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 # define CURMAD(slot,sv) if (PL_madskills) { curmad(slot,sv); sv = 0; }
40 # define NEXTVAL_NEXTTOKE PL_nexttoke[PL_curforce].next_val
42 # define CURMAD(slot,sv)
43 # define NEXTVAL_NEXTTOKE PL_nextval[PL_nexttoke]
46 #define XFAKEBRACK 128
49 #ifdef USE_UTF8_SCRIPTS
50 # define UTF (!IN_BYTES)
52 # define UTF ((PL_linestr && DO_UTF8(PL_linestr)) || (PL_hints & HINT_UTF8))
55 /* In variables named $^X, these are the legal values for X.
56 * 1999-02-27 mjd-perl-patch@plover.com */
57 #define isCONTROLVAR(x) (isUPPER(x) || strchr("[\\]^_?", (x)))
59 /* On MacOS, respect nonbreaking spaces */
60 #ifdef MACOS_TRADITIONAL
61 #define SPACE_OR_TAB(c) ((c)==' '||(c)=='\312'||(c)=='\t')
63 #define SPACE_OR_TAB(c) ((c)==' '||(c)=='\t')
66 /* LEX_* are values for PL_lex_state, the state of the lexer.
67 * They are arranged oddly so that the guard on the switch statement
68 * can get by with a single comparison (if the compiler is smart enough).
71 /* #define LEX_NOTPARSING 11 is done in perl.h. */
73 #define LEX_NORMAL 10 /* normal code (ie not within "...") */
74 #define LEX_INTERPNORMAL 9 /* code within a string, eg "$foo[$x+1]" */
75 #define LEX_INTERPCASEMOD 8 /* expecting a \U, \Q or \E etc */
76 #define LEX_INTERPPUSH 7 /* starting a new sublex parse level */
77 #define LEX_INTERPSTART 6 /* expecting the start of a $var */
79 /* at end of code, eg "$x" followed by: */
80 #define LEX_INTERPEND 5 /* ... eg not one of [, { or -> */
81 #define LEX_INTERPENDMAYBE 4 /* ... eg one of [, { or -> */
83 #define LEX_INTERPCONCAT 3 /* expecting anything, eg at start of
84 string or after \E, $foo, etc */
85 #define LEX_INTERPCONST 2 /* NOT USED */
86 #define LEX_FORMLINE 1 /* expecting a format line */
87 #define LEX_KNOWNEXT 0 /* next token known; just return it */
91 static const char* const lex_state_names[] = {
110 #include "keywords.h"
112 /* CLINE is a macro that ensures PL_copline has a sane value */
117 #define CLINE (PL_copline = (CopLINE(PL_curcop) < PL_copline ? CopLINE(PL_curcop) : PL_copline))
120 # define SKIPSPACE0(s) skipspace0(s)
121 # define SKIPSPACE1(s) skipspace1(s)
122 # define SKIPSPACE2(s,tsv) skipspace2(s,&tsv)
123 # define PEEKSPACE(s) skipspace2(s,0)
125 # define SKIPSPACE0(s) skipspace(s)
126 # define SKIPSPACE1(s) skipspace(s)
127 # define SKIPSPACE2(s,tsv) skipspace(s)
128 # define PEEKSPACE(s) skipspace(s)
132 * Convenience functions to return different tokens and prime the
133 * lexer for the next token. They all take an argument.
135 * TOKEN : generic token (used for '(', DOLSHARP, etc)
136 * OPERATOR : generic operator
137 * AOPERATOR : assignment operator
138 * PREBLOCK : beginning the block after an if, while, foreach, ...
139 * PRETERMBLOCK : beginning a non-code-defining {} block (eg, hash ref)
140 * PREREF : *EXPR where EXPR is not a simple identifier
141 * TERM : expression term
142 * LOOPX : loop exiting command (goto, last, dump, etc)
143 * FTST : file test operator
144 * FUN0 : zero-argument function
145 * FUN1 : not used, except for not, which isn't a UNIOP
146 * BOop : bitwise or or xor
148 * SHop : shift operator
149 * PWop : power operator
150 * PMop : pattern-matching operator
151 * Aop : addition-level operator
152 * Mop : multiplication-level operator
153 * Eop : equality-testing operator
154 * Rop : relational operator <= != gt
156 * Also see LOP and lop() below.
159 #ifdef DEBUGGING /* Serve -DT. */
160 # define REPORT(retval) tokereport((I32)retval)
162 # define REPORT(retval) (retval)
165 #define TOKEN(retval) return ( PL_bufptr = s, REPORT(retval))
166 #define OPERATOR(retval) return (PL_expect = XTERM, PL_bufptr = s, REPORT(retval))
167 #define AOPERATOR(retval) return ao((PL_expect = XTERM, PL_bufptr = s, REPORT(retval)))
168 #define PREBLOCK(retval) return (PL_expect = XBLOCK,PL_bufptr = s, REPORT(retval))
169 #define PRETERMBLOCK(retval) return (PL_expect = XTERMBLOCK,PL_bufptr = s, REPORT(retval))
170 #define PREREF(retval) return (PL_expect = XREF,PL_bufptr = s, REPORT(retval))
171 #define TERM(retval) return (CLINE, PL_expect = XOPERATOR, PL_bufptr = s, REPORT(retval))
172 #define LOOPX(f) return (yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)LOOPEX))
173 #define FTST(f) return (yylval.ival=f, PL_expect=XTERMORDORDOR, PL_bufptr=s, REPORT((int)UNIOP))
174 #define FUN0(f) return (yylval.ival=f, PL_expect=XOPERATOR, PL_bufptr=s, REPORT((int)FUNC0))
175 #define FUN1(f) return (yylval.ival=f, PL_expect=XOPERATOR, PL_bufptr=s, REPORT((int)FUNC1))
176 #define BOop(f) return ao((yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)BITOROP)))
177 #define BAop(f) return ao((yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)BITANDOP)))
178 #define SHop(f) return ao((yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)SHIFTOP)))
179 #define PWop(f) return ao((yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)POWOP)))
180 #define PMop(f) return(yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)MATCHOP))
181 #define Aop(f) return ao((yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)ADDOP)))
182 #define Mop(f) return ao((yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)MULOP)))
183 #define Eop(f) return (yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)EQOP))
184 #define Rop(f) return (yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)RELOP))
186 /* This bit of chicanery makes a unary function followed by
187 * a parenthesis into a function with one argument, highest precedence.
188 * The UNIDOR macro is for unary functions that can be followed by the //
189 * operator (such as C<shift // 0>).
191 #define UNI2(f,x) { \
195 PL_last_uni = PL_oldbufptr; \
196 PL_last_lop_op = f; \
198 return REPORT( (int)FUNC1 ); \
200 return REPORT( *s=='(' ? (int)FUNC1 : (int)UNIOP ); \
202 #define UNI(f) UNI2(f,XTERM)
203 #define UNIDOR(f) UNI2(f,XTERMORDORDOR)
205 #define UNIBRACK(f) { \
208 PL_last_uni = PL_oldbufptr; \
210 return REPORT( (int)FUNC1 ); \
212 return REPORT( (*s == '(') ? (int)FUNC1 : (int)UNIOP ); \
215 /* grandfather return to old style */
216 #define OLDLOP(f) return(yylval.ival=f,PL_expect = XTERM,PL_bufptr = s,(int)LSTOP)
220 /* how to interpret the yylval associated with the token */
224 TOKENTYPE_OPNUM, /* yylval.ival contains an opcode number */
230 static struct debug_tokens {
232 enum token_type type;
234 } const debug_tokens[] =
236 { ADDOP, TOKENTYPE_OPNUM, "ADDOP" },
237 { ANDAND, TOKENTYPE_NONE, "ANDAND" },
238 { ANDOP, TOKENTYPE_NONE, "ANDOP" },
239 { ANONSUB, TOKENTYPE_IVAL, "ANONSUB" },
240 { ARROW, TOKENTYPE_NONE, "ARROW" },
241 { ASSIGNOP, TOKENTYPE_OPNUM, "ASSIGNOP" },
242 { BITANDOP, TOKENTYPE_OPNUM, "BITANDOP" },
243 { BITOROP, TOKENTYPE_OPNUM, "BITOROP" },
244 { COLONATTR, TOKENTYPE_NONE, "COLONATTR" },
245 { CONTINUE, TOKENTYPE_NONE, "CONTINUE" },
246 { DEFAULT, TOKENTYPE_NONE, "DEFAULT" },
247 { DO, TOKENTYPE_NONE, "DO" },
248 { DOLSHARP, TOKENTYPE_NONE, "DOLSHARP" },
249 { DORDOR, TOKENTYPE_NONE, "DORDOR" },
250 { DOROP, TOKENTYPE_OPNUM, "DOROP" },
251 { DOTDOT, TOKENTYPE_IVAL, "DOTDOT" },
252 { ELSE, TOKENTYPE_NONE, "ELSE" },
253 { ELSIF, TOKENTYPE_IVAL, "ELSIF" },
254 { EQOP, TOKENTYPE_OPNUM, "EQOP" },
255 { FOR, TOKENTYPE_IVAL, "FOR" },
256 { FORMAT, TOKENTYPE_NONE, "FORMAT" },
257 { FUNC, TOKENTYPE_OPNUM, "FUNC" },
258 { FUNC0, TOKENTYPE_OPNUM, "FUNC0" },
259 { FUNC0SUB, TOKENTYPE_OPVAL, "FUNC0SUB" },
260 { FUNC1, TOKENTYPE_OPNUM, "FUNC1" },
261 { FUNCMETH, TOKENTYPE_OPVAL, "FUNCMETH" },
262 { GIVEN, TOKENTYPE_IVAL, "GIVEN" },
263 { HASHBRACK, TOKENTYPE_NONE, "HASHBRACK" },
264 { IF, TOKENTYPE_IVAL, "IF" },
265 { LABEL, TOKENTYPE_PVAL, "LABEL" },
266 { LOCAL, TOKENTYPE_IVAL, "LOCAL" },
267 { LOOPEX, TOKENTYPE_OPNUM, "LOOPEX" },
268 { LSTOP, TOKENTYPE_OPNUM, "LSTOP" },
269 { LSTOPSUB, TOKENTYPE_OPVAL, "LSTOPSUB" },
270 { MATCHOP, TOKENTYPE_OPNUM, "MATCHOP" },
271 { METHOD, TOKENTYPE_OPVAL, "METHOD" },
272 { MULOP, TOKENTYPE_OPNUM, "MULOP" },
273 { MY, TOKENTYPE_IVAL, "MY" },
274 { MYSUB, TOKENTYPE_NONE, "MYSUB" },
275 { NOAMP, TOKENTYPE_NONE, "NOAMP" },
276 { NOTOP, TOKENTYPE_NONE, "NOTOP" },
277 { OROP, TOKENTYPE_IVAL, "OROP" },
278 { OROR, TOKENTYPE_NONE, "OROR" },
279 { PACKAGE, TOKENTYPE_NONE, "PACKAGE" },
280 { PMFUNC, TOKENTYPE_OPVAL, "PMFUNC" },
281 { POSTDEC, TOKENTYPE_NONE, "POSTDEC" },
282 { POSTINC, TOKENTYPE_NONE, "POSTINC" },
283 { POWOP, TOKENTYPE_OPNUM, "POWOP" },
284 { PREDEC, TOKENTYPE_NONE, "PREDEC" },
285 { PREINC, TOKENTYPE_NONE, "PREINC" },
286 { PRIVATEREF, TOKENTYPE_OPVAL, "PRIVATEREF" },
287 { REFGEN, TOKENTYPE_NONE, "REFGEN" },
288 { RELOP, TOKENTYPE_OPNUM, "RELOP" },
289 { SHIFTOP, TOKENTYPE_OPNUM, "SHIFTOP" },
290 { SUB, TOKENTYPE_NONE, "SUB" },
291 { THING, TOKENTYPE_OPVAL, "THING" },
292 { UMINUS, TOKENTYPE_NONE, "UMINUS" },
293 { UNIOP, TOKENTYPE_OPNUM, "UNIOP" },
294 { UNIOPSUB, TOKENTYPE_OPVAL, "UNIOPSUB" },
295 { UNLESS, TOKENTYPE_IVAL, "UNLESS" },
296 { UNTIL, TOKENTYPE_IVAL, "UNTIL" },
297 { USE, TOKENTYPE_IVAL, "USE" },
298 { WHEN, TOKENTYPE_IVAL, "WHEN" },
299 { WHILE, TOKENTYPE_IVAL, "WHILE" },
300 { WORD, TOKENTYPE_OPVAL, "WORD" },
301 { 0, TOKENTYPE_NONE, 0 }
304 /* dump the returned token in rv, plus any optional arg in yylval */
307 S_tokereport(pTHX_ I32 rv)
311 const char *name = NULL;
312 enum token_type type = TOKENTYPE_NONE;
313 const struct debug_tokens *p;
314 SV* const report = newSVpvs("<== ");
316 for (p = debug_tokens; p->token; p++) {
317 if (p->token == (int)rv) {
324 Perl_sv_catpv(aTHX_ report, name);
325 else if ((char)rv > ' ' && (char)rv < '~')
326 Perl_sv_catpvf(aTHX_ report, "'%c'", (char)rv);
328 sv_catpvs(report, "EOF");
330 Perl_sv_catpvf(aTHX_ report, "?? %"IVdf, (IV)rv);
333 case TOKENTYPE_GVVAL: /* doesn't appear to be used */
336 Perl_sv_catpvf(aTHX_ report, "(ival=%"IVdf")", (IV)yylval.ival);
338 case TOKENTYPE_OPNUM:
339 Perl_sv_catpvf(aTHX_ report, "(ival=op_%s)",
340 PL_op_name[yylval.ival]);
343 Perl_sv_catpvf(aTHX_ report, "(pval=\"%s\")", yylval.pval);
345 case TOKENTYPE_OPVAL:
347 Perl_sv_catpvf(aTHX_ report, "(opval=op_%s)",
348 PL_op_name[yylval.opval->op_type]);
349 if (yylval.opval->op_type == OP_CONST) {
350 Perl_sv_catpvf(aTHX_ report, " %s",
351 SvPEEK(cSVOPx_sv(yylval.opval)));
356 sv_catpvs(report, "(opval=null)");
359 PerlIO_printf(Perl_debug_log, "### %s\n\n", SvPV_nolen_const(report));
365 /* print the buffer with suitable escapes */
368 S_printbuf(pTHX_ const char* fmt, const char* s)
370 SV* const tmp = newSVpvs("");
371 PerlIO_printf(Perl_debug_log, fmt, pv_display(tmp, s, strlen(s), 0, 60));
380 * This subroutine detects &&=, ||=, and //= and turns an ANDAND, OROR or DORDOR
381 * into an OP_ANDASSIGN, OP_ORASSIGN, or OP_DORASSIGN
385 S_ao(pTHX_ int toketype)
388 if (*PL_bufptr == '=') {
390 if (toketype == ANDAND)
391 yylval.ival = OP_ANDASSIGN;
392 else if (toketype == OROR)
393 yylval.ival = OP_ORASSIGN;
394 else if (toketype == DORDOR)
395 yylval.ival = OP_DORASSIGN;
403 * When Perl expects an operator and finds something else, no_op
404 * prints the warning. It always prints "<something> found where
405 * operator expected. It prints "Missing semicolon on previous line?"
406 * if the surprise occurs at the start of the line. "do you need to
407 * predeclare ..." is printed out for code like "sub bar; foo bar $x"
408 * where the compiler doesn't know if foo is a method call or a function.
409 * It prints "Missing operator before end of line" if there's nothing
410 * after the missing operator, or "... before <...>" if there is something
411 * after the missing operator.
415 S_no_op(pTHX_ const char *what, char *s)
418 char * const oldbp = PL_bufptr;
419 const bool is_first = (PL_oldbufptr == PL_linestart);
425 yywarn(Perl_form(aTHX_ "%s found where operator expected", what));
426 if (ckWARN_d(WARN_SYNTAX)) {
428 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
429 "\t(Missing semicolon on previous line?)\n");
430 else if (PL_oldoldbufptr && isIDFIRST_lazy_if(PL_oldoldbufptr,UTF)) {
432 for (t = PL_oldoldbufptr; *t && (isALNUM_lazy_if(t,UTF) || *t == ':'); t++)
434 if (t < PL_bufptr && isSPACE(*t))
435 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
436 "\t(Do you need to predeclare %.*s?)\n",
437 (int)(t - PL_oldoldbufptr), PL_oldoldbufptr);
441 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
442 "\t(Missing operator before %.*s?)\n", (int)(s - oldbp), oldbp);
450 * Complain about missing quote/regexp/heredoc terminator.
451 * If it's called with NULL then it cauterizes the line buffer.
452 * If we're in a delimited string and the delimiter is a control
453 * character, it's reformatted into a two-char sequence like ^C.
458 S_missingterm(pTHX_ char *s)
464 char * const nl = strrchr(s,'\n');
470 iscntrl(PL_multi_close)
472 PL_multi_close < 32 || PL_multi_close == 127
476 tmpbuf[1] = (char)toCTRL(PL_multi_close);
481 *tmpbuf = (char)PL_multi_close;
485 q = strchr(s,'"') ? '\'' : '"';
486 Perl_croak(aTHX_ "Can't find string terminator %c%s%c anywhere before EOF",q,s,q);
489 #define FEATURE_IS_ENABLED(name) \
490 ((0 != (PL_hints & HINT_LOCALIZE_HH)) \
491 && S_feature_is_enabled(aTHX_ STR_WITH_LEN(name)))
493 * S_feature_is_enabled
494 * Check whether the named feature is enabled.
497 S_feature_is_enabled(pTHX_ const char *name, STRLEN namelen)
500 HV * const hinthv = GvHV(PL_hintgv);
501 char he_name[32] = "feature_";
502 (void) strncpy(&he_name[8], name, 24);
504 return (hinthv && hv_exists(hinthv, he_name, 8 + namelen));
512 Perl_deprecate(pTHX_ const char *s)
514 if (ckWARN(WARN_DEPRECATED))
515 Perl_warner(aTHX_ packWARN(WARN_DEPRECATED), "Use of %s is deprecated", s);
519 Perl_deprecate_old(pTHX_ const char *s)
521 /* This function should NOT be called for any new deprecated warnings */
522 /* Use Perl_deprecate instead */
524 /* It is here to maintain backward compatibility with the pre-5.8 */
525 /* warnings category hierarchy. The "deprecated" category used to */
526 /* live under the "syntax" category. It is now a top-level category */
527 /* in its own right. */
529 if (ckWARN2(WARN_DEPRECATED, WARN_SYNTAX))
530 Perl_warner(aTHX_ packWARN2(WARN_DEPRECATED, WARN_SYNTAX),
531 "Use of %s is deprecated", s);
535 * experimental text filters for win32 carriage-returns, utf16-to-utf8 and
536 * utf16-to-utf8-reversed.
539 #ifdef PERL_CR_FILTER
543 register const char *s = SvPVX_const(sv);
544 register const char * const e = s + SvCUR(sv);
545 /* outer loop optimized to do nothing if there are no CR-LFs */
547 if (*s++ == '\r' && *s == '\n') {
548 /* hit a CR-LF, need to copy the rest */
549 register char *d = s - 1;
552 if (*s == '\r' && s[1] == '\n')
563 S_cr_textfilter(pTHX_ int idx, SV *sv, int maxlen)
565 const I32 count = FILTER_READ(idx+1, sv, maxlen);
566 if (count > 0 && !maxlen)
574 * Initialize variables. Uses the Perl save_stack to save its state (for
575 * recursive calls to the parser).
579 Perl_lex_start(pTHX_ SV *line)
585 SAVEI32(PL_lex_dojoin);
586 SAVEI32(PL_lex_brackets);
587 SAVEI32(PL_lex_casemods);
588 SAVEI32(PL_lex_starts);
589 SAVEI32(PL_lex_state);
590 SAVEVPTR(PL_lex_inpat);
591 SAVEI32(PL_lex_inwhat);
593 if (PL_lex_state == LEX_KNOWNEXT) {
594 I32 toke = PL_lasttoke;
595 while (--toke >= 0) {
596 SAVEI32(PL_nexttoke[toke].next_type);
597 SAVEVPTR(PL_nexttoke[toke].next_val);
599 SAVEVPTR(PL_nexttoke[toke].next_mad);
601 SAVEI32(PL_lasttoke);
604 SAVESPTR(PL_thistoken);
605 SAVESPTR(PL_thiswhite);
606 SAVESPTR(PL_nextwhite);
607 SAVESPTR(PL_thisopen);
608 SAVESPTR(PL_thisclose);
609 SAVESPTR(PL_thisstuff);
610 SAVEVPTR(PL_thismad);
611 SAVEI32(PL_realtokenstart);
612 SAVEI32(PL_faketokens);
614 SAVEI32(PL_curforce);
616 if (PL_lex_state == LEX_KNOWNEXT) {
617 I32 toke = PL_nexttoke;
618 while (--toke >= 0) {
619 SAVEI32(PL_nexttype[toke]);
620 SAVEVPTR(PL_nextval[toke]);
622 SAVEI32(PL_nexttoke);
625 SAVECOPLINE(PL_curcop);
628 SAVEPPTR(PL_oldbufptr);
629 SAVEPPTR(PL_oldoldbufptr);
630 SAVEPPTR(PL_last_lop);
631 SAVEPPTR(PL_last_uni);
632 SAVEPPTR(PL_linestart);
633 SAVESPTR(PL_linestr);
634 SAVEGENERICPV(PL_lex_brackstack);
635 SAVEGENERICPV(PL_lex_casestack);
636 SAVEDESTRUCTOR_X(restore_rsfp, PL_rsfp);
637 SAVESPTR(PL_lex_stuff);
638 SAVEI32(PL_lex_defer);
639 SAVEI32(PL_sublex_info.sub_inwhat);
640 SAVESPTR(PL_lex_repl);
642 SAVEINT(PL_lex_expect);
644 PL_lex_state = LEX_NORMAL;
648 Newx(PL_lex_brackstack, 120, char);
649 Newx(PL_lex_casestack, 12, char);
651 *PL_lex_casestack = '\0';
663 PL_sublex_info.sub_inwhat = 0;
665 if (SvREADONLY(PL_linestr))
666 PL_linestr = sv_2mortal(newSVsv(PL_linestr));
667 s = SvPV_const(PL_linestr, len);
668 if (!len || s[len-1] != ';') {
669 if (!(SvFLAGS(PL_linestr) & SVs_TEMP))
670 PL_linestr = sv_2mortal(newSVsv(PL_linestr));
671 sv_catpvs(PL_linestr, "\n;");
673 SvTEMP_off(PL_linestr);
674 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = SvPVX(PL_linestr);
675 PL_bufend = PL_bufptr + SvCUR(PL_linestr);
676 PL_last_lop = PL_last_uni = NULL;
682 * Finalizer for lexing operations. Must be called when the parser is
683 * done with the lexer.
690 PL_doextract = FALSE;
695 * This subroutine has nothing to do with tilting, whether at windmills
696 * or pinball tables. Its name is short for "increment line". It
697 * increments the current line number in CopLINE(PL_curcop) and checks
698 * to see whether the line starts with a comment of the form
699 * # line 500 "foo.pm"
700 * If so, it sets the current line number and file to the values in the comment.
704 S_incline(pTHX_ char *s)
712 CopLINE_inc(PL_curcop);
715 while (SPACE_OR_TAB(*s))
717 if (strnEQ(s, "line", 4))
721 if (SPACE_OR_TAB(*s))
725 while (SPACE_OR_TAB(*s))
733 while (SPACE_OR_TAB(*s))
735 if (*s == '"' && (t = strchr(s+1, '"'))) {
740 for (t = s; !isSPACE(*t); t++) ;
743 while (SPACE_OR_TAB(*e) || *e == '\r' || *e == '\f')
745 if (*e != '\n' && *e != '\0')
746 return; /* false alarm */
752 const char * const cf = CopFILE(PL_curcop);
753 STRLEN tmplen = cf ? strlen(cf) : 0;
754 if (tmplen > 7 && strnEQ(cf, "(eval ", 6)) {
755 /* must copy *{"::_<(eval N)[oldfilename:L]"}
756 * to *{"::_<newfilename"} */
757 char smallbuf[256], smallbuf2[256];
758 char *tmpbuf, *tmpbuf2;
760 STRLEN tmplen2 = strlen(s);
761 if (tmplen + 3 < sizeof smallbuf)
764 Newx(tmpbuf, tmplen + 3, char);
765 if (tmplen2 + 3 < sizeof smallbuf2)
768 Newx(tmpbuf2, tmplen2 + 3, char);
769 tmpbuf[0] = tmpbuf2[0] = '_';
770 tmpbuf[1] = tmpbuf2[1] = '<';
771 memcpy(tmpbuf + 2, cf, ++tmplen);
772 memcpy(tmpbuf2 + 2, s, ++tmplen2);
774 gvp = (GV**)hv_fetch(PL_defstash, tmpbuf, tmplen, FALSE);
776 gv2 = *(GV**)hv_fetch(PL_defstash, tmpbuf2, tmplen2, TRUE);
778 gv_init(gv2, PL_defstash, tmpbuf2, tmplen2, FALSE);
779 /* adjust ${"::_<newfilename"} to store the new file name */
780 GvSV(gv2) = newSVpvn(tmpbuf2 + 2, tmplen2 - 2);
781 GvHV(gv2) = (HV*)SvREFCNT_inc(GvHV(*gvp));
782 GvAV(gv2) = (AV*)SvREFCNT_inc(GvAV(*gvp));
784 if (tmpbuf != smallbuf) Safefree(tmpbuf);
785 if (tmpbuf2 != smallbuf2) Safefree(tmpbuf2);
788 CopFILE_free(PL_curcop);
789 CopFILE_set(PL_curcop, s);
792 CopLINE_set(PL_curcop, atoi(n)-1);
796 /* skip space before PL_thistoken */
799 S_skipspace0(pTHX_ register char *s)
806 PL_thiswhite = newSVpvn("",0);
807 sv_catsv(PL_thiswhite, PL_skipwhite);
808 sv_free(PL_skipwhite);
811 PL_realtokenstart = s - SvPVX(PL_linestr);
815 /* skip space after PL_thistoken */
818 S_skipspace1(pTHX_ register char *s)
820 const char *start = s;
821 I32 startoff = start - SvPVX(PL_linestr);
826 start = SvPVX(PL_linestr) + startoff;
827 if (!PL_thistoken && PL_realtokenstart >= 0) {
828 const char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
829 PL_thistoken = newSVpvn(tstart, start - tstart);
831 PL_realtokenstart = -1;
834 PL_nextwhite = newSVpvn("",0);
835 sv_catsv(PL_nextwhite, PL_skipwhite);
836 sv_free(PL_skipwhite);
843 S_skipspace2(pTHX_ register char *s, SV **svp)
846 I32 bufptroff = PL_bufptr - SvPVX(PL_linestr);
847 I32 startoff = start - SvPVX(PL_linestr);
849 PL_bufptr = SvPVX(PL_linestr) + bufptroff;
850 if (!PL_madskills || !svp)
852 start = SvPVX(PL_linestr) + startoff;
853 if (!PL_thistoken && PL_realtokenstart >= 0) {
854 char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
855 PL_thistoken = newSVpvn(tstart, start - tstart);
856 PL_realtokenstart = -1;
860 *svp = newSVpvn("",0);
861 sv_setsv(*svp, PL_skipwhite);
862 sv_free(PL_skipwhite);
872 * Called to gobble the appropriate amount and type of whitespace.
873 * Skips comments as well.
877 S_skipspace(pTHX_ register char *s)
882 int startoff = s - SvPVX(PL_linestr);
885 sv_free(PL_skipwhite);
890 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
891 while (s < PL_bufend && SPACE_OR_TAB(*s))
901 SSize_t oldprevlen, oldoldprevlen;
902 SSize_t oldloplen = 0, oldunilen = 0;
903 while (s < PL_bufend && isSPACE(*s)) {
904 if (*s++ == '\n' && PL_in_eval && !PL_rsfp)
909 if (s < PL_bufend && *s == '#') {
910 while (s < PL_bufend && *s != '\n')
914 if (PL_in_eval && !PL_rsfp) {
921 /* only continue to recharge the buffer if we're at the end
922 * of the buffer, we're not reading from a source filter, and
923 * we're in normal lexing mode
925 if (s < PL_bufend || !PL_rsfp || PL_sublex_info.sub_inwhat ||
926 PL_lex_state == LEX_FORMLINE)
933 /* try to recharge the buffer */
935 curoff = s - SvPVX(PL_linestr);
938 if ((s = filter_gets(PL_linestr, PL_rsfp,
939 (prevlen = SvCUR(PL_linestr)))) == NULL)
942 if (PL_madskills && curoff != startoff) {
944 PL_skipwhite = newSVpvn("",0);
945 sv_catpvn(PL_skipwhite, SvPVX(PL_linestr) + startoff,
949 /* mustn't throw out old stuff yet if madpropping */
950 SvCUR(PL_linestr) = curoff;
951 s = SvPVX(PL_linestr) + curoff;
953 if (curoff && s[-1] == '\n')
957 /* end of file. Add on the -p or -n magic */
958 /* XXX these shouldn't really be added here, can't set PL_faketokens */
962 ";}continue{print or die qq(-p destination: $!\\n);}");
965 ";}continue{print or die qq(-p destination: $!\\n);}");
967 PL_minus_n = PL_minus_p = 0;
969 else if (PL_minus_n) {
971 sv_catpvn(PL_linestr, ";}", 2);
973 sv_setpvn(PL_linestr, ";}", 2);
979 sv_catpvn(PL_linestr,";", 1);
981 sv_setpvn(PL_linestr,";", 1);
984 /* reset variables for next time we lex */
985 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = s = PL_linestart
991 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
992 PL_last_lop = PL_last_uni = NULL;
994 /* Close the filehandle. Could be from -P preprocessor,
995 * STDIN, or a regular file. If we were reading code from
996 * STDIN (because the commandline held no -e or filename)
997 * then we don't close it, we reset it so the code can
998 * read from STDIN too.
1001 if (PL_preprocess && !PL_in_eval)
1002 (void)PerlProc_pclose(PL_rsfp);
1003 else if ((PerlIO*)PL_rsfp == PerlIO_stdin())
1004 PerlIO_clearerr(PL_rsfp);
1006 (void)PerlIO_close(PL_rsfp);
1011 /* not at end of file, so we only read another line */
1012 /* make corresponding updates to old pointers, for yyerror() */
1013 oldprevlen = PL_oldbufptr - PL_bufend;
1014 oldoldprevlen = PL_oldoldbufptr - PL_bufend;
1016 oldunilen = PL_last_uni - PL_bufend;
1018 oldloplen = PL_last_lop - PL_bufend;
1019 PL_linestart = PL_bufptr = s + prevlen;
1020 PL_bufend = s + SvCUR(PL_linestr);
1022 PL_oldbufptr = s + oldprevlen;
1023 PL_oldoldbufptr = s + oldoldprevlen;
1025 PL_last_uni = s + oldunilen;
1027 PL_last_lop = s + oldloplen;
1030 /* debugger active and we're not compiling the debugger code,
1031 * so store the line into the debugger's array of lines
1033 if (PERLDB_LINE && PL_curstash != PL_debstash) {
1034 SV * const sv = newSV(0);
1036 sv_upgrade(sv, SVt_PVMG);
1037 sv_setpvn(sv,PL_bufptr,PL_bufend-PL_bufptr);
1040 av_store(CopFILEAVx(PL_curcop),(I32)CopLINE(PL_curcop),sv);
1048 PL_skipwhite = newSVpvn("",0);
1049 curoff = s - SvPVX(PL_linestr);
1050 if (curoff - startoff)
1051 sv_catpvn(PL_skipwhite, SvPVX(PL_linestr) + startoff,
1060 * Check the unary operators to ensure there's no ambiguity in how they're
1061 * used. An ambiguous piece of code would be:
1063 * This doesn't mean rand() + 5. Because rand() is a unary operator,
1064 * the +5 is its argument.
1074 if (PL_oldoldbufptr != PL_last_uni)
1076 while (isSPACE(*PL_last_uni))
1078 for (s = PL_last_uni; isALNUM_lazy_if(s,UTF) || *s == '-'; s++)
1080 if ((t = strchr(s, '(')) && t < PL_bufptr)
1083 if (ckWARN_d(WARN_AMBIGUOUS)){
1084 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
1085 "Warning: Use of \"%.*s\" without parentheses is ambiguous",
1086 (int)(s - PL_last_uni), PL_last_uni);
1091 * LOP : macro to build a list operator. Its behaviour has been replaced
1092 * with a subroutine, S_lop() for which LOP is just another name.
1095 #define LOP(f,x) return lop(f,x,s)
1099 * Build a list operator (or something that might be one). The rules:
1100 * - if we have a next token, then it's a list operator [why?]
1101 * - if the next thing is an opening paren, then it's a function
1102 * - else it's a list operator
1106 S_lop(pTHX_ I32 f, int x, char *s)
1113 PL_last_lop = PL_oldbufptr;
1114 PL_last_lop_op = (OPCODE)f;
1117 return REPORT(LSTOP);
1120 return REPORT(LSTOP);
1123 return REPORT(FUNC);
1126 return REPORT(FUNC);
1128 return REPORT(LSTOP);
1134 * Sets up for an eventual force_next(). start_force(0) basically does
1135 * an unshift, while start_force(-1) does a push. yylex removes items
1140 S_start_force(pTHX_ int where)
1144 if (where < 0) /* so people can duplicate start_force(PL_curforce) */
1145 where = PL_lasttoke;
1146 assert(PL_curforce < 0 || PL_curforce == where);
1147 if (PL_curforce != where) {
1148 for (i = PL_lasttoke; i > where; --i) {
1149 PL_nexttoke[i] = PL_nexttoke[i-1];
1153 if (PL_curforce < 0) /* in case of duplicate start_force() */
1154 Zero(&PL_nexttoke[where], 1, NEXTTOKE);
1155 PL_curforce = where;
1158 curmad('^', newSVpvn("",0));
1159 CURMAD('_', PL_nextwhite);
1164 S_curmad(pTHX_ char slot, SV *sv)
1170 if (PL_curforce < 0)
1171 where = &PL_thismad;
1173 where = &PL_nexttoke[PL_curforce].next_mad;
1176 sv_setpvn(sv, "", 0);
1179 if (UTF && is_utf8_string((U8*)SvPVX(sv), SvCUR(sv)))
1181 else if (PL_encoding) {
1182 sv_recode_to_utf8(sv, PL_encoding);
1187 /* keep a slot open for the head of the list? */
1188 if (slot != '_' && *where && (*where)->mad_key == '^') {
1189 (*where)->mad_key = slot;
1190 sv_free((*where)->mad_val);
1191 (*where)->mad_val = (void*)sv;
1194 addmad(newMADsv(slot, sv), where, 0);
1197 # define start_force(where) NOOP
1198 # define curmad(slot, sv) NOOP
1203 * When the lexer realizes it knows the next token (for instance,
1204 * it is reordering tokens for the parser) then it can call S_force_next
1205 * to know what token to return the next time the lexer is called. Caller
1206 * will need to set PL_nextval[] (or PL_nexttoke[].next_val with PERL_MAD),
1207 * and possibly PL_expect to ensure the lexer handles the token correctly.
1211 S_force_next(pTHX_ I32 type)
1215 if (PL_curforce < 0)
1216 start_force(PL_lasttoke);
1217 PL_nexttoke[PL_curforce].next_type = type;
1218 if (PL_lex_state != LEX_KNOWNEXT)
1219 PL_lex_defer = PL_lex_state;
1220 PL_lex_state = LEX_KNOWNEXT;
1221 PL_lex_expect = PL_expect;
1224 PL_nexttype[PL_nexttoke] = type;
1226 if (PL_lex_state != LEX_KNOWNEXT) {
1227 PL_lex_defer = PL_lex_state;
1228 PL_lex_expect = PL_expect;
1229 PL_lex_state = LEX_KNOWNEXT;
1235 S_newSV_maybe_utf8(pTHX_ const char *start, STRLEN len)
1238 SV * const sv = newSVpvn(start,len);
1239 if (UTF && !IN_BYTES && is_utf8_string((const U8*)start, len))
1246 * When the lexer knows the next thing is a word (for instance, it has
1247 * just seen -> and it knows that the next char is a word char, then
1248 * it calls S_force_word to stick the next word into the PL_next lookahead.
1251 * char *start : buffer position (must be within PL_linestr)
1252 * int token : PL_next will be this type of bare word (e.g., METHOD,WORD)
1253 * int check_keyword : if true, Perl checks to make sure the word isn't
1254 * a keyword (do this if the word is a label, e.g. goto FOO)
1255 * int allow_pack : if true, : characters will also be allowed (require,
1256 * use, etc. do this)
1257 * int allow_initial_tick : used by the "sub" lexer only.
1261 S_force_word(pTHX_ register char *start, int token, int check_keyword, int allow_pack, int allow_initial_tick)
1267 start = SKIPSPACE1(start);
1269 if (isIDFIRST_lazy_if(s,UTF) ||
1270 (allow_pack && *s == ':') ||
1271 (allow_initial_tick && *s == '\'') )
1273 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, allow_pack, &len);
1274 if (check_keyword && keyword(PL_tokenbuf, len))
1276 start_force(PL_curforce);
1278 curmad('X', newSVpvn(start,s-start));
1279 if (token == METHOD) {
1284 PL_expect = XOPERATOR;
1287 NEXTVAL_NEXTTOKE.opval
1288 = (OP*)newSVOP(OP_CONST,0,
1289 S_newSV_maybe_utf8(aTHX_ PL_tokenbuf, len));
1290 NEXTVAL_NEXTTOKE.opval->op_private |= OPpCONST_BARE;
1298 * Called when the lexer wants $foo *foo &foo etc, but the program
1299 * text only contains the "foo" portion. The first argument is a pointer
1300 * to the "foo", and the second argument is the type symbol to prefix.
1301 * Forces the next token to be a "WORD".
1302 * Creates the symbol if it didn't already exist (via gv_fetchpv()).
1306 S_force_ident(pTHX_ register const char *s, int kind)
1310 const STRLEN len = strlen(s);
1311 OP* const o = (OP*)newSVOP(OP_CONST, 0, newSVpvn(s, len));
1312 start_force(PL_curforce);
1313 NEXTVAL_NEXTTOKE.opval = o;
1316 o->op_private = OPpCONST_ENTERED;
1317 /* XXX see note in pp_entereval() for why we forgo typo
1318 warnings if the symbol must be introduced in an eval.
1320 gv_fetchpvn_flags(s, len,
1321 PL_in_eval ? (GV_ADDMULTI | GV_ADDINEVAL)
1323 kind == '$' ? SVt_PV :
1324 kind == '@' ? SVt_PVAV :
1325 kind == '%' ? SVt_PVHV :
1333 Perl_str_to_version(pTHX_ SV *sv)
1338 const char *start = SvPV_const(sv,len);
1339 const char * const end = start + len;
1340 const bool utf = SvUTF8(sv) ? TRUE : FALSE;
1341 while (start < end) {
1345 n = utf8n_to_uvchr((U8*)start, len, &skip, 0);
1350 retval += ((NV)n)/nshift;
1359 * Forces the next token to be a version number.
1360 * If the next token appears to be an invalid version number, (e.g. "v2b"),
1361 * and if "guessing" is TRUE, then no new token is created (and the caller
1362 * must use an alternative parsing method).
1366 S_force_version(pTHX_ char *s, int guessing)
1372 I32 startoff = s - SvPVX(PL_linestr);
1381 while (isDIGIT(*d) || *d == '_' || *d == '.')
1385 start_force(PL_curforce);
1386 curmad('X', newSVpvn(s,d-s));
1389 if (*d == ';' || isSPACE(*d) || *d == '}' || !*d) {
1391 s = scan_num(s, &yylval);
1392 version = yylval.opval;
1393 ver = cSVOPx(version)->op_sv;
1394 if (SvPOK(ver) && !SvNIOK(ver)) {
1395 SvUPGRADE(ver, SVt_PVNV);
1396 SvNV_set(ver, str_to_version(ver));
1397 SvNOK_on(ver); /* hint that it is a version */
1400 else if (guessing) {
1403 sv_free(PL_nextwhite); /* let next token collect whitespace */
1405 s = SvPVX(PL_linestr) + startoff;
1413 if (PL_madskills && !version) {
1414 sv_free(PL_nextwhite); /* let next token collect whitespace */
1416 s = SvPVX(PL_linestr) + startoff;
1419 /* NOTE: The parser sees the package name and the VERSION swapped */
1420 start_force(PL_curforce);
1421 NEXTVAL_NEXTTOKE.opval = version;
1429 * Tokenize a quoted string passed in as an SV. It finds the next
1430 * chunk, up to end of string or a backslash. It may make a new
1431 * SV containing that chunk (if HINT_NEW_STRING is on). It also
1436 S_tokeq(pTHX_ SV *sv)
1440 register char *send;
1448 s = SvPV_force(sv, len);
1449 if (SvTYPE(sv) >= SVt_PVIV && SvIVX(sv) == -1)
1452 while (s < send && *s != '\\')
1457 if ( PL_hints & HINT_NEW_STRING ) {
1458 pv = sv_2mortal(newSVpvn(SvPVX_const(pv), len));
1464 if (s + 1 < send && (s[1] == '\\'))
1465 s++; /* all that, just for this */
1470 SvCUR_set(sv, d - SvPVX_const(sv));
1472 if ( PL_hints & HINT_NEW_STRING )
1473 return new_constant(NULL, 0, "q", sv, pv, "q");
1478 * Now come three functions related to double-quote context,
1479 * S_sublex_start, S_sublex_push, and S_sublex_done. They're used when
1480 * converting things like "\u\Lgnat" into ucfirst(lc("gnat")). They
1481 * interact with PL_lex_state, and create fake ( ... ) argument lists
1482 * to handle functions and concatenation.
1483 * They assume that whoever calls them will be setting up a fake
1484 * join call, because each subthing puts a ',' after it. This lets
1487 * join($, , 'lower ', lcfirst( 'uPpEr', ) ,)
1489 * (I'm not sure whether the spurious commas at the end of lcfirst's
1490 * arguments and join's arguments are created or not).
1495 * Assumes that yylval.ival is the op we're creating (e.g. OP_LCFIRST).
1497 * Pattern matching will set PL_lex_op to the pattern-matching op to
1498 * make (we return THING if yylval.ival is OP_NULL, PMFUNC otherwise).
1500 * OP_CONST and OP_READLINE are easy--just make the new op and return.
1502 * Everything else becomes a FUNC.
1504 * Sets PL_lex_state to LEX_INTERPPUSH unless (ival was OP_NULL or we
1505 * had an OP_CONST or OP_READLINE). This just sets us up for a
1506 * call to S_sublex_push().
1510 S_sublex_start(pTHX)
1513 register const I32 op_type = yylval.ival;
1515 if (op_type == OP_NULL) {
1516 yylval.opval = PL_lex_op;
1520 if (op_type == OP_CONST || op_type == OP_READLINE) {
1521 SV *sv = tokeq(PL_lex_stuff);
1523 if (SvTYPE(sv) == SVt_PVIV) {
1524 /* Overloaded constants, nothing fancy: Convert to SVt_PV: */
1526 const char * const p = SvPV_const(sv, len);
1527 SV * const nsv = newSVpvn(p, len);
1533 yylval.opval = (OP*)newSVOP(op_type, 0, sv);
1534 PL_lex_stuff = NULL;
1535 /* Allow <FH> // "foo" */
1536 if (op_type == OP_READLINE)
1537 PL_expect = XTERMORDORDOR;
1541 PL_sublex_info.super_state = PL_lex_state;
1542 PL_sublex_info.sub_inwhat = op_type;
1543 PL_sublex_info.sub_op = PL_lex_op;
1544 PL_lex_state = LEX_INTERPPUSH;
1548 yylval.opval = PL_lex_op;
1558 * Create a new scope to save the lexing state. The scope will be
1559 * ended in S_sublex_done. Returns a '(', starting the function arguments
1560 * to the uc, lc, etc. found before.
1561 * Sets PL_lex_state to LEX_INTERPCONCAT.
1570 PL_lex_state = PL_sublex_info.super_state;
1571 SAVEI32(PL_lex_dojoin);
1572 SAVEI32(PL_lex_brackets);
1573 SAVEI32(PL_lex_casemods);
1574 SAVEI32(PL_lex_starts);
1575 SAVEI32(PL_lex_state);
1576 SAVEVPTR(PL_lex_inpat);
1577 SAVEI32(PL_lex_inwhat);
1578 SAVECOPLINE(PL_curcop);
1579 SAVEPPTR(PL_bufptr);
1580 SAVEPPTR(PL_bufend);
1581 SAVEPPTR(PL_oldbufptr);
1582 SAVEPPTR(PL_oldoldbufptr);
1583 SAVEPPTR(PL_last_lop);
1584 SAVEPPTR(PL_last_uni);
1585 SAVEPPTR(PL_linestart);
1586 SAVESPTR(PL_linestr);
1587 SAVEGENERICPV(PL_lex_brackstack);
1588 SAVEGENERICPV(PL_lex_casestack);
1590 PL_linestr = PL_lex_stuff;
1591 PL_lex_stuff = NULL;
1593 PL_bufend = PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart
1594 = SvPVX(PL_linestr);
1595 PL_bufend += SvCUR(PL_linestr);
1596 PL_last_lop = PL_last_uni = NULL;
1597 SAVEFREESV(PL_linestr);
1599 PL_lex_dojoin = FALSE;
1600 PL_lex_brackets = 0;
1601 Newx(PL_lex_brackstack, 120, char);
1602 Newx(PL_lex_casestack, 12, char);
1603 PL_lex_casemods = 0;
1604 *PL_lex_casestack = '\0';
1606 PL_lex_state = LEX_INTERPCONCAT;
1607 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
1609 PL_lex_inwhat = PL_sublex_info.sub_inwhat;
1610 if (PL_lex_inwhat == OP_MATCH || PL_lex_inwhat == OP_QR || PL_lex_inwhat == OP_SUBST)
1611 PL_lex_inpat = PL_sublex_info.sub_op;
1613 PL_lex_inpat = NULL;
1620 * Restores lexer state after a S_sublex_push.
1627 if (!PL_lex_starts++) {
1628 SV * const sv = newSVpvs("");
1629 if (SvUTF8(PL_linestr))
1631 PL_expect = XOPERATOR;
1632 yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
1636 if (PL_lex_casemods) { /* oops, we've got some unbalanced parens */
1637 PL_lex_state = LEX_INTERPCASEMOD;
1641 /* Is there a right-hand side to take care of? (s//RHS/ or tr//RHS/) */
1642 if (PL_lex_repl && (PL_lex_inwhat == OP_SUBST || PL_lex_inwhat == OP_TRANS)) {
1643 PL_linestr = PL_lex_repl;
1645 PL_bufend = PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart = SvPVX(PL_linestr);
1646 PL_bufend += SvCUR(PL_linestr);
1647 PL_last_lop = PL_last_uni = NULL;
1648 SAVEFREESV(PL_linestr);
1649 PL_lex_dojoin = FALSE;
1650 PL_lex_brackets = 0;
1651 PL_lex_casemods = 0;
1652 *PL_lex_casestack = '\0';
1654 if (SvEVALED(PL_lex_repl)) {
1655 PL_lex_state = LEX_INTERPNORMAL;
1657 /* we don't clear PL_lex_repl here, so that we can check later
1658 whether this is an evalled subst; that means we rely on the
1659 logic to ensure sublex_done() is called again only via the
1660 branch (in yylex()) that clears PL_lex_repl, else we'll loop */
1663 PL_lex_state = LEX_INTERPCONCAT;
1673 PL_endwhite = newSVpvn("",0);
1674 sv_catsv(PL_endwhite, PL_thiswhite);
1678 sv_setpvn(PL_thistoken,"",0);
1680 PL_realtokenstart = -1;
1684 PL_bufend = SvPVX(PL_linestr);
1685 PL_bufend += SvCUR(PL_linestr);
1686 PL_expect = XOPERATOR;
1687 PL_sublex_info.sub_inwhat = 0;
1695 Extracts a pattern, double-quoted string, or transliteration. This
1698 It looks at lex_inwhat and PL_lex_inpat to find out whether it's
1699 processing a pattern (PL_lex_inpat is true), a transliteration
1700 (lex_inwhat & OP_TRANS is true), or a double-quoted string.
1702 Returns a pointer to the character scanned up to. Iff this is
1703 advanced from the start pointer supplied (ie if anything was
1704 successfully parsed), will leave an OP for the substring scanned
1705 in yylval. Caller must intuit reason for not parsing further
1706 by looking at the next characters herself.
1710 double-quoted style: \r and \n
1711 regexp special ones: \D \s
1713 backrefs: \1 (deprecated in substitution replacements)
1714 case and quoting: \U \Q \E
1715 stops on @ and $, but not for $ as tail anchor
1717 In transliterations:
1718 characters are VERY literal, except for - not at the start or end
1719 of the string, which indicates a range. scan_const expands the
1720 range to the full set of intermediate characters.
1722 In double-quoted strings:
1724 double-quoted style: \r and \n
1726 backrefs: \1 (deprecated)
1727 case and quoting: \U \Q \E
1730 scan_const does *not* construct ops to handle interpolated strings.
1731 It stops processing as soon as it finds an embedded $ or @ variable
1732 and leaves it to the caller to work out what's going on.
1734 @ in pattern could be: @foo, @{foo}, @$foo, @'foo, @::foo.
1736 $ in pattern could be $foo or could be tail anchor. Assumption:
1737 it's a tail anchor if $ is the last thing in the string, or if it's
1738 followed by one of ")| \n\t"
1740 \1 (backreferences) are turned into $1
1742 The structure of the code is
1743 while (there's a character to process) {
1744 handle transliteration ranges
1745 skip regexp comments
1746 skip # initiated comments in //x patterns
1747 check for embedded @foo
1748 check for embedded scalars
1750 leave intact backslashes from leave (below)
1751 deprecate \1 in strings and sub replacements
1752 handle string-changing backslashes \l \U \Q \E, etc.
1753 switch (what was escaped) {
1754 handle - in a transliteration (becomes a literal -)
1755 handle \132 octal characters
1756 handle 0x15 hex characters
1757 handle \cV (control V)
1758 handle printf backslashes (\f, \r, \n, etc)
1760 } (end if backslash)
1761 } (end while character to read)
1766 S_scan_const(pTHX_ char *start)
1769 register char *send = PL_bufend; /* end of the constant */
1770 SV *sv = newSV(send - start); /* sv for the constant */
1771 register char *s = start; /* start of the constant */
1772 register char *d = SvPVX(sv); /* destination for copies */
1773 bool dorange = FALSE; /* are we in a translit range? */
1774 bool didrange = FALSE; /* did we just finish a range? */
1775 I32 has_utf8 = FALSE; /* Output constant is UTF8 */
1776 I32 this_utf8 = UTF; /* The source string is assumed to be UTF8 */
1779 UV literal_endpoint = 0;
1782 const char * const leaveit = /* set of acceptably-backslashed characters */
1784 ? "\\.^$@AGZdDwWsSbBpPXC+*?|()-nrtfeaxz0123456789[{]} \t\n\r\f\v#"
1787 if (PL_lex_inwhat == OP_TRANS && PL_sublex_info.sub_op) {
1788 /* If we are doing a trans and we know we want UTF8 set expectation */
1789 has_utf8 = PL_sublex_info.sub_op->op_private & (OPpTRANS_FROM_UTF|OPpTRANS_TO_UTF);
1790 this_utf8 = PL_sublex_info.sub_op->op_private & (PL_lex_repl ? OPpTRANS_FROM_UTF : OPpTRANS_TO_UTF);
1794 while (s < send || dorange) {
1795 /* get transliterations out of the way (they're most literal) */
1796 if (PL_lex_inwhat == OP_TRANS) {
1797 /* expand a range A-Z to the full set of characters. AIE! */
1799 I32 i; /* current expanded character */
1800 I32 min; /* first character in range */
1801 I32 max; /* last character in range */
1804 char * const c = (char*)utf8_hop((U8*)d, -1);
1808 *c = (char)UTF_TO_NATIVE(0xff);
1809 /* mark the range as done, and continue */
1815 i = d - SvPVX_const(sv); /* remember current offset */
1816 SvGROW(sv, SvLEN(sv) + 256); /* never more than 256 chars in a range */
1817 d = SvPVX(sv) + i; /* refresh d after realloc */
1818 d -= 2; /* eat the first char and the - */
1820 min = (U8)*d; /* first char in range */
1821 max = (U8)d[1]; /* last char in range */
1825 "Invalid range \"%c-%c\" in transliteration operator",
1826 (char)min, (char)max);
1830 if (literal_endpoint == 2 &&
1831 ((isLOWER(min) && isLOWER(max)) ||
1832 (isUPPER(min) && isUPPER(max)))) {
1834 for (i = min; i <= max; i++)
1836 *d++ = NATIVE_TO_NEED(has_utf8,i);
1838 for (i = min; i <= max; i++)
1840 *d++ = NATIVE_TO_NEED(has_utf8,i);
1845 for (i = min; i <= max; i++)
1848 /* mark the range as done, and continue */
1852 literal_endpoint = 0;
1857 /* range begins (ignore - as first or last char) */
1858 else if (*s == '-' && s+1 < send && s != start) {
1860 Perl_croak(aTHX_ "Ambiguous range in transliteration operator");
1863 *d++ = (char)UTF_TO_NATIVE(0xff); /* use illegal utf8 byte--see pmtrans */
1873 literal_endpoint = 0;
1878 /* if we get here, we're not doing a transliteration */
1880 /* skip for regexp comments /(?#comment)/ and code /(?{code})/,
1881 except for the last char, which will be done separately. */
1882 else if (*s == '(' && PL_lex_inpat && s[1] == '?') {
1884 while (s+1 < send && *s != ')')
1885 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
1887 else if (s[2] == '{' /* This should match regcomp.c */
1888 || ((s[2] == 'p' || s[2] == '?') && s[3] == '{'))
1891 char *regparse = s + (s[2] == '{' ? 3 : 4);
1894 while (count && (c = *regparse)) {
1895 if (c == '\\' && regparse[1])
1903 if (*regparse != ')')
1904 regparse--; /* Leave one char for continuation. */
1905 while (s < regparse)
1906 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
1910 /* likewise skip #-initiated comments in //x patterns */
1911 else if (*s == '#' && PL_lex_inpat &&
1912 ((PMOP*)PL_lex_inpat)->op_pmflags & PMf_EXTENDED) {
1913 while (s+1 < send && *s != '\n')
1914 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
1917 /* check for embedded arrays
1918 (@foo, @::foo, @'foo, @{foo}, @$foo, @+, @-)
1920 else if (*s == '@' && s[1]
1921 && (isALNUM_lazy_if(s+1,UTF) || strchr(":'{$+-", s[1])))
1924 /* check for embedded scalars. only stop if we're sure it's a
1927 else if (*s == '$') {
1928 if (!PL_lex_inpat) /* not a regexp, so $ must be var */
1930 if (s + 1 < send && !strchr("()| \r\n\t", s[1]))
1931 break; /* in regexp, $ might be tail anchor */
1934 /* End of else if chain - OP_TRANS rejoin rest */
1937 if (*s == '\\' && s+1 < send) {
1940 /* some backslashes we leave behind */
1941 if (*leaveit && *s && strchr(leaveit, *s)) {
1942 *d++ = NATIVE_TO_NEED(has_utf8,'\\');
1943 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
1947 /* deprecate \1 in strings and substitution replacements */
1948 if (PL_lex_inwhat == OP_SUBST && !PL_lex_inpat &&
1949 isDIGIT(*s) && *s != '0' && !isDIGIT(s[1]))
1951 if (ckWARN(WARN_SYNTAX))
1952 Perl_warner(aTHX_ packWARN(WARN_SYNTAX), "\\%c better written as $%c", *s, *s);
1957 /* string-change backslash escapes */
1958 if (PL_lex_inwhat != OP_TRANS && *s && strchr("lLuUEQ", *s)) {
1963 /* if we get here, it's either a quoted -, or a digit */
1966 /* quoted - in transliterations */
1968 if (PL_lex_inwhat == OP_TRANS) {
1978 Perl_warner(aTHX_ packWARN(WARN_MISC),
1979 "Unrecognized escape \\%c passed through",
1981 /* default action is to copy the quoted character */
1982 goto default_action;
1985 /* \132 indicates an octal constant */
1986 case '0': case '1': case '2': case '3':
1987 case '4': case '5': case '6': case '7':
1991 uv = grok_oct(s, &len, &flags, NULL);
1994 goto NUM_ESCAPE_INSERT;
1996 /* \x24 indicates a hex constant */
2000 char* const e = strchr(s, '}');
2001 I32 flags = PERL_SCAN_ALLOW_UNDERSCORES |
2002 PERL_SCAN_DISALLOW_PREFIX;
2007 yyerror("Missing right brace on \\x{}");
2011 uv = grok_hex(s, &len, &flags, NULL);
2017 I32 flags = PERL_SCAN_DISALLOW_PREFIX;
2018 uv = grok_hex(s, &len, &flags, NULL);
2024 /* Insert oct or hex escaped character.
2025 * There will always enough room in sv since such
2026 * escapes will be longer than any UTF-8 sequence
2027 * they can end up as. */
2029 /* We need to map to chars to ASCII before doing the tests
2032 if (!UNI_IS_INVARIANT(NATIVE_TO_UNI(uv))) {
2033 if (!has_utf8 && uv > 255) {
2034 /* Might need to recode whatever we have
2035 * accumulated so far if it contains any
2038 * (Can't we keep track of that and avoid
2039 * this rescan? --jhi)
2043 for (c = (U8 *) SvPVX(sv); c < (U8 *)d; c++) {
2044 if (!NATIVE_IS_INVARIANT(*c)) {
2049 const STRLEN offset = d - SvPVX_const(sv);
2051 d = SvGROW(sv, SvLEN(sv) + hicount + 1) + offset;
2055 while (src >= (const U8 *)SvPVX_const(sv)) {
2056 if (!NATIVE_IS_INVARIANT(*src)) {
2057 const U8 ch = NATIVE_TO_ASCII(*src);
2058 *dst-- = (U8)UTF8_EIGHT_BIT_LO(ch);
2059 *dst-- = (U8)UTF8_EIGHT_BIT_HI(ch);
2069 if (has_utf8 || uv > 255) {
2070 d = (char*)uvchr_to_utf8((U8*)d, uv);
2072 if (PL_lex_inwhat == OP_TRANS &&
2073 PL_sublex_info.sub_op) {
2074 PL_sublex_info.sub_op->op_private |=
2075 (PL_lex_repl ? OPpTRANS_FROM_UTF
2088 /* \N{LATIN SMALL LETTER A} is a named character */
2092 char* e = strchr(s, '}');
2098 yyerror("Missing right brace on \\N{}");
2102 if (e > s + 2 && s[1] == 'U' && s[2] == '+') {
2104 I32 flags = PERL_SCAN_ALLOW_UNDERSCORES |
2105 PERL_SCAN_DISALLOW_PREFIX;
2108 uv = grok_hex(s, &len, &flags, NULL);
2110 goto NUM_ESCAPE_INSERT;
2112 res = newSVpvn(s + 1, e - s - 1);
2113 res = new_constant( NULL, 0, "charnames",
2114 res, NULL, "\\N{...}" );
2116 sv_utf8_upgrade(res);
2117 str = SvPV_const(res,len);
2118 #ifdef EBCDIC_NEVER_MIND
2119 /* charnames uses pack U and that has been
2120 * recently changed to do the below uni->native
2121 * mapping, so this would be redundant (and wrong,
2122 * the code point would be doubly converted).
2123 * But leave this in just in case the pack U change
2124 * gets revoked, but the semantics is still
2125 * desireable for charnames. --jhi */
2127 UV uv = utf8_to_uvchr((const U8*)str, 0);
2130 U8 tmpbuf[UTF8_MAXBYTES+1], *d;
2132 d = uvchr_to_utf8(tmpbuf, UNI_TO_NATIVE(uv));
2133 sv_setpvn(res, (char *)tmpbuf, d - tmpbuf);
2134 str = SvPV_const(res, len);
2138 if (!has_utf8 && SvUTF8(res)) {
2139 const char * const ostart = SvPVX_const(sv);
2140 SvCUR_set(sv, d - ostart);
2143 sv_utf8_upgrade(sv);
2144 /* this just broke our allocation above... */
2145 SvGROW(sv, (STRLEN)(send - start));
2146 d = SvPVX(sv) + SvCUR(sv);
2149 if (len > (STRLEN)(e - s + 4)) { /* I _guess_ 4 is \N{} --jhi */
2150 const char * const odest = SvPVX_const(sv);
2152 SvGROW(sv, (SvLEN(sv) + len - (e - s + 4)));
2153 d = SvPVX(sv) + (d - odest);
2155 Copy(str, d, len, char);
2162 yyerror("Missing braces on \\N{}");
2165 /* \c is a control character */
2174 *d++ = NATIVE_TO_NEED(has_utf8,toCTRL(c));
2177 yyerror("Missing control char name in \\c");
2181 /* printf-style backslashes, formfeeds, newlines, etc */
2183 *d++ = NATIVE_TO_NEED(has_utf8,'\b');
2186 *d++ = NATIVE_TO_NEED(has_utf8,'\n');
2189 *d++ = NATIVE_TO_NEED(has_utf8,'\r');
2192 *d++ = NATIVE_TO_NEED(has_utf8,'\f');
2195 *d++ = NATIVE_TO_NEED(has_utf8,'\t');
2198 *d++ = ASCII_TO_NEED(has_utf8,'\033');
2201 *d++ = ASCII_TO_NEED(has_utf8,'\007');
2207 } /* end if (backslash) */
2214 /* If we started with encoded form, or already know we want it
2215 and then encode the next character */
2216 if ((has_utf8 || this_utf8) && !NATIVE_IS_INVARIANT((U8)(*s))) {
2218 const UV nextuv = (this_utf8) ? utf8n_to_uvchr((U8*)s, send - s, &len, 0) : (UV) ((U8) *s);
2219 const STRLEN need = UNISKIP(NATIVE_TO_UNI(nextuv));
2222 /* encoded value larger than old, need extra space (NOTE: SvCUR() not set here) */
2223 const STRLEN off = d - SvPVX_const(sv);
2224 d = SvGROW(sv, SvLEN(sv) + (need-len)) + off;
2226 d = (char*)uvchr_to_utf8((U8*)d, nextuv);
2230 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2232 } /* while loop to process each character */
2234 /* terminate the string and set up the sv */
2236 SvCUR_set(sv, d - SvPVX_const(sv));
2237 if (SvCUR(sv) >= SvLEN(sv))
2238 Perl_croak(aTHX_ "panic: constant overflowed allocated space");
2241 if (PL_encoding && !has_utf8) {
2242 sv_recode_to_utf8(sv, PL_encoding);
2248 if (PL_lex_inwhat == OP_TRANS && PL_sublex_info.sub_op) {
2249 PL_sublex_info.sub_op->op_private |=
2250 (PL_lex_repl ? OPpTRANS_FROM_UTF : OPpTRANS_TO_UTF);
2254 /* shrink the sv if we allocated more than we used */
2255 if (SvCUR(sv) + 5 < SvLEN(sv)) {
2256 SvPV_shrink_to_cur(sv);
2259 /* return the substring (via yylval) only if we parsed anything */
2260 if (s > PL_bufptr) {
2261 if ( PL_hints & ( PL_lex_inpat ? HINT_NEW_RE : HINT_NEW_STRING ) )
2262 sv = new_constant(start, s - start, (PL_lex_inpat ? "qr" : "q"),
2264 ( PL_lex_inwhat == OP_TRANS
2266 : ( (PL_lex_inwhat == OP_SUBST && !PL_lex_inpat)
2269 yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
2276 * Returns TRUE if there's more to the expression (e.g., a subscript),
2279 * It deals with "$foo[3]" and /$foo[3]/ and /$foo[0123456789$]+/
2281 * ->[ and ->{ return TRUE
2282 * { and [ outside a pattern are always subscripts, so return TRUE
2283 * if we're outside a pattern and it's not { or [, then return FALSE
2284 * if we're in a pattern and the first char is a {
2285 * {4,5} (any digits around the comma) returns FALSE
2286 * if we're in a pattern and the first char is a [
2288 * [SOMETHING] has a funky algorithm to decide whether it's a
2289 * character class or not. It has to deal with things like
2290 * /$foo[-3]/ and /$foo[$bar]/ as well as /$foo[$\d]+/
2291 * anything else returns TRUE
2294 /* This is the one truly awful dwimmer necessary to conflate C and sed. */
2297 S_intuit_more(pTHX_ register char *s)
2300 if (PL_lex_brackets)
2302 if (*s == '-' && s[1] == '>' && (s[2] == '[' || s[2] == '{'))
2304 if (*s != '{' && *s != '[')
2309 /* In a pattern, so maybe we have {n,m}. */
2326 /* On the other hand, maybe we have a character class */
2329 if (*s == ']' || *s == '^')
2332 /* this is terrifying, and it works */
2333 int weight = 2; /* let's weigh the evidence */
2335 unsigned char un_char = 255, last_un_char;
2336 const char * const send = strchr(s,']');
2337 char tmpbuf[sizeof PL_tokenbuf * 4];
2339 if (!send) /* has to be an expression */
2342 Zero(seen,256,char);
2345 else if (isDIGIT(*s)) {
2347 if (isDIGIT(s[1]) && s[2] == ']')
2353 for (; s < send; s++) {
2354 last_un_char = un_char;
2355 un_char = (unsigned char)*s;
2360 weight -= seen[un_char] * 10;
2361 if (isALNUM_lazy_if(s+1,UTF)) {
2363 scan_ident(s, send, tmpbuf, sizeof tmpbuf, FALSE);
2364 len = (int)strlen(tmpbuf);
2365 if (len > 1 && gv_fetchpvn_flags(tmpbuf, len, 0, SVt_PV))
2370 else if (*s == '$' && s[1] &&
2371 strchr("[#!%*<>()-=",s[1])) {
2372 if (/*{*/ strchr("])} =",s[2]))
2381 if (strchr("wds]",s[1]))
2383 else if (seen['\''] || seen['"'])
2385 else if (strchr("rnftbxcav",s[1]))
2387 else if (isDIGIT(s[1])) {
2389 while (s[1] && isDIGIT(s[1]))
2399 if (strchr("aA01! ",last_un_char))
2401 if (strchr("zZ79~",s[1]))
2403 if (last_un_char == 255 && (isDIGIT(s[1]) || s[1] == '$'))
2404 weight -= 5; /* cope with negative subscript */
2407 if (!isALNUM(last_un_char)
2408 && !(last_un_char == '$' || last_un_char == '@'
2409 || last_un_char == '&')
2410 && isALPHA(*s) && s[1] && isALPHA(s[1])) {
2415 if (keyword(tmpbuf, d - tmpbuf))
2418 if (un_char == last_un_char + 1)
2420 weight -= seen[un_char];
2425 if (weight >= 0) /* probably a character class */
2435 * Does all the checking to disambiguate
2437 * between foo(bar) and bar->foo. Returns 0 if not a method, otherwise
2438 * FUNCMETH (bar->foo(args)) or METHOD (bar->foo args).
2440 * First argument is the stuff after the first token, e.g. "bar".
2442 * Not a method if bar is a filehandle.
2443 * Not a method if foo is a subroutine prototyped to take a filehandle.
2444 * Not a method if it's really "Foo $bar"
2445 * Method if it's "foo $bar"
2446 * Not a method if it's really "print foo $bar"
2447 * Method if it's really "foo package::" (interpreted as package->foo)
2448 * Not a method if bar is known to be a subroutine ("sub bar; foo bar")
2449 * Not a method if bar is a filehandle or package, but is quoted with
2454 S_intuit_method(pTHX_ char *start, GV *gv, CV *cv)
2457 char *s = start + (*start == '$');
2458 char tmpbuf[sizeof PL_tokenbuf];
2466 if (SvTYPE(gv) == SVt_PVGV && GvIO(gv))
2470 const char *proto = SvPVX_const(cv);
2481 s = scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
2482 /* start is the beginning of the possible filehandle/object,
2483 * and s is the end of it
2484 * tmpbuf is a copy of it
2487 if (*start == '$') {
2488 if (gv || PL_last_lop_op == OP_PRINT || isUPPER(*PL_tokenbuf))
2491 len = start - SvPVX(PL_linestr);
2495 start = SvPVX(PL_linestr) + len;
2499 return *s == '(' ? FUNCMETH : METHOD;
2501 if (!keyword(tmpbuf, len)) {
2502 if (len > 2 && tmpbuf[len - 2] == ':' && tmpbuf[len - 1] == ':') {
2506 soff = s - SvPVX(PL_linestr);
2510 indirgv = gv_fetchpvn_flags(tmpbuf, len, 0, SVt_PVCV);
2511 if (indirgv && GvCVu(indirgv))
2513 /* filehandle or package name makes it a method */
2514 if (!gv || GvIO(indirgv) || gv_stashpvn(tmpbuf, len, FALSE)) {
2516 soff = s - SvPVX(PL_linestr);
2519 if ((PL_bufend - s) >= 2 && *s == '=' && *(s+1) == '>')
2520 return 0; /* no assumptions -- "=>" quotes bearword */
2522 start_force(PL_curforce);
2523 NEXTVAL_NEXTTOKE.opval = (OP*)newSVOP(OP_CONST, 0,
2524 newSVpvn(tmpbuf,len));
2525 NEXTVAL_NEXTTOKE.opval->op_private = OPpCONST_BARE;
2527 curmad('X', newSVpvn(start,SvPVX(PL_linestr) + soff - start));
2532 PL_bufptr = SvPVX(PL_linestr) + soff; /* restart before space */
2534 return *s == '(' ? FUNCMETH : METHOD;
2542 * Return a string of Perl code to load the debugger. If PERL5DB
2543 * is set, it will return the contents of that, otherwise a
2544 * compile-time require of perl5db.pl.
2552 const char * const pdb = PerlEnv_getenv("PERL5DB");
2556 SETERRNO(0,SS_NORMAL);
2557 return "BEGIN { require 'perl5db.pl' }";
2563 /* Encoded script support. filter_add() effectively inserts a
2564 * 'pre-processing' function into the current source input stream.
2565 * Note that the filter function only applies to the current source file
2566 * (e.g., it will not affect files 'require'd or 'use'd by this one).
2568 * The datasv parameter (which may be NULL) can be used to pass
2569 * private data to this instance of the filter. The filter function
2570 * can recover the SV using the FILTER_DATA macro and use it to
2571 * store private buffers and state information.
2573 * The supplied datasv parameter is upgraded to a PVIO type
2574 * and the IoDIRP/IoANY field is used to store the function pointer,
2575 * and IOf_FAKE_DIRP is enabled on datasv to mark this as such.
2576 * Note that IoTOP_NAME, IoFMT_NAME, IoBOTTOM_NAME, if set for
2577 * private use must be set using malloc'd pointers.
2581 Perl_filter_add(pTHX_ filter_t funcp, SV *datasv)
2587 if (!PL_rsfp_filters)
2588 PL_rsfp_filters = newAV();
2591 SvUPGRADE(datasv, SVt_PVIO);
2592 IoANY(datasv) = FPTR2DPTR(void *, funcp); /* stash funcp into spare field */
2593 IoFLAGS(datasv) |= IOf_FAKE_DIRP;
2594 DEBUG_P(PerlIO_printf(Perl_debug_log, "filter_add func %p (%s)\n",
2595 FPTR2DPTR(void *, IoANY(datasv)),
2596 SvPV_nolen(datasv)));
2597 av_unshift(PL_rsfp_filters, 1);
2598 av_store(PL_rsfp_filters, 0, datasv) ;
2603 /* Delete most recently added instance of this filter function. */
2605 Perl_filter_del(pTHX_ filter_t funcp)
2611 DEBUG_P(PerlIO_printf(Perl_debug_log, "filter_del func %p",
2612 FPTR2DPTR(void*, funcp)));
2614 if (!PL_rsfp_filters || AvFILLp(PL_rsfp_filters)<0)
2616 /* if filter is on top of stack (usual case) just pop it off */
2617 datasv = FILTER_DATA(AvFILLp(PL_rsfp_filters));
2618 if (IoANY(datasv) == FPTR2DPTR(void *, funcp)) {
2619 IoFLAGS(datasv) &= ~IOf_FAKE_DIRP;
2620 IoANY(datasv) = (void *)NULL;
2621 sv_free(av_pop(PL_rsfp_filters));
2625 /* we need to search for the correct entry and clear it */
2626 Perl_die(aTHX_ "filter_del can only delete in reverse order (currently)");
2630 /* Invoke the idxth filter function for the current rsfp. */
2631 /* maxlen 0 = read one text line */
2633 Perl_filter_read(pTHX_ int idx, SV *buf_sv, int maxlen)
2638 /* This API is bad. It should have been using unsigned int for maxlen.
2639 Not sure if we want to change the API, but if not we should sanity
2640 check the value here. */
2641 const unsigned int correct_length
2650 if (!PL_rsfp_filters)
2652 if (idx > AvFILLp(PL_rsfp_filters)) { /* Any more filters? */
2653 /* Provide a default input filter to make life easy. */
2654 /* Note that we append to the line. This is handy. */
2655 DEBUG_P(PerlIO_printf(Perl_debug_log,
2656 "filter_read %d: from rsfp\n", idx));
2657 if (correct_length) {
2660 const int old_len = SvCUR(buf_sv);
2662 /* ensure buf_sv is large enough */
2663 SvGROW(buf_sv, (STRLEN)(old_len + correct_length)) ;
2664 if ((len = PerlIO_read(PL_rsfp, SvPVX(buf_sv) + old_len,
2665 correct_length)) <= 0) {
2666 if (PerlIO_error(PL_rsfp))
2667 return -1; /* error */
2669 return 0 ; /* end of file */
2671 SvCUR_set(buf_sv, old_len + len) ;
2674 if (sv_gets(buf_sv, PL_rsfp, SvCUR(buf_sv)) == NULL) {
2675 if (PerlIO_error(PL_rsfp))
2676 return -1; /* error */
2678 return 0 ; /* end of file */
2681 return SvCUR(buf_sv);
2683 /* Skip this filter slot if filter has been deleted */
2684 if ( (datasv = FILTER_DATA(idx)) == &PL_sv_undef) {
2685 DEBUG_P(PerlIO_printf(Perl_debug_log,
2686 "filter_read %d: skipped (filter deleted)\n",
2688 return FILTER_READ(idx+1, buf_sv, correct_length); /* recurse */
2690 /* Get function pointer hidden within datasv */
2691 funcp = DPTR2FPTR(filter_t, IoANY(datasv));
2692 DEBUG_P(PerlIO_printf(Perl_debug_log,
2693 "filter_read %d: via function %p (%s)\n",
2694 idx, (void*)datasv, SvPV_nolen_const(datasv)));
2695 /* Call function. The function is expected to */
2696 /* call "FILTER_READ(idx+1, buf_sv)" first. */
2697 /* Return: <0:error, =0:eof, >0:not eof */
2698 return (*funcp)(aTHX_ idx, buf_sv, correct_length);
2702 S_filter_gets(pTHX_ register SV *sv, register PerlIO *fp, STRLEN append)
2705 #ifdef PERL_CR_FILTER
2706 if (!PL_rsfp_filters) {
2707 filter_add(S_cr_textfilter,NULL);
2710 if (PL_rsfp_filters) {
2712 SvCUR_set(sv, 0); /* start with empty line */
2713 if (FILTER_READ(0, sv, 0) > 0)
2714 return ( SvPVX(sv) ) ;
2719 return (sv_gets(sv, fp, append));
2723 S_find_in_my_stash(pTHX_ const char *pkgname, I32 len)
2728 if (len == 11 && *pkgname == '_' && strEQ(pkgname, "__PACKAGE__"))
2732 (pkgname[len - 2] == ':' && pkgname[len - 1] == ':') &&
2733 (gv = gv_fetchpvn_flags(pkgname, len, 0, SVt_PVHV)))
2735 return GvHV(gv); /* Foo:: */
2738 /* use constant CLASS => 'MyClass' */
2739 if ((gv = gv_fetchpvn_flags(pkgname, len, 0, SVt_PVCV))) {
2741 if (GvCV(gv) && (sv = cv_const_sv(GvCV(gv)))) {
2742 pkgname = SvPV_nolen_const(sv);
2746 return gv_stashpv(pkgname, FALSE);
2752 * The intent of this yylex wrapper is to minimize the changes to the
2753 * tokener when we aren't interested in collecting madprops. It remains
2754 * to be seen how successful this strategy will be...
2761 char *s = PL_bufptr;
2763 /* make sure PL_thiswhite is initialized */
2767 /* just do what yylex would do on pending identifier; leave PL_thiswhite alone */
2768 if (PL_pending_ident)
2769 return S_pending_ident(aTHX);
2771 /* previous token ate up our whitespace? */
2772 if (!PL_lasttoke && PL_nextwhite) {
2773 PL_thiswhite = PL_nextwhite;
2777 /* isolate the token, and figure out where it is without whitespace */
2778 PL_realtokenstart = -1;
2782 assert(PL_curforce < 0);
2784 if (!PL_thismad || PL_thismad->mad_key == '^') { /* not forced already? */
2785 if (!PL_thistoken) {
2786 if (PL_realtokenstart < 0 || !CopLINE(PL_curcop))
2787 PL_thistoken = newSVpvn("",0);
2789 char *tstart = SvPVX(PL_linestr) + PL_realtokenstart;
2790 PL_thistoken = newSVpvn(tstart, s - tstart);
2793 if (PL_thismad) /* install head */
2794 CURMAD('X', PL_thistoken);
2797 /* last whitespace of a sublex? */
2798 if (optype == ')' && PL_endwhite) {
2799 CURMAD('X', PL_endwhite);
2804 /* if no whitespace and we're at EOF, bail. Otherwise fake EOF below. */
2805 if (!PL_thiswhite && !PL_endwhite && !optype) {
2806 sv_free(PL_thistoken);
2811 /* put off final whitespace till peg */
2812 if (optype == ';' && !PL_rsfp) {
2813 PL_nextwhite = PL_thiswhite;
2816 else if (PL_thisopen) {
2817 CURMAD('q', PL_thisopen);
2819 sv_free(PL_thistoken);
2823 /* Store actual token text as madprop X */
2824 CURMAD('X', PL_thistoken);
2828 /* add preceding whitespace as madprop _ */
2829 CURMAD('_', PL_thiswhite);
2833 /* add quoted material as madprop = */
2834 CURMAD('=', PL_thisstuff);
2838 /* add terminating quote as madprop Q */
2839 CURMAD('Q', PL_thisclose);
2843 /* special processing based on optype */
2847 /* opval doesn't need a TOKEN since it can already store mp */
2858 append_madprops(PL_thismad, yylval.opval, 0);
2866 addmad(newMADsv('p', PL_endwhite), &PL_thismad, 0);
2875 /* remember any fake bracket that lexer is about to discard */
2876 if (PL_lex_brackets == 1 &&
2877 ((expectation)PL_lex_brackstack[0] & XFAKEBRACK))
2880 while (s < PL_bufend && (*s == ' ' || *s == '\t'))
2883 PL_thiswhite = newSVpvn(PL_bufptr, ++s - PL_bufptr);
2884 addmad(newMADsv('#', PL_thiswhite), &PL_thismad, 0);
2887 break; /* don't bother looking for trailing comment */
2896 /* attach a trailing comment to its statement instead of next token */
2900 if (PL_bufptr > PL_oldbufptr && PL_bufptr[-1] == optype) {
2902 while (s < PL_bufend && (*s == ' ' || *s == '\t'))
2904 if (*s == '\n' || *s == '#') {
2905 while (s < PL_bufend && *s != '\n')
2909 PL_thiswhite = newSVpvn(PL_bufptr, s - PL_bufptr);
2910 addmad(newMADsv('#', PL_thiswhite), &PL_thismad, 0);
2927 /* Create new token struct. Note: opvals return early above. */
2928 yylval.tkval = newTOKEN(optype, yylval, PL_thismad);
2935 S_tokenize_use(pTHX_ int is_use, char *s) {
2937 if (PL_expect != XSTATE)
2938 yyerror(Perl_form(aTHX_ "\"%s\" not allowed in expression",
2939 is_use ? "use" : "no"));
2941 if (isDIGIT(*s) || (*s == 'v' && isDIGIT(s[1]))) {
2942 s = force_version(s, TRUE);
2943 if (*s == ';' || (s = SKIPSPACE1(s), *s == ';')) {
2944 start_force(PL_curforce);
2945 NEXTVAL_NEXTTOKE.opval = NULL;
2948 else if (*s == 'v') {
2949 s = force_word(s,WORD,FALSE,TRUE,FALSE);
2950 s = force_version(s, FALSE);
2954 s = force_word(s,WORD,FALSE,TRUE,FALSE);
2955 s = force_version(s, FALSE);
2957 yylval.ival = is_use;
2961 static const char* const exp_name[] =
2962 { "OPERATOR", "TERM", "REF", "STATE", "BLOCK", "ATTRBLOCK",
2963 "ATTRTERM", "TERMBLOCK", "TERMORDORDOR"
2970 Works out what to call the token just pulled out of the input
2971 stream. The yacc parser takes care of taking the ops we return and
2972 stitching them into a tree.
2978 if read an identifier
2979 if we're in a my declaration
2980 croak if they tried to say my($foo::bar)
2981 build the ops for a my() declaration
2982 if it's an access to a my() variable
2983 are we in a sort block?
2984 croak if my($a); $a <=> $b
2985 build ops for access to a my() variable
2986 if in a dq string, and they've said @foo and we can't find @foo
2988 build ops for a bareword
2989 if we already built the token before, use it.
2994 #pragma segment Perl_yylex
3000 register char *s = PL_bufptr;
3006 SV* tmp = newSVpvs("");
3007 PerlIO_printf(Perl_debug_log, "### %"IVdf":LEX_%s/X%s %s\n",
3008 (IV)CopLINE(PL_curcop),
3009 lex_state_names[PL_lex_state],
3010 exp_name[PL_expect],
3011 pv_display(tmp, s, strlen(s), 0, 60));
3014 /* check if there's an identifier for us to look at */
3015 if (PL_pending_ident)
3016 return REPORT(S_pending_ident(aTHX));
3018 /* no identifier pending identification */
3020 switch (PL_lex_state) {
3022 case LEX_NORMAL: /* Some compilers will produce faster */
3023 case LEX_INTERPNORMAL: /* code if we comment these out. */
3027 /* when we've already built the next token, just pull it out of the queue */
3031 yylval = PL_nexttoke[PL_lasttoke].next_val;
3033 PL_thismad = PL_nexttoke[PL_lasttoke].next_mad;
3034 PL_nexttoke[PL_lasttoke].next_mad = 0;
3035 if (PL_thismad && PL_thismad->mad_key == '_') {
3036 PL_thiswhite = (SV*)PL_thismad->mad_val;
3037 PL_thismad->mad_val = 0;
3038 mad_free(PL_thismad);
3043 PL_lex_state = PL_lex_defer;
3044 PL_expect = PL_lex_expect;
3045 PL_lex_defer = LEX_NORMAL;
3046 if (!PL_nexttoke[PL_lasttoke].next_type)
3051 yylval = PL_nextval[PL_nexttoke];
3053 PL_lex_state = PL_lex_defer;
3054 PL_expect = PL_lex_expect;
3055 PL_lex_defer = LEX_NORMAL;
3059 /* FIXME - can these be merged? */
3060 return(PL_nexttoke[PL_lasttoke].next_type);
3062 return REPORT(PL_nexttype[PL_nexttoke]);
3065 /* interpolated case modifiers like \L \U, including \Q and \E.
3066 when we get here, PL_bufptr is at the \
3068 case LEX_INTERPCASEMOD:
3070 if (PL_bufptr != PL_bufend && *PL_bufptr != '\\')
3071 Perl_croak(aTHX_ "panic: INTERPCASEMOD");
3073 /* handle \E or end of string */
3074 if (PL_bufptr == PL_bufend || PL_bufptr[1] == 'E') {
3076 if (PL_lex_casemods) {
3077 const char oldmod = PL_lex_casestack[--PL_lex_casemods];
3078 PL_lex_casestack[PL_lex_casemods] = '\0';
3080 if (PL_bufptr != PL_bufend
3081 && (oldmod == 'L' || oldmod == 'U' || oldmod == 'Q')) {
3083 PL_lex_state = LEX_INTERPCONCAT;
3086 PL_thistoken = newSVpvn("\\E",2);
3092 while (PL_bufptr != PL_bufend &&
3093 PL_bufptr[0] == '\\' && PL_bufptr[1] == 'E') {
3095 PL_thiswhite = newSVpvn("",0);
3096 sv_catpvn(PL_thiswhite, PL_bufptr, 2);
3100 if (PL_bufptr != PL_bufend)
3103 PL_lex_state = LEX_INTERPCONCAT;
3107 DEBUG_T({ PerlIO_printf(Perl_debug_log,
3108 "### Saw case modifier\n"); });
3110 if (s[1] == '\\' && s[2] == 'E') {
3113 PL_thiswhite = newSVpvn("",0);
3114 sv_catpvn(PL_thiswhite, PL_bufptr, 4);
3117 PL_lex_state = LEX_INTERPCONCAT;
3122 if (!PL_madskills) /* when just compiling don't need correct */
3123 if (strnEQ(s, "L\\u", 3) || strnEQ(s, "U\\l", 3))
3124 tmp = *s, *s = s[2], s[2] = (char)tmp; /* misordered... */
3125 if ((*s == 'L' || *s == 'U') &&
3126 (strchr(PL_lex_casestack, 'L') || strchr(PL_lex_casestack, 'U'))) {
3127 PL_lex_casestack[--PL_lex_casemods] = '\0';
3130 if (PL_lex_casemods > 10)
3131 Renew(PL_lex_casestack, PL_lex_casemods + 2, char);
3132 PL_lex_casestack[PL_lex_casemods++] = *s;
3133 PL_lex_casestack[PL_lex_casemods] = '\0';
3134 PL_lex_state = LEX_INTERPCONCAT;
3135 start_force(PL_curforce);
3136 NEXTVAL_NEXTTOKE.ival = 0;
3138 start_force(PL_curforce);
3140 NEXTVAL_NEXTTOKE.ival = OP_LCFIRST;
3142 NEXTVAL_NEXTTOKE.ival = OP_UCFIRST;
3144 NEXTVAL_NEXTTOKE.ival = OP_LC;
3146 NEXTVAL_NEXTTOKE.ival = OP_UC;
3148 NEXTVAL_NEXTTOKE.ival = OP_QUOTEMETA;
3150 Perl_croak(aTHX_ "panic: yylex");
3152 SV* const tmpsv = newSVpvn("",0);
3153 Perl_sv_catpvf(aTHX_ tmpsv, "\\%c", *s);
3159 if (PL_lex_starts) {
3165 sv_free(PL_thistoken);
3166 PL_thistoken = newSVpvn("",0);
3169 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
3170 if (PL_lex_casemods == 1 && PL_lex_inpat)
3179 case LEX_INTERPPUSH:
3180 return REPORT(sublex_push());
3182 case LEX_INTERPSTART:
3183 if (PL_bufptr == PL_bufend)
3184 return REPORT(sublex_done());
3185 DEBUG_T({ PerlIO_printf(Perl_debug_log,
3186 "### Interpolated variable\n"); });
3188 PL_lex_dojoin = (*PL_bufptr == '@');
3189 PL_lex_state = LEX_INTERPNORMAL;
3190 if (PL_lex_dojoin) {
3191 start_force(PL_curforce);
3192 NEXTVAL_NEXTTOKE.ival = 0;
3194 start_force(PL_curforce);
3195 force_ident("\"", '$');
3196 start_force(PL_curforce);
3197 NEXTVAL_NEXTTOKE.ival = 0;
3199 start_force(PL_curforce);
3200 NEXTVAL_NEXTTOKE.ival = 0;
3202 start_force(PL_curforce);
3203 NEXTVAL_NEXTTOKE.ival = OP_JOIN; /* emulate join($", ...) */
3206 if (PL_lex_starts++) {
3211 sv_free(PL_thistoken);
3212 PL_thistoken = newSVpvn("",0);
3215 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
3216 if (!PL_lex_casemods && PL_lex_inpat)
3223 case LEX_INTERPENDMAYBE:
3224 if (intuit_more(PL_bufptr)) {
3225 PL_lex_state = LEX_INTERPNORMAL; /* false alarm, more expr */
3231 if (PL_lex_dojoin) {
3232 PL_lex_dojoin = FALSE;
3233 PL_lex_state = LEX_INTERPCONCAT;
3237 sv_free(PL_thistoken);
3238 PL_thistoken = newSVpvn("",0);
3243 if (PL_lex_inwhat == OP_SUBST && PL_linestr == PL_lex_repl
3244 && SvEVALED(PL_lex_repl))
3246 if (PL_bufptr != PL_bufend)
3247 Perl_croak(aTHX_ "Bad evalled substitution pattern");
3251 case LEX_INTERPCONCAT:
3253 if (PL_lex_brackets)
3254 Perl_croak(aTHX_ "panic: INTERPCONCAT");
3256 if (PL_bufptr == PL_bufend)
3257 return REPORT(sublex_done());
3259 if (SvIVX(PL_linestr) == '\'') {
3260 SV *sv = newSVsv(PL_linestr);
3263 else if ( PL_hints & HINT_NEW_RE )
3264 sv = new_constant(NULL, 0, "qr", sv, sv, "q");
3265 yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
3269 s = scan_const(PL_bufptr);
3271 PL_lex_state = LEX_INTERPCASEMOD;
3273 PL_lex_state = LEX_INTERPSTART;
3276 if (s != PL_bufptr) {
3277 start_force(PL_curforce);
3279 curmad('X', newSVpvn(PL_bufptr,s-PL_bufptr));
3281 NEXTVAL_NEXTTOKE = yylval;
3284 if (PL_lex_starts++) {
3288 sv_free(PL_thistoken);
3289 PL_thistoken = newSVpvn("",0);
3292 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
3293 if (!PL_lex_casemods && PL_lex_inpat)
3306 PL_lex_state = LEX_NORMAL;
3307 s = scan_formline(PL_bufptr);
3308 if (!PL_lex_formbrack)
3314 PL_oldoldbufptr = PL_oldbufptr;
3320 sv_free(PL_thistoken);
3323 PL_realtokenstart = s - SvPVX(PL_linestr); /* assume but undo on ws */
3327 if (isIDFIRST_lazy_if(s,UTF))
3329 Perl_croak(aTHX_ "Unrecognized character \\x%02X", *s & 255);
3332 goto fake_eof; /* emulate EOF on ^D or ^Z */
3341 if (PL_lex_brackets) {
3342 yyerror(PL_lex_formbrack
3343 ? "Format not terminated"
3344 : "Missing right curly or square bracket");
3346 DEBUG_T( { PerlIO_printf(Perl_debug_log,
3347 "### Tokener got EOF\n");
3351 if (s++ < PL_bufend)
3352 goto retry; /* ignore stray nulls */
3355 if (!PL_in_eval && !PL_preambled) {
3356 PL_preambled = TRUE;
3361 sv_setpv(PL_linestr,incl_perldb());
3362 if (SvCUR(PL_linestr))
3363 sv_catpvs(PL_linestr,";");
3365 while(AvFILLp(PL_preambleav) >= 0) {
3366 SV *tmpsv = av_shift(PL_preambleav);
3367 sv_catsv(PL_linestr, tmpsv);
3368 sv_catpvs(PL_linestr, ";");
3371 sv_free((SV*)PL_preambleav);
3372 PL_preambleav = NULL;
3374 if (PL_minus_n || PL_minus_p) {
3375 sv_catpvs(PL_linestr, "LINE: while (<>) {");
3377 sv_catpvs(PL_linestr,"chomp;");
3380 if ((*PL_splitstr == '/' || *PL_splitstr == '\''
3381 || *PL_splitstr == '"')
3382 && strchr(PL_splitstr + 1, *PL_splitstr))
3383 Perl_sv_catpvf(aTHX_ PL_linestr, "our @F=split(%s);", PL_splitstr);
3385 /* "q\0${splitstr}\0" is legal perl. Yes, even NUL
3386 bytes can be used as quoting characters. :-) */
3387 const char *splits = PL_splitstr;
3388 sv_catpvs(PL_linestr, "our @F=split(q\0");
3391 if (*splits == '\\')
3392 sv_catpvn(PL_linestr, splits, 1);
3393 sv_catpvn(PL_linestr, splits, 1);
3394 } while (*splits++);
3395 /* This loop will embed the trailing NUL of
3396 PL_linestr as the last thing it does before
3398 sv_catpvs(PL_linestr, ");");
3402 sv_catpvs(PL_linestr,"our @F=split(' ');");
3406 sv_catpvs(PL_linestr,"use feature ':5.10';");
3407 sv_catpvs(PL_linestr, "\n");
3408 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3409 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3410 PL_last_lop = PL_last_uni = NULL;
3411 if (PERLDB_LINE && PL_curstash != PL_debstash) {
3412 SV * const sv = newSV(0);
3414 sv_upgrade(sv, SVt_PVMG);
3415 sv_setsv(sv,PL_linestr);
3418 av_store(CopFILEAVx(PL_curcop),(I32)CopLINE(PL_curcop),sv);
3423 bof = PL_rsfp ? TRUE : FALSE;
3424 if ((s = filter_gets(PL_linestr, PL_rsfp, 0)) == NULL) {
3427 PL_realtokenstart = -1;
3430 if (PL_preprocess && !PL_in_eval)
3431 (void)PerlProc_pclose(PL_rsfp);
3432 else if ((PerlIO *)PL_rsfp == PerlIO_stdin())
3433 PerlIO_clearerr(PL_rsfp);
3435 (void)PerlIO_close(PL_rsfp);
3437 PL_doextract = FALSE;
3439 if (!PL_in_eval && (PL_minus_n || PL_minus_p)) {
3444 sv_setpv(PL_linestr,PL_minus_p
3445 ? ";}continue{print;}" : ";}");
3446 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3447 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3448 PL_last_lop = PL_last_uni = NULL;
3449 PL_minus_n = PL_minus_p = 0;
3452 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3453 PL_last_lop = PL_last_uni = NULL;
3454 sv_setpvn(PL_linestr,"",0);
3455 TOKEN(';'); /* not infinite loop because rsfp is NULL now */
3457 /* If it looks like the start of a BOM or raw UTF-16,
3458 * check if it in fact is. */
3464 #ifdef PERLIO_IS_STDIO
3465 # ifdef __GNU_LIBRARY__
3466 # if __GNU_LIBRARY__ == 1 /* Linux glibc5 */
3467 # define FTELL_FOR_PIPE_IS_BROKEN
3471 # if __GLIBC__ == 1 /* maybe some glibc5 release had it like this? */
3472 # define FTELL_FOR_PIPE_IS_BROKEN
3477 #ifdef FTELL_FOR_PIPE_IS_BROKEN
3478 /* This loses the possibility to detect the bof
3479 * situation on perl -P when the libc5 is being used.
3480 * Workaround? Maybe attach some extra state to PL_rsfp?
3483 bof = PerlIO_tell(PL_rsfp) == SvCUR(PL_linestr);
3485 bof = PerlIO_tell(PL_rsfp) == (Off_t)SvCUR(PL_linestr);
3488 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3489 s = swallow_bom((U8*)s);
3493 /* Incest with pod. */
3496 sv_catsv(PL_thiswhite, PL_linestr);
3498 if (*s == '=' && strnEQ(s, "=cut", 4)) {
3499 sv_setpvn(PL_linestr, "", 0);
3500 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3501 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3502 PL_last_lop = PL_last_uni = NULL;
3503 PL_doextract = FALSE;
3507 } while (PL_doextract);
3508 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = s;
3509 if (PERLDB_LINE && PL_curstash != PL_debstash) {
3510 SV * const sv = newSV(0);
3512 sv_upgrade(sv, SVt_PVMG);
3513 sv_setsv(sv,PL_linestr);
3516 av_store(CopFILEAVx(PL_curcop),(I32)CopLINE(PL_curcop),sv);
3518 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3519 PL_last_lop = PL_last_uni = NULL;
3520 if (CopLINE(PL_curcop) == 1) {
3521 while (s < PL_bufend && isSPACE(*s))
3523 if (*s == ':' && s[1] != ':') /* for csh execing sh scripts */
3527 PL_thiswhite = newSVpvn(PL_linestart, s - PL_linestart);
3531 if (*s == '#' && *(s+1) == '!')
3533 #ifdef ALTERNATE_SHEBANG
3535 static char const as[] = ALTERNATE_SHEBANG;
3536 if (*s == as[0] && strnEQ(s, as, sizeof(as) - 1))
3537 d = s + (sizeof(as) - 1);
3539 #endif /* ALTERNATE_SHEBANG */
3548 while (*d && !isSPACE(*d))
3552 #ifdef ARG_ZERO_IS_SCRIPT
3553 if (ipathend > ipath) {
3555 * HP-UX (at least) sets argv[0] to the script name,
3556 * which makes $^X incorrect. And Digital UNIX and Linux,
3557 * at least, set argv[0] to the basename of the Perl
3558 * interpreter. So, having found "#!", we'll set it right.
3560 SV * const x = GvSV(gv_fetchpvs("\030", GV_ADD|GV_NOTQUAL,
3562 assert(SvPOK(x) || SvGMAGICAL(x));
3563 if (sv_eq(x, CopFILESV(PL_curcop))) {
3564 sv_setpvn(x, ipath, ipathend - ipath);
3570 const char *bstart = SvPV_const(CopFILESV(PL_curcop),blen);
3571 const char * const lstart = SvPV_const(x,llen);
3573 bstart += blen - llen;
3574 if (strnEQ(bstart, lstart, llen) && bstart[-1] == '/') {
3575 sv_setpvn(x, ipath, ipathend - ipath);
3580 TAINT_NOT; /* $^X is always tainted, but that's OK */
3582 #endif /* ARG_ZERO_IS_SCRIPT */
3587 d = instr(s,"perl -");
3589 d = instr(s,"perl");
3591 /* avoid getting into infinite loops when shebang
3592 * line contains "Perl" rather than "perl" */
3594 for (d = ipathend-4; d >= ipath; --d) {
3595 if ((*d == 'p' || *d == 'P')
3596 && !ibcmp(d, "perl", 4))
3606 #ifdef ALTERNATE_SHEBANG
3608 * If the ALTERNATE_SHEBANG on this system starts with a
3609 * character that can be part of a Perl expression, then if
3610 * we see it but not "perl", we're probably looking at the
3611 * start of Perl code, not a request to hand off to some
3612 * other interpreter. Similarly, if "perl" is there, but
3613 * not in the first 'word' of the line, we assume the line
3614 * contains the start of the Perl program.
3616 if (d && *s != '#') {
3617 const char *c = ipath;
3618 while (*c && !strchr("; \t\r\n\f\v#", *c))
3621 d = NULL; /* "perl" not in first word; ignore */
3623 *s = '#'; /* Don't try to parse shebang line */
3625 #endif /* ALTERNATE_SHEBANG */
3626 #ifndef MACOS_TRADITIONAL
3631 !instr(s,"indir") &&
3632 instr(PL_origargv[0],"perl"))
3639 while (s < PL_bufend && isSPACE(*s))
3641 if (s < PL_bufend) {
3642 Newxz(newargv,PL_origargc+3,char*);
3644 while (s < PL_bufend && !isSPACE(*s))
3647 Copy(PL_origargv+1, newargv+2, PL_origargc+1, char*);
3650 newargv = PL_origargv;
3653 PerlProc_execv(ipath, EXEC_ARGV_CAST(newargv));
3655 Perl_croak(aTHX_ "Can't exec %s", ipath);
3659 while (*d && !isSPACE(*d)) d++;
3660 while (SPACE_OR_TAB(*d)) d++;
3663 const bool switches_done = PL_doswitches;
3664 const U32 oldpdb = PL_perldb;
3665 const bool oldn = PL_minus_n;
3666 const bool oldp = PL_minus_p;
3669 if (*d == 'M' || *d == 'm' || *d == 'C') {
3670 const char * const m = d;
3671 while (*d && !isSPACE(*d))
3673 Perl_croak(aTHX_ "Too late for \"-%.*s\" option",
3676 d = moreswitches(d);
3678 if (PL_doswitches && !switches_done) {
3679 int argc = PL_origargc;
3680 char **argv = PL_origargv;
3683 } while (argc && argv[0][0] == '-' && argv[0][1]);
3684 init_argv_symbols(argc,argv);
3686 if ((PERLDB_LINE && !oldpdb) ||
3687 ((PL_minus_n || PL_minus_p) && !(oldn || oldp)))
3688 /* if we have already added "LINE: while (<>) {",
3689 we must not do it again */
3691 sv_setpvn(PL_linestr, "", 0);
3692 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3693 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3694 PL_last_lop = PL_last_uni = NULL;
3695 PL_preambled = FALSE;
3697 (void)gv_fetchfile(PL_origfilename);
3704 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
3706 PL_lex_state = LEX_FORMLINE;
3711 #ifdef PERL_STRICT_CR
3712 Perl_warn(aTHX_ "Illegal character \\%03o (carriage return)", '\r');
3714 "\t(Maybe you didn't strip carriage returns after a network transfer?)\n");
3716 case ' ': case '\t': case '\f': case 013:
3717 #ifdef MACOS_TRADITIONAL
3721 PL_realtokenstart = -1;
3730 PL_realtokenstart = -1;
3734 if (PL_lex_state != LEX_NORMAL || (PL_in_eval && !PL_rsfp)) {
3735 if (*s == '#' && s == PL_linestart && PL_in_eval && !PL_rsfp) {
3736 /* handle eval qq[#line 1 "foo"\n ...] */
3737 CopLINE_dec(PL_curcop);
3740 if (PL_madskills && !PL_lex_formbrack && !PL_in_eval) {
3742 if (!PL_in_eval || PL_rsfp)
3747 while (d < PL_bufend && *d != '\n')
3751 else if (d > PL_bufend) /* Found by Ilya: feed random input to Perl. */
3752 Perl_croak(aTHX_ "panic: input overflow");
3755 PL_thiswhite = newSVpvn(s, d - s);
3760 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
3762 PL_lex_state = LEX_FORMLINE;
3768 if (PL_madskills && CopLINE(PL_curcop) >= 1 && !PL_lex_formbrack) {
3769 if (CopLINE(PL_curcop) == 1 && s[0] == '#' && s[1] == '!') {
3772 TOKEN(PEG); /* make sure any #! line is accessible */
3777 /* if (PL_madskills && PL_lex_formbrack) { */
3779 while (d < PL_bufend && *d != '\n')
3783 else if (d > PL_bufend) /* Found by Ilya: feed random input to Perl. */
3784 Perl_croak(aTHX_ "panic: input overflow");
3785 if (PL_madskills && CopLINE(PL_curcop) >= 1) {
3787 PL_thiswhite = newSVpvn("",0);
3788 if (CopLINE(PL_curcop) == 1) {
3789 sv_setpvn(PL_thiswhite, "", 0);
3792 sv_catpvn(PL_thiswhite, s, d - s);
3806 if (s[1] && isALPHA(s[1]) && !isALNUM(s[2])) {
3814 while (s < PL_bufend && SPACE_OR_TAB(*s))
3817 if (strnEQ(s,"=>",2)) {
3818 s = force_word(PL_bufptr,WORD,FALSE,FALSE,FALSE);
3819 DEBUG_T( { S_printbuf(aTHX_
3820 "### Saw unary minus before =>, forcing word %s\n", s);
3822 OPERATOR('-'); /* unary minus */
3824 PL_last_uni = PL_oldbufptr;
3826 case 'r': ftst = OP_FTEREAD; break;
3827 case 'w': ftst = OP_FTEWRITE; break;
3828 case 'x': ftst = OP_FTEEXEC; break;
3829 case 'o': ftst = OP_FTEOWNED; break;
3830 case 'R': ftst = OP_FTRREAD; break;
3831 case 'W': ftst = OP_FTRWRITE; break;
3832 case 'X': ftst = OP_FTREXEC; break;
3833 case 'O': ftst = OP_FTROWNED; break;
3834 case 'e': ftst = OP_FTIS; break;
3835 case 'z': ftst = OP_FTZERO; break;
3836 case 's': ftst = OP_FTSIZE; break;
3837 case 'f': ftst = OP_FTFILE; break;
3838 case 'd': ftst = OP_FTDIR; break;
3839 case 'l': ftst = OP_FTLINK; break;
3840 case 'p': ftst = OP_FTPIPE; break;
3841 case 'S': ftst = OP_FTSOCK; break;
3842 case 'u': ftst = OP_FTSUID; break;
3843 case 'g': ftst = OP_FTSGID; break;
3844 case 'k': ftst = OP_FTSVTX; break;
3845 case 'b': ftst = OP_FTBLK; break;
3846 case 'c': ftst = OP_FTCHR; break;
3847 case 't': ftst = OP_FTTTY; break;
3848 case 'T': ftst = OP_FTTEXT; break;
3849 case 'B': ftst = OP_FTBINARY; break;
3850 case 'M': case 'A': case 'C':
3851 gv_fetchpvs("\024", GV_ADD|GV_NOTQUAL, SVt_PV);
3853 case 'M': ftst = OP_FTMTIME; break;
3854 case 'A': ftst = OP_FTATIME; break;
3855 case 'C': ftst = OP_FTCTIME; break;
3863 PL_last_lop_op = (OPCODE)ftst;
3864 DEBUG_T( { PerlIO_printf(Perl_debug_log,
3865 "### Saw file test %c\n", (int)tmp);
3870 /* Assume it was a minus followed by a one-letter named
3871 * subroutine call (or a -bareword), then. */
3872 DEBUG_T( { PerlIO_printf(Perl_debug_log,
3873 "### '-%c' looked like a file test but was not\n",
3880 const char tmp = *s++;
3883 if (PL_expect == XOPERATOR)
3888 else if (*s == '>') {
3891 if (isIDFIRST_lazy_if(s,UTF)) {
3892 s = force_word(s,METHOD,FALSE,TRUE,FALSE);
3900 if (PL_expect == XOPERATOR)
3903 if (isSPACE(*s) || !isSPACE(*PL_bufptr))
3905 OPERATOR('-'); /* unary minus */
3911 const char tmp = *s++;
3914 if (PL_expect == XOPERATOR)
3919 if (PL_expect == XOPERATOR)
3922 if (isSPACE(*s) || !isSPACE(*PL_bufptr))
3929 if (PL_expect != XOPERATOR) {
3930 s = scan_ident(s, PL_bufend, PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
3931 PL_expect = XOPERATOR;
3932 force_ident(PL_tokenbuf, '*');
3945 if (PL_expect == XOPERATOR) {
3949 PL_tokenbuf[0] = '%';
3950 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1, TRUE);
3951 if (!PL_tokenbuf[1]) {
3954 PL_pending_ident = '%';
3965 && (PL_expect == XOPERATOR || PL_expect == XTERMORDORDOR)
3966 && FEATURE_IS_ENABLED("~~"))
3973 const char tmp = *s++;
3979 goto just_a_word_zero_gv;
3982 switch (PL_expect) {
3988 if (!PL_in_my || PL_lex_state != LEX_NORMAL)
3990 PL_bufptr = s; /* update in case we back off */
3996 PL_expect = XTERMBLOCK;
3999 stuffstart = s - SvPVX(PL_linestr) - 1;
4003 while (isIDFIRST_lazy_if(s,UTF)) {
4006 d = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
4007 if (isLOWER(*s) && (tmp = keyword(PL_tokenbuf, len))) {
4008 if (tmp < 0) tmp = -tmp;
4023 sv = newSVpvn(s, len);
4025 d = scan_str(d,TRUE,TRUE);
4027 /* MUST advance bufptr here to avoid bogus
4028 "at end of line" context messages from yyerror().
4030 PL_bufptr = s + len;
4031 yyerror("Unterminated attribute parameter in attribute list");
4035 return REPORT(0); /* EOF indicator */
4039 sv_catsv(sv, PL_lex_stuff);
4040 attrs = append_elem(OP_LIST, attrs,
4041 newSVOP(OP_CONST, 0, sv));
4042 SvREFCNT_dec(PL_lex_stuff);
4043 PL_lex_stuff = NULL;
4046 if (len == 6 && strnEQ(SvPVX(sv), "unique", len)) {
4048 if (PL_in_my == KEY_our) {
4050 GvUNIQUE_on(cGVOPx_gv(yylval.opval));
4052 /* skip to avoid loading attributes.pm */
4054 deprecate(":unique");
4057 Perl_croak(aTHX_ "The 'unique' attribute may only be applied to 'our' variables");
4060 /* NOTE: any CV attrs applied here need to be part of
4061 the CVf_BUILTIN_ATTRS define in cv.h! */
4062 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "lvalue", len)) {
4064 CvLVALUE_on(PL_compcv);
4066 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "locked", len)) {
4068 CvLOCKED_on(PL_compcv);
4070 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "method", len)) {
4072 CvMETHOD_on(PL_compcv);
4074 else if (!PL_in_my && len == 9 && strnEQ(SvPVX(sv), "assertion", len)) {
4076 CvASSERTION_on(PL_compcv);
4078 /* After we've set the flags, it could be argued that
4079 we don't need to do the attributes.pm-based setting
4080 process, and shouldn't bother appending recognized
4081 flags. To experiment with that, uncomment the
4082 following "else". (Note that's already been
4083 uncommented. That keeps the above-applied built-in
4084 attributes from being intercepted (and possibly
4085 rejected) by a package's attribute routines, but is
4086 justified by the performance win for the common case
4087 of applying only built-in attributes.) */
4089 attrs = append_elem(OP_LIST, attrs,
4090 newSVOP(OP_CONST, 0,
4094 if (*s == ':' && s[1] != ':')
4097 break; /* require real whitespace or :'s */
4098 /* XXX losing whitespace on sequential attributes here */
4102 = (PL_expect == XOPERATOR ? '=' : '{'); /*'}(' for vi */
4103 if (*s != ';' && *s != '}' && *s != tmp
4104 && (tmp != '=' || *s != ')')) {
4105 const char q = ((*s == '\'') ? '"' : '\'');
4106 /* If here for an expression, and parsed no attrs, back
4108 if (tmp == '=' && !attrs) {
4112 /* MUST advance bufptr here to avoid bogus "at end of line"
4113 context messages from yyerror().
4117 ? Perl_form(aTHX_ "Invalid separator character "
4118 "%c%c%c in attribute list", q, *s, q)
4119 : "Unterminated attribute list" );
4127 start_force(PL_curforce);
4128 NEXTVAL_NEXTTOKE.opval = attrs;
4129 CURMAD('_', PL_nextwhite);
4134 PL_thistoken = newSVpvn(SvPVX(PL_linestr) + stuffstart,
4135 (s - SvPVX(PL_linestr)) - stuffstart);
4143 if (PL_last_lop == PL_oldoldbufptr || PL_last_uni == PL_oldoldbufptr)
4144 PL_oldbufptr = PL_oldoldbufptr; /* allow print(STDOUT 123) */
4152 const char tmp = *s++;
4157 const char tmp = *s++;
4165 if (PL_lex_brackets <= 0)
4166 yyerror("Unmatched right square bracket");
4169 if (PL_lex_state == LEX_INTERPNORMAL) {
4170 if (PL_lex_brackets == 0) {
4171 if (*s != '[' && *s != '{' && (*s != '-' || s[1] != '>'))
4172 PL_lex_state = LEX_INTERPEND;
4179 if (PL_lex_brackets > 100) {
4180 Renew(PL_lex_brackstack, PL_lex_brackets + 10, char);
4182 switch (PL_expect) {
4184 if (PL_lex_formbrack) {
4188 if (PL_oldoldbufptr == PL_last_lop)
4189 PL_lex_brackstack[PL_lex_brackets++] = XTERM;
4191 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
4192 OPERATOR(HASHBRACK);
4194 while (s < PL_bufend && SPACE_OR_TAB(*s))
4197 PL_tokenbuf[0] = '\0';
4198 if (d < PL_bufend && *d == '-') {
4199 PL_tokenbuf[0] = '-';
4201 while (d < PL_bufend && SPACE_OR_TAB(*d))
4204 if (d < PL_bufend && isIDFIRST_lazy_if(d,UTF)) {
4205 d = scan_word(d, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1,
4207 while (d < PL_bufend && SPACE_OR_TAB(*d))
4210 const char minus = (PL_tokenbuf[0] == '-');
4211 s = force_word(s + minus, WORD, FALSE, TRUE, FALSE);
4219 PL_lex_brackstack[PL_lex_brackets++] = XSTATE;
4224 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
4229 if (PL_oldoldbufptr == PL_last_lop)
4230 PL_lex_brackstack[PL_lex_brackets++] = XTERM;
4232 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
4235 if (PL_expect == XREF && PL_lex_state == LEX_INTERPNORMAL) {
4237 /* This hack is to get the ${} in the message. */
4239 yyerror("syntax error");
4242 OPERATOR(HASHBRACK);
4244 /* This hack serves to disambiguate a pair of curlies
4245 * as being a block or an anon hash. Normally, expectation
4246 * determines that, but in cases where we're not in a
4247 * position to expect anything in particular (like inside
4248 * eval"") we have to resolve the ambiguity. This code
4249 * covers the case where the first term in the curlies is a
4250 * quoted string. Most other cases need to be explicitly
4251 * disambiguated by prepending a "+" before the opening
4252 * curly in order to force resolution as an anon hash.
4254 * XXX should probably propagate the outer expectation
4255 * into eval"" to rely less on this hack, but that could
4256 * potentially break current behavior of eval"".
4260 if (*s == '\'' || *s == '"' || *s == '`') {
4261 /* common case: get past first string, handling escapes */
4262 for (t++; t < PL_bufend && *t != *s;)
4263 if (*t++ == '\\' && (*t == '\\' || *t == *s))
4267 else if (*s == 'q') {
4270 || ((*t == 'q' || *t == 'x') && ++t < PL_bufend
4273 /* skip q//-like construct */
4275 char open, close, term;
4278 while (t < PL_bufend && isSPACE(*t))
4280 /* check for q => */
4281 if (t+1 < PL_bufend && t[0] == '=' && t[1] == '>') {
4282 OPERATOR(HASHBRACK);
4286 if (term && (tmps = strchr("([{< )]}> )]}>",term)))
4290 for (t++; t < PL_bufend; t++) {
4291 if (*t == '\\' && t+1 < PL_bufend && open != '\\')
4293 else if (*t == open)
4297 for (t++; t < PL_bufend; t++) {
4298 if (*t == '\\' && t+1 < PL_bufend)
4300 else if (*t == close && --brackets <= 0)
4302 else if (*t == open)
4309 /* skip plain q word */
4310 while (t < PL_bufend && isALNUM_lazy_if(t,UTF))
4313 else if (isALNUM_lazy_if(t,UTF)) {
4315 while (t < PL_bufend && isALNUM_lazy_if(t,UTF))
4318 while (t < PL_bufend && isSPACE(*t))
4320 /* if comma follows first term, call it an anon hash */
4321 /* XXX it could be a comma expression with loop modifiers */
4322 if (t < PL_bufend && ((*t == ',' && (*s == 'q' || !isLOWER(*s)))
4323 || (*t == '=' && t[1] == '>')))
4324 OPERATOR(HASHBRACK);
4325 if (PL_expect == XREF)
4328 PL_lex_brackstack[PL_lex_brackets-1] = XSTATE;
4334 yylval.ival = CopLINE(PL_curcop);
4335 if (isSPACE(*s) || *s == '#')
4336 PL_copline = NOLINE; /* invalidate current command line number */
4341 if (PL_lex_brackets <= 0)
4342 yyerror("Unmatched right curly bracket");
4344 PL_expect = (expectation)PL_lex_brackstack[--PL_lex_brackets];
4345 if (PL_lex_brackets < PL_lex_formbrack && PL_lex_state != LEX_INTERPNORMAL)
4346 PL_lex_formbrack = 0;
4347 if (PL_lex_state == LEX_INTERPNORMAL) {
4348 if (PL_lex_brackets == 0) {
4349 if (PL_expect & XFAKEBRACK) {
4350 PL_expect &= XENUMMASK;
4351 PL_lex_state = LEX_INTERPEND;
4356 PL_thiswhite = newSVpvn("",0);
4357 sv_catpvn(PL_thiswhite,"}",1);
4360 return yylex(); /* ignore fake brackets */
4362 if (*s == '-' && s[1] == '>')
4363 PL_lex_state = LEX_INTERPENDMAYBE;
4364 else if (*s != '[' && *s != '{')
4365 PL_lex_state = LEX_INTERPEND;
4368 if (PL_expect & XFAKEBRACK) {
4369 PL_expect &= XENUMMASK;
4371 return yylex(); /* ignore fake brackets */
4373 start_force(PL_curforce);
4375 curmad('X', newSVpvn(s-1,1));
4376 CURMAD('_', PL_thiswhite);
4381 PL_thistoken = newSVpvn("",0);
4389 if (PL_expect == XOPERATOR) {
4390 if (PL_bufptr == PL_linestart && ckWARN(WARN_SEMICOLON)
4391 && isIDFIRST_lazy_if(s,UTF))
4393 CopLINE_dec(PL_curcop);
4394 Perl_warner(aTHX_ packWARN(WARN_SEMICOLON), PL_warn_nosemi);
4395 CopLINE_inc(PL_curcop);
4400 s = scan_ident(s - 1, PL_bufend, PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
4402 PL_expect = XOPERATOR;
4403 force_ident(PL_tokenbuf, '&');
4407 yylval.ival = (OPpENTERSUB_AMPER<<8);
4419 const char tmp = *s++;
4426 if (tmp && isSPACE(*s) && ckWARN(WARN_SYNTAX)
4427 && strchr("+-*/%.^&|<",tmp))
4428 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
4429 "Reversed %c= operator",(int)tmp);
4431 if (PL_expect == XSTATE && isALPHA(tmp) &&
4432 (s == PL_linestart+1 || s[-2] == '\n') )
4434 if (PL_in_eval && !PL_rsfp) {
4439 if (strnEQ(s,"=cut",4)) {
4455 PL_thiswhite = newSVpvn("",0);
4456 sv_catpvn(PL_thiswhite, PL_linestart,
4457 PL_bufend - PL_linestart);
4461 PL_doextract = TRUE;
4465 if (PL_lex_brackets < PL_lex_formbrack) {
4467 #ifdef PERL_STRICT_CR
4468 for (t = s; SPACE_OR_TAB(*t); t++) ;
4470 for (t = s; SPACE_OR_TAB(*t) || *t == '\r'; t++) ;
4472 if (*t == '\n' || *t == '#') {
4483 const char tmp = *s++;
4485 /* was this !=~ where !~ was meant?
4486 * warn on m:!=~\s+([/?]|[msy]\W|tr\W): */
4488 if (*s == '~' && ckWARN(WARN_SYNTAX)) {
4489 const char *t = s+1;
4491 while (t < PL_bufend && isSPACE(*t))
4494 if (*t == '/' || *t == '?' ||
4495 ((*t == 'm' || *t == 's' || *t == 'y')
4496 && !isALNUM(t[1])) ||
4497 (*t == 't' && t[1] == 'r' && !isALNUM(t[2])))
4498 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
4499 "!=~ should be !~");
4509 if (PL_expect != XOPERATOR) {
4510 if (s[1] != '<' && !strchr(s,'>'))
4513 s = scan_heredoc(s);
4515 s = scan_inputsymbol(s);
4516 TERM(sublex_start());
4522 SHop(OP_LEFT_SHIFT);
4536 const char tmp = *s++;
4538 SHop(OP_RIGHT_SHIFT);
4539 else if (tmp == '=')
4548 if (PL_expect == XOPERATOR) {
4549 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
4551 deprecate_old(commaless_variable_list);
4552 return REPORT(','); /* grandfather non-comma-format format */
4556 if (s[1] == '#' && (isIDFIRST_lazy_if(s+2,UTF) || strchr("{$:+-", s[2]))) {
4557 PL_tokenbuf[0] = '@';
4558 s = scan_ident(s + 1, PL_bufend, PL_tokenbuf + 1,
4559 sizeof PL_tokenbuf - 1, FALSE);
4560 if (PL_expect == XOPERATOR)
4561 no_op("Array length", s);
4562 if (!PL_tokenbuf[1])
4564 PL_expect = XOPERATOR;
4565 PL_pending_ident = '#';
4569 PL_tokenbuf[0] = '$';
4570 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1,
4571 sizeof PL_tokenbuf - 1, FALSE);
4572 if (PL_expect == XOPERATOR)
4574 if (!PL_tokenbuf[1]) {
4576 yyerror("Final $ should be \\$ or $name");
4580 /* This kludge not intended to be bulletproof. */
4581 if (PL_tokenbuf[1] == '[' && !PL_tokenbuf[2]) {
4582 yylval.opval = newSVOP(OP_CONST, 0,
4583 newSViv(CopARYBASE_get(&PL_compiling)));
4584 yylval.opval->op_private = OPpCONST_ARYBASE;
4590 const char tmp = *s;
4591 if (PL_lex_state == LEX_NORMAL)
4594 if ((PL_expect != XREF || PL_oldoldbufptr == PL_last_lop)
4595 && intuit_more(s)) {
4597 PL_tokenbuf[0] = '@';
4598 if (ckWARN(WARN_SYNTAX)) {
4601 isSPACE(*t) || isALNUM_lazy_if(t,UTF) || *t == '$';
4604 PL_bufptr = PEEKSPACE(PL_bufptr); /* XXX can realloc */
4605 while (t < PL_bufend && *t != ']')
4607 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
4608 "Multidimensional syntax %.*s not supported",
4609 (int)((t - PL_bufptr) + 1), PL_bufptr);
4613 else if (*s == '{') {
4615 PL_tokenbuf[0] = '%';
4616 if (strEQ(PL_tokenbuf+1, "SIG") && ckWARN(WARN_SYNTAX)
4617 && (t = strchr(s, '}')) && (t = strchr(t, '=')))
4619 char tmpbuf[sizeof PL_tokenbuf];
4620 for (t++; isSPACE(*t); t++) ;
4621 if (isIDFIRST_lazy_if(t,UTF)) {
4623 t = scan_word(t, tmpbuf, sizeof tmpbuf, TRUE,
4625 for (; isSPACE(*t); t++) ;
4626 if (*t == ';' && get_cv(tmpbuf, FALSE))
4627 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
4628 "You need to quote \"%s\"",
4635 PL_expect = XOPERATOR;
4636 if (PL_lex_state == LEX_NORMAL && isSPACE((char)tmp)) {
4637 const bool islop = (PL_last_lop == PL_oldoldbufptr);
4638 if (!islop || PL_last_lop_op == OP_GREPSTART)
4639 PL_expect = XOPERATOR;
4640 else if (strchr("$@\"'`q", *s))
4641 PL_expect = XTERM; /* e.g. print $fh "foo" */
4642 else if (strchr("&*<%", *s) && isIDFIRST_lazy_if(s+1,UTF))
4643 PL_expect = XTERM; /* e.g. print $fh &sub */
4644 else if (isIDFIRST_lazy_if(s,UTF)) {
4645 char tmpbuf[sizeof PL_tokenbuf];
4647 scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
4648 if ((t2 = keyword(tmpbuf, len))) {
4649 /* binary operators exclude handle interpretations */
4661 PL_expect = XTERM; /* e.g. print $fh length() */
4666 PL_expect = XTERM; /* e.g. print $fh subr() */
4669 else if (isDIGIT(*s))
4670 PL_expect = XTERM; /* e.g. print $fh 3 */
4671 else if (*s == '.' && isDIGIT(s[1]))
4672 PL_expect = XTERM; /* e.g. print $fh .3 */
4673 else if ((*s == '?' || *s == '-' || *s == '+')
4674 && !isSPACE(s[1]) && s[1] != '=')
4675 PL_expect = XTERM; /* e.g. print $fh -1 */
4676 else if (*s == '/' && !isSPACE(s[1]) && s[1] != '='
4678 PL_expect = XTERM; /* e.g. print $fh /.../
4679 XXX except DORDOR operator
4681 else if (*s == '<' && s[1] == '<' && !isSPACE(s[2])
4683 PL_expect = XTERM; /* print $fh <<"EOF" */
4686 PL_pending_ident = '$';
4690 if (PL_expect == XOPERATOR)
4692 PL_tokenbuf[0] = '@';
4693 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1, FALSE);
4694 if (!PL_tokenbuf[1]) {
4697 if (PL_lex_state == LEX_NORMAL)
4699 if ((PL_expect != XREF || PL_oldoldbufptr == PL_last_lop) && intuit_more(s)) {
4701 PL_tokenbuf[0] = '%';
4703 /* Warn about @ where they meant $. */
4704 if (*s == '[' || *s == '{') {
4705 if (ckWARN(WARN_SYNTAX)) {
4706 const char *t = s + 1;
4707 while (*t && (isALNUM_lazy_if(t,UTF) || strchr(" \t$#+-'\"", *t)))
4709 if (*t == '}' || *t == ']') {
4711 PL_bufptr = PEEKSPACE(PL_bufptr); /* XXX can realloc */
4712 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
4713 "Scalar value %.*s better written as $%.*s",
4714 (int)(t-PL_bufptr), PL_bufptr,
4715 (int)(t-PL_bufptr-1), PL_bufptr+1);
4720 PL_pending_ident = '@';
4723 case '/': /* may be division, defined-or, or pattern */
4724 if (PL_expect == XTERMORDORDOR && s[1] == '/') {
4728 case '?': /* may either be conditional or pattern */
4729 if(PL_expect == XOPERATOR) {
4737 /* A // operator. */
4747 /* Disable warning on "study /blah/" */
4748 if (PL_oldoldbufptr == PL_last_uni
4749 && (*PL_last_uni != 's' || s - PL_last_uni < 5
4750 || memNE(PL_last_uni, "study", 5)
4751 || isALNUM_lazy_if(PL_last_uni+5,UTF)
4754 s = scan_pat(s,OP_MATCH);
4755 TERM(sublex_start());
4759 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack
4760 #ifdef PERL_STRICT_CR
4763 && (s[1] == '\n' || (s[1] == '\r' && s[2] == '\n'))
4765 && (s == PL_linestart || s[-1] == '\n') )
4767 PL_lex_formbrack = 0;
4771 if (PL_expect == XOPERATOR || !isDIGIT(s[1])) {
4777 yylval.ival = OPf_SPECIAL;
4783 if (PL_expect != XOPERATOR)
4788 case '0': case '1': case '2': case '3': case '4':
4789 case '5': case '6': case '7': case '8': case '9':
4790 s = scan_num(s, &yylval);
4791 DEBUG_T( { S_printbuf(aTHX_ "### Saw number in %s\n", s); } );
4792 if (PL_expect == XOPERATOR)
4797 s = scan_str(s,!!PL_madskills,FALSE);
4798 DEBUG_T( { S_printbuf(aTHX_ "### Saw string before %s\n", s); } );
4799 if (PL_expect == XOPERATOR) {
4800 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
4802 deprecate_old(commaless_variable_list);
4803 return REPORT(','); /* grandfather non-comma-format format */
4810 yylval.ival = OP_CONST;
4811 TERM(sublex_start());
4814 s = scan_str(s,!!PL_madskills,FALSE);
4815 DEBUG_T( { S_printbuf(aTHX_ "### Saw string before %s\n", s); } );
4816 if (PL_expect == XOPERATOR) {
4817 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
4819 deprecate_old(commaless_variable_list);
4820 return REPORT(','); /* grandfather non-comma-format format */
4827 yylval.ival = OP_CONST;
4828 /* FIXME. I think that this can be const if char *d is replaced by
4829 more localised variables. */
4830 for (d = SvPV(PL_lex_stuff, len); len; len--, d++) {
4831 if (*d == '$' || *d == '@' || *d == '\\' || !UTF8_IS_INVARIANT((U8)*d)) {
4832 yylval.ival = OP_STRINGIFY;
4836 TERM(sublex_start());
4839 s = scan_str(s,!!PL_madskills,FALSE);
4840 DEBUG_T( { S_printbuf(aTHX_ "### Saw backtick string before %s\n", s); } );
4841 if (PL_expect == XOPERATOR)
4842 no_op("Backticks",s);
4845 yylval.ival = OP_BACKTICK;
4847 TERM(sublex_start());
4851 if (PL_lex_inwhat && isDIGIT(*s) && ckWARN(WARN_SYNTAX))
4852 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),"Can't use \\%c to mean $%c in expression",
4854 if (PL_expect == XOPERATOR)
4855 no_op("Backslash",s);
4859 if (isDIGIT(s[1]) && PL_expect != XOPERATOR) {
4860 char *start = s + 2;
4861 while (isDIGIT(*start) || *start == '_')
4863 if (*start == '.' && isDIGIT(start[1])) {
4864 s = scan_num(s, &yylval);
4867 /* avoid v123abc() or $h{v1}, allow C<print v10;> */
4868 else if (!isALPHA(*start) && (PL_expect == XTERM
4869 || PL_expect == XREF || PL_expect == XSTATE
4870 || PL_expect == XTERMORDORDOR)) {
4871 /* XXX Use gv_fetchpvn rather than stomping on a const string */
4872 const char c = *start;
4875 gv = gv_fetchpv(s, 0, SVt_PVCV);
4878 s = scan_num(s, &yylval);
4885 if (isDIGIT(s[1]) && PL_expect == XOPERATOR) {
4921 I32 orig_keyword = 0;
4926 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
4928 /* Some keywords can be followed by any delimiter, including ':' */
4929 tmp = ((len == 1 && strchr("msyq", PL_tokenbuf[0])) ||
4930 (len == 2 && ((PL_tokenbuf[0] == 't' && PL_tokenbuf[1] == 'r') ||
4931 (PL_tokenbuf[0] == 'q' &&
4932 strchr("qwxr", PL_tokenbuf[1])))));
4934 /* x::* is just a word, unless x is "CORE" */
4935 if (!tmp && *s == ':' && s[1] == ':' && strNE(PL_tokenbuf, "CORE"))
4939 while (d < PL_bufend && isSPACE(*d))
4940 d++; /* no comments skipped here, or s### is misparsed */
4942 /* Is this a label? */
4943 if (!tmp && PL_expect == XSTATE
4944 && d < PL_bufend && *d == ':' && *(d + 1) != ':') {
4946 yylval.pval = savepv(PL_tokenbuf);
4951 /* Check for keywords */
4952 tmp = keyword(PL_tokenbuf, len);
4954 /* Is this a word before a => operator? */
4955 if (*d == '=' && d[1] == '>') {
4958 = (OP*)newSVOP(OP_CONST, 0,
4959 S_newSV_maybe_utf8(aTHX_ PL_tokenbuf, len));
4960 yylval.opval->op_private = OPpCONST_BARE;
4964 if (tmp < 0) { /* second-class keyword? */
4965 GV *ogv = NULL; /* override (winner) */
4966 GV *hgv = NULL; /* hidden (loser) */
4967 if (PL_expect != XOPERATOR && (*s != ':' || s[1] != ':')) {
4969 if ((gv = gv_fetchpvn_flags(PL_tokenbuf, len, 0, SVt_PVCV)) &&
4972 if (GvIMPORTED_CV(gv))
4974 else if (! CvMETHOD(cv))
4978 (gvp = (GV**)hv_fetch(PL_globalstash,PL_tokenbuf,len,FALSE)) &&
4979 (gv = *gvp) != (GV*)&PL_sv_undef &&
4980 GvCVu(gv) && GvIMPORTED_CV(gv))
4987 tmp = 0; /* overridden by import or by GLOBAL */
4990 && -tmp==KEY_lock /* XXX generalizable kludge */
4992 && !hv_fetchs(GvHVn(PL_incgv), "Thread.pm", FALSE))
4994 tmp = 0; /* any sub overrides "weak" keyword */
4996 else { /* no override */
4998 if (tmp == KEY_dump && ckWARN(WARN_MISC)) {
4999 Perl_warner(aTHX_ packWARN(WARN_MISC),
5000 "dump() better written as CORE::dump()");
5004 if (hgv && tmp != KEY_x && tmp != KEY_CORE
5005 && ckWARN(WARN_AMBIGUOUS)) /* never ambiguous */
5006 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
5007 "Ambiguous call resolved as CORE::%s(), %s",
5008 GvENAME(hgv), "qualify as such or use &");
5015 default: /* not a keyword */
5016 /* Trade off - by using this evil construction we can pull the
5017 variable gv into the block labelled keylookup. If not, then
5018 we have to give it function scope so that the goto from the
5019 earlier ':' case doesn't bypass the initialisation. */
5021 just_a_word_zero_gv:
5029 const char lastchar = (PL_bufptr == PL_oldoldbufptr ? 0 : PL_bufptr[-1]);
5032 SV *nextPL_nextwhite = 0;
5036 /* Get the rest if it looks like a package qualifier */
5038 if (*s == '\'' || (*s == ':' && s[1] == ':')) {
5040 s = scan_word(s, PL_tokenbuf + len, sizeof PL_tokenbuf - len,
5043 Perl_croak(aTHX_ "Bad name after %s%s", PL_tokenbuf,
5044 *s == '\'' ? "'" : "::");
5049 if (PL_expect == XOPERATOR) {
5050 if (PL_bufptr == PL_linestart) {
5051 CopLINE_dec(PL_curcop);
5052 Perl_warner(aTHX_ packWARN(WARN_SEMICOLON), PL_warn_nosemi);
5053 CopLINE_inc(PL_curcop);
5056 no_op("Bareword",s);
5059 /* Look for a subroutine with this name in current package,
5060 unless name is "Foo::", in which case Foo is a bearword
5061 (and a package name). */
5063 if (len > 2 && !PL_madskills &&
5064 PL_tokenbuf[len - 2] == ':' && PL_tokenbuf[len - 1] == ':')
5066 if (ckWARN(WARN_BAREWORD)
5067 && ! gv_fetchpvn_flags(PL_tokenbuf, len, 0, SVt_PVHV))
5068 Perl_warner(aTHX_ packWARN(WARN_BAREWORD),
5069 "Bareword \"%s\" refers to nonexistent package",
5072 PL_tokenbuf[len] = '\0';
5078 /* Mustn't actually add anything to a symbol table.
5079 But also don't want to "initialise" any placeholder
5080 constants that might already be there into full
5081 blown PVGVs with attached PVCV. */
5082 gv = gv_fetchpvn_flags(PL_tokenbuf, len,
5083 GV_NOADD_NOINIT, SVt_PVCV);
5088 /* if we saw a global override before, get the right name */
5091 sv = newSVpvs("CORE::GLOBAL::");
5092 sv_catpv(sv,PL_tokenbuf);
5095 /* If len is 0, newSVpv does strlen(), which is correct.
5096 If len is non-zero, then it will be the true length,
5097 and so the scalar will be created correctly. */
5098 sv = newSVpv(PL_tokenbuf,len);
5101 if (PL_madskills && !PL_thistoken) {
5102 char *start = SvPVX(PL_linestr) + PL_realtokenstart;
5103 PL_thistoken = newSVpv(start,s - start);
5104 PL_realtokenstart = s - SvPVX(PL_linestr);
5108 /* Presume this is going to be a bareword of some sort. */
5111 yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
5112 yylval.opval->op_private = OPpCONST_BARE;
5113 /* UTF-8 package name? */
5114 if (UTF && !IN_BYTES &&
5115 is_utf8_string((U8*)SvPVX_const(sv), SvCUR(sv)))
5118 /* And if "Foo::", then that's what it certainly is. */
5123 /* Do the explicit type check so that we don't need to force
5124 the initialisation of the symbol table to have a real GV.
5125 Beware - gv may not really be a PVGV, cv may not really be
5126 a PVCV, (because of the space optimisations that gv_init
5127 understands) But they're true if for this symbol there is
5128 respectively a typeglob and a subroutine.
5130 cv = gv ? ((SvTYPE(gv) == SVt_PVGV)
5131 /* Real typeglob, so get the real subroutine: */
5133 /* A proxy for a subroutine in this package? */
5134 : SvOK(gv) ? (CV *) gv : NULL)
5137 /* See if it's the indirect object for a list operator. */
5139 if (PL_oldoldbufptr &&
5140 PL_oldoldbufptr < PL_bufptr &&
5141 (PL_oldoldbufptr == PL_last_lop
5142 || PL_oldoldbufptr == PL_last_uni) &&
5143 /* NO SKIPSPACE BEFORE HERE! */
5144 (PL_expect == XREF ||
5145 ((PL_opargs[PL_last_lop_op] >> OASHIFT)& 7) == OA_FILEREF))
5147 bool immediate_paren = *s == '(';
5149 /* (Now we can afford to cross potential line boundary.) */
5150 s = SKIPSPACE2(s,nextPL_nextwhite);
5152 PL_nextwhite = nextPL_nextwhite; /* assume no & deception */
5155 /* Two barewords in a row may indicate method call. */
5157 if ((isIDFIRST_lazy_if(s,UTF) || *s == '$') &&
5158 (tmp = intuit_method(s, gv, cv)))
5161 /* If not a declared subroutine, it's an indirect object. */
5162 /* (But it's an indir obj regardless for sort.) */
5163 /* Also, if "_" follows a filetest operator, it's a bareword */
5166 ( !immediate_paren && (PL_last_lop_op == OP_SORT ||
5168 (PL_last_lop_op != OP_MAPSTART &&
5169 PL_last_lop_op != OP_GREPSTART))))
5170 || (PL_tokenbuf[0] == '_' && PL_tokenbuf[1] == '\0'
5171 && ((PL_opargs[PL_last_lop_op] & OA_CLASS_MASK) == OA_FILESTATOP))
5174 PL_expect = (PL_last_lop == PL_oldoldbufptr) ? XTERM : XOPERATOR;
5179 PL_expect = XOPERATOR;
5182 s = SKIPSPACE2(s,nextPL_nextwhite);
5183 PL_nextwhite = nextPL_nextwhite;
5188 /* Is this a word before a => operator? */
5189 if (*s == '=' && s[1] == '>' && !pkgname) {
5191 sv_setpv(((SVOP*)yylval.opval)->op_sv, PL_tokenbuf);
5192 if (UTF && !IN_BYTES && is_utf8_string((U8*)PL_tokenbuf, len))
5193 SvUTF8_on(((SVOP*)yylval.opval)->op_sv);
5197 /* If followed by a paren, it's certainly a subroutine. */
5201 for (d = s + 1; SPACE_OR_TAB(*d); d++) ;
5202 if (*d == ')' && (sv = gv_const_sv(gv))) {
5206 char *par = SvPVX(PL_linestr) + PL_realtokenstart;
5207 sv_catpvn(PL_thistoken, par, s - par);
5209 sv_free(PL_nextwhite);
5219 PL_nextwhite = PL_thiswhite;
5222 start_force(PL_curforce);
5224 NEXTVAL_NEXTTOKE.opval = yylval.opval;
5225 PL_expect = XOPERATOR;
5228 PL_nextwhite = nextPL_nextwhite;
5229 curmad('X', PL_thistoken);
5230 PL_thistoken = newSVpvn("",0);
5238 /* If followed by var or block, call it a method (unless sub) */
5240 if ((*s == '$' || *s == '{') && (!gv || !cv)) {
5241 PL_last_lop = PL_oldbufptr;
5242 PL_last_lop_op = OP_METHOD;
5246 /* If followed by a bareword, see if it looks like indir obj. */
5249 && (isIDFIRST_lazy_if(s,UTF) || *s == '$')
5250 && (tmp = intuit_method(s, gv, cv)))
5253 /* Not a method, so call it a subroutine (if defined) */
5256 if (lastchar == '-' && ckWARN_d(WARN_AMBIGUOUS))
5257 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
5258 "Ambiguous use of -%s resolved as -&%s()",
5259 PL_tokenbuf, PL_tokenbuf);
5260 /* Check for a constant sub */
5261 if ((sv = gv_const_sv(gv))) {
5263 SvREFCNT_dec(((SVOP*)yylval.opval)->op_sv);
5264 ((SVOP*)yylval.opval)->op_sv = SvREFCNT_inc_simple(sv);
5265 yylval.opval->op_private = 0;
5269 /* Resolve to GV now. */
5270 if (SvTYPE(gv) != SVt_PVGV) {
5271 gv = gv_fetchpv(PL_tokenbuf, 0, SVt_PVCV);
5272 assert (SvTYPE(gv) == SVt_PVGV);
5273 /* cv must have been some sort of placeholder, so
5274 now needs replacing with a real code reference. */
5278 op_free(yylval.opval);
5279 yylval.opval = newCVREF(0, newGVOP(OP_GV, 0, gv));
5280 yylval.opval->op_private |= OPpENTERSUB_NOPAREN;
5281 PL_last_lop = PL_oldbufptr;
5282 PL_last_lop_op = OP_ENTERSUB;
5283 /* Is there a prototype? */
5290 const char *proto = SvPV_const((SV*)cv, protolen);
5293 if (*proto == '$' && proto[1] == '\0')
5295 while (*proto == ';')
5297 if (*proto == '&' && *s == '{') {
5298 sv_setpv(PL_subname, PL_curstash ?
5299 "__ANON__" : "__ANON__::__ANON__");
5306 PL_nextwhite = PL_thiswhite;
5309 start_force(PL_curforce);
5310 NEXTVAL_NEXTTOKE.opval = yylval.opval;
5313 PL_nextwhite = nextPL_nextwhite;
5314 curmad('X', PL_thistoken);
5315 PL_thistoken = newSVpvn("",0);
5322 /* Guess harder when madskills require "best effort". */
5323 if (PL_madskills && (!gv || !GvCVu(gv))) {
5324 int probable_sub = 0;
5325 if (strchr("\"'`$@%0123456789!*+{[<", *s))
5327 else if (isALPHA(*s)) {
5331 d = scan_word(d, tmpbuf, sizeof tmpbuf, TRUE, &tmplen);
5332 if (!keyword(tmpbuf,tmplen))
5335 while (d < PL_bufend && isSPACE(*d))
5337 if (*d == '=' && d[1] == '>')
5342 gv = gv_fetchpv(PL_tokenbuf, TRUE, SVt_PVCV);
5343 op_free(yylval.opval);
5344 yylval.opval = newCVREF(0, newGVOP(OP_GV, 0, gv));
5345 yylval.opval->op_private |= OPpENTERSUB_NOPAREN;
5346 PL_last_lop = PL_oldbufptr;
5347 PL_last_lop_op = OP_ENTERSUB;
5348 PL_nextwhite = PL_thiswhite;
5350 start_force(PL_curforce);
5351 NEXTVAL_NEXTTOKE.opval = yylval.opval;
5353 PL_nextwhite = nextPL_nextwhite;
5354 curmad('X', PL_thistoken);
5355 PL_thistoken = newSVpvn("",0);
5360 NEXTVAL_NEXTTOKE.opval = yylval.opval;
5367 /* Call it a bare word */
5369 if (PL_hints & HINT_STRICT_SUBS)
5370 yylval.opval->op_private |= OPpCONST_STRICT;
5373 if (lastchar != '-') {
5374 if (ckWARN(WARN_RESERVED)) {
5375 for (d = PL_tokenbuf; *d && isLOWER(*d); d++) ;
5376 if (!*d && !gv_stashpv(PL_tokenbuf,FALSE))
5377 Perl_warner(aTHX_ packWARN(WARN_RESERVED), PL_warn_reserved,
5384 if ((lastchar == '*' || lastchar == '%' || lastchar == '&')
5385 && ckWARN_d(WARN_AMBIGUOUS)) {
5386 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
5387 "Operator or semicolon missing before %c%s",
5388 lastchar, PL_tokenbuf);
5389 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
5390 "Ambiguous use of %c resolved as operator %c",
5391 lastchar, lastchar);
5397 yylval.opval = (OP*)newSVOP(OP_CONST, 0,
5398 newSVpv(CopFILE(PL_curcop),0));
5402 yylval.opval = (OP*)newSVOP(OP_CONST, 0,
5403 Perl_newSVpvf(aTHX_ "%"IVdf, (IV)CopLINE(PL_curcop)));
5406 case KEY___PACKAGE__:
5407 yylval.opval = (OP*)newSVOP(OP_CONST, 0,
5409 ? newSVhek(HvNAME_HEK(PL_curstash))
5416 if (PL_rsfp && (!PL_in_eval || PL_tokenbuf[2] == 'D')) {
5417 const char *pname = "main";
5418 if (PL_tokenbuf[2] == 'D')
5419 pname = HvNAME_get(PL_curstash ? PL_curstash : PL_defstash);
5420 gv = gv_fetchpv(Perl_form(aTHX_ "%s::DATA", pname), GV_ADD,
5424 GvIOp(gv) = newIO();
5425 IoIFP(GvIOp(gv)) = PL_rsfp;
5426 #if defined(HAS_FCNTL) && defined(F_SETFD)
5428 const int fd = PerlIO_fileno(PL_rsfp);
5429 fcntl(fd,F_SETFD,fd >= 3);
5432 /* Mark this internal pseudo-handle as clean */
5433 IoFLAGS(GvIOp(gv)) |= IOf_UNTAINT;
5435 IoTYPE(GvIOp(gv)) = IoTYPE_PIPE;
5436 else if ((PerlIO*)PL_rsfp == PerlIO_stdin())
5437 IoTYPE(GvIOp(gv)) = IoTYPE_STD;
5439 IoTYPE(GvIOp(gv)) = IoTYPE_RDONLY;
5440 #if defined(WIN32) && !defined(PERL_TEXTMODE_SCRIPTS)
5441 /* if the script was opened in binmode, we need to revert
5442 * it to text mode for compatibility; but only iff it has CRs
5443 * XXX this is a questionable hack at best. */
5444 if (PL_bufend-PL_bufptr > 2
5445 && PL_bufend[-1] == '\n' && PL_bufend[-2] == '\r')
5448 if (IoTYPE(GvIOp(gv)) == IoTYPE_RDONLY) {
5449 loc = PerlIO_tell(PL_rsfp);
5450 (void)PerlIO_seek(PL_rsfp, 0L, 0);
5453 if (PerlLIO_setmode(PL_rsfp, O_TEXT) != -1) {
5455 if (PerlLIO_setmode(PerlIO_fileno(PL_rsfp), O_TEXT) != -1) {
5456 #endif /* NETWARE */
5457 #ifdef PERLIO_IS_STDIO /* really? */
5458 # if defined(__BORLANDC__)
5459 /* XXX see note in do_binmode() */
5460 ((FILE*)PL_rsfp)->flags &= ~_F_BIN;
5464 PerlIO_seek(PL_rsfp, loc, 0);
5468 #ifdef PERLIO_LAYERS
5471 PerlIO_apply_layers(aTHX_ PL_rsfp, NULL, ":utf8");
5472 else if (PL_encoding) {
5479 XPUSHs(PL_encoding);
5481 call_method("name", G_SCALAR);
5485 PerlIO_apply_layers(aTHX_ PL_rsfp, NULL,
5486 Perl_form(aTHX_ ":encoding(%"SVf")",
5495 if (PL_realtokenstart >= 0) {
5496 char *tstart = SvPVX(PL_linestr) + PL_realtokenstart;
5498 PL_endwhite = newSVpvn("",0);
5499 sv_catsv(PL_endwhite, PL_thiswhite);
5501 sv_catpvn(PL_endwhite, tstart, PL_bufend - tstart);
5502 PL_realtokenstart = -1;
5504 while ((s = filter_gets(PL_endwhite, PL_rsfp,
5505 SvCUR(PL_endwhite))) != Nullch) ;
5519 if (PL_expect == XSTATE) {
5526 if (*s == ':' && s[1] == ':') {
5529 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
5530 if (!(tmp = keyword(PL_tokenbuf, len)))
5531 Perl_croak(aTHX_ "CORE::%s is not a keyword", PL_tokenbuf);
5534 else if (tmp == KEY_require || tmp == KEY_do)
5535 /* that's a way to remember we saw "CORE::" */
5548 LOP(OP_ACCEPT,XTERM);
5554 LOP(OP_ATAN2,XTERM);
5560 LOP(OP_BINMODE,XTERM);
5563 LOP(OP_BLESS,XTERM);
5572 /* When 'use switch' is in effect, continue has a dual
5573 life as a control operator. */
5575 if (!FEATURE_IS_ENABLED("switch"))
5578 /* We have to disambiguate the two senses of
5579 "continue". If the next token is a '{' then
5580 treat it as the start of a continue block;
5581 otherwise treat it as a control operator.
5593 (void)gv_fetchpvs("ENV", GV_ADD|GV_NOTQUAL, SVt_PVHV);
5610 if (!PL_cryptseen) {
5611 PL_cryptseen = TRUE;
5615 LOP(OP_CRYPT,XTERM);
5618 LOP(OP_CHMOD,XTERM);
5621 LOP(OP_CHOWN,XTERM);
5624 LOP(OP_CONNECT,XTERM);
5643 s = force_word(s,WORD,TRUE,TRUE,FALSE);
5644 if (orig_keyword == KEY_do) {
5653 PL_hints |= HINT_BLOCK_SCOPE;
5663 gv_fetchpvs("AnyDBM_File::ISA", GV_ADDMULTI, SVt_PVAV);
5664 LOP(OP_DBMOPEN,XTERM);
5670 s = force_word(s,WORD,TRUE,FALSE,FALSE);
5677 yylval.ival = CopLINE(PL_curcop);
5693 PL_expect = (*s == '{') ? XTERMBLOCK : XTERM;
5694 UNIBRACK(OP_ENTEREVAL);
5712 case KEY_endhostent:
5718 case KEY_endservent:
5721 case KEY_endprotoent:
5732 yylval.ival = CopLINE(PL_curcop);
5734 if (PL_expect == XSTATE && isIDFIRST_lazy_if(s,UTF)) {
5737 int soff = s - SvPVX(PL_linestr); /* for skipspace realloc */
5740 if ((PL_bufend - p) >= 3 &&
5741 strnEQ(p, "my", 2) && isSPACE(*(p + 2)))
5743 else if ((PL_bufend - p) >= 4 &&
5744 strnEQ(p, "our", 3) && isSPACE(*(p + 3)))
5747 if (isIDFIRST_lazy_if(p,UTF)) {
5748 p = scan_ident(p, PL_bufend,
5749 PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
5753 Perl_croak(aTHX_ "Missing $ on loop variable");
5755 s = SvPVX(PL_linestr) + soff;
5761 LOP(OP_FORMLINE,XTERM);
5767 LOP(OP_FCNTL,XTERM);
5773 LOP(OP_FLOCK,XTERM);
5782 LOP(OP_GREPSTART, XREF);
5785 s = force_word(s,WORD,TRUE,FALSE,FALSE);
5800 case KEY_getpriority:
5801 LOP(OP_GETPRIORITY,XTERM);
5803 case KEY_getprotobyname:
5806 case KEY_getprotobynumber:
5807 LOP(OP_GPBYNUMBER,XTERM);
5809 case KEY_getprotoent:
5821 case KEY_getpeername:
5822 UNI(OP_GETPEERNAME);
5824 case KEY_gethostbyname:
5827 case KEY_gethostbyaddr:
5828 LOP(OP_GHBYADDR,XTERM);
5830 case KEY_gethostent:
5833 case KEY_getnetbyname:
5836 case KEY_getnetbyaddr:
5837 LOP(OP_GNBYADDR,XTERM);
5842 case KEY_getservbyname:
5843 LOP(OP_GSBYNAME,XTERM);
5845 case KEY_getservbyport:
5846 LOP(OP_GSBYPORT,XTERM);
5848 case KEY_getservent:
5851 case KEY_getsockname:
5852 UNI(OP_GETSOCKNAME);
5854 case KEY_getsockopt:
5855 LOP(OP_GSOCKOPT,XTERM);
5870 yylval.ival = CopLINE(PL_curcop);
5881 yylval.ival = CopLINE(PL_curcop);
5885 LOP(OP_INDEX,XTERM);
5891 LOP(OP_IOCTL,XTERM);
5903 s = force_word(s,WORD,TRUE,FALSE,FALSE);
5935 LOP(OP_LISTEN,XTERM);
5944 s = scan_pat(s,OP_MATCH);
5945 TERM(sublex_start());
5948 LOP(OP_MAPSTART, XREF);
5951 LOP(OP_MKDIR,XTERM);
5954 LOP(OP_MSGCTL,XTERM);
5957 LOP(OP_MSGGET,XTERM);
5960 LOP(OP_MSGRCV,XTERM);
5963 LOP(OP_MSGSND,XTERM);
5970 if (isIDFIRST_lazy_if(s,UTF)) {
5974 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, TRUE, &len);
5975 if (len == 3 && strnEQ(PL_tokenbuf, "sub", 3))
5977 PL_in_my_stash = find_in_my_stash(PL_tokenbuf, len);
5978 if (!PL_in_my_stash) {
5982 snprintf(tmpbuf, sizeof(tmpbuf), "No such class %.1000s", PL_tokenbuf);
5984 sprintf(tmpbuf, "No such class %.1000s", PL_tokenbuf);
5985 #endif /* #ifdef USE_SNPRINTF */
5989 if (PL_madskills) { /* just add type to declarator token */
5990 sv_catsv(PL_thistoken, PL_nextwhite);
5992 sv_catpvn(PL_thistoken, start, s - start);
6000 s = force_word(s,WORD,TRUE,FALSE,FALSE);
6007 s = tokenize_use(0, s);
6011 if (*s == '(' || (s = SKIPSPACE1(s), *s == '('))
6018 if (isIDFIRST_lazy_if(s,UTF)) {
6020 for (d = s; isALNUM_lazy_if(d,UTF); d++) ;
6021 for (t=d; *t && isSPACE(*t); t++) ;
6022 if ( *t && strchr("|&*+-=!?:.", *t) && ckWARN_d(WARN_PRECEDENCE)
6024 && !(t[0] == '=' && t[1] == '>')
6026 int parms_len = (int)(d-s);
6027 Perl_warner(aTHX_ packWARN(WARN_PRECEDENCE),
6028 "Precedence problem: open %.*s should be open(%.*s)",
6029 parms_len, s, parms_len, s);
6035 yylval.ival = OP_OR;
6045 LOP(OP_OPEN_DIR,XTERM);
6048 checkcomma(s,PL_tokenbuf,"filehandle");
6052 checkcomma(s,PL_tokenbuf,"filehandle");
6071 s = force_word(s,WORD,FALSE,TRUE,FALSE);
6075 LOP(OP_PIPE_OP,XTERM);
6078 s = scan_str(s,!!PL_madskills,FALSE);
6081 yylval.ival = OP_CONST;
6082 TERM(sublex_start());
6088 s = scan_str(s,!!PL_madskills,FALSE);
6091 PL_expect = XOPERATOR;
6093 if (SvCUR(PL_lex_stuff)) {
6096 d = SvPV_force(PL_lex_stuff, len);
6098 for (; isSPACE(*d) && len; --len, ++d)
6103 if (!warned && ckWARN(WARN_QW)) {
6104 for (; !isSPACE(*d) && len; --len, ++d) {
6106 Perl_warner(aTHX_ packWARN(WARN_QW),
6107 "Possible attempt to separate words with commas");
6110 else if (*d == '#') {
6111 Perl_warner(aTHX_ packWARN(WARN_QW),
6112 "Possible attempt to put comments in qw() list");
6118 for (; !isSPACE(*d) && len; --len, ++d)
6121 sv = newSVpvn(b, d-b);
6122 if (DO_UTF8(PL_lex_stuff))
6124 words = append_elem(OP_LIST, words,
6125 newSVOP(OP_CONST, 0, tokeq(sv)));
6129 start_force(PL_curforce);
6130 NEXTVAL_NEXTTOKE.opval = words;
6135 SvREFCNT_dec(PL_lex_stuff);
6136 PL_lex_stuff = NULL;
6142 s = scan_str(s,!!PL_madskills,FALSE);
6145 yylval.ival = OP_STRINGIFY;
6146 if (SvIVX(PL_lex_stuff) == '\'')
6147 SvIV_set(PL_lex_stuff, 0); /* qq'$foo' should intepolate */
6148 TERM(sublex_start());
6151 s = scan_pat(s,OP_QR);
6152 TERM(sublex_start());
6155 s = scan_str(s,!!PL_madskills,FALSE);
6158 yylval.ival = OP_BACKTICK;
6160 TERM(sublex_start());
6168 s = force_version(s, FALSE);
6170 else if (*s != 'v' || !isDIGIT(s[1])
6171 || (s = force_version(s, TRUE), *s == 'v'))
6173 *PL_tokenbuf = '\0';
6174 s = force_word(s,WORD,TRUE,TRUE,FALSE);
6175 if (isIDFIRST_lazy_if(PL_tokenbuf,UTF))
6176 gv_stashpvn(PL_tokenbuf, strlen(PL_tokenbuf), TRUE);
6178 yyerror("<> should be quotes");
6180 if (orig_keyword == KEY_require) {
6188 PL_last_uni = PL_oldbufptr;
6189 PL_last_lop_op = OP_REQUIRE;
6191 return REPORT( (int)REQUIRE );
6197 s = force_word(s,WORD,TRUE,FALSE,FALSE);
6201 LOP(OP_RENAME,XTERM);
6210 LOP(OP_RINDEX,XTERM);
6220 UNIDOR(OP_READLINE);
6233 LOP(OP_REVERSE,XTERM);
6236 UNIDOR(OP_READLINK);
6244 TERM(sublex_start());
6246 TOKEN(1); /* force error */
6249 checkcomma(s,PL_tokenbuf,"filehandle");
6259 LOP(OP_SELECT,XTERM);
6265 LOP(OP_SEMCTL,XTERM);
6268 LOP(OP_SEMGET,XTERM);
6271 LOP(OP_SEMOP,XTERM);
6277 LOP(OP_SETPGRP,XTERM);
6279 case KEY_setpriority:
6280 LOP(OP_SETPRIORITY,XTERM);
6282 case KEY_sethostent:
6288 case KEY_setservent:
6291 case KEY_setprotoent:
6301 LOP(OP_SEEKDIR,XTERM);
6303 case KEY_setsockopt:
6304 LOP(OP_SSOCKOPT,XTERM);
6310 LOP(OP_SHMCTL,XTERM);
6313 LOP(OP_SHMGET,XTERM);
6316 LOP(OP_SHMREAD,XTERM);
6319 LOP(OP_SHMWRITE,XTERM);
6322 LOP(OP_SHUTDOWN,XTERM);
6331 LOP(OP_SOCKET,XTERM);
6333 case KEY_socketpair:
6334 LOP(OP_SOCKPAIR,XTERM);
6337 checkcomma(s,PL_tokenbuf,"subroutine name");
6339 if (*s == ';' || *s == ')') /* probably a close */
6340 Perl_croak(aTHX_ "sort is now a reserved word");
6342 s = force_word(s,WORD,TRUE,TRUE,FALSE);
6346 LOP(OP_SPLIT,XTERM);
6349 LOP(OP_SPRINTF,XTERM);
6352 LOP(OP_SPLICE,XTERM);
6367 LOP(OP_SUBSTR,XTERM);
6373 char tmpbuf[sizeof PL_tokenbuf];
6374 SSize_t tboffset = 0;
6375 expectation attrful;
6376 bool have_name, have_proto, bad_proto;
6377 const int key = tmp;
6382 char *tstart = SvPVX(PL_linestr) + PL_realtokenstart;
6383 SV *subtoken = newSVpvn(tstart, s - tstart);
6387 s = SKIPSPACE2(s,tmpwhite);
6392 if (isIDFIRST_lazy_if(s,UTF) || *s == '\'' ||
6393 (*s == ':' && s[1] == ':'))
6400 attrful = XATTRBLOCK;
6401 /* remember buffer pos'n for later force_word */
6402 tboffset = s - PL_oldbufptr;
6403 d = scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
6406 nametoke = newSVpvn(s, d - s);
6408 if (strchr(tmpbuf, ':'))
6409 sv_setpv(PL_subname, tmpbuf);
6411 sv_setsv(PL_subname,PL_curstname);
6412 sv_catpvs(PL_subname,"::");
6413 sv_catpvn(PL_subname,tmpbuf,len);
6420 CURMAD('X', nametoke);
6421 CURMAD('_', tmpwhite);
6422 (void) force_word(PL_oldbufptr + tboffset, WORD,
6425 s = SKIPSPACE2(d,tmpwhite);
6432 Perl_croak(aTHX_ "Missing name in \"my sub\"");
6433 PL_expect = XTERMBLOCK;
6434 attrful = XATTRTERM;
6435 sv_setpvn(PL_subname,"?",1);
6439 if (key == KEY_format) {
6441 PL_lex_formbrack = PL_lex_brackets + 1;
6443 PL_thistoken = subtoken;
6447 (void) force_word(PL_oldbufptr + tboffset, WORD,
6453 /* Look for a prototype */
6457 s = scan_str(s,!!PL_madskills,FALSE);
6459 Perl_croak(aTHX_ "Prototype not terminated");
6460 /* strip spaces and check for bad characters */
6461 d = SvPVX(PL_lex_stuff);
6464 for (p = d; *p; ++p) {
6467 if (!strchr("$@%*;[]&\\", *p))
6472 if (bad_proto && ckWARN(WARN_SYNTAX))
6473 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
6474 "Illegal character in prototype for %"SVf" : %s",
6475 (void*)PL_subname, d);
6476 SvCUR_set(PL_lex_stuff, tmp);
6481 CURMAD('q', PL_thisopen);
6482 CURMAD('_', tmpwhite);
6483 CURMAD('=', PL_thisstuff);
6484 CURMAD('Q', PL_thisclose);
6485 NEXTVAL_NEXTTOKE.opval =
6486 (OP*)newSVOP(OP_CONST, 0, PL_lex_stuff);
6487 PL_lex_stuff = Nullsv;
6490 s = SKIPSPACE2(s,tmpwhite);
6498 if (*s == ':' && s[1] != ':')
6499 PL_expect = attrful;
6500 else if (*s != '{' && key == KEY_sub) {
6502 Perl_croak(aTHX_ "Illegal declaration of anonymous subroutine");
6504 Perl_croak(aTHX_ "Illegal declaration of subroutine %"SVf, (void*)PL_subname);
6511 curmad('^', newSVpvn("",0));
6512 CURMAD('_', tmpwhite);
6516 PL_thistoken = subtoken;
6519 NEXTVAL_NEXTTOKE.opval =
6520 (OP*)newSVOP(OP_CONST, 0, PL_lex_stuff);
6521 PL_lex_stuff = NULL;
6526 sv_setpv(PL_subname,
6527 PL_curstash ? "__ANON__" : "__ANON__::__ANON__");
6531 (void) force_word(PL_oldbufptr + tboffset, WORD,
6541 LOP(OP_SYSTEM,XREF);
6544 LOP(OP_SYMLINK,XTERM);
6547 LOP(OP_SYSCALL,XTERM);
6550 LOP(OP_SYSOPEN,XTERM);
6553 LOP(OP_SYSSEEK,XTERM);
6556 LOP(OP_SYSREAD,XTERM);
6559 LOP(OP_SYSWRITE,XTERM);
6563 TERM(sublex_start());
6584 LOP(OP_TRUNCATE,XTERM);
6596 yylval.ival = CopLINE(PL_curcop);
6600 yylval.ival = CopLINE(PL_curcop);
6604 LOP(OP_UNLINK,XTERM);
6610 LOP(OP_UNPACK,XTERM);
6613 LOP(OP_UTIME,XTERM);
6619 LOP(OP_UNSHIFT,XTERM);
6622 s = tokenize_use(1, s);
6632 yylval.ival = CopLINE(PL_curcop);
6636 yylval.ival = CopLINE(PL_curcop);
6640 PL_hints |= HINT_BLOCK_SCOPE;
6647 LOP(OP_WAITPID,XTERM);
6656 ctl_l[0] = toCTRL('L');
6658 gv_fetchpvn_flags(ctl_l, 1, GV_ADD|GV_NOTQUAL, SVt_PV);
6661 /* Make sure $^L is defined */
6662 gv_fetchpvs("\f", GV_ADD|GV_NOTQUAL, SVt_PV);
6667 if (PL_expect == XOPERATOR)
6673 yylval.ival = OP_XOR;
6678 TERM(sublex_start());
6683 #pragma segment Main
6687 S_pending_ident(pTHX)
6691 register PADOFFSET tmp = 0;
6692 /* pit holds the identifier we read and pending_ident is reset */
6693 char pit = PL_pending_ident;
6694 PL_pending_ident = 0;
6696 /* PL_realtokenstart = realtokenend = PL_bufptr - SvPVX(PL_linestr); */
6697 DEBUG_T({ PerlIO_printf(Perl_debug_log,
6698 "### Pending identifier '%s'\n", PL_tokenbuf); });
6700 /* if we're in a my(), we can't allow dynamics here.
6701 $foo'bar has already been turned into $foo::bar, so
6702 just check for colons.
6704 if it's a legal name, the OP is a PADANY.
6707 if (PL_in_my == KEY_our) { /* "our" is merely analogous to "my" */
6708 if (strchr(PL_tokenbuf,':'))
6709 yyerror(Perl_form(aTHX_ "No package name allowed for "
6710 "variable %s in \"our\"",
6712 tmp = allocmy(PL_tokenbuf);
6715 if (strchr(PL_tokenbuf,':'))
6716 yyerror(Perl_form(aTHX_ PL_no_myglob,
6717 PL_in_my == KEY_my ? "my" : "state", PL_tokenbuf));
6719 yylval.opval = newOP(OP_PADANY, 0);
6720 yylval.opval->op_targ = allocmy(PL_tokenbuf);
6726 build the ops for accesses to a my() variable.
6728 Deny my($a) or my($b) in a sort block, *if* $a or $b is
6729 then used in a comparison. This catches most, but not
6730 all cases. For instance, it catches
6731 sort { my($a); $a <=> $b }
6733 sort { my($a); $a < $b ? -1 : $a == $b ? 0 : 1; }
6734 (although why you'd do that is anyone's guess).
6737 if (!strchr(PL_tokenbuf,':')) {
6739 tmp = pad_findmy(PL_tokenbuf);
6740 if (tmp != NOT_IN_PAD) {
6741 /* might be an "our" variable" */
6742 if (PAD_COMPNAME_FLAGS_isOUR(tmp)) {
6743 /* build ops for a bareword */
6744 HV * const stash = PAD_COMPNAME_OURSTASH(tmp);
6745 HEK * const stashname = HvNAME_HEK(stash);
6746 SV * const sym = newSVhek(stashname);
6747 sv_catpvs(sym, "::");
6748 sv_catpv(sym, PL_tokenbuf+1);
6749 yylval.opval = (OP*)newSVOP(OP_CONST, 0, sym);
6750 yylval.opval->op_private = OPpCONST_ENTERED;
6753 ? (GV_ADDMULTI | GV_ADDINEVAL)
6756 ((PL_tokenbuf[0] == '$') ? SVt_PV
6757 : (PL_tokenbuf[0] == '@') ? SVt_PVAV
6762 /* if it's a sort block and they're naming $a or $b */
6763 if (PL_last_lop_op == OP_SORT &&
6764 PL_tokenbuf[0] == '$' &&
6765 (PL_tokenbuf[1] == 'a' || PL_tokenbuf[1] == 'b')
6768 for (d = PL_in_eval ? PL_oldoldbufptr : PL_linestart;
6769 d < PL_bufend && *d != '\n';
6772 if (strnEQ(d,"<=>",3) || strnEQ(d,"cmp",3)) {
6773 Perl_croak(aTHX_ "Can't use \"my %s\" in sort comparison",
6779 yylval.opval = newOP(OP_PADANY, 0);
6780 yylval.opval->op_targ = tmp;
6786 Whine if they've said @foo in a doublequoted string,
6787 and @foo isn't a variable we can find in the symbol
6790 if (pit == '@' && PL_lex_state != LEX_NORMAL && !PL_lex_brackets) {
6791 GV *gv = gv_fetchpv(PL_tokenbuf+1, 0, SVt_PVAV);
6792 if ((!gv || ((PL_tokenbuf[0] == '@') ? !GvAV(gv) : !GvHV(gv)))
6793 && ckWARN(WARN_AMBIGUOUS))
6795 /* Downgraded from fatal to warning 20000522 mjd */
6796 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
6797 "Possible unintended interpolation of %s in string",
6802 /* build ops for a bareword */
6803 yylval.opval = (OP*)newSVOP(OP_CONST, 0, newSVpv(PL_tokenbuf+1, 0));
6804 yylval.opval->op_private = OPpCONST_ENTERED;
6807 /* If the identifier refers to a stash, don't autovivify it.
6808 * Change 24660 had the side effect of causing symbol table
6809 * hashes to always be defined, even if they were freshly
6810 * created and the only reference in the entire program was
6811 * the single statement with the defined %foo::bar:: test.
6812 * It appears that all code in the wild doing this actually
6813 * wants to know whether sub-packages have been loaded, so
6814 * by avoiding auto-vivifying symbol tables, we ensure that
6815 * defined %foo::bar:: continues to be false, and the existing
6816 * tests still give the expected answers, even though what
6817 * they're actually testing has now changed subtly.
6819 (*PL_tokenbuf == '%' && *(d = PL_tokenbuf + strlen(PL_tokenbuf) - 1) == ':' && d[-1] == ':'
6821 : PL_in_eval ? (GV_ADDMULTI | GV_ADDINEVAL) : GV_ADD),
6822 ((PL_tokenbuf[0] == '$') ? SVt_PV
6823 : (PL_tokenbuf[0] == '@') ? SVt_PVAV
6829 * The following code was generated by perl_keyword.pl.
6833 Perl_keyword (pTHX_ const char *name, I32 len)
6838 case 1: /* 5 tokens of length 1 */
6870 case 2: /* 18 tokens of length 2 */
7016 case 3: /* 29 tokens of length 3 */
7020 if (name[1] == 'N' &&
7083 if (name[1] == 'i' &&
7105 return (FEATURE_IS_ENABLED("err") ? -KEY_err : 0);
7123 if (name[1] == 'o' &&
7132 if (name[1] == 'e' &&
7141 if (name[1] == 'n' &&
7150 if (name[1] == 'o' &&
7159 if (name[1] == 'a' &&
7168 if (name[1] == 'o' &&
7230 if (name[1] == 'e' &&
7244 return (FEATURE_IS_ENABLED("say") ? -KEY_say : 0);
7270 if (name[1] == 'i' &&
7279 if (name[1] == 's' &&
7288 if (name[1] == 'e' &&
7297 if (name[1] == 'o' &&
7309 case 4: /* 41 tokens of length 4 */
7313 if (name[1] == 'O' &&
7323 if (name[1] == 'N' &&
7333 if (name[1] == 'i' &&
7343 if (name[1] == 'h' &&
7353 if (name[1] == 'u' &&
7366 if (name[2] == 'c' &&
7375 if (name[2] == 's' &&
7384 if (name[2] == 'a' &&
7420 if (name[1] == 'o' &&
7433 if (name[2] == 't' &&
7442 if (name[2] == 'o' &&
7451 if (name[2] == 't' &&
7460 if (name[2] == 'e' &&
7473 if (name[1] == 'o' &&
7486 if (name[2] == 'y' &&
7495 if (name[2] == 'l' &&
7511 if (name[2] == 's' &&
7520 if (name[2] == 'n' &&
7529 if (name[2] == 'c' &&
7542 if (name[1] == 'e' &&
7552 if (name[1] == 'p' &&
7565 if (name[2] == 'c' &&
7574 if (name[2] == 'p' &&
7583 if (name[2] == 's' &&
7599 if (name[2] == 'n' &&
7669 if (name[2] == 'r' &&
7678 if (name[2] == 'r' &&
7687 if (name[2] == 'a' &&
7703 if (name[2] == 'l' &&
7765 if (name[2] == 'e' &&
7768 return (FEATURE_IS_ENABLED("switch") ? KEY_when : 0);
7781 case 5: /* 39 tokens of length 5 */
7785 if (name[1] == 'E' &&
7796 if (name[1] == 'H' &&
7810 if (name[2] == 'a' &&
7820 if (name[2] == 'a' &&
7837 if (name[2] == 'e' &&
7847 if (name[2] == 'e' &&
7851 return (FEATURE_IS_ENABLED("switch") ? -KEY_break : 0);
7867 if (name[3] == 'i' &&
7876 if (name[3] == 'o' &&
7912 if (name[2] == 'o' &&
7922 if (name[2] == 'y' &&
7936 if (name[1] == 'l' &&
7950 if (name[2] == 'n' &&
7960 if (name[2] == 'o' &&
7974 if (name[1] == 'i' &&
7979 return (FEATURE_IS_ENABLED("switch") ? KEY_given : 0);
7988 if (name[2] == 'd' &&
7998 if (name[2] == 'c' &&
8015 if (name[2] == 'c' &&
8025 if (name[2] == 't' &&
8039 if (name[1] == 'k' &&
8050 if (name[1] == 'r' &&
8064 if (name[2] == 's' &&
8074 if (name[2] == 'd' &&
8091 if (name[2] == 'm' &&
8101 if (name[2] == 'i' &&
8111 if (name[2] == 'e' &&
8121 if (name[2] == 'l' &&
8131 if (name[2] == 'a' &&
8144 if (name[3] == 't' &&
8147 return (FEATURE_IS_ENABLED("state") ? KEY_state : 0);
8153 if (name[3] == 'd' &&
8170 if (name[1] == 'i' &&
8184 if (name[2] == 'a' &&
8197 if (name[3] == 'e' &&
8232 if (name[2] == 'i' &&
8249 if (name[2] == 'i' &&
8259 if (name[2] == 'i' &&
8276 case 6: /* 33 tokens of length 6 */
8280 if (name[1] == 'c' &&
8295 if (name[2] == 'l' &&
8306 if (name[2] == 'r' &&
8321 if (name[1] == 'e' &&
8336 if (name[2] == 's' &&
8341 if(ckWARN_d(WARN_SYNTAX))
8342 Perl_warner(aTHX_ packWARN(WARN_SYNTAX), "elseif should be elsif");
8348 if (name[2] == 'i' &&
8366 if (name[2] == 'l' &&
8377 if (name[2] == 'r' &&
8392 if (name[1] == 'm' &&
8407 if (name[2] == 'n' &&
8418 if (name[2] == 's' &&
8433 if (name[1] == 's' &&
8439 if (name[4] == 't' &&
8448 if (name[4] == 'e' &&
8457 if (name[4] == 'c' &&
8466 if (name[4] == 'n' &&
8482 if (name[1] == 'r' &&
8500 if (name[3] == 'a' &&
8510 if (name[3] == 'u' &&
8524 if (name[2] == 'n' &&
8542 if (name[2] == 'a' &&
8556 if (name[3] == 'e' &&
8569 if (name[4] == 't' &&
8578 if (name[4] == 'e' &&
8600 if (name[4] == 't' &&
8609 if (name[4] == 'e' &&
8625 if (name[2] == 'c' &&
8636 if (name[2] == 'l' &&
8647 if (name[2] == 'b' &&
8658 if (name[2] == 's' &&
8681 if (name[4] == 's' &&
8690 if (name[4] == 'n' &&
8703 if (name[3] == 'a' &&
8720 if (name[1] == 'a' &&
8735 case 7: /* 29 tokens of length 7 */
8739 if (name[1] == 'E' &&
8752 if (name[1] == '_' &&
8765 if (name[1] == 'i' &&
8772 return -KEY_binmode;
8778 if (name[1] == 'o' &&
8785 return -KEY_connect;
8794 if (name[2] == 'm' &&
8800 return -KEY_dbmopen;
8811 if (name[4] == 'u' &&
8815 return (FEATURE_IS_ENABLED("switch") ? KEY_default : 0);
8821 if (name[4] == 'n' &&
8842 if (name[1] == 'o' &&
8855 if (name[1] == 'e' &&
8862 if (name[5] == 'r' &&
8865 return -KEY_getpgrp;
8871 if (name[5] == 'i' &&
8874 return -KEY_getppid;
8887 if (name[1] == 'c' &&
8894 return -KEY_lcfirst;
8900 if (name[1] == 'p' &&
8907 return -KEY_opendir;
8913 if (name[1] == 'a' &&
8931 if (name[3] == 'd' &&
8936 return -KEY_readdir;
8942 if (name[3] == 'u' &&
8953 if (name[3] == 'e' &&
8958 return -KEY_reverse;
8977 if (name[3] == 'k' &&
8982 return -KEY_seekdir;
8988 if (name[3] == 'p' &&
8993 return -KEY_setpgrp;
9003 if (name[2] == 'm' &&
9009 return -KEY_shmread;
9015 if (name[2] == 'r' &&
9021 return -KEY_sprintf;
9030 if (name[3] == 'l' &&
9035 return -KEY_symlink;
9044 if (name[4] == 'a' &&
9048 return -KEY_syscall;
9054 if (name[4] == 'p' &&
9058 return -KEY_sysopen;
9064 if (name[4] == 'e' &&
9068 return -KEY_sysread;
9074 if (name[4] == 'e' &&
9078 return -KEY_sysseek;
9096 if (name[1] == 'e' &&
9103 return -KEY_telldir;
9112 if (name[2] == 'f' &&
9118 return -KEY_ucfirst;
9124 if (name[2] == 's' &&
9130 return -KEY_unshift;
9140 if (name[1] == 'a' &&
9147 return -KEY_waitpid;
9156 case 8: /* 26 tokens of length 8 */
9160 if (name[1] == 'U' &&
9168 return KEY_AUTOLOAD;
9179 if (name[3] == 'A' &&
9185 return KEY___DATA__;
9191 if (name[3] == 'I' &&
9197 return -KEY___FILE__;
9203 if (name[3] == 'I' &&
9209 return -KEY___LINE__;
9225 if (name[2] == 'o' &&
9232 return -KEY_closedir;
9238 if (name[2] == 'n' &&
9245 return -KEY_continue;
9255 if (name[1] == 'b' &&
9263 return -KEY_dbmclose;
9269 if (name[1] == 'n' &&
9275 if (name[4] == 'r' &&
9280 return -KEY_endgrent;
9286 if (name[4] == 'w' &&
9291 return -KEY_endpwent;
9304 if (name[1] == 'o' &&
9312 return -KEY_formline;
9318 if (name[1] == 'e' &&
9329 if (name[6] == 'n' &&
9332 return -KEY_getgrent;
9338 if (name[6] == 'i' &&
9341 return -KEY_getgrgid;
9347 if (name[6] == 'a' &&
9350 return -KEY_getgrnam;
9363 if (name[4] == 'o' &&
9368 return -KEY_getlogin;
9379 if (name[6] == 'n' &&
9382 return -KEY_getpwent;
9388 if (name[6] == 'a' &&
9391 return -KEY_getpwnam;
9397 if (name[6] == 'i' &&
9400 return -KEY_getpwuid;
9420 if (name[1] == 'e' &&
9427 if (name[5] == 'i' &&
9434 return -KEY_readline;
9439 return -KEY_readlink;
9450 if (name[5] == 'i' &&
9454 return -KEY_readpipe;
9475 if (name[4] == 'r' &&
9480 return -KEY_setgrent;
9486 if (name[4] == 'w' &&
9491 return -KEY_setpwent;
9507 if (name[3] == 'w' &&
9513 return -KEY_shmwrite;
9519 if (name[3] == 't' &&
9525 return -KEY_shutdown;
9535 if (name[2] == 's' &&
9542 return -KEY_syswrite;
9552 if (name[1] == 'r' &&
9560 return -KEY_truncate;
9569 case 9: /* 8 tokens of length 9 */
9573 if (name[1] == 'n' &&
9582 return -KEY_endnetent;
9588 if (name[1] == 'e' &&
9597 return -KEY_getnetent;
9603 if (name[1] == 'o' &&
9612 return -KEY_localtime;
9618 if (name[1] == 'r' &&
9627 return KEY_prototype;
9633 if (name[1] == 'u' &&
9642 return -KEY_quotemeta;
9648 if (name[1] == 'e' &&
9657 return -KEY_rewinddir;
9663 if (name[1] == 'e' &&
9672 return -KEY_setnetent;
9678 if (name[1] == 'a' &&
9687 return -KEY_wantarray;
9696 case 10: /* 9 tokens of length 10 */
9700 if (name[1] == 'n' &&
9706 if (name[4] == 'o' &&
9713 return -KEY_endhostent;
9719 if (name[4] == 'e' &&
9726 return -KEY_endservent;
9739 if (name[1] == 'e' &&
9745 if (name[4] == 'o' &&
9752 return -KEY_gethostent;
9761 if (name[5] == 'r' &&
9767 return -KEY_getservent;
9773 if (name[5] == 'c' &&
9779 return -KEY_getsockopt;
9804 if (name[4] == 'o' &&
9811 return -KEY_sethostent;
9820 if (name[5] == 'r' &&
9826 return -KEY_setservent;
9832 if (name[5] == 'c' &&
9838 return -KEY_setsockopt;
9855 if (name[2] == 'c' &&
9864 return -KEY_socketpair;
9877 case 11: /* 8 tokens of length 11 */
9881 if (name[1] == '_' &&
9892 return -KEY___PACKAGE__;
9898 if (name[1] == 'n' &&
9909 return -KEY_endprotoent;
9915 if (name[1] == 'e' &&
9924 if (name[5] == 'e' &&
9931 return -KEY_getpeername;
9940 if (name[6] == 'o' &&
9946 return -KEY_getpriority;
9952 if (name[6] == 't' &&
9958 return -KEY_getprotoent;
9972 if (name[4] == 'o' &&
9980 return -KEY_getsockname;
9993 if (name[1] == 'e' &&
10001 if (name[6] == 'o' &&
10006 { /* setpriority */
10007 return -KEY_setpriority;
10013 if (name[6] == 't' &&
10018 { /* setprotoent */
10019 return -KEY_setprotoent;
10035 case 12: /* 2 tokens of length 12 */
10036 if (name[0] == 'g' &&
10048 if (name[9] == 'd' &&
10051 { /* getnetbyaddr */
10052 return -KEY_getnetbyaddr;
10058 if (name[9] == 'a' &&
10061 { /* getnetbyname */
10062 return -KEY_getnetbyname;
10074 case 13: /* 4 tokens of length 13 */
10075 if (name[0] == 'g' &&
10082 if (name[4] == 'o' &&
10091 if (name[10] == 'd' &&
10094 { /* gethostbyaddr */
10095 return -KEY_gethostbyaddr;
10101 if (name[10] == 'a' &&
10104 { /* gethostbyname */
10105 return -KEY_gethostbyname;
10118 if (name[4] == 'e' &&
10127 if (name[10] == 'a' &&
10130 { /* getservbyname */
10131 return -KEY_getservbyname;
10137 if (name[10] == 'o' &&
10140 { /* getservbyport */
10141 return -KEY_getservbyport;
10160 case 14: /* 1 tokens of length 14 */
10161 if (name[0] == 'g' &&
10175 { /* getprotobyname */
10176 return -KEY_getprotobyname;
10181 case 16: /* 1 tokens of length 16 */
10182 if (name[0] == 'g' &&
10198 { /* getprotobynumber */
10199 return -KEY_getprotobynumber;
10213 S_checkcomma(pTHX_ const char *s, const char *name, const char *what)
10217 if (*s == ' ' && s[1] == '(') { /* XXX gotta be a better way */
10218 if (ckWARN(WARN_SYNTAX)) {
10221 for (w = s+2; *w && level; w++) {
10224 else if (*w == ')')
10227 while (isSPACE(*w))
10229 if (!*w || !strchr(";|})]oaiuw!=", *w)) /* an advisory hack only... */
10230 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
10231 "%s (...) interpreted as function",name);
10234 while (s < PL_bufend && isSPACE(*s))
10238 while (s < PL_bufend && isSPACE(*s))
10240 if (isIDFIRST_lazy_if(s,UTF)) {
10241 const char * const w = s++;
10242 while (isALNUM_lazy_if(s,UTF))
10244 while (s < PL_bufend && isSPACE(*s))
10248 if (keyword(w, s - w))
10251 gv = gv_fetchpvn_flags(w, s - w, 0, SVt_PVCV);
10252 if (gv && GvCVu(gv))
10254 Perl_croak(aTHX_ "No comma allowed after %s", what);
10259 /* Either returns sv, or mortalizes sv and returns a new SV*.
10260 Best used as sv=new_constant(..., sv, ...).
10261 If s, pv are NULL, calls subroutine with one argument,
10262 and type is used with error messages only. */
10265 S_new_constant(pTHX_ const char *s, STRLEN len, const char *key, SV *sv, SV *pv,
10269 HV * const table = GvHV(PL_hintgv); /* ^H */
10273 const char *why1 = "", *why2 = "", *why3 = "";
10275 if (!table || !(PL_hints & HINT_LOCALIZE_HH)) {
10278 why2 = strEQ(key,"charnames")
10279 ? "(possibly a missing \"use charnames ...\")"
10281 msg = Perl_newSVpvf(aTHX_ "Constant(%s) unknown: %s",
10282 (type ? type: "undef"), why2);
10284 /* This is convoluted and evil ("goto considered harmful")
10285 * but I do not understand the intricacies of all the different
10286 * failure modes of %^H in here. The goal here is to make
10287 * the most probable error message user-friendly. --jhi */
10292 msg = Perl_newSVpvf(aTHX_ "Constant(%s): %s%s%s",
10293 (type ? type: "undef"), why1, why2, why3);
10295 yyerror(SvPVX_const(msg));
10299 cvp = hv_fetch(table, key, strlen(key), FALSE);
10300 if (!cvp || !SvOK(*cvp)) {
10303 why3 = "} is not defined";
10306 sv_2mortal(sv); /* Parent created it permanently */
10309 pv = sv_2mortal(newSVpvn(s, len));
10311 typesv = sv_2mortal(newSVpv(type, 0));
10313 typesv = &PL_sv_undef;
10315 PUSHSTACKi(PERLSI_OVERLOAD);
10327 call_sv(cv, G_SCALAR | ( PL_in_eval ? 0 : G_EVAL));
10331 /* Check the eval first */
10332 if (!PL_in_eval && SvTRUE(ERRSV)) {
10333 sv_catpvs(ERRSV, "Propagated");
10334 yyerror(SvPV_nolen_const(ERRSV)); /* Duplicates the message inside eval */
10336 res = SvREFCNT_inc_simple(sv);
10340 SvREFCNT_inc_simple_void(res);
10349 why1 = "Call to &{$^H{";
10351 why3 = "}} did not return a defined value";
10359 /* Returns a NUL terminated string, with the length of the string written to
10363 S_scan_word(pTHX_ register char *s, char *dest, STRLEN destlen, int allow_package, STRLEN *slp)
10366 register char *d = dest;
10367 register char * const e = d + destlen - 3; /* two-character token, ending NUL */
10370 Perl_croak(aTHX_ ident_too_long);
10371 if (isALNUM(*s)) /* UTF handled below */
10373 else if (*s == '\'' && allow_package && isIDFIRST_lazy_if(s+1,UTF)) {
10378 else if (*s == ':' && s[1] == ':' && allow_package && s[2] != '$') {
10382 else if (UTF && UTF8_IS_START(*s) && isALNUM_utf8((U8*)s)) {
10383 char *t = s + UTF8SKIP(s);
10384 while (UTF8_IS_CONTINUED(*t) && is_utf8_mark((U8*)t))
10386 if (d + (t - s) > e)
10387 Perl_croak(aTHX_ ident_too_long);
10388 Copy(s, d, t - s, char);
10401 S_scan_ident(pTHX_ register char *s, register const char *send, char *dest, STRLEN destlen, I32 ck_uni)
10404 char *bracket = NULL;
10406 register char *d = dest;
10407 register char * const e = d + destlen + 3; /* two-character token, ending NUL */
10412 while (isDIGIT(*s)) {
10414 Perl_croak(aTHX_ ident_too_long);
10421 Perl_croak(aTHX_ ident_too_long);
10422 if (isALNUM(*s)) /* UTF handled below */
10424 else if (*s == '\'' && isIDFIRST_lazy_if(s+1,UTF)) {
10429 else if (*s == ':' && s[1] == ':') {
10433 else if (UTF && UTF8_IS_START(*s) && isALNUM_utf8((U8*)s)) {
10434 char *t = s + UTF8SKIP(s);
10435 while (UTF8_IS_CONTINUED(*t) && is_utf8_mark((U8*)t))
10437 if (d + (t - s) > e)
10438 Perl_croak(aTHX_ ident_too_long);
10439 Copy(s, d, t - s, char);
10450 if (PL_lex_state != LEX_NORMAL)
10451 PL_lex_state = LEX_INTERPENDMAYBE;
10454 if (*s == '$' && s[1] &&
10455 (isALNUM_lazy_if(s+1,UTF) || s[1] == '$' || s[1] == '{' || strnEQ(s+1,"::",2)) )
10468 if (*d == '^' && *s && isCONTROLVAR(*s)) {
10473 if (isSPACE(s[-1])) {
10475 const char ch = *s++;
10476 if (!SPACE_OR_TAB(ch)) {
10482 if (isIDFIRST_lazy_if(d,UTF)) {
10486 while ((end < send && isALNUM_lazy_if(end,UTF)) || *end == ':') {
10487 end += UTF8SKIP(end);
10488 while (end < send && UTF8_IS_CONTINUED(*end) && is_utf8_mark((U8*)end))
10489 end += UTF8SKIP(end);
10491 Copy(s, d, end - s, char);
10496 while ((isALNUM(*s) || *s == ':') && d < e)
10499 Perl_croak(aTHX_ ident_too_long);
10502 while (s < send && SPACE_OR_TAB(*s)) s++;
10503 if ((*s == '[' || (*s == '{' && strNE(dest, "sub")))) {
10504 if (ckWARN(WARN_AMBIGUOUS) && keyword(dest, d - dest)) {
10505 const char *brack = *s == '[' ? "[...]" : "{...}";
10506 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
10507 "Ambiguous use of %c{%s%s} resolved to %c%s%s",
10508 funny, dest, brack, funny, dest, brack);
10511 PL_lex_brackstack[PL_lex_brackets++] = (char)(XOPERATOR | XFAKEBRACK);
10515 /* Handle extended ${^Foo} variables
10516 * 1999-02-27 mjd-perl-patch@plover.com */
10517 else if (!isALNUM(*d) && !isPRINT(*d) /* isCTRL(d) */
10521 while (isALNUM(*s) && d < e) {
10525 Perl_croak(aTHX_ ident_too_long);
10530 if (PL_lex_state == LEX_INTERPNORMAL && !PL_lex_brackets) {
10531 PL_lex_state = LEX_INTERPEND;
10536 if (PL_lex_state == LEX_NORMAL) {
10537 if (ckWARN(WARN_AMBIGUOUS) &&
10538 (keyword(dest, d - dest) || get_cv(dest, FALSE)))
10540 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
10541 "Ambiguous use of %c{%s} resolved to %c%s",
10542 funny, dest, funny, dest);
10547 s = bracket; /* let the parser handle it */
10551 else if (PL_lex_state == LEX_INTERPNORMAL && !PL_lex_brackets && !intuit_more(s))
10552 PL_lex_state = LEX_INTERPEND;
10557 Perl_pmflag(pTHX_ U32* pmfl, int ch)
10559 PERL_UNUSED_CONTEXT;
10562 else if (ch == 'g')
10563 *pmfl |= PMf_GLOBAL;
10564 else if (ch == 'c')
10565 *pmfl |= PMf_CONTINUE;
10566 else if (ch == 'o')
10568 else if (ch == 'm')
10569 *pmfl |= PMf_MULTILINE;
10570 else if (ch == 's')
10571 *pmfl |= PMf_SINGLELINE;
10572 else if (ch == 'x')
10573 *pmfl |= PMf_EXTENDED;
10577 S_scan_pat(pTHX_ char *start, I32 type)
10581 char *s = scan_str(start,!!PL_madskills,FALSE);
10582 const char * const valid_flags = (type == OP_QR) ? "iomsx" : "iogcmsx";
10589 const char * const delimiter = skipspace(start);
10590 Perl_croak(aTHX_ *delimiter == '?'
10591 ? "Search pattern not terminated or ternary operator parsed as search pattern"
10592 : "Search pattern not terminated" );
10595 pm = (PMOP*)newPMOP(type, 0);
10596 if (PL_multi_open == '?')
10597 pm->op_pmflags |= PMf_ONCE;
10601 while (*s && strchr(valid_flags, *s))
10602 pmflag(&pm->op_pmflags,*s++);
10604 if (PL_madskills && modstart != s) {
10605 SV* tmptoken = newSVpvn(modstart, s - modstart);
10606 append_madprops(newMADPROP('m', MAD_SV, tmptoken, 0), (OP*)pm, 0);
10609 /* issue a warning if /c is specified,but /g is not */
10610 if ((pm->op_pmflags & PMf_CONTINUE) && !(pm->op_pmflags & PMf_GLOBAL)
10611 && ckWARN(WARN_REGEXP))
10613 Perl_warner(aTHX_ packWARN(WARN_REGEXP), "Use of /c modifier is meaningless without /g" );
10616 pm->op_pmpermflags = pm->op_pmflags;
10618 PL_lex_op = (OP*)pm;
10619 yylval.ival = OP_MATCH;
10624 S_scan_subst(pTHX_ char *start)
10635 yylval.ival = OP_NULL;
10637 s = scan_str(start,!!PL_madskills,FALSE);
10640 Perl_croak(aTHX_ "Substitution pattern not terminated");
10642 if (s[-1] == PL_multi_open)
10645 if (PL_madskills) {
10646 CURMAD('q', PL_thisopen);
10647 CURMAD('_', PL_thiswhite);
10648 CURMAD('E', PL_thisstuff);
10649 CURMAD('Q', PL_thisclose);
10650 PL_realtokenstart = s - SvPVX(PL_linestr);
10654 first_start = PL_multi_start;
10655 s = scan_str(s,!!PL_madskills,FALSE);
10657 if (PL_lex_stuff) {
10658 SvREFCNT_dec(PL_lex_stuff);
10659 PL_lex_stuff = NULL;
10661 Perl_croak(aTHX_ "Substitution replacement not terminated");
10663 PL_multi_start = first_start; /* so whole substitution is taken together */
10665 pm = (PMOP*)newPMOP(OP_SUBST, 0);
10668 if (PL_madskills) {
10669 CURMAD('z', PL_thisopen);
10670 CURMAD('R', PL_thisstuff);
10671 CURMAD('Z', PL_thisclose);
10681 else if (strchr("iogcmsx", *s))
10682 pmflag(&pm->op_pmflags,*s++);
10688 if (PL_madskills) {
10690 curmad('m', newSVpvn(modstart, s - modstart));
10691 append_madprops(PL_thismad, (OP*)pm, 0);
10695 if ((pm->op_pmflags & PMf_CONTINUE) && ckWARN(WARN_REGEXP)) {
10696 Perl_warner(aTHX_ packWARN(WARN_REGEXP), "Use of /c modifier is meaningless in s///" );
10700 SV * const repl = newSVpvs("");
10702 PL_sublex_info.super_bufptr = s;
10703 PL_sublex_info.super_bufend = PL_bufend;
10705 pm->op_pmflags |= PMf_EVAL;
10707 sv_catpv(repl, es ? "eval " : "do ");
10708 sv_catpvs(repl, "{");
10709 sv_catsv(repl, PL_lex_repl);
10710 sv_catpvs(repl, "}");
10712 SvREFCNT_dec(PL_lex_repl);
10713 PL_lex_repl = repl;
10716 pm->op_pmpermflags = pm->op_pmflags;
10717 PL_lex_op = (OP*)pm;
10718 yylval.ival = OP_SUBST;
10723 S_scan_trans(pTHX_ char *start)
10736 yylval.ival = OP_NULL;
10738 s = scan_str(start,!!PL_madskills,FALSE);
10740 Perl_croak(aTHX_ "Transliteration pattern not terminated");
10742 if (s[-1] == PL_multi_open)
10745 if (PL_madskills) {
10746 CURMAD('q', PL_thisopen);
10747 CURMAD('_', PL_thiswhite);
10748 CURMAD('E', PL_thisstuff);
10749 CURMAD('Q', PL_thisclose);
10750 PL_realtokenstart = s - SvPVX(PL_linestr);
10754 s = scan_str(s,!!PL_madskills,FALSE);
10756 if (PL_lex_stuff) {
10757 SvREFCNT_dec(PL_lex_stuff);
10758 PL_lex_stuff = NULL;
10760 Perl_croak(aTHX_ "Transliteration replacement not terminated");
10762 if (PL_madskills) {
10763 CURMAD('z', PL_thisopen);
10764 CURMAD('R', PL_thisstuff);
10765 CURMAD('Z', PL_thisclose);
10768 complement = del = squash = 0;
10775 complement = OPpTRANS_COMPLEMENT;
10778 del = OPpTRANS_DELETE;
10781 squash = OPpTRANS_SQUASH;
10790 Newx(tbl, complement&&!del?258:256, short);
10791 o = newPVOP(OP_TRANS, 0, (char*)tbl);
10792 o->op_private &= ~OPpTRANS_ALL;
10793 o->op_private |= del|squash|complement|
10794 (DO_UTF8(PL_lex_stuff)? OPpTRANS_FROM_UTF : 0)|
10795 (DO_UTF8(PL_lex_repl) ? OPpTRANS_TO_UTF : 0);
10798 yylval.ival = OP_TRANS;
10801 if (PL_madskills) {
10803 curmad('m', newSVpvn(modstart, s - modstart));
10804 append_madprops(PL_thismad, o, 0);
10813 S_scan_heredoc(pTHX_ register char *s)
10817 I32 op_type = OP_SCALAR;
10821 const char *found_newline;
10825 const int outer = (PL_rsfp && !(PL_lex_inwhat == OP_SCALAR));
10827 I32 stuffstart = s - SvPVX(PL_linestr);
10830 PL_realtokenstart = -1;
10835 e = PL_tokenbuf + sizeof PL_tokenbuf - 1;
10838 for (peek = s; SPACE_OR_TAB(*peek); peek++) ;
10839 if (*peek == '`' || *peek == '\'' || *peek =='"') {
10842 s = delimcpy(d, e, s, PL_bufend, term, &len);
10852 if (!isALNUM_lazy_if(s,UTF))
10853 deprecate_old("bare << to mean <<\"\"");
10854 for (; isALNUM_lazy_if(s,UTF); s++) {
10859 if (d >= PL_tokenbuf + sizeof PL_tokenbuf - 1)
10860 Perl_croak(aTHX_ "Delimiter for here document is too long");
10863 len = d - PL_tokenbuf;
10866 if (PL_madskills) {
10867 tstart = PL_tokenbuf + !outer;
10868 PL_thisclose = newSVpvn(tstart, len - !outer);
10869 tstart = SvPVX(PL_linestr) + stuffstart;
10870 PL_thisopen = newSVpvn(tstart, s - tstart);
10871 stuffstart = s - SvPVX(PL_linestr);
10874 #ifndef PERL_STRICT_CR
10875 d = strchr(s, '\r');
10877 char * const olds = s;
10879 while (s < PL_bufend) {
10885 else if (*s == '\n' && s[1] == '\r') { /* \015\013 on a mac? */
10894 SvCUR_set(PL_linestr, PL_bufend - SvPVX_const(PL_linestr));
10901 if ( outer || !(found_newline = memchr(s, '\n', PL_bufend - s)) ) {
10902 herewas = newSVpvn(s,PL_bufend-s);
10906 herewas = newSVpvn(s-1,found_newline-s+1);
10909 herewas = newSVpvn(s,found_newline-s);
10913 if (PL_madskills) {
10914 tstart = SvPVX(PL_linestr) + stuffstart;
10916 sv_catpvn(PL_thisstuff, tstart, s - tstart);
10918 PL_thisstuff = newSVpvn(tstart, s - tstart);
10921 s += SvCUR(herewas);
10924 stuffstart = s - SvPVX(PL_linestr);
10930 tmpstr = newSV(79);
10931 sv_upgrade(tmpstr, SVt_PVIV);
10932 if (term == '\'') {
10933 op_type = OP_CONST;
10934 SvIV_set(tmpstr, -1);
10936 else if (term == '`') {
10937 op_type = OP_BACKTICK;
10938 SvIV_set(tmpstr, '\\');
10942 PL_multi_start = CopLINE(PL_curcop);
10943 PL_multi_open = PL_multi_close = '<';
10944 term = *PL_tokenbuf;
10945 if (PL_lex_inwhat == OP_SUBST && PL_in_eval && !PL_rsfp) {
10946 char * const bufptr = PL_sublex_info.super_bufptr;
10947 char * const bufend = PL_sublex_info.super_bufend;
10948 char * const olds = s - SvCUR(herewas);
10949 s = strchr(bufptr, '\n');
10953 while (s < bufend &&
10954 (*s != term || memNE(s,PL_tokenbuf,len)) ) {
10956 CopLINE_inc(PL_curcop);
10959 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
10960 missingterm(PL_tokenbuf);
10962 sv_setpvn(herewas,bufptr,d-bufptr+1);
10963 sv_setpvn(tmpstr,d+1,s-d);
10965 sv_catpvn(herewas,s,bufend-s);
10966 Copy(SvPVX_const(herewas),bufptr,SvCUR(herewas) + 1,char);
10973 while (s < PL_bufend &&
10974 (*s != term || memNE(s,PL_tokenbuf,len)) ) {
10976 CopLINE_inc(PL_curcop);
10978 if (s >= PL_bufend) {
10979 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
10980 missingterm(PL_tokenbuf);
10982 sv_setpvn(tmpstr,d+1,s-d);
10984 if (PL_madskills) {
10986 sv_catpvn(PL_thisstuff, d + 1, s - d);
10988 PL_thisstuff = newSVpvn(d + 1, s - d);
10989 stuffstart = s - SvPVX(PL_linestr);
10993 CopLINE_inc(PL_curcop); /* the preceding stmt passes a newline */
10995 sv_catpvn(herewas,s,PL_bufend-s);
10996 sv_setsv(PL_linestr,herewas);
10997 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = s = PL_linestart = SvPVX(PL_linestr);
10998 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
10999 PL_last_lop = PL_last_uni = NULL;
11002 sv_setpvn(tmpstr,"",0); /* avoid "uninitialized" warning */
11003 while (s >= PL_bufend) { /* multiple line string? */
11005 if (PL_madskills) {
11006 tstart = SvPVX(PL_linestr) + stuffstart;
11008 sv_catpvn(PL_thisstuff, tstart, PL_bufend - tstart);
11010 PL_thisstuff = newSVpvn(tstart, PL_bufend - tstart);
11014 !(PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = filter_gets(PL_linestr, PL_rsfp, 0))) {
11015 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
11016 missingterm(PL_tokenbuf);
11019 stuffstart = s - SvPVX(PL_linestr);
11021 CopLINE_inc(PL_curcop);
11022 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
11023 PL_last_lop = PL_last_uni = NULL;
11024 #ifndef PERL_STRICT_CR
11025 if (PL_bufend - PL_linestart >= 2) {
11026 if ((PL_bufend[-2] == '\r' && PL_bufend[-1] == '\n') ||
11027 (PL_bufend[-2] == '\n' && PL_bufend[-1] == '\r'))
11029 PL_bufend[-2] = '\n';
11031 SvCUR_set(PL_linestr, PL_bufend - SvPVX_const(PL_linestr));
11033 else if (PL_bufend[-1] == '\r')
11034 PL_bufend[-1] = '\n';
11036 else if (PL_bufend - PL_linestart == 1 && PL_bufend[-1] == '\r')
11037 PL_bufend[-1] = '\n';
11039 if (PERLDB_LINE && PL_curstash != PL_debstash) {
11040 SV * const sv = newSV(0);
11042 sv_upgrade(sv, SVt_PVMG);
11043 sv_setsv(sv,PL_linestr);
11044 (void)SvIOK_on(sv);
11046 av_store(CopFILEAVx(PL_curcop), (I32)CopLINE(PL_curcop),sv);
11048 if (*s == term && memEQ(s,PL_tokenbuf,len)) {
11049 STRLEN off = PL_bufend - 1 - SvPVX_const(PL_linestr);
11050 *(SvPVX(PL_linestr) + off ) = ' ';
11051 sv_catsv(PL_linestr,herewas);
11052 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
11053 s = SvPVX(PL_linestr) + off; /* In case PV of PL_linestr moved. */
11057 sv_catsv(tmpstr,PL_linestr);
11062 PL_multi_end = CopLINE(PL_curcop);
11063 if (SvCUR(tmpstr) + 5 < SvLEN(tmpstr)) {
11064 SvPV_shrink_to_cur(tmpstr);
11066 SvREFCNT_dec(herewas);
11068 if (UTF && is_utf8_string((U8*)SvPVX_const(tmpstr), SvCUR(tmpstr)))
11070 else if (PL_encoding)
11071 sv_recode_to_utf8(tmpstr, PL_encoding);
11073 PL_lex_stuff = tmpstr;
11074 yylval.ival = op_type;
11078 /* scan_inputsymbol
11079 takes: current position in input buffer
11080 returns: new position in input buffer
11081 side-effects: yylval and lex_op are set.
11086 <FH> read from filehandle
11087 <pkg::FH> read from package qualified filehandle
11088 <pkg'FH> read from package qualified filehandle
11089 <$fh> read from filehandle in $fh
11090 <*.h> filename glob
11095 S_scan_inputsymbol(pTHX_ char *start)
11098 register char *s = start; /* current position in buffer */
11102 char *d = PL_tokenbuf; /* start of temp holding space */
11103 const char * const e = PL_tokenbuf + sizeof PL_tokenbuf; /* end of temp holding space */
11105 end = strchr(s, '\n');
11108 s = delimcpy(d, e, s + 1, end, '>', &len); /* extract until > */
11110 /* die if we didn't have space for the contents of the <>,
11111 or if it didn't end, or if we see a newline
11114 if (len >= (I32)sizeof PL_tokenbuf)
11115 Perl_croak(aTHX_ "Excessively long <> operator");
11117 Perl_croak(aTHX_ "Unterminated <> operator");
11122 Remember, only scalar variables are interpreted as filehandles by
11123 this code. Anything more complex (e.g., <$fh{$num}>) will be
11124 treated as a glob() call.
11125 This code makes use of the fact that except for the $ at the front,
11126 a scalar variable and a filehandle look the same.
11128 if (*d == '$' && d[1]) d++;
11130 /* allow <Pkg'VALUE> or <Pkg::VALUE> */
11131 while (*d && (isALNUM_lazy_if(d,UTF) || *d == '\'' || *d == ':'))
11134 /* If we've tried to read what we allow filehandles to look like, and
11135 there's still text left, then it must be a glob() and not a getline.
11136 Use scan_str to pull out the stuff between the <> and treat it
11137 as nothing more than a string.
11140 if (d - PL_tokenbuf != len) {
11141 yylval.ival = OP_GLOB;
11143 s = scan_str(start,!!PL_madskills,FALSE);
11145 Perl_croak(aTHX_ "Glob not terminated");
11149 bool readline_overriden = FALSE;
11152 /* we're in a filehandle read situation */
11155 /* turn <> into <ARGV> */
11157 Copy("ARGV",d,5,char);
11159 /* Check whether readline() is overriden */
11160 gv_readline = gv_fetchpvs("readline", GV_NOTQUAL, SVt_PVCV);
11162 && GvCVu(gv_readline) && GvIMPORTED_CV(gv_readline))
11164 ((gvp = (GV**)hv_fetchs(PL_globalstash, "readline", FALSE))
11165 && (gv_readline = *gvp) != (GV*)&PL_sv_undef
11166 && GvCVu(gv_readline) && GvIMPORTED_CV(gv_readline)))
11167 readline_overriden = TRUE;
11169 /* if <$fh>, create the ops to turn the variable into a
11175 /* try to find it in the pad for this block, otherwise find
11176 add symbol table ops
11178 if ((tmp = pad_findmy(d)) != NOT_IN_PAD) {
11179 if (PAD_COMPNAME_FLAGS_isOUR(tmp)) {
11180 HV * const stash = PAD_COMPNAME_OURSTASH(tmp);
11181 HEK * const stashname = HvNAME_HEK(stash);
11182 SV * const sym = sv_2mortal(newSVhek(stashname));
11183 sv_catpvs(sym, "::");
11184 sv_catpv(sym, d+1);
11189 OP * const o = newOP(OP_PADSV, 0);
11191 PL_lex_op = readline_overriden
11192 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
11193 append_elem(OP_LIST, o,
11194 newCVREF(0, newGVOP(OP_GV,0,gv_readline))))
11195 : (OP*)newUNOP(OP_READLINE, 0, o);
11204 ? (GV_ADDMULTI | GV_ADDINEVAL)
11207 PL_lex_op = readline_overriden
11208 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
11209 append_elem(OP_LIST,
11210 newUNOP(OP_RV2SV, 0, newGVOP(OP_GV, 0, gv)),
11211 newCVREF(0, newGVOP(OP_GV, 0, gv_readline))))
11212 : (OP*)newUNOP(OP_READLINE, 0,
11213 newUNOP(OP_RV2SV, 0,
11214 newGVOP(OP_GV, 0, gv)));
11216 if (!readline_overriden)
11217 PL_lex_op->op_flags |= OPf_SPECIAL;
11218 /* we created the ops in PL_lex_op, so make yylval.ival a null op */
11219 yylval.ival = OP_NULL;
11222 /* If it's none of the above, it must be a literal filehandle
11223 (<Foo::BAR> or <FOO>) so build a simple readline OP */
11225 GV * const gv = gv_fetchpv(d, GV_ADD, SVt_PVIO);
11226 PL_lex_op = readline_overriden
11227 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
11228 append_elem(OP_LIST,
11229 newGVOP(OP_GV, 0, gv),
11230 newCVREF(0, newGVOP(OP_GV, 0, gv_readline))))
11231 : (OP*)newUNOP(OP_READLINE, 0, newGVOP(OP_GV, 0, gv));
11232 yylval.ival = OP_NULL;
11241 takes: start position in buffer
11242 keep_quoted preserve \ on the embedded delimiter(s)
11243 keep_delims preserve the delimiters around the string
11244 returns: position to continue reading from buffer
11245 side-effects: multi_start, multi_close, lex_repl or lex_stuff, and
11246 updates the read buffer.
11248 This subroutine pulls a string out of the input. It is called for:
11249 q single quotes q(literal text)
11250 ' single quotes 'literal text'
11251 qq double quotes qq(interpolate $here please)
11252 " double quotes "interpolate $here please"
11253 qx backticks qx(/bin/ls -l)
11254 ` backticks `/bin/ls -l`
11255 qw quote words @EXPORT_OK = qw( func() $spam )
11256 m// regexp match m/this/
11257 s/// regexp substitute s/this/that/
11258 tr/// string transliterate tr/this/that/
11259 y/// string transliterate y/this/that/
11260 ($*@) sub prototypes sub foo ($)
11261 (stuff) sub attr parameters sub foo : attr(stuff)
11262 <> readline or globs <FOO>, <>, <$fh>, or <*.c>
11264 In most of these cases (all but <>, patterns and transliterate)
11265 yylex() calls scan_str(). m// makes yylex() call scan_pat() which
11266 calls scan_str(). s/// makes yylex() call scan_subst() which calls
11267 scan_str(). tr/// and y/// make yylex() call scan_trans() which
11270 It skips whitespace before the string starts, and treats the first
11271 character as the delimiter. If the delimiter is one of ([{< then
11272 the corresponding "close" character )]}> is used as the closing
11273 delimiter. It allows quoting of delimiters, and if the string has
11274 balanced delimiters ([{<>}]) it allows nesting.
11276 On success, the SV with the resulting string is put into lex_stuff or,
11277 if that is already non-NULL, into lex_repl. The second case occurs only
11278 when parsing the RHS of the special constructs s/// and tr/// (y///).
11279 For convenience, the terminating delimiter character is stuffed into
11284 S_scan_str(pTHX_ char *start, int keep_quoted, int keep_delims)
11287 SV *sv; /* scalar value: string */
11288 char *tmps; /* temp string, used for delimiter matching */
11289 register char *s = start; /* current position in the buffer */
11290 register char term; /* terminating character */
11291 register char *to; /* current position in the sv's data */
11292 I32 brackets = 1; /* bracket nesting level */
11293 bool has_utf8 = FALSE; /* is there any utf8 content? */
11294 I32 termcode; /* terminating char. code */
11295 U8 termstr[UTF8_MAXBYTES]; /* terminating string */
11296 STRLEN termlen; /* length of terminating string */
11297 char *last = NULL; /* last position for nesting bracket */
11303 /* skip space before the delimiter */
11309 if (PL_realtokenstart >= 0) {
11310 stuffstart = PL_realtokenstart;
11311 PL_realtokenstart = -1;
11314 stuffstart = start - SvPVX(PL_linestr);
11316 /* mark where we are, in case we need to report errors */
11319 /* after skipping whitespace, the next character is the terminator */
11322 termcode = termstr[0] = term;
11326 termcode = utf8_to_uvchr((U8*)s, &termlen);
11327 Copy(s, termstr, termlen, U8);
11328 if (!UTF8_IS_INVARIANT(term))
11332 /* mark where we are */
11333 PL_multi_start = CopLINE(PL_curcop);
11334 PL_multi_open = term;
11336 /* find corresponding closing delimiter */
11337 if (term && (tmps = strchr("([{< )]}> )]}>",term)))
11338 termcode = termstr[0] = term = tmps[5];
11340 PL_multi_close = term;
11342 /* create a new SV to hold the contents. 79 is the SV's initial length.
11343 What a random number. */
11345 sv_upgrade(sv, SVt_PVIV);
11346 SvIV_set(sv, termcode);
11347 (void)SvPOK_only(sv); /* validate pointer */
11349 /* move past delimiter and try to read a complete string */
11351 sv_catpvn(sv, s, termlen);
11354 tstart = SvPVX(PL_linestr) + stuffstart;
11355 if (!PL_thisopen && !keep_delims) {
11356 PL_thisopen = newSVpvn(tstart, s - tstart);
11357 stuffstart = s - SvPVX(PL_linestr);
11361 if (PL_encoding && !UTF) {
11365 int offset = s - SvPVX_const(PL_linestr);
11366 const bool found = sv_cat_decode(sv, PL_encoding, PL_linestr,
11367 &offset, (char*)termstr, termlen);
11368 const char * const ns = SvPVX_const(PL_linestr) + offset;
11369 char * const svlast = SvEND(sv) - 1;
11371 for (; s < ns; s++) {
11372 if (*s == '\n' && !PL_rsfp)
11373 CopLINE_inc(PL_curcop);
11376 goto read_more_line;
11378 /* handle quoted delimiters */
11379 if (SvCUR(sv) > 1 && *(svlast-1) == '\\') {
11381 for (t = svlast-2; t >= SvPVX_const(sv) && *t == '\\';)
11383 if ((svlast-1 - t) % 2) {
11384 if (!keep_quoted) {
11385 *(svlast-1) = term;
11387 SvCUR_set(sv, SvCUR(sv) - 1);
11392 if (PL_multi_open == PL_multi_close) {
11400 for (t = w = last; t < svlast; w++, t++) {
11401 /* At here, all closes are "was quoted" one,
11402 so we don't check PL_multi_close. */
11404 if (!keep_quoted && *(t+1) == PL_multi_open)
11409 else if (*t == PL_multi_open)
11417 SvCUR_set(sv, w - SvPVX_const(sv));
11420 if (--brackets <= 0)
11425 if (!keep_delims) {
11426 SvCUR_set(sv, SvCUR(sv) - 1);
11432 /* extend sv if need be */
11433 SvGROW(sv, SvCUR(sv) + (PL_bufend - s) + 1);
11434 /* set 'to' to the next character in the sv's string */
11435 to = SvPVX(sv)+SvCUR(sv);
11437 /* if open delimiter is the close delimiter read unbridle */
11438 if (PL_multi_open == PL_multi_close) {
11439 for (; s < PL_bufend; s++,to++) {
11440 /* embedded newlines increment the current line number */
11441 if (*s == '\n' && !PL_rsfp)
11442 CopLINE_inc(PL_curcop);
11443 /* handle quoted delimiters */
11444 if (*s == '\\' && s+1 < PL_bufend && term != '\\') {
11445 if (!keep_quoted && s[1] == term)
11447 /* any other quotes are simply copied straight through */
11451 /* terminate when run out of buffer (the for() condition), or
11452 have found the terminator */
11453 else if (*s == term) {
11456 if (s+termlen <= PL_bufend && memEQ(s, (char*)termstr, termlen))
11459 else if (!has_utf8 && !UTF8_IS_INVARIANT((U8)*s) && UTF)
11465 /* if the terminator isn't the same as the start character (e.g.,
11466 matched brackets), we have to allow more in the quoting, and
11467 be prepared for nested brackets.
11470 /* read until we run out of string, or we find the terminator */
11471 for (; s < PL_bufend; s++,to++) {
11472 /* embedded newlines increment the line count */
11473 if (*s == '\n' && !PL_rsfp)
11474 CopLINE_inc(PL_curcop);
11475 /* backslashes can escape the open or closing characters */
11476 if (*s == '\\' && s+1 < PL_bufend) {
11477 if (!keep_quoted &&
11478 ((s[1] == PL_multi_open) || (s[1] == PL_multi_close)))
11483 /* allow nested opens and closes */
11484 else if (*s == PL_multi_close && --brackets <= 0)
11486 else if (*s == PL_multi_open)
11488 else if (!has_utf8 && !UTF8_IS_INVARIANT((U8)*s) && UTF)
11493 /* terminate the copied string and update the sv's end-of-string */
11495 SvCUR_set(sv, to - SvPVX_const(sv));
11498 * this next chunk reads more into the buffer if we're not done yet
11502 break; /* handle case where we are done yet :-) */
11504 #ifndef PERL_STRICT_CR
11505 if (to - SvPVX_const(sv) >= 2) {
11506 if ((to[-2] == '\r' && to[-1] == '\n') ||
11507 (to[-2] == '\n' && to[-1] == '\r'))
11511 SvCUR_set(sv, to - SvPVX_const(sv));
11513 else if (to[-1] == '\r')
11516 else if (to - SvPVX_const(sv) == 1 && to[-1] == '\r')
11521 /* if we're out of file, or a read fails, bail and reset the current
11522 line marker so we can report where the unterminated string began
11525 if (PL_madskills) {
11526 char *tstart = SvPVX(PL_linestr) + stuffstart;
11528 sv_catpvn(PL_thisstuff, tstart, PL_bufend - tstart);
11530 PL_thisstuff = newSVpvn(tstart, PL_bufend - tstart);
11534 !(PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = filter_gets(PL_linestr, PL_rsfp, 0))) {
11536 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
11542 /* we read a line, so increment our line counter */
11543 CopLINE_inc(PL_curcop);
11545 /* update debugger info */
11546 if (PERLDB_LINE && PL_curstash != PL_debstash) {
11547 SV * const line_sv = newSV(0);
11549 sv_upgrade(line_sv, SVt_PVMG);
11550 sv_setsv(line_sv,PL_linestr);
11551 (void)SvIOK_on(line_sv);
11552 SvIV_set(line_sv, 0);
11553 av_store(CopFILEAVx(PL_curcop), (I32)CopLINE(PL_curcop), line_sv);
11556 /* having changed the buffer, we must update PL_bufend */
11557 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
11558 PL_last_lop = PL_last_uni = NULL;
11561 /* at this point, we have successfully read the delimited string */
11563 if (!PL_encoding || UTF) {
11565 if (PL_madskills) {
11566 char *tstart = SvPVX(PL_linestr) + stuffstart;
11568 sv_catpvn(PL_thisstuff, tstart, s - tstart);
11570 PL_thisstuff = newSVpvn(tstart, s - tstart);
11571 if (!PL_thisclose && !keep_delims)
11572 PL_thisclose = newSVpvn(s,termlen);
11577 sv_catpvn(sv, s, termlen);
11582 if (PL_madskills) {
11583 char *tstart = SvPVX(PL_linestr) + stuffstart;
11585 sv_catpvn(PL_thisstuff, tstart, s - tstart - termlen);
11587 PL_thisstuff = newSVpvn(tstart, s - tstart - termlen);
11588 if (!PL_thisclose && !keep_delims)
11589 PL_thisclose = newSVpvn(s - termlen,termlen);
11593 if (has_utf8 || PL_encoding)
11596 PL_multi_end = CopLINE(PL_curcop);
11598 /* if we allocated too much space, give some back */
11599 if (SvCUR(sv) + 5 < SvLEN(sv)) {
11600 SvLEN_set(sv, SvCUR(sv) + 1);
11601 SvPV_renew(sv, SvLEN(sv));
11604 /* decide whether this is the first or second quoted string we've read
11617 takes: pointer to position in buffer
11618 returns: pointer to new position in buffer
11619 side-effects: builds ops for the constant in yylval.op
11621 Read a number in any of the formats that Perl accepts:
11623 \d(_?\d)*(\.(\d(_?\d)*)?)?[Ee][\+\-]?(\d(_?\d)*) 12 12.34 12.
11624 \.\d(_?\d)*[Ee][\+\-]?(\d(_?\d)*) .34
11627 0x[0-9A-Fa-f](_?[0-9A-Fa-f])*
11629 Like most scan_ routines, it uses the PL_tokenbuf buffer to hold the
11632 If it reads a number without a decimal point or an exponent, it will
11633 try converting the number to an integer and see if it can do so
11634 without loss of precision.
11638 Perl_scan_num(pTHX_ const char *start, YYSTYPE* lvalp)
11641 register const char *s = start; /* current position in buffer */
11642 register char *d; /* destination in temp buffer */
11643 register char *e; /* end of temp buffer */
11644 NV nv; /* number read, as a double */
11645 SV *sv = NULL; /* place to put the converted number */
11646 bool floatit; /* boolean: int or float? */
11647 const char *lastub = NULL; /* position of last underbar */
11648 static char const number_too_long[] = "Number too long";
11650 /* We use the first character to decide what type of number this is */
11654 Perl_croak(aTHX_ "panic: scan_num");
11656 /* if it starts with a 0, it could be an octal number, a decimal in
11657 0.13 disguise, or a hexadecimal number, or a binary number. */
11661 u holds the "number so far"
11662 shift the power of 2 of the base
11663 (hex == 4, octal == 3, binary == 1)
11664 overflowed was the number more than we can hold?
11666 Shift is used when we add a digit. It also serves as an "are
11667 we in octal/hex/binary?" indicator to disallow hex characters
11668 when in octal mode.
11673 bool overflowed = FALSE;
11674 bool just_zero = TRUE; /* just plain 0 or binary number? */
11675 static const NV nvshift[5] = { 1.0, 2.0, 4.0, 8.0, 16.0 };
11676 static const char* const bases[5] =
11677 { "", "binary", "", "octal", "hexadecimal" };
11678 static const char* const Bases[5] =
11679 { "", "Binary", "", "Octal", "Hexadecimal" };
11680 static const char* const maxima[5] =
11682 "0b11111111111111111111111111111111",
11686 const char *base, *Base, *max;
11688 /* check for hex */
11693 } else if (s[1] == 'b') {
11698 /* check for a decimal in disguise */
11699 else if (s[1] == '.' || s[1] == 'e' || s[1] == 'E')
11701 /* so it must be octal */
11708 if (ckWARN(WARN_SYNTAX))
11709 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
11710 "Misplaced _ in number");
11714 base = bases[shift];
11715 Base = Bases[shift];
11716 max = maxima[shift];
11718 /* read the rest of the number */
11720 /* x is used in the overflow test,
11721 b is the digit we're adding on. */
11726 /* if we don't mention it, we're done */
11730 /* _ are ignored -- but warned about if consecutive */
11732 if (lastub && s == lastub + 1 && ckWARN(WARN_SYNTAX))
11733 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
11734 "Misplaced _ in number");
11738 /* 8 and 9 are not octal */
11739 case '8': case '9':
11741 yyerror(Perl_form(aTHX_ "Illegal octal digit '%c'", *s));
11745 case '2': case '3': case '4':
11746 case '5': case '6': case '7':
11748 yyerror(Perl_form(aTHX_ "Illegal binary digit '%c'", *s));
11751 case '0': case '1':
11752 b = *s++ & 15; /* ASCII digit -> value of digit */
11756 case 'a': case 'b': case 'c': case 'd': case 'e': case 'f':
11757 case 'A': case 'B': case 'C': case 'D': case 'E': case 'F':
11758 /* make sure they said 0x */
11761 b = (*s++ & 7) + 9;
11763 /* Prepare to put the digit we have onto the end
11764 of the number so far. We check for overflows.
11770 x = u << shift; /* make room for the digit */
11772 if ((x >> shift) != u
11773 && !(PL_hints & HINT_NEW_BINARY)) {
11776 if (ckWARN_d(WARN_OVERFLOW))
11777 Perl_warner(aTHX_ packWARN(WARN_OVERFLOW),
11778 "Integer overflow in %s number",
11781 u = x | b; /* add the digit to the end */
11784 n *= nvshift[shift];
11785 /* If an NV has not enough bits in its
11786 * mantissa to represent an UV this summing of
11787 * small low-order numbers is a waste of time
11788 * (because the NV cannot preserve the
11789 * low-order bits anyway): we could just
11790 * remember when did we overflow and in the
11791 * end just multiply n by the right
11799 /* if we get here, we had success: make a scalar value from
11804 /* final misplaced underbar check */
11805 if (s[-1] == '_') {
11806 if (ckWARN(WARN_SYNTAX))
11807 Perl_warner(aTHX_ packWARN(WARN_SYNTAX), "Misplaced _ in number");
11812 if (n > 4294967295.0 && ckWARN(WARN_PORTABLE))
11813 Perl_warner(aTHX_ packWARN(WARN_PORTABLE),
11814 "%s number > %s non-portable",
11820 if (u > 0xffffffff && ckWARN(WARN_PORTABLE))
11821 Perl_warner(aTHX_ packWARN(WARN_PORTABLE),
11822 "%s number > %s non-portable",
11827 if (just_zero && (PL_hints & HINT_NEW_INTEGER))
11828 sv = new_constant(start, s - start, "integer",
11830 else if (PL_hints & HINT_NEW_BINARY)
11831 sv = new_constant(start, s - start, "binary", sv, NULL, NULL);
11836 handle decimal numbers.
11837 we're also sent here when we read a 0 as the first digit
11839 case '1': case '2': case '3': case '4': case '5':
11840 case '6': case '7': case '8': case '9': case '.':
11843 e = PL_tokenbuf + sizeof PL_tokenbuf - 6; /* room for various punctuation */
11846 /* read next group of digits and _ and copy into d */
11847 while (isDIGIT(*s) || *s == '_') {
11848 /* skip underscores, checking for misplaced ones
11852 if (lastub && s == lastub + 1 && ckWARN(WARN_SYNTAX))
11853 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
11854 "Misplaced _ in number");
11858 /* check for end of fixed-length buffer */
11860 Perl_croak(aTHX_ number_too_long);
11861 /* if we're ok, copy the character */
11866 /* final misplaced underbar check */
11867 if (lastub && s == lastub + 1) {
11868 if (ckWARN(WARN_SYNTAX))
11869 Perl_warner(aTHX_ packWARN(WARN_SYNTAX), "Misplaced _ in number");
11872 /* read a decimal portion if there is one. avoid
11873 3..5 being interpreted as the number 3. followed
11876 if (*s == '.' && s[1] != '.') {
11881 if (ckWARN(WARN_SYNTAX))
11882 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
11883 "Misplaced _ in number");
11887 /* copy, ignoring underbars, until we run out of digits.
11889 for (; isDIGIT(*s) || *s == '_'; s++) {
11890 /* fixed length buffer check */
11892 Perl_croak(aTHX_ number_too_long);
11894 if (lastub && s == lastub + 1 && ckWARN(WARN_SYNTAX))
11895 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
11896 "Misplaced _ in number");
11902 /* fractional part ending in underbar? */
11903 if (s[-1] == '_') {
11904 if (ckWARN(WARN_SYNTAX))
11905 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
11906 "Misplaced _ in number");
11908 if (*s == '.' && isDIGIT(s[1])) {
11909 /* oops, it's really a v-string, but without the "v" */
11915 /* read exponent part, if present */
11916 if ((*s == 'e' || *s == 'E') && strchr("+-0123456789_", s[1])) {
11920 /* regardless of whether user said 3E5 or 3e5, use lower 'e' */
11921 *d++ = 'e'; /* At least some Mach atof()s don't grok 'E' */
11923 /* stray preinitial _ */
11925 if (ckWARN(WARN_SYNTAX))
11926 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
11927 "Misplaced _ in number");
11931 /* allow positive or negative exponent */
11932 if (*s == '+' || *s == '-')
11935 /* stray initial _ */
11937 if (ckWARN(WARN_SYNTAX))
11938 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
11939 "Misplaced _ in number");
11943 /* read digits of exponent */
11944 while (isDIGIT(*s) || *s == '_') {
11947 Perl_croak(aTHX_ number_too_long);
11951 if (((lastub && s == lastub + 1) ||
11952 (!isDIGIT(s[1]) && s[1] != '_'))
11953 && ckWARN(WARN_SYNTAX))
11954 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
11955 "Misplaced _ in number");
11962 /* make an sv from the string */
11966 We try to do an integer conversion first if no characters
11967 indicating "float" have been found.
11972 const int flags = grok_number (PL_tokenbuf, d - PL_tokenbuf, &uv);
11974 if (flags == IS_NUMBER_IN_UV) {
11976 sv_setiv(sv, uv); /* Prefer IVs over UVs. */
11979 } else if (flags == (IS_NUMBER_IN_UV | IS_NUMBER_NEG)) {
11980 if (uv <= (UV) IV_MIN)
11981 sv_setiv(sv, -(IV)uv);
11988 /* terminate the string */
11990 nv = Atof(PL_tokenbuf);
11994 if ( floatit ? (PL_hints & HINT_NEW_FLOAT) :
11995 (PL_hints & HINT_NEW_INTEGER) )
11996 sv = new_constant(PL_tokenbuf, d - PL_tokenbuf,
11997 (floatit ? "float" : "integer"),
12001 /* if it starts with a v, it could be a v-string */
12004 sv = newSV(5); /* preallocate storage space */
12005 s = scan_vstring(s,sv);
12009 /* make the op for the constant and return */
12012 lvalp->opval = newSVOP(OP_CONST, 0, sv);
12014 lvalp->opval = NULL;
12020 S_scan_formline(pTHX_ register char *s)
12023 register char *eol;
12025 SV * const stuff = newSVpvs("");
12026 bool needargs = FALSE;
12027 bool eofmt = FALSE;
12029 char *tokenstart = s;
12032 if (PL_madskills) {
12033 savewhite = PL_thiswhite;
12038 while (!needargs) {
12040 #ifdef PERL_STRICT_CR
12041 for (t = s+1;SPACE_OR_TAB(*t); t++) ;
12043 for (t = s+1;SPACE_OR_TAB(*t) || *t == '\r'; t++) ;
12045 if (*t == '\n' || t == PL_bufend) {
12050 if (PL_in_eval && !PL_rsfp) {
12051 eol = (char *) memchr(s,'\n',PL_bufend-s);
12056 eol = PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
12058 for (t = s; t < eol; t++) {
12059 if (*t == '~' && t[1] == '~' && SvCUR(stuff)) {
12061 goto enough; /* ~~ must be first line in formline */
12063 if (*t == '@' || *t == '^')
12067 sv_catpvn(stuff, s, eol-s);
12068 #ifndef PERL_STRICT_CR
12069 if (eol-s > 1 && eol[-2] == '\r' && eol[-1] == '\n') {
12070 char *end = SvPVX(stuff) + SvCUR(stuff);
12073 SvCUR_set(stuff, SvCUR(stuff) - 1);
12083 if (PL_madskills) {
12085 sv_catpvn(PL_thistoken, tokenstart, PL_bufend - tokenstart);
12087 PL_thistoken = newSVpvn(tokenstart, PL_bufend - tokenstart);
12090 s = filter_gets(PL_linestr, PL_rsfp, 0);
12092 tokenstart = PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = SvPVX(PL_linestr);
12094 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = SvPVX(PL_linestr);
12096 PL_bufend = PL_bufptr + SvCUR(PL_linestr);
12097 PL_last_lop = PL_last_uni = NULL;
12106 if (SvCUR(stuff)) {
12109 PL_lex_state = LEX_NORMAL;
12110 start_force(PL_curforce);
12111 NEXTVAL_NEXTTOKE.ival = 0;
12115 PL_lex_state = LEX_FORMLINE;
12117 if (UTF && is_utf8_string((U8*)SvPVX_const(stuff), SvCUR(stuff)))
12119 else if (PL_encoding)
12120 sv_recode_to_utf8(stuff, PL_encoding);
12122 start_force(PL_curforce);
12123 NEXTVAL_NEXTTOKE.opval = (OP*)newSVOP(OP_CONST, 0, stuff);
12125 start_force(PL_curforce);
12126 NEXTVAL_NEXTTOKE.ival = OP_FORMLINE;
12130 SvREFCNT_dec(stuff);
12132 PL_lex_formbrack = 0;
12136 if (PL_madskills) {
12138 sv_catpvn(PL_thistoken, tokenstart, s - tokenstart);
12140 PL_thistoken = newSVpvn(tokenstart, s - tokenstart);
12141 PL_thiswhite = savewhite;
12153 PL_cshlen = strlen(PL_cshname);
12155 #if defined(USE_ITHREADS)
12156 PERL_UNUSED_CONTEXT;
12162 Perl_start_subparse(pTHX_ I32 is_format, U32 flags)
12165 const I32 oldsavestack_ix = PL_savestack_ix;
12166 CV* const outsidecv = PL_compcv;
12169 assert(SvTYPE(PL_compcv) == SVt_PVCV);
12171 SAVEI32(PL_subline);
12172 save_item(PL_subname);
12173 SAVESPTR(PL_compcv);
12175 PL_compcv = (CV*)newSV(0);
12176 sv_upgrade((SV *)PL_compcv, is_format ? SVt_PVFM : SVt_PVCV);
12177 CvFLAGS(PL_compcv) |= flags;
12179 PL_subline = CopLINE(PL_curcop);
12180 CvPADLIST(PL_compcv) = pad_new(padnew_SAVE|padnew_SAVESUB);
12181 CvOUTSIDE(PL_compcv) = (CV*)SvREFCNT_inc_simple(outsidecv);
12182 CvOUTSIDE_SEQ(PL_compcv) = PL_cop_seqmax;
12184 return oldsavestack_ix;
12188 #pragma segment Perl_yylex
12191 Perl_yywarn(pTHX_ const char *s)
12194 PL_in_eval |= EVAL_WARNONLY;
12196 PL_in_eval &= ~EVAL_WARNONLY;
12201 Perl_yyerror(pTHX_ const char *s)
12204 const char *where = NULL;
12205 const char *context = NULL;
12209 if (!yychar || (yychar == ';' && !PL_rsfp))
12211 else if (PL_oldoldbufptr && PL_bufptr > PL_oldoldbufptr &&
12212 PL_bufptr - PL_oldoldbufptr < 200 && PL_oldoldbufptr != PL_oldbufptr &&
12213 PL_oldbufptr != PL_bufptr) {
12216 The code below is removed for NetWare because it abends/crashes on NetWare
12217 when the script has error such as not having the closing quotes like:
12218 if ($var eq "value)
12219 Checking of white spaces is anyway done in NetWare code.
12222 while (isSPACE(*PL_oldoldbufptr))
12225 context = PL_oldoldbufptr;
12226 contlen = PL_bufptr - PL_oldoldbufptr;
12228 else if (PL_oldbufptr && PL_bufptr > PL_oldbufptr &&
12229 PL_bufptr - PL_oldbufptr < 200 && PL_oldbufptr != PL_bufptr) {
12232 The code below is removed for NetWare because it abends/crashes on NetWare
12233 when the script has error such as not having the closing quotes like:
12234 if ($var eq "value)
12235 Checking of white spaces is anyway done in NetWare code.
12238 while (isSPACE(*PL_oldbufptr))
12241 context = PL_oldbufptr;
12242 contlen = PL_bufptr - PL_oldbufptr;
12244 else if (yychar > 255)
12245 where = "next token ???";
12246 else if (yychar == -2) { /* YYEMPTY */
12247 if (PL_lex_state == LEX_NORMAL ||
12248 (PL_lex_state == LEX_KNOWNEXT && PL_lex_defer == LEX_NORMAL))
12249 where = "at end of line";
12250 else if (PL_lex_inpat)
12251 where = "within pattern";
12253 where = "within string";
12256 SV * const where_sv = sv_2mortal(newSVpvs("next char "));
12258 Perl_sv_catpvf(aTHX_ where_sv, "^%c", toCTRL(yychar));
12259 else if (isPRINT_LC(yychar))
12260 Perl_sv_catpvf(aTHX_ where_sv, "%c", yychar);
12262 Perl_sv_catpvf(aTHX_ where_sv, "\\%03o", yychar & 255);
12263 where = SvPVX_const(where_sv);
12265 msg = sv_2mortal(newSVpv(s, 0));
12266 Perl_sv_catpvf(aTHX_ msg, " at %s line %"IVdf", ",
12267 OutCopFILE(PL_curcop), (IV)CopLINE(PL_curcop));
12269 Perl_sv_catpvf(aTHX_ msg, "near \"%.*s\"\n", contlen, context);
12271 Perl_sv_catpvf(aTHX_ msg, "%s\n", where);
12272 if (PL_multi_start < PL_multi_end && (U32)(CopLINE(PL_curcop) - PL_multi_end) <= 1) {
12273 Perl_sv_catpvf(aTHX_ msg,
12274 " (Might be a runaway multi-line %c%c string starting on line %"IVdf")\n",
12275 (int)PL_multi_open,(int)PL_multi_close,(IV)PL_multi_start);
12278 if (PL_in_eval & EVAL_WARNONLY && ckWARN_d(WARN_SYNTAX))
12279 Perl_warner(aTHX_ packWARN(WARN_SYNTAX), "%"SVf, (void*)msg);
12282 if (PL_error_count >= 10) {
12283 if (PL_in_eval && SvCUR(ERRSV))
12284 Perl_croak(aTHX_ "%"SVf"%s has too many errors.\n",
12285 (void*)ERRSV, OutCopFILE(PL_curcop));
12287 Perl_croak(aTHX_ "%s has too many errors.\n",
12288 OutCopFILE(PL_curcop));
12291 PL_in_my_stash = NULL;
12295 #pragma segment Main
12299 S_swallow_bom(pTHX_ U8 *s)
12302 const STRLEN slen = SvCUR(PL_linestr);
12305 if (s[1] == 0xFE) {
12306 /* UTF-16 little-endian? (or UTF32-LE?) */
12307 if (s[2] == 0 && s[3] == 0) /* UTF-32 little-endian */
12308 Perl_croak(aTHX_ "Unsupported script encoding UTF32-LE");
12309 #ifndef PERL_NO_UTF16_FILTER
12310 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF16-LE script encoding (BOM)\n");
12313 if (PL_bufend > (char*)s) {
12317 filter_add(utf16rev_textfilter, NULL);
12318 Newx(news, (PL_bufend - (char*)s) * 3 / 2 + 1, U8);
12319 utf16_to_utf8_reversed(s, news,
12320 PL_bufend - (char*)s - 1,
12322 sv_setpvn(PL_linestr, (const char*)news, newlen);
12324 s = (U8*)SvPVX(PL_linestr);
12325 Copy(news, s, newlen, U8);
12329 SvUTF8_on(PL_linestr);
12330 s = (U8*)SvPVX(PL_linestr);
12332 /* FIXME - is this a general bug fix? */
12335 PL_bufend = SvPVX(PL_linestr) + newlen;
12338 Perl_croak(aTHX_ "Unsupported script encoding UTF16-LE");
12343 if (s[1] == 0xFF) { /* UTF-16 big-endian? */
12344 #ifndef PERL_NO_UTF16_FILTER
12345 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16BE script encoding (BOM)\n");
12348 if (PL_bufend > (char *)s) {
12352 filter_add(utf16_textfilter, NULL);
12353 Newx(news, (PL_bufend - (char*)s) * 3 / 2 + 1, U8);
12354 utf16_to_utf8(s, news,
12355 PL_bufend - (char*)s,
12357 sv_setpvn(PL_linestr, (const char*)news, newlen);
12359 SvUTF8_on(PL_linestr);
12360 s = (U8*)SvPVX(PL_linestr);
12361 PL_bufend = SvPVX(PL_linestr) + newlen;
12364 Perl_croak(aTHX_ "Unsupported script encoding UTF16-BE");
12369 if (slen > 2 && s[1] == 0xBB && s[2] == 0xBF) {
12370 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-8 script encoding (BOM)\n");
12371 s += 3; /* UTF-8 */
12377 if (s[2] == 0xFE && s[3] == 0xFF) {
12378 /* UTF-32 big-endian */
12379 Perl_croak(aTHX_ "Unsupported script encoding UTF32-BE");
12382 else if (s[2] == 0 && s[3] != 0) {
12385 * are a good indicator of UTF-16BE. */
12386 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16BE script encoding (no BOM)\n");
12391 if (slen > 3 && s[1] == 0 && s[2] != 0 && s[3] == 0) {
12394 * are a good indicator of UTF-16LE. */
12395 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16LE script encoding (no BOM)\n");
12404 * Restore a source filter.
12408 restore_rsfp(pTHX_ void *f)
12411 PerlIO * const fp = (PerlIO*)f;
12413 if (PL_rsfp == PerlIO_stdin())
12414 PerlIO_clearerr(PL_rsfp);
12415 else if (PL_rsfp && (PL_rsfp != fp))
12416 PerlIO_close(PL_rsfp);
12420 #ifndef PERL_NO_UTF16_FILTER
12422 utf16_textfilter(pTHX_ int idx, SV *sv, int maxlen)
12425 const STRLEN old = SvCUR(sv);
12426 const I32 count = FILTER_READ(idx+1, sv, maxlen);
12427 DEBUG_P(PerlIO_printf(Perl_debug_log,
12428 "utf16_textfilter(%p): %d %d (%d)\n",
12429 FPTR2DPTR(void *, utf16_textfilter),
12430 idx, maxlen, (int) count));
12434 Newx(tmps, SvCUR(sv) * 3 / 2 + 1, U8);
12435 Copy(SvPVX_const(sv), tmps, old, char);
12436 utf16_to_utf8((U8*)SvPVX_const(sv) + old, tmps + old,
12437 SvCUR(sv) - old, &newlen);
12438 sv_usepvn(sv, (char*)tmps, (STRLEN)newlen + old);
12440 DEBUG_P({sv_dump(sv);});
12445 utf16rev_textfilter(pTHX_ int idx, SV *sv, int maxlen)
12448 const STRLEN old = SvCUR(sv);
12449 const I32 count = FILTER_READ(idx+1, sv, maxlen);
12450 DEBUG_P(PerlIO_printf(Perl_debug_log,
12451 "utf16rev_textfilter(%p): %d %d (%d)\n",
12452 FPTR2DPTR(void *, utf16rev_textfilter),
12453 idx, maxlen, (int) count));
12457 Newx(tmps, SvCUR(sv) * 3 / 2 + 1, U8);
12458 Copy(SvPVX_const(sv), tmps, old, char);
12459 utf16_to_utf8((U8*)SvPVX_const(sv) + old, tmps + old,
12460 SvCUR(sv) - old, &newlen);
12461 sv_usepvn(sv, (char*)tmps, (STRLEN)newlen + old);
12463 DEBUG_P({ sv_dump(sv); });
12469 Returns a pointer to the next character after the parsed
12470 vstring, as well as updating the passed in sv.
12472 Function must be called like
12475 s = scan_vstring(s,sv);
12477 The sv should already be large enough to store the vstring
12478 passed in, for performance reasons.
12483 Perl_scan_vstring(pTHX_ const char *s, SV *sv)
12486 const char *pos = s;
12487 const char *start = s;
12488 if (*pos == 'v') pos++; /* get past 'v' */
12489 while (pos < PL_bufend && (isDIGIT(*pos) || *pos == '_'))
12491 if ( *pos != '.') {
12492 /* this may not be a v-string if followed by => */
12493 const char *next = pos;
12494 while (next < PL_bufend && isSPACE(*next))
12496 if ((PL_bufend - next) >= 2 && *next == '=' && next[1] == '>' ) {
12497 /* return string not v-string */
12498 sv_setpvn(sv,(char *)s,pos-s);
12499 return (char *)pos;
12503 if (!isALPHA(*pos)) {
12504 U8 tmpbuf[UTF8_MAXBYTES+1];
12507 s++; /* get past 'v' */
12509 sv_setpvn(sv, "", 0);
12512 /* this is atoi() that tolerates underscores */
12515 const char *end = pos;
12517 while (--end >= s) {
12519 const UV orev = rev;
12520 rev += (*end - '0') * mult;
12522 if (orev > rev && ckWARN_d(WARN_OVERFLOW))
12523 Perl_warner(aTHX_ packWARN(WARN_OVERFLOW),
12524 "Integer overflow in decimal number");
12528 if (rev > 0x7FFFFFFF)
12529 Perl_croak(aTHX_ "In EBCDIC the v-string components cannot exceed 2147483647");
12531 /* Append native character for the rev point */
12532 tmpend = uvchr_to_utf8(tmpbuf, rev);
12533 sv_catpvn(sv, (const char*)tmpbuf, tmpend - tmpbuf);
12534 if (!UNI_IS_INVARIANT(NATIVE_TO_UNI(rev)))
12536 if (pos + 1 < PL_bufend && *pos == '.' && isDIGIT(pos[1]))
12542 while (pos < PL_bufend && (isDIGIT(*pos) || *pos == '_'))
12546 sv_magic(sv,NULL,PERL_MAGIC_vstring,(const char*)start, pos-start);
12554 * c-indentation-style: bsd
12555 * c-basic-offset: 4
12556 * indent-tabs-mode: t
12559 * ex: set ts=8 sts=4 sw=4 noet: