = d + destlen - 3; /* two-character token, ending NUL */
10590 Perl_croak(aTHX_ ident_too_long);
10591 if (isALNUM(*s)) /* UTF handled below */
10593 else if (allow_package && (*s == '\'') && isIDFIRST_lazy_if(s+1,UTF)) {
10598 else if (allow_package && (s[0] == ':') && (s[1] == ':') && (s[2] != '$')) {
10602 else if (UTF && UTF8_IS_START(*s) && isALNUM_utf8((U8*)s)) {
10603 char *t = s + UTF8SKIP(s);
10605 while (UTF8_IS_CONTINUED(*t) && is_utf8_mark((U8*)t))
10609 Perl_croak(aTHX_ ident_too_long);
10610 Copy(s, d, len, char);
10623 S_scan_ident(pTHX_ register char *s, register const char *send, char *dest, STRLEN destlen, I32 ck_uni)
10626 char *bracket = NULL;
10628 register char *d = dest;
10629 register char * const e = d + destlen + 3; /* two-character token, ending NUL */
10634 while (isDIGIT(*s)) {
10636 Perl_croak(aTHX_ ident_too_long);
10643 Perl_croak(aTHX_ ident_too_long);
10644 if (isALNUM(*s)) /* UTF handled below */
10646 else if (*s == '\'' && isIDFIRST_lazy_if(s+1,UTF)) {
10651 else if (*s == ':' && s[1] == ':') {
10655 else if (UTF && UTF8_IS_START(*s) && isALNUM_utf8((U8*)s)) {
10656 char *t = s + UTF8SKIP(s);
10657 while (UTF8_IS_CONTINUED(*t) && is_utf8_mark((U8*)t))
10659 if (d + (t - s) > e)
10660 Perl_croak(aTHX_ ident_too_long);
10661 Copy(s, d, t - s, char);
10672 if (PL_lex_state != LEX_NORMAL)
10673 PL_lex_state = LEX_INTERPENDMAYBE;
10676 if (*s == '$' && s[1] &&
10677 (isALNUM_lazy_if(s+1,UTF) || s[1] == '$' || s[1] == '{' || strnEQ(s+1,"::",2)) )
10690 if (*d == '^' && *s && isCONTROLVAR(*s)) {
10695 if (isSPACE(s[-1])) {
10697 const char ch = *s++;
10698 if (!SPACE_OR_TAB(ch)) {
10704 if (isIDFIRST_lazy_if(d,UTF)) {
10708 while ((end < send && isALNUM_lazy_if(end,UTF)) || *end == ':') {
10709 end += UTF8SKIP(end);
10710 while (end < send && UTF8_IS_CONTINUED(*end) && is_utf8_mark((U8*)end))
10711 end += UTF8SKIP(end);
10713 Copy(s, d, end - s, char);
10718 while ((isALNUM(*s) || *s == ':') && d < e)
10721 Perl_croak(aTHX_ ident_too_long);
10724 while (s < send && SPACE_OR_TAB(*s))
10726 if ((*s == '[' || (*s == '{' && strNE(dest, "sub")))) {
10727 if (ckWARN(WARN_AMBIGUOUS) && keyword(dest, d - dest, 0)) {
10728 const char * const brack =
10730 ((*s == '[') ? "[...]" : "{...}");
10731 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
10732 "Ambiguous use of %c{%s%s} resolved to %c%s%s",
10733 funny, dest, brack, funny, dest, brack);
10736 PL_lex_brackstack[PL_lex_brackets++] = (char)(XOPERATOR | XFAKEBRACK);
10740 /* Handle extended ${^Foo} variables
10741 * 1999-02-27 mjd-perl-patch@plover.com */
10742 else if (!isALNUM(*d) && !isPRINT(*d) /* isCTRL(d) */
10746 while (isALNUM(*s) && d < e) {
10750 Perl_croak(aTHX_ ident_too_long);
10755 if (PL_lex_state == LEX_INTERPNORMAL && !PL_lex_brackets) {
10756 PL_lex_state = LEX_INTERPEND;
10759 if (PL_lex_state == LEX_NORMAL) {
10760 if (ckWARN(WARN_AMBIGUOUS) &&
10761 (keyword(dest, d - dest, 0) || get_cv(dest, FALSE)))
10765 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
10766 "Ambiguous use of %c{%s} resolved to %c%s",
10767 funny, dest, funny, dest);
10772 s = bracket; /* let the parser handle it */
10776 else if (PL_lex_state == LEX_INTERPNORMAL && !PL_lex_brackets && !intuit_more(s))
10777 PL_lex_state = LEX_INTERPEND;
10782 Perl_pmflag(pTHX_ U32* pmfl, int ch)
10784 PERL_UNUSED_CONTEXT;
10787 else if (ch == 'g')
10788 *pmfl |= PMf_GLOBAL;
10789 else if (ch == 'c')
10790 *pmfl |= PMf_CONTINUE;
10791 else if (ch == 'o')
10793 else if (ch == 'm')
10794 *pmfl |= PMf_MULTILINE;
10795 else if (ch == 's')
10796 *pmfl |= PMf_SINGLELINE;
10797 else if (ch == 'x')
10798 *pmfl |= PMf_EXTENDED;
10802 S_scan_pat(pTHX_ char *start, I32 type)
10806 char *s = scan_str(start,!!PL_madskills,FALSE);
10807 const char * const valid_flags =
10808 (const char *)((type == OP_QR) ? "iomsx" : "iogcmsx");
10815 const char * const delimiter = skipspace(start);
10819 ? "Search pattern not terminated or ternary operator parsed as search pattern"
10820 : "Search pattern not terminated" ));
10823 pm = (PMOP*)newPMOP(type, 0);
10824 if (PL_multi_open == '?')
10825 pm->op_pmflags |= PMf_ONCE;
10829 while (*s && strchr(valid_flags, *s))
10830 pmflag(&pm->op_pmflags,*s++);
10832 if (PL_madskills && modstart != s) {
10833 SV* tmptoken = newSVpvn(modstart, s - modstart);
10834 append_madprops(newMADPROP('m', MAD_SV, tmptoken, 0), (OP*)pm, 0);
10837 /* issue a warning if /c is specified,but /g is not */
10838 if ((pm->op_pmflags & PMf_CONTINUE) && !(pm->op_pmflags & PMf_GLOBAL)
10839 && ckWARN(WARN_REGEXP))
10841 Perl_warner(aTHX_ packWARN(WARN_REGEXP), "Use of /c modifier is meaningless without /g" );
10844 pm->op_pmpermflags = pm->op_pmflags;
10846 PL_lex_op = (OP*)pm;
10847 yylval.ival = OP_MATCH;
10852 S_scan_subst(pTHX_ char *start)
10863 yylval.ival = OP_NULL;
10865 s = scan_str(start,!!PL_madskills,FALSE);
10868 Perl_croak(aTHX_ "Substitution pattern not terminated");
10870 if (s[-1] == PL_multi_open)
10873 if (PL_madskills) {
10874 CURMAD('q', PL_thisopen);
10875 CURMAD('_', PL_thiswhite);
10876 CURMAD('E', PL_thisstuff);
10877 CURMAD('Q', PL_thisclose);
10878 PL_realtokenstart = s - SvPVX(PL_linestr);
10882 first_start = PL_multi_start;
10883 s = scan_str(s,!!PL_madskills,FALSE);
10885 if (PL_lex_stuff) {
10886 SvREFCNT_dec(PL_lex_stuff);
10887 PL_lex_stuff = NULL;
10889 Perl_croak(aTHX_ "Substitution replacement not terminated");
10891 PL_multi_start = first_start; /* so whole substitution is taken together */
10893 pm = (PMOP*)newPMOP(OP_SUBST, 0);
10896 if (PL_madskills) {
10897 CURMAD('z', PL_thisopen);
10898 CURMAD('R', PL_thisstuff);
10899 CURMAD('Z', PL_thisclose);
10909 else if (strchr("iogcmsx", *s))
10910 pmflag(&pm->op_pmflags,*s++);
10916 if (PL_madskills) {
10918 curmad('m', newSVpvn(modstart, s - modstart));
10919 append_madprops(PL_thismad, (OP*)pm, 0);
10923 if ((pm->op_pmflags & PMf_CONTINUE) && ckWARN(WARN_REGEXP)) {
10924 Perl_warner(aTHX_ packWARN(WARN_REGEXP), "Use of /c modifier is meaningless in s///" );
10928 SV * const repl = newSVpvs("");
10930 PL_sublex_info.super_bufptr = s;
10931 PL_sublex_info.super_bufend = PL_bufend;
10933 pm->op_pmflags |= PMf_EVAL;
10935 sv_catpv(repl, (const char *)(es ? "eval " : "do "));
10936 sv_catpvs(repl, "{");
10937 sv_catsv(repl, PL_lex_repl);
10938 if (strchr(SvPVX(PL_lex_repl), '#'))
10939 sv_catpvs(repl, "\n");
10940 sv_catpvs(repl, "}");
10942 SvREFCNT_dec(PL_lex_repl);
10943 PL_lex_repl = repl;
10946 pm->op_pmpermflags = pm->op_pmflags;
10947 PL_lex_op = (OP*)pm;
10948 yylval.ival = OP_SUBST;
10953 S_scan_trans(pTHX_ char *start)
10966 yylval.ival = OP_NULL;
10968 s = scan_str(start,!!PL_madskills,FALSE);
10970 Perl_croak(aTHX_ "Transliteration pattern not terminated");
10972 if (s[-1] == PL_multi_open)
10975 if (PL_madskills) {
10976 CURMAD('q', PL_thisopen);
10977 CURMAD('_', PL_thiswhite);
10978 CURMAD('E', PL_thisstuff);
10979 CURMAD('Q', PL_thisclose);
10980 PL_realtokenstart = s - SvPVX(PL_linestr);
10984 s = scan_str(s,!!PL_madskills,FALSE);
10986 if (PL_lex_stuff) {
10987 SvREFCNT_dec(PL_lex_stuff);
10988 PL_lex_stuff = NULL;
10990 Perl_croak(aTHX_ "Transliteration replacement not terminated");
10992 if (PL_madskills) {
10993 CURMAD('z', PL_thisopen);
10994 CURMAD('R', PL_thisstuff);
10995 CURMAD('Z', PL_thisclose);
10998 complement = del = squash = 0;
11005 complement = OPpTRANS_COMPLEMENT;
11008 del = OPpTRANS_DELETE;
11011 squash = OPpTRANS_SQUASH;
11020 Newx(tbl, complement&&!del?258:256, short);
11021 o = newPVOP(OP_TRANS, 0, (char*)tbl);
11022 o->op_private &= ~OPpTRANS_ALL;
11023 o->op_private |= del|squash|complement|
11024 (DO_UTF8(PL_lex_stuff)? OPpTRANS_FROM_UTF : 0)|
11025 (DO_UTF8(PL_lex_repl) ? OPpTRANS_TO_UTF : 0);
11028 yylval.ival = OP_TRANS;
11031 if (PL_madskills) {
11033 curmad('m', newSVpvn(modstart, s - modstart));
11034 append_madprops(PL_thismad, o, 0);
11043 S_scan_heredoc(pTHX_ register char *s)
11047 I32 op_type = OP_SCALAR;
11051 const char *found_newline;
11055 const int outer = (PL_rsfp && !(PL_lex_inwhat == OP_SCALAR));
11057 I32 stuffstart = s - SvPVX(PL_linestr);
11060 PL_realtokenstart = -1;
11065 e = PL_tokenbuf + sizeof PL_tokenbuf - 1;
11069 while (SPACE_OR_TAB(*peek))
11071 if (*peek == '`' || *peek == '\'' || *peek =='"') {
11074 s = delimcpy(d, e, s, PL_bufend, term, &len);
11084 if (!isALNUM_lazy_if(s,UTF))
11085 deprecate_old("bare << to mean <<\"\"");
11086 for (; isALNUM_lazy_if(s,UTF); s++) {
11091 if (d >= PL_tokenbuf + sizeof PL_tokenbuf - 1)
11092 Perl_croak(aTHX_ "Delimiter for here document is too long");
11095 len = d - PL_tokenbuf;
11098 if (PL_madskills) {
11099 tstart = PL_tokenbuf + !outer;
11100 PL_thisclose = newSVpvn(tstart, len - !outer);
11101 tstart = SvPVX(PL_linestr) + stuffstart;
11102 PL_thisopen = newSVpvn(tstart, s - tstart);
11103 stuffstart = s - SvPVX(PL_linestr);
11106 #ifndef PERL_STRICT_CR
11107 d = strchr(s, '\r');
11109 char * const olds = s;
11111 while (s < PL_bufend) {
11117 else if (*s == '\n' && s[1] == '\r') { /* \015\013 on a mac? */
11126 SvCUR_set(PL_linestr, PL_bufend - SvPVX_const(PL_linestr));
11133 if ( outer || !(found_newline = (char*)memchr((void*)s, '\n', PL_bufend - s)) ) {
11134 herewas = newSVpvn(s,PL_bufend-s);
11138 herewas = newSVpvn(s-1,found_newline-s+1);
11141 herewas = newSVpvn(s,found_newline-s);
11145 if (PL_madskills) {
11146 tstart = SvPVX(PL_linestr) + stuffstart;
11148 sv_catpvn(PL_thisstuff, tstart, s - tstart);
11150 PL_thisstuff = newSVpvn(tstart, s - tstart);
11153 s += SvCUR(herewas);
11156 stuffstart = s - SvPVX(PL_linestr);
11162 tmpstr = newSV(79);
11163 sv_upgrade(tmpstr, SVt_PVIV);
11164 if (term == '\'') {
11165 op_type = OP_CONST;
11166 SvIV_set(tmpstr, -1);
11168 else if (term == '`') {
11169 op_type = OP_BACKTICK;
11170 SvIV_set(tmpstr, '\\');
11174 PL_multi_start = CopLINE(PL_curcop);
11175 PL_multi_open = PL_multi_close = '<';
11176 term = *PL_tokenbuf;
11177 if (PL_lex_inwhat == OP_SUBST && PL_in_eval && !PL_rsfp) {
11178 char * const bufptr = PL_sublex_info.super_bufptr;
11179 char * const bufend = PL_sublex_info.super_bufend;
11180 char * const olds = s - SvCUR(herewas);
11181 s = strchr(bufptr, '\n');
11185 while (s < bufend &&
11186 (*s != term || memNE(s,PL_tokenbuf,len)) ) {
11188 CopLINE_inc(PL_curcop);
11191 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
11192 missingterm(PL_tokenbuf);
11194 sv_setpvn(herewas,bufptr,d-bufptr+1);
11195 sv_setpvn(tmpstr,d+1,s-d);
11197 sv_catpvn(herewas,s,bufend-s);
11198 Copy(SvPVX_const(herewas),bufptr,SvCUR(herewas) + 1,char);
11205 while (s < PL_bufend &&
11206 (*s != term || memNE(s,PL_tokenbuf,len)) ) {
11208 CopLINE_inc(PL_curcop);
11210 if (s >= PL_bufend) {
11211 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
11212 missingterm(PL_tokenbuf);
11214 sv_setpvn(tmpstr,d+1,s-d);
11216 if (PL_madskills) {
11218 sv_catpvn(PL_thisstuff, d + 1, s - d);
11220 PL_thisstuff = newSVpvn(d + 1, s - d);
11221 stuffstart = s - SvPVX(PL_linestr);
11225 CopLINE_inc(PL_curcop); /* the preceding stmt passes a newline */
11227 sv_catpvn(herewas,s,PL_bufend-s);
11228 sv_setsv(PL_linestr,herewas);
11229 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = s = PL_linestart = SvPVX(PL_linestr);
11230 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
11231 PL_last_lop = PL_last_uni = NULL;
11234 sv_setpvn(tmpstr,"",0); /* avoid "uninitialized" warning */
11235 while (s >= PL_bufend) { /* multiple line string? */
11237 if (PL_madskills) {
11238 tstart = SvPVX(PL_linestr) + stuffstart;
11240 sv_catpvn(PL_thisstuff, tstart, PL_bufend - tstart);
11242 PL_thisstuff = newSVpvn(tstart, PL_bufend - tstart);
11246 !(PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = filter_gets(PL_linestr, PL_rsfp, 0))) {
11247 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
11248 missingterm(PL_tokenbuf);
11251 stuffstart = s - SvPVX(PL_linestr);
11253 CopLINE_inc(PL_curcop);
11254 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
11255 PL_last_lop = PL_last_uni = NULL;
11256 #ifndef PERL_STRICT_CR
11257 if (PL_bufend - PL_linestart >= 2) {
11258 if ((PL_bufend[-2] == '\r' && PL_bufend[-1] == '\n') ||
11259 (PL_bufend[-2] == '\n' && PL_bufend[-1] == '\r'))
11261 PL_bufend[-2] = '\n';
11263 SvCUR_set(PL_linestr, PL_bufend - SvPVX_const(PL_linestr));
11265 else if (PL_bufend[-1] == '\r')
11266 PL_bufend[-1] = '\n';
11268 else if (PL_bufend - PL_linestart == 1 && PL_bufend[-1] == '\r')
11269 PL_bufend[-1] = '\n';
11271 if (PERLDB_LINE && PL_curstash != PL_debstash)
11272 update_debugger_info_sv(PL_linestr);
11273 if (*s == term && memEQ(s,PL_tokenbuf,len)) {
11274 STRLEN off = PL_bufend - 1 - SvPVX_const(PL_linestr);
11275 *(SvPVX(PL_linestr) + off ) = ' ';
11276 sv_catsv(PL_linestr,herewas);
11277 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
11278 s = SvPVX(PL_linestr) + off; /* In case PV of PL_linestr moved. */
11282 sv_catsv(tmpstr,PL_linestr);
11287 PL_multi_end = CopLINE(PL_curcop);
11288 if (SvCUR(tmpstr) + 5 < SvLEN(tmpstr)) {
11289 SvPV_shrink_to_cur(tmpstr);
11291 SvREFCNT_dec(herewas);
11293 if (UTF && is_utf8_string((U8*)SvPVX_const(tmpstr), SvCUR(tmpstr)))
11295 else if (PL_encoding)
11296 sv_recode_to_utf8(tmpstr, PL_encoding);
11298 PL_lex_stuff = tmpstr;
11299 yylval.ival = op_type;
11303 /* scan_inputsymbol
11304 takes: current position in input buffer
11305 returns: new position in input buffer
11306 side-effects: yylval and lex_op are set.
11311 <FH> read from filehandle
11312 <pkg::FH> read from package qualified filehandle
11313 <pkg'FH> read from package qualified filehandle
11314 <$fh> read from filehandle in $fh
11315 <*.h> filename glob
11320 S_scan_inputsymbol(pTHX_ char *start)
11323 register char *s = start; /* current position in buffer */
11327 char *d = PL_tokenbuf; /* start of temp holding space */
11328 const char * const e = PL_tokenbuf + sizeof PL_tokenbuf; /* end of temp holding space */
11330 end = strchr(s, '\n');
11333 s = delimcpy(d, e, s + 1, end, '>', &len); /* extract until > */
11335 /* die if we didn't have space for the contents of the <>,
11336 or if it didn't end, or if we see a newline
11339 if (len >= (I32)sizeof PL_tokenbuf)
11340 Perl_croak(aTHX_ "Excessively long <> operator");
11342 Perl_croak(aTHX_ "Unterminated <> operator");
11347 Remember, only scalar variables are interpreted as filehandles by
11348 this code. Anything more complex (e.g., <$fh{$num}>) will be
11349 treated as a glob() call.
11350 This code makes use of the fact that except for the $ at the front,
11351 a scalar variable and a filehandle look the same.
11353 if (*d == '$' && d[1]) d++;
11355 /* allow <Pkg'VALUE> or <Pkg::VALUE> */
11356 while (*d && (isALNUM_lazy_if(d,UTF) || *d == '\'' || *d == ':'))
11359 /* If we've tried to read what we allow filehandles to look like, and
11360 there's still text left, then it must be a glob() and not a getline.
11361 Use scan_str to pull out the stuff between the <> and treat it
11362 as nothing more than a string.
11365 if (d - PL_tokenbuf != len) {
11366 yylval.ival = OP_GLOB;
11368 s = scan_str(start,!!PL_madskills,FALSE);
11370 Perl_croak(aTHX_ "Glob not terminated");
11374 bool readline_overriden = FALSE;
11377 /* we're in a filehandle read situation */
11380 /* turn <> into <ARGV> */
11382 Copy("ARGV",d,5,char);
11384 /* Check whether readline() is overriden */
11385 gv_readline = gv_fetchpvs("readline", GV_NOTQUAL, SVt_PVCV);
11387 && GvCVu(gv_readline) && GvIMPORTED_CV(gv_readline))
11389 ((gvp = (GV**)hv_fetchs(PL_globalstash, "readline", FALSE))
11390 && (gv_readline = *gvp) != (GV*)&PL_sv_undef
11391 && GvCVu(gv_readline) && GvIMPORTED_CV(gv_readline)))
11392 readline_overriden = TRUE;
11394 /* if <$fh>, create the ops to turn the variable into a
11398 /* try to find it in the pad for this block, otherwise find
11399 add symbol table ops
11401 const PADOFFSET tmp = pad_findmy(d);
11402 if (tmp != NOT_IN_PAD) {
11403 if (PAD_COMPNAME_FLAGS_isOUR(tmp)) {
11404 HV * const stash = PAD_COMPNAME_OURSTASH(tmp);
11405 HEK * const stashname = HvNAME_HEK(stash);
11406 SV * const sym = sv_2mortal(newSVhek(stashname));
11407 sv_catpvs(sym, "::");
11408 sv_catpv(sym, d+1);
11413 OP * const o = newOP(OP_PADSV, 0);
11415 PL_lex_op = readline_overriden
11416 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
11417 append_elem(OP_LIST, o,
11418 newCVREF(0, newGVOP(OP_GV,0,gv_readline))))
11419 : (OP*)newUNOP(OP_READLINE, 0, o);
11428 ? (GV_ADDMULTI | GV_ADDINEVAL)
11431 PL_lex_op = readline_overriden
11432 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
11433 append_elem(OP_LIST,
11434 newUNOP(OP_RV2SV, 0, newGVOP(OP_GV, 0, gv)),
11435 newCVREF(0, newGVOP(OP_GV, 0, gv_readline))))
11436 : (OP*)newUNOP(OP_READLINE, 0,
11437 newUNOP(OP_RV2SV, 0,
11438 newGVOP(OP_GV, 0, gv)));
11440 if (!readline_overriden)
11441 PL_lex_op->op_flags |= OPf_SPECIAL;
11442 /* we created the ops in PL_lex_op, so make yylval.ival a null op */
11443 yylval.ival = OP_NULL;
11446 /* If it's none of the above, it must be a literal filehandle
11447 (<Foo::BAR> or <FOO>) so build a simple readline OP */
11449 GV * const gv = gv_fetchpv(d, GV_ADD, SVt_PVIO);
11450 PL_lex_op = readline_overriden
11451 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
11452 append_elem(OP_LIST,
11453 newGVOP(OP_GV, 0, gv),
11454 newCVREF(0, newGVOP(OP_GV, 0, gv_readline))))
11455 : (OP*)newUNOP(OP_READLINE, 0, newGVOP(OP_GV, 0, gv));
11456 yylval.ival = OP_NULL;
11465 takes: start position in buffer
11466 keep_quoted preserve \ on the embedded delimiter(s)
11467 keep_delims preserve the delimiters around the string
11468 returns: position to continue reading from buffer
11469 side-effects: multi_start, multi_close, lex_repl or lex_stuff, and
11470 updates the read buffer.
11472 This subroutine pulls a string out of the input. It is called for:
11473 q single quotes q(literal text)
11474 ' single quotes 'literal text'
11475 qq double quotes qq(interpolate $here please)
11476 " double quotes "interpolate $here please"
11477 qx backticks qx(/bin/ls -l)
11478 ` backticks `/bin/ls -l`
11479 qw quote words @EXPORT_OK = qw( func() $spam )
11480 m// regexp match m/this/
11481 s/// regexp substitute s/this/that/
11482 tr/// string transliterate tr/this/that/
11483 y/// string transliterate y/this/that/
11484 ($*@) sub prototypes sub foo ($)
11485 (stuff) sub attr parameters sub foo : attr(stuff)
11486 <> readline or globs <FOO>, <>, <$fh>, or <*.c>
11488 In most of these cases (all but <>, patterns and transliterate)
11489 yylex() calls scan_str(). m// makes yylex() call scan_pat() which
11490 calls scan_str(). s/// makes yylex() call scan_subst() which calls
11491 scan_str(). tr/// and y/// make yylex() call scan_trans() which
11494 It skips whitespace before the string starts, and treats the first
11495 character as the delimiter. If the delimiter is one of ([{< then
11496 the corresponding "close" character )]}> is used as the closing
11497 delimiter. It allows quoting of delimiters, and if the string has
11498 balanced delimiters ([{<>}]) it allows nesting.
11500 On success, the SV with the resulting string is put into lex_stuff or,
11501 if that is already non-NULL, into lex_repl. The second case occurs only
11502 when parsing the RHS of the special constructs s/// and tr/// (y///).
11503 For convenience, the terminating delimiter character is stuffed into
11508 S_scan_str(pTHX_ char *start, int keep_quoted, int keep_delims)
11511 SV *sv; /* scalar value: string */
11512 const char *tmps; /* temp string, used for delimiter matching */
11513 register char *s = start; /* current position in the buffer */
11514 register char term; /* terminating character */
11515 register char *to; /* current position in the sv's data */
11516 I32 brackets = 1; /* bracket nesting level */
11517 bool has_utf8 = FALSE; /* is there any utf8 content? */
11518 I32 termcode; /* terminating char. code */
11519 U8 termstr[UTF8_MAXBYTES]; /* terminating string */
11520 STRLEN termlen; /* length of terminating string */
11521 int last_off = 0; /* last position for nesting bracket */
11527 /* skip space before the delimiter */
11533 if (PL_realtokenstart >= 0) {
11534 stuffstart = PL_realtokenstart;
11535 PL_realtokenstart = -1;
11538 stuffstart = start - SvPVX(PL_linestr);
11540 /* mark where we are, in case we need to report errors */
11543 /* after skipping whitespace, the next character is the terminator */
11546 termcode = termstr[0] = term;
11550 termcode = utf8_to_uvchr((U8*)s, &termlen);
11551 Copy(s, termstr, termlen, U8);
11552 if (!UTF8_IS_INVARIANT(term))
11556 /* mark where we are */
11557 PL_multi_start = CopLINE(PL_curcop);
11558 PL_multi_open = term;
11560 /* find corresponding closing delimiter */
11561 if (term && (tmps = strchr("([{< )]}> )]}>",term)))
11562 termcode = termstr[0] = term = tmps[5];
11564 PL_multi_close = term;
11566 /* create a new SV to hold the contents. 79 is the SV's initial length.
11567 What a random number. */
11569 sv_upgrade(sv, SVt_PVIV);
11570 SvIV_set(sv, termcode);
11571 (void)SvPOK_only(sv); /* validate pointer */
11573 /* move past delimiter and try to read a complete string */
11575 sv_catpvn(sv, s, termlen);
11578 tstart = SvPVX(PL_linestr) + stuffstart;
11579 if (!PL_thisopen && !keep_delims) {
11580 PL_thisopen = newSVpvn(tstart, s - tstart);
11581 stuffstart = s - SvPVX(PL_linestr);
11585 if (PL_encoding && !UTF) {
11589 int offset = s - SvPVX_const(PL_linestr);
11590 const bool found = sv_cat_decode(sv, PL_encoding, PL_linestr,
11591 &offset, (char*)termstr, termlen);
11592 const char * const ns = SvPVX_const(PL_linestr) + offset;
11593 char * const svlast = SvEND(sv) - 1;
11595 for (; s < ns; s++) {
11596 if (*s == '\n' && !PL_rsfp)
11597 CopLINE_inc(PL_curcop);
11600 goto read_more_line;
11602 /* handle quoted delimiters */
11603 if (SvCUR(sv) > 1 && *(svlast-1) == '\\') {
11605 for (t = svlast-2; t >= SvPVX_const(sv) && *t == '\\';)
11607 if ((svlast-1 - t) % 2) {
11608 if (!keep_quoted) {
11609 *(svlast-1) = term;
11611 SvCUR_set(sv, SvCUR(sv) - 1);
11616 if (PL_multi_open == PL_multi_close) {
11622 for (t = w = SvPVX(sv)+last_off; t < svlast; w++, t++) {
11623 /* At here, all closes are "was quoted" one,
11624 so we don't check PL_multi_close. */
11626 if (!keep_quoted && *(t+1) == PL_multi_open)
11631 else if (*t == PL_multi_open)
11639 SvCUR_set(sv, w - SvPVX_const(sv));
11641 last_off = w - SvPVX(sv);
11642 if (--brackets <= 0)
11647 if (!keep_delims) {
11648 SvCUR_set(sv, SvCUR(sv) - 1);
11654 /* extend sv if need be */
11655 SvGROW(sv, SvCUR(sv) + (PL_bufend - s) + 1);
11656 /* set 'to' to the next character in the sv's string */
11657 to = SvPVX(sv)+SvCUR(sv);
11659 /* if open delimiter is the close delimiter read unbridle */
11660 if (PL_multi_open == PL_multi_close) {
11661 for (; s < PL_bufend; s++,to++) {
11662 /* embedded newlines increment the current line number */
11663 if (*s == '\n' && !PL_rsfp)
11664 CopLINE_inc(PL_curcop);
11665 /* handle quoted delimiters */
11666 if (*s == '\\' && s+1 < PL_bufend && term != '\\') {
11667 if (!keep_quoted && s[1] == term)
11669 /* any other quotes are simply copied straight through */
11673 /* terminate when run out of buffer (the for() condition), or
11674 have found the terminator */
11675 else if (*s == term) {
11678 if (s+termlen <= PL_bufend && memEQ(s, (char*)termstr, termlen))
11681 else if (!has_utf8 && !UTF8_IS_INVARIANT((U8)*s) && UTF)
11687 /* if the terminator isn't the same as the start character (e.g.,
11688 matched brackets), we have to allow more in the quoting, and
11689 be prepared for nested brackets.
11692 /* read until we run out of string, or we find the terminator */
11693 for (; s < PL_bufend; s++,to++) {
11694 /* embedded newlines increment the line count */
11695 if (*s == '\n' && !PL_rsfp)
11696 CopLINE_inc(PL_curcop);
11697 /* backslashes can escape the open or closing characters */
11698 if (*s == '\\' && s+1 < PL_bufend) {
11699 if (!keep_quoted &&
11700 ((s[1] == PL_multi_open) || (s[1] == PL_multi_close)))
11705 /* allow nested opens and closes */
11706 else if (*s == PL_multi_close && --brackets <= 0)
11708 else if (*s == PL_multi_open)
11710 else if (!has_utf8 && !UTF8_IS_INVARIANT((U8)*s) && UTF)
11715 /* terminate the copied string and update the sv's end-of-string */
11717 SvCUR_set(sv, to - SvPVX_const(sv));
11720 * this next chunk reads more into the buffer if we're not done yet
11724 break; /* handle case where we are done yet :-) */
11726 #ifndef PERL_STRICT_CR
11727 if (to - SvPVX_const(sv) >= 2) {
11728 if ((to[-2] == '\r' && to[-1] == '\n') ||
11729 (to[-2] == '\n' && to[-1] == '\r'))
11733 SvCUR_set(sv, to - SvPVX_const(sv));
11735 else if (to[-1] == '\r')
11738 else if (to - SvPVX_const(sv) == 1 && to[-1] == '\r')
11743 /* if we're out of file, or a read fails, bail and reset the current
11744 line marker so we can report where the unterminated string began
11747 if (PL_madskills) {
11748 char * const tstart = SvPVX(PL_linestr) + stuffstart;
11750 sv_catpvn(PL_thisstuff, tstart, PL_bufend - tstart);
11752 PL_thisstuff = newSVpvn(tstart, PL_bufend - tstart);
11756 !(PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = filter_gets(PL_linestr, PL_rsfp, 0))) {
11758 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
11764 /* we read a line, so increment our line counter */
11765 CopLINE_inc(PL_curcop);
11767 /* update debugger info */
11768 if (PERLDB_LINE && PL_curstash != PL_debstash)
11769 update_debugger_info_sv(PL_linestr);
11771 /* having changed the buffer, we must update PL_bufend */
11772 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
11773 PL_last_lop = PL_last_uni = NULL;
11776 /* at this point, we have successfully read the delimited string */
11778 if (!PL_encoding || UTF) {
11780 if (PL_madskills) {
11781 char * const tstart = SvPVX(PL_linestr) + stuffstart;
11782 const int len = s - tstart;
11784 sv_catpvn(PL_thisstuff, tstart, len);
11786 PL_thisstuff = newSVpvn(tstart, len);
11787 if (!PL_thisclose && !keep_delims)
11788 PL_thisclose = newSVpvn(s,termlen);
11793 sv_catpvn(sv, s, termlen);
11798 if (PL_madskills) {
11799 char * const tstart = SvPVX(PL_linestr) + stuffstart;
11800 const int len = s - tstart - termlen;
11802 sv_catpvn(PL_thisstuff, tstart, len);
11804 PL_thisstuff = newSVpvn(tstart, len);
11805 if (!PL_thisclose && !keep_delims)
11806 PL_thisclose = newSVpvn(s - termlen,termlen);
11810 if (has_utf8 || PL_encoding)
11813 PL_multi_end = CopLINE(PL_curcop);
11815 /* if we allocated too much space, give some back */
11816 if (SvCUR(sv) + 5 < SvLEN(sv)) {
11817 SvLEN_set(sv, SvCUR(sv) + 1);
11818 SvPV_renew(sv, SvLEN(sv));
11821 /* decide whether this is the first or second quoted string we've read
11834 takes: pointer to position in buffer
11835 returns: pointer to new position in buffer
11836 side-effects: builds ops for the constant in yylval.op
11838 Read a number in any of the formats that Perl accepts:
11840 \d(_?\d)*(\.(\d(_?\d)*)?)?[Ee][\+\-]?(\d(_?\d)*) 12 12.34 12.
11841 \.\d(_?\d)*[Ee][\+\-]?(\d(_?\d)*) .34
11844 0x[0-9A-Fa-f](_?[0-9A-Fa-f])*
11846 Like most scan_ routines, it uses the PL_tokenbuf buffer to hold the
11849 If it reads a number without a decimal point or an exponent, it will
11850 try converting the number to an integer and see if it can do so
11851 without loss of precision.
11855 Perl_scan_num(pTHX_ const char *start, YYSTYPE* lvalp)
11858 register const char *s = start; /* current position in buffer */
11859 register char *d; /* destination in temp buffer */
11860 register char *e; /* end of temp buffer */
11861 NV nv; /* number read, as a double */
11862 SV *sv = NULL; /* place to put the converted number */
11863 bool floatit; /* boolean: int or float? */
11864 const char *lastub = NULL; /* position of last underbar */
11865 static char const number_too_long[] = "Number too long";
11867 /* We use the first character to decide what type of number this is */
11871 Perl_croak(aTHX_ "panic: scan_num");
11873 /* if it starts with a 0, it could be an octal number, a decimal in
11874 0.13 disguise, or a hexadecimal number, or a binary number. */
11878 u holds the "number so far"
11879 shift the power of 2 of the base
11880 (hex == 4, octal == 3, binary == 1)
11881 overflowed was the number more than we can hold?
11883 Shift is used when we add a digit. It also serves as an "are
11884 we in octal/hex/binary?" indicator to disallow hex characters
11885 when in octal mode.
11890 bool overflowed = FALSE;
11891 bool just_zero = TRUE; /* just plain 0 or binary number? */
11892 static const NV nvshift[5] = { 1.0, 2.0, 4.0, 8.0, 16.0 };
11893 static const char* const bases[5] =
11894 { "", "binary", "", "octal", "hexadecimal" };
11895 static const char* const Bases[5] =
11896 { "", "Binary", "", "Octal", "Hexadecimal" };
11897 static const char* const maxima[5] =
11899 "0b11111111111111111111111111111111",
11903 const char *base, *Base, *max;
11905 /* check for hex */
11910 } else if (s[1] == 'b') {
11915 /* check for a decimal in disguise */
11916 else if (s[1] == '.' || s[1] == 'e' || s[1] == 'E')
11918 /* so it must be octal */
11925 if (ckWARN(WARN_SYNTAX))
11926 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
11927 "Misplaced _ in number");
11931 base = bases[shift];
11932 Base = Bases[shift];
11933 max = maxima[shift];
11935 /* read the rest of the number */
11937 /* x is used in the overflow test,
11938 b is the digit we're adding on. */
11943 /* if we don't mention it, we're done */
11947 /* _ are ignored -- but warned about if consecutive */
11949 if (lastub && s == lastub + 1 && ckWARN(WARN_SYNTAX))
11950 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
11951 "Misplaced _ in number");
11955 /* 8 and 9 are not octal */
11956 case '8': case '9':
11958 yyerror(Perl_form(aTHX_ "Illegal octal digit '%c'", *s));
11962 case '2': case '3': case '4':
11963 case '5': case '6': case '7':
11965 yyerror(Perl_form(aTHX_ "Illegal binary digit '%c'", *s));
11968 case '0': case '1':
11969 b = *s++ & 15; /* ASCII digit -> value of digit */
11973 case 'a': case 'b': case 'c': case 'd': case 'e': case 'f':
11974 case 'A': case 'B': case 'C': case 'D': case 'E': case 'F':
11975 /* make sure they said 0x */
11978 b = (*s++ & 7) + 9;
11980 /* Prepare to put the digit we have onto the end
11981 of the number so far. We check for overflows.
11987 x = u << shift; /* make room for the digit */
11989 if ((x >> shift) != u
11990 && !(PL_hints & HINT_NEW_BINARY)) {
11993 if (ckWARN_d(WARN_OVERFLOW))
11994 Perl_warner(aTHX_ packWARN(WARN_OVERFLOW),
11995 "Integer overflow in %s number",
11998 u = x | b; /* add the digit to the end */
12001 n *= nvshift[shift];
12002 /* If an NV has not enough bits in its
12003 * mantissa to represent an UV this summing of
12004 * small low-order numbers is a waste of time
12005 * (because the NV cannot preserve the
12006 * low-order bits anyway): we could just
12007 * remember when did we overflow and in the
12008 * end just multiply n by the right
12016 /* if we get here, we had success: make a scalar value from
12021 /* final misplaced underbar check */
12022 if (s[-1] == '_') {
12023 if (ckWARN(WARN_SYNTAX))
12024 Perl_warner(aTHX_ packWARN(WARN_SYNTAX), "Misplaced _ in number");
12029 if (n > 4294967295.0 && ckWARN(WARN_PORTABLE))
12030 Perl_warner(aTHX_ packWARN(WARN_PORTABLE),
12031 "%s number > %s non-portable",
12037 if (u > 0xffffffff && ckWARN(WARN_PORTABLE))
12038 Perl_warner(aTHX_ packWARN(WARN_PORTABLE),
12039 "%s number > %s non-portable",
12044 if (just_zero && (PL_hints & HINT_NEW_INTEGER))
12045 sv = new_constant(start, s - start, "integer",
12047 else if (PL_hints & HINT_NEW_BINARY)
12048 sv = new_constant(start, s - start, "binary", sv, NULL, NULL);
12053 handle decimal numbers.
12054 we're also sent here when we read a 0 as the first digit
12056 case '1': case '2': case '3': case '4': case '5':
12057 case '6': case '7': case '8': case '9': case '.':
12060 e = PL_tokenbuf + sizeof PL_tokenbuf - 6; /* room for various punctuation */
12063 /* read next group of digits and _ and copy into d */
12064 while (isDIGIT(*s) || *s == '_') {
12065 /* skip underscores, checking for misplaced ones
12069 if (lastub && s == lastub + 1 && ckWARN(WARN_SYNTAX))
12070 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
12071 "Misplaced _ in number");
12075 /* check for end of fixed-length buffer */
12077 Perl_croak(aTHX_ number_too_long);
12078 /* if we're ok, copy the character */
12083 /* final misplaced underbar check */
12084 if (lastub && s == lastub + 1) {
12085 if (ckWARN(WARN_SYNTAX))
12086 Perl_warner(aTHX_ packWARN(WARN_SYNTAX), "Misplaced _ in number");
12089 /* read a decimal portion if there is one. avoid
12090 3..5 being interpreted as the number 3. followed
12093 if (*s == '.' && s[1] != '.') {
12098 if (ckWARN(WARN_SYNTAX))
12099 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
12100 "Misplaced _ in number");
12104 /* copy, ignoring underbars, until we run out of digits.
12106 for (; isDIGIT(*s) || *s == '_'; s++) {
12107 /* fixed length buffer check */
12109 Perl_croak(aTHX_ number_too_long);
12111 if (lastub && s == lastub + 1 && ckWARN(WARN_SYNTAX))
12112 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
12113 "Misplaced _ in number");
12119 /* fractional part ending in underbar? */
12120 if (s[-1] == '_') {
12121 if (ckWARN(WARN_SYNTAX))
12122 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
12123 "Misplaced _ in number");
12125 if (*s == '.' && isDIGIT(s[1])) {
12126 /* oops, it's really a v-string, but without the "v" */
12132 /* read exponent part, if present */
12133 if ((*s == 'e' || *s == 'E') && strchr("+-0123456789_", s[1])) {
12137 /* regardless of whether user said 3E5 or 3e5, use lower 'e' */
12138 *d++ = 'e'; /* At least some Mach atof()s don't grok 'E' */
12140 /* stray preinitial _ */
12142 if (ckWARN(WARN_SYNTAX))
12143 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
12144 "Misplaced _ in number");
12148 /* allow positive or negative exponent */
12149 if (*s == '+' || *s == '-')
12152 /* stray initial _ */
12154 if (ckWARN(WARN_SYNTAX))
12155 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
12156 "Misplaced _ in number");
12160 /* read digits of exponent */
12161 while (isDIGIT(*s) || *s == '_') {
12164 Perl_croak(aTHX_ number_too_long);
12168 if (((lastub && s == lastub + 1) ||
12169 (!isDIGIT(s[1]) && s[1] != '_'))
12170 && ckWARN(WARN_SYNTAX))
12171 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
12172 "Misplaced _ in number");
12179 /* make an sv from the string */
12183 We try to do an integer conversion first if no characters
12184 indicating "float" have been found.
12189 const int flags = grok_number (PL_tokenbuf, d - PL_tokenbuf, &uv);
12191 if (flags == IS_NUMBER_IN_UV) {
12193 sv_setiv(sv, uv); /* Prefer IVs over UVs. */
12196 } else if (flags == (IS_NUMBER_IN_UV | IS_NUMBER_NEG)) {
12197 if (uv <= (UV) IV_MIN)
12198 sv_setiv(sv, -(IV)uv);
12205 /* terminate the string */
12207 nv = Atof(PL_tokenbuf);
12211 if ( floatit ? (PL_hints & HINT_NEW_FLOAT) :
12212 (PL_hints & HINT_NEW_INTEGER) )
12213 sv = new_constant(PL_tokenbuf,
12216 (floatit ? "float" : "integer"),
12220 /* if it starts with a v, it could be a v-string */
12223 sv = newSV(5); /* preallocate storage space */
12224 s = scan_vstring(s,sv);
12228 /* make the op for the constant and return */
12231 lvalp->opval = newSVOP(OP_CONST, 0, sv);
12233 lvalp->opval = NULL;
12239 S_scan_formline(pTHX_ register char *s)
12242 register char *eol;
12244 SV * const stuff = newSVpvs("");
12245 bool needargs = FALSE;
12246 bool eofmt = FALSE;
12248 char *tokenstart = s;
12251 if (PL_madskills) {
12252 savewhite = PL_thiswhite;
12257 while (!needargs) {
12260 #ifdef PERL_STRICT_CR
12261 while (SPACE_OR_TAB(*t))
12264 while (SPACE_OR_TAB(*t) || *t == '\r')
12267 if (*t == '\n' || t == PL_bufend) {
12272 if (PL_in_eval && !PL_rsfp) {
12273 eol = (char *) memchr(s,'\n',PL_bufend-s);
12278 eol = PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
12280 for (t = s; t < eol; t++) {
12281 if (*t == '~' && t[1] == '~' && SvCUR(stuff)) {
12283 goto enough; /* ~~ must be first line in formline */
12285 if (*t == '@' || *t == '^')
12289 sv_catpvn(stuff, s, eol-s);
12290 #ifndef PERL_STRICT_CR
12291 if (eol-s > 1 && eol[-2] == '\r' && eol[-1] == '\n') {
12292 char *end = SvPVX(stuff) + SvCUR(stuff);
12295 SvCUR_set(stuff, SvCUR(stuff) - 1);
12305 if (PL_madskills) {
12307 sv_catpvn(PL_thistoken, tokenstart, PL_bufend - tokenstart);
12309 PL_thistoken = newSVpvn(tokenstart, PL_bufend - tokenstart);
12312 s = filter_gets(PL_linestr, PL_rsfp, 0);
12314 tokenstart = PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = SvPVX(PL_linestr);
12316 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = SvPVX(PL_linestr);
12318 PL_bufend = PL_bufptr + SvCUR(PL_linestr);
12319 PL_last_lop = PL_last_uni = NULL;
12328 if (SvCUR(stuff)) {
12331 PL_lex_state = LEX_NORMAL;
12332 start_force(PL_curforce);
12333 NEXTVAL_NEXTTOKE.ival = 0;
12337 PL_lex_state = LEX_FORMLINE;
12339 if (UTF && is_utf8_string((U8*)SvPVX_const(stuff), SvCUR(stuff)))
12341 else if (PL_encoding)
12342 sv_recode_to_utf8(stuff, PL_encoding);
12344 start_force(PL_curforce);
12345 NEXTVAL_NEXTTOKE.opval = (OP*)newSVOP(OP_CONST, 0, stuff);
12347 start_force(PL_curforce);
12348 NEXTVAL_NEXTTOKE.ival = OP_FORMLINE;
12352 SvREFCNT_dec(stuff);
12354 PL_lex_formbrack = 0;
12358 if (PL_madskills) {
12360 sv_catpvn(PL_thistoken, tokenstart, s - tokenstart);
12362 PL_thistoken = newSVpvn(tokenstart, s - tokenstart);
12363 PL_thiswhite = savewhite;
12375 PL_cshlen = strlen(PL_cshname);
12377 #if defined(USE_ITHREADS)
12378 PERL_UNUSED_CONTEXT;
12384 Perl_start_subparse(pTHX_ I32 is_format, U32 flags)
12387 const I32 oldsavestack_ix = PL_savestack_ix;
12388 CV* const outsidecv = PL_compcv;
12391 assert(SvTYPE(PL_compcv) == SVt_PVCV);
12393 SAVEI32(PL_subline);
12394 save_item(PL_subname);
12395 SAVESPTR(PL_compcv);
12397 PL_compcv = (CV*)newSV(0);
12398 sv_upgrade((SV *)PL_compcv, is_format ? SVt_PVFM : SVt_PVCV);
12399 CvFLAGS(PL_compcv) |= flags;
12401 PL_subline = CopLINE(PL_curcop);
12402 CvPADLIST(PL_compcv) = pad_new(padnew_SAVE|padnew_SAVESUB);
12403 CvOUTSIDE(PL_compcv) = (CV*)SvREFCNT_inc_simple(outsidecv);
12404 CvOUTSIDE_SEQ(PL_compcv) = PL_cop_seqmax;
12406 return oldsavestack_ix;
12410 #pragma segment Perl_yylex
12413 Perl_yywarn(pTHX_ const char *s)
12416 PL_in_eval |= EVAL_WARNONLY;
12418 PL_in_eval &= ~EVAL_WARNONLY;
12423 Perl_yyerror(pTHX_ const char *s)
12426 const char *where = NULL;
12427 const char *context = NULL;
12430 int yychar = PL_parser->yychar;
12432 if (!yychar || (yychar == ';' && !PL_rsfp))
12434 else if (PL_oldoldbufptr && PL_bufptr > PL_oldoldbufptr &&
12435 PL_bufptr - PL_oldoldbufptr < 200 && PL_oldoldbufptr != PL_oldbufptr &&
12436 PL_oldbufptr != PL_bufptr) {
12439 The code below is removed for NetWare because it abends/crashes on NetWare
12440 when the script has error such as not having the closing quotes like:
12441 if ($var eq "value)
12442 Checking of white spaces is anyway done in NetWare code.
12445 while (isSPACE(*PL_oldoldbufptr))
12448 context = PL_oldoldbufptr;
12449 contlen = PL_bufptr - PL_oldoldbufptr;
12451 else if (PL_oldbufptr && PL_bufptr > PL_oldbufptr &&
12452 PL_bufptr - PL_oldbufptr < 200 && PL_oldbufptr != PL_bufptr) {
12455 The code below is removed for NetWare because it abends/crashes on NetWare
12456 when the script has error such as not having the closing quotes like:
12457 if ($var eq "value)
12458 Checking of white spaces is anyway done in NetWare code.
12461 while (isSPACE(*PL_oldbufptr))
12464 context = PL_oldbufptr;
12465 contlen = PL_bufptr - PL_oldbufptr;
12467 else if (yychar > 255)
12468 where = "next token ???";
12469 else if (yychar == -2) { /* YYEMPTY */
12470 if (PL_lex_state == LEX_NORMAL ||
12471 (PL_lex_state == LEX_KNOWNEXT && PL_lex_defer == LEX_NORMAL))
12472 where = "at end of line";
12473 else if (PL_lex_inpat)
12474 where = "within pattern";
12476 where = "within string";
12479 SV * const where_sv = sv_2mortal(newSVpvs("next char "));
12481 Perl_sv_catpvf(aTHX_ where_sv, "^%c", toCTRL(yychar));
12482 else if (isPRINT_LC(yychar))
12483 Perl_sv_catpvf(aTHX_ where_sv, "%c", yychar);
12485 Perl_sv_catpvf(aTHX_ where_sv, "\\%03o", yychar & 255);
12486 where = SvPVX_const(where_sv);
12488 msg = sv_2mortal(newSVpv(s, 0));
12489 Perl_sv_catpvf(aTHX_ msg, " at %s line %"IVdf", ",
12490 OutCopFILE(PL_curcop), (IV)CopLINE(PL_curcop));
12492 Perl_sv_catpvf(aTHX_ msg, "near \"%.*s\"\n", contlen, context);
12494 Perl_sv_catpvf(aTHX_ msg, "%s\n", where);
12495 if (PL_multi_start < PL_multi_end && (U32)(CopLINE(PL_curcop) - PL_multi_end) <= 1) {
12496 Perl_sv_catpvf(aTHX_ msg,
12497 " (Might be a runaway multi-line %c%c string starting on line %"IVdf")\n",
12498 (int)PL_multi_open,(int)PL_multi_close,(IV)PL_multi_start);
12501 if (PL_in_eval & EVAL_WARNONLY && ckWARN_d(WARN_SYNTAX))
12502 Perl_warner(aTHX_ packWARN(WARN_SYNTAX), "%"SVf, (void*)msg);
12505 if (PL_error_count >= 10) {
12506 if (PL_in_eval && SvCUR(ERRSV))
12507 Perl_croak(aTHX_ "%"SVf"%s has too many errors.\n",
12508 (void*)ERRSV, OutCopFILE(PL_curcop));
12510 Perl_croak(aTHX_ "%s has too many errors.\n",
12511 OutCopFILE(PL_curcop));
12514 PL_in_my_stash = NULL;
12518 #pragma segment Main
12522 S_swallow_bom(pTHX_ U8 *s)
12525 const STRLEN slen = SvCUR(PL_linestr);
12528 if (s[1] == 0xFE) {
12529 /* UTF-16 little-endian? (or UTF32-LE?) */
12530 if (s[2] == 0 && s[3] == 0) /* UTF-32 little-endian */
12531 Perl_croak(aTHX_ "Unsupported script encoding UTF32-LE");
12532 #ifndef PERL_NO_UTF16_FILTER
12533 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF16-LE script encoding (BOM)\n");
12536 if (PL_bufend > (char*)s) {
12540 filter_add(utf16rev_textfilter, NULL);
12541 Newx(news, (PL_bufend - (char*)s) * 3 / 2 + 1, U8);
12542 utf16_to_utf8_reversed(s, news,
12543 PL_bufend - (char*)s - 1,
12545 sv_setpvn(PL_linestr, (const char*)news, newlen);
12547 s = (U8*)SvPVX(PL_linestr);
12548 Copy(news, s, newlen, U8);
12552 SvUTF8_on(PL_linestr);
12553 s = (U8*)SvPVX(PL_linestr);
12555 /* FIXME - is this a general bug fix? */
12558 PL_bufend = SvPVX(PL_linestr) + newlen;
12561 Perl_croak(aTHX_ "Unsupported script encoding UTF16-LE");
12566 if (s[1] == 0xFF) { /* UTF-16 big-endian? */
12567 #ifndef PERL_NO_UTF16_FILTER
12568 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16BE script encoding (BOM)\n");
12571 if (PL_bufend > (char *)s) {
12575 filter_add(utf16_textfilter, NULL);
12576 Newx(news, (PL_bufend - (char*)s) * 3 / 2 + 1, U8);
12577 utf16_to_utf8(s, news,
12578 PL_bufend - (char*)s,
12580 sv_setpvn(PL_linestr, (const char*)news, newlen);
12582 SvUTF8_on(PL_linestr);
12583 s = (U8*)SvPVX(PL_linestr);
12584 PL_bufend = SvPVX(PL_linestr) + newlen;
12587 Perl_croak(aTHX_ "Unsupported script encoding UTF16-BE");
12592 if (slen > 2 && s[1] == 0xBB && s[2] == 0xBF) {
12593 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-8 script encoding (BOM)\n");
12594 s += 3; /* UTF-8 */
12600 if (s[2] == 0xFE && s[3] == 0xFF) {
12601 /* UTF-32 big-endian */
12602 Perl_croak(aTHX_ "Unsupported script encoding UTF32-BE");
12605 else if (s[2] == 0 && s[3] != 0) {
12608 * are a good indicator of UTF-16BE. */
12609 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16BE script encoding (no BOM)\n");
12615 if (slen > 3 && s[1] == 0x73 && s[2] == 0x66 && s[3] == 0x73) {
12616 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-8 script encoding (BOM)\n");
12617 s += 4; /* UTF-8 */
12623 if (slen > 3 && s[1] == 0 && s[2] != 0 && s[3] == 0) {
12626 * are a good indicator of UTF-16LE. */
12627 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16LE script encoding (no BOM)\n");
12636 * Restore a source filter.
12640 restore_rsfp(pTHX_ void *f)
12643 PerlIO * const fp = (PerlIO*)f;
12645 if (PL_rsfp == PerlIO_stdin())
12646 PerlIO_clearerr(PL_rsfp);
12647 else if (PL_rsfp && (PL_rsfp != fp))
12648 PerlIO_close(PL_rsfp);
12652 #ifndef PERL_NO_UTF16_FILTER
12654 utf16_textfilter(pTHX_ int idx, SV *sv, int maxlen)
12657 const STRLEN old = SvCUR(sv);
12658 const I32 count = FILTER_READ(idx+1, sv, maxlen);
12659 DEBUG_P(PerlIO_printf(Perl_debug_log,
12660 "utf16_textfilter(%p): %d %d (%d)\n",
12661 FPTR2DPTR(void *, utf16_textfilter),
12662 idx, maxlen, (int) count));
12666 Newx(tmps, SvCUR(sv) * 3 / 2 + 1, U8);
12667 Copy(SvPVX_const(sv), tmps, old, char);
12668 utf16_to_utf8((U8*)SvPVX_const(sv) + old, tmps + old,
12669 SvCUR(sv) - old, &newlen);
12670 sv_usepvn(sv, (char*)tmps, (STRLEN)newlen + old);
12672 DEBUG_P({sv_dump(sv);});
12677 utf16rev_textfilter(pTHX_ int idx, SV *sv, int maxlen)
12680 const STRLEN old = SvCUR(sv);
12681 const I32 count = FILTER_READ(idx+1, sv, maxlen);
12682 DEBUG_P(PerlIO_printf(Perl_debug_log,
12683 "utf16rev_textfilter(%p): %d %d (%d)\n",
12684 FPTR2DPTR(void *, utf16rev_textfilter),
12685 idx, maxlen, (int) count));
12689 Newx(tmps, SvCUR(sv) * 3 / 2 + 1, U8);
12690 Copy(SvPVX_const(sv), tmps, old, char);
12691 utf16_to_utf8((U8*)SvPVX_const(sv) + old, tmps + old,
12692 SvCUR(sv) - old, &newlen);
12693 sv_usepvn(sv, (char*)tmps, (STRLEN)newlen + old);
12695 DEBUG_P({ sv_dump(sv); });
12701 Returns a pointer to the next character after the parsed
12702 vstring, as well as updating the passed in sv.
12704 Function must be called like
12707 s = scan_vstring(s,sv);
12709 The sv should already be large enough to store the vstring
12710 passed in, for performance reasons.
12715 Perl_scan_vstring(pTHX_ const char *s, SV *sv)
12718 const char *pos = s;
12719 const char *start = s;
12720 if (*pos == 'v') pos++; /* get past 'v' */
12721 while (pos < PL_bufend && (isDIGIT(*pos) || *pos == '_'))
12723 if ( *pos != '.') {
12724 /* this may not be a v-string if followed by => */
12725 const char *next = pos;
12726 while (next < PL_bufend && isSPACE(*next))
12728 if ((PL_bufend - next) >= 2 && *next == '=' && next[1] == '>' ) {
12729 /* return string not v-string */
12730 sv_setpvn(sv,(char *)s,pos-s);
12731 return (char *)pos;
12735 if (!isALPHA(*pos)) {
12736 U8 tmpbuf[UTF8_MAXBYTES+1];
12739 s++; /* get past 'v' */
12741 sv_setpvn(sv, "", 0);
12744 /* this is atoi() that tolerates underscores */
12747 const char *end = pos;
12749 while (--end >= s) {
12751 const UV orev = rev;
12752 rev += (*end - '0') * mult;
12754 if (orev > rev && ckWARN_d(WARN_OVERFLOW))
12755 Perl_warner(aTHX_ packWARN(WARN_OVERFLOW),
12756 "Integer overflow in decimal number");
12760 if (rev > 0x7FFFFFFF)
12761 Perl_croak(aTHX_ "In EBCDIC the v-string components cannot exceed 2147483647");
12763 /* Append native character for the rev point */
12764 tmpend = uvchr_to_utf8(tmpbuf, rev);
12765 sv_catpvn(sv, (const char*)tmpbuf, tmpend - tmpbuf);
12766 if (!UNI_IS_INVARIANT(NATIVE_TO_UNI(rev)))
12768 if (pos + 1 < PL_bufend && *pos == '.' && isDIGIT(pos[1]))
12774 while (pos < PL_bufend && (isDIGIT(*pos) || *pos == '_'))
12778 sv_magic(sv,NULL,PERL_MAGIC_vstring,(const char*)start, pos-start);
12786 * c-indentation-style: bsd
12787 * c-basic-offset: 4
12788 * indent-tabs-mode: t
12791 * ex: set ts=8 sts=4 sw=4 noet: