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 yychar (*PL_yycharp)
27 #define yylval (*PL_yylvalp)
29 static const char ident_too_long[] = "Identifier too long";
30 static const char commaless_variable_list[] = "comma-less variable list";
32 static void restore_rsfp(pTHX_ void *f);
33 #ifndef PERL_NO_UTF16_FILTER
34 static I32 utf16_textfilter(pTHX_ int idx, SV *sv, int maxlen);
35 static I32 utf16rev_textfilter(pTHX_ int idx, SV *sv, int maxlen);
39 /* XXX these probably need to be made into PL vars */
40 static I32 realtokenstart;
41 static I32 faketokens = 0;
42 static MADPROP *thismad;
51 static I32 curforce = -1;
53 # define CURMAD(slot,sv) if (PL_madskills) { curmad(slot,sv); sv = 0; }
54 # define NEXTVAL_NEXTTOKE PL_nexttoke[curforce].next_val
56 # define CURMAD(slot,sv)
57 # define NEXTVAL_NEXTTOKE PL_nextval[PL_nexttoke]
60 #define XFAKEBRACK 128
63 #ifdef USE_UTF8_SCRIPTS
64 # define UTF (!IN_BYTES)
66 # define UTF ((PL_linestr && DO_UTF8(PL_linestr)) || (PL_hints & HINT_UTF8))
69 /* In variables named $^X, these are the legal values for X.
70 * 1999-02-27 mjd-perl-patch@plover.com */
71 #define isCONTROLVAR(x) (isUPPER(x) || strchr("[\\]^_?", (x)))
73 /* On MacOS, respect nonbreaking spaces */
74 #ifdef MACOS_TRADITIONAL
75 #define SPACE_OR_TAB(c) ((c)==' '||(c)=='\312'||(c)=='\t')
77 #define SPACE_OR_TAB(c) ((c)==' '||(c)=='\t')
80 /* LEX_* are values for PL_lex_state, the state of the lexer.
81 * They are arranged oddly so that the guard on the switch statement
82 * can get by with a single comparison (if the compiler is smart enough).
85 /* #define LEX_NOTPARSING 11 is done in perl.h. */
87 #define LEX_NORMAL 10 /* normal code (ie not within "...") */
88 #define LEX_INTERPNORMAL 9 /* code within a string, eg "$foo[$x+1]" */
89 #define LEX_INTERPCASEMOD 8 /* expecting a \U, \Q or \E etc */
90 #define LEX_INTERPPUSH 7 /* starting a new sublex parse level */
91 #define LEX_INTERPSTART 6 /* expecting the start of a $var */
93 /* at end of code, eg "$x" followed by: */
94 #define LEX_INTERPEND 5 /* ... eg not one of [, { or -> */
95 #define LEX_INTERPENDMAYBE 4 /* ... eg one of [, { or -> */
97 #define LEX_INTERPCONCAT 3 /* expecting anything, eg at start of
98 string or after \E, $foo, etc */
99 #define LEX_INTERPCONST 2 /* NOT USED */
100 #define LEX_FORMLINE 1 /* expecting a format line */
101 #define LEX_KNOWNEXT 0 /* next token known; just return it */
105 static const char* const lex_state_names[] = {
124 #include "keywords.h"
126 /* CLINE is a macro that ensures PL_copline has a sane value */
131 #define CLINE (PL_copline = (CopLINE(PL_curcop) < PL_copline ? CopLINE(PL_curcop) : PL_copline))
134 # define SKIPSPACE0(s) skipspace0(s)
135 # define SKIPSPACE1(s) skipspace1(s)
136 # define SKIPSPACE2(s,tsv) skipspace2(s,&tsv)
137 # define PEEKSPACE(s) skipspace2(s,0)
139 # define SKIPSPACE0(s) skipspace(s)
140 # define SKIPSPACE1(s) skipspace(s)
141 # define SKIPSPACE2(s,tsv) skipspace(s)
142 # define PEEKSPACE(s) skipspace(s)
146 * Convenience functions to return different tokens and prime the
147 * lexer for the next token. They all take an argument.
149 * TOKEN : generic token (used for '(', DOLSHARP, etc)
150 * OPERATOR : generic operator
151 * AOPERATOR : assignment operator
152 * PREBLOCK : beginning the block after an if, while, foreach, ...
153 * PRETERMBLOCK : beginning a non-code-defining {} block (eg, hash ref)
154 * PREREF : *EXPR where EXPR is not a simple identifier
155 * TERM : expression term
156 * LOOPX : loop exiting command (goto, last, dump, etc)
157 * FTST : file test operator
158 * FUN0 : zero-argument function
159 * FUN1 : not used, except for not, which isn't a UNIOP
160 * BOop : bitwise or or xor
162 * SHop : shift operator
163 * PWop : power operator
164 * PMop : pattern-matching operator
165 * Aop : addition-level operator
166 * Mop : multiplication-level operator
167 * Eop : equality-testing operator
168 * Rop : relational operator <= != gt
170 * Also see LOP and lop() below.
173 #ifdef DEBUGGING /* Serve -DT. */
174 # define REPORT(retval) tokereport((I32)retval)
176 # define REPORT(retval) (retval)
179 #define TOKEN(retval) return ( PL_bufptr = s, REPORT(retval))
180 #define OPERATOR(retval) return (PL_expect = XTERM, PL_bufptr = s, REPORT(retval))
181 #define AOPERATOR(retval) return ao((PL_expect = XTERM, PL_bufptr = s, REPORT(retval)))
182 #define PREBLOCK(retval) return (PL_expect = XBLOCK,PL_bufptr = s, REPORT(retval))
183 #define PRETERMBLOCK(retval) return (PL_expect = XTERMBLOCK,PL_bufptr = s, REPORT(retval))
184 #define PREREF(retval) return (PL_expect = XREF,PL_bufptr = s, REPORT(retval))
185 #define TERM(retval) return (CLINE, PL_expect = XOPERATOR, PL_bufptr = s, REPORT(retval))
186 #define LOOPX(f) return (yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)LOOPEX))
187 #define FTST(f) return (yylval.ival=f, PL_expect=XTERMORDORDOR, PL_bufptr=s, REPORT((int)UNIOP))
188 #define FUN0(f) return (yylval.ival=f, PL_expect=XOPERATOR, PL_bufptr=s, REPORT((int)FUNC0))
189 #define FUN1(f) return (yylval.ival=f, PL_expect=XOPERATOR, PL_bufptr=s, REPORT((int)FUNC1))
190 #define BOop(f) return ao((yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)BITOROP)))
191 #define BAop(f) return ao((yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)BITANDOP)))
192 #define SHop(f) return ao((yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)SHIFTOP)))
193 #define PWop(f) return ao((yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)POWOP)))
194 #define PMop(f) return(yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)MATCHOP))
195 #define Aop(f) return ao((yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)ADDOP)))
196 #define Mop(f) return ao((yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)MULOP)))
197 #define Eop(f) return (yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)EQOP))
198 #define Rop(f) return (yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)RELOP))
200 /* This bit of chicanery makes a unary function followed by
201 * a parenthesis into a function with one argument, highest precedence.
202 * The UNIDOR macro is for unary functions that can be followed by the //
203 * operator (such as C<shift // 0>).
205 #define UNI2(f,x) { \
209 PL_last_uni = PL_oldbufptr; \
210 PL_last_lop_op = f; \
212 return REPORT( (int)FUNC1 ); \
214 return REPORT( *s=='(' ? (int)FUNC1 : (int)UNIOP ); \
216 #define UNI(f) UNI2(f,XTERM)
217 #define UNIDOR(f) UNI2(f,XTERMORDORDOR)
219 #define UNIBRACK(f) { \
222 PL_last_uni = PL_oldbufptr; \
224 return REPORT( (int)FUNC1 ); \
226 return REPORT( (*s == '(') ? (int)FUNC1 : (int)UNIOP ); \
229 /* grandfather return to old style */
230 #define OLDLOP(f) return(yylval.ival=f,PL_expect = XTERM,PL_bufptr = s,(int)LSTOP)
234 /* how to interpret the yylval associated with the token */
238 TOKENTYPE_OPNUM, /* yylval.ival contains an opcode number */
244 static struct debug_tokens {
246 enum token_type type;
248 } const debug_tokens[] =
250 { ADDOP, TOKENTYPE_OPNUM, "ADDOP" },
251 { ANDAND, TOKENTYPE_NONE, "ANDAND" },
252 { ANDOP, TOKENTYPE_NONE, "ANDOP" },
253 { ANONSUB, TOKENTYPE_IVAL, "ANONSUB" },
254 { ARROW, TOKENTYPE_NONE, "ARROW" },
255 { ASSIGNOP, TOKENTYPE_OPNUM, "ASSIGNOP" },
256 { BITANDOP, TOKENTYPE_OPNUM, "BITANDOP" },
257 { BITOROP, TOKENTYPE_OPNUM, "BITOROP" },
258 { COLONATTR, TOKENTYPE_NONE, "COLONATTR" },
259 { CONTINUE, TOKENTYPE_NONE, "CONTINUE" },
260 { DEFAULT, TOKENTYPE_NONE, "DEFAULT" },
261 { DO, TOKENTYPE_NONE, "DO" },
262 { DOLSHARP, TOKENTYPE_NONE, "DOLSHARP" },
263 { DORDOR, TOKENTYPE_NONE, "DORDOR" },
264 { DOROP, TOKENTYPE_OPNUM, "DOROP" },
265 { DOTDOT, TOKENTYPE_IVAL, "DOTDOT" },
266 { ELSE, TOKENTYPE_NONE, "ELSE" },
267 { ELSIF, TOKENTYPE_IVAL, "ELSIF" },
268 { EQOP, TOKENTYPE_OPNUM, "EQOP" },
269 { FOR, TOKENTYPE_IVAL, "FOR" },
270 { FORMAT, TOKENTYPE_NONE, "FORMAT" },
271 { FUNC, TOKENTYPE_OPNUM, "FUNC" },
272 { FUNC0, TOKENTYPE_OPNUM, "FUNC0" },
273 { FUNC0SUB, TOKENTYPE_OPVAL, "FUNC0SUB" },
274 { FUNC1, TOKENTYPE_OPNUM, "FUNC1" },
275 { FUNCMETH, TOKENTYPE_OPVAL, "FUNCMETH" },
276 { GIVEN, TOKENTYPE_IVAL, "GIVEN" },
277 { HASHBRACK, TOKENTYPE_NONE, "HASHBRACK" },
278 { IF, TOKENTYPE_IVAL, "IF" },
279 { LABEL, TOKENTYPE_PVAL, "LABEL" },
280 { LOCAL, TOKENTYPE_IVAL, "LOCAL" },
281 { LOOPEX, TOKENTYPE_OPNUM, "LOOPEX" },
282 { LSTOP, TOKENTYPE_OPNUM, "LSTOP" },
283 { LSTOPSUB, TOKENTYPE_OPVAL, "LSTOPSUB" },
284 { MATCHOP, TOKENTYPE_OPNUM, "MATCHOP" },
285 { METHOD, TOKENTYPE_OPVAL, "METHOD" },
286 { MULOP, TOKENTYPE_OPNUM, "MULOP" },
287 { MY, TOKENTYPE_IVAL, "MY" },
288 { MYSUB, TOKENTYPE_NONE, "MYSUB" },
289 { NOAMP, TOKENTYPE_NONE, "NOAMP" },
290 { NOTOP, TOKENTYPE_NONE, "NOTOP" },
291 { OROP, TOKENTYPE_IVAL, "OROP" },
292 { OROR, TOKENTYPE_NONE, "OROR" },
293 { PACKAGE, TOKENTYPE_NONE, "PACKAGE" },
294 { PMFUNC, TOKENTYPE_OPVAL, "PMFUNC" },
295 { POSTDEC, TOKENTYPE_NONE, "POSTDEC" },
296 { POSTINC, TOKENTYPE_NONE, "POSTINC" },
297 { POWOP, TOKENTYPE_OPNUM, "POWOP" },
298 { PREDEC, TOKENTYPE_NONE, "PREDEC" },
299 { PREINC, TOKENTYPE_NONE, "PREINC" },
300 { PRIVATEREF, TOKENTYPE_OPVAL, "PRIVATEREF" },
301 { REFGEN, TOKENTYPE_NONE, "REFGEN" },
302 { RELOP, TOKENTYPE_OPNUM, "RELOP" },
303 { SHIFTOP, TOKENTYPE_OPNUM, "SHIFTOP" },
304 { SUB, TOKENTYPE_NONE, "SUB" },
305 { THING, TOKENTYPE_OPVAL, "THING" },
306 { UMINUS, TOKENTYPE_NONE, "UMINUS" },
307 { UNIOP, TOKENTYPE_OPNUM, "UNIOP" },
308 { UNIOPSUB, TOKENTYPE_OPVAL, "UNIOPSUB" },
309 { UNLESS, TOKENTYPE_IVAL, "UNLESS" },
310 { UNTIL, TOKENTYPE_IVAL, "UNTIL" },
311 { USE, TOKENTYPE_IVAL, "USE" },
312 { WHEN, TOKENTYPE_IVAL, "WHEN" },
313 { WHILE, TOKENTYPE_IVAL, "WHILE" },
314 { WORD, TOKENTYPE_OPVAL, "WORD" },
315 { 0, TOKENTYPE_NONE, 0 }
318 /* dump the returned token in rv, plus any optional arg in yylval */
321 S_tokereport(pTHX_ I32 rv)
325 const char *name = NULL;
326 enum token_type type = TOKENTYPE_NONE;
327 const struct debug_tokens *p;
328 SV* const report = newSVpvs("<== ");
330 for (p = debug_tokens; p->token; p++) {
331 if (p->token == (int)rv) {
338 Perl_sv_catpv(aTHX_ report, name);
339 else if ((char)rv > ' ' && (char)rv < '~')
340 Perl_sv_catpvf(aTHX_ report, "'%c'", (char)rv);
342 sv_catpvs(report, "EOF");
344 Perl_sv_catpvf(aTHX_ report, "?? %"IVdf, (IV)rv);
347 case TOKENTYPE_GVVAL: /* doesn't appear to be used */
350 Perl_sv_catpvf(aTHX_ report, "(ival=%"IVdf")", (IV)yylval.ival);
352 case TOKENTYPE_OPNUM:
353 Perl_sv_catpvf(aTHX_ report, "(ival=op_%s)",
354 PL_op_name[yylval.ival]);
357 Perl_sv_catpvf(aTHX_ report, "(pval=\"%s\")", yylval.pval);
359 case TOKENTYPE_OPVAL:
361 Perl_sv_catpvf(aTHX_ report, "(opval=op_%s)",
362 PL_op_name[yylval.opval->op_type]);
363 if (yylval.opval->op_type == OP_CONST) {
364 Perl_sv_catpvf(aTHX_ report, " %s",
365 SvPEEK(cSVOPx_sv(yylval.opval)));
370 sv_catpvs(report, "(opval=null)");
373 PerlIO_printf(Perl_debug_log, "### %s\n\n", SvPV_nolen_const(report));
379 /* print the buffer with suitable escapes */
382 S_printbuf(pTHX_ const char* fmt, const char* s)
384 SV* const tmp = newSVpvs("");
385 PerlIO_printf(Perl_debug_log, fmt, pv_display(tmp, s, strlen(s), 0, 60));
394 * This subroutine detects &&=, ||=, and //= and turns an ANDAND, OROR or DORDOR
395 * into an OP_ANDASSIGN, OP_ORASSIGN, or OP_DORASSIGN
399 S_ao(pTHX_ int toketype)
402 if (*PL_bufptr == '=') {
404 if (toketype == ANDAND)
405 yylval.ival = OP_ANDASSIGN;
406 else if (toketype == OROR)
407 yylval.ival = OP_ORASSIGN;
408 else if (toketype == DORDOR)
409 yylval.ival = OP_DORASSIGN;
417 * When Perl expects an operator and finds something else, no_op
418 * prints the warning. It always prints "<something> found where
419 * operator expected. It prints "Missing semicolon on previous line?"
420 * if the surprise occurs at the start of the line. "do you need to
421 * predeclare ..." is printed out for code like "sub bar; foo bar $x"
422 * where the compiler doesn't know if foo is a method call or a function.
423 * It prints "Missing operator before end of line" if there's nothing
424 * after the missing operator, or "... before <...>" if there is something
425 * after the missing operator.
429 S_no_op(pTHX_ const char *what, char *s)
432 char * const oldbp = PL_bufptr;
433 const bool is_first = (PL_oldbufptr == PL_linestart);
439 yywarn(Perl_form(aTHX_ "%s found where operator expected", what));
440 if (ckWARN_d(WARN_SYNTAX)) {
442 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
443 "\t(Missing semicolon on previous line?)\n");
444 else if (PL_oldoldbufptr && isIDFIRST_lazy_if(PL_oldoldbufptr,UTF)) {
446 for (t = PL_oldoldbufptr; *t && (isALNUM_lazy_if(t,UTF) || *t == ':'); t++) ;
447 if (t < PL_bufptr && isSPACE(*t))
448 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
449 "\t(Do you need to predeclare %.*s?)\n",
450 (int)(t - PL_oldoldbufptr), PL_oldoldbufptr);
454 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
455 "\t(Missing operator before %.*s?)\n", (int)(s - oldbp), oldbp);
463 * Complain about missing quote/regexp/heredoc terminator.
464 * If it's called with (char *)NULL then it cauterizes the line buffer.
465 * If we're in a delimited string and the delimiter is a control
466 * character, it's reformatted into a two-char sequence like ^C.
471 S_missingterm(pTHX_ char *s)
477 char * const nl = strrchr(s,'\n');
483 iscntrl(PL_multi_close)
485 PL_multi_close < 32 || PL_multi_close == 127
489 tmpbuf[1] = (char)toCTRL(PL_multi_close);
494 *tmpbuf = (char)PL_multi_close;
498 q = strchr(s,'"') ? '\'' : '"';
499 Perl_croak(aTHX_ "Can't find string terminator %c%s%c anywhere before EOF",q,s,q);
502 #define FEATURE_IS_ENABLED(name) \
503 ((0 != (PL_hints & HINT_LOCALIZE_HH)) \
504 && S_feature_is_enabled(aTHX_ STR_WITH_LEN(name)))
506 * S_feature_is_enabled
507 * Check whether the named feature is enabled.
510 S_feature_is_enabled(pTHX_ char *name, STRLEN namelen)
513 HV * const hinthv = GvHV(PL_hintgv);
514 char he_name[32] = "feature_";
515 (void) strncpy(&he_name[8], name, 24);
517 return (hinthv && hv_exists(hinthv, he_name, 8 + namelen));
525 Perl_deprecate(pTHX_ const char *s)
527 if (ckWARN(WARN_DEPRECATED))
528 Perl_warner(aTHX_ packWARN(WARN_DEPRECATED), "Use of %s is deprecated", s);
532 Perl_deprecate_old(pTHX_ const char *s)
534 /* This function should NOT be called for any new deprecated warnings */
535 /* Use Perl_deprecate instead */
537 /* It is here to maintain backward compatibility with the pre-5.8 */
538 /* warnings category hierarchy. The "deprecated" category used to */
539 /* live under the "syntax" category. It is now a top-level category */
540 /* in its own right. */
542 if (ckWARN2(WARN_DEPRECATED, WARN_SYNTAX))
543 Perl_warner(aTHX_ packWARN2(WARN_DEPRECATED, WARN_SYNTAX),
544 "Use of %s is deprecated", s);
548 * experimental text filters for win32 carriage-returns, utf16-to-utf8 and
549 * utf16-to-utf8-reversed.
552 #ifdef PERL_CR_FILTER
556 register const char *s = SvPVX_const(sv);
557 register const char * const e = s + SvCUR(sv);
558 /* outer loop optimized to do nothing if there are no CR-LFs */
560 if (*s++ == '\r' && *s == '\n') {
561 /* hit a CR-LF, need to copy the rest */
562 register char *d = s - 1;
565 if (*s == '\r' && s[1] == '\n')
576 S_cr_textfilter(pTHX_ int idx, SV *sv, int maxlen)
578 const I32 count = FILTER_READ(idx+1, sv, maxlen);
579 if (count > 0 && !maxlen)
587 * Initialize variables. Uses the Perl save_stack to save its state (for
588 * recursive calls to the parser).
592 Perl_lex_start(pTHX_ SV *line)
598 SAVEI32(PL_lex_dojoin);
599 SAVEI32(PL_lex_brackets);
600 SAVEI32(PL_lex_casemods);
601 SAVEI32(PL_lex_starts);
602 SAVEI32(PL_lex_state);
603 SAVEVPTR(PL_lex_inpat);
604 SAVEI32(PL_lex_inwhat);
606 if (PL_lex_state == LEX_KNOWNEXT) {
607 I32 toke = PL_lasttoke;
608 while (--toke >= 0) {
609 SAVEI32(PL_nexttoke[toke].next_type);
610 SAVEVPTR(PL_nexttoke[toke].next_val);
612 SAVEVPTR(PL_nexttoke[toke].next_mad);
614 SAVEI32(PL_lasttoke);
624 SAVEI32(realtokenstart);
629 if (PL_lex_state == LEX_KNOWNEXT) {
630 I32 toke = PL_nexttoke;
631 while (--toke >= 0) {
632 SAVEI32(PL_nexttype[toke]);
633 SAVEVPTR(PL_nextval[toke]);
635 SAVEI32(PL_nexttoke);
638 SAVECOPLINE(PL_curcop);
641 SAVEPPTR(PL_oldbufptr);
642 SAVEPPTR(PL_oldoldbufptr);
643 SAVEPPTR(PL_last_lop);
644 SAVEPPTR(PL_last_uni);
645 SAVEPPTR(PL_linestart);
646 SAVESPTR(PL_linestr);
647 SAVEGENERICPV(PL_lex_brackstack);
648 SAVEGENERICPV(PL_lex_casestack);
649 SAVEDESTRUCTOR_X(restore_rsfp, PL_rsfp);
650 SAVESPTR(PL_lex_stuff);
651 SAVEI32(PL_lex_defer);
652 SAVEI32(PL_sublex_info.sub_inwhat);
653 SAVESPTR(PL_lex_repl);
655 SAVEINT(PL_lex_expect);
657 PL_lex_state = LEX_NORMAL;
661 Newx(PL_lex_brackstack, 120, char);
662 Newx(PL_lex_casestack, 12, char);
664 *PL_lex_casestack = '\0';
676 PL_sublex_info.sub_inwhat = 0;
678 if (SvREADONLY(PL_linestr))
679 PL_linestr = sv_2mortal(newSVsv(PL_linestr));
680 s = SvPV_const(PL_linestr, len);
681 if (!len || s[len-1] != ';') {
682 if (!(SvFLAGS(PL_linestr) & SVs_TEMP))
683 PL_linestr = sv_2mortal(newSVsv(PL_linestr));
684 sv_catpvs(PL_linestr, "\n;");
686 SvTEMP_off(PL_linestr);
687 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = SvPVX(PL_linestr);
688 PL_bufend = PL_bufptr + SvCUR(PL_linestr);
689 PL_last_lop = PL_last_uni = NULL;
695 * Finalizer for lexing operations. Must be called when the parser is
696 * done with the lexer.
703 PL_doextract = FALSE;
708 * This subroutine has nothing to do with tilting, whether at windmills
709 * or pinball tables. Its name is short for "increment line". It
710 * increments the current line number in CopLINE(PL_curcop) and checks
711 * to see whether the line starts with a comment of the form
712 * # line 500 "foo.pm"
713 * If so, it sets the current line number and file to the values in the comment.
717 S_incline(pTHX_ char *s)
725 CopLINE_inc(PL_curcop);
728 while (SPACE_OR_TAB(*s)) s++;
729 if (strnEQ(s, "line", 4))
733 if (SPACE_OR_TAB(*s))
737 while (SPACE_OR_TAB(*s)) s++;
743 while (SPACE_OR_TAB(*s))
745 if (*s == '"' && (t = strchr(s+1, '"'))) {
750 for (t = s; !isSPACE(*t); t++) ;
753 while (SPACE_OR_TAB(*e) || *e == '\r' || *e == '\f')
755 if (*e != '\n' && *e != '\0')
756 return; /* false alarm */
762 const char * const cf = CopFILE(PL_curcop);
763 STRLEN tmplen = cf ? strlen(cf) : 0;
764 if (tmplen > 7 && strnEQ(cf, "(eval ", 6)) {
765 /* must copy *{"::_<(eval N)[oldfilename:L]"}
766 * to *{"::_<newfilename"} */
767 char smallbuf[256], smallbuf2[256];
768 char *tmpbuf, *tmpbuf2;
770 STRLEN tmplen2 = strlen(s);
771 if (tmplen + 3 < sizeof smallbuf)
774 Newx(tmpbuf, tmplen + 3, char);
775 if (tmplen2 + 3 < sizeof smallbuf2)
778 Newx(tmpbuf2, tmplen2 + 3, char);
779 tmpbuf[0] = tmpbuf2[0] = '_';
780 tmpbuf[1] = tmpbuf2[1] = '<';
781 memcpy(tmpbuf + 2, cf, ++tmplen);
782 memcpy(tmpbuf2 + 2, s, ++tmplen2);
784 gvp = (GV**)hv_fetch(PL_defstash, tmpbuf, tmplen, FALSE);
786 gv2 = *(GV**)hv_fetch(PL_defstash, tmpbuf2, tmplen2, TRUE);
788 gv_init(gv2, PL_defstash, tmpbuf2, tmplen2, FALSE);
789 /* adjust ${"::_<newfilename"} to store the new file name */
790 GvSV(gv2) = newSVpvn(tmpbuf2 + 2, tmplen2 - 2);
791 GvHV(gv2) = (HV*)SvREFCNT_inc(GvHV(*gvp));
792 GvAV(gv2) = (AV*)SvREFCNT_inc(GvAV(*gvp));
794 if (tmpbuf != smallbuf) Safefree(tmpbuf);
795 if (tmpbuf2 != smallbuf2) Safefree(tmpbuf2);
798 CopFILE_free(PL_curcop);
799 CopFILE_set(PL_curcop, s);
802 CopLINE_set(PL_curcop, atoi(n)-1);
806 /* skip space before thistoken */
809 S_skipspace0(pTHX_ register char *s)
816 thiswhite = newSVpvn("",0);
817 sv_catsv(thiswhite, skipwhite);
821 realtokenstart = s - SvPVX(PL_linestr);
825 /* skip space after thistoken */
828 S_skipspace1(pTHX_ register char *s)
831 I32 startoff = start - SvPVX(PL_linestr);
836 start = SvPVX(PL_linestr) + startoff;
837 if (!thistoken && realtokenstart >= 0) {
838 char *tstart = SvPVX(PL_linestr) + realtokenstart;
839 thistoken = newSVpvn(tstart, start - tstart);
844 nextwhite = newSVpvn("",0);
845 sv_catsv(nextwhite, skipwhite);
853 S_skipspace2(pTHX_ register char *s, SV **svp)
856 I32 bufptroff = PL_bufptr - SvPVX(PL_linestr);
857 I32 startoff = start - SvPVX(PL_linestr);
859 PL_bufptr = SvPVX(PL_linestr) + bufptroff;
860 if (!PL_madskills || !svp)
862 start = SvPVX(PL_linestr) + startoff;
863 if (!thistoken && realtokenstart >= 0) {
864 char *tstart = SvPVX(PL_linestr) + realtokenstart;
865 thistoken = newSVpvn(tstart, start - tstart);
870 *svp = newSVpvn("",0);
871 sv_setsv(*svp, skipwhite);
882 * Called to gobble the appropriate amount and type of whitespace.
883 * Skips comments as well.
887 S_skipspace(pTHX_ register char *s)
892 int startoff = s - SvPVX(PL_linestr);
900 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
901 while (s < PL_bufend && SPACE_OR_TAB(*s))
911 SSize_t oldprevlen, oldoldprevlen;
912 SSize_t oldloplen = 0, oldunilen = 0;
913 while (s < PL_bufend && isSPACE(*s)) {
914 if (*s++ == '\n' && PL_in_eval && !PL_rsfp)
919 if (s < PL_bufend && *s == '#') {
920 while (s < PL_bufend && *s != '\n')
924 if (PL_in_eval && !PL_rsfp) {
931 /* only continue to recharge the buffer if we're at the end
932 * of the buffer, we're not reading from a source filter, and
933 * we're in normal lexing mode
935 if (s < PL_bufend || !PL_rsfp || PL_sublex_info.sub_inwhat ||
936 PL_lex_state == LEX_FORMLINE)
943 /* try to recharge the buffer */
945 curoff = s - SvPVX(PL_linestr);
948 if ((s = filter_gets(PL_linestr, PL_rsfp,
949 (prevlen = SvCUR(PL_linestr)))) == NULL)
952 if (PL_madskills && curoff != startoff) {
954 skipwhite = newSVpvn("",0);
955 sv_catpvn(skipwhite, SvPVX(PL_linestr) + startoff,
959 /* mustn't throw out old stuff yet if madpropping */
960 SvCUR(PL_linestr) = curoff;
961 s = SvPVX(PL_linestr) + curoff;
963 if (curoff && s[-1] == '\n')
967 /* end of file. Add on the -p or -n magic */
968 /* XXX these shouldn't really be added here, can't set faketokens */
972 ";}continue{print or die qq(-p destination: $!\\n);}");
975 ";}continue{print or die qq(-p destination: $!\\n);}");
977 PL_minus_n = PL_minus_p = 0;
979 else if (PL_minus_n) {
981 sv_catpvn(PL_linestr, ";}", 2);
983 sv_setpvn(PL_linestr, ";}", 2);
989 sv_catpvn(PL_linestr,";", 1);
991 sv_setpvn(PL_linestr,";", 1);
994 /* reset variables for next time we lex */
995 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = s = PL_linestart
997 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
998 PL_last_lop = PL_last_uni = NULL;
1000 /* Close the filehandle. Could be from -P preprocessor,
1001 * STDIN, or a regular file. If we were reading code from
1002 * STDIN (because the commandline held no -e or filename)
1003 * then we don't close it, we reset it so the code can
1004 * read from STDIN too.
1007 if (PL_preprocess && !PL_in_eval)
1008 (void)PerlProc_pclose(PL_rsfp);
1009 else if ((PerlIO*)PL_rsfp == PerlIO_stdin())
1010 PerlIO_clearerr(PL_rsfp);
1012 (void)PerlIO_close(PL_rsfp);
1017 /* not at end of file, so we only read another line */
1018 /* make corresponding updates to old pointers, for yyerror() */
1019 oldprevlen = PL_oldbufptr - PL_bufend;
1020 oldoldprevlen = PL_oldoldbufptr - PL_bufend;
1022 oldunilen = PL_last_uni - PL_bufend;
1024 oldloplen = PL_last_lop - PL_bufend;
1025 PL_linestart = PL_bufptr = s + prevlen;
1026 PL_bufend = s + SvCUR(PL_linestr);
1028 PL_oldbufptr = s + oldprevlen;
1029 PL_oldoldbufptr = s + oldoldprevlen;
1031 PL_last_uni = s + oldunilen;
1033 PL_last_lop = s + oldloplen;
1036 /* debugger active and we're not compiling the debugger code,
1037 * so store the line into the debugger's array of lines
1039 if (PERLDB_LINE && PL_curstash != PL_debstash) {
1040 SV * const sv = newSV(0);
1042 sv_upgrade(sv, SVt_PVMG);
1043 sv_setpvn(sv,PL_bufptr,PL_bufend-PL_bufptr);
1046 av_store(CopFILEAVx(PL_curcop),(I32)CopLINE(PL_curcop),sv);
1054 skipwhite = newSVpvn("",0);
1055 curoff = s - SvPVX(PL_linestr);
1056 if (curoff - startoff)
1057 sv_catpvn(skipwhite, SvPVX(PL_linestr) + startoff,
1066 * Check the unary operators to ensure there's no ambiguity in how they're
1067 * used. An ambiguous piece of code would be:
1069 * This doesn't mean rand() + 5. Because rand() is a unary operator,
1070 * the +5 is its argument.
1080 if (PL_oldoldbufptr != PL_last_uni)
1082 while (isSPACE(*PL_last_uni))
1084 for (s = PL_last_uni; isALNUM_lazy_if(s,UTF) || *s == '-'; s++) ;
1085 if ((t = strchr(s, '(')) && t < PL_bufptr)
1088 /* XXX Things like this are just so nasty. We shouldn't be modifying
1089 source code, even if we realquick set it back. */
1090 if (ckWARN_d(WARN_AMBIGUOUS)){
1093 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
1094 "Warning: Use of \"%s\" without parentheses is ambiguous",
1101 * LOP : macro to build a list operator. Its behaviour has been replaced
1102 * with a subroutine, S_lop() for which LOP is just another name.
1105 #define LOP(f,x) return lop(f,x,s)
1109 * Build a list operator (or something that might be one). The rules:
1110 * - if we have a next token, then it's a list operator [why?]
1111 * - if the next thing is an opening paren, then it's a function
1112 * - else it's a list operator
1116 S_lop(pTHX_ I32 f, int x, char *s)
1123 PL_last_lop = PL_oldbufptr;
1124 PL_last_lop_op = (OPCODE)f;
1127 return REPORT(LSTOP);
1130 return REPORT(LSTOP);
1133 return REPORT(FUNC);
1136 return REPORT(FUNC);
1138 return REPORT(LSTOP);
1144 * Sets up for an eventual force_next(). start_force(0) basically does
1145 * an unshift, while start_force(-1) does a push. yylex removes items
1150 S_start_force(pTHX_ int where)
1154 if (where < 0) /* so people can duplicate start_force(curforce) */
1155 where = PL_lasttoke;
1156 assert(curforce < 0 || curforce == where);
1157 if (curforce != where) {
1158 for (i = PL_lasttoke; i > where; --i) {
1159 PL_nexttoke[i] = PL_nexttoke[i-1];
1163 if (curforce < 0) /* in case of duplicate start_force() */
1164 Zero(&PL_nexttoke[where], 1, NEXTTOKE);
1168 curmad('^', newSVpvn("",0));
1169 CURMAD('_', nextwhite);
1174 S_curmad(pTHX_ char slot, SV *sv)
1183 where = &PL_nexttoke[curforce].next_mad;
1186 sv_setpvn(sv, "", 0);
1189 if (UTF && is_utf8_string((U8*)SvPVX(sv), SvCUR(sv)))
1191 else if (PL_encoding) {
1192 sv_recode_to_utf8(sv, PL_encoding);
1197 /* keep a slot open for the head of the list? */
1198 if (slot != '_' && *where && (*where)->mad_key == '^') {
1199 (*where)->mad_key = slot;
1200 sv_free((*where)->mad_val);
1201 (*where)->mad_val = (void*)sv;
1204 addmad(newMADsv(slot, sv), where, 0);
1207 # define start_force(where)
1208 # define curmad(slot, sv)
1213 * When the lexer realizes it knows the next token (for instance,
1214 * it is reordering tokens for the parser) then it can call S_force_next
1215 * to know what token to return the next time the lexer is called. Caller
1216 * will need to set PL_nextval[] (or PL_nexttoke[].next_val with PERL_MAD),
1217 * and possibly PL_expect to ensure the lexer handles the token correctly.
1221 S_force_next(pTHX_ I32 type)
1226 start_force(PL_lasttoke);
1227 PL_nexttoke[curforce].next_type = type;
1228 if (PL_lex_state != LEX_KNOWNEXT)
1229 PL_lex_defer = PL_lex_state;
1230 PL_lex_state = LEX_KNOWNEXT;
1231 PL_lex_expect = PL_expect;
1234 PL_nexttype[PL_nexttoke] = type;
1236 if (PL_lex_state != LEX_KNOWNEXT) {
1237 PL_lex_defer = PL_lex_state;
1238 PL_lex_expect = PL_expect;
1239 PL_lex_state = LEX_KNOWNEXT;
1245 S_newSV_maybe_utf8(pTHX_ const char *start, STRLEN len)
1248 SV * const sv = newSVpvn(start,len);
1249 if (UTF && !IN_BYTES && is_utf8_string((const U8*)start, len))
1256 * When the lexer knows the next thing is a word (for instance, it has
1257 * just seen -> and it knows that the next char is a word char, then
1258 * it calls S_force_word to stick the next word into the PL_next lookahead.
1261 * char *start : buffer position (must be within PL_linestr)
1262 * int token : PL_next will be this type of bare word (e.g., METHOD,WORD)
1263 * int check_keyword : if true, Perl checks to make sure the word isn't
1264 * a keyword (do this if the word is a label, e.g. goto FOO)
1265 * int allow_pack : if true, : characters will also be allowed (require,
1266 * use, etc. do this)
1267 * int allow_initial_tick : used by the "sub" lexer only.
1271 S_force_word(pTHX_ register char *start, int token, int check_keyword, int allow_pack, int allow_initial_tick)
1277 start = SKIPSPACE1(start);
1279 if (isIDFIRST_lazy_if(s,UTF) ||
1280 (allow_pack && *s == ':') ||
1281 (allow_initial_tick && *s == '\'') )
1283 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, allow_pack, &len);
1284 if (check_keyword && keyword(PL_tokenbuf, len))
1286 start_force(curforce);
1288 curmad('X', newSVpvn(start,s-start));
1289 if (token == METHOD) {
1294 PL_expect = XOPERATOR;
1297 NEXTVAL_NEXTTOKE.opval
1298 = (OP*)newSVOP(OP_CONST,0,
1299 S_newSV_maybe_utf8(aTHX_ PL_tokenbuf, len));
1300 NEXTVAL_NEXTTOKE.opval->op_private |= OPpCONST_BARE;
1308 * Called when the lexer wants $foo *foo &foo etc, but the program
1309 * text only contains the "foo" portion. The first argument is a pointer
1310 * to the "foo", and the second argument is the type symbol to prefix.
1311 * Forces the next token to be a "WORD".
1312 * Creates the symbol if it didn't already exist (via gv_fetchpv()).
1316 S_force_ident(pTHX_ register const char *s, int kind)
1320 const STRLEN len = strlen(s);
1321 OP* const o = (OP*)newSVOP(OP_CONST, 0, newSVpvn(s, len));
1322 start_force(curforce);
1323 NEXTVAL_NEXTTOKE.opval = o;
1326 o->op_private = OPpCONST_ENTERED;
1327 /* XXX see note in pp_entereval() for why we forgo typo
1328 warnings if the symbol must be introduced in an eval.
1330 gv_fetchpvn_flags(s, len,
1331 PL_in_eval ? (GV_ADDMULTI | GV_ADDINEVAL)
1333 kind == '$' ? SVt_PV :
1334 kind == '@' ? SVt_PVAV :
1335 kind == '%' ? SVt_PVHV :
1343 Perl_str_to_version(pTHX_ SV *sv)
1348 const char *start = SvPV_const(sv,len);
1349 const char * const end = start + len;
1350 const bool utf = SvUTF8(sv) ? TRUE : FALSE;
1351 while (start < end) {
1355 n = utf8n_to_uvchr((U8*)start, len, &skip, 0);
1360 retval += ((NV)n)/nshift;
1369 * Forces the next token to be a version number.
1370 * If the next token appears to be an invalid version number, (e.g. "v2b"),
1371 * and if "guessing" is TRUE, then no new token is created (and the caller
1372 * must use an alternative parsing method).
1376 S_force_version(pTHX_ char *s, int guessing)
1382 I32 startoff = s - SvPVX(PL_linestr);
1391 while (isDIGIT(*d) || *d == '_' || *d == '.')
1395 start_force(curforce);
1396 curmad('X', newSVpvn(s,d-s));
1399 if (*d == ';' || isSPACE(*d) || *d == '}' || !*d) {
1401 s = scan_num(s, &yylval);
1402 version = yylval.opval;
1403 ver = cSVOPx(version)->op_sv;
1404 if (SvPOK(ver) && !SvNIOK(ver)) {
1405 SvUPGRADE(ver, SVt_PVNV);
1406 SvNV_set(ver, str_to_version(ver));
1407 SvNOK_on(ver); /* hint that it is a version */
1410 else if (guessing) {
1413 sv_free(nextwhite); /* let next token collect whitespace */
1415 s = SvPVX(PL_linestr) + startoff;
1423 if (PL_madskills && !version) {
1424 sv_free(nextwhite); /* let next token collect whitespace */
1426 s = SvPVX(PL_linestr) + startoff;
1429 /* NOTE: The parser sees the package name and the VERSION swapped */
1430 start_force(curforce);
1431 NEXTVAL_NEXTTOKE.opval = version;
1439 * Tokenize a quoted string passed in as an SV. It finds the next
1440 * chunk, up to end of string or a backslash. It may make a new
1441 * SV containing that chunk (if HINT_NEW_STRING is on). It also
1446 S_tokeq(pTHX_ SV *sv)
1450 register char *send;
1458 s = SvPV_force(sv, len);
1459 if (SvTYPE(sv) >= SVt_PVIV && SvIVX(sv) == -1)
1462 while (s < send && *s != '\\')
1467 if ( PL_hints & HINT_NEW_STRING ) {
1468 pv = sv_2mortal(newSVpvn(SvPVX_const(pv), len));
1474 if (s + 1 < send && (s[1] == '\\'))
1475 s++; /* all that, just for this */
1480 SvCUR_set(sv, d - SvPVX_const(sv));
1482 if ( PL_hints & HINT_NEW_STRING )
1483 return new_constant(NULL, 0, "q", sv, pv, "q");
1488 * Now come three functions related to double-quote context,
1489 * S_sublex_start, S_sublex_push, and S_sublex_done. They're used when
1490 * converting things like "\u\Lgnat" into ucfirst(lc("gnat")). They
1491 * interact with PL_lex_state, and create fake ( ... ) argument lists
1492 * to handle functions and concatenation.
1493 * They assume that whoever calls them will be setting up a fake
1494 * join call, because each subthing puts a ',' after it. This lets
1497 * join($, , 'lower ', lcfirst( 'uPpEr', ) ,)
1499 * (I'm not sure whether the spurious commas at the end of lcfirst's
1500 * arguments and join's arguments are created or not).
1505 * Assumes that yylval.ival is the op we're creating (e.g. OP_LCFIRST).
1507 * Pattern matching will set PL_lex_op to the pattern-matching op to
1508 * make (we return THING if yylval.ival is OP_NULL, PMFUNC otherwise).
1510 * OP_CONST and OP_READLINE are easy--just make the new op and return.
1512 * Everything else becomes a FUNC.
1514 * Sets PL_lex_state to LEX_INTERPPUSH unless (ival was OP_NULL or we
1515 * had an OP_CONST or OP_READLINE). This just sets us up for a
1516 * call to S_sublex_push().
1520 S_sublex_start(pTHX)
1523 register const I32 op_type = yylval.ival;
1525 if (op_type == OP_NULL) {
1526 yylval.opval = PL_lex_op;
1530 if (op_type == OP_CONST || op_type == OP_READLINE) {
1531 SV *sv = tokeq(PL_lex_stuff);
1533 if (SvTYPE(sv) == SVt_PVIV) {
1534 /* Overloaded constants, nothing fancy: Convert to SVt_PV: */
1536 const char * const p = SvPV_const(sv, len);
1537 SV * const nsv = newSVpvn(p, len);
1543 yylval.opval = (OP*)newSVOP(op_type, 0, sv);
1544 PL_lex_stuff = NULL;
1545 /* Allow <FH> // "foo" */
1546 if (op_type == OP_READLINE)
1547 PL_expect = XTERMORDORDOR;
1551 PL_sublex_info.super_state = PL_lex_state;
1552 PL_sublex_info.sub_inwhat = op_type;
1553 PL_sublex_info.sub_op = PL_lex_op;
1554 PL_lex_state = LEX_INTERPPUSH;
1558 yylval.opval = PL_lex_op;
1568 * Create a new scope to save the lexing state. The scope will be
1569 * ended in S_sublex_done. Returns a '(', starting the function arguments
1570 * to the uc, lc, etc. found before.
1571 * Sets PL_lex_state to LEX_INTERPCONCAT.
1580 PL_lex_state = PL_sublex_info.super_state;
1581 SAVEI32(PL_lex_dojoin);
1582 SAVEI32(PL_lex_brackets);
1583 SAVEI32(PL_lex_casemods);
1584 SAVEI32(PL_lex_starts);
1585 SAVEI32(PL_lex_state);
1586 SAVEVPTR(PL_lex_inpat);
1587 SAVEI32(PL_lex_inwhat);
1588 SAVECOPLINE(PL_curcop);
1589 SAVEPPTR(PL_bufptr);
1590 SAVEPPTR(PL_bufend);
1591 SAVEPPTR(PL_oldbufptr);
1592 SAVEPPTR(PL_oldoldbufptr);
1593 SAVEPPTR(PL_last_lop);
1594 SAVEPPTR(PL_last_uni);
1595 SAVEPPTR(PL_linestart);
1596 SAVESPTR(PL_linestr);
1597 SAVEGENERICPV(PL_lex_brackstack);
1598 SAVEGENERICPV(PL_lex_casestack);
1600 PL_linestr = PL_lex_stuff;
1601 PL_lex_stuff = NULL;
1603 PL_bufend = PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart
1604 = SvPVX(PL_linestr);
1605 PL_bufend += SvCUR(PL_linestr);
1606 PL_last_lop = PL_last_uni = NULL;
1607 SAVEFREESV(PL_linestr);
1609 PL_lex_dojoin = FALSE;
1610 PL_lex_brackets = 0;
1611 Newx(PL_lex_brackstack, 120, char);
1612 Newx(PL_lex_casestack, 12, char);
1613 PL_lex_casemods = 0;
1614 *PL_lex_casestack = '\0';
1616 PL_lex_state = LEX_INTERPCONCAT;
1617 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
1619 PL_lex_inwhat = PL_sublex_info.sub_inwhat;
1620 if (PL_lex_inwhat == OP_MATCH || PL_lex_inwhat == OP_QR || PL_lex_inwhat == OP_SUBST)
1621 PL_lex_inpat = PL_sublex_info.sub_op;
1623 PL_lex_inpat = NULL;
1630 * Restores lexer state after a S_sublex_push.
1637 if (!PL_lex_starts++) {
1638 SV * const sv = newSVpvs("");
1639 if (SvUTF8(PL_linestr))
1641 PL_expect = XOPERATOR;
1642 yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
1646 if (PL_lex_casemods) { /* oops, we've got some unbalanced parens */
1647 PL_lex_state = LEX_INTERPCASEMOD;
1651 /* Is there a right-hand side to take care of? (s//RHS/ or tr//RHS/) */
1652 if (PL_lex_repl && (PL_lex_inwhat == OP_SUBST || PL_lex_inwhat == OP_TRANS)) {
1653 PL_linestr = PL_lex_repl;
1655 PL_bufend = PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart = SvPVX(PL_linestr);
1656 PL_bufend += SvCUR(PL_linestr);
1657 PL_last_lop = PL_last_uni = NULL;
1658 SAVEFREESV(PL_linestr);
1659 PL_lex_dojoin = FALSE;
1660 PL_lex_brackets = 0;
1661 PL_lex_casemods = 0;
1662 *PL_lex_casestack = '\0';
1664 if (SvEVALED(PL_lex_repl)) {
1665 PL_lex_state = LEX_INTERPNORMAL;
1667 /* we don't clear PL_lex_repl here, so that we can check later
1668 whether this is an evalled subst; that means we rely on the
1669 logic to ensure sublex_done() is called again only via the
1670 branch (in yylex()) that clears PL_lex_repl, else we'll loop */
1673 PL_lex_state = LEX_INTERPCONCAT;
1683 endwhite = newSVpvn("",0);
1684 sv_catsv(endwhite, thiswhite);
1688 sv_setpvn(thistoken,"",0);
1690 realtokenstart = -1;
1694 PL_bufend = SvPVX(PL_linestr);
1695 PL_bufend += SvCUR(PL_linestr);
1696 PL_expect = XOPERATOR;
1697 PL_sublex_info.sub_inwhat = 0;
1705 Extracts a pattern, double-quoted string, or transliteration. This
1708 It looks at lex_inwhat and PL_lex_inpat to find out whether it's
1709 processing a pattern (PL_lex_inpat is true), a transliteration
1710 (lex_inwhat & OP_TRANS is true), or a double-quoted string.
1712 Returns a pointer to the character scanned up to. Iff this is
1713 advanced from the start pointer supplied (ie if anything was
1714 successfully parsed), will leave an OP for the substring scanned
1715 in yylval. Caller must intuit reason for not parsing further
1716 by looking at the next characters herself.
1720 double-quoted style: \r and \n
1721 regexp special ones: \D \s
1723 backrefs: \1 (deprecated in substitution replacements)
1724 case and quoting: \U \Q \E
1725 stops on @ and $, but not for $ as tail anchor
1727 In transliterations:
1728 characters are VERY literal, except for - not at the start or end
1729 of the string, which indicates a range. scan_const expands the
1730 range to the full set of intermediate characters.
1732 In double-quoted strings:
1734 double-quoted style: \r and \n
1736 backrefs: \1 (deprecated)
1737 case and quoting: \U \Q \E
1740 scan_const does *not* construct ops to handle interpolated strings.
1741 It stops processing as soon as it finds an embedded $ or @ variable
1742 and leaves it to the caller to work out what's going on.
1744 @ in pattern could be: @foo, @{foo}, @$foo, @'foo, @::foo.
1746 $ in pattern could be $foo or could be tail anchor. Assumption:
1747 it's a tail anchor if $ is the last thing in the string, or if it's
1748 followed by one of ")| \n\t"
1750 \1 (backreferences) are turned into $1
1752 The structure of the code is
1753 while (there's a character to process) {
1754 handle transliteration ranges
1755 skip regexp comments
1756 skip # initiated comments in //x patterns
1757 check for embedded @foo
1758 check for embedded scalars
1760 leave intact backslashes from leave (below)
1761 deprecate \1 in strings and sub replacements
1762 handle string-changing backslashes \l \U \Q \E, etc.
1763 switch (what was escaped) {
1764 handle - in a transliteration (becomes a literal -)
1765 handle \132 octal characters
1766 handle 0x15 hex characters
1767 handle \cV (control V)
1768 handle printf backslashes (\f, \r, \n, etc)
1770 } (end if backslash)
1771 } (end while character to read)
1776 S_scan_const(pTHX_ char *start)
1779 register char *send = PL_bufend; /* end of the constant */
1780 SV *sv = newSV(send - start); /* sv for the constant */
1781 register char *s = start; /* start of the constant */
1782 register char *d = SvPVX(sv); /* destination for copies */
1783 bool dorange = FALSE; /* are we in a translit range? */
1784 bool didrange = FALSE; /* did we just finish a range? */
1785 I32 has_utf8 = FALSE; /* Output constant is UTF8 */
1786 I32 this_utf8 = UTF; /* The source string is assumed to be UTF8 */
1789 UV literal_endpoint = 0;
1792 const char *leaveit = /* set of acceptably-backslashed characters */
1794 ? "\\.^$@AGZdDwWsSbBpPXC+*?|()-nrtfeaxz0123456789[{]} \t\n\r\f\v#"
1797 if (PL_lex_inwhat == OP_TRANS && PL_sublex_info.sub_op) {
1798 /* If we are doing a trans and we know we want UTF8 set expectation */
1799 has_utf8 = PL_sublex_info.sub_op->op_private & (OPpTRANS_FROM_UTF|OPpTRANS_TO_UTF);
1800 this_utf8 = PL_sublex_info.sub_op->op_private & (PL_lex_repl ? OPpTRANS_FROM_UTF : OPpTRANS_TO_UTF);
1804 while (s < send || dorange) {
1805 /* get transliterations out of the way (they're most literal) */
1806 if (PL_lex_inwhat == OP_TRANS) {
1807 /* expand a range A-Z to the full set of characters. AIE! */
1809 I32 i; /* current expanded character */
1810 I32 min; /* first character in range */
1811 I32 max; /* last character in range */
1814 char * const c = (char*)utf8_hop((U8*)d, -1);
1818 *c = (char)UTF_TO_NATIVE(0xff);
1819 /* mark the range as done, and continue */
1825 i = d - SvPVX_const(sv); /* remember current offset */
1826 SvGROW(sv, SvLEN(sv) + 256); /* never more than 256 chars in a range */
1827 d = SvPVX(sv) + i; /* refresh d after realloc */
1828 d -= 2; /* eat the first char and the - */
1830 min = (U8)*d; /* first char in range */
1831 max = (U8)d[1]; /* last char in range */
1835 "Invalid range \"%c-%c\" in transliteration operator",
1836 (char)min, (char)max);
1840 if (literal_endpoint == 2 &&
1841 ((isLOWER(min) && isLOWER(max)) ||
1842 (isUPPER(min) && isUPPER(max)))) {
1844 for (i = min; i <= max; i++)
1846 *d++ = NATIVE_TO_NEED(has_utf8,i);
1848 for (i = min; i <= max; i++)
1850 *d++ = NATIVE_TO_NEED(has_utf8,i);
1855 for (i = min; i <= max; i++)
1858 /* mark the range as done, and continue */
1862 literal_endpoint = 0;
1867 /* range begins (ignore - as first or last char) */
1868 else if (*s == '-' && s+1 < send && s != start) {
1870 Perl_croak(aTHX_ "Ambiguous range in transliteration operator");
1873 *d++ = (char)UTF_TO_NATIVE(0xff); /* use illegal utf8 byte--see pmtrans */
1883 literal_endpoint = 0;
1888 /* if we get here, we're not doing a transliteration */
1890 /* skip for regexp comments /(?#comment)/ and code /(?{code})/,
1891 except for the last char, which will be done separately. */
1892 else if (*s == '(' && PL_lex_inpat && s[1] == '?') {
1894 while (s+1 < send && *s != ')')
1895 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
1897 else if (s[2] == '{' /* This should match regcomp.c */
1898 || ((s[2] == 'p' || s[2] == '?') && s[3] == '{'))
1901 char *regparse = s + (s[2] == '{' ? 3 : 4);
1904 while (count && (c = *regparse)) {
1905 if (c == '\\' && regparse[1])
1913 if (*regparse != ')')
1914 regparse--; /* Leave one char for continuation. */
1915 while (s < regparse)
1916 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
1920 /* likewise skip #-initiated comments in //x patterns */
1921 else if (*s == '#' && PL_lex_inpat &&
1922 ((PMOP*)PL_lex_inpat)->op_pmflags & PMf_EXTENDED) {
1923 while (s+1 < send && *s != '\n')
1924 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
1927 /* check for embedded arrays
1928 (@foo, @::foo, @'foo, @{foo}, @$foo, @+, @-)
1930 else if (*s == '@' && s[1]
1931 && (isALNUM_lazy_if(s+1,UTF) || strchr(":'{$+-", s[1])))
1934 /* check for embedded scalars. only stop if we're sure it's a
1937 else if (*s == '$') {
1938 if (!PL_lex_inpat) /* not a regexp, so $ must be var */
1940 if (s + 1 < send && !strchr("()| \r\n\t", s[1]))
1941 break; /* in regexp, $ might be tail anchor */
1944 /* End of else if chain - OP_TRANS rejoin rest */
1947 if (*s == '\\' && s+1 < send) {
1950 /* some backslashes we leave behind */
1951 if (*leaveit && *s && strchr(leaveit, *s)) {
1952 *d++ = NATIVE_TO_NEED(has_utf8,'\\');
1953 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
1957 /* deprecate \1 in strings and substitution replacements */
1958 if (PL_lex_inwhat == OP_SUBST && !PL_lex_inpat &&
1959 isDIGIT(*s) && *s != '0' && !isDIGIT(s[1]))
1961 if (ckWARN(WARN_SYNTAX))
1962 Perl_warner(aTHX_ packWARN(WARN_SYNTAX), "\\%c better written as $%c", *s, *s);
1967 /* string-change backslash escapes */
1968 if (PL_lex_inwhat != OP_TRANS && *s && strchr("lLuUEQ", *s)) {
1973 /* if we get here, it's either a quoted -, or a digit */
1976 /* quoted - in transliterations */
1978 if (PL_lex_inwhat == OP_TRANS) {
1988 Perl_warner(aTHX_ packWARN(WARN_MISC),
1989 "Unrecognized escape \\%c passed through",
1991 /* default action is to copy the quoted character */
1992 goto default_action;
1995 /* \132 indicates an octal constant */
1996 case '0': case '1': case '2': case '3':
1997 case '4': case '5': case '6': case '7':
2001 uv = grok_oct(s, &len, &flags, NULL);
2004 goto NUM_ESCAPE_INSERT;
2006 /* \x24 indicates a hex constant */
2010 char* const e = strchr(s, '}');
2011 I32 flags = PERL_SCAN_ALLOW_UNDERSCORES |
2012 PERL_SCAN_DISALLOW_PREFIX;
2017 yyerror("Missing right brace on \\x{}");
2021 uv = grok_hex(s, &len, &flags, NULL);
2027 I32 flags = PERL_SCAN_DISALLOW_PREFIX;
2028 uv = grok_hex(s, &len, &flags, NULL);
2034 /* Insert oct or hex escaped character.
2035 * There will always enough room in sv since such
2036 * escapes will be longer than any UTF-8 sequence
2037 * they can end up as. */
2039 /* We need to map to chars to ASCII before doing the tests
2042 if (!UNI_IS_INVARIANT(NATIVE_TO_UNI(uv))) {
2043 if (!has_utf8 && uv > 255) {
2044 /* Might need to recode whatever we have
2045 * accumulated so far if it contains any
2048 * (Can't we keep track of that and avoid
2049 * this rescan? --jhi)
2053 for (c = (U8 *) SvPVX(sv); c < (U8 *)d; c++) {
2054 if (!NATIVE_IS_INVARIANT(*c)) {
2059 const STRLEN offset = d - SvPVX_const(sv);
2061 d = SvGROW(sv, SvLEN(sv) + hicount + 1) + offset;
2065 while (src >= (const U8 *)SvPVX_const(sv)) {
2066 if (!NATIVE_IS_INVARIANT(*src)) {
2067 const U8 ch = NATIVE_TO_ASCII(*src);
2068 *dst-- = (U8)UTF8_EIGHT_BIT_LO(ch);
2069 *dst-- = (U8)UTF8_EIGHT_BIT_HI(ch);
2079 if (has_utf8 || uv > 255) {
2080 d = (char*)uvchr_to_utf8((U8*)d, uv);
2082 if (PL_lex_inwhat == OP_TRANS &&
2083 PL_sublex_info.sub_op) {
2084 PL_sublex_info.sub_op->op_private |=
2085 (PL_lex_repl ? OPpTRANS_FROM_UTF
2098 /* \N{LATIN SMALL LETTER A} is a named character */
2102 char* e = strchr(s, '}');
2108 yyerror("Missing right brace on \\N{}");
2112 if (e > s + 2 && s[1] == 'U' && s[2] == '+') {
2114 I32 flags = PERL_SCAN_ALLOW_UNDERSCORES |
2115 PERL_SCAN_DISALLOW_PREFIX;
2118 uv = grok_hex(s, &len, &flags, NULL);
2120 goto NUM_ESCAPE_INSERT;
2122 res = newSVpvn(s + 1, e - s - 1);
2123 res = new_constant( NULL, 0, "charnames",
2124 res, NULL, "\\N{...}" );
2126 sv_utf8_upgrade(res);
2127 str = SvPV_const(res,len);
2128 #ifdef EBCDIC_NEVER_MIND
2129 /* charnames uses pack U and that has been
2130 * recently changed to do the below uni->native
2131 * mapping, so this would be redundant (and wrong,
2132 * the code point would be doubly converted).
2133 * But leave this in just in case the pack U change
2134 * gets revoked, but the semantics is still
2135 * desireable for charnames. --jhi */
2137 UV uv = utf8_to_uvchr((const U8*)str, 0);
2140 U8 tmpbuf[UTF8_MAXBYTES+1], *d;
2142 d = uvchr_to_utf8(tmpbuf, UNI_TO_NATIVE(uv));
2143 sv_setpvn(res, (char *)tmpbuf, d - tmpbuf);
2144 str = SvPV_const(res, len);
2148 if (!has_utf8 && SvUTF8(res)) {
2149 const char * const ostart = SvPVX_const(sv);
2150 SvCUR_set(sv, d - ostart);
2153 sv_utf8_upgrade(sv);
2154 /* this just broke our allocation above... */
2155 SvGROW(sv, (STRLEN)(send - start));
2156 d = SvPVX(sv) + SvCUR(sv);
2159 if (len > (STRLEN)(e - s + 4)) { /* I _guess_ 4 is \N{} --jhi */
2160 const char * const odest = SvPVX_const(sv);
2162 SvGROW(sv, (SvLEN(sv) + len - (e - s + 4)));
2163 d = SvPVX(sv) + (d - odest);
2165 Copy(str, d, len, char);
2172 yyerror("Missing braces on \\N{}");
2175 /* \c is a control character */
2184 *d++ = NATIVE_TO_NEED(has_utf8,toCTRL(c));
2187 yyerror("Missing control char name in \\c");
2191 /* printf-style backslashes, formfeeds, newlines, etc */
2193 *d++ = NATIVE_TO_NEED(has_utf8,'\b');
2196 *d++ = NATIVE_TO_NEED(has_utf8,'\n');
2199 *d++ = NATIVE_TO_NEED(has_utf8,'\r');
2202 *d++ = NATIVE_TO_NEED(has_utf8,'\f');
2205 *d++ = NATIVE_TO_NEED(has_utf8,'\t');
2208 *d++ = ASCII_TO_NEED(has_utf8,'\033');
2211 *d++ = ASCII_TO_NEED(has_utf8,'\007');
2217 } /* end if (backslash) */
2224 /* If we started with encoded form, or already know we want it
2225 and then encode the next character */
2226 if ((has_utf8 || this_utf8) && !NATIVE_IS_INVARIANT((U8)(*s))) {
2228 const UV nextuv = (this_utf8) ? utf8n_to_uvchr((U8*)s, send - s, &len, 0) : (UV) ((U8) *s);
2229 const STRLEN need = UNISKIP(NATIVE_TO_UNI(nextuv));
2232 /* encoded value larger than old, need extra space (NOTE: SvCUR() not set here) */
2233 const STRLEN off = d - SvPVX_const(sv);
2234 d = SvGROW(sv, SvLEN(sv) + (need-len)) + off;
2236 d = (char*)uvchr_to_utf8((U8*)d, nextuv);
2240 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2242 } /* while loop to process each character */
2244 /* terminate the string and set up the sv */
2246 SvCUR_set(sv, d - SvPVX_const(sv));
2247 if (SvCUR(sv) >= SvLEN(sv))
2248 Perl_croak(aTHX_ "panic: constant overflowed allocated space");
2251 if (PL_encoding && !has_utf8) {
2252 sv_recode_to_utf8(sv, PL_encoding);
2258 if (PL_lex_inwhat == OP_TRANS && PL_sublex_info.sub_op) {
2259 PL_sublex_info.sub_op->op_private |=
2260 (PL_lex_repl ? OPpTRANS_FROM_UTF : OPpTRANS_TO_UTF);
2264 /* shrink the sv if we allocated more than we used */
2265 if (SvCUR(sv) + 5 < SvLEN(sv)) {
2266 SvPV_shrink_to_cur(sv);
2269 /* return the substring (via yylval) only if we parsed anything */
2270 if (s > PL_bufptr) {
2271 if ( PL_hints & ( PL_lex_inpat ? HINT_NEW_RE : HINT_NEW_STRING ) )
2272 sv = new_constant(start, s - start, (PL_lex_inpat ? "qr" : "q"),
2274 ( PL_lex_inwhat == OP_TRANS
2276 : ( (PL_lex_inwhat == OP_SUBST && !PL_lex_inpat)
2279 yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
2286 * Returns TRUE if there's more to the expression (e.g., a subscript),
2289 * It deals with "$foo[3]" and /$foo[3]/ and /$foo[0123456789$]+/
2291 * ->[ and ->{ return TRUE
2292 * { and [ outside a pattern are always subscripts, so return TRUE
2293 * if we're outside a pattern and it's not { or [, then return FALSE
2294 * if we're in a pattern and the first char is a {
2295 * {4,5} (any digits around the comma) returns FALSE
2296 * if we're in a pattern and the first char is a [
2298 * [SOMETHING] has a funky algorithm to decide whether it's a
2299 * character class or not. It has to deal with things like
2300 * /$foo[-3]/ and /$foo[$bar]/ as well as /$foo[$\d]+/
2301 * anything else returns TRUE
2304 /* This is the one truly awful dwimmer necessary to conflate C and sed. */
2307 S_intuit_more(pTHX_ register char *s)
2310 if (PL_lex_brackets)
2312 if (*s == '-' && s[1] == '>' && (s[2] == '[' || s[2] == '{'))
2314 if (*s != '{' && *s != '[')
2319 /* In a pattern, so maybe we have {n,m}. */
2336 /* On the other hand, maybe we have a character class */
2339 if (*s == ']' || *s == '^')
2342 /* this is terrifying, and it works */
2343 int weight = 2; /* let's weigh the evidence */
2345 unsigned char un_char = 255, last_un_char;
2346 const char * const send = strchr(s,']');
2347 char tmpbuf[sizeof PL_tokenbuf * 4];
2349 if (!send) /* has to be an expression */
2352 Zero(seen,256,char);
2355 else if (isDIGIT(*s)) {
2357 if (isDIGIT(s[1]) && s[2] == ']')
2363 for (; s < send; s++) {
2364 last_un_char = un_char;
2365 un_char = (unsigned char)*s;
2370 weight -= seen[un_char] * 10;
2371 if (isALNUM_lazy_if(s+1,UTF)) {
2373 scan_ident(s, send, tmpbuf, sizeof tmpbuf, FALSE);
2374 len = (int)strlen(tmpbuf);
2375 if (len > 1 && gv_fetchpvn_flags(tmpbuf, len, 0, SVt_PV))
2380 else if (*s == '$' && s[1] &&
2381 strchr("[#!%*<>()-=",s[1])) {
2382 if (/*{*/ strchr("])} =",s[2]))
2391 if (strchr("wds]",s[1]))
2393 else if (seen['\''] || seen['"'])
2395 else if (strchr("rnftbxcav",s[1]))
2397 else if (isDIGIT(s[1])) {
2399 while (s[1] && isDIGIT(s[1]))
2409 if (strchr("aA01! ",last_un_char))
2411 if (strchr("zZ79~",s[1]))
2413 if (last_un_char == 255 && (isDIGIT(s[1]) || s[1] == '$'))
2414 weight -= 5; /* cope with negative subscript */
2417 if (!isALNUM(last_un_char)
2418 && !(last_un_char == '$' || last_un_char == '@'
2419 || last_un_char == '&')
2420 && isALPHA(*s) && s[1] && isALPHA(s[1])) {
2425 if (keyword(tmpbuf, d - tmpbuf))
2428 if (un_char == last_un_char + 1)
2430 weight -= seen[un_char];
2435 if (weight >= 0) /* probably a character class */
2445 * Does all the checking to disambiguate
2447 * between foo(bar) and bar->foo. Returns 0 if not a method, otherwise
2448 * FUNCMETH (bar->foo(args)) or METHOD (bar->foo args).
2450 * First argument is the stuff after the first token, e.g. "bar".
2452 * Not a method if bar is a filehandle.
2453 * Not a method if foo is a subroutine prototyped to take a filehandle.
2454 * Not a method if it's really "Foo $bar"
2455 * Method if it's "foo $bar"
2456 * Not a method if it's really "print foo $bar"
2457 * Method if it's really "foo package::" (interpreted as package->foo)
2458 * Not a method if bar is known to be a subroutine ("sub bar; foo bar")
2459 * Not a method if bar is a filehandle or package, but is quoted with
2464 S_intuit_method(pTHX_ char *start, GV *gv, CV *cv)
2467 char *s = start + (*start == '$');
2468 char tmpbuf[sizeof PL_tokenbuf];
2476 if (SvTYPE(gv) == SVt_PVGV && GvIO(gv))
2480 const char *proto = SvPVX_const(cv);
2491 s = scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
2492 /* start is the beginning of the possible filehandle/object,
2493 * and s is the end of it
2494 * tmpbuf is a copy of it
2497 if (*start == '$') {
2498 if (gv || PL_last_lop_op == OP_PRINT || isUPPER(*PL_tokenbuf))
2501 len = start - SvPVX(PL_linestr);
2505 start = SvPVX(PL_linestr) + len;
2509 return *s == '(' ? FUNCMETH : METHOD;
2511 if (!keyword(tmpbuf, len)) {
2512 if (len > 2 && tmpbuf[len - 2] == ':' && tmpbuf[len - 1] == ':') {
2516 soff = s - SvPVX(PL_linestr);
2520 indirgv = gv_fetchpvn_flags(tmpbuf, len, 0, SVt_PVCV);
2521 if (indirgv && GvCVu(indirgv))
2523 /* filehandle or package name makes it a method */
2524 if (!gv || GvIO(indirgv) || gv_stashpvn(tmpbuf, len, FALSE)) {
2526 soff = s - SvPVX(PL_linestr);
2529 if ((PL_bufend - s) >= 2 && *s == '=' && *(s+1) == '>')
2530 return 0; /* no assumptions -- "=>" quotes bearword */
2532 start_force(curforce);
2533 NEXTVAL_NEXTTOKE.opval = (OP*)newSVOP(OP_CONST, 0,
2534 newSVpvn(tmpbuf,len));
2535 NEXTVAL_NEXTTOKE.opval->op_private = OPpCONST_BARE;
2537 curmad('X', newSVpvn(start,SvPVX(PL_linestr) + soff - start));
2542 PL_bufptr = SvPVX(PL_linestr) + soff; /* restart before space */
2544 return *s == '(' ? FUNCMETH : METHOD;
2552 * Return a string of Perl code to load the debugger. If PERL5DB
2553 * is set, it will return the contents of that, otherwise a
2554 * compile-time require of perl5db.pl.
2562 const char * const pdb = PerlEnv_getenv("PERL5DB");
2566 SETERRNO(0,SS_NORMAL);
2567 return "BEGIN { require 'perl5db.pl' }";
2573 /* Encoded script support. filter_add() effectively inserts a
2574 * 'pre-processing' function into the current source input stream.
2575 * Note that the filter function only applies to the current source file
2576 * (e.g., it will not affect files 'require'd or 'use'd by this one).
2578 * The datasv parameter (which may be NULL) can be used to pass
2579 * private data to this instance of the filter. The filter function
2580 * can recover the SV using the FILTER_DATA macro and use it to
2581 * store private buffers and state information.
2583 * The supplied datasv parameter is upgraded to a PVIO type
2584 * and the IoDIRP/IoANY field is used to store the function pointer,
2585 * and IOf_FAKE_DIRP is enabled on datasv to mark this as such.
2586 * Note that IoTOP_NAME, IoFMT_NAME, IoBOTTOM_NAME, if set for
2587 * private use must be set using malloc'd pointers.
2591 Perl_filter_add(pTHX_ filter_t funcp, SV *datasv)
2597 if (!PL_rsfp_filters)
2598 PL_rsfp_filters = newAV();
2601 SvUPGRADE(datasv, SVt_PVIO);
2602 IoANY(datasv) = FPTR2DPTR(void *, funcp); /* stash funcp into spare field */
2603 IoFLAGS(datasv) |= IOf_FAKE_DIRP;
2604 DEBUG_P(PerlIO_printf(Perl_debug_log, "filter_add func %p (%s)\n",
2605 IoANY(datasv), SvPV_nolen(datasv)));
2606 av_unshift(PL_rsfp_filters, 1);
2607 av_store(PL_rsfp_filters, 0, datasv) ;
2612 /* Delete most recently added instance of this filter function. */
2614 Perl_filter_del(pTHX_ filter_t funcp)
2620 DEBUG_P(PerlIO_printf(Perl_debug_log, "filter_del func %p", FPTR2DPTR(XPVIO *, funcp)));
2622 if (!PL_rsfp_filters || AvFILLp(PL_rsfp_filters)<0)
2624 /* if filter is on top of stack (usual case) just pop it off */
2625 datasv = FILTER_DATA(AvFILLp(PL_rsfp_filters));
2626 if (IoANY(datasv) == FPTR2DPTR(void *, funcp)) {
2627 IoFLAGS(datasv) &= ~IOf_FAKE_DIRP;
2628 IoANY(datasv) = (void *)NULL;
2629 sv_free(av_pop(PL_rsfp_filters));
2633 /* we need to search for the correct entry and clear it */
2634 Perl_die(aTHX_ "filter_del can only delete in reverse order (currently)");
2638 /* Invoke the idxth filter function for the current rsfp. */
2639 /* maxlen 0 = read one text line */
2641 Perl_filter_read(pTHX_ int idx, SV *buf_sv, int maxlen)
2647 if (!PL_rsfp_filters)
2649 if (idx > AvFILLp(PL_rsfp_filters)) { /* Any more filters? */
2650 /* Provide a default input filter to make life easy. */
2651 /* Note that we append to the line. This is handy. */
2652 DEBUG_P(PerlIO_printf(Perl_debug_log,
2653 "filter_read %d: from rsfp\n", idx));
2657 const int old_len = SvCUR(buf_sv);
2659 /* ensure buf_sv is large enough */
2660 SvGROW(buf_sv, (STRLEN)(old_len + maxlen)) ;
2661 if ((len = PerlIO_read(PL_rsfp, SvPVX(buf_sv) + old_len, maxlen)) <= 0){
2662 if (PerlIO_error(PL_rsfp))
2663 return -1; /* error */
2665 return 0 ; /* end of file */
2667 SvCUR_set(buf_sv, old_len + len) ;
2670 if (sv_gets(buf_sv, PL_rsfp, SvCUR(buf_sv)) == NULL) {
2671 if (PerlIO_error(PL_rsfp))
2672 return -1; /* error */
2674 return 0 ; /* end of file */
2677 return SvCUR(buf_sv);
2679 /* Skip this filter slot if filter has been deleted */
2680 if ( (datasv = FILTER_DATA(idx)) == &PL_sv_undef) {
2681 DEBUG_P(PerlIO_printf(Perl_debug_log,
2682 "filter_read %d: skipped (filter deleted)\n",
2684 return FILTER_READ(idx+1, buf_sv, maxlen); /* recurse */
2686 /* Get function pointer hidden within datasv */
2687 funcp = DPTR2FPTR(filter_t, IoANY(datasv));
2688 DEBUG_P(PerlIO_printf(Perl_debug_log,
2689 "filter_read %d: via function %p (%s)\n",
2690 idx, datasv, SvPV_nolen_const(datasv)));
2691 /* Call function. The function is expected to */
2692 /* call "FILTER_READ(idx+1, buf_sv)" first. */
2693 /* Return: <0:error, =0:eof, >0:not eof */
2694 return (*funcp)(aTHX_ idx, buf_sv, maxlen);
2698 S_filter_gets(pTHX_ register SV *sv, register PerlIO *fp, STRLEN append)
2701 #ifdef PERL_CR_FILTER
2702 if (!PL_rsfp_filters) {
2703 filter_add(S_cr_textfilter,NULL);
2706 if (PL_rsfp_filters) {
2708 SvCUR_set(sv, 0); /* start with empty line */
2709 if (FILTER_READ(0, sv, 0) > 0)
2710 return ( SvPVX(sv) ) ;
2715 return (sv_gets(sv, fp, append));
2719 S_find_in_my_stash(pTHX_ const char *pkgname, I32 len)
2724 if (len == 11 && *pkgname == '_' && strEQ(pkgname, "__PACKAGE__"))
2728 (pkgname[len - 2] == ':' && pkgname[len - 1] == ':') &&
2729 (gv = gv_fetchpvn_flags(pkgname, len, 0, SVt_PVHV)))
2731 return GvHV(gv); /* Foo:: */
2734 /* use constant CLASS => 'MyClass' */
2735 if ((gv = gv_fetchpvn_flags(pkgname, len, 0, SVt_PVCV))) {
2737 if (GvCV(gv) && (sv = cv_const_sv(GvCV(gv)))) {
2738 pkgname = SvPV_nolen_const(sv);
2742 return gv_stashpv(pkgname, FALSE);
2748 * The intent of this yylex wrapper is to minimize the changes to the
2749 * tokener when we aren't interested in collecting madprops. It remains
2750 * to be seen how successful this strategy will be...
2757 char *s = PL_bufptr;
2759 /* make sure thiswhite is initialized */
2763 /* just do what yylex would do on pending identifier; leave thiswhite alone */
2764 if (PL_pending_ident)
2765 return S_pending_ident(aTHX);
2767 /* previous token ate up our whitespace? */
2768 if (!PL_lasttoke && nextwhite) {
2769 thiswhite = nextwhite;
2773 /* isolate the token, and figure out where it is without whitespace */
2774 realtokenstart = -1;
2778 assert(curforce < 0);
2780 if (!thismad || thismad->mad_key == '^') { /* not forced already? */
2782 if (realtokenstart < 0 || !CopLINE(PL_curcop))
2783 thistoken = newSVpvn("",0);
2785 char *tstart = SvPVX(PL_linestr) + realtokenstart;
2786 thistoken = newSVpvn(tstart, s - tstart);
2789 if (thismad) /* install head */
2790 CURMAD('X', thistoken);
2793 /* last whitespace of a sublex? */
2794 if (optype == ')' && endwhite) {
2795 CURMAD('X', endwhite);
2800 /* if no whitespace and we're at EOF, bail. Otherwise fake EOF below. */
2801 if (!thiswhite && !endwhite && !optype) {
2807 /* put off final whitespace till peg */
2808 if (optype == ';' && !PL_rsfp) {
2809 nextwhite = thiswhite;
2812 else if (thisopen) {
2813 CURMAD('q', thisopen);
2819 /* Store actual token text as madprop X */
2820 CURMAD('X', thistoken);
2824 /* add preceding whitespace as madprop _ */
2825 CURMAD('_', thiswhite);
2829 /* add quoted material as madprop = */
2830 CURMAD('=', thisstuff);
2834 /* add terminating quote as madprop Q */
2835 CURMAD('Q', thisclose);
2839 /* special processing based on optype */
2843 /* opval doesn't need a TOKEN since it can already store mp */
2854 append_madprops(thismad, yylval.opval, 0);
2862 addmad(newMADsv('p', endwhite), &thismad, 0);
2871 /* remember any fake bracket that lexer is about to discard */
2872 if (PL_lex_brackets == 1 &&
2873 ((expectation)PL_lex_brackstack[0] & XFAKEBRACK))
2876 while (s < PL_bufend && (*s == ' ' || *s == '\t'))
2879 thiswhite = newSVpvn(PL_bufptr, ++s - PL_bufptr);
2880 addmad(newMADsv('#', thiswhite), &thismad, 0);
2883 break; /* don't bother looking for trailing comment */
2892 /* attach a trailing comment to its statement instead of next token */
2896 if (PL_bufptr > PL_oldbufptr && PL_bufptr[-1] == optype) {
2898 while (s < PL_bufend && (*s == ' ' || *s == '\t'))
2900 if (*s == '\n' || *s == '#') {
2901 while (s < PL_bufend && *s != '\n')
2905 thiswhite = newSVpvn(PL_bufptr, s - PL_bufptr);
2906 addmad(newMADsv('#', thiswhite), &thismad, 0);
2923 /* Create new token struct. Note: opvals return early above. */
2924 yylval.tkval = newTOKEN(optype, yylval, thismad);
2931 S_tokenize_use(pTHX_ int is_use, char *s) {
2933 if (PL_expect != XSTATE)
2934 yyerror(Perl_form(aTHX_ "\"%s\" not allowed in expression",
2935 is_use ? "use" : "no"));
2937 if (isDIGIT(*s) || (*s == 'v' && isDIGIT(s[1]))) {
2938 s = force_version(s, TRUE);
2939 if (*s == ';' || (s = SKIPSPACE1(s), *s == ';')) {
2940 start_force(curforce);
2941 NEXTVAL_NEXTTOKE.opval = NULL;
2944 else if (*s == 'v') {
2945 s = force_word(s,WORD,FALSE,TRUE,FALSE);
2946 s = force_version(s, FALSE);
2950 s = force_word(s,WORD,FALSE,TRUE,FALSE);
2951 s = force_version(s, FALSE);
2953 yylval.ival = is_use;
2957 static const char* const exp_name[] =
2958 { "OPERATOR", "TERM", "REF", "STATE", "BLOCK", "ATTRBLOCK",
2959 "ATTRTERM", "TERMBLOCK", "TERMORDORDOR"
2966 Works out what to call the token just pulled out of the input
2967 stream. The yacc parser takes care of taking the ops we return and
2968 stitching them into a tree.
2974 if read an identifier
2975 if we're in a my declaration
2976 croak if they tried to say my($foo::bar)
2977 build the ops for a my() declaration
2978 if it's an access to a my() variable
2979 are we in a sort block?
2980 croak if my($a); $a <=> $b
2981 build ops for access to a my() variable
2982 if in a dq string, and they've said @foo and we can't find @foo
2984 build ops for a bareword
2985 if we already built the token before, use it.
2990 #pragma segment Perl_yylex
2996 register char *s = PL_bufptr;
3002 SV* tmp = newSVpvs("");
3003 PerlIO_printf(Perl_debug_log, "### %"IVdf":LEX_%s/X%s %s\n",
3004 (IV)CopLINE(PL_curcop),
3005 lex_state_names[PL_lex_state],
3006 exp_name[PL_expect],
3007 pv_display(tmp, s, strlen(s), 0, 60));
3010 /* check if there's an identifier for us to look at */
3011 if (PL_pending_ident)
3012 return REPORT(S_pending_ident(aTHX));
3014 /* no identifier pending identification */
3016 switch (PL_lex_state) {
3018 case LEX_NORMAL: /* Some compilers will produce faster */
3019 case LEX_INTERPNORMAL: /* code if we comment these out. */
3023 /* when we've already built the next token, just pull it out of the queue */
3027 yylval = PL_nexttoke[PL_lasttoke].next_val;
3029 thismad = PL_nexttoke[PL_lasttoke].next_mad;
3030 PL_nexttoke[PL_lasttoke].next_mad = 0;
3031 if (thismad && thismad->mad_key == '_') {
3032 thiswhite = (SV*)thismad->mad_val;
3033 thismad->mad_val = 0;
3039 PL_lex_state = PL_lex_defer;
3040 PL_expect = PL_lex_expect;
3041 PL_lex_defer = LEX_NORMAL;
3042 if (!PL_nexttoke[PL_lasttoke].next_type)
3047 yylval = PL_nextval[PL_nexttoke];
3049 PL_lex_state = PL_lex_defer;
3050 PL_expect = PL_lex_expect;
3051 PL_lex_defer = LEX_NORMAL;
3055 /* FIXME - can these be merged? */
3056 return(PL_nexttoke[PL_lasttoke].next_type);
3058 return REPORT(PL_nexttype[PL_nexttoke]);
3061 /* interpolated case modifiers like \L \U, including \Q and \E.
3062 when we get here, PL_bufptr is at the \
3064 case LEX_INTERPCASEMOD:
3066 if (PL_bufptr != PL_bufend && *PL_bufptr != '\\')
3067 Perl_croak(aTHX_ "panic: INTERPCASEMOD");
3069 /* handle \E or end of string */
3070 if (PL_bufptr == PL_bufend || PL_bufptr[1] == 'E') {
3072 if (PL_lex_casemods) {
3073 const char oldmod = PL_lex_casestack[--PL_lex_casemods];
3074 PL_lex_casestack[PL_lex_casemods] = '\0';
3076 if (PL_bufptr != PL_bufend
3077 && (oldmod == 'L' || oldmod == 'U' || oldmod == 'Q')) {
3079 PL_lex_state = LEX_INTERPCONCAT;
3082 thistoken = newSVpvn("\\E",2);
3088 while (PL_bufptr != PL_bufend &&
3089 PL_bufptr[0] == '\\' && PL_bufptr[1] == 'E') {
3091 thiswhite = newSVpvn("",0);
3092 sv_catpvn(thiswhite, PL_bufptr, 2);
3096 if (PL_bufptr != PL_bufend)
3099 PL_lex_state = LEX_INTERPCONCAT;
3103 DEBUG_T({ PerlIO_printf(Perl_debug_log,
3104 "### Saw case modifier\n"); });
3106 if (s[1] == '\\' && s[2] == 'E') {
3110 thiswhite = newSVpvn("",0);
3111 sv_catpvn(thiswhite, PL_bufptr, 4);
3113 PL_lex_state = LEX_INTERPCONCAT;
3118 if (!PL_madskills) /* when just compiling don't need correct */
3119 if (strnEQ(s, "L\\u", 3) || strnEQ(s, "U\\l", 3))
3120 tmp = *s, *s = s[2], s[2] = (char)tmp; /* misordered... */
3121 if ((*s == 'L' || *s == 'U') &&
3122 (strchr(PL_lex_casestack, 'L') || strchr(PL_lex_casestack, 'U'))) {
3123 PL_lex_casestack[--PL_lex_casemods] = '\0';
3126 if (PL_lex_casemods > 10)
3127 Renew(PL_lex_casestack, PL_lex_casemods + 2, char);
3128 PL_lex_casestack[PL_lex_casemods++] = *s;
3129 PL_lex_casestack[PL_lex_casemods] = '\0';
3130 PL_lex_state = LEX_INTERPCONCAT;
3131 start_force(curforce);
3132 NEXTVAL_NEXTTOKE.ival = 0;
3134 start_force(curforce);
3136 NEXTVAL_NEXTTOKE.ival = OP_LCFIRST;
3138 NEXTVAL_NEXTTOKE.ival = OP_UCFIRST;
3140 NEXTVAL_NEXTTOKE.ival = OP_LC;
3142 NEXTVAL_NEXTTOKE.ival = OP_UC;
3144 NEXTVAL_NEXTTOKE.ival = OP_QUOTEMETA;
3146 Perl_croak(aTHX_ "panic: yylex");
3148 SV* tmpsv = newSVpvn("",0);
3149 Perl_sv_catpvf(aTHX_ tmpsv, "\\%c", *s);
3155 if (PL_lex_starts) {
3162 thistoken = newSVpvn("",0);
3165 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
3166 if (PL_lex_casemods == 1 && PL_lex_inpat)
3175 case LEX_INTERPPUSH:
3176 return REPORT(sublex_push());
3178 case LEX_INTERPSTART:
3179 if (PL_bufptr == PL_bufend)
3180 return REPORT(sublex_done());
3181 DEBUG_T({ PerlIO_printf(Perl_debug_log,
3182 "### Interpolated variable\n"); });
3184 PL_lex_dojoin = (*PL_bufptr == '@');
3185 PL_lex_state = LEX_INTERPNORMAL;
3186 if (PL_lex_dojoin) {
3187 start_force(curforce);
3188 NEXTVAL_NEXTTOKE.ival = 0;
3190 start_force(curforce);
3191 force_ident("\"", '$');
3192 start_force(curforce);
3193 NEXTVAL_NEXTTOKE.ival = 0;
3195 start_force(curforce);
3196 NEXTVAL_NEXTTOKE.ival = 0;
3198 start_force(curforce);
3199 NEXTVAL_NEXTTOKE.ival = OP_JOIN; /* emulate join($", ...) */
3202 if (PL_lex_starts++) {
3208 thistoken = newSVpvn("",0);
3211 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
3212 if (!PL_lex_casemods && PL_lex_inpat)
3219 case LEX_INTERPENDMAYBE:
3220 if (intuit_more(PL_bufptr)) {
3221 PL_lex_state = LEX_INTERPNORMAL; /* false alarm, more expr */
3227 if (PL_lex_dojoin) {
3228 PL_lex_dojoin = FALSE;
3229 PL_lex_state = LEX_INTERPCONCAT;
3234 thistoken = newSVpvn("",0);
3239 if (PL_lex_inwhat == OP_SUBST && PL_linestr == PL_lex_repl
3240 && SvEVALED(PL_lex_repl))
3242 if (PL_bufptr != PL_bufend)
3243 Perl_croak(aTHX_ "Bad evalled substitution pattern");
3247 case LEX_INTERPCONCAT:
3249 if (PL_lex_brackets)
3250 Perl_croak(aTHX_ "panic: INTERPCONCAT");
3252 if (PL_bufptr == PL_bufend)
3253 return REPORT(sublex_done());
3255 if (SvIVX(PL_linestr) == '\'') {
3256 SV *sv = newSVsv(PL_linestr);
3259 else if ( PL_hints & HINT_NEW_RE )
3260 sv = new_constant(NULL, 0, "qr", sv, sv, "q");
3261 yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
3265 s = scan_const(PL_bufptr);
3267 PL_lex_state = LEX_INTERPCASEMOD;
3269 PL_lex_state = LEX_INTERPSTART;
3272 if (s != PL_bufptr) {
3273 start_force(curforce);
3275 curmad('X', newSVpvn(PL_bufptr,s-PL_bufptr));
3277 NEXTVAL_NEXTTOKE = yylval;
3280 if (PL_lex_starts++) {
3285 thistoken = newSVpvn("",0);
3288 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
3289 if (!PL_lex_casemods && PL_lex_inpat)
3302 PL_lex_state = LEX_NORMAL;
3303 s = scan_formline(PL_bufptr);
3304 if (!PL_lex_formbrack)
3310 PL_oldoldbufptr = PL_oldbufptr;
3319 realtokenstart = s - SvPVX(PL_linestr); /* assume but undo on ws */
3323 if (isIDFIRST_lazy_if(s,UTF))
3325 Perl_croak(aTHX_ "Unrecognized character \\x%02X", *s & 255);
3328 goto fake_eof; /* emulate EOF on ^D or ^Z */
3337 if (PL_lex_brackets) {
3338 yyerror(PL_lex_formbrack
3339 ? "Format not terminated"
3340 : "Missing right curly or square bracket");
3342 DEBUG_T( { PerlIO_printf(Perl_debug_log,
3343 "### Tokener got EOF\n");
3347 if (s++ < PL_bufend)
3348 goto retry; /* ignore stray nulls */
3351 if (!PL_in_eval && !PL_preambled) {
3352 PL_preambled = TRUE;
3357 sv_setpv(PL_linestr,incl_perldb());
3358 if (SvCUR(PL_linestr))
3359 sv_catpvs(PL_linestr,";");
3361 while(AvFILLp(PL_preambleav) >= 0) {
3362 SV *tmpsv = av_shift(PL_preambleav);
3363 sv_catsv(PL_linestr, tmpsv);
3364 sv_catpvs(PL_linestr, ";");
3367 sv_free((SV*)PL_preambleav);
3368 PL_preambleav = NULL;
3370 if (PL_minus_n || PL_minus_p) {
3371 sv_catpvs(PL_linestr, "LINE: while (<>) {");
3373 sv_catpvs(PL_linestr,"chomp;");
3376 if ((*PL_splitstr == '/' || *PL_splitstr == '\''
3377 || *PL_splitstr == '"')
3378 && strchr(PL_splitstr + 1, *PL_splitstr))
3379 Perl_sv_catpvf(aTHX_ PL_linestr, "our @F=split(%s);", PL_splitstr);
3381 /* "q\0${splitstr}\0" is legal perl. Yes, even NUL
3382 bytes can be used as quoting characters. :-) */
3383 const char *splits = PL_splitstr;
3384 sv_catpvs(PL_linestr, "our @F=split(q\0");
3387 if (*splits == '\\')
3388 sv_catpvn(PL_linestr, splits, 1);
3389 sv_catpvn(PL_linestr, splits, 1);
3390 } while (*splits++);
3391 /* This loop will embed the trailing NUL of
3392 PL_linestr as the last thing it does before
3394 sv_catpvs(PL_linestr, ");");
3398 sv_catpvs(PL_linestr,"our @F=split(' ');");
3402 sv_catpvs(PL_linestr,"use feature ':5.10';");
3403 sv_catpvs(PL_linestr, "\n");
3404 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3405 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3406 PL_last_lop = PL_last_uni = NULL;
3407 if (PERLDB_LINE && PL_curstash != PL_debstash) {
3408 SV * const sv = newSV(0);
3410 sv_upgrade(sv, SVt_PVMG);
3411 sv_setsv(sv,PL_linestr);
3414 av_store(CopFILEAVx(PL_curcop),(I32)CopLINE(PL_curcop),sv);
3419 bof = PL_rsfp ? TRUE : FALSE;
3420 if ((s = filter_gets(PL_linestr, PL_rsfp, 0)) == NULL) {
3423 realtokenstart = -1;
3426 if (PL_preprocess && !PL_in_eval)
3427 (void)PerlProc_pclose(PL_rsfp);
3428 else if ((PerlIO *)PL_rsfp == PerlIO_stdin())
3429 PerlIO_clearerr(PL_rsfp);
3431 (void)PerlIO_close(PL_rsfp);
3433 PL_doextract = FALSE;
3435 if (!PL_in_eval && (PL_minus_n || PL_minus_p)) {
3440 sv_setpv(PL_linestr,PL_minus_p
3441 ? ";}continue{print;}" : ";}");
3442 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3443 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3444 PL_last_lop = PL_last_uni = NULL;
3445 PL_minus_n = PL_minus_p = 0;
3448 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3449 PL_last_lop = PL_last_uni = NULL;
3450 sv_setpvn(PL_linestr,"",0);
3451 TOKEN(';'); /* not infinite loop because rsfp is NULL now */
3453 /* If it looks like the start of a BOM or raw UTF-16,
3454 * check if it in fact is. */
3460 #ifdef PERLIO_IS_STDIO
3461 # ifdef __GNU_LIBRARY__
3462 # if __GNU_LIBRARY__ == 1 /* Linux glibc5 */
3463 # define FTELL_FOR_PIPE_IS_BROKEN
3467 # if __GLIBC__ == 1 /* maybe some glibc5 release had it like this? */
3468 # define FTELL_FOR_PIPE_IS_BROKEN
3473 #ifdef FTELL_FOR_PIPE_IS_BROKEN
3474 /* This loses the possibility to detect the bof
3475 * situation on perl -P when the libc5 is being used.
3476 * Workaround? Maybe attach some extra state to PL_rsfp?
3479 bof = PerlIO_tell(PL_rsfp) == SvCUR(PL_linestr);
3481 bof = PerlIO_tell(PL_rsfp) == (Off_t)SvCUR(PL_linestr);
3484 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3485 s = swallow_bom((U8*)s);
3489 /* Incest with pod. */
3492 sv_catsv(thiswhite, PL_linestr);
3494 if (*s == '=' && strnEQ(s, "=cut", 4)) {
3495 sv_setpvn(PL_linestr, "", 0);
3496 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3497 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3498 PL_last_lop = PL_last_uni = NULL;
3499 PL_doextract = FALSE;
3503 } while (PL_doextract);
3504 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = s;
3505 if (PERLDB_LINE && PL_curstash != PL_debstash) {
3506 SV * const sv = newSV(0);
3508 sv_upgrade(sv, SVt_PVMG);
3509 sv_setsv(sv,PL_linestr);
3512 av_store(CopFILEAVx(PL_curcop),(I32)CopLINE(PL_curcop),sv);
3514 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3515 PL_last_lop = PL_last_uni = NULL;
3516 if (CopLINE(PL_curcop) == 1) {
3517 while (s < PL_bufend && isSPACE(*s))
3519 if (*s == ':' && s[1] != ':') /* for csh execing sh scripts */
3523 thiswhite = newSVpvn(PL_linestart, s - PL_linestart);
3527 if (*s == '#' && *(s+1) == '!')
3529 #ifdef ALTERNATE_SHEBANG
3531 static char const as[] = ALTERNATE_SHEBANG;
3532 if (*s == as[0] && strnEQ(s, as, sizeof(as) - 1))
3533 d = s + (sizeof(as) - 1);
3535 #endif /* ALTERNATE_SHEBANG */
3544 while (*d && !isSPACE(*d))
3548 #ifdef ARG_ZERO_IS_SCRIPT
3549 if (ipathend > ipath) {
3551 * HP-UX (at least) sets argv[0] to the script name,
3552 * which makes $^X incorrect. And Digital UNIX and Linux,
3553 * at least, set argv[0] to the basename of the Perl
3554 * interpreter. So, having found "#!", we'll set it right.
3556 SV * const x = GvSV(gv_fetchpvs("\030", GV_ADD|GV_NOTQUAL,
3558 assert(SvPOK(x) || SvGMAGICAL(x));
3559 if (sv_eq(x, CopFILESV(PL_curcop))) {
3560 sv_setpvn(x, ipath, ipathend - ipath);
3566 const char *bstart = SvPV_const(CopFILESV(PL_curcop),blen);
3567 const char * const lstart = SvPV_const(x,llen);
3569 bstart += blen - llen;
3570 if (strnEQ(bstart, lstart, llen) && bstart[-1] == '/') {
3571 sv_setpvn(x, ipath, ipathend - ipath);
3576 TAINT_NOT; /* $^X is always tainted, but that's OK */
3578 #endif /* ARG_ZERO_IS_SCRIPT */
3583 d = instr(s,"perl -");
3585 d = instr(s,"perl");
3587 /* avoid getting into infinite loops when shebang
3588 * line contains "Perl" rather than "perl" */
3590 for (d = ipathend-4; d >= ipath; --d) {
3591 if ((*d == 'p' || *d == 'P')
3592 && !ibcmp(d, "perl", 4))
3602 #ifdef ALTERNATE_SHEBANG
3604 * If the ALTERNATE_SHEBANG on this system starts with a
3605 * character that can be part of a Perl expression, then if
3606 * we see it but not "perl", we're probably looking at the
3607 * start of Perl code, not a request to hand off to some
3608 * other interpreter. Similarly, if "perl" is there, but
3609 * not in the first 'word' of the line, we assume the line
3610 * contains the start of the Perl program.
3612 if (d && *s != '#') {
3613 const char *c = ipath;
3614 while (*c && !strchr("; \t\r\n\f\v#", *c))
3617 d = NULL; /* "perl" not in first word; ignore */
3619 *s = '#'; /* Don't try to parse shebang line */
3621 #endif /* ALTERNATE_SHEBANG */
3622 #ifndef MACOS_TRADITIONAL
3627 !instr(s,"indir") &&
3628 instr(PL_origargv[0],"perl"))
3635 while (s < PL_bufend && isSPACE(*s))
3637 if (s < PL_bufend) {
3638 Newxz(newargv,PL_origargc+3,char*);
3640 while (s < PL_bufend && !isSPACE(*s))
3643 Copy(PL_origargv+1, newargv+2, PL_origargc+1, char*);
3646 newargv = PL_origargv;
3649 PerlProc_execv(ipath, EXEC_ARGV_CAST(newargv));
3651 Perl_croak(aTHX_ "Can't exec %s", ipath);
3655 while (*d && !isSPACE(*d)) d++;
3656 while (SPACE_OR_TAB(*d)) d++;
3659 const bool switches_done = PL_doswitches;
3660 const U32 oldpdb = PL_perldb;
3661 const bool oldn = PL_minus_n;
3662 const bool oldp = PL_minus_p;
3665 if (*d == 'M' || *d == 'm' || *d == 'C') {
3666 const char * const m = d;
3667 while (*d && !isSPACE(*d)) d++;
3668 Perl_croak(aTHX_ "Too late for \"-%.*s\" option",
3671 d = moreswitches(d);
3673 if (PL_doswitches && !switches_done) {
3674 int argc = PL_origargc;
3675 char **argv = PL_origargv;
3678 } while (argc && argv[0][0] == '-' && argv[0][1]);
3679 init_argv_symbols(argc,argv);
3681 if ((PERLDB_LINE && !oldpdb) ||
3682 ((PL_minus_n || PL_minus_p) && !(oldn || oldp)))
3683 /* if we have already added "LINE: while (<>) {",
3684 we must not do it again */
3686 sv_setpvn(PL_linestr, "", 0);
3687 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3688 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3689 PL_last_lop = PL_last_uni = NULL;
3690 PL_preambled = FALSE;
3692 (void)gv_fetchfile(PL_origfilename);
3699 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
3701 PL_lex_state = LEX_FORMLINE;
3706 #ifdef PERL_STRICT_CR
3707 Perl_warn(aTHX_ "Illegal character \\%03o (carriage return)", '\r');
3709 "\t(Maybe you didn't strip carriage returns after a network transfer?)\n");
3711 case ' ': case '\t': case '\f': case 013:
3712 #ifdef MACOS_TRADITIONAL
3716 realtokenstart = -1;
3725 realtokenstart = -1;
3729 if (PL_lex_state != LEX_NORMAL || (PL_in_eval && !PL_rsfp)) {
3730 if (*s == '#' && s == PL_linestart && PL_in_eval && !PL_rsfp) {
3731 /* handle eval qq[#line 1 "foo"\n ...] */
3732 CopLINE_dec(PL_curcop);
3735 if (PL_madskills && !PL_lex_formbrack && !PL_in_eval) {
3737 if (!PL_in_eval || PL_rsfp)
3742 while (d < PL_bufend && *d != '\n')
3746 else if (d > PL_bufend) /* Found by Ilya: feed random input to Perl. */
3747 Perl_croak(aTHX_ "panic: input overflow");
3750 thiswhite = newSVpvn(s, d - s);
3755 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
3757 PL_lex_state = LEX_FORMLINE;
3763 if (PL_madskills && CopLINE(PL_curcop) >= 1 && !PL_lex_formbrack) {
3764 if (CopLINE(PL_curcop) == 1 && s[0] == '#' && s[1] == '!') {
3767 TOKEN(PEG); /* make sure any #! line is accessible */
3772 /* if (PL_madskills && PL_lex_formbrack) { */
3774 while (d < PL_bufend && *d != '\n')
3778 else if (d > PL_bufend) /* Found by Ilya: feed random input to Perl. */
3779 Perl_croak(aTHX_ "panic: input overflow");
3780 if (PL_madskills && CopLINE(PL_curcop) >= 1) {
3782 thiswhite = newSVpvn("",0);
3783 if (CopLINE(PL_curcop) == 1) {
3784 sv_setpvn(thiswhite, "", 0);
3787 sv_catpvn(thiswhite, s, d - s);
3801 if (s[1] && isALPHA(s[1]) && !isALNUM(s[2])) {
3809 while (s < PL_bufend && SPACE_OR_TAB(*s))
3812 if (strnEQ(s,"=>",2)) {
3813 s = force_word(PL_bufptr,WORD,FALSE,FALSE,FALSE);
3814 DEBUG_T( { S_printbuf(aTHX_
3815 "### Saw unary minus before =>, forcing word %s\n", s);
3817 OPERATOR('-'); /* unary minus */
3819 PL_last_uni = PL_oldbufptr;
3821 case 'r': ftst = OP_FTEREAD; break;
3822 case 'w': ftst = OP_FTEWRITE; break;
3823 case 'x': ftst = OP_FTEEXEC; break;
3824 case 'o': ftst = OP_FTEOWNED; break;
3825 case 'R': ftst = OP_FTRREAD; break;
3826 case 'W': ftst = OP_FTRWRITE; break;
3827 case 'X': ftst = OP_FTREXEC; break;
3828 case 'O': ftst = OP_FTROWNED; break;
3829 case 'e': ftst = OP_FTIS; break;
3830 case 'z': ftst = OP_FTZERO; break;
3831 case 's': ftst = OP_FTSIZE; break;
3832 case 'f': ftst = OP_FTFILE; break;
3833 case 'd': ftst = OP_FTDIR; break;
3834 case 'l': ftst = OP_FTLINK; break;
3835 case 'p': ftst = OP_FTPIPE; break;
3836 case 'S': ftst = OP_FTSOCK; break;
3837 case 'u': ftst = OP_FTSUID; break;
3838 case 'g': ftst = OP_FTSGID; break;
3839 case 'k': ftst = OP_FTSVTX; break;
3840 case 'b': ftst = OP_FTBLK; break;
3841 case 'c': ftst = OP_FTCHR; break;
3842 case 't': ftst = OP_FTTTY; break;
3843 case 'T': ftst = OP_FTTEXT; break;
3844 case 'B': ftst = OP_FTBINARY; break;
3845 case 'M': case 'A': case 'C':
3846 gv_fetchpvs("\024", GV_ADD|GV_NOTQUAL, SVt_PV);
3848 case 'M': ftst = OP_FTMTIME; break;
3849 case 'A': ftst = OP_FTATIME; break;
3850 case 'C': ftst = OP_FTCTIME; break;
3858 PL_last_lop_op = (OPCODE)ftst;
3859 DEBUG_T( { PerlIO_printf(Perl_debug_log,
3860 "### Saw file test %c\n", (int)tmp);
3865 /* Assume it was a minus followed by a one-letter named
3866 * subroutine call (or a -bareword), then. */
3867 DEBUG_T( { PerlIO_printf(Perl_debug_log,
3868 "### '-%c' looked like a file test but was not\n",
3875 const char tmp = *s++;
3878 if (PL_expect == XOPERATOR)
3883 else if (*s == '>') {
3886 if (isIDFIRST_lazy_if(s,UTF)) {
3887 s = force_word(s,METHOD,FALSE,TRUE,FALSE);
3895 if (PL_expect == XOPERATOR)
3898 if (isSPACE(*s) || !isSPACE(*PL_bufptr))
3900 OPERATOR('-'); /* unary minus */
3906 const char tmp = *s++;
3909 if (PL_expect == XOPERATOR)
3914 if (PL_expect == XOPERATOR)
3917 if (isSPACE(*s) || !isSPACE(*PL_bufptr))
3924 if (PL_expect != XOPERATOR) {
3925 s = scan_ident(s, PL_bufend, PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
3926 PL_expect = XOPERATOR;
3927 force_ident(PL_tokenbuf, '*');
3940 if (PL_expect == XOPERATOR) {
3944 PL_tokenbuf[0] = '%';
3945 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1, TRUE);
3946 if (!PL_tokenbuf[1]) {
3949 PL_pending_ident = '%';
3960 && (PL_expect == XOPERATOR || PL_expect == XTERMORDORDOR)
3961 && FEATURE_IS_ENABLED("~~"))
3968 const char tmp = *s++;
3974 goto just_a_word_zero_gv;
3977 switch (PL_expect) {
3983 if (!PL_in_my || PL_lex_state != LEX_NORMAL)
3985 PL_bufptr = s; /* update in case we back off */
3991 PL_expect = XTERMBLOCK;
3994 stuffstart = s - SvPVX(PL_linestr) - 1;
3998 while (isIDFIRST_lazy_if(s,UTF)) {
4000 d = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
4001 if (isLOWER(*s) && (tmp = keyword(PL_tokenbuf, len))) {
4002 if (tmp < 0) tmp = -tmp;
4018 d = scan_str(d,TRUE,TRUE);
4020 /* MUST advance bufptr here to avoid bogus
4021 "at end of line" context messages from yyerror().
4023 PL_bufptr = s + len;
4024 yyerror("Unterminated attribute parameter in attribute list");
4027 return REPORT(0); /* EOF indicator */
4031 SV *sv = newSVpvn(s, len);
4032 sv_catsv(sv, PL_lex_stuff);
4033 attrs = append_elem(OP_LIST, attrs,
4034 newSVOP(OP_CONST, 0, sv));
4035 SvREFCNT_dec(PL_lex_stuff);
4036 PL_lex_stuff = NULL;
4039 if (len == 6 && strnEQ(s, "unique", len)) {
4040 if (PL_in_my == KEY_our)
4042 GvUNIQUE_on(cGVOPx_gv(yylval.opval));
4044 /*EMPTY*/; /* skip to avoid loading attributes.pm */
4047 Perl_croak(aTHX_ "The 'unique' attribute may only be applied to 'our' variables");
4050 /* NOTE: any CV attrs applied here need to be part of
4051 the CVf_BUILTIN_ATTRS define in cv.h! */
4052 else if (!PL_in_my && len == 6 && strnEQ(s, "lvalue", len))
4053 CvLVALUE_on(PL_compcv);
4054 else if (!PL_in_my && len == 6 && strnEQ(s, "locked", len))
4055 CvLOCKED_on(PL_compcv);
4056 else if (!PL_in_my && len == 6 && strnEQ(s, "method", len))
4057 CvMETHOD_on(PL_compcv);
4058 else if (!PL_in_my && len == 9 && strnEQ(s, "assertion", len))
4059 CvASSERTION_on(PL_compcv);
4060 /* After we've set the flags, it could be argued that
4061 we don't need to do the attributes.pm-based setting
4062 process, and shouldn't bother appending recognized
4063 flags. To experiment with that, uncomment the
4064 following "else". (Note that's already been
4065 uncommented. That keeps the above-applied built-in
4066 attributes from being intercepted (and possibly
4067 rejected) by a package's attribute routines, but is
4068 justified by the performance win for the common case
4069 of applying only built-in attributes.) */
4071 attrs = append_elem(OP_LIST, attrs,
4072 newSVOP(OP_CONST, 0,
4076 if (*s == ':' && s[1] != ':')
4079 break; /* require real whitespace or :'s */
4080 /* XXX losing whitespace on sequential attributes here */
4084 = (PL_expect == XOPERATOR ? '=' : '{'); /*'}(' for vi */
4085 if (*s != ';' && *s != '}' && *s != tmp
4086 && (tmp != '=' || *s != ')')) {
4087 const char q = ((*s == '\'') ? '"' : '\'');
4088 /* If here for an expression, and parsed no attrs, back
4090 if (tmp == '=' && !attrs) {
4094 /* MUST advance bufptr here to avoid bogus "at end of line"
4095 context messages from yyerror().
4099 ? Perl_form(aTHX_ "Invalid separator character "
4100 "%c%c%c in attribute list", q, *s, q)
4101 : "Unterminated attribute list" );
4109 start_force(curforce);
4110 NEXTVAL_NEXTTOKE.opval = attrs;
4111 CURMAD('_', nextwhite);
4116 thistoken = newSVpvn(SvPVX(PL_linestr) + stuffstart,
4117 (s - SvPVX(PL_linestr)) - stuffstart);
4125 if (PL_last_lop == PL_oldoldbufptr || PL_last_uni == PL_oldoldbufptr)
4126 PL_oldbufptr = PL_oldoldbufptr; /* allow print(STDOUT 123) */
4134 const char tmp = *s++;
4139 const char tmp = *s++;
4147 if (PL_lex_brackets <= 0)
4148 yyerror("Unmatched right square bracket");
4151 if (PL_lex_state == LEX_INTERPNORMAL) {
4152 if (PL_lex_brackets == 0) {
4153 if (*s != '[' && *s != '{' && (*s != '-' || s[1] != '>'))
4154 PL_lex_state = LEX_INTERPEND;
4161 if (PL_lex_brackets > 100) {
4162 Renew(PL_lex_brackstack, PL_lex_brackets + 10, char);
4164 switch (PL_expect) {
4166 if (PL_lex_formbrack) {
4170 if (PL_oldoldbufptr == PL_last_lop)
4171 PL_lex_brackstack[PL_lex_brackets++] = XTERM;
4173 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
4174 OPERATOR(HASHBRACK);
4176 while (s < PL_bufend && SPACE_OR_TAB(*s))
4179 PL_tokenbuf[0] = '\0';
4180 if (d < PL_bufend && *d == '-') {
4181 PL_tokenbuf[0] = '-';
4183 while (d < PL_bufend && SPACE_OR_TAB(*d))
4186 if (d < PL_bufend && isIDFIRST_lazy_if(d,UTF)) {
4187 d = scan_word(d, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1,
4189 while (d < PL_bufend && SPACE_OR_TAB(*d))
4192 const char minus = (PL_tokenbuf[0] == '-');
4193 s = force_word(s + minus, WORD, FALSE, TRUE, FALSE);
4201 PL_lex_brackstack[PL_lex_brackets++] = XSTATE;
4206 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
4211 if (PL_oldoldbufptr == PL_last_lop)
4212 PL_lex_brackstack[PL_lex_brackets++] = XTERM;
4214 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
4217 if (PL_expect == XREF && PL_lex_state == LEX_INTERPNORMAL) {
4219 /* This hack is to get the ${} in the message. */
4221 yyerror("syntax error");
4224 OPERATOR(HASHBRACK);
4226 /* This hack serves to disambiguate a pair of curlies
4227 * as being a block or an anon hash. Normally, expectation
4228 * determines that, but in cases where we're not in a
4229 * position to expect anything in particular (like inside
4230 * eval"") we have to resolve the ambiguity. This code
4231 * covers the case where the first term in the curlies is a
4232 * quoted string. Most other cases need to be explicitly
4233 * disambiguated by prepending a "+" before the opening
4234 * curly in order to force resolution as an anon hash.
4236 * XXX should probably propagate the outer expectation
4237 * into eval"" to rely less on this hack, but that could
4238 * potentially break current behavior of eval"".
4242 if (*s == '\'' || *s == '"' || *s == '`') {
4243 /* common case: get past first string, handling escapes */
4244 for (t++; t < PL_bufend && *t != *s;)
4245 if (*t++ == '\\' && (*t == '\\' || *t == *s))
4249 else if (*s == 'q') {
4252 || ((*t == 'q' || *t == 'x') && ++t < PL_bufend
4255 /* skip q//-like construct */
4257 char open, close, term;
4260 while (t < PL_bufend && isSPACE(*t))
4262 /* check for q => */
4263 if (t+1 < PL_bufend && t[0] == '=' && t[1] == '>') {
4264 OPERATOR(HASHBRACK);
4268 if (term && (tmps = strchr("([{< )]}> )]}>",term)))
4272 for (t++; t < PL_bufend; t++) {
4273 if (*t == '\\' && t+1 < PL_bufend && open != '\\')
4275 else if (*t == open)
4279 for (t++; t < PL_bufend; t++) {
4280 if (*t == '\\' && t+1 < PL_bufend)
4282 else if (*t == close && --brackets <= 0)
4284 else if (*t == open)
4291 /* skip plain q word */
4292 while (t < PL_bufend && isALNUM_lazy_if(t,UTF))
4295 else if (isALNUM_lazy_if(t,UTF)) {
4297 while (t < PL_bufend && isALNUM_lazy_if(t,UTF))
4300 while (t < PL_bufend && isSPACE(*t))
4302 /* if comma follows first term, call it an anon hash */
4303 /* XXX it could be a comma expression with loop modifiers */
4304 if (t < PL_bufend && ((*t == ',' && (*s == 'q' || !isLOWER(*s)))
4305 || (*t == '=' && t[1] == '>')))
4306 OPERATOR(HASHBRACK);
4307 if (PL_expect == XREF)
4310 PL_lex_brackstack[PL_lex_brackets-1] = XSTATE;
4316 yylval.ival = CopLINE(PL_curcop);
4317 if (isSPACE(*s) || *s == '#')
4318 PL_copline = NOLINE; /* invalidate current command line number */
4323 if (PL_lex_brackets <= 0)
4324 yyerror("Unmatched right curly bracket");
4326 PL_expect = (expectation)PL_lex_brackstack[--PL_lex_brackets];
4327 if (PL_lex_brackets < PL_lex_formbrack && PL_lex_state != LEX_INTERPNORMAL)
4328 PL_lex_formbrack = 0;
4329 if (PL_lex_state == LEX_INTERPNORMAL) {
4330 if (PL_lex_brackets == 0) {
4331 if (PL_expect & XFAKEBRACK) {
4332 PL_expect &= XENUMMASK;
4333 PL_lex_state = LEX_INTERPEND;
4338 thiswhite = newSVpvn("",0);
4339 sv_catpvn(thiswhite,"}",1);
4342 return yylex(); /* ignore fake brackets */
4344 if (*s == '-' && s[1] == '>')
4345 PL_lex_state = LEX_INTERPENDMAYBE;
4346 else if (*s != '[' && *s != '{')
4347 PL_lex_state = LEX_INTERPEND;
4350 if (PL_expect & XFAKEBRACK) {
4351 PL_expect &= XENUMMASK;
4353 return yylex(); /* ignore fake brackets */
4355 start_force(curforce);
4357 curmad('X', newSVpvn(s-1,1));
4358 CURMAD('_', thiswhite);
4363 thistoken = newSVpvn("",0);
4371 if (PL_expect == XOPERATOR) {
4372 if (PL_bufptr == PL_linestart && ckWARN(WARN_SEMICOLON)
4373 && isIDFIRST_lazy_if(s,UTF))
4375 CopLINE_dec(PL_curcop);
4376 Perl_warner(aTHX_ packWARN(WARN_SEMICOLON), PL_warn_nosemi);
4377 CopLINE_inc(PL_curcop);
4382 s = scan_ident(s - 1, PL_bufend, PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
4384 PL_expect = XOPERATOR;
4385 force_ident(PL_tokenbuf, '&');
4389 yylval.ival = (OPpENTERSUB_AMPER<<8);
4401 const char tmp = *s++;
4408 if (tmp && isSPACE(*s) && ckWARN(WARN_SYNTAX)
4409 && strchr("+-*/%.^&|<",tmp))
4410 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
4411 "Reversed %c= operator",(int)tmp);
4413 if (PL_expect == XSTATE && isALPHA(tmp) &&
4414 (s == PL_linestart+1 || s[-2] == '\n') )
4416 if (PL_in_eval && !PL_rsfp) {
4421 if (strnEQ(s,"=cut",4)) {
4437 thiswhite = newSVpvn("",0);
4438 sv_catpvn(thiswhite, PL_linestart,
4439 PL_bufend - PL_linestart);
4443 PL_doextract = TRUE;
4447 if (PL_lex_brackets < PL_lex_formbrack) {
4449 #ifdef PERL_STRICT_CR
4450 for (t = s; SPACE_OR_TAB(*t); t++) ;
4452 for (t = s; SPACE_OR_TAB(*t) || *t == '\r'; t++) ;
4454 if (*t == '\n' || *t == '#') {
4465 const char tmp = *s++;
4467 /* was this !=~ where !~ was meant?
4468 * warn on m:!=~\s+([/?]|[msy]\W|tr\W): */
4470 if (*s == '~' && ckWARN(WARN_SYNTAX)) {
4471 const char *t = s+1;
4473 while (t < PL_bufend && isSPACE(*t))
4476 if (*t == '/' || *t == '?' ||
4477 ((*t == 'm' || *t == 's' || *t == 'y')
4478 && !isALNUM(t[1])) ||
4479 (*t == 't' && t[1] == 'r' && !isALNUM(t[2])))
4480 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
4481 "!=~ should be !~");
4491 if (PL_expect != XOPERATOR) {
4492 if (s[1] != '<' && !strchr(s,'>'))
4495 s = scan_heredoc(s);
4497 s = scan_inputsymbol(s);
4498 TERM(sublex_start());
4504 SHop(OP_LEFT_SHIFT);
4518 const char tmp = *s++;
4520 SHop(OP_RIGHT_SHIFT);
4530 if (PL_expect == XOPERATOR) {
4531 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
4533 deprecate_old(commaless_variable_list);
4534 return REPORT(','); /* grandfather non-comma-format format */
4538 if (s[1] == '#' && (isIDFIRST_lazy_if(s+2,UTF) || strchr("{$:+-", s[2]))) {
4539 PL_tokenbuf[0] = '@';
4540 s = scan_ident(s + 1, PL_bufend, PL_tokenbuf + 1,
4541 sizeof PL_tokenbuf - 1, FALSE);
4542 if (PL_expect == XOPERATOR)
4543 no_op("Array length", s);
4544 if (!PL_tokenbuf[1])
4546 PL_expect = XOPERATOR;
4547 PL_pending_ident = '#';
4551 PL_tokenbuf[0] = '$';
4552 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1,
4553 sizeof PL_tokenbuf - 1, FALSE);
4554 if (PL_expect == XOPERATOR)
4556 if (!PL_tokenbuf[1]) {
4558 yyerror("Final $ should be \\$ or $name");
4562 /* This kludge not intended to be bulletproof. */
4563 if (PL_tokenbuf[1] == '[' && !PL_tokenbuf[2]) {
4564 yylval.opval = newSVOP(OP_CONST, 0,
4565 newSViv(PL_compiling.cop_arybase));
4566 yylval.opval->op_private = OPpCONST_ARYBASE;
4572 const char tmp = *s;
4573 if (PL_lex_state == LEX_NORMAL)
4576 if ((PL_expect != XREF || PL_oldoldbufptr == PL_last_lop)
4577 && intuit_more(s)) {
4579 PL_tokenbuf[0] = '@';
4580 if (ckWARN(WARN_SYNTAX)) {
4583 isSPACE(*t) || isALNUM_lazy_if(t,UTF) || *t == '$';
4586 PL_bufptr = PEEKSPACE(PL_bufptr); /* XXX can realloc */
4587 while (t < PL_bufend && *t != ']')
4589 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
4590 "Multidimensional syntax %.*s not supported",
4591 (int)((t - PL_bufptr) + 1), PL_bufptr);
4595 else if (*s == '{') {
4597 PL_tokenbuf[0] = '%';
4598 if (strEQ(PL_tokenbuf+1, "SIG") && ckWARN(WARN_SYNTAX)
4599 && (t = strchr(s, '}')) && (t = strchr(t, '=')))
4601 char tmpbuf[sizeof PL_tokenbuf];
4602 for (t++; isSPACE(*t); t++) ;
4603 if (isIDFIRST_lazy_if(t,UTF)) {
4605 t = scan_word(t, tmpbuf, sizeof tmpbuf, TRUE,
4607 for (; isSPACE(*t); t++) ;
4608 if (*t == ';' && get_cv(tmpbuf, FALSE))
4609 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
4610 "You need to quote \"%s\"",
4617 PL_expect = XOPERATOR;
4618 if (PL_lex_state == LEX_NORMAL && isSPACE((char)tmp)) {
4619 const bool islop = (PL_last_lop == PL_oldoldbufptr);
4620 if (!islop || PL_last_lop_op == OP_GREPSTART)
4621 PL_expect = XOPERATOR;
4622 else if (strchr("$@\"'`q", *s))
4623 PL_expect = XTERM; /* e.g. print $fh "foo" */
4624 else if (strchr("&*<%", *s) && isIDFIRST_lazy_if(s+1,UTF))
4625 PL_expect = XTERM; /* e.g. print $fh &sub */
4626 else if (isIDFIRST_lazy_if(s,UTF)) {
4627 char tmpbuf[sizeof PL_tokenbuf];
4629 scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
4630 if ((t2 = keyword(tmpbuf, len))) {
4631 /* binary operators exclude handle interpretations */
4643 PL_expect = XTERM; /* e.g. print $fh length() */
4648 PL_expect = XTERM; /* e.g. print $fh subr() */
4651 else if (isDIGIT(*s))
4652 PL_expect = XTERM; /* e.g. print $fh 3 */
4653 else if (*s == '.' && isDIGIT(s[1]))
4654 PL_expect = XTERM; /* e.g. print $fh .3 */
4655 else if ((*s == '?' || *s == '-' || *s == '+')
4656 && !isSPACE(s[1]) && s[1] != '=')
4657 PL_expect = XTERM; /* e.g. print $fh -1 */
4658 else if (*s == '/' && !isSPACE(s[1]) && s[1] != '='
4660 PL_expect = XTERM; /* e.g. print $fh /.../
4661 XXX except DORDOR operator
4663 else if (*s == '<' && s[1] == '<' && !isSPACE(s[2])
4665 PL_expect = XTERM; /* print $fh <<"EOF" */
4668 PL_pending_ident = '$';
4672 if (PL_expect == XOPERATOR)
4674 PL_tokenbuf[0] = '@';
4675 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1, FALSE);
4676 if (!PL_tokenbuf[1]) {
4679 if (PL_lex_state == LEX_NORMAL)
4681 if ((PL_expect != XREF || PL_oldoldbufptr == PL_last_lop) && intuit_more(s)) {
4683 PL_tokenbuf[0] = '%';
4685 /* Warn about @ where they meant $. */
4686 if (*s == '[' || *s == '{') {
4687 if (ckWARN(WARN_SYNTAX)) {
4688 const char *t = s + 1;
4689 while (*t && (isALNUM_lazy_if(t,UTF) || strchr(" \t$#+-'\"", *t)))
4691 if (*t == '}' || *t == ']') {
4693 PL_bufptr = PEEKSPACE(PL_bufptr); /* XXX can realloc */
4694 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
4695 "Scalar value %.*s better written as $%.*s",
4696 (int)(t-PL_bufptr), PL_bufptr,
4697 (int)(t-PL_bufptr-1), PL_bufptr+1);
4702 PL_pending_ident = '@';
4705 case '/': /* may be division, defined-or, or pattern */
4706 if (PL_expect == XTERMORDORDOR && s[1] == '/') {
4710 case '?': /* may either be conditional or pattern */
4711 if(PL_expect == XOPERATOR) {
4719 /* A // operator. */
4729 /* Disable warning on "study /blah/" */
4730 if (PL_oldoldbufptr == PL_last_uni
4731 && (*PL_last_uni != 's' || s - PL_last_uni < 5
4732 || memNE(PL_last_uni, "study", 5)
4733 || isALNUM_lazy_if(PL_last_uni+5,UTF)
4736 s = scan_pat(s,OP_MATCH);
4737 TERM(sublex_start());
4741 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack
4742 #ifdef PERL_STRICT_CR
4745 && (s[1] == '\n' || (s[1] == '\r' && s[2] == '\n'))
4747 && (s == PL_linestart || s[-1] == '\n') )
4749 PL_lex_formbrack = 0;
4753 if (PL_expect == XOPERATOR || !isDIGIT(s[1])) {
4759 yylval.ival = OPf_SPECIAL;
4765 if (PL_expect != XOPERATOR)
4770 case '0': case '1': case '2': case '3': case '4':
4771 case '5': case '6': case '7': case '8': case '9':
4772 s = scan_num(s, &yylval);
4773 DEBUG_T( { S_printbuf(aTHX_ "### Saw number in %s\n", s); } );
4774 if (PL_expect == XOPERATOR)
4779 s = scan_str(s,!!PL_madskills,FALSE);
4780 DEBUG_T( { S_printbuf(aTHX_ "### Saw string before %s\n", s); } );
4781 if (PL_expect == XOPERATOR) {
4782 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
4784 deprecate_old(commaless_variable_list);
4785 return REPORT(','); /* grandfather non-comma-format format */
4791 missingterm((char*)0);
4792 yylval.ival = OP_CONST;
4793 TERM(sublex_start());
4796 s = scan_str(s,!!PL_madskills,FALSE);
4797 DEBUG_T( { S_printbuf(aTHX_ "### Saw string before %s\n", s); } );
4798 if (PL_expect == XOPERATOR) {
4799 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
4801 deprecate_old(commaless_variable_list);
4802 return REPORT(','); /* grandfather non-comma-format format */
4808 missingterm((char*)0);
4809 yylval.ival = OP_CONST;
4810 /* FIXME. I think that this can be const if char *d is replaced by
4811 more localised variables. */
4812 for (d = SvPV(PL_lex_stuff, len); len; len--, d++) {
4813 if (*d == '$' || *d == '@' || *d == '\\' || !UTF8_IS_INVARIANT((U8)*d)) {
4814 yylval.ival = OP_STRINGIFY;
4818 TERM(sublex_start());
4821 s = scan_str(s,!!PL_madskills,FALSE);
4822 DEBUG_T( { S_printbuf(aTHX_ "### Saw backtick string before %s\n", s); } );
4823 if (PL_expect == XOPERATOR)
4824 no_op("Backticks",s);
4826 missingterm((char*)0);
4827 yylval.ival = OP_BACKTICK;
4829 TERM(sublex_start());
4833 if (PL_lex_inwhat && isDIGIT(*s) && ckWARN(WARN_SYNTAX))
4834 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),"Can't use \\%c to mean $%c in expression",
4836 if (PL_expect == XOPERATOR)
4837 no_op("Backslash",s);
4841 if (isDIGIT(s[1]) && PL_expect != XOPERATOR) {
4842 char *start = s + 2;
4843 while (isDIGIT(*start) || *start == '_')
4845 if (*start == '.' && isDIGIT(start[1])) {
4846 s = scan_num(s, &yylval);
4849 /* avoid v123abc() or $h{v1}, allow C<print v10;> */
4850 else if (!isALPHA(*start) && (PL_expect == XTERM
4851 || PL_expect == XREF || PL_expect == XSTATE
4852 || PL_expect == XTERMORDORDOR)) {
4853 const char c = *start;
4856 gv = gv_fetchpv(s, 0, SVt_PVCV);
4859 s = scan_num(s, &yylval);
4866 if (isDIGIT(s[1]) && PL_expect == XOPERATOR) {
4902 I32 orig_keyword = 0;
4907 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
4909 /* Some keywords can be followed by any delimiter, including ':' */
4910 tmp = ((len == 1 && strchr("msyq", PL_tokenbuf[0])) ||
4911 (len == 2 && ((PL_tokenbuf[0] == 't' && PL_tokenbuf[1] == 'r') ||
4912 (PL_tokenbuf[0] == 'q' &&
4913 strchr("qwxr", PL_tokenbuf[1])))));
4915 /* x::* is just a word, unless x is "CORE" */
4916 if (!tmp && *s == ':' && s[1] == ':' && strNE(PL_tokenbuf, "CORE"))
4920 while (d < PL_bufend && isSPACE(*d))
4921 d++; /* no comments skipped here, or s### is misparsed */
4923 /* Is this a label? */
4924 if (!tmp && PL_expect == XSTATE
4925 && d < PL_bufend && *d == ':' && *(d + 1) != ':') {
4927 yylval.pval = savepv(PL_tokenbuf);
4932 /* Check for keywords */
4933 tmp = keyword(PL_tokenbuf, len);
4935 /* Is this a word before a => operator? */
4936 if (*d == '=' && d[1] == '>') {
4939 = (OP*)newSVOP(OP_CONST, 0,
4940 S_newSV_maybe_utf8(aTHX_ PL_tokenbuf, len));
4941 yylval.opval->op_private = OPpCONST_BARE;
4945 if (tmp < 0) { /* second-class keyword? */
4946 GV *ogv = NULL; /* override (winner) */
4947 GV *hgv = NULL; /* hidden (loser) */
4948 if (PL_expect != XOPERATOR && (*s != ':' || s[1] != ':')) {
4950 if ((gv = gv_fetchpvn_flags(PL_tokenbuf, len, 0, SVt_PVCV)) &&
4953 if (GvIMPORTED_CV(gv))
4955 else if (! CvMETHOD(cv))
4959 (gvp = (GV**)hv_fetch(PL_globalstash,PL_tokenbuf,len,FALSE)) &&
4960 (gv = *gvp) != (GV*)&PL_sv_undef &&
4961 GvCVu(gv) && GvIMPORTED_CV(gv))
4968 tmp = 0; /* overridden by import or by GLOBAL */
4971 && -tmp==KEY_lock /* XXX generalizable kludge */
4973 && !hv_fetchs(GvHVn(PL_incgv), "Thread.pm", FALSE))
4975 tmp = 0; /* any sub overrides "weak" keyword */
4977 else { /* no override */
4979 if (tmp == KEY_dump && ckWARN(WARN_MISC)) {
4980 Perl_warner(aTHX_ packWARN(WARN_MISC),
4981 "dump() better written as CORE::dump()");
4985 if (hgv && tmp != KEY_x && tmp != KEY_CORE
4986 && ckWARN(WARN_AMBIGUOUS)) /* never ambiguous */
4987 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
4988 "Ambiguous call resolved as CORE::%s(), %s",
4989 GvENAME(hgv), "qualify as such or use &");
4996 default: /* not a keyword */
4997 /* Trade off - by using this evil construction we can pull the
4998 variable gv into the block labelled keylookup. If not, then
4999 we have to give it function scope so that the goto from the
5000 earlier ':' case doesn't bypass the initialisation. */
5002 just_a_word_zero_gv:
5010 const char lastchar = (PL_bufptr == PL_oldoldbufptr ? 0 : PL_bufptr[-1]);
5013 SV *nextnextwhite = 0;
5017 /* Get the rest if it looks like a package qualifier */
5019 if (*s == '\'' || (*s == ':' && s[1] == ':')) {
5021 s = scan_word(s, PL_tokenbuf + len, sizeof PL_tokenbuf - len,
5024 Perl_croak(aTHX_ "Bad name after %s%s", PL_tokenbuf,
5025 *s == '\'' ? "'" : "::");
5030 if (PL_expect == XOPERATOR) {
5031 if (PL_bufptr == PL_linestart) {
5032 CopLINE_dec(PL_curcop);
5033 Perl_warner(aTHX_ packWARN(WARN_SEMICOLON), PL_warn_nosemi);
5034 CopLINE_inc(PL_curcop);
5037 no_op("Bareword",s);
5040 /* Look for a subroutine with this name in current package,
5041 unless name is "Foo::", in which case Foo is a bearword
5042 (and a package name). */
5044 if (len > 2 && !PL_madskills &&
5045 PL_tokenbuf[len - 2] == ':' && PL_tokenbuf[len - 1] == ':')
5047 if (ckWARN(WARN_BAREWORD)
5048 && ! gv_fetchpvn_flags(PL_tokenbuf, len, 0, SVt_PVHV))
5049 Perl_warner(aTHX_ packWARN(WARN_BAREWORD),
5050 "Bareword \"%s\" refers to nonexistent package",
5053 PL_tokenbuf[len] = '\0';
5059 /* Mustn't actually add anything to a symbol table.
5060 But also don't want to "initialise" any placeholder
5061 constants that might already be there into full
5062 blown PVGVs with attached PVCV. */
5063 gv = gv_fetchpvn_flags(PL_tokenbuf, len,
5064 GV_NOADD_NOINIT, SVt_PVCV);
5069 /* if we saw a global override before, get the right name */
5072 sv = newSVpvs("CORE::GLOBAL::");
5073 sv_catpv(sv,PL_tokenbuf);
5076 /* If len is 0, newSVpv does strlen(), which is correct.
5077 If len is non-zero, then it will be the true length,
5078 and so the scalar will be created correctly. */
5079 sv = newSVpv(PL_tokenbuf,len);
5082 if (PL_madskills && !thistoken) {
5083 char *start = SvPVX(PL_linestr) + realtokenstart;
5084 thistoken = newSVpv(start,s - start);
5085 realtokenstart = s - SvPVX(PL_linestr);
5089 /* Presume this is going to be a bareword of some sort. */
5092 yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
5093 yylval.opval->op_private = OPpCONST_BARE;
5094 /* UTF-8 package name? */
5095 if (UTF && !IN_BYTES &&
5096 is_utf8_string((U8*)SvPVX_const(sv), SvCUR(sv)))
5099 /* And if "Foo::", then that's what it certainly is. */
5104 /* Do the explicit type check so that we don't need to force
5105 the initialisation of the symbol table to have a real GV.
5106 Beware - gv may not really be a PVGV, cv may not really be
5107 a PVCV, (because of the space optimisations that gv_init
5108 understands) But they're true if for this symbol there is
5109 respectively a typeglob and a subroutine.
5111 cv = gv ? ((SvTYPE(gv) == SVt_PVGV)
5112 /* Real typeglob, so get the real subroutine: */
5114 /* A proxy for a subroutine in this package? */
5115 : SvOK(gv) ? (CV *) gv : NULL)
5118 /* See if it's the indirect object for a list operator. */
5120 if (PL_oldoldbufptr &&
5121 PL_oldoldbufptr < PL_bufptr &&
5122 (PL_oldoldbufptr == PL_last_lop
5123 || PL_oldoldbufptr == PL_last_uni) &&
5124 /* NO SKIPSPACE BEFORE HERE! */
5125 (PL_expect == XREF ||
5126 ((PL_opargs[PL_last_lop_op] >> OASHIFT)& 7) == OA_FILEREF))
5128 bool immediate_paren = *s == '(';
5130 /* (Now we can afford to cross potential line boundary.) */
5131 s = SKIPSPACE2(s,nextnextwhite);
5133 nextwhite = nextnextwhite; /* assume no & deception */
5136 /* Two barewords in a row may indicate method call. */
5138 if ((isIDFIRST_lazy_if(s,UTF) || *s == '$') &&
5139 (tmp = intuit_method(s, gv, cv)))
5142 /* If not a declared subroutine, it's an indirect object. */
5143 /* (But it's an indir obj regardless for sort.) */
5144 /* Also, if "_" follows a filetest operator, it's a bareword */
5147 ( !immediate_paren && (PL_last_lop_op == OP_SORT ||
5149 (PL_last_lop_op != OP_MAPSTART &&
5150 PL_last_lop_op != OP_GREPSTART))))
5151 || (PL_tokenbuf[0] == '_' && PL_tokenbuf[1] == '\0'
5152 && ((PL_opargs[PL_last_lop_op] & OA_CLASS_MASK) == OA_FILESTATOP))
5155 PL_expect = (PL_last_lop == PL_oldoldbufptr) ? XTERM : XOPERATOR;
5160 PL_expect = XOPERATOR;
5163 s = SKIPSPACE2(s,nextnextwhite);
5164 nextwhite = nextnextwhite;
5169 /* Is this a word before a => operator? */
5170 if (*s == '=' && s[1] == '>' && !pkgname) {
5172 sv_setpv(((SVOP*)yylval.opval)->op_sv, PL_tokenbuf);
5173 if (UTF && !IN_BYTES && is_utf8_string((U8*)PL_tokenbuf, len))
5174 SvUTF8_on(((SVOP*)yylval.opval)->op_sv);
5178 /* If followed by a paren, it's certainly a subroutine. */
5182 for (d = s + 1; SPACE_OR_TAB(*d); d++) ;
5183 if (*d == ')' && (sv = gv_const_sv(gv))) {
5187 char *par = SvPVX(PL_linestr) + realtokenstart;
5188 sv_catpvn(thistoken, par, s - par);
5200 nextwhite = thiswhite;
5203 start_force(curforce);
5205 NEXTVAL_NEXTTOKE.opval = yylval.opval;
5206 PL_expect = XOPERATOR;
5209 nextwhite = nextnextwhite;
5210 curmad('X', thistoken);
5211 thistoken = newSVpvn("",0);
5219 /* If followed by var or block, call it a method (unless sub) */
5221 if ((*s == '$' || *s == '{') && (!gv || !cv)) {
5222 PL_last_lop = PL_oldbufptr;
5223 PL_last_lop_op = OP_METHOD;
5227 /* If followed by a bareword, see if it looks like indir obj. */
5230 && (isIDFIRST_lazy_if(s,UTF) || *s == '$')
5231 && (tmp = intuit_method(s, gv, cv)))
5234 /* Not a method, so call it a subroutine (if defined) */
5237 if (lastchar == '-' && ckWARN_d(WARN_AMBIGUOUS))
5238 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
5239 "Ambiguous use of -%s resolved as -&%s()",
5240 PL_tokenbuf, PL_tokenbuf);
5241 /* Check for a constant sub */
5242 if ((sv = gv_const_sv(gv))) {
5244 SvREFCNT_dec(((SVOP*)yylval.opval)->op_sv);
5245 ((SVOP*)yylval.opval)->op_sv = SvREFCNT_inc_simple(sv);
5246 yylval.opval->op_private = 0;
5250 /* Resolve to GV now. */
5251 if (SvTYPE(gv) != SVt_PVGV) {
5252 gv = gv_fetchpv(PL_tokenbuf, 0, SVt_PVCV);
5253 assert (SvTYPE(gv) == SVt_PVGV);
5254 /* cv must have been some sort of placeholder, so
5255 now needs replacing with a real code reference. */
5259 op_free(yylval.opval);
5260 yylval.opval = newCVREF(0, newGVOP(OP_GV, 0, gv));
5261 yylval.opval->op_private |= OPpENTERSUB_NOPAREN;
5262 PL_last_lop = PL_oldbufptr;
5263 PL_last_lop_op = OP_ENTERSUB;
5264 /* Is there a prototype? */
5271 const char *proto = SvPV_const((SV*)cv, protolen);
5274 if (*proto == '$' && proto[1] == '\0')
5276 while (*proto == ';')
5278 if (*proto == '&' && *s == '{') {
5279 sv_setpv(PL_subname, PL_curstash ?
5280 "__ANON__" : "__ANON__::__ANON__");
5287 nextwhite = thiswhite;
5290 start_force(curforce);
5291 NEXTVAL_NEXTTOKE.opval = yylval.opval;
5294 nextwhite = nextnextwhite;
5295 curmad('X', thistoken);
5296 thistoken = newSVpvn("",0);
5303 /* Guess harder when madskills require "best effort". */
5304 if (PL_madskills && (!gv || !GvCVu(gv))) {
5305 int probable_sub = 0;
5306 if (strchr("\"'`$@%0123456789!*+{[<", *s))
5308 else if (isALPHA(*s)) {
5312 d = scan_word(d, tmpbuf, sizeof tmpbuf, TRUE, &tmplen);
5313 if (!keyword(tmpbuf,tmplen))
5316 while (d < PL_bufend && isSPACE(*d))
5318 if (*d == '=' && d[1] == '>')
5323 gv = gv_fetchpv(PL_tokenbuf, TRUE, SVt_PVCV);
5324 op_free(yylval.opval);
5325 yylval.opval = newCVREF(0, newGVOP(OP_GV, 0, gv));
5326 yylval.opval->op_private |= OPpENTERSUB_NOPAREN;
5327 PL_last_lop = PL_oldbufptr;
5328 PL_last_lop_op = OP_ENTERSUB;
5329 nextwhite = thiswhite;
5331 start_force(curforce);
5332 NEXTVAL_NEXTTOKE.opval = yylval.opval;
5334 nextwhite = nextnextwhite;
5335 curmad('X', thistoken);
5336 thistoken = newSVpvn("",0);
5341 NEXTVAL_NEXTTOKE.opval = yylval.opval;
5348 /* Call it a bare word */
5350 if (PL_hints & HINT_STRICT_SUBS)
5351 yylval.opval->op_private |= OPpCONST_STRICT;
5354 if (lastchar != '-') {
5355 if (ckWARN(WARN_RESERVED)) {
5356 for (d = PL_tokenbuf; *d && isLOWER(*d); d++) ;
5357 if (!*d && !gv_stashpv(PL_tokenbuf,FALSE))
5358 Perl_warner(aTHX_ packWARN(WARN_RESERVED), PL_warn_reserved,
5365 if ((lastchar == '*' || lastchar == '%' || lastchar == '&')
5366 && ckWARN_d(WARN_AMBIGUOUS)) {
5367 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
5368 "Operator or semicolon missing before %c%s",
5369 lastchar, PL_tokenbuf);
5370 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
5371 "Ambiguous use of %c resolved as operator %c",
5372 lastchar, lastchar);
5378 yylval.opval = (OP*)newSVOP(OP_CONST, 0,
5379 newSVpv(CopFILE(PL_curcop),0));
5383 yylval.opval = (OP*)newSVOP(OP_CONST, 0,
5384 Perl_newSVpvf(aTHX_ "%"IVdf, (IV)CopLINE(PL_curcop)));
5387 case KEY___PACKAGE__:
5388 yylval.opval = (OP*)newSVOP(OP_CONST, 0,
5390 ? newSVhek(HvNAME_HEK(PL_curstash))
5397 if (PL_rsfp && (!PL_in_eval || PL_tokenbuf[2] == 'D')) {
5398 const char *pname = "main";
5399 if (PL_tokenbuf[2] == 'D')
5400 pname = HvNAME_get(PL_curstash ? PL_curstash : PL_defstash);
5401 gv = gv_fetchpv(Perl_form(aTHX_ "%s::DATA", pname), GV_ADD,
5405 GvIOp(gv) = newIO();
5406 IoIFP(GvIOp(gv)) = PL_rsfp;
5407 #if defined(HAS_FCNTL) && defined(F_SETFD)
5409 const int fd = PerlIO_fileno(PL_rsfp);
5410 fcntl(fd,F_SETFD,fd >= 3);
5413 /* Mark this internal pseudo-handle as clean */
5414 IoFLAGS(GvIOp(gv)) |= IOf_UNTAINT;
5416 IoTYPE(GvIOp(gv)) = IoTYPE_PIPE;
5417 else if ((PerlIO*)PL_rsfp == PerlIO_stdin())
5418 IoTYPE(GvIOp(gv)) = IoTYPE_STD;
5420 IoTYPE(GvIOp(gv)) = IoTYPE_RDONLY;
5421 #if defined(WIN32) && !defined(PERL_TEXTMODE_SCRIPTS)
5422 /* if the script was opened in binmode, we need to revert
5423 * it to text mode for compatibility; but only iff it has CRs
5424 * XXX this is a questionable hack at best. */
5425 if (PL_bufend-PL_bufptr > 2
5426 && PL_bufend[-1] == '\n' && PL_bufend[-2] == '\r')
5429 if (IoTYPE(GvIOp(gv)) == IoTYPE_RDONLY) {
5430 loc = PerlIO_tell(PL_rsfp);
5431 (void)PerlIO_seek(PL_rsfp, 0L, 0);
5434 if (PerlLIO_setmode(PL_rsfp, O_TEXT) != -1) {
5436 if (PerlLIO_setmode(PerlIO_fileno(PL_rsfp), O_TEXT) != -1) {
5437 #endif /* NETWARE */
5438 #ifdef PERLIO_IS_STDIO /* really? */
5439 # if defined(__BORLANDC__)
5440 /* XXX see note in do_binmode() */
5441 ((FILE*)PL_rsfp)->flags &= ~_F_BIN;
5445 PerlIO_seek(PL_rsfp, loc, 0);
5449 #ifdef PERLIO_LAYERS
5452 PerlIO_apply_layers(aTHX_ PL_rsfp, NULL, ":utf8");
5453 else if (PL_encoding) {
5460 XPUSHs(PL_encoding);
5462 call_method("name", G_SCALAR);
5466 PerlIO_apply_layers(aTHX_ PL_rsfp, NULL,
5467 Perl_form(aTHX_ ":encoding(%"SVf")",
5476 if (realtokenstart >= 0) {
5477 char *tstart = SvPVX(PL_linestr) + realtokenstart;
5479 endwhite = newSVpvn("",0);
5480 sv_catsv(endwhite, thiswhite);
5482 sv_catpvn(endwhite, tstart, PL_bufend - tstart);
5483 realtokenstart = -1;
5485 while ((s = filter_gets(endwhite, PL_rsfp,
5486 SvCUR(endwhite))) != Nullch) ;
5500 if (PL_expect == XSTATE) {
5507 if (*s == ':' && s[1] == ':') {
5510 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
5511 if (!(tmp = keyword(PL_tokenbuf, len)))
5512 Perl_croak(aTHX_ "CORE::%s is not a keyword", PL_tokenbuf);
5515 else if (tmp == KEY_require || tmp == KEY_do)
5516 /* that's a way to remember we saw "CORE::" */
5529 LOP(OP_ACCEPT,XTERM);
5535 LOP(OP_ATAN2,XTERM);
5541 LOP(OP_BINMODE,XTERM);
5544 LOP(OP_BLESS,XTERM);
5553 /* When 'use switch' is in effect, continue has a dual
5554 life as a control operator. */
5556 if (!FEATURE_IS_ENABLED("switch"))
5559 /* We have to disambiguate the two senses of
5560 "continue". If the next token is a '{' then
5561 treat it as the start of a continue block;
5562 otherwise treat it as a control operator.
5574 (void)gv_fetchpvs("ENV", GV_ADD|GV_NOTQUAL, SVt_PVHV);
5591 if (!PL_cryptseen) {
5592 PL_cryptseen = TRUE;
5596 LOP(OP_CRYPT,XTERM);
5599 LOP(OP_CHMOD,XTERM);
5602 LOP(OP_CHOWN,XTERM);
5605 LOP(OP_CONNECT,XTERM);
5624 s = force_word(s,WORD,TRUE,TRUE,FALSE);
5625 if (orig_keyword == KEY_do) {
5634 PL_hints |= HINT_BLOCK_SCOPE;
5644 gv_fetchpvs("AnyDBM_File::ISA", GV_ADDMULTI, SVt_PVAV);
5645 LOP(OP_DBMOPEN,XTERM);
5651 s = force_word(s,WORD,TRUE,FALSE,FALSE);
5658 yylval.ival = CopLINE(PL_curcop);
5674 PL_expect = (*s == '{') ? XTERMBLOCK : XTERM;
5675 UNIBRACK(OP_ENTEREVAL);
5693 case KEY_endhostent:
5699 case KEY_endservent:
5702 case KEY_endprotoent:
5713 yylval.ival = CopLINE(PL_curcop);
5715 if (PL_expect == XSTATE && isIDFIRST_lazy_if(s,UTF)) {
5718 int soff = s - SvPVX(PL_linestr); /* for skipspace realloc */
5721 if ((PL_bufend - p) >= 3 &&
5722 strnEQ(p, "my", 2) && isSPACE(*(p + 2)))
5724 else if ((PL_bufend - p) >= 4 &&
5725 strnEQ(p, "our", 3) && isSPACE(*(p + 3)))
5728 if (isIDFIRST_lazy_if(p,UTF)) {
5729 p = scan_ident(p, PL_bufend,
5730 PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
5734 Perl_croak(aTHX_ "Missing $ on loop variable");
5736 s = SvPVX(PL_linestr) + soff;
5742 LOP(OP_FORMLINE,XTERM);
5748 LOP(OP_FCNTL,XTERM);
5754 LOP(OP_FLOCK,XTERM);
5763 LOP(OP_GREPSTART, XREF);
5766 s = force_word(s,WORD,TRUE,FALSE,FALSE);
5781 case KEY_getpriority:
5782 LOP(OP_GETPRIORITY,XTERM);
5784 case KEY_getprotobyname:
5787 case KEY_getprotobynumber:
5788 LOP(OP_GPBYNUMBER,XTERM);
5790 case KEY_getprotoent:
5802 case KEY_getpeername:
5803 UNI(OP_GETPEERNAME);
5805 case KEY_gethostbyname:
5808 case KEY_gethostbyaddr:
5809 LOP(OP_GHBYADDR,XTERM);
5811 case KEY_gethostent:
5814 case KEY_getnetbyname:
5817 case KEY_getnetbyaddr:
5818 LOP(OP_GNBYADDR,XTERM);
5823 case KEY_getservbyname:
5824 LOP(OP_GSBYNAME,XTERM);
5826 case KEY_getservbyport:
5827 LOP(OP_GSBYPORT,XTERM);
5829 case KEY_getservent:
5832 case KEY_getsockname:
5833 UNI(OP_GETSOCKNAME);
5835 case KEY_getsockopt:
5836 LOP(OP_GSOCKOPT,XTERM);
5851 yylval.ival = CopLINE(PL_curcop);
5862 yylval.ival = CopLINE(PL_curcop);
5866 LOP(OP_INDEX,XTERM);
5872 LOP(OP_IOCTL,XTERM);
5884 s = force_word(s,WORD,TRUE,FALSE,FALSE);
5916 LOP(OP_LISTEN,XTERM);
5925 s = scan_pat(s,OP_MATCH);
5926 TERM(sublex_start());
5929 LOP(OP_MAPSTART, XREF);
5932 LOP(OP_MKDIR,XTERM);
5935 LOP(OP_MSGCTL,XTERM);
5938 LOP(OP_MSGGET,XTERM);
5941 LOP(OP_MSGRCV,XTERM);
5944 LOP(OP_MSGSND,XTERM);
5950 if (isIDFIRST_lazy_if(s,UTF)) {
5954 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, TRUE, &len);
5955 if (len == 3 && strnEQ(PL_tokenbuf, "sub", 3))
5957 PL_in_my_stash = find_in_my_stash(PL_tokenbuf, len);
5958 if (!PL_in_my_stash) {
5961 sprintf(tmpbuf, "No such class %.1000s", PL_tokenbuf);
5965 if (PL_madskills) { /* just add type to declarator token */
5966 sv_catsv(thistoken, nextwhite);
5968 sv_catpvn(thistoken, start, s - start);
5976 s = force_word(s,WORD,TRUE,FALSE,FALSE);
5983 s = tokenize_use(0, s);
5987 if (*s == '(' || (s = SKIPSPACE1(s), *s == '('))
5994 if (isIDFIRST_lazy_if(s,UTF)) {
5996 for (d = s; isALNUM_lazy_if(d,UTF); d++) ;
5997 for (t=d; *t && isSPACE(*t); t++) ;
5998 if ( *t && strchr("|&*+-=!?:.", *t) && ckWARN_d(WARN_PRECEDENCE)
6000 && !(t[0] == '=' && t[1] == '>')
6002 int parms_len = (int)(d-s);
6003 Perl_warner(aTHX_ packWARN(WARN_PRECEDENCE),
6004 "Precedence problem: open %.*s should be open(%.*s)",
6005 parms_len, s, parms_len, s);
6011 yylval.ival = OP_OR;
6021 LOP(OP_OPEN_DIR,XTERM);
6024 checkcomma(s,PL_tokenbuf,"filehandle");
6028 checkcomma(s,PL_tokenbuf,"filehandle");
6047 s = force_word(s,WORD,FALSE,TRUE,FALSE);
6051 LOP(OP_PIPE_OP,XTERM);
6054 s = scan_str(s,!!PL_madskills,FALSE);
6056 missingterm((char*)0);
6057 yylval.ival = OP_CONST;
6058 TERM(sublex_start());
6064 s = scan_str(s,!!PL_madskills,FALSE);
6066 missingterm((char*)0);
6067 PL_expect = XOPERATOR;
6069 if (SvCUR(PL_lex_stuff)) {
6072 d = SvPV_force(PL_lex_stuff, len);
6075 for (; isSPACE(*d) && len; --len, ++d) ;
6078 if (!warned && ckWARN(WARN_QW)) {
6079 for (; !isSPACE(*d) && len; --len, ++d) {
6081 Perl_warner(aTHX_ packWARN(WARN_QW),
6082 "Possible attempt to separate words with commas");
6085 else if (*d == '#') {
6086 Perl_warner(aTHX_ packWARN(WARN_QW),
6087 "Possible attempt to put comments in qw() list");
6093 for (; !isSPACE(*d) && len; --len, ++d) ;
6095 sv = newSVpvn(b, d-b);
6096 if (DO_UTF8(PL_lex_stuff))
6098 words = append_elem(OP_LIST, words,
6099 newSVOP(OP_CONST, 0, tokeq(sv)));
6103 start_force(curforce);
6104 NEXTVAL_NEXTTOKE.opval = words;
6109 SvREFCNT_dec(PL_lex_stuff);
6110 PL_lex_stuff = NULL;
6116 s = scan_str(s,!!PL_madskills,FALSE);
6118 missingterm((char*)0);
6119 yylval.ival = OP_STRINGIFY;
6120 if (SvIVX(PL_lex_stuff) == '\'')
6121 SvIV_set(PL_lex_stuff, 0); /* qq'$foo' should intepolate */
6122 TERM(sublex_start());
6125 s = scan_pat(s,OP_QR);
6126 TERM(sublex_start());
6129 s = scan_str(s,!!PL_madskills,FALSE);
6131 missingterm((char*)0);
6132 yylval.ival = OP_BACKTICK;
6134 TERM(sublex_start());
6142 s = force_version(s, FALSE);
6144 else if (*s != 'v' || !isDIGIT(s[1])
6145 || (s = force_version(s, TRUE), *s == 'v'))
6147 *PL_tokenbuf = '\0';
6148 s = force_word(s,WORD,TRUE,TRUE,FALSE);
6149 if (isIDFIRST_lazy_if(PL_tokenbuf,UTF))
6150 gv_stashpvn(PL_tokenbuf, strlen(PL_tokenbuf), TRUE);
6152 yyerror("<> should be quotes");
6154 if (orig_keyword == KEY_require) {
6162 PL_last_uni = PL_oldbufptr;
6163 PL_last_lop_op = OP_REQUIRE;
6165 return REPORT( (int)REQUIRE );
6171 s = force_word(s,WORD,TRUE,FALSE,FALSE);
6175 LOP(OP_RENAME,XTERM);
6184 LOP(OP_RINDEX,XTERM);
6194 UNIDOR(OP_READLINE);
6207 LOP(OP_REVERSE,XTERM);
6210 UNIDOR(OP_READLINK);
6218 TERM(sublex_start());
6220 TOKEN(1); /* force error */
6223 checkcomma(s,PL_tokenbuf,"filehandle");
6233 LOP(OP_SELECT,XTERM);
6239 LOP(OP_SEMCTL,XTERM);
6242 LOP(OP_SEMGET,XTERM);
6245 LOP(OP_SEMOP,XTERM);
6251 LOP(OP_SETPGRP,XTERM);
6253 case KEY_setpriority:
6254 LOP(OP_SETPRIORITY,XTERM);
6256 case KEY_sethostent:
6262 case KEY_setservent:
6265 case KEY_setprotoent:
6275 LOP(OP_SEEKDIR,XTERM);
6277 case KEY_setsockopt:
6278 LOP(OP_SSOCKOPT,XTERM);
6284 LOP(OP_SHMCTL,XTERM);
6287 LOP(OP_SHMGET,XTERM);
6290 LOP(OP_SHMREAD,XTERM);
6293 LOP(OP_SHMWRITE,XTERM);
6296 LOP(OP_SHUTDOWN,XTERM);
6305 LOP(OP_SOCKET,XTERM);
6307 case KEY_socketpair:
6308 LOP(OP_SOCKPAIR,XTERM);
6311 checkcomma(s,PL_tokenbuf,"subroutine name");
6313 if (*s == ';' || *s == ')') /* probably a close */
6314 Perl_croak(aTHX_ "sort is now a reserved word");
6316 s = force_word(s,WORD,TRUE,TRUE,FALSE);
6320 LOP(OP_SPLIT,XTERM);
6323 LOP(OP_SPRINTF,XTERM);
6326 LOP(OP_SPLICE,XTERM);
6341 LOP(OP_SUBSTR,XTERM);
6347 char tmpbuf[sizeof PL_tokenbuf];
6348 SSize_t tboffset = 0;
6349 expectation attrful;
6350 bool have_name, have_proto, bad_proto;
6351 const int key = tmp;
6356 char *tstart = SvPVX(PL_linestr) + realtokenstart;
6357 SV *subtoken = newSVpvn(tstart, s - tstart);
6361 s = SKIPSPACE2(s,tmpwhite);
6366 if (isIDFIRST_lazy_if(s,UTF) || *s == '\'' ||
6367 (*s == ':' && s[1] == ':'))
6374 attrful = XATTRBLOCK;
6375 /* remember buffer pos'n for later force_word */
6376 tboffset = s - PL_oldbufptr;
6377 d = scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
6380 nametoke = newSVpvn(s, d - s);
6382 if (strchr(tmpbuf, ':'))
6383 sv_setpv(PL_subname, tmpbuf);
6385 sv_setsv(PL_subname,PL_curstname);
6386 sv_catpvs(PL_subname,"::");
6387 sv_catpvn(PL_subname,tmpbuf,len);
6394 CURMAD('X', nametoke);
6395 CURMAD('_', tmpwhite);
6396 (void) force_word(PL_oldbufptr + tboffset, WORD,
6399 s = SKIPSPACE2(d,tmpwhite);
6406 Perl_croak(aTHX_ "Missing name in \"my sub\"");
6407 PL_expect = XTERMBLOCK;
6408 attrful = XATTRTERM;
6409 sv_setpvn(PL_subname,"?",1);
6413 if (key == KEY_format) {
6415 PL_lex_formbrack = PL_lex_brackets + 1;
6417 thistoken = subtoken;
6421 (void) force_word(PL_oldbufptr + tboffset, WORD,
6427 /* Look for a prototype */
6431 s = scan_str(s,!!PL_madskills,FALSE);
6433 Perl_croak(aTHX_ "Prototype not terminated");
6434 /* strip spaces and check for bad characters */
6435 d = SvPVX(PL_lex_stuff);
6438 for (p = d; *p; ++p) {
6441 if (!strchr("$@%*;[]&\\", *p))
6446 if (bad_proto && ckWARN(WARN_SYNTAX))
6447 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
6448 "Illegal character in prototype for %"SVf" : %s",
6450 SvCUR_set(PL_lex_stuff, tmp);
6455 CURMAD('q', thisopen);
6456 CURMAD('_', tmpwhite);
6457 CURMAD('=', thisstuff);
6458 CURMAD('Q', thisclose);
6459 NEXTVAL_NEXTTOKE.opval =
6460 (OP*)newSVOP(OP_CONST, 0, PL_lex_stuff);
6461 PL_lex_stuff = Nullsv;
6464 s = SKIPSPACE2(s,tmpwhite);
6472 if (*s == ':' && s[1] != ':')
6473 PL_expect = attrful;
6474 else if (*s != '{' && key == KEY_sub) {
6476 Perl_croak(aTHX_ "Illegal declaration of anonymous subroutine");
6478 Perl_croak(aTHX_ "Illegal declaration of subroutine %"SVf, PL_subname);
6485 curmad('^', newSVpvn("",0));
6486 CURMAD('_', tmpwhite);
6490 thistoken = subtoken;
6493 NEXTVAL_NEXTTOKE.opval =
6494 (OP*)newSVOP(OP_CONST, 0, PL_lex_stuff);
6495 PL_lex_stuff = NULL;
6500 sv_setpv(PL_subname,
6501 PL_curstash ? "__ANON__" : "__ANON__::__ANON__");
6505 (void) force_word(PL_oldbufptr + tboffset, WORD,
6515 LOP(OP_SYSTEM,XREF);
6518 LOP(OP_SYMLINK,XTERM);
6521 LOP(OP_SYSCALL,XTERM);
6524 LOP(OP_SYSOPEN,XTERM);
6527 LOP(OP_SYSSEEK,XTERM);
6530 LOP(OP_SYSREAD,XTERM);
6533 LOP(OP_SYSWRITE,XTERM);
6537 TERM(sublex_start());
6558 LOP(OP_TRUNCATE,XTERM);
6570 yylval.ival = CopLINE(PL_curcop);
6574 yylval.ival = CopLINE(PL_curcop);
6578 LOP(OP_UNLINK,XTERM);
6584 LOP(OP_UNPACK,XTERM);
6587 LOP(OP_UTIME,XTERM);
6593 LOP(OP_UNSHIFT,XTERM);
6596 s = tokenize_use(1, s);
6606 yylval.ival = CopLINE(PL_curcop);
6610 yylval.ival = CopLINE(PL_curcop);
6614 PL_hints |= HINT_BLOCK_SCOPE;
6621 LOP(OP_WAITPID,XTERM);
6630 ctl_l[0] = toCTRL('L');
6632 gv_fetchpvn_flags(ctl_l, 1, GV_ADD|GV_NOTQUAL, SVt_PV);
6635 /* Make sure $^L is defined */
6636 gv_fetchpvs("\f", GV_ADD|GV_NOTQUAL, SVt_PV);
6641 if (PL_expect == XOPERATOR)
6647 yylval.ival = OP_XOR;
6652 TERM(sublex_start());
6657 #pragma segment Main
6661 S_pending_ident(pTHX)
6665 register I32 tmp = 0;
6666 /* pit holds the identifier we read and pending_ident is reset */
6667 char pit = PL_pending_ident;
6668 PL_pending_ident = 0;
6670 /* realtokenstart = realtokenend = PL_bufptr - SvPVX(PL_linestr); */
6671 DEBUG_T({ PerlIO_printf(Perl_debug_log,
6672 "### Pending identifier '%s'\n", PL_tokenbuf); });
6674 /* if we're in a my(), we can't allow dynamics here.
6675 $foo'bar has already been turned into $foo::bar, so
6676 just check for colons.
6678 if it's a legal name, the OP is a PADANY.
6681 if (PL_in_my == KEY_our) { /* "our" is merely analogous to "my" */
6682 if (strchr(PL_tokenbuf,':'))
6683 yyerror(Perl_form(aTHX_ "No package name allowed for "
6684 "variable %s in \"our\"",
6686 tmp = allocmy(PL_tokenbuf);
6689 if (strchr(PL_tokenbuf,':'))
6690 yyerror(Perl_form(aTHX_ PL_no_myglob,PL_tokenbuf));
6692 yylval.opval = newOP(OP_PADANY, 0);
6693 yylval.opval->op_targ = allocmy(PL_tokenbuf);
6699 build the ops for accesses to a my() variable.
6701 Deny my($a) or my($b) in a sort block, *if* $a or $b is
6702 then used in a comparison. This catches most, but not
6703 all cases. For instance, it catches
6704 sort { my($a); $a <=> $b }
6706 sort { my($a); $a < $b ? -1 : $a == $b ? 0 : 1; }
6707 (although why you'd do that is anyone's guess).
6710 if (!strchr(PL_tokenbuf,':')) {
6712 tmp = pad_findmy(PL_tokenbuf);
6713 if (tmp != NOT_IN_PAD) {
6714 /* might be an "our" variable" */
6715 if (PAD_COMPNAME_FLAGS_isOUR(tmp)) {
6716 /* build ops for a bareword */
6717 HV * const stash = PAD_COMPNAME_OURSTASH(tmp);
6718 HEK * const stashname = HvNAME_HEK(stash);
6719 SV * const sym = newSVhek(stashname);
6720 sv_catpvs(sym, "::");
6721 sv_catpv(sym, PL_tokenbuf+1);
6722 yylval.opval = (OP*)newSVOP(OP_CONST, 0, sym);
6723 yylval.opval->op_private = OPpCONST_ENTERED;
6726 ? (GV_ADDMULTI | GV_ADDINEVAL)
6729 ((PL_tokenbuf[0] == '$') ? SVt_PV
6730 : (PL_tokenbuf[0] == '@') ? SVt_PVAV
6735 /* if it's a sort block and they're naming $a or $b */
6736 if (PL_last_lop_op == OP_SORT &&
6737 PL_tokenbuf[0] == '$' &&
6738 (PL_tokenbuf[1] == 'a' || PL_tokenbuf[1] == 'b')
6741 for (d = PL_in_eval ? PL_oldoldbufptr : PL_linestart;
6742 d < PL_bufend && *d != '\n';
6745 if (strnEQ(d,"<=>",3) || strnEQ(d,"cmp",3)) {
6746 Perl_croak(aTHX_ "Can't use \"my %s\" in sort comparison",
6752 yylval.opval = newOP(OP_PADANY, 0);
6753 yylval.opval->op_targ = tmp;
6759 Whine if they've said @foo in a doublequoted string,
6760 and @foo isn't a variable we can find in the symbol
6763 if (pit == '@' && PL_lex_state != LEX_NORMAL && !PL_lex_brackets) {
6764 GV *gv = gv_fetchpv(PL_tokenbuf+1, 0, SVt_PVAV);
6765 if ((!gv || ((PL_tokenbuf[0] == '@') ? !GvAV(gv) : !GvHV(gv)))
6766 && ckWARN(WARN_AMBIGUOUS))
6768 /* Downgraded from fatal to warning 20000522 mjd */
6769 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
6770 "Possible unintended interpolation of %s in string",
6775 /* build ops for a bareword */
6776 yylval.opval = (OP*)newSVOP(OP_CONST, 0, newSVpv(PL_tokenbuf+1, 0));
6777 yylval.opval->op_private = OPpCONST_ENTERED;
6780 /* If the identifier refers to a stash, don't autovivify it.
6781 * Change 24660 had the side effect of causing symbol table
6782 * hashes to always be defined, even if they were freshly
6783 * created and the only reference in the entire program was
6784 * the single statement with the defined %foo::bar:: test.
6785 * It appears that all code in the wild doing this actually
6786 * wants to know whether sub-packages have been loaded, so
6787 * by avoiding auto-vivifying symbol tables, we ensure that
6788 * defined %foo::bar:: continues to be false, and the existing
6789 * tests still give the expected answers, even though what
6790 * they're actually testing has now changed subtly.
6792 (*PL_tokenbuf == '%' && *(d = PL_tokenbuf + strlen(PL_tokenbuf) - 1) == ':' && d[-1] == ':'
6794 : PL_in_eval ? (GV_ADDMULTI | GV_ADDINEVAL) : GV_ADD),
6795 ((PL_tokenbuf[0] == '$') ? SVt_PV
6796 : (PL_tokenbuf[0] == '@') ? SVt_PVAV
6802 * The following code was generated by perl_keyword.pl.
6806 Perl_keyword (pTHX_ const char *name, I32 len)
6811 case 1: /* 5 tokens of length 1 */
6843 case 2: /* 18 tokens of length 2 */
6989 case 3: /* 29 tokens of length 3 */
6993 if (name[1] == 'N' &&
7056 if (name[1] == 'i' &&
7078 return (FEATURE_IS_ENABLED("err") ? -KEY_err : 0);
7096 if (name[1] == 'o' &&
7105 if (name[1] == 'e' &&
7114 if (name[1] == 'n' &&
7123 if (name[1] == 'o' &&
7132 if (name[1] == 'a' &&
7141 if (name[1] == 'o' &&
7203 if (name[1] == 'e' &&
7217 return (FEATURE_IS_ENABLED("say") ? -KEY_say : 0);
7243 if (name[1] == 'i' &&
7252 if (name[1] == 's' &&
7261 if (name[1] == 'e' &&
7270 if (name[1] == 'o' &&
7282 case 4: /* 41 tokens of length 4 */
7286 if (name[1] == 'O' &&
7296 if (name[1] == 'N' &&
7306 if (name[1] == 'i' &&
7316 if (name[1] == 'h' &&
7326 if (name[1] == 'u' &&
7339 if (name[2] == 'c' &&
7348 if (name[2] == 's' &&
7357 if (name[2] == 'a' &&
7393 if (name[1] == 'o' &&
7406 if (name[2] == 't' &&
7415 if (name[2] == 'o' &&
7424 if (name[2] == 't' &&
7433 if (name[2] == 'e' &&
7446 if (name[1] == 'o' &&
7459 if (name[2] == 'y' &&
7468 if (name[2] == 'l' &&
7484 if (name[2] == 's' &&
7493 if (name[2] == 'n' &&
7502 if (name[2] == 'c' &&
7515 if (name[1] == 'e' &&
7525 if (name[1] == 'p' &&
7538 if (name[2] == 'c' &&
7547 if (name[2] == 'p' &&
7556 if (name[2] == 's' &&
7572 if (name[2] == 'n' &&
7642 if (name[2] == 'r' &&
7651 if (name[2] == 'r' &&
7660 if (name[2] == 'a' &&
7676 if (name[2] == 'l' &&
7738 if (name[2] == 'e' &&
7741 return (FEATURE_IS_ENABLED("switch") ? KEY_when : 0);
7754 case 5: /* 38 tokens of length 5 */
7758 if (name[1] == 'E' &&
7769 if (name[1] == 'H' &&
7783 if (name[2] == 'a' &&
7793 if (name[2] == 'a' &&
7810 if (name[2] == 'e' &&
7820 if (name[2] == 'e' &&
7824 return (FEATURE_IS_ENABLED("switch") ? -KEY_break : 0);
7840 if (name[3] == 'i' &&
7849 if (name[3] == 'o' &&
7885 if (name[2] == 'o' &&
7895 if (name[2] == 'y' &&
7909 if (name[1] == 'l' &&
7923 if (name[2] == 'n' &&
7933 if (name[2] == 'o' &&
7947 if (name[1] == 'i' &&
7952 return (FEATURE_IS_ENABLED("switch") ? KEY_given : 0);
7961 if (name[2] == 'd' &&
7971 if (name[2] == 'c' &&
7988 if (name[2] == 'c' &&
7998 if (name[2] == 't' &&
8012 if (name[1] == 'k' &&
8023 if (name[1] == 'r' &&
8037 if (name[2] == 's' &&
8047 if (name[2] == 'd' &&
8064 if (name[2] == 'm' &&
8074 if (name[2] == 'i' &&
8084 if (name[2] == 'e' &&
8094 if (name[2] == 'l' &&
8104 if (name[2] == 'a' &&
8114 if (name[2] == 'u' &&
8128 if (name[1] == 'i' &&
8142 if (name[2] == 'a' &&
8155 if (name[3] == 'e' &&
8190 if (name[2] == 'i' &&
8207 if (name[2] == 'i' &&
8217 if (name[2] == 'i' &&
8234 case 6: /* 33 tokens of length 6 */
8238 if (name[1] == 'c' &&
8253 if (name[2] == 'l' &&
8264 if (name[2] == 'r' &&
8279 if (name[1] == 'e' &&
8294 if (name[2] == 's' &&
8299 if(ckWARN_d(WARN_SYNTAX))
8300 Perl_warner(aTHX_ packWARN(WARN_SYNTAX), "elseif should be elsif");
8306 if (name[2] == 'i' &&
8324 if (name[2] == 'l' &&
8335 if (name[2] == 'r' &&
8350 if (name[1] == 'm' &&
8365 if (name[2] == 'n' &&
8376 if (name[2] == 's' &&
8391 if (name[1] == 's' &&
8397 if (name[4] == 't' &&
8406 if (name[4] == 'e' &&
8415 if (name[4] == 'c' &&
8424 if (name[4] == 'n' &&
8440 if (name[1] == 'r' &&
8458 if (name[3] == 'a' &&
8468 if (name[3] == 'u' &&
8482 if (name[2] == 'n' &&
8500 if (name[2] == 'a' &&
8514 if (name[3] == 'e' &&
8527 if (name[4] == 't' &&
8536 if (name[4] == 'e' &&
8558 if (name[4] == 't' &&
8567 if (name[4] == 'e' &&
8583 if (name[2] == 'c' &&
8594 if (name[2] == 'l' &&
8605 if (name[2] == 'b' &&
8616 if (name[2] == 's' &&
8639 if (name[4] == 's' &&
8648 if (name[4] == 'n' &&
8661 if (name[3] == 'a' &&
8678 if (name[1] == 'a' &&
8693 case 7: /* 29 tokens of length 7 */
8697 if (name[1] == 'E' &&
8710 if (name[1] == '_' &&
8723 if (name[1] == 'i' &&
8730 return -KEY_binmode;
8736 if (name[1] == 'o' &&
8743 return -KEY_connect;
8752 if (name[2] == 'm' &&
8758 return -KEY_dbmopen;
8769 if (name[4] == 'u' &&
8773 return (FEATURE_IS_ENABLED("switch") ? KEY_default : 0);
8779 if (name[4] == 'n' &&
8800 if (name[1] == 'o' &&
8813 if (name[1] == 'e' &&
8820 if (name[5] == 'r' &&
8823 return -KEY_getpgrp;
8829 if (name[5] == 'i' &&
8832 return -KEY_getppid;
8845 if (name[1] == 'c' &&
8852 return -KEY_lcfirst;
8858 if (name[1] == 'p' &&
8865 return -KEY_opendir;
8871 if (name[1] == 'a' &&
8889 if (name[3] == 'd' &&
8894 return -KEY_readdir;
8900 if (name[3] == 'u' &&
8911 if (name[3] == 'e' &&
8916 return -KEY_reverse;
8935 if (name[3] == 'k' &&
8940 return -KEY_seekdir;
8946 if (name[3] == 'p' &&
8951 return -KEY_setpgrp;
8961 if (name[2] == 'm' &&
8967 return -KEY_shmread;
8973 if (name[2] == 'r' &&
8979 return -KEY_sprintf;
8988 if (name[3] == 'l' &&
8993 return -KEY_symlink;
9002 if (name[4] == 'a' &&
9006 return -KEY_syscall;
9012 if (name[4] == 'p' &&
9016 return -KEY_sysopen;
9022 if (name[4] == 'e' &&
9026 return -KEY_sysread;
9032 if (name[4] == 'e' &&
9036 return -KEY_sysseek;
9054 if (name[1] == 'e' &&
9061 return -KEY_telldir;
9070 if (name[2] == 'f' &&
9076 return -KEY_ucfirst;
9082 if (name[2] == 's' &&
9088 return -KEY_unshift;
9098 if (name[1] == 'a' &&
9105 return -KEY_waitpid;
9114 case 8: /* 26 tokens of length 8 */
9118 if (name[1] == 'U' &&
9126 return KEY_AUTOLOAD;
9137 if (name[3] == 'A' &&
9143 return KEY___DATA__;
9149 if (name[3] == 'I' &&
9155 return -KEY___FILE__;
9161 if (name[3] == 'I' &&
9167 return -KEY___LINE__;
9183 if (name[2] == 'o' &&
9190 return -KEY_closedir;
9196 if (name[2] == 'n' &&
9203 return -KEY_continue;
9213 if (name[1] == 'b' &&
9221 return -KEY_dbmclose;
9227 if (name[1] == 'n' &&
9233 if (name[4] == 'r' &&
9238 return -KEY_endgrent;
9244 if (name[4] == 'w' &&
9249 return -KEY_endpwent;
9262 if (name[1] == 'o' &&
9270 return -KEY_formline;
9276 if (name[1] == 'e' &&
9287 if (name[6] == 'n' &&
9290 return -KEY_getgrent;
9296 if (name[6] == 'i' &&
9299 return -KEY_getgrgid;
9305 if (name[6] == 'a' &&
9308 return -KEY_getgrnam;
9321 if (name[4] == 'o' &&
9326 return -KEY_getlogin;
9337 if (name[6] == 'n' &&
9340 return -KEY_getpwent;
9346 if (name[6] == 'a' &&
9349 return -KEY_getpwnam;
9355 if (name[6] == 'i' &&
9358 return -KEY_getpwuid;
9378 if (name[1] == 'e' &&
9385 if (name[5] == 'i' &&
9392 return -KEY_readline;
9397 return -KEY_readlink;
9408 if (name[5] == 'i' &&
9412 return -KEY_readpipe;
9433 if (name[4] == 'r' &&
9438 return -KEY_setgrent;
9444 if (name[4] == 'w' &&
9449 return -KEY_setpwent;
9465 if (name[3] == 'w' &&
9471 return -KEY_shmwrite;
9477 if (name[3] == 't' &&
9483 return -KEY_shutdown;
9493 if (name[2] == 's' &&
9500 return -KEY_syswrite;
9510 if (name[1] == 'r' &&
9518 return -KEY_truncate;
9527 case 9: /* 8 tokens of length 9 */
9531 if (name[1] == 'n' &&
9540 return -KEY_endnetent;
9546 if (name[1] == 'e' &&
9555 return -KEY_getnetent;
9561 if (name[1] == 'o' &&
9570 return -KEY_localtime;
9576 if (name[1] == 'r' &&
9585 return KEY_prototype;
9591 if (name[1] == 'u' &&
9600 return -KEY_quotemeta;
9606 if (name[1] == 'e' &&
9615 return -KEY_rewinddir;
9621 if (name[1] == 'e' &&
9630 return -KEY_setnetent;
9636 if (name[1] == 'a' &&
9645 return -KEY_wantarray;
9654 case 10: /* 9 tokens of length 10 */
9658 if (name[1] == 'n' &&
9664 if (name[4] == 'o' &&
9671 return -KEY_endhostent;
9677 if (name[4] == 'e' &&
9684 return -KEY_endservent;
9697 if (name[1] == 'e' &&
9703 if (name[4] == 'o' &&
9710 return -KEY_gethostent;
9719 if (name[5] == 'r' &&
9725 return -KEY_getservent;
9731 if (name[5] == 'c' &&
9737 return -KEY_getsockopt;
9762 if (name[4] == 'o' &&
9769 return -KEY_sethostent;
9778 if (name[5] == 'r' &&
9784 return -KEY_setservent;
9790 if (name[5] == 'c' &&
9796 return -KEY_setsockopt;
9813 if (name[2] == 'c' &&
9822 return -KEY_socketpair;
9835 case 11: /* 8 tokens of length 11 */
9839 if (name[1] == '_' &&
9850 return -KEY___PACKAGE__;
9856 if (name[1] == 'n' &&
9867 return -KEY_endprotoent;
9873 if (name[1] == 'e' &&
9882 if (name[5] == 'e' &&
9889 return -KEY_getpeername;
9898 if (name[6] == 'o' &&
9904 return -KEY_getpriority;
9910 if (name[6] == 't' &&
9916 return -KEY_getprotoent;
9930 if (name[4] == 'o' &&
9938 return -KEY_getsockname;
9951 if (name[1] == 'e' &&
9959 if (name[6] == 'o' &&
9965 return -KEY_setpriority;
9971 if (name[6] == 't' &&
9977 return -KEY_setprotoent;
9993 case 12: /* 2 tokens of length 12 */
9994 if (name[0] == 'g' &&
10006 if (name[9] == 'd' &&
10009 { /* getnetbyaddr */
10010 return -KEY_getnetbyaddr;
10016 if (name[9] == 'a' &&
10019 { /* getnetbyname */
10020 return -KEY_getnetbyname;
10032 case 13: /* 4 tokens of length 13 */
10033 if (name[0] == 'g' &&
10040 if (name[4] == 'o' &&
10049 if (name[10] == 'd' &&
10052 { /* gethostbyaddr */
10053 return -KEY_gethostbyaddr;
10059 if (name[10] == 'a' &&
10062 { /* gethostbyname */
10063 return -KEY_gethostbyname;
10076 if (name[4] == 'e' &&
10085 if (name[10] == 'a' &&
10088 { /* getservbyname */
10089 return -KEY_getservbyname;
10095 if (name[10] == 'o' &&
10098 { /* getservbyport */
10099 return -KEY_getservbyport;
10118 case 14: /* 1 tokens of length 14 */
10119 if (name[0] == 'g' &&
10133 { /* getprotobyname */
10134 return -KEY_getprotobyname;
10139 case 16: /* 1 tokens of length 16 */
10140 if (name[0] == 'g' &&
10156 { /* getprotobynumber */
10157 return -KEY_getprotobynumber;
10171 S_checkcomma(pTHX_ register char *s, const char *name, const char *what)
10176 if (*s == ' ' && s[1] == '(') { /* XXX gotta be a better way */
10177 if (ckWARN(WARN_SYNTAX)) {
10179 for (w = s+2; *w && level; w++) {
10182 else if (*w == ')')
10186 for (; *w && isSPACE(*w); w++) ;
10187 if (!*w || !strchr(";|})]oaiuw!=", *w)) /* an advisory hack only... */
10188 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
10189 "%s (...) interpreted as function",name);
10192 while (s < PL_bufend && isSPACE(*s))
10196 while (s < PL_bufend && isSPACE(*s))
10198 if (isIDFIRST_lazy_if(s,UTF)) {
10200 while (isALNUM_lazy_if(s,UTF))
10202 while (s < PL_bufend && isSPACE(*s))
10206 *s = '\0'; /* XXX If we didn't do this, we could const a lot of toke.c */
10207 kw = keyword(w, s - w) || get_cv(w, FALSE) != 0;
10211 Perl_croak(aTHX_ "No comma allowed after %s", what);
10216 /* Either returns sv, or mortalizes sv and returns a new SV*.
10217 Best used as sv=new_constant(..., sv, ...).
10218 If s, pv are NULL, calls subroutine with one argument,
10219 and type is used with error messages only. */
10222 S_new_constant(pTHX_ const char *s, STRLEN len, const char *key, SV *sv, SV *pv,
10226 HV * const table = GvHV(PL_hintgv); /* ^H */
10230 const char *why1 = "", *why2 = "", *why3 = "";
10232 if (!table || !(PL_hints & HINT_LOCALIZE_HH)) {
10235 why2 = strEQ(key,"charnames")
10236 ? "(possibly a missing \"use charnames ...\")"
10238 msg = Perl_newSVpvf(aTHX_ "Constant(%s) unknown: %s",
10239 (type ? type: "undef"), why2);
10241 /* This is convoluted and evil ("goto considered harmful")
10242 * but I do not understand the intricacies of all the different
10243 * failure modes of %^H in here. The goal here is to make
10244 * the most probable error message user-friendly. --jhi */
10249 msg = Perl_newSVpvf(aTHX_ "Constant(%s): %s%s%s",
10250 (type ? type: "undef"), why1, why2, why3);
10252 yyerror(SvPVX_const(msg));
10256 cvp = hv_fetch(table, key, strlen(key), FALSE);
10257 if (!cvp || !SvOK(*cvp)) {
10260 why3 = "} is not defined";
10263 sv_2mortal(sv); /* Parent created it permanently */
10266 pv = sv_2mortal(newSVpvn(s, len));
10268 typesv = sv_2mortal(newSVpv(type, 0));
10270 typesv = &PL_sv_undef;
10272 PUSHSTACKi(PERLSI_OVERLOAD);
10284 call_sv(cv, G_SCALAR | ( PL_in_eval ? 0 : G_EVAL));
10288 /* Check the eval first */
10289 if (!PL_in_eval && SvTRUE(ERRSV)) {
10290 sv_catpvs(ERRSV, "Propagated");
10291 yyerror(SvPV_nolen_const(ERRSV)); /* Duplicates the message inside eval */
10293 res = SvREFCNT_inc_simple(sv);
10297 SvREFCNT_inc_simple_void(res);
10306 why1 = "Call to &{$^H{";
10308 why3 = "}} did not return a defined value";
10316 /* Returns a NUL terminated string, with the length of the string written to
10320 S_scan_word(pTHX_ register char *s, char *dest, STRLEN destlen, int allow_package, STRLEN *slp)
10323 register char *d = dest;
10324 register char * const e = d + destlen - 3; /* two-character token, ending NUL */
10327 Perl_croak(aTHX_ ident_too_long);
10328 if (isALNUM(*s)) /* UTF handled below */
10330 else if (*s == '\'' && allow_package && isIDFIRST_lazy_if(s+1,UTF)) {
10335 else if (*s == ':' && s[1] == ':' && allow_package && s[2] != '$') {
10339 else if (UTF && UTF8_IS_START(*s) && isALNUM_utf8((U8*)s)) {
10340 char *t = s + UTF8SKIP(s);
10341 while (UTF8_IS_CONTINUED(*t) && is_utf8_mark((U8*)t))
10343 if (d + (t - s) > e)
10344 Perl_croak(aTHX_ ident_too_long);
10345 Copy(s, d, t - s, char);
10358 S_scan_ident(pTHX_ register char *s, register const char *send, char *dest, STRLEN destlen, I32 ck_uni)
10361 char *bracket = NULL;
10363 register char *d = dest;
10364 register char * const e = d + destlen + 3; /* two-character token, ending NUL */
10369 while (isDIGIT(*s)) {
10371 Perl_croak(aTHX_ ident_too_long);
10378 Perl_croak(aTHX_ ident_too_long);
10379 if (isALNUM(*s)) /* UTF handled below */
10381 else if (*s == '\'' && isIDFIRST_lazy_if(s+1,UTF)) {
10386 else if (*s == ':' && s[1] == ':') {
10390 else if (UTF && UTF8_IS_START(*s) && isALNUM_utf8((U8*)s)) {
10391 char *t = s + UTF8SKIP(s);
10392 while (UTF8_IS_CONTINUED(*t) && is_utf8_mark((U8*)t))
10394 if (d + (t - s) > e)
10395 Perl_croak(aTHX_ ident_too_long);
10396 Copy(s, d, t - s, char);
10407 if (PL_lex_state != LEX_NORMAL)
10408 PL_lex_state = LEX_INTERPENDMAYBE;
10411 if (*s == '$' && s[1] &&
10412 (isALNUM_lazy_if(s+1,UTF) || s[1] == '$' || s[1] == '{' || strnEQ(s+1,"::",2)) )
10425 if (*d == '^' && *s && isCONTROLVAR(*s)) {
10430 if (isSPACE(s[-1])) {
10432 const char ch = *s++;
10433 if (!SPACE_OR_TAB(ch)) {
10439 if (isIDFIRST_lazy_if(d,UTF)) {
10443 while ((end < send && isALNUM_lazy_if(end,UTF)) || *end == ':') {
10444 end += UTF8SKIP(end);
10445 while (end < send && UTF8_IS_CONTINUED(*end) && is_utf8_mark((U8*)end))
10446 end += UTF8SKIP(end);
10448 Copy(s, d, end - s, char);
10453 while ((isALNUM(*s) || *s == ':') && d < e)
10456 Perl_croak(aTHX_ ident_too_long);
10459 while (s < send && SPACE_OR_TAB(*s)) s++;
10460 if ((*s == '[' || (*s == '{' && strNE(dest, "sub")))) {
10461 if (ckWARN(WARN_AMBIGUOUS) && keyword(dest, d - dest)) {
10462 const char *brack = *s == '[' ? "[...]" : "{...}";
10463 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
10464 "Ambiguous use of %c{%s%s} resolved to %c%s%s",
10465 funny, dest, brack, funny, dest, brack);
10468 PL_lex_brackstack[PL_lex_brackets++] = (char)(XOPERATOR | XFAKEBRACK);
10472 /* Handle extended ${^Foo} variables
10473 * 1999-02-27 mjd-perl-patch@plover.com */
10474 else if (!isALNUM(*d) && !isPRINT(*d) /* isCTRL(d) */
10478 while (isALNUM(*s) && d < e) {
10482 Perl_croak(aTHX_ ident_too_long);
10487 if (PL_lex_state == LEX_INTERPNORMAL && !PL_lex_brackets) {
10488 PL_lex_state = LEX_INTERPEND;
10493 if (PL_lex_state == LEX_NORMAL) {
10494 if (ckWARN(WARN_AMBIGUOUS) &&
10495 (keyword(dest, d - dest) || get_cv(dest, FALSE)))
10497 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
10498 "Ambiguous use of %c{%s} resolved to %c%s",
10499 funny, dest, funny, dest);
10504 s = bracket; /* let the parser handle it */
10508 else if (PL_lex_state == LEX_INTERPNORMAL && !PL_lex_brackets && !intuit_more(s))
10509 PL_lex_state = LEX_INTERPEND;
10514 Perl_pmflag(pTHX_ U32* pmfl, int ch)
10516 PERL_UNUSED_CONTEXT;
10519 else if (ch == 'g')
10520 *pmfl |= PMf_GLOBAL;
10521 else if (ch == 'c')
10522 *pmfl |= PMf_CONTINUE;
10523 else if (ch == 'o')
10525 else if (ch == 'm')
10526 *pmfl |= PMf_MULTILINE;
10527 else if (ch == 's')
10528 *pmfl |= PMf_SINGLELINE;
10529 else if (ch == 'x')
10530 *pmfl |= PMf_EXTENDED;
10534 S_scan_pat(pTHX_ char *start, I32 type)
10538 char *s = scan_str(start,!!PL_madskills,FALSE);
10539 const char * const valid_flags = (type == OP_QR) ? "iomsx" : "iogcmsx";
10546 const char * const delimiter = skipspace(start);
10547 Perl_croak(aTHX_ *delimiter == '?'
10548 ? "Search pattern not terminated or ternary operator parsed as search pattern"
10549 : "Search pattern not terminated" );
10552 pm = (PMOP*)newPMOP(type, 0);
10553 if (PL_multi_open == '?')
10554 pm->op_pmflags |= PMf_ONCE;
10558 while (*s && strchr(valid_flags, *s))
10559 pmflag(&pm->op_pmflags,*s++);
10561 if (PL_madskills && modstart != s) {
10562 SV* tmptoken = newSVpvn(modstart, s - modstart);
10563 append_madprops(newMADPROP('m', MAD_SV, tmptoken, 0), (OP*)pm, 0);
10566 /* issue a warning if /c is specified,but /g is not */
10567 if ((pm->op_pmflags & PMf_CONTINUE) && !(pm->op_pmflags & PMf_GLOBAL)
10568 && ckWARN(WARN_REGEXP))
10570 Perl_warner(aTHX_ packWARN(WARN_REGEXP), "Use of /c modifier is meaningless without /g" );
10573 pm->op_pmpermflags = pm->op_pmflags;
10575 PL_lex_op = (OP*)pm;
10576 yylval.ival = OP_MATCH;
10581 S_scan_subst(pTHX_ char *start)
10592 yylval.ival = OP_NULL;
10594 s = scan_str(start,!!PL_madskills,FALSE);
10597 Perl_croak(aTHX_ "Substitution pattern not terminated");
10599 if (s[-1] == PL_multi_open)
10602 if (PL_madskills) {
10603 CURMAD('q', thisopen);
10604 CURMAD('_', thiswhite);
10605 CURMAD('E', thisstuff);
10606 CURMAD('Q', thisclose);
10607 realtokenstart = s - SvPVX(PL_linestr);
10611 first_start = PL_multi_start;
10612 s = scan_str(s,!!PL_madskills,FALSE);
10614 if (PL_lex_stuff) {
10615 SvREFCNT_dec(PL_lex_stuff);
10616 PL_lex_stuff = NULL;
10618 Perl_croak(aTHX_ "Substitution replacement not terminated");
10620 PL_multi_start = first_start; /* so whole substitution is taken together */
10622 pm = (PMOP*)newPMOP(OP_SUBST, 0);
10625 if (PL_madskills) {
10626 CURMAD('z', thisopen);
10627 CURMAD('R', thisstuff);
10628 CURMAD('Z', thisclose);
10638 else if (strchr("iogcmsx", *s))
10639 pmflag(&pm->op_pmflags,*s++);
10645 if (PL_madskills) {
10647 curmad('m', newSVpvn(modstart, s - modstart));
10648 append_madprops(thismad, (OP*)pm, 0);
10652 if ((pm->op_pmflags & PMf_CONTINUE) && ckWARN(WARN_REGEXP)) {
10653 Perl_warner(aTHX_ packWARN(WARN_REGEXP), "Use of /c modifier is meaningless in s///" );
10657 SV * const repl = newSVpvs("");
10659 PL_sublex_info.super_bufptr = s;
10660 PL_sublex_info.super_bufend = PL_bufend;
10662 pm->op_pmflags |= PMf_EVAL;
10664 sv_catpv(repl, es ? "eval " : "do ");
10665 sv_catpvs(repl, "{");
10666 sv_catsv(repl, PL_lex_repl);
10667 sv_catpvs(repl, "}");
10669 SvREFCNT_dec(PL_lex_repl);
10670 PL_lex_repl = repl;
10673 pm->op_pmpermflags = pm->op_pmflags;
10674 PL_lex_op = (OP*)pm;
10675 yylval.ival = OP_SUBST;
10680 S_scan_trans(pTHX_ char *start)
10693 yylval.ival = OP_NULL;
10695 s = scan_str(start,!!PL_madskills,FALSE);
10697 Perl_croak(aTHX_ "Transliteration pattern not terminated");
10699 if (s[-1] == PL_multi_open)
10702 if (PL_madskills) {
10703 CURMAD('q', thisopen);
10704 CURMAD('_', thiswhite);
10705 CURMAD('E', thisstuff);
10706 CURMAD('Q', thisclose);
10707 realtokenstart = s - SvPVX(PL_linestr);
10711 s = scan_str(s,!!PL_madskills,FALSE);
10713 if (PL_lex_stuff) {
10714 SvREFCNT_dec(PL_lex_stuff);
10715 PL_lex_stuff = NULL;
10717 Perl_croak(aTHX_ "Transliteration replacement not terminated");
10719 if (PL_madskills) {
10720 CURMAD('z', thisopen);
10721 CURMAD('R', thisstuff);
10722 CURMAD('Z', thisclose);
10725 complement = del = squash = 0;
10732 complement = OPpTRANS_COMPLEMENT;
10735 del = OPpTRANS_DELETE;
10738 squash = OPpTRANS_SQUASH;
10747 Newx(tbl, complement&&!del?258:256, short);
10748 o = newPVOP(OP_TRANS, 0, (char*)tbl);
10749 o->op_private &= ~OPpTRANS_ALL;
10750 o->op_private |= del|squash|complement|
10751 (DO_UTF8(PL_lex_stuff)? OPpTRANS_FROM_UTF : 0)|
10752 (DO_UTF8(PL_lex_repl) ? OPpTRANS_TO_UTF : 0);
10755 yylval.ival = OP_TRANS;
10758 if (PL_madskills) {
10760 curmad('m', newSVpvn(modstart, s - modstart));
10761 append_madprops(thismad, o, 0);
10770 S_scan_heredoc(pTHX_ register char *s)
10774 I32 op_type = OP_SCALAR;
10778 const char *found_newline;
10782 const int outer = (PL_rsfp && !(PL_lex_inwhat == OP_SCALAR));
10784 I32 stuffstart = s - SvPVX(PL_linestr);
10787 realtokenstart = -1;
10792 e = PL_tokenbuf + sizeof PL_tokenbuf - 1;
10795 for (peek = s; SPACE_OR_TAB(*peek); peek++) ;
10796 if (*peek == '`' || *peek == '\'' || *peek =='"') {
10799 s = delimcpy(d, e, s, PL_bufend, term, &len);
10809 if (!isALNUM_lazy_if(s,UTF))
10810 deprecate_old("bare << to mean <<\"\"");
10811 for (; isALNUM_lazy_if(s,UTF); s++) {
10816 if (d >= PL_tokenbuf + sizeof PL_tokenbuf - 1)
10817 Perl_croak(aTHX_ "Delimiter for here document is too long");
10820 len = d - PL_tokenbuf;
10823 if (PL_madskills) {
10824 tstart = PL_tokenbuf + !outer;
10825 thisclose = newSVpvn(tstart, len - !outer);
10826 tstart = SvPVX(PL_linestr) + stuffstart;
10827 thisopen = newSVpvn(tstart, s - tstart);
10828 stuffstart = s - SvPVX(PL_linestr);
10831 #ifndef PERL_STRICT_CR
10832 d = strchr(s, '\r');
10834 char * const olds = s;
10836 while (s < PL_bufend) {
10842 else if (*s == '\n' && s[1] == '\r') { /* \015\013 on a mac? */
10851 SvCUR_set(PL_linestr, PL_bufend - SvPVX_const(PL_linestr));
10858 if ( outer || !(found_newline = memchr(s, '\n', PL_bufend - s)) ) {
10859 herewas = newSVpvn(s,PL_bufend-s);
10863 herewas = newSVpvn(s-1,found_newline-s+1);
10866 herewas = newSVpvn(s,found_newline-s);
10870 if (PL_madskills) {
10871 tstart = SvPVX(PL_linestr) + stuffstart;
10873 sv_catpvn(thisstuff, tstart, s - tstart);
10875 thisstuff = newSVpvn(tstart, s - tstart);
10878 s += SvCUR(herewas);
10881 stuffstart = s - SvPVX(PL_linestr);
10887 tmpstr = newSV(79);
10888 sv_upgrade(tmpstr, SVt_PVIV);
10889 if (term == '\'') {
10890 op_type = OP_CONST;
10891 SvIV_set(tmpstr, -1);
10893 else if (term == '`') {
10894 op_type = OP_BACKTICK;
10895 SvIV_set(tmpstr, '\\');
10899 PL_multi_start = CopLINE(PL_curcop);
10900 PL_multi_open = PL_multi_close = '<';
10901 term = *PL_tokenbuf;
10902 if (PL_lex_inwhat == OP_SUBST && PL_in_eval && !PL_rsfp) {
10903 char * const bufptr = PL_sublex_info.super_bufptr;
10904 char * const bufend = PL_sublex_info.super_bufend;
10905 char * const olds = s - SvCUR(herewas);
10906 s = strchr(bufptr, '\n');
10910 while (s < bufend &&
10911 (*s != term || memNE(s,PL_tokenbuf,len)) ) {
10913 CopLINE_inc(PL_curcop);
10916 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
10917 missingterm(PL_tokenbuf);
10919 sv_setpvn(herewas,bufptr,d-bufptr+1);
10920 sv_setpvn(tmpstr,d+1,s-d);
10922 sv_catpvn(herewas,s,bufend-s);
10923 Copy(SvPVX_const(herewas),bufptr,SvCUR(herewas) + 1,char);
10930 while (s < PL_bufend &&
10931 (*s != term || memNE(s,PL_tokenbuf,len)) ) {
10933 CopLINE_inc(PL_curcop);
10935 if (s >= PL_bufend) {
10936 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
10937 missingterm(PL_tokenbuf);
10939 sv_setpvn(tmpstr,d+1,s-d);
10941 if (PL_madskills) {
10943 sv_catpvn(thisstuff, d + 1, s - d);
10945 thisstuff = newSVpvn(d + 1, s - d);
10946 stuffstart = s - SvPVX(PL_linestr);
10950 CopLINE_inc(PL_curcop); /* the preceding stmt passes a newline */
10952 sv_catpvn(herewas,s,PL_bufend-s);
10953 sv_setsv(PL_linestr,herewas);
10954 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = s = PL_linestart = SvPVX(PL_linestr);
10955 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
10956 PL_last_lop = PL_last_uni = NULL;
10959 sv_setpvn(tmpstr,"",0); /* avoid "uninitialized" warning */
10960 while (s >= PL_bufend) { /* multiple line string? */
10962 if (PL_madskills) {
10963 tstart = SvPVX(PL_linestr) + stuffstart;
10965 sv_catpvn(thisstuff, tstart, PL_bufend - tstart);
10967 thisstuff = newSVpvn(tstart, PL_bufend - tstart);
10971 !(PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = filter_gets(PL_linestr, PL_rsfp, 0))) {
10972 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
10973 missingterm(PL_tokenbuf);
10976 stuffstart = s - SvPVX(PL_linestr);
10978 CopLINE_inc(PL_curcop);
10979 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
10980 PL_last_lop = PL_last_uni = NULL;
10981 #ifndef PERL_STRICT_CR
10982 if (PL_bufend - PL_linestart >= 2) {
10983 if ((PL_bufend[-2] == '\r' && PL_bufend[-1] == '\n') ||
10984 (PL_bufend[-2] == '\n' && PL_bufend[-1] == '\r'))
10986 PL_bufend[-2] = '\n';
10988 SvCUR_set(PL_linestr, PL_bufend - SvPVX_const(PL_linestr));
10990 else if (PL_bufend[-1] == '\r')
10991 PL_bufend[-1] = '\n';
10993 else if (PL_bufend - PL_linestart == 1 && PL_bufend[-1] == '\r')
10994 PL_bufend[-1] = '\n';
10996 if (PERLDB_LINE && PL_curstash != PL_debstash) {
10997 SV * const sv = newSV(0);
10999 sv_upgrade(sv, SVt_PVMG);
11000 sv_setsv(sv,PL_linestr);
11001 (void)SvIOK_on(sv);
11003 av_store(CopFILEAVx(PL_curcop), (I32)CopLINE(PL_curcop),sv);
11005 if (*s == term && memEQ(s,PL_tokenbuf,len)) {
11006 STRLEN off = PL_bufend - 1 - SvPVX_const(PL_linestr);
11007 *(SvPVX(PL_linestr) + off ) = ' ';
11008 sv_catsv(PL_linestr,herewas);
11009 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
11010 s = SvPVX(PL_linestr) + off; /* In case PV of PL_linestr moved. */
11014 sv_catsv(tmpstr,PL_linestr);
11019 PL_multi_end = CopLINE(PL_curcop);
11020 if (SvCUR(tmpstr) + 5 < SvLEN(tmpstr)) {
11021 SvPV_shrink_to_cur(tmpstr);
11023 SvREFCNT_dec(herewas);
11025 if (UTF && is_utf8_string((U8*)SvPVX_const(tmpstr), SvCUR(tmpstr)))
11027 else if (PL_encoding)
11028 sv_recode_to_utf8(tmpstr, PL_encoding);
11030 PL_lex_stuff = tmpstr;
11031 yylval.ival = op_type;
11035 /* scan_inputsymbol
11036 takes: current position in input buffer
11037 returns: new position in input buffer
11038 side-effects: yylval and lex_op are set.
11043 <FH> read from filehandle
11044 <pkg::FH> read from package qualified filehandle
11045 <pkg'FH> read from package qualified filehandle
11046 <$fh> read from filehandle in $fh
11047 <*.h> filename glob
11052 S_scan_inputsymbol(pTHX_ char *start)
11055 register char *s = start; /* current position in buffer */
11059 char *d = PL_tokenbuf; /* start of temp holding space */
11060 const char * const e = PL_tokenbuf + sizeof PL_tokenbuf; /* end of temp holding space */
11062 end = strchr(s, '\n');
11065 s = delimcpy(d, e, s + 1, end, '>', &len); /* extract until > */
11067 /* die if we didn't have space for the contents of the <>,
11068 or if it didn't end, or if we see a newline
11071 if (len >= sizeof PL_tokenbuf)
11072 Perl_croak(aTHX_ "Excessively long <> operator");
11074 Perl_croak(aTHX_ "Unterminated <> operator");
11079 Remember, only scalar variables are interpreted as filehandles by
11080 this code. Anything more complex (e.g., <$fh{$num}>) will be
11081 treated as a glob() call.
11082 This code makes use of the fact that except for the $ at the front,
11083 a scalar variable and a filehandle look the same.
11085 if (*d == '$' && d[1]) d++;
11087 /* allow <Pkg'VALUE> or <Pkg::VALUE> */
11088 while (*d && (isALNUM_lazy_if(d,UTF) || *d == '\'' || *d == ':'))
11091 /* If we've tried to read what we allow filehandles to look like, and
11092 there's still text left, then it must be a glob() and not a getline.
11093 Use scan_str to pull out the stuff between the <> and treat it
11094 as nothing more than a string.
11097 if (d - PL_tokenbuf != len) {
11098 yylval.ival = OP_GLOB;
11100 s = scan_str(start,!!PL_madskills,FALSE);
11102 Perl_croak(aTHX_ "Glob not terminated");
11106 bool readline_overriden = FALSE;
11109 /* we're in a filehandle read situation */
11112 /* turn <> into <ARGV> */
11114 Copy("ARGV",d,5,char);
11116 /* Check whether readline() is overriden */
11117 gv_readline = gv_fetchpvs("readline", GV_NOTQUAL, SVt_PVCV);
11119 && GvCVu(gv_readline) && GvIMPORTED_CV(gv_readline))
11121 ((gvp = (GV**)hv_fetchs(PL_globalstash, "readline", FALSE))
11122 && (gv_readline = *gvp) != (GV*)&PL_sv_undef
11123 && GvCVu(gv_readline) && GvIMPORTED_CV(gv_readline)))
11124 readline_overriden = TRUE;
11126 /* if <$fh>, create the ops to turn the variable into a
11132 /* try to find it in the pad for this block, otherwise find
11133 add symbol table ops
11135 if ((tmp = pad_findmy(d)) != NOT_IN_PAD) {
11136 if (PAD_COMPNAME_FLAGS_isOUR(tmp)) {
11137 HV * const stash = PAD_COMPNAME_OURSTASH(tmp);
11138 HEK * const stashname = HvNAME_HEK(stash);
11139 SV * const sym = sv_2mortal(newSVhek(stashname));
11140 sv_catpvs(sym, "::");
11141 sv_catpv(sym, d+1);
11146 OP * const o = newOP(OP_PADSV, 0);
11148 PL_lex_op = readline_overriden
11149 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
11150 append_elem(OP_LIST, o,
11151 newCVREF(0, newGVOP(OP_GV,0,gv_readline))))
11152 : (OP*)newUNOP(OP_READLINE, 0, o);
11161 ? (GV_ADDMULTI | GV_ADDINEVAL)
11164 PL_lex_op = readline_overriden
11165 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
11166 append_elem(OP_LIST,
11167 newUNOP(OP_RV2SV, 0, newGVOP(OP_GV, 0, gv)),
11168 newCVREF(0, newGVOP(OP_GV, 0, gv_readline))))
11169 : (OP*)newUNOP(OP_READLINE, 0,
11170 newUNOP(OP_RV2SV, 0,
11171 newGVOP(OP_GV, 0, gv)));
11173 if (!readline_overriden)
11174 PL_lex_op->op_flags |= OPf_SPECIAL;
11175 /* we created the ops in PL_lex_op, so make yylval.ival a null op */
11176 yylval.ival = OP_NULL;
11179 /* If it's none of the above, it must be a literal filehandle
11180 (<Foo::BAR> or <FOO>) so build a simple readline OP */
11182 GV * const gv = gv_fetchpv(d, GV_ADD, SVt_PVIO);
11183 PL_lex_op = readline_overriden
11184 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
11185 append_elem(OP_LIST,
11186 newGVOP(OP_GV, 0, gv),
11187 newCVREF(0, newGVOP(OP_GV, 0, gv_readline))))
11188 : (OP*)newUNOP(OP_READLINE, 0, newGVOP(OP_GV, 0, gv));
11189 yylval.ival = OP_NULL;
11198 takes: start position in buffer
11199 keep_quoted preserve \ on the embedded delimiter(s)
11200 keep_delims preserve the delimiters around the string
11201 returns: position to continue reading from buffer
11202 side-effects: multi_start, multi_close, lex_repl or lex_stuff, and
11203 updates the read buffer.
11205 This subroutine pulls a string out of the input. It is called for:
11206 q single quotes q(literal text)
11207 ' single quotes 'literal text'
11208 qq double quotes qq(interpolate $here please)
11209 " double quotes "interpolate $here please"
11210 qx backticks qx(/bin/ls -l)
11211 ` backticks `/bin/ls -l`
11212 qw quote words @EXPORT_OK = qw( func() $spam )
11213 m// regexp match m/this/
11214 s/// regexp substitute s/this/that/
11215 tr/// string transliterate tr/this/that/
11216 y/// string transliterate y/this/that/
11217 ($*@) sub prototypes sub foo ($)
11218 (stuff) sub attr parameters sub foo : attr(stuff)
11219 <> readline or globs <FOO>, <>, <$fh>, or <*.c>
11221 In most of these cases (all but <>, patterns and transliterate)
11222 yylex() calls scan_str(). m// makes yylex() call scan_pat() which
11223 calls scan_str(). s/// makes yylex() call scan_subst() which calls
11224 scan_str(). tr/// and y/// make yylex() call scan_trans() which
11227 It skips whitespace before the string starts, and treats the first
11228 character as the delimiter. If the delimiter is one of ([{< then
11229 the corresponding "close" character )]}> is used as the closing
11230 delimiter. It allows quoting of delimiters, and if the string has
11231 balanced delimiters ([{<>}]) it allows nesting.
11233 On success, the SV with the resulting string is put into lex_stuff or,
11234 if that is already non-NULL, into lex_repl. The second case occurs only
11235 when parsing the RHS of the special constructs s/// and tr/// (y///).
11236 For convenience, the terminating delimiter character is stuffed into
11241 S_scan_str(pTHX_ char *start, int keep_quoted, int keep_delims)
11244 SV *sv; /* scalar value: string */
11245 char *tmps; /* temp string, used for delimiter matching */
11246 register char *s = start; /* current position in the buffer */
11247 register char term; /* terminating character */
11248 register char *to; /* current position in the sv's data */
11249 I32 brackets = 1; /* bracket nesting level */
11250 bool has_utf8 = FALSE; /* is there any utf8 content? */
11251 I32 termcode; /* terminating char. code */
11252 U8 termstr[UTF8_MAXBYTES]; /* terminating string */
11253 STRLEN termlen; /* length of terminating string */
11254 char *last = NULL; /* last position for nesting bracket */
11260 /* skip space before the delimiter */
11266 if (realtokenstart >= 0) {
11267 stuffstart = realtokenstart;
11268 realtokenstart = -1;
11271 stuffstart = start - SvPVX(PL_linestr);
11273 /* mark where we are, in case we need to report errors */
11276 /* after skipping whitespace, the next character is the terminator */
11279 termcode = termstr[0] = term;
11283 termcode = utf8_to_uvchr((U8*)s, &termlen);
11284 Copy(s, termstr, termlen, U8);
11285 if (!UTF8_IS_INVARIANT(term))
11289 /* mark where we are */
11290 PL_multi_start = CopLINE(PL_curcop);
11291 PL_multi_open = term;
11293 /* find corresponding closing delimiter */
11294 if (term && (tmps = strchr("([{< )]}> )]}>",term)))
11295 termcode = termstr[0] = term = tmps[5];
11297 PL_multi_close = term;
11299 /* create a new SV to hold the contents. 79 is the SV's initial length.
11300 What a random number. */
11302 sv_upgrade(sv, SVt_PVIV);
11303 SvIV_set(sv, termcode);
11304 (void)SvPOK_only(sv); /* validate pointer */
11306 /* move past delimiter and try to read a complete string */
11308 sv_catpvn(sv, s, termlen);
11311 tstart = SvPVX(PL_linestr) + stuffstart;
11312 if (!thisopen && !keep_delims) {
11313 thisopen = newSVpvn(tstart, s - tstart);
11314 stuffstart = s - SvPVX(PL_linestr);
11318 if (PL_encoding && !UTF) {
11322 int offset = s - SvPVX_const(PL_linestr);
11323 const bool found = sv_cat_decode(sv, PL_encoding, PL_linestr,
11324 &offset, (char*)termstr, termlen);
11325 const char * const ns = SvPVX_const(PL_linestr) + offset;
11326 char * const svlast = SvEND(sv) - 1;
11328 for (; s < ns; s++) {
11329 if (*s == '\n' && !PL_rsfp)
11330 CopLINE_inc(PL_curcop);
11333 goto read_more_line;
11335 /* handle quoted delimiters */
11336 if (SvCUR(sv) > 1 && *(svlast-1) == '\\') {
11338 for (t = svlast-2; t >= SvPVX_const(sv) && *t == '\\';)
11340 if ((svlast-1 - t) % 2) {
11341 if (!keep_quoted) {
11342 *(svlast-1) = term;
11344 SvCUR_set(sv, SvCUR(sv) - 1);
11349 if (PL_multi_open == PL_multi_close) {
11357 for (t = w = last; t < svlast; w++, t++) {
11358 /* At here, all closes are "was quoted" one,
11359 so we don't check PL_multi_close. */
11361 if (!keep_quoted && *(t+1) == PL_multi_open)
11366 else if (*t == PL_multi_open)
11374 SvCUR_set(sv, w - SvPVX_const(sv));
11377 if (--brackets <= 0)
11382 if (!keep_delims) {
11383 SvCUR_set(sv, SvCUR(sv) - 1);
11389 /* extend sv if need be */
11390 SvGROW(sv, SvCUR(sv) + (PL_bufend - s) + 1);
11391 /* set 'to' to the next character in the sv's string */
11392 to = SvPVX(sv)+SvCUR(sv);
11394 /* if open delimiter is the close delimiter read unbridle */
11395 if (PL_multi_open == PL_multi_close) {
11396 for (; s < PL_bufend; s++,to++) {
11397 /* embedded newlines increment the current line number */
11398 if (*s == '\n' && !PL_rsfp)
11399 CopLINE_inc(PL_curcop);
11400 /* handle quoted delimiters */
11401 if (*s == '\\' && s+1 < PL_bufend && term != '\\') {
11402 if (!keep_quoted && s[1] == term)
11404 /* any other quotes are simply copied straight through */
11408 /* terminate when run out of buffer (the for() condition), or
11409 have found the terminator */
11410 else if (*s == term) {
11413 if (s+termlen <= PL_bufend && memEQ(s, (char*)termstr, termlen))
11416 else if (!has_utf8 && !UTF8_IS_INVARIANT((U8)*s) && UTF)
11422 /* if the terminator isn't the same as the start character (e.g.,
11423 matched brackets), we have to allow more in the quoting, and
11424 be prepared for nested brackets.
11427 /* read until we run out of string, or we find the terminator */
11428 for (; s < PL_bufend; s++,to++) {
11429 /* embedded newlines increment the line count */
11430 if (*s == '\n' && !PL_rsfp)
11431 CopLINE_inc(PL_curcop);
11432 /* backslashes can escape the open or closing characters */
11433 if (*s == '\\' && s+1 < PL_bufend) {
11434 if (!keep_quoted &&
11435 ((s[1] == PL_multi_open) || (s[1] == PL_multi_close)))
11440 /* allow nested opens and closes */
11441 else if (*s == PL_multi_close && --brackets <= 0)
11443 else if (*s == PL_multi_open)
11445 else if (!has_utf8 && !UTF8_IS_INVARIANT((U8)*s) && UTF)
11450 /* terminate the copied string and update the sv's end-of-string */
11452 SvCUR_set(sv, to - SvPVX_const(sv));
11455 * this next chunk reads more into the buffer if we're not done yet
11459 break; /* handle case where we are done yet :-) */
11461 #ifndef PERL_STRICT_CR
11462 if (to - SvPVX_const(sv) >= 2) {
11463 if ((to[-2] == '\r' && to[-1] == '\n') ||
11464 (to[-2] == '\n' && to[-1] == '\r'))
11468 SvCUR_set(sv, to - SvPVX_const(sv));
11470 else if (to[-1] == '\r')
11473 else if (to - SvPVX_const(sv) == 1 && to[-1] == '\r')
11478 /* if we're out of file, or a read fails, bail and reset the current
11479 line marker so we can report where the unterminated string began
11482 if (PL_madskills) {
11483 char *tstart = SvPVX(PL_linestr) + stuffstart;
11485 sv_catpvn(thisstuff, tstart, PL_bufend - tstart);
11487 thisstuff = newSVpvn(tstart, PL_bufend - tstart);
11491 !(PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = filter_gets(PL_linestr, PL_rsfp, 0))) {
11493 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
11499 /* we read a line, so increment our line counter */
11500 CopLINE_inc(PL_curcop);
11502 /* update debugger info */
11503 if (PERLDB_LINE && PL_curstash != PL_debstash) {
11504 SV * const line_sv = newSV(0);
11506 sv_upgrade(line_sv, SVt_PVMG);
11507 sv_setsv(line_sv,PL_linestr);
11508 (void)SvIOK_on(line_sv);
11509 SvIV_set(line_sv, 0);
11510 av_store(CopFILEAVx(PL_curcop), (I32)CopLINE(PL_curcop), line_sv);
11513 /* having changed the buffer, we must update PL_bufend */
11514 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
11515 PL_last_lop = PL_last_uni = NULL;
11518 /* at this point, we have successfully read the delimited string */
11520 if (!PL_encoding || UTF) {
11522 if (PL_madskills) {
11523 char *tstart = SvPVX(PL_linestr) + stuffstart;
11525 sv_catpvn(thisstuff, tstart, s - tstart);
11527 thisstuff = newSVpvn(tstart, s - tstart);
11528 if (!thisclose && !keep_delims)
11529 thisclose = newSVpvn(s,termlen);
11534 sv_catpvn(sv, s, termlen);
11539 if (PL_madskills) {
11540 char *tstart = SvPVX(PL_linestr) + stuffstart;
11542 sv_catpvn(thisstuff, tstart, s - tstart - termlen);
11544 thisstuff = newSVpvn(tstart, s - tstart - termlen);
11545 if (!thisclose && !keep_delims)
11546 thisclose = newSVpvn(s - termlen,termlen);
11550 if (has_utf8 || PL_encoding)
11553 PL_multi_end = CopLINE(PL_curcop);
11555 /* if we allocated too much space, give some back */
11556 if (SvCUR(sv) + 5 < SvLEN(sv)) {
11557 SvLEN_set(sv, SvCUR(sv) + 1);
11558 SvPV_renew(sv, SvLEN(sv));
11561 /* decide whether this is the first or second quoted string we've read
11574 takes: pointer to position in buffer
11575 returns: pointer to new position in buffer
11576 side-effects: builds ops for the constant in yylval.op
11578 Read a number in any of the formats that Perl accepts:
11580 \d(_?\d)*(\.(\d(_?\d)*)?)?[Ee][\+\-]?(\d(_?\d)*) 12 12.34 12.
11581 \.\d(_?\d)*[Ee][\+\-]?(\d(_?\d)*) .34
11584 0x[0-9A-Fa-f](_?[0-9A-Fa-f])*
11586 Like most scan_ routines, it uses the PL_tokenbuf buffer to hold the
11589 If it reads a number without a decimal point or an exponent, it will
11590 try converting the number to an integer and see if it can do so
11591 without loss of precision.
11595 Perl_scan_num(pTHX_ const char *start, YYSTYPE* lvalp)
11598 register const char *s = start; /* current position in buffer */
11599 register char *d; /* destination in temp buffer */
11600 register char *e; /* end of temp buffer */
11601 NV nv; /* number read, as a double */
11602 SV *sv = NULL; /* place to put the converted number */
11603 bool floatit; /* boolean: int or float? */
11604 const char *lastub = NULL; /* position of last underbar */
11605 static char const number_too_long[] = "Number too long";
11607 /* We use the first character to decide what type of number this is */
11611 Perl_croak(aTHX_ "panic: scan_num");
11613 /* if it starts with a 0, it could be an octal number, a decimal in
11614 0.13 disguise, or a hexadecimal number, or a binary number. */
11618 u holds the "number so far"
11619 shift the power of 2 of the base
11620 (hex == 4, octal == 3, binary == 1)
11621 overflowed was the number more than we can hold?
11623 Shift is used when we add a digit. It also serves as an "are
11624 we in octal/hex/binary?" indicator to disallow hex characters
11625 when in octal mode.
11630 bool overflowed = FALSE;
11631 bool just_zero = TRUE; /* just plain 0 or binary number? */
11632 static const NV nvshift[5] = { 1.0, 2.0, 4.0, 8.0, 16.0 };
11633 static const char* const bases[5] =
11634 { "", "binary", "", "octal", "hexadecimal" };
11635 static const char* const Bases[5] =
11636 { "", "Binary", "", "Octal", "Hexadecimal" };
11637 static const char* const maxima[5] =
11639 "0b11111111111111111111111111111111",
11643 const char *base, *Base, *max;
11645 /* check for hex */
11650 } else if (s[1] == 'b') {
11655 /* check for a decimal in disguise */
11656 else if (s[1] == '.' || s[1] == 'e' || s[1] == 'E')
11658 /* so it must be octal */
11665 if (ckWARN(WARN_SYNTAX))
11666 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
11667 "Misplaced _ in number");
11671 base = bases[shift];
11672 Base = Bases[shift];
11673 max = maxima[shift];
11675 /* read the rest of the number */
11677 /* x is used in the overflow test,
11678 b is the digit we're adding on. */
11683 /* if we don't mention it, we're done */
11687 /* _ are ignored -- but warned about if consecutive */
11689 if (lastub && s == lastub + 1 && ckWARN(WARN_SYNTAX))
11690 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
11691 "Misplaced _ in number");
11695 /* 8 and 9 are not octal */
11696 case '8': case '9':
11698 yyerror(Perl_form(aTHX_ "Illegal octal digit '%c'", *s));
11702 case '2': case '3': case '4':
11703 case '5': case '6': case '7':
11705 yyerror(Perl_form(aTHX_ "Illegal binary digit '%c'", *s));
11708 case '0': case '1':
11709 b = *s++ & 15; /* ASCII digit -> value of digit */
11713 case 'a': case 'b': case 'c': case 'd': case 'e': case 'f':
11714 case 'A': case 'B': case 'C': case 'D': case 'E': case 'F':
11715 /* make sure they said 0x */
11718 b = (*s++ & 7) + 9;
11720 /* Prepare to put the digit we have onto the end
11721 of the number so far. We check for overflows.
11727 x = u << shift; /* make room for the digit */
11729 if ((x >> shift) != u
11730 && !(PL_hints & HINT_NEW_BINARY)) {
11733 if (ckWARN_d(WARN_OVERFLOW))
11734 Perl_warner(aTHX_ packWARN(WARN_OVERFLOW),
11735 "Integer overflow in %s number",
11738 u = x | b; /* add the digit to the end */
11741 n *= nvshift[shift];
11742 /* If an NV has not enough bits in its
11743 * mantissa to represent an UV this summing of
11744 * small low-order numbers is a waste of time
11745 * (because the NV cannot preserve the
11746 * low-order bits anyway): we could just
11747 * remember when did we overflow and in the
11748 * end just multiply n by the right
11756 /* if we get here, we had success: make a scalar value from
11761 /* final misplaced underbar check */
11762 if (s[-1] == '_') {
11763 if (ckWARN(WARN_SYNTAX))
11764 Perl_warner(aTHX_ packWARN(WARN_SYNTAX), "Misplaced _ in number");
11769 if (n > 4294967295.0 && ckWARN(WARN_PORTABLE))
11770 Perl_warner(aTHX_ packWARN(WARN_PORTABLE),
11771 "%s number > %s non-portable",
11777 if (u > 0xffffffff && ckWARN(WARN_PORTABLE))
11778 Perl_warner(aTHX_ packWARN(WARN_PORTABLE),
11779 "%s number > %s non-portable",
11784 if (just_zero && (PL_hints & HINT_NEW_INTEGER))
11785 sv = new_constant(start, s - start, "integer",
11787 else if (PL_hints & HINT_NEW_BINARY)
11788 sv = new_constant(start, s - start, "binary", sv, NULL, NULL);
11793 handle decimal numbers.
11794 we're also sent here when we read a 0 as the first digit
11796 case '1': case '2': case '3': case '4': case '5':
11797 case '6': case '7': case '8': case '9': case '.':
11800 e = PL_tokenbuf + sizeof PL_tokenbuf - 6; /* room for various punctuation */
11803 /* read next group of digits and _ and copy into d */
11804 while (isDIGIT(*s) || *s == '_') {
11805 /* skip underscores, checking for misplaced ones
11809 if (lastub && s == lastub + 1 && ckWARN(WARN_SYNTAX))
11810 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
11811 "Misplaced _ in number");
11815 /* check for end of fixed-length buffer */
11817 Perl_croak(aTHX_ number_too_long);
11818 /* if we're ok, copy the character */
11823 /* final misplaced underbar check */
11824 if (lastub && s == lastub + 1) {
11825 if (ckWARN(WARN_SYNTAX))
11826 Perl_warner(aTHX_ packWARN(WARN_SYNTAX), "Misplaced _ in number");
11829 /* read a decimal portion if there is one. avoid
11830 3..5 being interpreted as the number 3. followed
11833 if (*s == '.' && s[1] != '.') {
11838 if (ckWARN(WARN_SYNTAX))
11839 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
11840 "Misplaced _ in number");
11844 /* copy, ignoring underbars, until we run out of digits.
11846 for (; isDIGIT(*s) || *s == '_'; s++) {
11847 /* fixed length buffer check */
11849 Perl_croak(aTHX_ number_too_long);
11851 if (lastub && s == lastub + 1 && ckWARN(WARN_SYNTAX))
11852 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
11853 "Misplaced _ in number");
11859 /* fractional part ending in underbar? */
11860 if (s[-1] == '_') {
11861 if (ckWARN(WARN_SYNTAX))
11862 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
11863 "Misplaced _ in number");
11865 if (*s == '.' && isDIGIT(s[1])) {
11866 /* oops, it's really a v-string, but without the "v" */
11872 /* read exponent part, if present */
11873 if ((*s == 'e' || *s == 'E') && strchr("+-0123456789_", s[1])) {
11877 /* regardless of whether user said 3E5 or 3e5, use lower 'e' */
11878 *d++ = 'e'; /* At least some Mach atof()s don't grok 'E' */
11880 /* stray preinitial _ */
11882 if (ckWARN(WARN_SYNTAX))
11883 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
11884 "Misplaced _ in number");
11888 /* allow positive or negative exponent */
11889 if (*s == '+' || *s == '-')
11892 /* stray initial _ */
11894 if (ckWARN(WARN_SYNTAX))
11895 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
11896 "Misplaced _ in number");
11900 /* read digits of exponent */
11901 while (isDIGIT(*s) || *s == '_') {
11904 Perl_croak(aTHX_ number_too_long);
11908 if (((lastub && s == lastub + 1) ||
11909 (!isDIGIT(s[1]) && s[1] != '_'))
11910 && ckWARN(WARN_SYNTAX))
11911 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
11912 "Misplaced _ in number");
11919 /* make an sv from the string */
11923 We try to do an integer conversion first if no characters
11924 indicating "float" have been found.
11929 const int flags = grok_number (PL_tokenbuf, d - PL_tokenbuf, &uv);
11931 if (flags == IS_NUMBER_IN_UV) {
11933 sv_setiv(sv, uv); /* Prefer IVs over UVs. */
11936 } else if (flags == (IS_NUMBER_IN_UV | IS_NUMBER_NEG)) {
11937 if (uv <= (UV) IV_MIN)
11938 sv_setiv(sv, -(IV)uv);
11945 /* terminate the string */
11947 nv = Atof(PL_tokenbuf);
11951 if ( floatit ? (PL_hints & HINT_NEW_FLOAT) :
11952 (PL_hints & HINT_NEW_INTEGER) )
11953 sv = new_constant(PL_tokenbuf, d - PL_tokenbuf,
11954 (floatit ? "float" : "integer"),
11958 /* if it starts with a v, it could be a v-string */
11961 sv = newSV(5); /* preallocate storage space */
11962 s = scan_vstring(s,sv);
11966 /* make the op for the constant and return */
11969 lvalp->opval = newSVOP(OP_CONST, 0, sv);
11971 lvalp->opval = NULL;
11977 S_scan_formline(pTHX_ register char *s)
11980 register char *eol;
11982 SV * const stuff = newSVpvs("");
11983 bool needargs = FALSE;
11984 bool eofmt = FALSE;
11986 char *tokenstart = s;
11989 if (PL_madskills) {
11990 savewhite = thiswhite;
11995 while (!needargs) {
11997 #ifdef PERL_STRICT_CR
11998 for (t = s+1;SPACE_OR_TAB(*t); t++) ;
12000 for (t = s+1;SPACE_OR_TAB(*t) || *t == '\r'; t++) ;
12002 if (*t == '\n' || t == PL_bufend) {
12007 if (PL_in_eval && !PL_rsfp) {
12008 eol = (char *) memchr(s,'\n',PL_bufend-s);
12013 eol = PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
12015 for (t = s; t < eol; t++) {
12016 if (*t == '~' && t[1] == '~' && SvCUR(stuff)) {
12018 goto enough; /* ~~ must be first line in formline */
12020 if (*t == '@' || *t == '^')
12024 sv_catpvn(stuff, s, eol-s);
12025 #ifndef PERL_STRICT_CR
12026 if (eol-s > 1 && eol[-2] == '\r' && eol[-1] == '\n') {
12027 char *end = SvPVX(stuff) + SvCUR(stuff);
12030 SvCUR_set(stuff, SvCUR(stuff) - 1);
12040 if (PL_madskills) {
12042 sv_catpvn(thistoken, tokenstart, PL_bufend - tokenstart);
12044 thistoken = newSVpvn(tokenstart, PL_bufend - tokenstart);
12047 s = filter_gets(PL_linestr, PL_rsfp, 0);
12049 tokenstart = PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = SvPVX(PL_linestr);
12051 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = SvPVX(PL_linestr);
12053 PL_bufend = PL_bufptr + SvCUR(PL_linestr);
12054 PL_last_lop = PL_last_uni = NULL;
12063 if (SvCUR(stuff)) {
12066 PL_lex_state = LEX_NORMAL;
12067 start_force(curforce);
12068 NEXTVAL_NEXTTOKE.ival = 0;
12072 PL_lex_state = LEX_FORMLINE;
12074 if (UTF && is_utf8_string((U8*)SvPVX_const(stuff), SvCUR(stuff)))
12076 else if (PL_encoding)
12077 sv_recode_to_utf8(stuff, PL_encoding);
12079 start_force(curforce);
12080 NEXTVAL_NEXTTOKE.opval = (OP*)newSVOP(OP_CONST, 0, stuff);
12082 start_force(curforce);
12083 NEXTVAL_NEXTTOKE.ival = OP_FORMLINE;
12087 SvREFCNT_dec(stuff);
12089 PL_lex_formbrack = 0;
12093 if (PL_madskills) {
12095 sv_catpvn(thistoken, tokenstart, s - tokenstart);
12097 thistoken = newSVpvn(tokenstart, s - tokenstart);
12098 thiswhite = savewhite;
12110 PL_cshlen = strlen(PL_cshname);
12112 #if defined(USE_ITHREADS)
12113 PERL_UNUSED_CONTEXT;
12119 Perl_start_subparse(pTHX_ I32 is_format, U32 flags)
12122 const I32 oldsavestack_ix = PL_savestack_ix;
12123 CV* const outsidecv = PL_compcv;
12126 assert(SvTYPE(PL_compcv) == SVt_PVCV);
12128 SAVEI32(PL_subline);
12129 save_item(PL_subname);
12130 SAVESPTR(PL_compcv);
12132 PL_compcv = (CV*)newSV(0);
12133 sv_upgrade((SV *)PL_compcv, is_format ? SVt_PVFM : SVt_PVCV);
12134 CvFLAGS(PL_compcv) |= flags;
12136 PL_subline = CopLINE(PL_curcop);
12137 CvPADLIST(PL_compcv) = pad_new(padnew_SAVE|padnew_SAVESUB);
12138 CvOUTSIDE(PL_compcv) = (CV*)SvREFCNT_inc_simple(outsidecv);
12139 CvOUTSIDE_SEQ(PL_compcv) = PL_cop_seqmax;
12141 return oldsavestack_ix;
12145 #pragma segment Perl_yylex
12148 Perl_yywarn(pTHX_ const char *s)
12151 PL_in_eval |= EVAL_WARNONLY;
12153 PL_in_eval &= ~EVAL_WARNONLY;
12158 Perl_yyerror(pTHX_ const char *s)
12161 const char *where = NULL;
12162 const char *context = NULL;
12166 if (!yychar || (yychar == ';' && !PL_rsfp))
12168 else if (PL_oldoldbufptr && PL_bufptr > PL_oldoldbufptr &&
12169 PL_bufptr - PL_oldoldbufptr < 200 && PL_oldoldbufptr != PL_oldbufptr &&
12170 PL_oldbufptr != PL_bufptr) {
12173 The code below is removed for NetWare because it abends/crashes on NetWare
12174 when the script has error such as not having the closing quotes like:
12175 if ($var eq "value)
12176 Checking of white spaces is anyway done in NetWare code.
12179 while (isSPACE(*PL_oldoldbufptr))
12182 context = PL_oldoldbufptr;
12183 contlen = PL_bufptr - PL_oldoldbufptr;
12185 else if (PL_oldbufptr && PL_bufptr > PL_oldbufptr &&
12186 PL_bufptr - PL_oldbufptr < 200 && PL_oldbufptr != PL_bufptr) {
12189 The code below is removed for NetWare because it abends/crashes on NetWare
12190 when the script has error such as not having the closing quotes like:
12191 if ($var eq "value)
12192 Checking of white spaces is anyway done in NetWare code.
12195 while (isSPACE(*PL_oldbufptr))
12198 context = PL_oldbufptr;
12199 contlen = PL_bufptr - PL_oldbufptr;
12201 else if (yychar > 255)
12202 where = "next token ???";
12203 else if (yychar == -2) { /* YYEMPTY */
12204 if (PL_lex_state == LEX_NORMAL ||
12205 (PL_lex_state == LEX_KNOWNEXT && PL_lex_defer == LEX_NORMAL))
12206 where = "at end of line";
12207 else if (PL_lex_inpat)
12208 where = "within pattern";
12210 where = "within string";
12213 SV * const where_sv = sv_2mortal(newSVpvs("next char "));
12215 Perl_sv_catpvf(aTHX_ where_sv, "^%c", toCTRL(yychar));
12216 else if (isPRINT_LC(yychar))
12217 Perl_sv_catpvf(aTHX_ where_sv, "%c", yychar);
12219 Perl_sv_catpvf(aTHX_ where_sv, "\\%03o", yychar & 255);
12220 where = SvPVX_const(where_sv);
12222 msg = sv_2mortal(newSVpv(s, 0));
12223 Perl_sv_catpvf(aTHX_ msg, " at %s line %"IVdf", ",
12224 OutCopFILE(PL_curcop), (IV)CopLINE(PL_curcop));
12226 Perl_sv_catpvf(aTHX_ msg, "near \"%.*s\"\n", contlen, context);
12228 Perl_sv_catpvf(aTHX_ msg, "%s\n", where);
12229 if (PL_multi_start < PL_multi_end && (U32)(CopLINE(PL_curcop) - PL_multi_end) <= 1) {
12230 Perl_sv_catpvf(aTHX_ msg,
12231 " (Might be a runaway multi-line %c%c string starting on line %"IVdf")\n",
12232 (int)PL_multi_open,(int)PL_multi_close,(IV)PL_multi_start);
12235 if (PL_in_eval & EVAL_WARNONLY && ckWARN_d(WARN_SYNTAX))
12236 Perl_warner(aTHX_ packWARN(WARN_SYNTAX), "%"SVf, msg);
12239 if (PL_error_count >= 10) {
12240 if (PL_in_eval && SvCUR(ERRSV))
12241 Perl_croak(aTHX_ "%"SVf"%s has too many errors.\n",
12242 ERRSV, OutCopFILE(PL_curcop));
12244 Perl_croak(aTHX_ "%s has too many errors.\n",
12245 OutCopFILE(PL_curcop));
12248 PL_in_my_stash = NULL;
12252 #pragma segment Main
12256 S_swallow_bom(pTHX_ U8 *s)
12259 const STRLEN slen = SvCUR(PL_linestr);
12262 if (s[1] == 0xFE) {
12263 /* UTF-16 little-endian? (or UTF32-LE?) */
12264 if (s[2] == 0 && s[3] == 0) /* UTF-32 little-endian */
12265 Perl_croak(aTHX_ "Unsupported script encoding UTF32-LE");
12266 #ifndef PERL_NO_UTF16_FILTER
12267 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF16-LE script encoding (BOM)\n");
12270 if (PL_bufend > (char*)s) {
12274 filter_add(utf16rev_textfilter, NULL);
12275 Newx(news, (PL_bufend - (char*)s) * 3 / 2 + 1, U8);
12276 utf16_to_utf8_reversed(s, news,
12277 PL_bufend - (char*)s - 1,
12279 sv_setpvn(PL_linestr, (const char*)news, newlen);
12281 s = (U8*)SvPVX(PL_linestr);
12282 Copy(news, s, newlen, U8);
12286 SvUTF8_on(PL_linestr);
12287 s = (U8*)SvPVX(PL_linestr);
12289 /* FIXME - is this a general bug fix? */
12292 PL_bufend = SvPVX(PL_linestr) + newlen;
12295 Perl_croak(aTHX_ "Unsupported script encoding UTF16-LE");
12300 if (s[1] == 0xFF) { /* UTF-16 big-endian? */
12301 #ifndef PERL_NO_UTF16_FILTER
12302 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16BE script encoding (BOM)\n");
12305 if (PL_bufend > (char *)s) {
12309 filter_add(utf16_textfilter, NULL);
12310 Newx(news, (PL_bufend - (char*)s) * 3 / 2 + 1, U8);
12311 utf16_to_utf8(s, news,
12312 PL_bufend - (char*)s,
12314 sv_setpvn(PL_linestr, (const char*)news, newlen);
12316 SvUTF8_on(PL_linestr);
12317 s = (U8*)SvPVX(PL_linestr);
12318 PL_bufend = SvPVX(PL_linestr) + newlen;
12321 Perl_croak(aTHX_ "Unsupported script encoding UTF16-BE");
12326 if (slen > 2 && s[1] == 0xBB && s[2] == 0xBF) {
12327 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-8 script encoding (BOM)\n");
12328 s += 3; /* UTF-8 */
12334 if (s[2] == 0xFE && s[3] == 0xFF) {
12335 /* UTF-32 big-endian */
12336 Perl_croak(aTHX_ "Unsupported script encoding UTF32-BE");
12339 else if (s[2] == 0 && s[3] != 0) {
12342 * are a good indicator of UTF-16BE. */
12343 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16BE script encoding (no BOM)\n");
12348 if (slen > 3 && s[1] == 0 && s[2] != 0 && s[3] == 0) {
12351 * are a good indicator of UTF-16LE. */
12352 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16LE script encoding (no BOM)\n");
12361 * Restore a source filter.
12365 restore_rsfp(pTHX_ void *f)
12368 PerlIO * const fp = (PerlIO*)f;
12370 if (PL_rsfp == PerlIO_stdin())
12371 PerlIO_clearerr(PL_rsfp);
12372 else if (PL_rsfp && (PL_rsfp != fp))
12373 PerlIO_close(PL_rsfp);
12377 #ifndef PERL_NO_UTF16_FILTER
12379 utf16_textfilter(pTHX_ int idx, SV *sv, int maxlen)
12382 const STRLEN old = SvCUR(sv);
12383 const I32 count = FILTER_READ(idx+1, sv, maxlen);
12384 DEBUG_P(PerlIO_printf(Perl_debug_log,
12385 "utf16_textfilter(%p): %d %d (%d)\n",
12386 utf16_textfilter, idx, maxlen, (int) count));
12390 Newx(tmps, SvCUR(sv) * 3 / 2 + 1, U8);
12391 Copy(SvPVX_const(sv), tmps, old, char);
12392 utf16_to_utf8((U8*)SvPVX_const(sv) + old, tmps + old,
12393 SvCUR(sv) - old, &newlen);
12394 sv_usepvn(sv, (char*)tmps, (STRLEN)newlen + old);
12396 DEBUG_P({sv_dump(sv);});
12401 utf16rev_textfilter(pTHX_ int idx, SV *sv, int maxlen)
12404 const STRLEN old = SvCUR(sv);
12405 const I32 count = FILTER_READ(idx+1, sv, maxlen);
12406 DEBUG_P(PerlIO_printf(Perl_debug_log,
12407 "utf16rev_textfilter(%p): %d %d (%d)\n",
12408 utf16rev_textfilter, idx, maxlen, (int) count));
12412 Newx(tmps, SvCUR(sv) * 3 / 2 + 1, U8);
12413 Copy(SvPVX_const(sv), tmps, old, char);
12414 utf16_to_utf8((U8*)SvPVX_const(sv) + old, tmps + old,
12415 SvCUR(sv) - old, &newlen);
12416 sv_usepvn(sv, (char*)tmps, (STRLEN)newlen + old);
12418 DEBUG_P({ sv_dump(sv); });
12424 Returns a pointer to the next character after the parsed
12425 vstring, as well as updating the passed in sv.
12427 Function must be called like
12430 s = scan_vstring(s,sv);
12432 The sv should already be large enough to store the vstring
12433 passed in, for performance reasons.
12438 Perl_scan_vstring(pTHX_ const char *s, SV *sv)
12441 const char *pos = s;
12442 const char *start = s;
12443 if (*pos == 'v') pos++; /* get past 'v' */
12444 while (pos < PL_bufend && (isDIGIT(*pos) || *pos == '_'))
12446 if ( *pos != '.') {
12447 /* this may not be a v-string if followed by => */
12448 const char *next = pos;
12449 while (next < PL_bufend && isSPACE(*next))
12451 if ((PL_bufend - next) >= 2 && *next == '=' && next[1] == '>' ) {
12452 /* return string not v-string */
12453 sv_setpvn(sv,(char *)s,pos-s);
12454 return (char *)pos;
12458 if (!isALPHA(*pos)) {
12459 U8 tmpbuf[UTF8_MAXBYTES+1];
12461 if (*s == 'v') s++; /* get past 'v' */
12463 sv_setpvn(sv, "", 0);
12469 /* this is atoi() that tolerates underscores */
12470 const char *end = pos;
12472 while (--end >= s) {
12477 rev += (*end - '0') * mult;
12479 if (orev > rev && ckWARN_d(WARN_OVERFLOW))
12480 Perl_warner(aTHX_ packWARN(WARN_OVERFLOW),
12481 "Integer overflow in decimal number");
12485 if (rev > 0x7FFFFFFF)
12486 Perl_croak(aTHX_ "In EBCDIC the v-string components cannot exceed 2147483647");
12488 /* Append native character for the rev point */
12489 tmpend = uvchr_to_utf8(tmpbuf, rev);
12490 sv_catpvn(sv, (const char*)tmpbuf, tmpend - tmpbuf);
12491 if (!UNI_IS_INVARIANT(NATIVE_TO_UNI(rev)))
12493 if (pos + 1 < PL_bufend && *pos == '.' && isDIGIT(pos[1]))
12499 while (pos < PL_bufend && (isDIGIT(*pos) || *pos == '_'))
12503 sv_magic(sv,NULL,PERL_MAGIC_vstring,(const char*)start, pos-start);
12511 * c-indentation-style: bsd
12512 * c-basic-offset: 4
12513 * indent-tabs-mode: t
12516 * ex: set ts=8 sts=4 sw=4 noet: