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( { printbuf("### Saw unary minus before =>, forcing word %s\n", s); } );
3820 OPERATOR('-'); /* unary minus */
3822 PL_last_uni = PL_oldbufptr;
3824 case 'r': ftst = OP_FTEREAD; break;
3825 case 'w': ftst = OP_FTEWRITE; break;
3826 case 'x': ftst = OP_FTEEXEC; break;
3827 case 'o': ftst = OP_FTEOWNED; break;
3828 case 'R': ftst = OP_FTRREAD; break;
3829 case 'W': ftst = OP_FTRWRITE; break;
3830 case 'X': ftst = OP_FTREXEC; break;
3831 case 'O': ftst = OP_FTROWNED; break;
3832 case 'e': ftst = OP_FTIS; break;
3833 case 'z': ftst = OP_FTZERO; break;
3834 case 's': ftst = OP_FTSIZE; break;
3835 case 'f': ftst = OP_FTFILE; break;
3836 case 'd': ftst = OP_FTDIR; break;
3837 case 'l': ftst = OP_FTLINK; break;
3838 case 'p': ftst = OP_FTPIPE; break;
3839 case 'S': ftst = OP_FTSOCK; break;
3840 case 'u': ftst = OP_FTSUID; break;
3841 case 'g': ftst = OP_FTSGID; break;
3842 case 'k': ftst = OP_FTSVTX; break;
3843 case 'b': ftst = OP_FTBLK; break;
3844 case 'c': ftst = OP_FTCHR; break;
3845 case 't': ftst = OP_FTTTY; break;
3846 case 'T': ftst = OP_FTTEXT; break;
3847 case 'B': ftst = OP_FTBINARY; break;
3848 case 'M': case 'A': case 'C':
3849 gv_fetchpvs("\024", GV_ADD|GV_NOTQUAL, SVt_PV);
3851 case 'M': ftst = OP_FTMTIME; break;
3852 case 'A': ftst = OP_FTATIME; break;
3853 case 'C': ftst = OP_FTCTIME; break;
3861 PL_last_lop_op = (OPCODE)ftst;
3862 DEBUG_T( { PerlIO_printf(Perl_debug_log,
3863 "### Saw file test %c\n", (int)tmp);
3868 /* Assume it was a minus followed by a one-letter named
3869 * subroutine call (or a -bareword), then. */
3870 DEBUG_T( { PerlIO_printf(Perl_debug_log,
3871 "### '-%c' looked like a file test but was not\n",
3878 const char tmp = *s++;
3881 if (PL_expect == XOPERATOR)
3886 else if (*s == '>') {
3889 if (isIDFIRST_lazy_if(s,UTF)) {
3890 s = force_word(s,METHOD,FALSE,TRUE,FALSE);
3898 if (PL_expect == XOPERATOR)
3901 if (isSPACE(*s) || !isSPACE(*PL_bufptr))
3903 OPERATOR('-'); /* unary minus */
3909 const char tmp = *s++;
3912 if (PL_expect == XOPERATOR)
3917 if (PL_expect == XOPERATOR)
3920 if (isSPACE(*s) || !isSPACE(*PL_bufptr))
3927 if (PL_expect != XOPERATOR) {
3928 s = scan_ident(s, PL_bufend, PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
3929 PL_expect = XOPERATOR;
3930 force_ident(PL_tokenbuf, '*');
3943 if (PL_expect == XOPERATOR) {
3947 PL_tokenbuf[0] = '%';
3948 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1, TRUE);
3949 if (!PL_tokenbuf[1]) {
3952 PL_pending_ident = '%';
3963 && (PL_expect == XOPERATOR || PL_expect == XTERMORDORDOR)
3964 && FEATURE_IS_ENABLED("~~"))
3971 const char tmp = *s++;
3977 goto just_a_word_zero_gv;
3980 switch (PL_expect) {
3986 if (!PL_in_my || PL_lex_state != LEX_NORMAL)
3988 PL_bufptr = s; /* update in case we back off */
3994 PL_expect = XTERMBLOCK;
3997 stuffstart = s - SvPVX(PL_linestr) - 1;
4001 while (isIDFIRST_lazy_if(s,UTF)) {
4004 d = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
4005 if (isLOWER(*s) && (tmp = keyword(PL_tokenbuf, len))) {
4006 if (tmp < 0) tmp = -tmp;
4021 sv = newSVpvn(s, len);
4023 d = scan_str(d,TRUE,TRUE);
4025 /* MUST advance bufptr here to avoid bogus
4026 "at end of line" context messages from yyerror().
4028 PL_bufptr = s + len;
4029 yyerror("Unterminated attribute parameter in attribute list");
4033 return REPORT(0); /* EOF indicator */
4037 sv_catsv(sv, PL_lex_stuff);
4038 attrs = append_elem(OP_LIST, attrs,
4039 newSVOP(OP_CONST, 0, sv));
4040 SvREFCNT_dec(PL_lex_stuff);
4041 PL_lex_stuff = NULL;
4044 if (len == 6 && strnEQ(SvPVX(sv), "unique", len)) {
4046 if (PL_in_my == KEY_our) {
4048 GvUNIQUE_on(cGVOPx_gv(yylval.opval));
4050 /* skip to avoid loading attributes.pm */
4052 deprecate(":unique");
4055 Perl_croak(aTHX_ "The 'unique' attribute may only be applied to 'our' variables");
4058 /* NOTE: any CV attrs applied here need to be part of
4059 the CVf_BUILTIN_ATTRS define in cv.h! */
4060 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "lvalue", len)) {
4062 CvLVALUE_on(PL_compcv);
4064 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "locked", len)) {
4066 CvLOCKED_on(PL_compcv);
4068 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "method", len)) {
4070 CvMETHOD_on(PL_compcv);
4072 else if (!PL_in_my && len == 9 && strnEQ(SvPVX(sv), "assertion", len)) {
4074 CvASSERTION_on(PL_compcv);
4076 /* After we've set the flags, it could be argued that
4077 we don't need to do the attributes.pm-based setting
4078 process, and shouldn't bother appending recognized
4079 flags. To experiment with that, uncomment the
4080 following "else". (Note that's already been
4081 uncommented. That keeps the above-applied built-in
4082 attributes from being intercepted (and possibly
4083 rejected) by a package's attribute routines, but is
4084 justified by the performance win for the common case
4085 of applying only built-in attributes.) */
4087 attrs = append_elem(OP_LIST, attrs,
4088 newSVOP(OP_CONST, 0,
4092 if (*s == ':' && s[1] != ':')
4095 break; /* require real whitespace or :'s */
4096 /* XXX losing whitespace on sequential attributes here */
4100 = (PL_expect == XOPERATOR ? '=' : '{'); /*'}(' for vi */
4101 if (*s != ';' && *s != '}' && *s != tmp
4102 && (tmp != '=' || *s != ')')) {
4103 const char q = ((*s == '\'') ? '"' : '\'');
4104 /* If here for an expression, and parsed no attrs, back
4106 if (tmp == '=' && !attrs) {
4110 /* MUST advance bufptr here to avoid bogus "at end of line"
4111 context messages from yyerror().
4115 ? Perl_form(aTHX_ "Invalid separator character "
4116 "%c%c%c in attribute list", q, *s, q)
4117 : "Unterminated attribute list" );
4125 start_force(PL_curforce);
4126 NEXTVAL_NEXTTOKE.opval = attrs;
4127 CURMAD('_', PL_nextwhite);
4132 PL_thistoken = newSVpvn(SvPVX(PL_linestr) + stuffstart,
4133 (s - SvPVX(PL_linestr)) - stuffstart);
4141 if (PL_last_lop == PL_oldoldbufptr || PL_last_uni == PL_oldoldbufptr)
4142 PL_oldbufptr = PL_oldoldbufptr; /* allow print(STDOUT 123) */
4150 const char tmp = *s++;
4155 const char tmp = *s++;
4163 if (PL_lex_brackets <= 0)
4164 yyerror("Unmatched right square bracket");
4167 if (PL_lex_state == LEX_INTERPNORMAL) {
4168 if (PL_lex_brackets == 0) {
4169 if (*s != '[' && *s != '{' && (*s != '-' || s[1] != '>'))
4170 PL_lex_state = LEX_INTERPEND;
4177 if (PL_lex_brackets > 100) {
4178 Renew(PL_lex_brackstack, PL_lex_brackets + 10, char);
4180 switch (PL_expect) {
4182 if (PL_lex_formbrack) {
4186 if (PL_oldoldbufptr == PL_last_lop)
4187 PL_lex_brackstack[PL_lex_brackets++] = XTERM;
4189 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
4190 OPERATOR(HASHBRACK);
4192 while (s < PL_bufend && SPACE_OR_TAB(*s))
4195 PL_tokenbuf[0] = '\0';
4196 if (d < PL_bufend && *d == '-') {
4197 PL_tokenbuf[0] = '-';
4199 while (d < PL_bufend && SPACE_OR_TAB(*d))
4202 if (d < PL_bufend && isIDFIRST_lazy_if(d,UTF)) {
4203 d = scan_word(d, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1,
4205 while (d < PL_bufend && SPACE_OR_TAB(*d))
4208 const char minus = (PL_tokenbuf[0] == '-');
4209 s = force_word(s + minus, WORD, FALSE, TRUE, FALSE);
4217 PL_lex_brackstack[PL_lex_brackets++] = XSTATE;
4222 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
4227 if (PL_oldoldbufptr == PL_last_lop)
4228 PL_lex_brackstack[PL_lex_brackets++] = XTERM;
4230 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
4233 if (PL_expect == XREF && PL_lex_state == LEX_INTERPNORMAL) {
4235 /* This hack is to get the ${} in the message. */
4237 yyerror("syntax error");
4240 OPERATOR(HASHBRACK);
4242 /* This hack serves to disambiguate a pair of curlies
4243 * as being a block or an anon hash. Normally, expectation
4244 * determines that, but in cases where we're not in a
4245 * position to expect anything in particular (like inside
4246 * eval"") we have to resolve the ambiguity. This code
4247 * covers the case where the first term in the curlies is a
4248 * quoted string. Most other cases need to be explicitly
4249 * disambiguated by prepending a "+" before the opening
4250 * curly in order to force resolution as an anon hash.
4252 * XXX should probably propagate the outer expectation
4253 * into eval"" to rely less on this hack, but that could
4254 * potentially break current behavior of eval"".
4258 if (*s == '\'' || *s == '"' || *s == '`') {
4259 /* common case: get past first string, handling escapes */
4260 for (t++; t < PL_bufend && *t != *s;)
4261 if (*t++ == '\\' && (*t == '\\' || *t == *s))
4265 else if (*s == 'q') {
4268 || ((*t == 'q' || *t == 'x') && ++t < PL_bufend
4271 /* skip q//-like construct */
4273 char open, close, term;
4276 while (t < PL_bufend && isSPACE(*t))
4278 /* check for q => */
4279 if (t+1 < PL_bufend && t[0] == '=' && t[1] == '>') {
4280 OPERATOR(HASHBRACK);
4284 if (term && (tmps = strchr("([{< )]}> )]}>",term)))
4288 for (t++; t < PL_bufend; t++) {
4289 if (*t == '\\' && t+1 < PL_bufend && open != '\\')
4291 else if (*t == open)
4295 for (t++; t < PL_bufend; t++) {
4296 if (*t == '\\' && t+1 < PL_bufend)
4298 else if (*t == close && --brackets <= 0)
4300 else if (*t == open)
4307 /* skip plain q word */
4308 while (t < PL_bufend && isALNUM_lazy_if(t,UTF))
4311 else if (isALNUM_lazy_if(t,UTF)) {
4313 while (t < PL_bufend && isALNUM_lazy_if(t,UTF))
4316 while (t < PL_bufend && isSPACE(*t))
4318 /* if comma follows first term, call it an anon hash */
4319 /* XXX it could be a comma expression with loop modifiers */
4320 if (t < PL_bufend && ((*t == ',' && (*s == 'q' || !isLOWER(*s)))
4321 || (*t == '=' && t[1] == '>')))
4322 OPERATOR(HASHBRACK);
4323 if (PL_expect == XREF)
4326 PL_lex_brackstack[PL_lex_brackets-1] = XSTATE;
4332 yylval.ival = CopLINE(PL_curcop);
4333 if (isSPACE(*s) || *s == '#')
4334 PL_copline = NOLINE; /* invalidate current command line number */
4339 if (PL_lex_brackets <= 0)
4340 yyerror("Unmatched right curly bracket");
4342 PL_expect = (expectation)PL_lex_brackstack[--PL_lex_brackets];
4343 if (PL_lex_brackets < PL_lex_formbrack && PL_lex_state != LEX_INTERPNORMAL)
4344 PL_lex_formbrack = 0;
4345 if (PL_lex_state == LEX_INTERPNORMAL) {
4346 if (PL_lex_brackets == 0) {
4347 if (PL_expect & XFAKEBRACK) {
4348 PL_expect &= XENUMMASK;
4349 PL_lex_state = LEX_INTERPEND;
4354 PL_thiswhite = newSVpvn("",0);
4355 sv_catpvn(PL_thiswhite,"}",1);
4358 return yylex(); /* ignore fake brackets */
4360 if (*s == '-' && s[1] == '>')
4361 PL_lex_state = LEX_INTERPENDMAYBE;
4362 else if (*s != '[' && *s != '{')
4363 PL_lex_state = LEX_INTERPEND;
4366 if (PL_expect & XFAKEBRACK) {
4367 PL_expect &= XENUMMASK;
4369 return yylex(); /* ignore fake brackets */
4371 start_force(PL_curforce);
4373 curmad('X', newSVpvn(s-1,1));
4374 CURMAD('_', PL_thiswhite);
4379 PL_thistoken = newSVpvn("",0);
4387 if (PL_expect == XOPERATOR) {
4388 if (PL_bufptr == PL_linestart && ckWARN(WARN_SEMICOLON)
4389 && isIDFIRST_lazy_if(s,UTF))
4391 CopLINE_dec(PL_curcop);
4392 Perl_warner(aTHX_ packWARN(WARN_SEMICOLON), PL_warn_nosemi);
4393 CopLINE_inc(PL_curcop);
4398 s = scan_ident(s - 1, PL_bufend, PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
4400 PL_expect = XOPERATOR;
4401 force_ident(PL_tokenbuf, '&');
4405 yylval.ival = (OPpENTERSUB_AMPER<<8);
4417 const char tmp = *s++;
4424 if (tmp && isSPACE(*s) && ckWARN(WARN_SYNTAX)
4425 && strchr("+-*/%.^&|<",tmp))
4426 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
4427 "Reversed %c= operator",(int)tmp);
4429 if (PL_expect == XSTATE && isALPHA(tmp) &&
4430 (s == PL_linestart+1 || s[-2] == '\n') )
4432 if (PL_in_eval && !PL_rsfp) {
4437 if (strnEQ(s,"=cut",4)) {
4453 PL_thiswhite = newSVpvn("",0);
4454 sv_catpvn(PL_thiswhite, PL_linestart,
4455 PL_bufend - PL_linestart);
4459 PL_doextract = TRUE;
4463 if (PL_lex_brackets < PL_lex_formbrack) {
4465 #ifdef PERL_STRICT_CR
4466 for (t = s; SPACE_OR_TAB(*t); t++) ;
4468 for (t = s; SPACE_OR_TAB(*t) || *t == '\r'; t++) ;
4470 if (*t == '\n' || *t == '#') {
4481 const char tmp = *s++;
4483 /* was this !=~ where !~ was meant?
4484 * warn on m:!=~\s+([/?]|[msy]\W|tr\W): */
4486 if (*s == '~' && ckWARN(WARN_SYNTAX)) {
4487 const char *t = s+1;
4489 while (t < PL_bufend && isSPACE(*t))
4492 if (*t == '/' || *t == '?' ||
4493 ((*t == 'm' || *t == 's' || *t == 'y')
4494 && !isALNUM(t[1])) ||
4495 (*t == 't' && t[1] == 'r' && !isALNUM(t[2])))
4496 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
4497 "!=~ should be !~");
4507 if (PL_expect != XOPERATOR) {
4508 if (s[1] != '<' && !strchr(s,'>'))
4511 s = scan_heredoc(s);
4513 s = scan_inputsymbol(s);
4514 TERM(sublex_start());
4520 SHop(OP_LEFT_SHIFT);
4534 const char tmp = *s++;
4536 SHop(OP_RIGHT_SHIFT);
4537 else if (tmp == '=')
4546 if (PL_expect == XOPERATOR) {
4547 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
4549 deprecate_old(commaless_variable_list);
4550 return REPORT(','); /* grandfather non-comma-format format */
4554 if (s[1] == '#' && (isIDFIRST_lazy_if(s+2,UTF) || strchr("{$:+-", s[2]))) {
4555 PL_tokenbuf[0] = '@';
4556 s = scan_ident(s + 1, PL_bufend, PL_tokenbuf + 1,
4557 sizeof PL_tokenbuf - 1, FALSE);
4558 if (PL_expect == XOPERATOR)
4559 no_op("Array length", s);
4560 if (!PL_tokenbuf[1])
4562 PL_expect = XOPERATOR;
4563 PL_pending_ident = '#';
4567 PL_tokenbuf[0] = '$';
4568 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1,
4569 sizeof PL_tokenbuf - 1, FALSE);
4570 if (PL_expect == XOPERATOR)
4572 if (!PL_tokenbuf[1]) {
4574 yyerror("Final $ should be \\$ or $name");
4578 /* This kludge not intended to be bulletproof. */
4579 if (PL_tokenbuf[1] == '[' && !PL_tokenbuf[2]) {
4580 yylval.opval = newSVOP(OP_CONST, 0,
4581 newSViv(CopARYBASE_get(&PL_compiling)));
4582 yylval.opval->op_private = OPpCONST_ARYBASE;
4588 const char tmp = *s;
4589 if (PL_lex_state == LEX_NORMAL)
4592 if ((PL_expect != XREF || PL_oldoldbufptr == PL_last_lop)
4593 && intuit_more(s)) {
4595 PL_tokenbuf[0] = '@';
4596 if (ckWARN(WARN_SYNTAX)) {
4599 isSPACE(*t) || isALNUM_lazy_if(t,UTF) || *t == '$';
4602 PL_bufptr = PEEKSPACE(PL_bufptr); /* XXX can realloc */
4603 while (t < PL_bufend && *t != ']')
4605 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
4606 "Multidimensional syntax %.*s not supported",
4607 (int)((t - PL_bufptr) + 1), PL_bufptr);
4611 else if (*s == '{') {
4613 PL_tokenbuf[0] = '%';
4614 if (strEQ(PL_tokenbuf+1, "SIG") && ckWARN(WARN_SYNTAX)
4615 && (t = strchr(s, '}')) && (t = strchr(t, '=')))
4617 char tmpbuf[sizeof PL_tokenbuf];
4618 for (t++; isSPACE(*t); t++) ;
4619 if (isIDFIRST_lazy_if(t,UTF)) {
4621 t = scan_word(t, tmpbuf, sizeof tmpbuf, TRUE,
4623 for (; isSPACE(*t); t++) ;
4624 if (*t == ';' && get_cv(tmpbuf, FALSE))
4625 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
4626 "You need to quote \"%s\"",
4633 PL_expect = XOPERATOR;
4634 if (PL_lex_state == LEX_NORMAL && isSPACE((char)tmp)) {
4635 const bool islop = (PL_last_lop == PL_oldoldbufptr);
4636 if (!islop || PL_last_lop_op == OP_GREPSTART)
4637 PL_expect = XOPERATOR;
4638 else if (strchr("$@\"'`q", *s))
4639 PL_expect = XTERM; /* e.g. print $fh "foo" */
4640 else if (strchr("&*<%", *s) && isIDFIRST_lazy_if(s+1,UTF))
4641 PL_expect = XTERM; /* e.g. print $fh &sub */
4642 else if (isIDFIRST_lazy_if(s,UTF)) {
4643 char tmpbuf[sizeof PL_tokenbuf];
4645 scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
4646 if ((t2 = keyword(tmpbuf, len))) {
4647 /* binary operators exclude handle interpretations */
4659 PL_expect = XTERM; /* e.g. print $fh length() */
4664 PL_expect = XTERM; /* e.g. print $fh subr() */
4667 else if (isDIGIT(*s))
4668 PL_expect = XTERM; /* e.g. print $fh 3 */
4669 else if (*s == '.' && isDIGIT(s[1]))
4670 PL_expect = XTERM; /* e.g. print $fh .3 */
4671 else if ((*s == '?' || *s == '-' || *s == '+')
4672 && !isSPACE(s[1]) && s[1] != '=')
4673 PL_expect = XTERM; /* e.g. print $fh -1 */
4674 else if (*s == '/' && !isSPACE(s[1]) && s[1] != '='
4676 PL_expect = XTERM; /* e.g. print $fh /.../
4677 XXX except DORDOR operator
4679 else if (*s == '<' && s[1] == '<' && !isSPACE(s[2])
4681 PL_expect = XTERM; /* print $fh <<"EOF" */
4684 PL_pending_ident = '$';
4688 if (PL_expect == XOPERATOR)
4690 PL_tokenbuf[0] = '@';
4691 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1, FALSE);
4692 if (!PL_tokenbuf[1]) {
4695 if (PL_lex_state == LEX_NORMAL)
4697 if ((PL_expect != XREF || PL_oldoldbufptr == PL_last_lop) && intuit_more(s)) {
4699 PL_tokenbuf[0] = '%';
4701 /* Warn about @ where they meant $. */
4702 if (*s == '[' || *s == '{') {
4703 if (ckWARN(WARN_SYNTAX)) {
4704 const char *t = s + 1;
4705 while (*t && (isALNUM_lazy_if(t,UTF) || strchr(" \t$#+-'\"", *t)))
4707 if (*t == '}' || *t == ']') {
4709 PL_bufptr = PEEKSPACE(PL_bufptr); /* XXX can realloc */
4710 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
4711 "Scalar value %.*s better written as $%.*s",
4712 (int)(t-PL_bufptr), PL_bufptr,
4713 (int)(t-PL_bufptr-1), PL_bufptr+1);
4718 PL_pending_ident = '@';
4721 case '/': /* may be division, defined-or, or pattern */
4722 if (PL_expect == XTERMORDORDOR && s[1] == '/') {
4726 case '?': /* may either be conditional or pattern */
4727 if(PL_expect == XOPERATOR) {
4735 /* A // operator. */
4745 /* Disable warning on "study /blah/" */
4746 if (PL_oldoldbufptr == PL_last_uni
4747 && (*PL_last_uni != 's' || s - PL_last_uni < 5
4748 || memNE(PL_last_uni, "study", 5)
4749 || isALNUM_lazy_if(PL_last_uni+5,UTF)
4752 s = scan_pat(s,OP_MATCH);
4753 TERM(sublex_start());
4757 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack
4758 #ifdef PERL_STRICT_CR
4761 && (s[1] == '\n' || (s[1] == '\r' && s[2] == '\n'))
4763 && (s == PL_linestart || s[-1] == '\n') )
4765 PL_lex_formbrack = 0;
4769 if (PL_expect == XOPERATOR || !isDIGIT(s[1])) {
4775 yylval.ival = OPf_SPECIAL;
4781 if (PL_expect != XOPERATOR)
4786 case '0': case '1': case '2': case '3': case '4':
4787 case '5': case '6': case '7': case '8': case '9':
4788 s = scan_num(s, &yylval);
4789 DEBUG_T( { printbuf("### Saw number in %s\n", s); } );
4790 if (PL_expect == XOPERATOR)
4795 s = scan_str(s,!!PL_madskills,FALSE);
4796 DEBUG_T( { printbuf("### Saw string before %s\n", s); } );
4797 if (PL_expect == XOPERATOR) {
4798 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
4800 deprecate_old(commaless_variable_list);
4801 return REPORT(','); /* grandfather non-comma-format format */
4808 yylval.ival = OP_CONST;
4809 TERM(sublex_start());
4812 s = scan_str(s,!!PL_madskills,FALSE);
4813 DEBUG_T( { printbuf("### Saw string before %s\n", s); } );
4814 if (PL_expect == XOPERATOR) {
4815 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
4817 deprecate_old(commaless_variable_list);
4818 return REPORT(','); /* grandfather non-comma-format format */
4825 yylval.ival = OP_CONST;
4826 /* FIXME. I think that this can be const if char *d is replaced by
4827 more localised variables. */
4828 for (d = SvPV(PL_lex_stuff, len); len; len--, d++) {
4829 if (*d == '$' || *d == '@' || *d == '\\' || !UTF8_IS_INVARIANT((U8)*d)) {
4830 yylval.ival = OP_STRINGIFY;
4834 TERM(sublex_start());
4837 s = scan_str(s,!!PL_madskills,FALSE);
4838 DEBUG_T( { printbuf("### Saw backtick string before %s\n", s); } );
4839 if (PL_expect == XOPERATOR)
4840 no_op("Backticks",s);
4843 yylval.ival = OP_BACKTICK;
4845 TERM(sublex_start());
4849 if (PL_lex_inwhat && isDIGIT(*s) && ckWARN(WARN_SYNTAX))
4850 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),"Can't use \\%c to mean $%c in expression",
4852 if (PL_expect == XOPERATOR)
4853 no_op("Backslash",s);
4857 if (isDIGIT(s[1]) && PL_expect != XOPERATOR) {
4858 char *start = s + 2;
4859 while (isDIGIT(*start) || *start == '_')
4861 if (*start == '.' && isDIGIT(start[1])) {
4862 s = scan_num(s, &yylval);
4865 /* avoid v123abc() or $h{v1}, allow C<print v10;> */
4866 else if (!isALPHA(*start) && (PL_expect == XTERM
4867 || PL_expect == XREF || PL_expect == XSTATE
4868 || PL_expect == XTERMORDORDOR)) {
4869 /* XXX Use gv_fetchpvn rather than stomping on a const string */
4870 const char c = *start;
4873 gv = gv_fetchpv(s, 0, SVt_PVCV);
4876 s = scan_num(s, &yylval);
4883 if (isDIGIT(s[1]) && PL_expect == XOPERATOR) {
4919 I32 orig_keyword = 0;
4924 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
4926 /* Some keywords can be followed by any delimiter, including ':' */
4927 tmp = ((len == 1 && strchr("msyq", PL_tokenbuf[0])) ||
4928 (len == 2 && ((PL_tokenbuf[0] == 't' && PL_tokenbuf[1] == 'r') ||
4929 (PL_tokenbuf[0] == 'q' &&
4930 strchr("qwxr", PL_tokenbuf[1])))));
4932 /* x::* is just a word, unless x is "CORE" */
4933 if (!tmp && *s == ':' && s[1] == ':' && strNE(PL_tokenbuf, "CORE"))
4937 while (d < PL_bufend && isSPACE(*d))
4938 d++; /* no comments skipped here, or s### is misparsed */
4940 /* Is this a label? */
4941 if (!tmp && PL_expect == XSTATE
4942 && d < PL_bufend && *d == ':' && *(d + 1) != ':') {
4944 yylval.pval = savepv(PL_tokenbuf);
4949 /* Check for keywords */
4950 tmp = keyword(PL_tokenbuf, len);
4952 /* Is this a word before a => operator? */
4953 if (*d == '=' && d[1] == '>') {
4956 = (OP*)newSVOP(OP_CONST, 0,
4957 S_newSV_maybe_utf8(aTHX_ PL_tokenbuf, len));
4958 yylval.opval->op_private = OPpCONST_BARE;
4962 if (tmp < 0) { /* second-class keyword? */
4963 GV *ogv = NULL; /* override (winner) */
4964 GV *hgv = NULL; /* hidden (loser) */
4965 if (PL_expect != XOPERATOR && (*s != ':' || s[1] != ':')) {
4967 if ((gv = gv_fetchpvn_flags(PL_tokenbuf, len, 0, SVt_PVCV)) &&
4970 if (GvIMPORTED_CV(gv))
4972 else if (! CvMETHOD(cv))
4976 (gvp = (GV**)hv_fetch(PL_globalstash,PL_tokenbuf,len,FALSE)) &&
4977 (gv = *gvp) != (GV*)&PL_sv_undef &&
4978 GvCVu(gv) && GvIMPORTED_CV(gv))
4985 tmp = 0; /* overridden by import or by GLOBAL */
4988 && -tmp==KEY_lock /* XXX generalizable kludge */
4990 && !hv_fetchs(GvHVn(PL_incgv), "Thread.pm", FALSE))
4992 tmp = 0; /* any sub overrides "weak" keyword */
4994 else { /* no override */
4996 if (tmp == KEY_dump && ckWARN(WARN_MISC)) {
4997 Perl_warner(aTHX_ packWARN(WARN_MISC),
4998 "dump() better written as CORE::dump()");
5002 if (hgv && tmp != KEY_x && tmp != KEY_CORE
5003 && ckWARN(WARN_AMBIGUOUS)) /* never ambiguous */
5004 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
5005 "Ambiguous call resolved as CORE::%s(), %s",
5006 GvENAME(hgv), "qualify as such or use &");
5013 default: /* not a keyword */
5014 /* Trade off - by using this evil construction we can pull the
5015 variable gv into the block labelled keylookup. If not, then
5016 we have to give it function scope so that the goto from the
5017 earlier ':' case doesn't bypass the initialisation. */
5019 just_a_word_zero_gv:
5027 const char lastchar = (PL_bufptr == PL_oldoldbufptr ? 0 : PL_bufptr[-1]);
5030 SV *nextPL_nextwhite = 0;
5034 /* Get the rest if it looks like a package qualifier */
5036 if (*s == '\'' || (*s == ':' && s[1] == ':')) {
5038 s = scan_word(s, PL_tokenbuf + len, sizeof PL_tokenbuf - len,
5041 Perl_croak(aTHX_ "Bad name after %s%s", PL_tokenbuf,
5042 *s == '\'' ? "'" : "::");
5047 if (PL_expect == XOPERATOR) {
5048 if (PL_bufptr == PL_linestart) {
5049 CopLINE_dec(PL_curcop);
5050 Perl_warner(aTHX_ packWARN(WARN_SEMICOLON), PL_warn_nosemi);
5051 CopLINE_inc(PL_curcop);
5054 no_op("Bareword",s);
5057 /* Look for a subroutine with this name in current package,
5058 unless name is "Foo::", in which case Foo is a bearword
5059 (and a package name). */
5061 if (len > 2 && !PL_madskills &&
5062 PL_tokenbuf[len - 2] == ':' && PL_tokenbuf[len - 1] == ':')
5064 if (ckWARN(WARN_BAREWORD)
5065 && ! gv_fetchpvn_flags(PL_tokenbuf, len, 0, SVt_PVHV))
5066 Perl_warner(aTHX_ packWARN(WARN_BAREWORD),
5067 "Bareword \"%s\" refers to nonexistent package",
5070 PL_tokenbuf[len] = '\0';
5076 /* Mustn't actually add anything to a symbol table.
5077 But also don't want to "initialise" any placeholder
5078 constants that might already be there into full
5079 blown PVGVs with attached PVCV. */
5080 gv = gv_fetchpvn_flags(PL_tokenbuf, len,
5081 GV_NOADD_NOINIT, SVt_PVCV);
5086 /* if we saw a global override before, get the right name */
5089 sv = newSVpvs("CORE::GLOBAL::");
5090 sv_catpv(sv,PL_tokenbuf);
5093 /* If len is 0, newSVpv does strlen(), which is correct.
5094 If len is non-zero, then it will be the true length,
5095 and so the scalar will be created correctly. */
5096 sv = newSVpv(PL_tokenbuf,len);
5099 if (PL_madskills && !PL_thistoken) {
5100 char *start = SvPVX(PL_linestr) + PL_realtokenstart;
5101 PL_thistoken = newSVpv(start,s - start);
5102 PL_realtokenstart = s - SvPVX(PL_linestr);
5106 /* Presume this is going to be a bareword of some sort. */
5109 yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
5110 yylval.opval->op_private = OPpCONST_BARE;
5111 /* UTF-8 package name? */
5112 if (UTF && !IN_BYTES &&
5113 is_utf8_string((U8*)SvPVX_const(sv), SvCUR(sv)))
5116 /* And if "Foo::", then that's what it certainly is. */
5121 /* Do the explicit type check so that we don't need to force
5122 the initialisation of the symbol table to have a real GV.
5123 Beware - gv may not really be a PVGV, cv may not really be
5124 a PVCV, (because of the space optimisations that gv_init
5125 understands) But they're true if for this symbol there is
5126 respectively a typeglob and a subroutine.
5128 cv = gv ? ((SvTYPE(gv) == SVt_PVGV)
5129 /* Real typeglob, so get the real subroutine: */
5131 /* A proxy for a subroutine in this package? */
5132 : SvOK(gv) ? (CV *) gv : NULL)
5135 /* See if it's the indirect object for a list operator. */
5137 if (PL_oldoldbufptr &&
5138 PL_oldoldbufptr < PL_bufptr &&
5139 (PL_oldoldbufptr == PL_last_lop
5140 || PL_oldoldbufptr == PL_last_uni) &&
5141 /* NO SKIPSPACE BEFORE HERE! */
5142 (PL_expect == XREF ||
5143 ((PL_opargs[PL_last_lop_op] >> OASHIFT)& 7) == OA_FILEREF))
5145 bool immediate_paren = *s == '(';
5147 /* (Now we can afford to cross potential line boundary.) */
5148 s = SKIPSPACE2(s,nextPL_nextwhite);
5150 PL_nextwhite = nextPL_nextwhite; /* assume no & deception */
5153 /* Two barewords in a row may indicate method call. */
5155 if ((isIDFIRST_lazy_if(s,UTF) || *s == '$') &&
5156 (tmp = intuit_method(s, gv, cv)))
5159 /* If not a declared subroutine, it's an indirect object. */
5160 /* (But it's an indir obj regardless for sort.) */
5161 /* Also, if "_" follows a filetest operator, it's a bareword */
5164 ( !immediate_paren && (PL_last_lop_op == OP_SORT ||
5166 (PL_last_lop_op != OP_MAPSTART &&
5167 PL_last_lop_op != OP_GREPSTART))))
5168 || (PL_tokenbuf[0] == '_' && PL_tokenbuf[1] == '\0'
5169 && ((PL_opargs[PL_last_lop_op] & OA_CLASS_MASK) == OA_FILESTATOP))
5172 PL_expect = (PL_last_lop == PL_oldoldbufptr) ? XTERM : XOPERATOR;
5177 PL_expect = XOPERATOR;
5180 s = SKIPSPACE2(s,nextPL_nextwhite);
5181 PL_nextwhite = nextPL_nextwhite;
5186 /* Is this a word before a => operator? */
5187 if (*s == '=' && s[1] == '>' && !pkgname) {
5189 sv_setpv(((SVOP*)yylval.opval)->op_sv, PL_tokenbuf);
5190 if (UTF && !IN_BYTES && is_utf8_string((U8*)PL_tokenbuf, len))
5191 SvUTF8_on(((SVOP*)yylval.opval)->op_sv);
5195 /* If followed by a paren, it's certainly a subroutine. */
5199 for (d = s + 1; SPACE_OR_TAB(*d); d++) ;
5200 if (*d == ')' && (sv = gv_const_sv(gv))) {
5204 char *par = SvPVX(PL_linestr) + PL_realtokenstart;
5205 sv_catpvn(PL_thistoken, par, s - par);
5207 sv_free(PL_nextwhite);
5217 PL_nextwhite = PL_thiswhite;
5220 start_force(PL_curforce);
5222 NEXTVAL_NEXTTOKE.opval = yylval.opval;
5223 PL_expect = XOPERATOR;
5226 PL_nextwhite = nextPL_nextwhite;
5227 curmad('X', PL_thistoken);
5228 PL_thistoken = newSVpvn("",0);
5236 /* If followed by var or block, call it a method (unless sub) */
5238 if ((*s == '$' || *s == '{') && (!gv || !cv)) {
5239 PL_last_lop = PL_oldbufptr;
5240 PL_last_lop_op = OP_METHOD;
5244 /* If followed by a bareword, see if it looks like indir obj. */
5247 && (isIDFIRST_lazy_if(s,UTF) || *s == '$')
5248 && (tmp = intuit_method(s, gv, cv)))
5251 /* Not a method, so call it a subroutine (if defined) */
5254 if (lastchar == '-' && ckWARN_d(WARN_AMBIGUOUS))
5255 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
5256 "Ambiguous use of -%s resolved as -&%s()",
5257 PL_tokenbuf, PL_tokenbuf);
5258 /* Check for a constant sub */
5259 if ((sv = gv_const_sv(gv))) {
5261 SvREFCNT_dec(((SVOP*)yylval.opval)->op_sv);
5262 ((SVOP*)yylval.opval)->op_sv = SvREFCNT_inc_simple(sv);
5263 yylval.opval->op_private = 0;
5267 /* Resolve to GV now. */
5268 if (SvTYPE(gv) != SVt_PVGV) {
5269 gv = gv_fetchpv(PL_tokenbuf, 0, SVt_PVCV);
5270 assert (SvTYPE(gv) == SVt_PVGV);
5271 /* cv must have been some sort of placeholder, so
5272 now needs replacing with a real code reference. */
5276 op_free(yylval.opval);
5277 yylval.opval = newCVREF(0, newGVOP(OP_GV, 0, gv));
5278 yylval.opval->op_private |= OPpENTERSUB_NOPAREN;
5279 PL_last_lop = PL_oldbufptr;
5280 PL_last_lop_op = OP_ENTERSUB;
5281 /* Is there a prototype? */
5288 const char *proto = SvPV_const((SV*)cv, protolen);
5291 if (*proto == '$' && proto[1] == '\0')
5293 while (*proto == ';')
5295 if (*proto == '&' && *s == '{') {
5296 sv_setpv(PL_subname, PL_curstash ?
5297 "__ANON__" : "__ANON__::__ANON__");
5304 PL_nextwhite = PL_thiswhite;
5307 start_force(PL_curforce);
5308 NEXTVAL_NEXTTOKE.opval = yylval.opval;
5311 PL_nextwhite = nextPL_nextwhite;
5312 curmad('X', PL_thistoken);
5313 PL_thistoken = newSVpvn("",0);
5320 /* Guess harder when madskills require "best effort". */
5321 if (PL_madskills && (!gv || !GvCVu(gv))) {
5322 int probable_sub = 0;
5323 if (strchr("\"'`$@%0123456789!*+{[<", *s))
5325 else if (isALPHA(*s)) {
5329 d = scan_word(d, tmpbuf, sizeof tmpbuf, TRUE, &tmplen);
5330 if (!keyword(tmpbuf,tmplen))
5333 while (d < PL_bufend && isSPACE(*d))
5335 if (*d == '=' && d[1] == '>')
5340 gv = gv_fetchpv(PL_tokenbuf, TRUE, SVt_PVCV);
5341 op_free(yylval.opval);
5342 yylval.opval = newCVREF(0, newGVOP(OP_GV, 0, gv));
5343 yylval.opval->op_private |= OPpENTERSUB_NOPAREN;
5344 PL_last_lop = PL_oldbufptr;
5345 PL_last_lop_op = OP_ENTERSUB;
5346 PL_nextwhite = PL_thiswhite;
5348 start_force(PL_curforce);
5349 NEXTVAL_NEXTTOKE.opval = yylval.opval;
5351 PL_nextwhite = nextPL_nextwhite;
5352 curmad('X', PL_thistoken);
5353 PL_thistoken = newSVpvn("",0);
5358 NEXTVAL_NEXTTOKE.opval = yylval.opval;
5365 /* Call it a bare word */
5367 if (PL_hints & HINT_STRICT_SUBS)
5368 yylval.opval->op_private |= OPpCONST_STRICT;
5371 if (lastchar != '-') {
5372 if (ckWARN(WARN_RESERVED)) {
5373 for (d = PL_tokenbuf; *d && isLOWER(*d); d++) ;
5374 if (!*d && !gv_stashpv(PL_tokenbuf,FALSE))
5375 Perl_warner(aTHX_ packWARN(WARN_RESERVED), PL_warn_reserved,
5382 if ((lastchar == '*' || lastchar == '%' || lastchar == '&')
5383 && ckWARN_d(WARN_AMBIGUOUS)) {
5384 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
5385 "Operator or semicolon missing before %c%s",
5386 lastchar, PL_tokenbuf);
5387 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
5388 "Ambiguous use of %c resolved as operator %c",
5389 lastchar, lastchar);
5395 yylval.opval = (OP*)newSVOP(OP_CONST, 0,
5396 newSVpv(CopFILE(PL_curcop),0));
5400 yylval.opval = (OP*)newSVOP(OP_CONST, 0,
5401 Perl_newSVpvf(aTHX_ "%"IVdf, (IV)CopLINE(PL_curcop)));
5404 case KEY___PACKAGE__:
5405 yylval.opval = (OP*)newSVOP(OP_CONST, 0,
5407 ? newSVhek(HvNAME_HEK(PL_curstash))
5414 if (PL_rsfp && (!PL_in_eval || PL_tokenbuf[2] == 'D')) {
5415 const char *pname = "main";
5416 if (PL_tokenbuf[2] == 'D')
5417 pname = HvNAME_get(PL_curstash ? PL_curstash : PL_defstash);
5418 gv = gv_fetchpv(Perl_form(aTHX_ "%s::DATA", pname), GV_ADD,
5422 GvIOp(gv) = newIO();
5423 IoIFP(GvIOp(gv)) = PL_rsfp;
5424 #if defined(HAS_FCNTL) && defined(F_SETFD)
5426 const int fd = PerlIO_fileno(PL_rsfp);
5427 fcntl(fd,F_SETFD,fd >= 3);
5430 /* Mark this internal pseudo-handle as clean */
5431 IoFLAGS(GvIOp(gv)) |= IOf_UNTAINT;
5433 IoTYPE(GvIOp(gv)) = IoTYPE_PIPE;
5434 else if ((PerlIO*)PL_rsfp == PerlIO_stdin())
5435 IoTYPE(GvIOp(gv)) = IoTYPE_STD;
5437 IoTYPE(GvIOp(gv)) = IoTYPE_RDONLY;
5438 #if defined(WIN32) && !defined(PERL_TEXTMODE_SCRIPTS)
5439 /* if the script was opened in binmode, we need to revert
5440 * it to text mode for compatibility; but only iff it has CRs
5441 * XXX this is a questionable hack at best. */
5442 if (PL_bufend-PL_bufptr > 2
5443 && PL_bufend[-1] == '\n' && PL_bufend[-2] == '\r')
5446 if (IoTYPE(GvIOp(gv)) == IoTYPE_RDONLY) {
5447 loc = PerlIO_tell(PL_rsfp);
5448 (void)PerlIO_seek(PL_rsfp, 0L, 0);
5451 if (PerlLIO_setmode(PL_rsfp, O_TEXT) != -1) {
5453 if (PerlLIO_setmode(PerlIO_fileno(PL_rsfp), O_TEXT) != -1) {
5454 #endif /* NETWARE */
5455 #ifdef PERLIO_IS_STDIO /* really? */
5456 # if defined(__BORLANDC__)
5457 /* XXX see note in do_binmode() */
5458 ((FILE*)PL_rsfp)->flags &= ~_F_BIN;
5462 PerlIO_seek(PL_rsfp, loc, 0);
5466 #ifdef PERLIO_LAYERS
5469 PerlIO_apply_layers(aTHX_ PL_rsfp, NULL, ":utf8");
5470 else if (PL_encoding) {
5477 XPUSHs(PL_encoding);
5479 call_method("name", G_SCALAR);
5483 PerlIO_apply_layers(aTHX_ PL_rsfp, NULL,
5484 Perl_form(aTHX_ ":encoding(%"SVf")",
5493 if (PL_realtokenstart >= 0) {
5494 char *tstart = SvPVX(PL_linestr) + PL_realtokenstart;
5496 PL_endwhite = newSVpvn("",0);
5497 sv_catsv(PL_endwhite, PL_thiswhite);
5499 sv_catpvn(PL_endwhite, tstart, PL_bufend - tstart);
5500 PL_realtokenstart = -1;
5502 while ((s = filter_gets(PL_endwhite, PL_rsfp,
5503 SvCUR(PL_endwhite))) != Nullch) ;
5517 if (PL_expect == XSTATE) {
5524 if (*s == ':' && s[1] == ':') {
5527 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
5528 if (!(tmp = keyword(PL_tokenbuf, len)))
5529 Perl_croak(aTHX_ "CORE::%s is not a keyword", PL_tokenbuf);
5532 else if (tmp == KEY_require || tmp == KEY_do)
5533 /* that's a way to remember we saw "CORE::" */
5546 LOP(OP_ACCEPT,XTERM);
5552 LOP(OP_ATAN2,XTERM);
5558 LOP(OP_BINMODE,XTERM);
5561 LOP(OP_BLESS,XTERM);
5570 /* When 'use switch' is in effect, continue has a dual
5571 life as a control operator. */
5573 if (!FEATURE_IS_ENABLED("switch"))
5576 /* We have to disambiguate the two senses of
5577 "continue". If the next token is a '{' then
5578 treat it as the start of a continue block;
5579 otherwise treat it as a control operator.
5591 (void)gv_fetchpvs("ENV", GV_ADD|GV_NOTQUAL, SVt_PVHV);
5608 if (!PL_cryptseen) {
5609 PL_cryptseen = TRUE;
5613 LOP(OP_CRYPT,XTERM);
5616 LOP(OP_CHMOD,XTERM);
5619 LOP(OP_CHOWN,XTERM);
5622 LOP(OP_CONNECT,XTERM);
5641 s = force_word(s,WORD,TRUE,TRUE,FALSE);
5642 if (orig_keyword == KEY_do) {
5651 PL_hints |= HINT_BLOCK_SCOPE;
5661 gv_fetchpvs("AnyDBM_File::ISA", GV_ADDMULTI, SVt_PVAV);
5662 LOP(OP_DBMOPEN,XTERM);
5668 s = force_word(s,WORD,TRUE,FALSE,FALSE);
5675 yylval.ival = CopLINE(PL_curcop);
5691 PL_expect = (*s == '{') ? XTERMBLOCK : XTERM;
5692 UNIBRACK(OP_ENTEREVAL);
5710 case KEY_endhostent:
5716 case KEY_endservent:
5719 case KEY_endprotoent:
5730 yylval.ival = CopLINE(PL_curcop);
5732 if (PL_expect == XSTATE && isIDFIRST_lazy_if(s,UTF)) {
5735 int soff = s - SvPVX(PL_linestr); /* for skipspace realloc */
5738 if ((PL_bufend - p) >= 3 &&
5739 strnEQ(p, "my", 2) && isSPACE(*(p + 2)))
5741 else if ((PL_bufend - p) >= 4 &&
5742 strnEQ(p, "our", 3) && isSPACE(*(p + 3)))
5745 if (isIDFIRST_lazy_if(p,UTF)) {
5746 p = scan_ident(p, PL_bufend,
5747 PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
5751 Perl_croak(aTHX_ "Missing $ on loop variable");
5753 s = SvPVX(PL_linestr) + soff;
5759 LOP(OP_FORMLINE,XTERM);
5765 LOP(OP_FCNTL,XTERM);
5771 LOP(OP_FLOCK,XTERM);
5780 LOP(OP_GREPSTART, XREF);
5783 s = force_word(s,WORD,TRUE,FALSE,FALSE);
5798 case KEY_getpriority:
5799 LOP(OP_GETPRIORITY,XTERM);
5801 case KEY_getprotobyname:
5804 case KEY_getprotobynumber:
5805 LOP(OP_GPBYNUMBER,XTERM);
5807 case KEY_getprotoent:
5819 case KEY_getpeername:
5820 UNI(OP_GETPEERNAME);
5822 case KEY_gethostbyname:
5825 case KEY_gethostbyaddr:
5826 LOP(OP_GHBYADDR,XTERM);
5828 case KEY_gethostent:
5831 case KEY_getnetbyname:
5834 case KEY_getnetbyaddr:
5835 LOP(OP_GNBYADDR,XTERM);
5840 case KEY_getservbyname:
5841 LOP(OP_GSBYNAME,XTERM);
5843 case KEY_getservbyport:
5844 LOP(OP_GSBYPORT,XTERM);
5846 case KEY_getservent:
5849 case KEY_getsockname:
5850 UNI(OP_GETSOCKNAME);
5852 case KEY_getsockopt:
5853 LOP(OP_GSOCKOPT,XTERM);
5868 yylval.ival = CopLINE(PL_curcop);
5879 yylval.ival = CopLINE(PL_curcop);
5883 LOP(OP_INDEX,XTERM);
5889 LOP(OP_IOCTL,XTERM);
5901 s = force_word(s,WORD,TRUE,FALSE,FALSE);
5933 LOP(OP_LISTEN,XTERM);
5942 s = scan_pat(s,OP_MATCH);
5943 TERM(sublex_start());
5946 LOP(OP_MAPSTART, XREF);
5949 LOP(OP_MKDIR,XTERM);
5952 LOP(OP_MSGCTL,XTERM);
5955 LOP(OP_MSGGET,XTERM);
5958 LOP(OP_MSGRCV,XTERM);
5961 LOP(OP_MSGSND,XTERM);
5968 if (isIDFIRST_lazy_if(s,UTF)) {
5972 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, TRUE, &len);
5973 if (len == 3 && strnEQ(PL_tokenbuf, "sub", 3))
5975 PL_in_my_stash = find_in_my_stash(PL_tokenbuf, len);
5976 if (!PL_in_my_stash) {
5980 snprintf(tmpbuf, sizeof(tmpbuf), "No such class %.1000s", PL_tokenbuf);
5982 sprintf(tmpbuf, "No such class %.1000s", PL_tokenbuf);
5983 #endif /* #ifdef USE_SNPRINTF */
5987 if (PL_madskills) { /* just add type to declarator token */
5988 sv_catsv(PL_thistoken, PL_nextwhite);
5990 sv_catpvn(PL_thistoken, start, s - start);
5998 s = force_word(s,WORD,TRUE,FALSE,FALSE);
6005 s = tokenize_use(0, s);
6009 if (*s == '(' || (s = SKIPSPACE1(s), *s == '('))
6016 if (isIDFIRST_lazy_if(s,UTF)) {
6018 for (d = s; isALNUM_lazy_if(d,UTF); d++) ;
6019 for (t=d; *t && isSPACE(*t); t++) ;
6020 if ( *t && strchr("|&*+-=!?:.", *t) && ckWARN_d(WARN_PRECEDENCE)
6022 && !(t[0] == '=' && t[1] == '>')
6024 int parms_len = (int)(d-s);
6025 Perl_warner(aTHX_ packWARN(WARN_PRECEDENCE),
6026 "Precedence problem: open %.*s should be open(%.*s)",
6027 parms_len, s, parms_len, s);
6033 yylval.ival = OP_OR;
6043 LOP(OP_OPEN_DIR,XTERM);
6046 checkcomma(s,PL_tokenbuf,"filehandle");
6050 checkcomma(s,PL_tokenbuf,"filehandle");
6069 s = force_word(s,WORD,FALSE,TRUE,FALSE);
6073 LOP(OP_PIPE_OP,XTERM);
6076 s = scan_str(s,!!PL_madskills,FALSE);
6079 yylval.ival = OP_CONST;
6080 TERM(sublex_start());
6086 s = scan_str(s,!!PL_madskills,FALSE);
6089 PL_expect = XOPERATOR;
6091 if (SvCUR(PL_lex_stuff)) {
6094 d = SvPV_force(PL_lex_stuff, len);
6096 for (; isSPACE(*d) && len; --len, ++d)
6101 if (!warned && ckWARN(WARN_QW)) {
6102 for (; !isSPACE(*d) && len; --len, ++d) {
6104 Perl_warner(aTHX_ packWARN(WARN_QW),
6105 "Possible attempt to separate words with commas");
6108 else if (*d == '#') {
6109 Perl_warner(aTHX_ packWARN(WARN_QW),
6110 "Possible attempt to put comments in qw() list");
6116 for (; !isSPACE(*d) && len; --len, ++d)
6119 sv = newSVpvn(b, d-b);
6120 if (DO_UTF8(PL_lex_stuff))
6122 words = append_elem(OP_LIST, words,
6123 newSVOP(OP_CONST, 0, tokeq(sv)));
6127 start_force(PL_curforce);
6128 NEXTVAL_NEXTTOKE.opval = words;
6133 SvREFCNT_dec(PL_lex_stuff);
6134 PL_lex_stuff = NULL;
6140 s = scan_str(s,!!PL_madskills,FALSE);
6143 yylval.ival = OP_STRINGIFY;
6144 if (SvIVX(PL_lex_stuff) == '\'')
6145 SvIV_set(PL_lex_stuff, 0); /* qq'$foo' should intepolate */
6146 TERM(sublex_start());
6149 s = scan_pat(s,OP_QR);
6150 TERM(sublex_start());
6153 s = scan_str(s,!!PL_madskills,FALSE);
6156 yylval.ival = OP_BACKTICK;
6158 TERM(sublex_start());
6166 s = force_version(s, FALSE);
6168 else if (*s != 'v' || !isDIGIT(s[1])
6169 || (s = force_version(s, TRUE), *s == 'v'))
6171 *PL_tokenbuf = '\0';
6172 s = force_word(s,WORD,TRUE,TRUE,FALSE);
6173 if (isIDFIRST_lazy_if(PL_tokenbuf,UTF))
6174 gv_stashpvn(PL_tokenbuf, strlen(PL_tokenbuf), TRUE);
6176 yyerror("<> should be quotes");
6178 if (orig_keyword == KEY_require) {
6186 PL_last_uni = PL_oldbufptr;
6187 PL_last_lop_op = OP_REQUIRE;
6189 return REPORT( (int)REQUIRE );
6195 s = force_word(s,WORD,TRUE,FALSE,FALSE);
6199 LOP(OP_RENAME,XTERM);
6208 LOP(OP_RINDEX,XTERM);
6218 UNIDOR(OP_READLINE);
6231 LOP(OP_REVERSE,XTERM);
6234 UNIDOR(OP_READLINK);
6242 TERM(sublex_start());
6244 TOKEN(1); /* force error */
6247 checkcomma(s,PL_tokenbuf,"filehandle");
6257 LOP(OP_SELECT,XTERM);
6263 LOP(OP_SEMCTL,XTERM);
6266 LOP(OP_SEMGET,XTERM);
6269 LOP(OP_SEMOP,XTERM);
6275 LOP(OP_SETPGRP,XTERM);
6277 case KEY_setpriority:
6278 LOP(OP_SETPRIORITY,XTERM);
6280 case KEY_sethostent:
6286 case KEY_setservent:
6289 case KEY_setprotoent:
6299 LOP(OP_SEEKDIR,XTERM);
6301 case KEY_setsockopt:
6302 LOP(OP_SSOCKOPT,XTERM);
6308 LOP(OP_SHMCTL,XTERM);
6311 LOP(OP_SHMGET,XTERM);
6314 LOP(OP_SHMREAD,XTERM);
6317 LOP(OP_SHMWRITE,XTERM);
6320 LOP(OP_SHUTDOWN,XTERM);
6329 LOP(OP_SOCKET,XTERM);
6331 case KEY_socketpair:
6332 LOP(OP_SOCKPAIR,XTERM);
6335 checkcomma(s,PL_tokenbuf,"subroutine name");
6337 if (*s == ';' || *s == ')') /* probably a close */
6338 Perl_croak(aTHX_ "sort is now a reserved word");
6340 s = force_word(s,WORD,TRUE,TRUE,FALSE);
6344 LOP(OP_SPLIT,XTERM);
6347 LOP(OP_SPRINTF,XTERM);
6350 LOP(OP_SPLICE,XTERM);
6365 LOP(OP_SUBSTR,XTERM);
6371 char tmpbuf[sizeof PL_tokenbuf];
6372 SSize_t tboffset = 0;
6373 expectation attrful;
6374 bool have_name, have_proto, bad_proto;
6375 const int key = tmp;
6380 char *tstart = SvPVX(PL_linestr) + PL_realtokenstart;
6381 SV *subtoken = newSVpvn(tstart, s - tstart);
6385 s = SKIPSPACE2(s,tmpwhite);
6390 if (isIDFIRST_lazy_if(s,UTF) || *s == '\'' ||
6391 (*s == ':' && s[1] == ':'))
6398 attrful = XATTRBLOCK;
6399 /* remember buffer pos'n for later force_word */
6400 tboffset = s - PL_oldbufptr;
6401 d = scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
6404 nametoke = newSVpvn(s, d - s);
6406 if (strchr(tmpbuf, ':'))
6407 sv_setpv(PL_subname, tmpbuf);
6409 sv_setsv(PL_subname,PL_curstname);
6410 sv_catpvs(PL_subname,"::");
6411 sv_catpvn(PL_subname,tmpbuf,len);
6418 CURMAD('X', nametoke);
6419 CURMAD('_', tmpwhite);
6420 (void) force_word(PL_oldbufptr + tboffset, WORD,
6423 s = SKIPSPACE2(d,tmpwhite);
6430 Perl_croak(aTHX_ "Missing name in \"my sub\"");
6431 PL_expect = XTERMBLOCK;
6432 attrful = XATTRTERM;
6433 sv_setpvn(PL_subname,"?",1);
6437 if (key == KEY_format) {
6439 PL_lex_formbrack = PL_lex_brackets + 1;
6441 PL_thistoken = subtoken;
6445 (void) force_word(PL_oldbufptr + tboffset, WORD,
6451 /* Look for a prototype */
6455 s = scan_str(s,!!PL_madskills,FALSE);
6457 Perl_croak(aTHX_ "Prototype not terminated");
6458 /* strip spaces and check for bad characters */
6459 d = SvPVX(PL_lex_stuff);
6462 for (p = d; *p; ++p) {
6465 if (!strchr("$@%*;[]&\\", *p))
6470 if (bad_proto && ckWARN(WARN_SYNTAX))
6471 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
6472 "Illegal character in prototype for %"SVf" : %s",
6473 (void*)PL_subname, d);
6474 SvCUR_set(PL_lex_stuff, tmp);
6479 CURMAD('q', PL_thisopen);
6480 CURMAD('_', tmpwhite);
6481 CURMAD('=', PL_thisstuff);
6482 CURMAD('Q', PL_thisclose);
6483 NEXTVAL_NEXTTOKE.opval =
6484 (OP*)newSVOP(OP_CONST, 0, PL_lex_stuff);
6485 PL_lex_stuff = Nullsv;
6488 s = SKIPSPACE2(s,tmpwhite);
6496 if (*s == ':' && s[1] != ':')
6497 PL_expect = attrful;
6498 else if (*s != '{' && key == KEY_sub) {
6500 Perl_croak(aTHX_ "Illegal declaration of anonymous subroutine");
6502 Perl_croak(aTHX_ "Illegal declaration of subroutine %"SVf, (void*)PL_subname);
6509 curmad('^', newSVpvn("",0));
6510 CURMAD('_', tmpwhite);
6514 PL_thistoken = subtoken;
6517 NEXTVAL_NEXTTOKE.opval =
6518 (OP*)newSVOP(OP_CONST, 0, PL_lex_stuff);
6519 PL_lex_stuff = NULL;
6524 sv_setpv(PL_subname,
6525 PL_curstash ? "__ANON__" : "__ANON__::__ANON__");
6529 (void) force_word(PL_oldbufptr + tboffset, WORD,
6539 LOP(OP_SYSTEM,XREF);
6542 LOP(OP_SYMLINK,XTERM);
6545 LOP(OP_SYSCALL,XTERM);
6548 LOP(OP_SYSOPEN,XTERM);
6551 LOP(OP_SYSSEEK,XTERM);
6554 LOP(OP_SYSREAD,XTERM);
6557 LOP(OP_SYSWRITE,XTERM);
6561 TERM(sublex_start());
6582 LOP(OP_TRUNCATE,XTERM);
6594 yylval.ival = CopLINE(PL_curcop);
6598 yylval.ival = CopLINE(PL_curcop);
6602 LOP(OP_UNLINK,XTERM);
6608 LOP(OP_UNPACK,XTERM);
6611 LOP(OP_UTIME,XTERM);
6617 LOP(OP_UNSHIFT,XTERM);
6620 s = tokenize_use(1, s);
6630 yylval.ival = CopLINE(PL_curcop);
6634 yylval.ival = CopLINE(PL_curcop);
6638 PL_hints |= HINT_BLOCK_SCOPE;
6645 LOP(OP_WAITPID,XTERM);
6654 ctl_l[0] = toCTRL('L');
6656 gv_fetchpvn_flags(ctl_l, 1, GV_ADD|GV_NOTQUAL, SVt_PV);
6659 /* Make sure $^L is defined */
6660 gv_fetchpvs("\f", GV_ADD|GV_NOTQUAL, SVt_PV);
6665 if (PL_expect == XOPERATOR)
6671 yylval.ival = OP_XOR;
6676 TERM(sublex_start());
6681 #pragma segment Main
6685 S_pending_ident(pTHX)
6689 register PADOFFSET tmp = 0;
6690 /* pit holds the identifier we read and pending_ident is reset */
6691 char pit = PL_pending_ident;
6692 PL_pending_ident = 0;
6694 /* PL_realtokenstart = realtokenend = PL_bufptr - SvPVX(PL_linestr); */
6695 DEBUG_T({ PerlIO_printf(Perl_debug_log,
6696 "### Pending identifier '%s'\n", PL_tokenbuf); });
6698 /* if we're in a my(), we can't allow dynamics here.
6699 $foo'bar has already been turned into $foo::bar, so
6700 just check for colons.
6702 if it's a legal name, the OP is a PADANY.
6705 if (PL_in_my == KEY_our) { /* "our" is merely analogous to "my" */
6706 if (strchr(PL_tokenbuf,':'))
6707 yyerror(Perl_form(aTHX_ "No package name allowed for "
6708 "variable %s in \"our\"",
6710 tmp = allocmy(PL_tokenbuf);
6713 if (strchr(PL_tokenbuf,':'))
6714 yyerror(Perl_form(aTHX_ PL_no_myglob,
6715 PL_in_my == KEY_my ? "my" : "state", PL_tokenbuf));
6717 yylval.opval = newOP(OP_PADANY, 0);
6718 yylval.opval->op_targ = allocmy(PL_tokenbuf);
6724 build the ops for accesses to a my() variable.
6726 Deny my($a) or my($b) in a sort block, *if* $a or $b is
6727 then used in a comparison. This catches most, but not
6728 all cases. For instance, it catches
6729 sort { my($a); $a <=> $b }
6731 sort { my($a); $a < $b ? -1 : $a == $b ? 0 : 1; }
6732 (although why you'd do that is anyone's guess).
6735 if (!strchr(PL_tokenbuf,':')) {
6737 tmp = pad_findmy(PL_tokenbuf);
6738 if (tmp != NOT_IN_PAD) {
6739 /* might be an "our" variable" */
6740 if (PAD_COMPNAME_FLAGS_isOUR(tmp)) {
6741 /* build ops for a bareword */
6742 HV * const stash = PAD_COMPNAME_OURSTASH(tmp);
6743 HEK * const stashname = HvNAME_HEK(stash);
6744 SV * const sym = newSVhek(stashname);
6745 sv_catpvs(sym, "::");
6746 sv_catpv(sym, PL_tokenbuf+1);
6747 yylval.opval = (OP*)newSVOP(OP_CONST, 0, sym);
6748 yylval.opval->op_private = OPpCONST_ENTERED;
6751 ? (GV_ADDMULTI | GV_ADDINEVAL)
6754 ((PL_tokenbuf[0] == '$') ? SVt_PV
6755 : (PL_tokenbuf[0] == '@') ? SVt_PVAV
6760 /* if it's a sort block and they're naming $a or $b */
6761 if (PL_last_lop_op == OP_SORT &&
6762 PL_tokenbuf[0] == '$' &&
6763 (PL_tokenbuf[1] == 'a' || PL_tokenbuf[1] == 'b')
6766 for (d = PL_in_eval ? PL_oldoldbufptr : PL_linestart;
6767 d < PL_bufend && *d != '\n';
6770 if (strnEQ(d,"<=>",3) || strnEQ(d,"cmp",3)) {
6771 Perl_croak(aTHX_ "Can't use \"my %s\" in sort comparison",
6777 yylval.opval = newOP(OP_PADANY, 0);
6778 yylval.opval->op_targ = tmp;
6784 Whine if they've said @foo in a doublequoted string,
6785 and @foo isn't a variable we can find in the symbol
6788 if (pit == '@' && PL_lex_state != LEX_NORMAL && !PL_lex_brackets) {
6789 GV *gv = gv_fetchpv(PL_tokenbuf+1, 0, SVt_PVAV);
6790 if ((!gv || ((PL_tokenbuf[0] == '@') ? !GvAV(gv) : !GvHV(gv)))
6791 && ckWARN(WARN_AMBIGUOUS))
6793 /* Downgraded from fatal to warning 20000522 mjd */
6794 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
6795 "Possible unintended interpolation of %s in string",
6800 /* build ops for a bareword */
6801 yylval.opval = (OP*)newSVOP(OP_CONST, 0, newSVpv(PL_tokenbuf+1, 0));
6802 yylval.opval->op_private = OPpCONST_ENTERED;
6805 /* If the identifier refers to a stash, don't autovivify it.
6806 * Change 24660 had the side effect of causing symbol table
6807 * hashes to always be defined, even if they were freshly
6808 * created and the only reference in the entire program was
6809 * the single statement with the defined %foo::bar:: test.
6810 * It appears that all code in the wild doing this actually
6811 * wants to know whether sub-packages have been loaded, so
6812 * by avoiding auto-vivifying symbol tables, we ensure that
6813 * defined %foo::bar:: continues to be false, and the existing
6814 * tests still give the expected answers, even though what
6815 * they're actually testing has now changed subtly.
6817 (*PL_tokenbuf == '%' && *(d = PL_tokenbuf + strlen(PL_tokenbuf) - 1) == ':' && d[-1] == ':'
6819 : PL_in_eval ? (GV_ADDMULTI | GV_ADDINEVAL) : GV_ADD),
6820 ((PL_tokenbuf[0] == '$') ? SVt_PV
6821 : (PL_tokenbuf[0] == '@') ? SVt_PVAV
6827 * The following code was generated by perl_keyword.pl.
6831 Perl_keyword (pTHX_ const char *name, I32 len)
6836 case 1: /* 5 tokens of length 1 */
6868 case 2: /* 18 tokens of length 2 */
7014 case 3: /* 29 tokens of length 3 */
7018 if (name[1] == 'N' &&
7081 if (name[1] == 'i' &&
7103 return (FEATURE_IS_ENABLED("err") ? -KEY_err : 0);
7121 if (name[1] == 'o' &&
7130 if (name[1] == 'e' &&
7139 if (name[1] == 'n' &&
7148 if (name[1] == 'o' &&
7157 if (name[1] == 'a' &&
7166 if (name[1] == 'o' &&
7228 if (name[1] == 'e' &&
7242 return (FEATURE_IS_ENABLED("say") ? -KEY_say : 0);
7268 if (name[1] == 'i' &&
7277 if (name[1] == 's' &&
7286 if (name[1] == 'e' &&
7295 if (name[1] == 'o' &&
7307 case 4: /* 41 tokens of length 4 */
7311 if (name[1] == 'O' &&
7321 if (name[1] == 'N' &&
7331 if (name[1] == 'i' &&
7341 if (name[1] == 'h' &&
7351 if (name[1] == 'u' &&
7364 if (name[2] == 'c' &&
7373 if (name[2] == 's' &&
7382 if (name[2] == 'a' &&
7418 if (name[1] == 'o' &&
7431 if (name[2] == 't' &&
7440 if (name[2] == 'o' &&
7449 if (name[2] == 't' &&
7458 if (name[2] == 'e' &&
7471 if (name[1] == 'o' &&
7484 if (name[2] == 'y' &&
7493 if (name[2] == 'l' &&
7509 if (name[2] == 's' &&
7518 if (name[2] == 'n' &&
7527 if (name[2] == 'c' &&
7540 if (name[1] == 'e' &&
7550 if (name[1] == 'p' &&
7563 if (name[2] == 'c' &&
7572 if (name[2] == 'p' &&
7581 if (name[2] == 's' &&
7597 if (name[2] == 'n' &&
7667 if (name[2] == 'r' &&
7676 if (name[2] == 'r' &&
7685 if (name[2] == 'a' &&
7701 if (name[2] == 'l' &&
7763 if (name[2] == 'e' &&
7766 return (FEATURE_IS_ENABLED("switch") ? KEY_when : 0);
7779 case 5: /* 39 tokens of length 5 */
7783 if (name[1] == 'E' &&
7794 if (name[1] == 'H' &&
7808 if (name[2] == 'a' &&
7818 if (name[2] == 'a' &&
7835 if (name[2] == 'e' &&
7845 if (name[2] == 'e' &&
7849 return (FEATURE_IS_ENABLED("switch") ? -KEY_break : 0);
7865 if (name[3] == 'i' &&
7874 if (name[3] == 'o' &&
7910 if (name[2] == 'o' &&
7920 if (name[2] == 'y' &&
7934 if (name[1] == 'l' &&
7948 if (name[2] == 'n' &&
7958 if (name[2] == 'o' &&
7972 if (name[1] == 'i' &&
7977 return (FEATURE_IS_ENABLED("switch") ? KEY_given : 0);
7986 if (name[2] == 'd' &&
7996 if (name[2] == 'c' &&
8013 if (name[2] == 'c' &&
8023 if (name[2] == 't' &&
8037 if (name[1] == 'k' &&
8048 if (name[1] == 'r' &&
8062 if (name[2] == 's' &&
8072 if (name[2] == 'd' &&
8089 if (name[2] == 'm' &&
8099 if (name[2] == 'i' &&
8109 if (name[2] == 'e' &&
8119 if (name[2] == 'l' &&
8129 if (name[2] == 'a' &&
8142 if (name[3] == 't' &&
8145 return (FEATURE_IS_ENABLED("state") ? KEY_state : 0);
8151 if (name[3] == 'd' &&
8168 if (name[1] == 'i' &&
8182 if (name[2] == 'a' &&
8195 if (name[3] == 'e' &&
8230 if (name[2] == 'i' &&
8247 if (name[2] == 'i' &&
8257 if (name[2] == 'i' &&
8274 case 6: /* 33 tokens of length 6 */
8278 if (name[1] == 'c' &&
8293 if (name[2] == 'l' &&
8304 if (name[2] == 'r' &&
8319 if (name[1] == 'e' &&
8334 if (name[2] == 's' &&
8339 if(ckWARN_d(WARN_SYNTAX))
8340 Perl_warner(aTHX_ packWARN(WARN_SYNTAX), "elseif should be elsif");
8346 if (name[2] == 'i' &&
8364 if (name[2] == 'l' &&
8375 if (name[2] == 'r' &&
8390 if (name[1] == 'm' &&
8405 if (name[2] == 'n' &&
8416 if (name[2] == 's' &&
8431 if (name[1] == 's' &&
8437 if (name[4] == 't' &&
8446 if (name[4] == 'e' &&
8455 if (name[4] == 'c' &&
8464 if (name[4] == 'n' &&
8480 if (name[1] == 'r' &&
8498 if (name[3] == 'a' &&
8508 if (name[3] == 'u' &&
8522 if (name[2] == 'n' &&
8540 if (name[2] == 'a' &&
8554 if (name[3] == 'e' &&
8567 if (name[4] == 't' &&
8576 if (name[4] == 'e' &&
8598 if (name[4] == 't' &&
8607 if (name[4] == 'e' &&
8623 if (name[2] == 'c' &&
8634 if (name[2] == 'l' &&
8645 if (name[2] == 'b' &&
8656 if (name[2] == 's' &&
8679 if (name[4] == 's' &&
8688 if (name[4] == 'n' &&
8701 if (name[3] == 'a' &&
8718 if (name[1] == 'a' &&
8733 case 7: /* 29 tokens of length 7 */
8737 if (name[1] == 'E' &&
8750 if (name[1] == '_' &&
8763 if (name[1] == 'i' &&
8770 return -KEY_binmode;
8776 if (name[1] == 'o' &&
8783 return -KEY_connect;
8792 if (name[2] == 'm' &&
8798 return -KEY_dbmopen;
8809 if (name[4] == 'u' &&
8813 return (FEATURE_IS_ENABLED("switch") ? KEY_default : 0);
8819 if (name[4] == 'n' &&
8840 if (name[1] == 'o' &&
8853 if (name[1] == 'e' &&
8860 if (name[5] == 'r' &&
8863 return -KEY_getpgrp;
8869 if (name[5] == 'i' &&
8872 return -KEY_getppid;
8885 if (name[1] == 'c' &&
8892 return -KEY_lcfirst;
8898 if (name[1] == 'p' &&
8905 return -KEY_opendir;
8911 if (name[1] == 'a' &&
8929 if (name[3] == 'd' &&
8934 return -KEY_readdir;
8940 if (name[3] == 'u' &&
8951 if (name[3] == 'e' &&
8956 return -KEY_reverse;
8975 if (name[3] == 'k' &&
8980 return -KEY_seekdir;
8986 if (name[3] == 'p' &&
8991 return -KEY_setpgrp;
9001 if (name[2] == 'm' &&
9007 return -KEY_shmread;
9013 if (name[2] == 'r' &&
9019 return -KEY_sprintf;
9028 if (name[3] == 'l' &&
9033 return -KEY_symlink;
9042 if (name[4] == 'a' &&
9046 return -KEY_syscall;
9052 if (name[4] == 'p' &&
9056 return -KEY_sysopen;
9062 if (name[4] == 'e' &&
9066 return -KEY_sysread;
9072 if (name[4] == 'e' &&
9076 return -KEY_sysseek;
9094 if (name[1] == 'e' &&
9101 return -KEY_telldir;
9110 if (name[2] == 'f' &&
9116 return -KEY_ucfirst;
9122 if (name[2] == 's' &&
9128 return -KEY_unshift;
9138 if (name[1] == 'a' &&
9145 return -KEY_waitpid;
9154 case 8: /* 26 tokens of length 8 */
9158 if (name[1] == 'U' &&
9166 return KEY_AUTOLOAD;
9177 if (name[3] == 'A' &&
9183 return KEY___DATA__;
9189 if (name[3] == 'I' &&
9195 return -KEY___FILE__;
9201 if (name[3] == 'I' &&
9207 return -KEY___LINE__;
9223 if (name[2] == 'o' &&
9230 return -KEY_closedir;
9236 if (name[2] == 'n' &&
9243 return -KEY_continue;
9253 if (name[1] == 'b' &&
9261 return -KEY_dbmclose;
9267 if (name[1] == 'n' &&
9273 if (name[4] == 'r' &&
9278 return -KEY_endgrent;
9284 if (name[4] == 'w' &&
9289 return -KEY_endpwent;
9302 if (name[1] == 'o' &&
9310 return -KEY_formline;
9316 if (name[1] == 'e' &&
9327 if (name[6] == 'n' &&
9330 return -KEY_getgrent;
9336 if (name[6] == 'i' &&
9339 return -KEY_getgrgid;
9345 if (name[6] == 'a' &&
9348 return -KEY_getgrnam;
9361 if (name[4] == 'o' &&
9366 return -KEY_getlogin;
9377 if (name[6] == 'n' &&
9380 return -KEY_getpwent;
9386 if (name[6] == 'a' &&
9389 return -KEY_getpwnam;
9395 if (name[6] == 'i' &&
9398 return -KEY_getpwuid;
9418 if (name[1] == 'e' &&
9425 if (name[5] == 'i' &&
9432 return -KEY_readline;
9437 return -KEY_readlink;
9448 if (name[5] == 'i' &&
9452 return -KEY_readpipe;
9473 if (name[4] == 'r' &&
9478 return -KEY_setgrent;
9484 if (name[4] == 'w' &&
9489 return -KEY_setpwent;
9505 if (name[3] == 'w' &&
9511 return -KEY_shmwrite;
9517 if (name[3] == 't' &&
9523 return -KEY_shutdown;
9533 if (name[2] == 's' &&
9540 return -KEY_syswrite;
9550 if (name[1] == 'r' &&
9558 return -KEY_truncate;
9567 case 9: /* 8 tokens of length 9 */
9571 if (name[1] == 'n' &&
9580 return -KEY_endnetent;
9586 if (name[1] == 'e' &&
9595 return -KEY_getnetent;
9601 if (name[1] == 'o' &&
9610 return -KEY_localtime;
9616 if (name[1] == 'r' &&
9625 return KEY_prototype;
9631 if (name[1] == 'u' &&
9640 return -KEY_quotemeta;
9646 if (name[1] == 'e' &&
9655 return -KEY_rewinddir;
9661 if (name[1] == 'e' &&
9670 return -KEY_setnetent;
9676 if (name[1] == 'a' &&
9685 return -KEY_wantarray;
9694 case 10: /* 9 tokens of length 10 */
9698 if (name[1] == 'n' &&
9704 if (name[4] == 'o' &&
9711 return -KEY_endhostent;
9717 if (name[4] == 'e' &&
9724 return -KEY_endservent;
9737 if (name[1] == 'e' &&
9743 if (name[4] == 'o' &&
9750 return -KEY_gethostent;
9759 if (name[5] == 'r' &&
9765 return -KEY_getservent;
9771 if (name[5] == 'c' &&
9777 return -KEY_getsockopt;
9802 if (name[4] == 'o' &&
9809 return -KEY_sethostent;
9818 if (name[5] == 'r' &&
9824 return -KEY_setservent;
9830 if (name[5] == 'c' &&
9836 return -KEY_setsockopt;
9853 if (name[2] == 'c' &&
9862 return -KEY_socketpair;
9875 case 11: /* 8 tokens of length 11 */
9879 if (name[1] == '_' &&
9890 return -KEY___PACKAGE__;
9896 if (name[1] == 'n' &&
9907 return -KEY_endprotoent;
9913 if (name[1] == 'e' &&
9922 if (name[5] == 'e' &&
9929 return -KEY_getpeername;
9938 if (name[6] == 'o' &&
9944 return -KEY_getpriority;
9950 if (name[6] == 't' &&
9956 return -KEY_getprotoent;
9970 if (name[4] == 'o' &&
9978 return -KEY_getsockname;
9991 if (name[1] == 'e' &&
9999 if (name[6] == 'o' &&
10004 { /* setpriority */
10005 return -KEY_setpriority;
10011 if (name[6] == 't' &&
10016 { /* setprotoent */
10017 return -KEY_setprotoent;
10033 case 12: /* 2 tokens of length 12 */
10034 if (name[0] == 'g' &&
10046 if (name[9] == 'd' &&
10049 { /* getnetbyaddr */
10050 return -KEY_getnetbyaddr;
10056 if (name[9] == 'a' &&
10059 { /* getnetbyname */
10060 return -KEY_getnetbyname;
10072 case 13: /* 4 tokens of length 13 */
10073 if (name[0] == 'g' &&
10080 if (name[4] == 'o' &&
10089 if (name[10] == 'd' &&
10092 { /* gethostbyaddr */
10093 return -KEY_gethostbyaddr;
10099 if (name[10] == 'a' &&
10102 { /* gethostbyname */
10103 return -KEY_gethostbyname;
10116 if (name[4] == 'e' &&
10125 if (name[10] == 'a' &&
10128 { /* getservbyname */
10129 return -KEY_getservbyname;
10135 if (name[10] == 'o' &&
10138 { /* getservbyport */
10139 return -KEY_getservbyport;
10158 case 14: /* 1 tokens of length 14 */
10159 if (name[0] == 'g' &&
10173 { /* getprotobyname */
10174 return -KEY_getprotobyname;
10179 case 16: /* 1 tokens of length 16 */
10180 if (name[0] == 'g' &&
10196 { /* getprotobynumber */
10197 return -KEY_getprotobynumber;
10211 S_checkcomma(pTHX_ const char *s, const char *name, const char *what)
10215 if (*s == ' ' && s[1] == '(') { /* XXX gotta be a better way */
10216 if (ckWARN(WARN_SYNTAX)) {
10219 for (w = s+2; *w && level; w++) {
10222 else if (*w == ')')
10225 while (isSPACE(*w))
10227 if (!*w || !strchr(";|})]oaiuw!=", *w)) /* an advisory hack only... */
10228 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
10229 "%s (...) interpreted as function",name);
10232 while (s < PL_bufend && isSPACE(*s))
10236 while (s < PL_bufend && isSPACE(*s))
10238 if (isIDFIRST_lazy_if(s,UTF)) {
10239 const char * const w = s++;
10240 while (isALNUM_lazy_if(s,UTF))
10242 while (s < PL_bufend && isSPACE(*s))
10246 if (keyword(w, s - w))
10249 gv = gv_fetchpvn_flags(w, s - w, 0, SVt_PVCV);
10250 if (gv && GvCVu(gv))
10252 Perl_croak(aTHX_ "No comma allowed after %s", what);
10257 /* Either returns sv, or mortalizes sv and returns a new SV*.
10258 Best used as sv=new_constant(..., sv, ...).
10259 If s, pv are NULL, calls subroutine with one argument,
10260 and type is used with error messages only. */
10263 S_new_constant(pTHX_ const char *s, STRLEN len, const char *key, SV *sv, SV *pv,
10267 HV * const table = GvHV(PL_hintgv); /* ^H */
10271 const char *why1 = "", *why2 = "", *why3 = "";
10273 if (!table || !(PL_hints & HINT_LOCALIZE_HH)) {
10276 why2 = strEQ(key,"charnames")
10277 ? "(possibly a missing \"use charnames ...\")"
10279 msg = Perl_newSVpvf(aTHX_ "Constant(%s) unknown: %s",
10280 (type ? type: "undef"), why2);
10282 /* This is convoluted and evil ("goto considered harmful")
10283 * but I do not understand the intricacies of all the different
10284 * failure modes of %^H in here. The goal here is to make
10285 * the most probable error message user-friendly. --jhi */
10290 msg = Perl_newSVpvf(aTHX_ "Constant(%s): %s%s%s",
10291 (type ? type: "undef"), why1, why2, why3);
10293 yyerror(SvPVX_const(msg));
10297 cvp = hv_fetch(table, key, strlen(key), FALSE);
10298 if (!cvp || !SvOK(*cvp)) {
10301 why3 = "} is not defined";
10304 sv_2mortal(sv); /* Parent created it permanently */
10307 pv = sv_2mortal(newSVpvn(s, len));
10309 typesv = sv_2mortal(newSVpv(type, 0));
10311 typesv = &PL_sv_undef;
10313 PUSHSTACKi(PERLSI_OVERLOAD);
10325 call_sv(cv, G_SCALAR | ( PL_in_eval ? 0 : G_EVAL));
10329 /* Check the eval first */
10330 if (!PL_in_eval && SvTRUE(ERRSV)) {
10331 sv_catpvs(ERRSV, "Propagated");
10332 yyerror(SvPV_nolen_const(ERRSV)); /* Duplicates the message inside eval */
10334 res = SvREFCNT_inc_simple(sv);
10338 SvREFCNT_inc_simple_void(res);
10347 why1 = "Call to &{$^H{";
10349 why3 = "}} did not return a defined value";
10357 /* Returns a NUL terminated string, with the length of the string written to
10361 S_scan_word(pTHX_ register char *s, char *dest, STRLEN destlen, int allow_package, STRLEN *slp)
10364 register char *d = dest;
10365 register char * const e = d + destlen - 3; /* two-character token, ending NUL */
10368 Perl_croak(aTHX_ ident_too_long);
10369 if (isALNUM(*s)) /* UTF handled below */
10371 else if (*s == '\'' && allow_package && isIDFIRST_lazy_if(s+1,UTF)) {
10376 else if (*s == ':' && s[1] == ':' && allow_package && s[2] != '$') {
10380 else if (UTF && UTF8_IS_START(*s) && isALNUM_utf8((U8*)s)) {
10381 char *t = s + UTF8SKIP(s);
10382 while (UTF8_IS_CONTINUED(*t) && is_utf8_mark((U8*)t))
10384 if (d + (t - s) > e)
10385 Perl_croak(aTHX_ ident_too_long);
10386 Copy(s, d, t - s, char);
10399 S_scan_ident(pTHX_ register char *s, register const char *send, char *dest, STRLEN destlen, I32 ck_uni)
10402 char *bracket = NULL;
10404 register char *d = dest;
10405 register char * const e = d + destlen + 3; /* two-character token, ending NUL */
10410 while (isDIGIT(*s)) {
10412 Perl_croak(aTHX_ ident_too_long);
10419 Perl_croak(aTHX_ ident_too_long);
10420 if (isALNUM(*s)) /* UTF handled below */
10422 else if (*s == '\'' && isIDFIRST_lazy_if(s+1,UTF)) {
10427 else if (*s == ':' && s[1] == ':') {
10431 else if (UTF && UTF8_IS_START(*s) && isALNUM_utf8((U8*)s)) {
10432 char *t = s + UTF8SKIP(s);
10433 while (UTF8_IS_CONTINUED(*t) && is_utf8_mark((U8*)t))
10435 if (d + (t - s) > e)
10436 Perl_croak(aTHX_ ident_too_long);
10437 Copy(s, d, t - s, char);
10448 if (PL_lex_state != LEX_NORMAL)
10449 PL_lex_state = LEX_INTERPENDMAYBE;
10452 if (*s == '$' && s[1] &&
10453 (isALNUM_lazy_if(s+1,UTF) || s[1] == '$' || s[1] == '{' || strnEQ(s+1,"::",2)) )
10466 if (*d == '^' && *s && isCONTROLVAR(*s)) {
10471 if (isSPACE(s[-1])) {
10473 const char ch = *s++;
10474 if (!SPACE_OR_TAB(ch)) {
10480 if (isIDFIRST_lazy_if(d,UTF)) {
10484 while ((end < send && isALNUM_lazy_if(end,UTF)) || *end == ':') {
10485 end += UTF8SKIP(end);
10486 while (end < send && UTF8_IS_CONTINUED(*end) && is_utf8_mark((U8*)end))
10487 end += UTF8SKIP(end);
10489 Copy(s, d, end - s, char);
10494 while ((isALNUM(*s) || *s == ':') && d < e)
10497 Perl_croak(aTHX_ ident_too_long);
10500 while (s < send && SPACE_OR_TAB(*s)) s++;
10501 if ((*s == '[' || (*s == '{' && strNE(dest, "sub")))) {
10502 if (ckWARN(WARN_AMBIGUOUS) && keyword(dest, d - dest)) {
10503 const char *brack = *s == '[' ? "[...]" : "{...}";
10504 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
10505 "Ambiguous use of %c{%s%s} resolved to %c%s%s",
10506 funny, dest, brack, funny, dest, brack);
10509 PL_lex_brackstack[PL_lex_brackets++] = (char)(XOPERATOR | XFAKEBRACK);
10513 /* Handle extended ${^Foo} variables
10514 * 1999-02-27 mjd-perl-patch@plover.com */
10515 else if (!isALNUM(*d) && !isPRINT(*d) /* isCTRL(d) */
10519 while (isALNUM(*s) && d < e) {
10523 Perl_croak(aTHX_ ident_too_long);
10528 if (PL_lex_state == LEX_INTERPNORMAL && !PL_lex_brackets) {
10529 PL_lex_state = LEX_INTERPEND;
10534 if (PL_lex_state == LEX_NORMAL) {
10535 if (ckWARN(WARN_AMBIGUOUS) &&
10536 (keyword(dest, d - dest) || get_cv(dest, FALSE)))
10538 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
10539 "Ambiguous use of %c{%s} resolved to %c%s",
10540 funny, dest, funny, dest);
10545 s = bracket; /* let the parser handle it */
10549 else if (PL_lex_state == LEX_INTERPNORMAL && !PL_lex_brackets && !intuit_more(s))
10550 PL_lex_state = LEX_INTERPEND;
10555 Perl_pmflag(pTHX_ U32* pmfl, int ch)
10557 PERL_UNUSED_CONTEXT;
10560 else if (ch == 'g')
10561 *pmfl |= PMf_GLOBAL;
10562 else if (ch == 'c')
10563 *pmfl |= PMf_CONTINUE;
10564 else if (ch == 'o')
10566 else if (ch == 'm')
10567 *pmfl |= PMf_MULTILINE;
10568 else if (ch == 's')
10569 *pmfl |= PMf_SINGLELINE;
10570 else if (ch == 'x')
10571 *pmfl |= PMf_EXTENDED;
10575 S_scan_pat(pTHX_ char *start, I32 type)
10579 char *s = scan_str(start,!!PL_madskills,FALSE);
10580 const char * const valid_flags = (type == OP_QR) ? "iomsx" : "iogcmsx";
10587 const char * const delimiter = skipspace(start);
10588 Perl_croak(aTHX_ *delimiter == '?'
10589 ? "Search pattern not terminated or ternary operator parsed as search pattern"
10590 : "Search pattern not terminated" );
10593 pm = (PMOP*)newPMOP(type, 0);
10594 if (PL_multi_open == '?')
10595 pm->op_pmflags |= PMf_ONCE;
10599 while (*s && strchr(valid_flags, *s))
10600 pmflag(&pm->op_pmflags,*s++);
10602 if (PL_madskills && modstart != s) {
10603 SV* tmptoken = newSVpvn(modstart, s - modstart);
10604 append_madprops(newMADPROP('m', MAD_SV, tmptoken, 0), (OP*)pm, 0);
10607 /* issue a warning if /c is specified,but /g is not */
10608 if ((pm->op_pmflags & PMf_CONTINUE) && !(pm->op_pmflags & PMf_GLOBAL)
10609 && ckWARN(WARN_REGEXP))
10611 Perl_warner(aTHX_ packWARN(WARN_REGEXP), "Use of /c modifier is meaningless without /g" );
10614 pm->op_pmpermflags = pm->op_pmflags;
10616 PL_lex_op = (OP*)pm;
10617 yylval.ival = OP_MATCH;
10622 S_scan_subst(pTHX_ char *start)
10633 yylval.ival = OP_NULL;
10635 s = scan_str(start,!!PL_madskills,FALSE);
10638 Perl_croak(aTHX_ "Substitution pattern not terminated");
10640 if (s[-1] == PL_multi_open)
10643 if (PL_madskills) {
10644 CURMAD('q', PL_thisopen);
10645 CURMAD('_', PL_thiswhite);
10646 CURMAD('E', PL_thisstuff);
10647 CURMAD('Q', PL_thisclose);
10648 PL_realtokenstart = s - SvPVX(PL_linestr);
10652 first_start = PL_multi_start;
10653 s = scan_str(s,!!PL_madskills,FALSE);
10655 if (PL_lex_stuff) {
10656 SvREFCNT_dec(PL_lex_stuff);
10657 PL_lex_stuff = NULL;
10659 Perl_croak(aTHX_ "Substitution replacement not terminated");
10661 PL_multi_start = first_start; /* so whole substitution is taken together */
10663 pm = (PMOP*)newPMOP(OP_SUBST, 0);
10666 if (PL_madskills) {
10667 CURMAD('z', PL_thisopen);
10668 CURMAD('R', PL_thisstuff);
10669 CURMAD('Z', PL_thisclose);
10679 else if (strchr("iogcmsx", *s))
10680 pmflag(&pm->op_pmflags,*s++);
10686 if (PL_madskills) {
10688 curmad('m', newSVpvn(modstart, s - modstart));
10689 append_madprops(PL_thismad, (OP*)pm, 0);
10693 if ((pm->op_pmflags & PMf_CONTINUE) && ckWARN(WARN_REGEXP)) {
10694 Perl_warner(aTHX_ packWARN(WARN_REGEXP), "Use of /c modifier is meaningless in s///" );
10698 SV * const repl = newSVpvs("");
10700 PL_sublex_info.super_bufptr = s;
10701 PL_sublex_info.super_bufend = PL_bufend;
10703 pm->op_pmflags |= PMf_EVAL;
10705 sv_catpv(repl, es ? "eval " : "do ");
10706 sv_catpvs(repl, "{");
10707 sv_catsv(repl, PL_lex_repl);
10708 sv_catpvs(repl, "}");
10710 SvREFCNT_dec(PL_lex_repl);
10711 PL_lex_repl = repl;
10714 pm->op_pmpermflags = pm->op_pmflags;
10715 PL_lex_op = (OP*)pm;
10716 yylval.ival = OP_SUBST;
10721 S_scan_trans(pTHX_ char *start)
10734 yylval.ival = OP_NULL;
10736 s = scan_str(start,!!PL_madskills,FALSE);
10738 Perl_croak(aTHX_ "Transliteration pattern not terminated");
10740 if (s[-1] == PL_multi_open)
10743 if (PL_madskills) {
10744 CURMAD('q', PL_thisopen);
10745 CURMAD('_', PL_thiswhite);
10746 CURMAD('E', PL_thisstuff);
10747 CURMAD('Q', PL_thisclose);
10748 PL_realtokenstart = s - SvPVX(PL_linestr);
10752 s = scan_str(s,!!PL_madskills,FALSE);
10754 if (PL_lex_stuff) {
10755 SvREFCNT_dec(PL_lex_stuff);
10756 PL_lex_stuff = NULL;
10758 Perl_croak(aTHX_ "Transliteration replacement not terminated");
10760 if (PL_madskills) {
10761 CURMAD('z', PL_thisopen);
10762 CURMAD('R', PL_thisstuff);
10763 CURMAD('Z', PL_thisclose);
10766 complement = del = squash = 0;
10773 complement = OPpTRANS_COMPLEMENT;
10776 del = OPpTRANS_DELETE;
10779 squash = OPpTRANS_SQUASH;
10788 Newx(tbl, complement&&!del?258:256, short);
10789 o = newPVOP(OP_TRANS, 0, (char*)tbl);
10790 o->op_private &= ~OPpTRANS_ALL;
10791 o->op_private |= del|squash|complement|
10792 (DO_UTF8(PL_lex_stuff)? OPpTRANS_FROM_UTF : 0)|
10793 (DO_UTF8(PL_lex_repl) ? OPpTRANS_TO_UTF : 0);
10796 yylval.ival = OP_TRANS;
10799 if (PL_madskills) {
10801 curmad('m', newSVpvn(modstart, s - modstart));
10802 append_madprops(PL_thismad, o, 0);
10811 S_scan_heredoc(pTHX_ register char *s)
10815 I32 op_type = OP_SCALAR;
10819 const char *found_newline;
10823 const int outer = (PL_rsfp && !(PL_lex_inwhat == OP_SCALAR));
10825 I32 stuffstart = s - SvPVX(PL_linestr);
10828 PL_realtokenstart = -1;
10833 e = PL_tokenbuf + sizeof PL_tokenbuf - 1;
10836 for (peek = s; SPACE_OR_TAB(*peek); peek++) ;
10837 if (*peek == '`' || *peek == '\'' || *peek =='"') {
10840 s = delimcpy(d, e, s, PL_bufend, term, &len);
10850 if (!isALNUM_lazy_if(s,UTF))
10851 deprecate_old("bare << to mean <<\"\"");
10852 for (; isALNUM_lazy_if(s,UTF); s++) {
10857 if (d >= PL_tokenbuf + sizeof PL_tokenbuf - 1)
10858 Perl_croak(aTHX_ "Delimiter for here document is too long");
10861 len = d - PL_tokenbuf;
10864 if (PL_madskills) {
10865 tstart = PL_tokenbuf + !outer;
10866 PL_thisclose = newSVpvn(tstart, len - !outer);
10867 tstart = SvPVX(PL_linestr) + stuffstart;
10868 PL_thisopen = newSVpvn(tstart, s - tstart);
10869 stuffstart = s - SvPVX(PL_linestr);
10872 #ifndef PERL_STRICT_CR
10873 d = strchr(s, '\r');
10875 char * const olds = s;
10877 while (s < PL_bufend) {
10883 else if (*s == '\n' && s[1] == '\r') { /* \015\013 on a mac? */
10892 SvCUR_set(PL_linestr, PL_bufend - SvPVX_const(PL_linestr));
10899 if ( outer || !(found_newline = memchr(s, '\n', PL_bufend - s)) ) {
10900 herewas = newSVpvn(s,PL_bufend-s);
10904 herewas = newSVpvn(s-1,found_newline-s+1);
10907 herewas = newSVpvn(s,found_newline-s);
10911 if (PL_madskills) {
10912 tstart = SvPVX(PL_linestr) + stuffstart;
10914 sv_catpvn(PL_thisstuff, tstart, s - tstart);
10916 PL_thisstuff = newSVpvn(tstart, s - tstart);
10919 s += SvCUR(herewas);
10922 stuffstart = s - SvPVX(PL_linestr);
10928 tmpstr = newSV(79);
10929 sv_upgrade(tmpstr, SVt_PVIV);
10930 if (term == '\'') {
10931 op_type = OP_CONST;
10932 SvIV_set(tmpstr, -1);
10934 else if (term == '`') {
10935 op_type = OP_BACKTICK;
10936 SvIV_set(tmpstr, '\\');
10940 PL_multi_start = CopLINE(PL_curcop);
10941 PL_multi_open = PL_multi_close = '<';
10942 term = *PL_tokenbuf;
10943 if (PL_lex_inwhat == OP_SUBST && PL_in_eval && !PL_rsfp) {
10944 char * const bufptr = PL_sublex_info.super_bufptr;
10945 char * const bufend = PL_sublex_info.super_bufend;
10946 char * const olds = s - SvCUR(herewas);
10947 s = strchr(bufptr, '\n');
10951 while (s < bufend &&
10952 (*s != term || memNE(s,PL_tokenbuf,len)) ) {
10954 CopLINE_inc(PL_curcop);
10957 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
10958 missingterm(PL_tokenbuf);
10960 sv_setpvn(herewas,bufptr,d-bufptr+1);
10961 sv_setpvn(tmpstr,d+1,s-d);
10963 sv_catpvn(herewas,s,bufend-s);
10964 Copy(SvPVX_const(herewas),bufptr,SvCUR(herewas) + 1,char);
10971 while (s < PL_bufend &&
10972 (*s != term || memNE(s,PL_tokenbuf,len)) ) {
10974 CopLINE_inc(PL_curcop);
10976 if (s >= PL_bufend) {
10977 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
10978 missingterm(PL_tokenbuf);
10980 sv_setpvn(tmpstr,d+1,s-d);
10982 if (PL_madskills) {
10984 sv_catpvn(PL_thisstuff, d + 1, s - d);
10986 PL_thisstuff = newSVpvn(d + 1, s - d);
10987 stuffstart = s - SvPVX(PL_linestr);
10991 CopLINE_inc(PL_curcop); /* the preceding stmt passes a newline */
10993 sv_catpvn(herewas,s,PL_bufend-s);
10994 sv_setsv(PL_linestr,herewas);
10995 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = s = PL_linestart = SvPVX(PL_linestr);
10996 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
10997 PL_last_lop = PL_last_uni = NULL;
11000 sv_setpvn(tmpstr,"",0); /* avoid "uninitialized" warning */
11001 while (s >= PL_bufend) { /* multiple line string? */
11003 if (PL_madskills) {
11004 tstart = SvPVX(PL_linestr) + stuffstart;
11006 sv_catpvn(PL_thisstuff, tstart, PL_bufend - tstart);
11008 PL_thisstuff = newSVpvn(tstart, PL_bufend - tstart);
11012 !(PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = filter_gets(PL_linestr, PL_rsfp, 0))) {
11013 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
11014 missingterm(PL_tokenbuf);
11017 stuffstart = s - SvPVX(PL_linestr);
11019 CopLINE_inc(PL_curcop);
11020 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
11021 PL_last_lop = PL_last_uni = NULL;
11022 #ifndef PERL_STRICT_CR
11023 if (PL_bufend - PL_linestart >= 2) {
11024 if ((PL_bufend[-2] == '\r' && PL_bufend[-1] == '\n') ||
11025 (PL_bufend[-2] == '\n' && PL_bufend[-1] == '\r'))
11027 PL_bufend[-2] = '\n';
11029 SvCUR_set(PL_linestr, PL_bufend - SvPVX_const(PL_linestr));
11031 else if (PL_bufend[-1] == '\r')
11032 PL_bufend[-1] = '\n';
11034 else if (PL_bufend - PL_linestart == 1 && PL_bufend[-1] == '\r')
11035 PL_bufend[-1] = '\n';
11037 if (PERLDB_LINE && PL_curstash != PL_debstash) {
11038 SV * const sv = newSV(0);
11040 sv_upgrade(sv, SVt_PVMG);
11041 sv_setsv(sv,PL_linestr);
11042 (void)SvIOK_on(sv);
11044 av_store(CopFILEAVx(PL_curcop), (I32)CopLINE(PL_curcop),sv);
11046 if (*s == term && memEQ(s,PL_tokenbuf,len)) {
11047 STRLEN off = PL_bufend - 1 - SvPVX_const(PL_linestr);
11048 *(SvPVX(PL_linestr) + off ) = ' ';
11049 sv_catsv(PL_linestr,herewas);
11050 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
11051 s = SvPVX(PL_linestr) + off; /* In case PV of PL_linestr moved. */
11055 sv_catsv(tmpstr,PL_linestr);
11060 PL_multi_end = CopLINE(PL_curcop);
11061 if (SvCUR(tmpstr) + 5 < SvLEN(tmpstr)) {
11062 SvPV_shrink_to_cur(tmpstr);
11064 SvREFCNT_dec(herewas);
11066 if (UTF && is_utf8_string((U8*)SvPVX_const(tmpstr), SvCUR(tmpstr)))
11068 else if (PL_encoding)
11069 sv_recode_to_utf8(tmpstr, PL_encoding);
11071 PL_lex_stuff = tmpstr;
11072 yylval.ival = op_type;
11076 /* scan_inputsymbol
11077 takes: current position in input buffer
11078 returns: new position in input buffer
11079 side-effects: yylval and lex_op are set.
11084 <FH> read from filehandle
11085 <pkg::FH> read from package qualified filehandle
11086 <pkg'FH> read from package qualified filehandle
11087 <$fh> read from filehandle in $fh
11088 <*.h> filename glob
11093 S_scan_inputsymbol(pTHX_ char *start)
11096 register char *s = start; /* current position in buffer */
11100 char *d = PL_tokenbuf; /* start of temp holding space */
11101 const char * const e = PL_tokenbuf + sizeof PL_tokenbuf; /* end of temp holding space */
11103 end = strchr(s, '\n');
11106 s = delimcpy(d, e, s + 1, end, '>', &len); /* extract until > */
11108 /* die if we didn't have space for the contents of the <>,
11109 or if it didn't end, or if we see a newline
11112 if (len >= (I32)sizeof PL_tokenbuf)
11113 Perl_croak(aTHX_ "Excessively long <> operator");
11115 Perl_croak(aTHX_ "Unterminated <> operator");
11120 Remember, only scalar variables are interpreted as filehandles by
11121 this code. Anything more complex (e.g., <$fh{$num}>) will be
11122 treated as a glob() call.
11123 This code makes use of the fact that except for the $ at the front,
11124 a scalar variable and a filehandle look the same.
11126 if (*d == '$' && d[1]) d++;
11128 /* allow <Pkg'VALUE> or <Pkg::VALUE> */
11129 while (*d && (isALNUM_lazy_if(d,UTF) || *d == '\'' || *d == ':'))
11132 /* If we've tried to read what we allow filehandles to look like, and
11133 there's still text left, then it must be a glob() and not a getline.
11134 Use scan_str to pull out the stuff between the <> and treat it
11135 as nothing more than a string.
11138 if (d - PL_tokenbuf != len) {
11139 yylval.ival = OP_GLOB;
11141 s = scan_str(start,!!PL_madskills,FALSE);
11143 Perl_croak(aTHX_ "Glob not terminated");
11147 bool readline_overriden = FALSE;
11150 /* we're in a filehandle read situation */
11153 /* turn <> into <ARGV> */
11155 Copy("ARGV",d,5,char);
11157 /* Check whether readline() is overriden */
11158 gv_readline = gv_fetchpvs("readline", GV_NOTQUAL, SVt_PVCV);
11160 && GvCVu(gv_readline) && GvIMPORTED_CV(gv_readline))
11162 ((gvp = (GV**)hv_fetchs(PL_globalstash, "readline", FALSE))
11163 && (gv_readline = *gvp) != (GV*)&PL_sv_undef
11164 && GvCVu(gv_readline) && GvIMPORTED_CV(gv_readline)))
11165 readline_overriden = TRUE;
11167 /* if <$fh>, create the ops to turn the variable into a
11173 /* try to find it in the pad for this block, otherwise find
11174 add symbol table ops
11176 if ((tmp = pad_findmy(d)) != NOT_IN_PAD) {
11177 if (PAD_COMPNAME_FLAGS_isOUR(tmp)) {
11178 HV * const stash = PAD_COMPNAME_OURSTASH(tmp);
11179 HEK * const stashname = HvNAME_HEK(stash);
11180 SV * const sym = sv_2mortal(newSVhek(stashname));
11181 sv_catpvs(sym, "::");
11182 sv_catpv(sym, d+1);
11187 OP * const o = newOP(OP_PADSV, 0);
11189 PL_lex_op = readline_overriden
11190 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
11191 append_elem(OP_LIST, o,
11192 newCVREF(0, newGVOP(OP_GV,0,gv_readline))))
11193 : (OP*)newUNOP(OP_READLINE, 0, o);
11202 ? (GV_ADDMULTI | GV_ADDINEVAL)
11205 PL_lex_op = readline_overriden
11206 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
11207 append_elem(OP_LIST,
11208 newUNOP(OP_RV2SV, 0, newGVOP(OP_GV, 0, gv)),
11209 newCVREF(0, newGVOP(OP_GV, 0, gv_readline))))
11210 : (OP*)newUNOP(OP_READLINE, 0,
11211 newUNOP(OP_RV2SV, 0,
11212 newGVOP(OP_GV, 0, gv)));
11214 if (!readline_overriden)
11215 PL_lex_op->op_flags |= OPf_SPECIAL;
11216 /* we created the ops in PL_lex_op, so make yylval.ival a null op */
11217 yylval.ival = OP_NULL;
11220 /* If it's none of the above, it must be a literal filehandle
11221 (<Foo::BAR> or <FOO>) so build a simple readline OP */
11223 GV * const gv = gv_fetchpv(d, GV_ADD, SVt_PVIO);
11224 PL_lex_op = readline_overriden
11225 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
11226 append_elem(OP_LIST,
11227 newGVOP(OP_GV, 0, gv),
11228 newCVREF(0, newGVOP(OP_GV, 0, gv_readline))))
11229 : (OP*)newUNOP(OP_READLINE, 0, newGVOP(OP_GV, 0, gv));
11230 yylval.ival = OP_NULL;
11239 takes: start position in buffer
11240 keep_quoted preserve \ on the embedded delimiter(s)
11241 keep_delims preserve the delimiters around the string
11242 returns: position to continue reading from buffer
11243 side-effects: multi_start, multi_close, lex_repl or lex_stuff, and
11244 updates the read buffer.
11246 This subroutine pulls a string out of the input. It is called for:
11247 q single quotes q(literal text)
11248 ' single quotes 'literal text'
11249 qq double quotes qq(interpolate $here please)
11250 " double quotes "interpolate $here please"
11251 qx backticks qx(/bin/ls -l)
11252 ` backticks `/bin/ls -l`
11253 qw quote words @EXPORT_OK = qw( func() $spam )
11254 m// regexp match m/this/
11255 s/// regexp substitute s/this/that/
11256 tr/// string transliterate tr/this/that/
11257 y/// string transliterate y/this/that/
11258 ($*@) sub prototypes sub foo ($)
11259 (stuff) sub attr parameters sub foo : attr(stuff)
11260 <> readline or globs <FOO>, <>, <$fh>, or <*.c>
11262 In most of these cases (all but <>, patterns and transliterate)
11263 yylex() calls scan_str(). m// makes yylex() call scan_pat() which
11264 calls scan_str(). s/// makes yylex() call scan_subst() which calls
11265 scan_str(). tr/// and y/// make yylex() call scan_trans() which
11268 It skips whitespace before the string starts, and treats the first
11269 character as the delimiter. If the delimiter is one of ([{< then
11270 the corresponding "close" character )]}> is used as the closing
11271 delimiter. It allows quoting of delimiters, and if the string has
11272 balanced delimiters ([{<>}]) it allows nesting.
11274 On success, the SV with the resulting string is put into lex_stuff or,
11275 if that is already non-NULL, into lex_repl. The second case occurs only
11276 when parsing the RHS of the special constructs s/// and tr/// (y///).
11277 For convenience, the terminating delimiter character is stuffed into
11282 S_scan_str(pTHX_ char *start, int keep_quoted, int keep_delims)
11285 SV *sv; /* scalar value: string */
11286 char *tmps; /* temp string, used for delimiter matching */
11287 register char *s = start; /* current position in the buffer */
11288 register char term; /* terminating character */
11289 register char *to; /* current position in the sv's data */
11290 I32 brackets = 1; /* bracket nesting level */
11291 bool has_utf8 = FALSE; /* is there any utf8 content? */
11292 I32 termcode; /* terminating char. code */
11293 U8 termstr[UTF8_MAXBYTES]; /* terminating string */
11294 STRLEN termlen; /* length of terminating string */
11295 char *last = NULL; /* last position for nesting bracket */
11301 /* skip space before the delimiter */
11307 if (PL_realtokenstart >= 0) {
11308 stuffstart = PL_realtokenstart;
11309 PL_realtokenstart = -1;
11312 stuffstart = start - SvPVX(PL_linestr);
11314 /* mark where we are, in case we need to report errors */
11317 /* after skipping whitespace, the next character is the terminator */
11320 termcode = termstr[0] = term;
11324 termcode = utf8_to_uvchr((U8*)s, &termlen);
11325 Copy(s, termstr, termlen, U8);
11326 if (!UTF8_IS_INVARIANT(term))
11330 /* mark where we are */
11331 PL_multi_start = CopLINE(PL_curcop);
11332 PL_multi_open = term;
11334 /* find corresponding closing delimiter */
11335 if (term && (tmps = strchr("([{< )]}> )]}>",term)))
11336 termcode = termstr[0] = term = tmps[5];
11338 PL_multi_close = term;
11340 /* create a new SV to hold the contents. 79 is the SV's initial length.
11341 What a random number. */
11343 sv_upgrade(sv, SVt_PVIV);
11344 SvIV_set(sv, termcode);
11345 (void)SvPOK_only(sv); /* validate pointer */
11347 /* move past delimiter and try to read a complete string */
11349 sv_catpvn(sv, s, termlen);
11352 tstart = SvPVX(PL_linestr) + stuffstart;
11353 if (!PL_thisopen && !keep_delims) {
11354 PL_thisopen = newSVpvn(tstart, s - tstart);
11355 stuffstart = s - SvPVX(PL_linestr);
11359 if (PL_encoding && !UTF) {
11363 int offset = s - SvPVX_const(PL_linestr);
11364 const bool found = sv_cat_decode(sv, PL_encoding, PL_linestr,
11365 &offset, (char*)termstr, termlen);
11366 const char * const ns = SvPVX_const(PL_linestr) + offset;
11367 char * const svlast = SvEND(sv) - 1;
11369 for (; s < ns; s++) {
11370 if (*s == '\n' && !PL_rsfp)
11371 CopLINE_inc(PL_curcop);
11374 goto read_more_line;
11376 /* handle quoted delimiters */
11377 if (SvCUR(sv) > 1 && *(svlast-1) == '\\') {
11379 for (t = svlast-2; t >= SvPVX_const(sv) && *t == '\\';)
11381 if ((svlast-1 - t) % 2) {
11382 if (!keep_quoted) {
11383 *(svlast-1) = term;
11385 SvCUR_set(sv, SvCUR(sv) - 1);
11390 if (PL_multi_open == PL_multi_close) {
11398 for (t = w = last; t < svlast; w++, t++) {
11399 /* At here, all closes are "was quoted" one,
11400 so we don't check PL_multi_close. */
11402 if (!keep_quoted && *(t+1) == PL_multi_open)
11407 else if (*t == PL_multi_open)
11415 SvCUR_set(sv, w - SvPVX_const(sv));
11418 if (--brackets <= 0)
11423 if (!keep_delims) {
11424 SvCUR_set(sv, SvCUR(sv) - 1);
11430 /* extend sv if need be */
11431 SvGROW(sv, SvCUR(sv) + (PL_bufend - s) + 1);
11432 /* set 'to' to the next character in the sv's string */
11433 to = SvPVX(sv)+SvCUR(sv);
11435 /* if open delimiter is the close delimiter read unbridle */
11436 if (PL_multi_open == PL_multi_close) {
11437 for (; s < PL_bufend; s++,to++) {
11438 /* embedded newlines increment the current line number */
11439 if (*s == '\n' && !PL_rsfp)
11440 CopLINE_inc(PL_curcop);
11441 /* handle quoted delimiters */
11442 if (*s == '\\' && s+1 < PL_bufend && term != '\\') {
11443 if (!keep_quoted && s[1] == term)
11445 /* any other quotes are simply copied straight through */
11449 /* terminate when run out of buffer (the for() condition), or
11450 have found the terminator */
11451 else if (*s == term) {
11454 if (s+termlen <= PL_bufend && memEQ(s, (char*)termstr, termlen))
11457 else if (!has_utf8 && !UTF8_IS_INVARIANT((U8)*s) && UTF)
11463 /* if the terminator isn't the same as the start character (e.g.,
11464 matched brackets), we have to allow more in the quoting, and
11465 be prepared for nested brackets.
11468 /* read until we run out of string, or we find the terminator */
11469 for (; s < PL_bufend; s++,to++) {
11470 /* embedded newlines increment the line count */
11471 if (*s == '\n' && !PL_rsfp)
11472 CopLINE_inc(PL_curcop);
11473 /* backslashes can escape the open or closing characters */
11474 if (*s == '\\' && s+1 < PL_bufend) {
11475 if (!keep_quoted &&
11476 ((s[1] == PL_multi_open) || (s[1] == PL_multi_close)))
11481 /* allow nested opens and closes */
11482 else if (*s == PL_multi_close && --brackets <= 0)
11484 else if (*s == PL_multi_open)
11486 else if (!has_utf8 && !UTF8_IS_INVARIANT((U8)*s) && UTF)
11491 /* terminate the copied string and update the sv's end-of-string */
11493 SvCUR_set(sv, to - SvPVX_const(sv));
11496 * this next chunk reads more into the buffer if we're not done yet
11500 break; /* handle case where we are done yet :-) */
11502 #ifndef PERL_STRICT_CR
11503 if (to - SvPVX_const(sv) >= 2) {
11504 if ((to[-2] == '\r' && to[-1] == '\n') ||
11505 (to[-2] == '\n' && to[-1] == '\r'))
11509 SvCUR_set(sv, to - SvPVX_const(sv));
11511 else if (to[-1] == '\r')
11514 else if (to - SvPVX_const(sv) == 1 && to[-1] == '\r')
11519 /* if we're out of file, or a read fails, bail and reset the current
11520 line marker so we can report where the unterminated string began
11523 if (PL_madskills) {
11524 char *tstart = SvPVX(PL_linestr) + stuffstart;
11526 sv_catpvn(PL_thisstuff, tstart, PL_bufend - tstart);
11528 PL_thisstuff = newSVpvn(tstart, PL_bufend - tstart);
11532 !(PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = filter_gets(PL_linestr, PL_rsfp, 0))) {
11534 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
11540 /* we read a line, so increment our line counter */
11541 CopLINE_inc(PL_curcop);
11543 /* update debugger info */
11544 if (PERLDB_LINE && PL_curstash != PL_debstash) {
11545 SV * const line_sv = newSV(0);
11547 sv_upgrade(line_sv, SVt_PVMG);
11548 sv_setsv(line_sv,PL_linestr);
11549 (void)SvIOK_on(line_sv);
11550 SvIV_set(line_sv, 0);
11551 av_store(CopFILEAVx(PL_curcop), (I32)CopLINE(PL_curcop), line_sv);
11554 /* having changed the buffer, we must update PL_bufend */
11555 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
11556 PL_last_lop = PL_last_uni = NULL;
11559 /* at this point, we have successfully read the delimited string */
11561 if (!PL_encoding || UTF) {
11563 if (PL_madskills) {
11564 char *tstart = SvPVX(PL_linestr) + stuffstart;
11566 sv_catpvn(PL_thisstuff, tstart, s - tstart);
11568 PL_thisstuff = newSVpvn(tstart, s - tstart);
11569 if (!PL_thisclose && !keep_delims)
11570 PL_thisclose = newSVpvn(s,termlen);
11575 sv_catpvn(sv, s, termlen);
11580 if (PL_madskills) {
11581 char *tstart = SvPVX(PL_linestr) + stuffstart;
11583 sv_catpvn(PL_thisstuff, tstart, s - tstart - termlen);
11585 PL_thisstuff = newSVpvn(tstart, s - tstart - termlen);
11586 if (!PL_thisclose && !keep_delims)
11587 PL_thisclose = newSVpvn(s - termlen,termlen);
11591 if (has_utf8 || PL_encoding)
11594 PL_multi_end = CopLINE(PL_curcop);
11596 /* if we allocated too much space, give some back */
11597 if (SvCUR(sv) + 5 < SvLEN(sv)) {
11598 SvLEN_set(sv, SvCUR(sv) + 1);
11599 SvPV_renew(sv, SvLEN(sv));
11602 /* decide whether this is the first or second quoted string we've read
11615 takes: pointer to position in buffer
11616 returns: pointer to new position in buffer
11617 side-effects: builds ops for the constant in yylval.op
11619 Read a number in any of the formats that Perl accepts:
11621 \d(_?\d)*(\.(\d(_?\d)*)?)?[Ee][\+\-]?(\d(_?\d)*) 12 12.34 12.
11622 \.\d(_?\d)*[Ee][\+\-]?(\d(_?\d)*) .34
11625 0x[0-9A-Fa-f](_?[0-9A-Fa-f])*
11627 Like most scan_ routines, it uses the PL_tokenbuf buffer to hold the
11630 If it reads a number without a decimal point or an exponent, it will
11631 try converting the number to an integer and see if it can do so
11632 without loss of precision.
11636 Perl_scan_num(pTHX_ const char *start, YYSTYPE* lvalp)
11639 register const char *s = start; /* current position in buffer */
11640 register char *d; /* destination in temp buffer */
11641 register char *e; /* end of temp buffer */
11642 NV nv; /* number read, as a double */
11643 SV *sv = NULL; /* place to put the converted number */
11644 bool floatit; /* boolean: int or float? */
11645 const char *lastub = NULL; /* position of last underbar */
11646 static char const number_too_long[] = "Number too long";
11648 /* We use the first character to decide what type of number this is */
11652 Perl_croak(aTHX_ "panic: scan_num");
11654 /* if it starts with a 0, it could be an octal number, a decimal in
11655 0.13 disguise, or a hexadecimal number, or a binary number. */
11659 u holds the "number so far"
11660 shift the power of 2 of the base
11661 (hex == 4, octal == 3, binary == 1)
11662 overflowed was the number more than we can hold?
11664 Shift is used when we add a digit. It also serves as an "are
11665 we in octal/hex/binary?" indicator to disallow hex characters
11666 when in octal mode.
11671 bool overflowed = FALSE;
11672 bool just_zero = TRUE; /* just plain 0 or binary number? */
11673 static const NV nvshift[5] = { 1.0, 2.0, 4.0, 8.0, 16.0 };
11674 static const char* const bases[5] =
11675 { "", "binary", "", "octal", "hexadecimal" };
11676 static const char* const Bases[5] =
11677 { "", "Binary", "", "Octal", "Hexadecimal" };
11678 static const char* const maxima[5] =
11680 "0b11111111111111111111111111111111",
11684 const char *base, *Base, *max;
11686 /* check for hex */
11691 } else if (s[1] == 'b') {
11696 /* check for a decimal in disguise */
11697 else if (s[1] == '.' || s[1] == 'e' || s[1] == 'E')
11699 /* so it must be octal */
11706 if (ckWARN(WARN_SYNTAX))
11707 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
11708 "Misplaced _ in number");
11712 base = bases[shift];
11713 Base = Bases[shift];
11714 max = maxima[shift];
11716 /* read the rest of the number */
11718 /* x is used in the overflow test,
11719 b is the digit we're adding on. */
11724 /* if we don't mention it, we're done */
11728 /* _ are ignored -- but warned about if consecutive */
11730 if (lastub && s == lastub + 1 && ckWARN(WARN_SYNTAX))
11731 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
11732 "Misplaced _ in number");
11736 /* 8 and 9 are not octal */
11737 case '8': case '9':
11739 yyerror(Perl_form(aTHX_ "Illegal octal digit '%c'", *s));
11743 case '2': case '3': case '4':
11744 case '5': case '6': case '7':
11746 yyerror(Perl_form(aTHX_ "Illegal binary digit '%c'", *s));
11749 case '0': case '1':
11750 b = *s++ & 15; /* ASCII digit -> value of digit */
11754 case 'a': case 'b': case 'c': case 'd': case 'e': case 'f':
11755 case 'A': case 'B': case 'C': case 'D': case 'E': case 'F':
11756 /* make sure they said 0x */
11759 b = (*s++ & 7) + 9;
11761 /* Prepare to put the digit we have onto the end
11762 of the number so far. We check for overflows.
11768 x = u << shift; /* make room for the digit */
11770 if ((x >> shift) != u
11771 && !(PL_hints & HINT_NEW_BINARY)) {
11774 if (ckWARN_d(WARN_OVERFLOW))
11775 Perl_warner(aTHX_ packWARN(WARN_OVERFLOW),
11776 "Integer overflow in %s number",
11779 u = x | b; /* add the digit to the end */
11782 n *= nvshift[shift];
11783 /* If an NV has not enough bits in its
11784 * mantissa to represent an UV this summing of
11785 * small low-order numbers is a waste of time
11786 * (because the NV cannot preserve the
11787 * low-order bits anyway): we could just
11788 * remember when did we overflow and in the
11789 * end just multiply n by the right
11797 /* if we get here, we had success: make a scalar value from
11802 /* final misplaced underbar check */
11803 if (s[-1] == '_') {
11804 if (ckWARN(WARN_SYNTAX))
11805 Perl_warner(aTHX_ packWARN(WARN_SYNTAX), "Misplaced _ in number");
11810 if (n > 4294967295.0 && ckWARN(WARN_PORTABLE))
11811 Perl_warner(aTHX_ packWARN(WARN_PORTABLE),
11812 "%s number > %s non-portable",
11818 if (u > 0xffffffff && ckWARN(WARN_PORTABLE))
11819 Perl_warner(aTHX_ packWARN(WARN_PORTABLE),
11820 "%s number > %s non-portable",
11825 if (just_zero && (PL_hints & HINT_NEW_INTEGER))
11826 sv = new_constant(start, s - start, "integer",
11828 else if (PL_hints & HINT_NEW_BINARY)
11829 sv = new_constant(start, s - start, "binary", sv, NULL, NULL);
11834 handle decimal numbers.
11835 we're also sent here when we read a 0 as the first digit
11837 case '1': case '2': case '3': case '4': case '5':
11838 case '6': case '7': case '8': case '9': case '.':
11841 e = PL_tokenbuf + sizeof PL_tokenbuf - 6; /* room for various punctuation */
11844 /* read next group of digits and _ and copy into d */
11845 while (isDIGIT(*s) || *s == '_') {
11846 /* skip underscores, checking for misplaced ones
11850 if (lastub && s == lastub + 1 && ckWARN(WARN_SYNTAX))
11851 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
11852 "Misplaced _ in number");
11856 /* check for end of fixed-length buffer */
11858 Perl_croak(aTHX_ number_too_long);
11859 /* if we're ok, copy the character */
11864 /* final misplaced underbar check */
11865 if (lastub && s == lastub + 1) {
11866 if (ckWARN(WARN_SYNTAX))
11867 Perl_warner(aTHX_ packWARN(WARN_SYNTAX), "Misplaced _ in number");
11870 /* read a decimal portion if there is one. avoid
11871 3..5 being interpreted as the number 3. followed
11874 if (*s == '.' && s[1] != '.') {
11879 if (ckWARN(WARN_SYNTAX))
11880 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
11881 "Misplaced _ in number");
11885 /* copy, ignoring underbars, until we run out of digits.
11887 for (; isDIGIT(*s) || *s == '_'; s++) {
11888 /* fixed length buffer check */
11890 Perl_croak(aTHX_ number_too_long);
11892 if (lastub && s == lastub + 1 && ckWARN(WARN_SYNTAX))
11893 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
11894 "Misplaced _ in number");
11900 /* fractional part ending in underbar? */
11901 if (s[-1] == '_') {
11902 if (ckWARN(WARN_SYNTAX))
11903 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
11904 "Misplaced _ in number");
11906 if (*s == '.' && isDIGIT(s[1])) {
11907 /* oops, it's really a v-string, but without the "v" */
11913 /* read exponent part, if present */
11914 if ((*s == 'e' || *s == 'E') && strchr("+-0123456789_", s[1])) {
11918 /* regardless of whether user said 3E5 or 3e5, use lower 'e' */
11919 *d++ = 'e'; /* At least some Mach atof()s don't grok 'E' */
11921 /* stray preinitial _ */
11923 if (ckWARN(WARN_SYNTAX))
11924 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
11925 "Misplaced _ in number");
11929 /* allow positive or negative exponent */
11930 if (*s == '+' || *s == '-')
11933 /* stray initial _ */
11935 if (ckWARN(WARN_SYNTAX))
11936 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
11937 "Misplaced _ in number");
11941 /* read digits of exponent */
11942 while (isDIGIT(*s) || *s == '_') {
11945 Perl_croak(aTHX_ number_too_long);
11949 if (((lastub && s == lastub + 1) ||
11950 (!isDIGIT(s[1]) && s[1] != '_'))
11951 && ckWARN(WARN_SYNTAX))
11952 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
11953 "Misplaced _ in number");
11960 /* make an sv from the string */
11964 We try to do an integer conversion first if no characters
11965 indicating "float" have been found.
11970 const int flags = grok_number (PL_tokenbuf, d - PL_tokenbuf, &uv);
11972 if (flags == IS_NUMBER_IN_UV) {
11974 sv_setiv(sv, uv); /* Prefer IVs over UVs. */
11977 } else if (flags == (IS_NUMBER_IN_UV | IS_NUMBER_NEG)) {
11978 if (uv <= (UV) IV_MIN)
11979 sv_setiv(sv, -(IV)uv);
11986 /* terminate the string */
11988 nv = Atof(PL_tokenbuf);
11992 if ( floatit ? (PL_hints & HINT_NEW_FLOAT) :
11993 (PL_hints & HINT_NEW_INTEGER) )
11994 sv = new_constant(PL_tokenbuf, d - PL_tokenbuf,
11995 (floatit ? "float" : "integer"),
11999 /* if it starts with a v, it could be a v-string */
12002 sv = newSV(5); /* preallocate storage space */
12003 s = scan_vstring(s,sv);
12007 /* make the op for the constant and return */
12010 lvalp->opval = newSVOP(OP_CONST, 0, sv);
12012 lvalp->opval = NULL;
12018 S_scan_formline(pTHX_ register char *s)
12021 register char *eol;
12023 SV * const stuff = newSVpvs("");
12024 bool needargs = FALSE;
12025 bool eofmt = FALSE;
12027 char *tokenstart = s;
12030 if (PL_madskills) {
12031 savewhite = PL_thiswhite;
12036 while (!needargs) {
12038 #ifdef PERL_STRICT_CR
12039 for (t = s+1;SPACE_OR_TAB(*t); t++) ;
12041 for (t = s+1;SPACE_OR_TAB(*t) || *t == '\r'; t++) ;
12043 if (*t == '\n' || t == PL_bufend) {
12048 if (PL_in_eval && !PL_rsfp) {
12049 eol = (char *) memchr(s,'\n',PL_bufend-s);
12054 eol = PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
12056 for (t = s; t < eol; t++) {
12057 if (*t == '~' && t[1] == '~' && SvCUR(stuff)) {
12059 goto enough; /* ~~ must be first line in formline */
12061 if (*t == '@' || *t == '^')
12065 sv_catpvn(stuff, s, eol-s);
12066 #ifndef PERL_STRICT_CR
12067 if (eol-s > 1 && eol[-2] == '\r' && eol[-1] == '\n') {
12068 char *end = SvPVX(stuff) + SvCUR(stuff);
12071 SvCUR_set(stuff, SvCUR(stuff) - 1);
12081 if (PL_madskills) {
12083 sv_catpvn(PL_thistoken, tokenstart, PL_bufend - tokenstart);
12085 PL_thistoken = newSVpvn(tokenstart, PL_bufend - tokenstart);
12088 s = filter_gets(PL_linestr, PL_rsfp, 0);
12090 tokenstart = PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = SvPVX(PL_linestr);
12092 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = SvPVX(PL_linestr);
12094 PL_bufend = PL_bufptr + SvCUR(PL_linestr);
12095 PL_last_lop = PL_last_uni = NULL;
12104 if (SvCUR(stuff)) {
12107 PL_lex_state = LEX_NORMAL;
12108 start_force(PL_curforce);
12109 NEXTVAL_NEXTTOKE.ival = 0;
12113 PL_lex_state = LEX_FORMLINE;
12115 if (UTF && is_utf8_string((U8*)SvPVX_const(stuff), SvCUR(stuff)))
12117 else if (PL_encoding)
12118 sv_recode_to_utf8(stuff, PL_encoding);
12120 start_force(PL_curforce);
12121 NEXTVAL_NEXTTOKE.opval = (OP*)newSVOP(OP_CONST, 0, stuff);
12123 start_force(PL_curforce);
12124 NEXTVAL_NEXTTOKE.ival = OP_FORMLINE;
12128 SvREFCNT_dec(stuff);
12130 PL_lex_formbrack = 0;
12134 if (PL_madskills) {
12136 sv_catpvn(PL_thistoken, tokenstart, s - tokenstart);
12138 PL_thistoken = newSVpvn(tokenstart, s - tokenstart);
12139 PL_thiswhite = savewhite;
12151 PL_cshlen = strlen(PL_cshname);
12153 #if defined(USE_ITHREADS)
12154 PERL_UNUSED_CONTEXT;
12160 Perl_start_subparse(pTHX_ I32 is_format, U32 flags)
12163 const I32 oldsavestack_ix = PL_savestack_ix;
12164 CV* const outsidecv = PL_compcv;
12167 assert(SvTYPE(PL_compcv) == SVt_PVCV);
12169 SAVEI32(PL_subline);
12170 save_item(PL_subname);
12171 SAVESPTR(PL_compcv);
12173 PL_compcv = (CV*)newSV(0);
12174 sv_upgrade((SV *)PL_compcv, is_format ? SVt_PVFM : SVt_PVCV);
12175 CvFLAGS(PL_compcv) |= flags;
12177 PL_subline = CopLINE(PL_curcop);
12178 CvPADLIST(PL_compcv) = pad_new(padnew_SAVE|padnew_SAVESUB);
12179 CvOUTSIDE(PL_compcv) = (CV*)SvREFCNT_inc_simple(outsidecv);
12180 CvOUTSIDE_SEQ(PL_compcv) = PL_cop_seqmax;
12182 return oldsavestack_ix;
12186 #pragma segment Perl_yylex
12189 Perl_yywarn(pTHX_ const char *s)
12192 PL_in_eval |= EVAL_WARNONLY;
12194 PL_in_eval &= ~EVAL_WARNONLY;
12199 Perl_yyerror(pTHX_ const char *s)
12202 const char *where = NULL;
12203 const char *context = NULL;
12207 if (!yychar || (yychar == ';' && !PL_rsfp))
12209 else if (PL_oldoldbufptr && PL_bufptr > PL_oldoldbufptr &&
12210 PL_bufptr - PL_oldoldbufptr < 200 && PL_oldoldbufptr != PL_oldbufptr &&
12211 PL_oldbufptr != PL_bufptr) {
12214 The code below is removed for NetWare because it abends/crashes on NetWare
12215 when the script has error such as not having the closing quotes like:
12216 if ($var eq "value)
12217 Checking of white spaces is anyway done in NetWare code.
12220 while (isSPACE(*PL_oldoldbufptr))
12223 context = PL_oldoldbufptr;
12224 contlen = PL_bufptr - PL_oldoldbufptr;
12226 else if (PL_oldbufptr && PL_bufptr > PL_oldbufptr &&
12227 PL_bufptr - PL_oldbufptr < 200 && PL_oldbufptr != PL_bufptr) {
12230 The code below is removed for NetWare because it abends/crashes on NetWare
12231 when the script has error such as not having the closing quotes like:
12232 if ($var eq "value)
12233 Checking of white spaces is anyway done in NetWare code.
12236 while (isSPACE(*PL_oldbufptr))
12239 context = PL_oldbufptr;
12240 contlen = PL_bufptr - PL_oldbufptr;
12242 else if (yychar > 255)
12243 where = "next token ???";
12244 else if (yychar == -2) { /* YYEMPTY */
12245 if (PL_lex_state == LEX_NORMAL ||
12246 (PL_lex_state == LEX_KNOWNEXT && PL_lex_defer == LEX_NORMAL))
12247 where = "at end of line";
12248 else if (PL_lex_inpat)
12249 where = "within pattern";
12251 where = "within string";
12254 SV * const where_sv = sv_2mortal(newSVpvs("next char "));
12256 Perl_sv_catpvf(aTHX_ where_sv, "^%c", toCTRL(yychar));
12257 else if (isPRINT_LC(yychar))
12258 Perl_sv_catpvf(aTHX_ where_sv, "%c", yychar);
12260 Perl_sv_catpvf(aTHX_ where_sv, "\\%03o", yychar & 255);
12261 where = SvPVX_const(where_sv);
12263 msg = sv_2mortal(newSVpv(s, 0));
12264 Perl_sv_catpvf(aTHX_ msg, " at %s line %"IVdf", ",
12265 OutCopFILE(PL_curcop), (IV)CopLINE(PL_curcop));
12267 Perl_sv_catpvf(aTHX_ msg, "near \"%.*s\"\n", contlen, context);
12269 Perl_sv_catpvf(aTHX_ msg, "%s\n", where);
12270 if (PL_multi_start < PL_multi_end && (U32)(CopLINE(PL_curcop) - PL_multi_end) <= 1) {
12271 Perl_sv_catpvf(aTHX_ msg,
12272 " (Might be a runaway multi-line %c%c string starting on line %"IVdf")\n",
12273 (int)PL_multi_open,(int)PL_multi_close,(IV)PL_multi_start);
12276 if (PL_in_eval & EVAL_WARNONLY && ckWARN_d(WARN_SYNTAX))
12277 Perl_warner(aTHX_ packWARN(WARN_SYNTAX), "%"SVf, (void*)msg);
12280 if (PL_error_count >= 10) {
12281 if (PL_in_eval && SvCUR(ERRSV))
12282 Perl_croak(aTHX_ "%"SVf"%s has too many errors.\n",
12283 (void*)ERRSV, OutCopFILE(PL_curcop));
12285 Perl_croak(aTHX_ "%s has too many errors.\n",
12286 OutCopFILE(PL_curcop));
12289 PL_in_my_stash = NULL;
12293 #pragma segment Main
12297 S_swallow_bom(pTHX_ U8 *s)
12300 const STRLEN slen = SvCUR(PL_linestr);
12303 if (s[1] == 0xFE) {
12304 /* UTF-16 little-endian? (or UTF32-LE?) */
12305 if (s[2] == 0 && s[3] == 0) /* UTF-32 little-endian */
12306 Perl_croak(aTHX_ "Unsupported script encoding UTF32-LE");
12307 #ifndef PERL_NO_UTF16_FILTER
12308 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF16-LE script encoding (BOM)\n");
12311 if (PL_bufend > (char*)s) {
12315 filter_add(utf16rev_textfilter, NULL);
12316 Newx(news, (PL_bufend - (char*)s) * 3 / 2 + 1, U8);
12317 utf16_to_utf8_reversed(s, news,
12318 PL_bufend - (char*)s - 1,
12320 sv_setpvn(PL_linestr, (const char*)news, newlen);
12322 s = (U8*)SvPVX(PL_linestr);
12323 Copy(news, s, newlen, U8);
12327 SvUTF8_on(PL_linestr);
12328 s = (U8*)SvPVX(PL_linestr);
12330 /* FIXME - is this a general bug fix? */
12333 PL_bufend = SvPVX(PL_linestr) + newlen;
12336 Perl_croak(aTHX_ "Unsupported script encoding UTF16-LE");
12341 if (s[1] == 0xFF) { /* UTF-16 big-endian? */
12342 #ifndef PERL_NO_UTF16_FILTER
12343 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16BE script encoding (BOM)\n");
12346 if (PL_bufend > (char *)s) {
12350 filter_add(utf16_textfilter, NULL);
12351 Newx(news, (PL_bufend - (char*)s) * 3 / 2 + 1, U8);
12352 utf16_to_utf8(s, news,
12353 PL_bufend - (char*)s,
12355 sv_setpvn(PL_linestr, (const char*)news, newlen);
12357 SvUTF8_on(PL_linestr);
12358 s = (U8*)SvPVX(PL_linestr);
12359 PL_bufend = SvPVX(PL_linestr) + newlen;
12362 Perl_croak(aTHX_ "Unsupported script encoding UTF16-BE");
12367 if (slen > 2 && s[1] == 0xBB && s[2] == 0xBF) {
12368 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-8 script encoding (BOM)\n");
12369 s += 3; /* UTF-8 */
12375 if (s[2] == 0xFE && s[3] == 0xFF) {
12376 /* UTF-32 big-endian */
12377 Perl_croak(aTHX_ "Unsupported script encoding UTF32-BE");
12380 else if (s[2] == 0 && s[3] != 0) {
12383 * are a good indicator of UTF-16BE. */
12384 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16BE script encoding (no BOM)\n");
12389 if (slen > 3 && s[1] == 0 && s[2] != 0 && s[3] == 0) {
12392 * are a good indicator of UTF-16LE. */
12393 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16LE script encoding (no BOM)\n");
12402 * Restore a source filter.
12406 restore_rsfp(pTHX_ void *f)
12409 PerlIO * const fp = (PerlIO*)f;
12411 if (PL_rsfp == PerlIO_stdin())
12412 PerlIO_clearerr(PL_rsfp);
12413 else if (PL_rsfp && (PL_rsfp != fp))
12414 PerlIO_close(PL_rsfp);
12418 #ifndef PERL_NO_UTF16_FILTER
12420 utf16_textfilter(pTHX_ int idx, SV *sv, int maxlen)
12423 const STRLEN old = SvCUR(sv);
12424 const I32 count = FILTER_READ(idx+1, sv, maxlen);
12425 DEBUG_P(PerlIO_printf(Perl_debug_log,
12426 "utf16_textfilter(%p): %d %d (%d)\n",
12427 FPTR2DPTR(void *, utf16_textfilter),
12428 idx, maxlen, (int) count));
12432 Newx(tmps, SvCUR(sv) * 3 / 2 + 1, U8);
12433 Copy(SvPVX_const(sv), tmps, old, char);
12434 utf16_to_utf8((U8*)SvPVX_const(sv) + old, tmps + old,
12435 SvCUR(sv) - old, &newlen);
12436 sv_usepvn(sv, (char*)tmps, (STRLEN)newlen + old);
12438 DEBUG_P({sv_dump(sv);});
12443 utf16rev_textfilter(pTHX_ int idx, SV *sv, int maxlen)
12446 const STRLEN old = SvCUR(sv);
12447 const I32 count = FILTER_READ(idx+1, sv, maxlen);
12448 DEBUG_P(PerlIO_printf(Perl_debug_log,
12449 "utf16rev_textfilter(%p): %d %d (%d)\n",
12450 FPTR2DPTR(void *, utf16rev_textfilter),
12451 idx, maxlen, (int) count));
12455 Newx(tmps, SvCUR(sv) * 3 / 2 + 1, U8);
12456 Copy(SvPVX_const(sv), tmps, old, char);
12457 utf16_to_utf8((U8*)SvPVX_const(sv) + old, tmps + old,
12458 SvCUR(sv) - old, &newlen);
12459 sv_usepvn(sv, (char*)tmps, (STRLEN)newlen + old);
12461 DEBUG_P({ sv_dump(sv); });
12467 Returns a pointer to the next character after the parsed
12468 vstring, as well as updating the passed in sv.
12470 Function must be called like
12473 s = scan_vstring(s,sv);
12475 The sv should already be large enough to store the vstring
12476 passed in, for performance reasons.
12481 Perl_scan_vstring(pTHX_ const char *s, SV *sv)
12484 const char *pos = s;
12485 const char *start = s;
12486 if (*pos == 'v') pos++; /* get past 'v' */
12487 while (pos < PL_bufend && (isDIGIT(*pos) || *pos == '_'))
12489 if ( *pos != '.') {
12490 /* this may not be a v-string if followed by => */
12491 const char *next = pos;
12492 while (next < PL_bufend && isSPACE(*next))
12494 if ((PL_bufend - next) >= 2 && *next == '=' && next[1] == '>' ) {
12495 /* return string not v-string */
12496 sv_setpvn(sv,(char *)s,pos-s);
12497 return (char *)pos;
12501 if (!isALPHA(*pos)) {
12502 U8 tmpbuf[UTF8_MAXBYTES+1];
12505 s++; /* get past 'v' */
12507 sv_setpvn(sv, "", 0);
12510 /* this is atoi() that tolerates underscores */
12513 const char *end = pos;
12515 while (--end >= s) {
12517 const UV orev = rev;
12518 rev += (*end - '0') * mult;
12520 if (orev > rev && ckWARN_d(WARN_OVERFLOW))
12521 Perl_warner(aTHX_ packWARN(WARN_OVERFLOW),
12522 "Integer overflow in decimal number");
12526 if (rev > 0x7FFFFFFF)
12527 Perl_croak(aTHX_ "In EBCDIC the v-string components cannot exceed 2147483647");
12529 /* Append native character for the rev point */
12530 tmpend = uvchr_to_utf8(tmpbuf, rev);
12531 sv_catpvn(sv, (const char*)tmpbuf, tmpend - tmpbuf);
12532 if (!UNI_IS_INVARIANT(NATIVE_TO_UNI(rev)))
12534 if (pos + 1 < PL_bufend && *pos == '.' && isDIGIT(pos[1]))
12540 while (pos < PL_bufend && (isDIGIT(*pos) || *pos == '_'))
12544 sv_magic(sv,NULL,PERL_MAGIC_vstring,(const char*)start, pos-start);
12552 * c-indentation-style: bsd
12553 * c-basic-offset: 4
12554 * indent-tabs-mode: t
12557 * ex: set ts=8 sts=4 sw=4 noet: