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, NULL }
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; (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) my_strlcpy(&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, '"'))) {
745 while (SPACE_OR_TAB(*e) || *e == '\r' || *e == '\f')
747 if (*e != '\n' && *e != '\0')
748 return; /* false alarm */
754 const char * const cf = CopFILE(PL_curcop);
755 STRLEN tmplen = cf ? strlen(cf) : 0;
756 if (tmplen > 7 && strnEQ(cf, "(eval ", 6)) {
757 /* must copy *{"::_<(eval N)[oldfilename:L]"}
758 * to *{"::_<newfilename"} */
759 char smallbuf[256], smallbuf2[256];
760 char *tmpbuf, *tmpbuf2;
762 STRLEN tmplen2 = strlen(s);
763 if (tmplen + 3 < sizeof smallbuf)
766 Newx(tmpbuf, tmplen + 3, char);
767 if (tmplen2 + 3 < sizeof smallbuf2)
770 Newx(tmpbuf2, tmplen2 + 3, char);
771 tmpbuf[0] = tmpbuf2[0] = '_';
772 tmpbuf[1] = tmpbuf2[1] = '<';
773 memcpy(tmpbuf + 2, cf, ++tmplen);
774 memcpy(tmpbuf2 + 2, s, ++tmplen2);
776 gvp = (GV**)hv_fetch(PL_defstash, tmpbuf, tmplen, FALSE);
778 gv2 = *(GV**)hv_fetch(PL_defstash, tmpbuf2, tmplen2, TRUE);
780 gv_init(gv2, PL_defstash, tmpbuf2, tmplen2, FALSE);
781 /* adjust ${"::_<newfilename"} to store the new file name */
782 GvSV(gv2) = newSVpvn(tmpbuf2 + 2, tmplen2 - 2);
783 GvHV(gv2) = (HV*)SvREFCNT_inc(GvHV(*gvp));
784 GvAV(gv2) = (AV*)SvREFCNT_inc(GvAV(*gvp));
786 if (tmpbuf != smallbuf) Safefree(tmpbuf);
787 if (tmpbuf2 != smallbuf2) Safefree(tmpbuf2);
790 CopFILE_free(PL_curcop);
791 CopFILE_set(PL_curcop, s);
794 CopLINE_set(PL_curcop, atoi(n)-1);
798 /* skip space before PL_thistoken */
801 S_skipspace0(pTHX_ register char *s)
808 PL_thiswhite = newSVpvn("",0);
809 sv_catsv(PL_thiswhite, PL_skipwhite);
810 sv_free(PL_skipwhite);
813 PL_realtokenstart = s - SvPVX(PL_linestr);
817 /* skip space after PL_thistoken */
820 S_skipspace1(pTHX_ register char *s)
822 const char *start = s;
823 I32 startoff = start - SvPVX(PL_linestr);
828 start = SvPVX(PL_linestr) + startoff;
829 if (!PL_thistoken && PL_realtokenstart >= 0) {
830 const char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
831 PL_thistoken = newSVpvn(tstart, start - tstart);
833 PL_realtokenstart = -1;
836 PL_nextwhite = newSVpvn("",0);
837 sv_catsv(PL_nextwhite, PL_skipwhite);
838 sv_free(PL_skipwhite);
845 S_skipspace2(pTHX_ register char *s, SV **svp)
848 const I32 bufptroff = PL_bufptr - SvPVX(PL_linestr);
849 const I32 startoff = s - SvPVX(PL_linestr);
852 PL_bufptr = SvPVX(PL_linestr) + bufptroff;
853 if (!PL_madskills || !svp)
855 start = SvPVX(PL_linestr) + startoff;
856 if (!PL_thistoken && PL_realtokenstart >= 0) {
857 char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
858 PL_thistoken = newSVpvn(tstart, start - tstart);
859 PL_realtokenstart = -1;
863 *svp = newSVpvn("",0);
864 sv_setsv(*svp, PL_skipwhite);
865 sv_free(PL_skipwhite);
875 * Called to gobble the appropriate amount and type of whitespace.
876 * Skips comments as well.
880 S_skipspace(pTHX_ register char *s)
885 int startoff = s - SvPVX(PL_linestr);
888 sv_free(PL_skipwhite);
893 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
894 while (s < PL_bufend && SPACE_OR_TAB(*s))
904 SSize_t oldprevlen, oldoldprevlen;
905 SSize_t oldloplen = 0, oldunilen = 0;
906 while (s < PL_bufend && isSPACE(*s)) {
907 if (*s++ == '\n' && PL_in_eval && !PL_rsfp)
912 if (s < PL_bufend && *s == '#') {
913 while (s < PL_bufend && *s != '\n')
917 if (PL_in_eval && !PL_rsfp) {
924 /* only continue to recharge the buffer if we're at the end
925 * of the buffer, we're not reading from a source filter, and
926 * we're in normal lexing mode
928 if (s < PL_bufend || !PL_rsfp || PL_sublex_info.sub_inwhat ||
929 PL_lex_state == LEX_FORMLINE)
936 /* try to recharge the buffer */
938 curoff = s - SvPVX(PL_linestr);
941 if ((s = filter_gets(PL_linestr, PL_rsfp,
942 (prevlen = SvCUR(PL_linestr)))) == NULL)
945 if (PL_madskills && curoff != startoff) {
947 PL_skipwhite = newSVpvn("",0);
948 sv_catpvn(PL_skipwhite, SvPVX(PL_linestr) + startoff,
952 /* mustn't throw out old stuff yet if madpropping */
953 SvCUR(PL_linestr) = curoff;
954 s = SvPVX(PL_linestr) + curoff;
956 if (curoff && s[-1] == '\n')
960 /* end of file. Add on the -p or -n magic */
961 /* XXX these shouldn't really be added here, can't set PL_faketokens */
965 ";}continue{print or die qq(-p destination: $!\\n);}");
968 ";}continue{print or die qq(-p destination: $!\\n);}");
970 PL_minus_n = PL_minus_p = 0;
972 else if (PL_minus_n) {
974 sv_catpvn(PL_linestr, ";}", 2);
976 sv_setpvn(PL_linestr, ";}", 2);
982 sv_catpvn(PL_linestr,";", 1);
984 sv_setpvn(PL_linestr,";", 1);
987 /* reset variables for next time we lex */
988 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = s = PL_linestart
994 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
995 PL_last_lop = PL_last_uni = NULL;
997 /* Close the filehandle. Could be from -P preprocessor,
998 * STDIN, or a regular file. If we were reading code from
999 * STDIN (because the commandline held no -e or filename)
1000 * then we don't close it, we reset it so the code can
1001 * read from STDIN too.
1004 if (PL_preprocess && !PL_in_eval)
1005 (void)PerlProc_pclose(PL_rsfp);
1006 else if ((PerlIO*)PL_rsfp == PerlIO_stdin())
1007 PerlIO_clearerr(PL_rsfp);
1009 (void)PerlIO_close(PL_rsfp);
1014 /* not at end of file, so we only read another line */
1015 /* make corresponding updates to old pointers, for yyerror() */
1016 oldprevlen = PL_oldbufptr - PL_bufend;
1017 oldoldprevlen = PL_oldoldbufptr - PL_bufend;
1019 oldunilen = PL_last_uni - PL_bufend;
1021 oldloplen = PL_last_lop - PL_bufend;
1022 PL_linestart = PL_bufptr = s + prevlen;
1023 PL_bufend = s + SvCUR(PL_linestr);
1025 PL_oldbufptr = s + oldprevlen;
1026 PL_oldoldbufptr = s + oldoldprevlen;
1028 PL_last_uni = s + oldunilen;
1030 PL_last_lop = s + oldloplen;
1033 /* debugger active and we're not compiling the debugger code,
1034 * so store the line into the debugger's array of lines
1036 if (PERLDB_LINE && PL_curstash != PL_debstash) {
1037 SV * const sv = newSV(0);
1039 sv_upgrade(sv, SVt_PVMG);
1040 sv_setpvn(sv,PL_bufptr,PL_bufend-PL_bufptr);
1043 av_store(CopFILEAVx(PL_curcop),(I32)CopLINE(PL_curcop),sv);
1051 PL_skipwhite = newSVpvn("",0);
1052 curoff = s - SvPVX(PL_linestr);
1053 if (curoff - startoff)
1054 sv_catpvn(PL_skipwhite, SvPVX(PL_linestr) + startoff,
1063 * Check the unary operators to ensure there's no ambiguity in how they're
1064 * used. An ambiguous piece of code would be:
1066 * This doesn't mean rand() + 5. Because rand() is a unary operator,
1067 * the +5 is its argument.
1077 if (PL_oldoldbufptr != PL_last_uni)
1079 while (isSPACE(*PL_last_uni))
1082 while (isALNUM_lazy_if(s,UTF) || *s == '-')
1084 if ((t = strchr(s, '(')) && t < PL_bufptr)
1087 if (ckWARN_d(WARN_AMBIGUOUS)){
1088 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
1089 "Warning: Use of \"%.*s\" without parentheses is ambiguous",
1090 (int)(s - PL_last_uni), PL_last_uni);
1095 * LOP : macro to build a list operator. Its behaviour has been replaced
1096 * with a subroutine, S_lop() for which LOP is just another name.
1099 #define LOP(f,x) return lop(f,x,s)
1103 * Build a list operator (or something that might be one). The rules:
1104 * - if we have a next token, then it's a list operator [why?]
1105 * - if the next thing is an opening paren, then it's a function
1106 * - else it's a list operator
1110 S_lop(pTHX_ I32 f, int x, char *s)
1117 PL_last_lop = PL_oldbufptr;
1118 PL_last_lop_op = (OPCODE)f;
1121 return REPORT(LSTOP);
1124 return REPORT(LSTOP);
1127 return REPORT(FUNC);
1130 return REPORT(FUNC);
1132 return REPORT(LSTOP);
1138 * Sets up for an eventual force_next(). start_force(0) basically does
1139 * an unshift, while start_force(-1) does a push. yylex removes items
1144 S_start_force(pTHX_ int where)
1148 if (where < 0) /* so people can duplicate start_force(PL_curforce) */
1149 where = PL_lasttoke;
1150 assert(PL_curforce < 0 || PL_curforce == where);
1151 if (PL_curforce != where) {
1152 for (i = PL_lasttoke; i > where; --i) {
1153 PL_nexttoke[i] = PL_nexttoke[i-1];
1157 if (PL_curforce < 0) /* in case of duplicate start_force() */
1158 Zero(&PL_nexttoke[where], 1, NEXTTOKE);
1159 PL_curforce = where;
1162 curmad('^', newSVpvn("",0));
1163 CURMAD('_', PL_nextwhite);
1168 S_curmad(pTHX_ char slot, SV *sv)
1174 if (PL_curforce < 0)
1175 where = &PL_thismad;
1177 where = &PL_nexttoke[PL_curforce].next_mad;
1180 sv_setpvn(sv, "", 0);
1183 if (UTF && is_utf8_string((U8*)SvPVX(sv), SvCUR(sv)))
1185 else if (PL_encoding) {
1186 sv_recode_to_utf8(sv, PL_encoding);
1191 /* keep a slot open for the head of the list? */
1192 if (slot != '_' && *where && (*where)->mad_key == '^') {
1193 (*where)->mad_key = slot;
1194 sv_free((*where)->mad_val);
1195 (*where)->mad_val = (void*)sv;
1198 addmad(newMADsv(slot, sv), where, 0);
1201 # define start_force(where) NOOP
1202 # define curmad(slot, sv) NOOP
1207 * When the lexer realizes it knows the next token (for instance,
1208 * it is reordering tokens for the parser) then it can call S_force_next
1209 * to know what token to return the next time the lexer is called. Caller
1210 * will need to set PL_nextval[] (or PL_nexttoke[].next_val with PERL_MAD),
1211 * and possibly PL_expect to ensure the lexer handles the token correctly.
1215 S_force_next(pTHX_ I32 type)
1219 if (PL_curforce < 0)
1220 start_force(PL_lasttoke);
1221 PL_nexttoke[PL_curforce].next_type = type;
1222 if (PL_lex_state != LEX_KNOWNEXT)
1223 PL_lex_defer = PL_lex_state;
1224 PL_lex_state = LEX_KNOWNEXT;
1225 PL_lex_expect = PL_expect;
1228 PL_nexttype[PL_nexttoke] = type;
1230 if (PL_lex_state != LEX_KNOWNEXT) {
1231 PL_lex_defer = PL_lex_state;
1232 PL_lex_expect = PL_expect;
1233 PL_lex_state = LEX_KNOWNEXT;
1239 S_newSV_maybe_utf8(pTHX_ const char *start, STRLEN len)
1242 SV * const sv = newSVpvn(start,len);
1243 if (UTF && !IN_BYTES && is_utf8_string((const U8*)start, len))
1250 * When the lexer knows the next thing is a word (for instance, it has
1251 * just seen -> and it knows that the next char is a word char, then
1252 * it calls S_force_word to stick the next word into the PL_next lookahead.
1255 * char *start : buffer position (must be within PL_linestr)
1256 * int token : PL_next will be this type of bare word (e.g., METHOD,WORD)
1257 * int check_keyword : if true, Perl checks to make sure the word isn't
1258 * a keyword (do this if the word is a label, e.g. goto FOO)
1259 * int allow_pack : if true, : characters will also be allowed (require,
1260 * use, etc. do this)
1261 * int allow_initial_tick : used by the "sub" lexer only.
1265 S_force_word(pTHX_ register char *start, int token, int check_keyword, int allow_pack, int allow_initial_tick)
1271 start = SKIPSPACE1(start);
1273 if (isIDFIRST_lazy_if(s,UTF) ||
1274 (allow_pack && *s == ':') ||
1275 (allow_initial_tick && *s == '\'') )
1277 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, allow_pack, &len);
1278 if (check_keyword && keyword(PL_tokenbuf, len))
1280 start_force(PL_curforce);
1282 curmad('X', newSVpvn(start,s-start));
1283 if (token == METHOD) {
1288 PL_expect = XOPERATOR;
1291 NEXTVAL_NEXTTOKE.opval
1292 = (OP*)newSVOP(OP_CONST,0,
1293 S_newSV_maybe_utf8(aTHX_ PL_tokenbuf, len));
1294 NEXTVAL_NEXTTOKE.opval->op_private |= OPpCONST_BARE;
1302 * Called when the lexer wants $foo *foo &foo etc, but the program
1303 * text only contains the "foo" portion. The first argument is a pointer
1304 * to the "foo", and the second argument is the type symbol to prefix.
1305 * Forces the next token to be a "WORD".
1306 * Creates the symbol if it didn't already exist (via gv_fetchpv()).
1310 S_force_ident(pTHX_ register const char *s, int kind)
1314 const STRLEN len = strlen(s);
1315 OP* const o = (OP*)newSVOP(OP_CONST, 0, newSVpvn(s, len));
1316 start_force(PL_curforce);
1317 NEXTVAL_NEXTTOKE.opval = o;
1320 o->op_private = OPpCONST_ENTERED;
1321 /* XXX see note in pp_entereval() for why we forgo typo
1322 warnings if the symbol must be introduced in an eval.
1324 gv_fetchpvn_flags(s, len,
1325 PL_in_eval ? (GV_ADDMULTI | GV_ADDINEVAL)
1327 kind == '$' ? SVt_PV :
1328 kind == '@' ? SVt_PVAV :
1329 kind == '%' ? SVt_PVHV :
1337 Perl_str_to_version(pTHX_ SV *sv)
1342 const char *start = SvPV_const(sv,len);
1343 const char * const end = start + len;
1344 const bool utf = SvUTF8(sv) ? TRUE : FALSE;
1345 while (start < end) {
1349 n = utf8n_to_uvchr((U8*)start, len, &skip, 0);
1354 retval += ((NV)n)/nshift;
1363 * Forces the next token to be a version number.
1364 * If the next token appears to be an invalid version number, (e.g. "v2b"),
1365 * and if "guessing" is TRUE, then no new token is created (and the caller
1366 * must use an alternative parsing method).
1370 S_force_version(pTHX_ char *s, int guessing)
1376 I32 startoff = s - SvPVX(PL_linestr);
1385 while (isDIGIT(*d) || *d == '_' || *d == '.')
1389 start_force(PL_curforce);
1390 curmad('X', newSVpvn(s,d-s));
1393 if (*d == ';' || isSPACE(*d) || *d == '}' || !*d) {
1395 s = scan_num(s, &yylval);
1396 version = yylval.opval;
1397 ver = cSVOPx(version)->op_sv;
1398 if (SvPOK(ver) && !SvNIOK(ver)) {
1399 SvUPGRADE(ver, SVt_PVNV);
1400 SvNV_set(ver, str_to_version(ver));
1401 SvNOK_on(ver); /* hint that it is a version */
1404 else if (guessing) {
1407 sv_free(PL_nextwhite); /* let next token collect whitespace */
1409 s = SvPVX(PL_linestr) + startoff;
1417 if (PL_madskills && !version) {
1418 sv_free(PL_nextwhite); /* let next token collect whitespace */
1420 s = SvPVX(PL_linestr) + startoff;
1423 /* NOTE: The parser sees the package name and the VERSION swapped */
1424 start_force(PL_curforce);
1425 NEXTVAL_NEXTTOKE.opval = version;
1433 * Tokenize a quoted string passed in as an SV. It finds the next
1434 * chunk, up to end of string or a backslash. It may make a new
1435 * SV containing that chunk (if HINT_NEW_STRING is on). It also
1440 S_tokeq(pTHX_ SV *sv)
1444 register char *send;
1452 s = SvPV_force(sv, len);
1453 if (SvTYPE(sv) >= SVt_PVIV && SvIVX(sv) == -1)
1456 while (s < send && *s != '\\')
1461 if ( PL_hints & HINT_NEW_STRING ) {
1462 pv = sv_2mortal(newSVpvn(SvPVX_const(pv), len));
1468 if (s + 1 < send && (s[1] == '\\'))
1469 s++; /* all that, just for this */
1474 SvCUR_set(sv, d - SvPVX_const(sv));
1476 if ( PL_hints & HINT_NEW_STRING )
1477 return new_constant(NULL, 0, "q", sv, pv, "q");
1482 * Now come three functions related to double-quote context,
1483 * S_sublex_start, S_sublex_push, and S_sublex_done. They're used when
1484 * converting things like "\u\Lgnat" into ucfirst(lc("gnat")). They
1485 * interact with PL_lex_state, and create fake ( ... ) argument lists
1486 * to handle functions and concatenation.
1487 * They assume that whoever calls them will be setting up a fake
1488 * join call, because each subthing puts a ',' after it. This lets
1491 * join($, , 'lower ', lcfirst( 'uPpEr', ) ,)
1493 * (I'm not sure whether the spurious commas at the end of lcfirst's
1494 * arguments and join's arguments are created or not).
1499 * Assumes that yylval.ival is the op we're creating (e.g. OP_LCFIRST).
1501 * Pattern matching will set PL_lex_op to the pattern-matching op to
1502 * make (we return THING if yylval.ival is OP_NULL, PMFUNC otherwise).
1504 * OP_CONST and OP_READLINE are easy--just make the new op and return.
1506 * Everything else becomes a FUNC.
1508 * Sets PL_lex_state to LEX_INTERPPUSH unless (ival was OP_NULL or we
1509 * had an OP_CONST or OP_READLINE). This just sets us up for a
1510 * call to S_sublex_push().
1514 S_sublex_start(pTHX)
1517 register const I32 op_type = yylval.ival;
1519 if (op_type == OP_NULL) {
1520 yylval.opval = PL_lex_op;
1524 if (op_type == OP_CONST || op_type == OP_READLINE) {
1525 SV *sv = tokeq(PL_lex_stuff);
1527 if (SvTYPE(sv) == SVt_PVIV) {
1528 /* Overloaded constants, nothing fancy: Convert to SVt_PV: */
1530 const char * const p = SvPV_const(sv, len);
1531 SV * const nsv = newSVpvn(p, len);
1537 yylval.opval = (OP*)newSVOP(op_type, 0, sv);
1538 PL_lex_stuff = NULL;
1539 /* Allow <FH> // "foo" */
1540 if (op_type == OP_READLINE)
1541 PL_expect = XTERMORDORDOR;
1545 PL_sublex_info.super_state = PL_lex_state;
1546 PL_sublex_info.sub_inwhat = op_type;
1547 PL_sublex_info.sub_op = PL_lex_op;
1548 PL_lex_state = LEX_INTERPPUSH;
1552 yylval.opval = PL_lex_op;
1562 * Create a new scope to save the lexing state. The scope will be
1563 * ended in S_sublex_done. Returns a '(', starting the function arguments
1564 * to the uc, lc, etc. found before.
1565 * Sets PL_lex_state to LEX_INTERPCONCAT.
1574 PL_lex_state = PL_sublex_info.super_state;
1575 SAVEI32(PL_lex_dojoin);
1576 SAVEI32(PL_lex_brackets);
1577 SAVEI32(PL_lex_casemods);
1578 SAVEI32(PL_lex_starts);
1579 SAVEI32(PL_lex_state);
1580 SAVEVPTR(PL_lex_inpat);
1581 SAVEI32(PL_lex_inwhat);
1582 SAVECOPLINE(PL_curcop);
1583 SAVEPPTR(PL_bufptr);
1584 SAVEPPTR(PL_bufend);
1585 SAVEPPTR(PL_oldbufptr);
1586 SAVEPPTR(PL_oldoldbufptr);
1587 SAVEPPTR(PL_last_lop);
1588 SAVEPPTR(PL_last_uni);
1589 SAVEPPTR(PL_linestart);
1590 SAVESPTR(PL_linestr);
1591 SAVEGENERICPV(PL_lex_brackstack);
1592 SAVEGENERICPV(PL_lex_casestack);
1594 PL_linestr = PL_lex_stuff;
1595 PL_lex_stuff = NULL;
1597 PL_bufend = PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart
1598 = SvPVX(PL_linestr);
1599 PL_bufend += SvCUR(PL_linestr);
1600 PL_last_lop = PL_last_uni = NULL;
1601 SAVEFREESV(PL_linestr);
1603 PL_lex_dojoin = FALSE;
1604 PL_lex_brackets = 0;
1605 Newx(PL_lex_brackstack, 120, char);
1606 Newx(PL_lex_casestack, 12, char);
1607 PL_lex_casemods = 0;
1608 *PL_lex_casestack = '\0';
1610 PL_lex_state = LEX_INTERPCONCAT;
1611 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
1613 PL_lex_inwhat = PL_sublex_info.sub_inwhat;
1614 if (PL_lex_inwhat == OP_MATCH || PL_lex_inwhat == OP_QR || PL_lex_inwhat == OP_SUBST)
1615 PL_lex_inpat = PL_sublex_info.sub_op;
1617 PL_lex_inpat = NULL;
1624 * Restores lexer state after a S_sublex_push.
1631 if (!PL_lex_starts++) {
1632 SV * const sv = newSVpvs("");
1633 if (SvUTF8(PL_linestr))
1635 PL_expect = XOPERATOR;
1636 yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
1640 if (PL_lex_casemods) { /* oops, we've got some unbalanced parens */
1641 PL_lex_state = LEX_INTERPCASEMOD;
1645 /* Is there a right-hand side to take care of? (s//RHS/ or tr//RHS/) */
1646 if (PL_lex_repl && (PL_lex_inwhat == OP_SUBST || PL_lex_inwhat == OP_TRANS)) {
1647 PL_linestr = PL_lex_repl;
1649 PL_bufend = PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart = SvPVX(PL_linestr);
1650 PL_bufend += SvCUR(PL_linestr);
1651 PL_last_lop = PL_last_uni = NULL;
1652 SAVEFREESV(PL_linestr);
1653 PL_lex_dojoin = FALSE;
1654 PL_lex_brackets = 0;
1655 PL_lex_casemods = 0;
1656 *PL_lex_casestack = '\0';
1658 if (SvEVALED(PL_lex_repl)) {
1659 PL_lex_state = LEX_INTERPNORMAL;
1661 /* we don't clear PL_lex_repl here, so that we can check later
1662 whether this is an evalled subst; that means we rely on the
1663 logic to ensure sublex_done() is called again only via the
1664 branch (in yylex()) that clears PL_lex_repl, else we'll loop */
1667 PL_lex_state = LEX_INTERPCONCAT;
1677 PL_endwhite = newSVpvn("",0);
1678 sv_catsv(PL_endwhite, PL_thiswhite);
1682 sv_setpvn(PL_thistoken,"",0);
1684 PL_realtokenstart = -1;
1688 PL_bufend = SvPVX(PL_linestr);
1689 PL_bufend += SvCUR(PL_linestr);
1690 PL_expect = XOPERATOR;
1691 PL_sublex_info.sub_inwhat = 0;
1699 Extracts a pattern, double-quoted string, or transliteration. This
1702 It looks at PL_lex_inwhat and PL_lex_inpat to find out whether it's
1703 processing a pattern (PL_lex_inpat is true), a transliteration
1704 (PL_lex_inwhat == OP_TRANS is true), or a double-quoted string.
1706 Returns a pointer to the character scanned up to. If this is
1707 advanced from the start pointer supplied (i.e. if anything was
1708 successfully parsed), will leave an OP for the substring scanned
1709 in yylval. Caller must intuit reason for not parsing further
1710 by looking at the next characters herself.
1714 double-quoted style: \r and \n
1715 regexp special ones: \D \s
1718 case and quoting: \U \Q \E
1719 stops on @ and $, but not for $ as tail anchor
1721 In transliterations:
1722 characters are VERY literal, except for - not at the start or end
1723 of the string, which indicates a range. If the range is in bytes,
1724 scan_const expands the range to the full set of intermediate
1725 characters. If the range is in utf8, the hyphen is replaced with
1726 a certain range mark which will be handled by pmtrans() in op.c.
1728 In double-quoted strings:
1730 double-quoted style: \r and \n
1732 deprecated backrefs: \1 (in substitution replacements)
1733 case and quoting: \U \Q \E
1736 scan_const does *not* construct ops to handle interpolated strings.
1737 It stops processing as soon as it finds an embedded $ or @ variable
1738 and leaves it to the caller to work out what's going on.
1740 embedded arrays (whether in pattern or not) could be:
1741 @foo, @::foo, @'foo, @{foo}, @$foo, @+, @-.
1743 $ in double-quoted strings must be the symbol of an embedded scalar.
1745 $ in pattern could be $foo or could be tail anchor. Assumption:
1746 it's a tail anchor if $ is the last thing in the string, or if it's
1747 followed by one of "()| \r\n\t"
1749 \1 (backreferences) are turned into $1
1751 The structure of the code is
1752 while (there's a character to process) {
1753 handle transliteration ranges
1754 skip regexp comments /(?#comment)/ and codes /(?{code})/
1755 skip #-initiated comments in //x patterns
1756 check for embedded arrays
1757 check for embedded scalars
1759 leave intact backslashes from leaveit (below)
1760 deprecate \1 in substitution replacements
1761 handle string-changing backslashes \l \U \Q \E, etc.
1762 switch (what was escaped) {
1763 handle \- in a transliteration (becomes a literal -)
1764 handle \132 (octal characters)
1765 handle \x15 and \x{1234} (hex characters)
1766 handle \N{name} (named characters)
1767 handle \cV (control characters)
1768 handle printf-style backslashes (\f, \r, \n, etc)
1770 } (end if backslash)
1771 } (end while character to read)
1776 S_scan_const(pTHX_ char *start)
1779 register char *send = PL_bufend; /* end of the constant */
1780 SV *sv = newSV(send - start); /* sv for the constant */
1781 register char *s = start; /* start of the constant */
1782 register char *d = SvPVX(sv); /* destination for copies */
1783 bool dorange = FALSE; /* are we in a translit range? */
1784 bool didrange = FALSE; /* did we just finish a range? */
1785 I32 has_utf8 = FALSE; /* Output constant is UTF8 */
1786 I32 this_utf8 = UTF; /* The source string is assumed to be UTF8 */
1789 UV literal_endpoint = 0;
1792 const char * const leaveit = /* set of acceptably-backslashed characters */
1794 ? "\\.^$@AGZdDwWsSbBpPXC+*?|()-nrtfeaxcz0123456789[{]} \t\n\r\f\v#"
1797 if (PL_lex_inwhat == OP_TRANS && PL_sublex_info.sub_op) {
1798 /* If we are doing a trans and we know we want UTF8 set expectation */
1799 has_utf8 = PL_sublex_info.sub_op->op_private & (OPpTRANS_FROM_UTF|OPpTRANS_TO_UTF);
1800 this_utf8 = PL_sublex_info.sub_op->op_private & (PL_lex_repl ? OPpTRANS_FROM_UTF : OPpTRANS_TO_UTF);
1804 while (s < send || dorange) {
1805 /* get transliterations out of the way (they're most literal) */
1806 if (PL_lex_inwhat == OP_TRANS) {
1807 /* expand a range A-Z to the full set of characters. AIE! */
1809 I32 i; /* current expanded character */
1810 I32 min; /* first character in range */
1811 I32 max; /* last character in range */
1814 char * const c = (char*)utf8_hop((U8*)d, -1);
1818 *c = (char)UTF_TO_NATIVE(0xff);
1819 /* mark the range as done, and continue */
1825 i = d - SvPVX_const(sv); /* remember current offset */
1826 SvGROW(sv, SvLEN(sv) + 256); /* never more than 256 chars in a range */
1827 d = SvPVX(sv) + i; /* refresh d after realloc */
1828 d -= 2; /* eat the first char and the - */
1830 min = (U8)*d; /* first char in range */
1831 max = (U8)d[1]; /* last char in range */
1835 "Invalid range \"%c-%c\" in transliteration operator",
1836 (char)min, (char)max);
1840 if (literal_endpoint == 2 &&
1841 ((isLOWER(min) && isLOWER(max)) ||
1842 (isUPPER(min) && isUPPER(max)))) {
1844 for (i = min; i <= max; i++)
1846 *d++ = NATIVE_TO_NEED(has_utf8,i);
1848 for (i = min; i <= max; i++)
1850 *d++ = NATIVE_TO_NEED(has_utf8,i);
1855 for (i = min; i <= max; i++)
1858 /* mark the range as done, and continue */
1862 literal_endpoint = 0;
1867 /* range begins (ignore - as first or last char) */
1868 else if (*s == '-' && s+1 < send && s != start) {
1870 Perl_croak(aTHX_ "Ambiguous range in transliteration operator");
1873 *d++ = (char)UTF_TO_NATIVE(0xff); /* use illegal utf8 byte--see pmtrans */
1883 literal_endpoint = 0;
1888 /* if we get here, we're not doing a transliteration */
1890 /* skip for regexp comments /(?#comment)/ and code /(?{code})/,
1891 except for the last char, which will be done separately. */
1892 else if (*s == '(' && PL_lex_inpat && s[1] == '?') {
1894 while (s+1 < send && *s != ')')
1895 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
1897 else if (s[2] == '{' /* This should match regcomp.c */
1898 || ((s[2] == 'p' || s[2] == '?') && s[3] == '{'))
1901 char *regparse = s + (s[2] == '{' ? 3 : 4);
1904 while (count && (c = *regparse)) {
1905 if (c == '\\' && regparse[1])
1913 if (*regparse != ')')
1914 regparse--; /* Leave one char for continuation. */
1915 while (s < regparse)
1916 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
1920 /* likewise skip #-initiated comments in //x patterns */
1921 else if (*s == '#' && PL_lex_inpat &&
1922 ((PMOP*)PL_lex_inpat)->op_pmflags & PMf_EXTENDED) {
1923 while (s+1 < send && *s != '\n')
1924 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
1927 /* check for embedded arrays
1928 (@foo, @::foo, @'foo, @{foo}, @$foo, @+, @-)
1930 else if (*s == '@' && s[1]
1931 && (isALNUM_lazy_if(s+1,UTF) || strchr(":'{$+-", s[1])))
1934 /* check for embedded scalars. only stop if we're sure it's a
1937 else if (*s == '$') {
1938 if (!PL_lex_inpat) /* not a regexp, so $ must be var */
1940 if (s + 1 < send && !strchr("()| \r\n\t", s[1]))
1941 break; /* in regexp, $ might be tail anchor */
1944 /* End of else if chain - OP_TRANS rejoin rest */
1947 if (*s == '\\' && s+1 < send) {
1950 /* some backslashes we leave behind */
1951 if (*leaveit && *s && strchr(leaveit, *s)) {
1952 *d++ = NATIVE_TO_NEED(has_utf8,'\\');
1953 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
1957 /* deprecate \1 in strings and substitution replacements */
1958 if (PL_lex_inwhat == OP_SUBST && !PL_lex_inpat &&
1959 isDIGIT(*s) && *s != '0' && !isDIGIT(s[1]))
1961 if (ckWARN(WARN_SYNTAX))
1962 Perl_warner(aTHX_ packWARN(WARN_SYNTAX), "\\%c better written as $%c", *s, *s);
1967 /* string-change backslash escapes */
1968 if (PL_lex_inwhat != OP_TRANS && *s && strchr("lLuUEQ", *s)) {
1973 /* if we get here, it's either a quoted -, or a digit */
1976 /* quoted - in transliterations */
1978 if (PL_lex_inwhat == OP_TRANS) {
1985 if ((isALPHA(*s) || isDIGIT(*s)) &&
1987 Perl_warner(aTHX_ packWARN(WARN_MISC),
1988 "Unrecognized escape \\%c passed through",
1990 /* default action is to copy the quoted character */
1991 goto default_action;
1994 /* \132 indicates an octal constant */
1995 case '0': case '1': case '2': case '3':
1996 case '4': case '5': case '6': case '7':
2000 uv = grok_oct(s, &len, &flags, NULL);
2003 goto NUM_ESCAPE_INSERT;
2005 /* \x24 indicates a hex constant */
2009 char* const e = strchr(s, '}');
2010 I32 flags = PERL_SCAN_ALLOW_UNDERSCORES |
2011 PERL_SCAN_DISALLOW_PREFIX;
2016 yyerror("Missing right brace on \\x{}");
2020 uv = grok_hex(s, &len, &flags, NULL);
2026 I32 flags = PERL_SCAN_DISALLOW_PREFIX;
2027 uv = grok_hex(s, &len, &flags, NULL);
2033 /* Insert oct or hex escaped character.
2034 * There will always enough room in sv since such
2035 * escapes will be longer than any UTF-8 sequence
2036 * they can end up as. */
2038 /* We need to map to chars to ASCII before doing the tests
2041 if (!UNI_IS_INVARIANT(NATIVE_TO_UNI(uv))) {
2042 if (!has_utf8 && uv > 255) {
2043 /* Might need to recode whatever we have
2044 * accumulated so far if it contains any
2047 * (Can't we keep track of that and avoid
2048 * this rescan? --jhi)
2052 for (c = (U8 *) SvPVX(sv); c < (U8 *)d; c++) {
2053 if (!NATIVE_IS_INVARIANT(*c)) {
2058 const STRLEN offset = d - SvPVX_const(sv);
2060 d = SvGROW(sv, SvLEN(sv) + hicount + 1) + offset;
2064 while (src >= (const U8 *)SvPVX_const(sv)) {
2065 if (!NATIVE_IS_INVARIANT(*src)) {
2066 const U8 ch = NATIVE_TO_ASCII(*src);
2067 *dst-- = (U8)UTF8_EIGHT_BIT_LO(ch);
2068 *dst-- = (U8)UTF8_EIGHT_BIT_HI(ch);
2078 if (has_utf8 || uv > 255) {
2079 d = (char*)uvchr_to_utf8((U8*)d, uv);
2081 if (PL_lex_inwhat == OP_TRANS &&
2082 PL_sublex_info.sub_op) {
2083 PL_sublex_info.sub_op->op_private |=
2084 (PL_lex_repl ? OPpTRANS_FROM_UTF
2097 /* \N{LATIN SMALL LETTER A} is a named character */
2101 char* e = strchr(s, '}');
2107 yyerror("Missing right brace on \\N{}");
2111 if (e > s + 2 && s[1] == 'U' && s[2] == '+') {
2113 I32 flags = PERL_SCAN_ALLOW_UNDERSCORES |
2114 PERL_SCAN_DISALLOW_PREFIX;
2117 uv = grok_hex(s, &len, &flags, NULL);
2119 goto NUM_ESCAPE_INSERT;
2121 res = newSVpvn(s + 1, e - s - 1);
2122 res = new_constant( NULL, 0, "charnames",
2123 res, NULL, "\\N{...}" );
2125 sv_utf8_upgrade(res);
2126 str = SvPV_const(res,len);
2127 #ifdef EBCDIC_NEVER_MIND
2128 /* charnames uses pack U and that has been
2129 * recently changed to do the below uni->native
2130 * mapping, so this would be redundant (and wrong,
2131 * the code point would be doubly converted).
2132 * But leave this in just in case the pack U change
2133 * gets revoked, but the semantics is still
2134 * desireable for charnames. --jhi */
2136 UV uv = utf8_to_uvchr((const U8*)str, 0);
2139 U8 tmpbuf[UTF8_MAXBYTES+1], *d;
2141 d = uvchr_to_utf8(tmpbuf, UNI_TO_NATIVE(uv));
2142 sv_setpvn(res, (char *)tmpbuf, d - tmpbuf);
2143 str = SvPV_const(res, len);
2147 if (!has_utf8 && SvUTF8(res)) {
2148 const char * const ostart = SvPVX_const(sv);
2149 SvCUR_set(sv, d - ostart);
2152 sv_utf8_upgrade(sv);
2153 /* this just broke our allocation above... */
2154 SvGROW(sv, (STRLEN)(send - start));
2155 d = SvPVX(sv) + SvCUR(sv);
2158 if (len > (STRLEN)(e - s + 4)) { /* I _guess_ 4 is \N{} --jhi */
2159 const char * const odest = SvPVX_const(sv);
2161 SvGROW(sv, (SvLEN(sv) + len - (e - s + 4)));
2162 d = SvPVX(sv) + (d - odest);
2164 Copy(str, d, len, char);
2171 yyerror("Missing braces on \\N{}");
2174 /* \c is a control character */
2183 *d++ = NATIVE_TO_NEED(has_utf8,toCTRL(c));
2186 yyerror("Missing control char name in \\c");
2190 /* printf-style backslashes, formfeeds, newlines, etc */
2192 *d++ = NATIVE_TO_NEED(has_utf8,'\b');
2195 *d++ = NATIVE_TO_NEED(has_utf8,'\n');
2198 *d++ = NATIVE_TO_NEED(has_utf8,'\r');
2201 *d++ = NATIVE_TO_NEED(has_utf8,'\f');
2204 *d++ = NATIVE_TO_NEED(has_utf8,'\t');
2207 *d++ = ASCII_TO_NEED(has_utf8,'\033');
2210 *d++ = ASCII_TO_NEED(has_utf8,'\007');
2216 } /* end if (backslash) */
2223 /* If we started with encoded form, or already know we want it
2224 and then encode the next character */
2225 if ((has_utf8 || this_utf8) && !NATIVE_IS_INVARIANT((U8)(*s))) {
2227 const UV nextuv = (this_utf8) ? utf8n_to_uvchr((U8*)s, send - s, &len, 0) : (UV) ((U8) *s);
2228 const STRLEN need = UNISKIP(NATIVE_TO_UNI(nextuv));
2231 /* encoded value larger than old, need extra space (NOTE: SvCUR() not set here) */
2232 const STRLEN off = d - SvPVX_const(sv);
2233 d = SvGROW(sv, SvLEN(sv) + (need-len)) + off;
2235 d = (char*)uvchr_to_utf8((U8*)d, nextuv);
2239 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2241 } /* while loop to process each character */
2243 /* terminate the string and set up the sv */
2245 SvCUR_set(sv, d - SvPVX_const(sv));
2246 if (SvCUR(sv) >= SvLEN(sv))
2247 Perl_croak(aTHX_ "panic: constant overflowed allocated space");
2250 if (PL_encoding && !has_utf8) {
2251 sv_recode_to_utf8(sv, PL_encoding);
2257 if (PL_lex_inwhat == OP_TRANS && PL_sublex_info.sub_op) {
2258 PL_sublex_info.sub_op->op_private |=
2259 (PL_lex_repl ? OPpTRANS_FROM_UTF : OPpTRANS_TO_UTF);
2263 /* shrink the sv if we allocated more than we used */
2264 if (SvCUR(sv) + 5 < SvLEN(sv)) {
2265 SvPV_shrink_to_cur(sv);
2268 /* return the substring (via yylval) only if we parsed anything */
2269 if (s > PL_bufptr) {
2270 if ( PL_hints & ( PL_lex_inpat ? HINT_NEW_RE : HINT_NEW_STRING ) )
2271 sv = new_constant(start, s - start, (PL_lex_inpat ? "qr" : "q"),
2273 ( PL_lex_inwhat == OP_TRANS
2275 : ( (PL_lex_inwhat == OP_SUBST && !PL_lex_inpat)
2278 yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
2285 * Returns TRUE if there's more to the expression (e.g., a subscript),
2288 * It deals with "$foo[3]" and /$foo[3]/ and /$foo[0123456789$]+/
2290 * ->[ and ->{ return TRUE
2291 * { and [ outside a pattern are always subscripts, so return TRUE
2292 * if we're outside a pattern and it's not { or [, then return FALSE
2293 * if we're in a pattern and the first char is a {
2294 * {4,5} (any digits around the comma) returns FALSE
2295 * if we're in a pattern and the first char is a [
2297 * [SOMETHING] has a funky algorithm to decide whether it's a
2298 * character class or not. It has to deal with things like
2299 * /$foo[-3]/ and /$foo[$bar]/ as well as /$foo[$\d]+/
2300 * anything else returns TRUE
2303 /* This is the one truly awful dwimmer necessary to conflate C and sed. */
2306 S_intuit_more(pTHX_ register char *s)
2309 if (PL_lex_brackets)
2311 if (*s == '-' && s[1] == '>' && (s[2] == '[' || s[2] == '{'))
2313 if (*s != '{' && *s != '[')
2318 /* In a pattern, so maybe we have {n,m}. */
2335 /* On the other hand, maybe we have a character class */
2338 if (*s == ']' || *s == '^')
2341 /* this is terrifying, and it works */
2342 int weight = 2; /* let's weigh the evidence */
2344 unsigned char un_char = 255, last_un_char;
2345 const char * const send = strchr(s,']');
2346 char tmpbuf[sizeof PL_tokenbuf * 4];
2348 if (!send) /* has to be an expression */
2351 Zero(seen,256,char);
2354 else if (isDIGIT(*s)) {
2356 if (isDIGIT(s[1]) && s[2] == ']')
2362 for (; s < send; s++) {
2363 last_un_char = un_char;
2364 un_char = (unsigned char)*s;
2369 weight -= seen[un_char] * 10;
2370 if (isALNUM_lazy_if(s+1,UTF)) {
2372 scan_ident(s, send, tmpbuf, sizeof tmpbuf, FALSE);
2373 len = (int)strlen(tmpbuf);
2374 if (len > 1 && gv_fetchpvn_flags(tmpbuf, len, 0, SVt_PV))
2379 else if (*s == '$' && s[1] &&
2380 strchr("[#!%*<>()-=",s[1])) {
2381 if (/*{*/ strchr("])} =",s[2]))
2390 if (strchr("wds]",s[1]))
2392 else if (seen['\''] || seen['"'])
2394 else if (strchr("rnftbxcav",s[1]))
2396 else if (isDIGIT(s[1])) {
2398 while (s[1] && isDIGIT(s[1]))
2408 if (strchr("aA01! ",last_un_char))
2410 if (strchr("zZ79~",s[1]))
2412 if (last_un_char == 255 && (isDIGIT(s[1]) || s[1] == '$'))
2413 weight -= 5; /* cope with negative subscript */
2416 if (!isALNUM(last_un_char)
2417 && !(last_un_char == '$' || last_un_char == '@'
2418 || last_un_char == '&')
2419 && isALPHA(*s) && s[1] && isALPHA(s[1])) {
2424 if (keyword(tmpbuf, d - tmpbuf))
2427 if (un_char == last_un_char + 1)
2429 weight -= seen[un_char];
2434 if (weight >= 0) /* probably a character class */
2444 * Does all the checking to disambiguate
2446 * between foo(bar) and bar->foo. Returns 0 if not a method, otherwise
2447 * FUNCMETH (bar->foo(args)) or METHOD (bar->foo args).
2449 * First argument is the stuff after the first token, e.g. "bar".
2451 * Not a method if bar is a filehandle.
2452 * Not a method if foo is a subroutine prototyped to take a filehandle.
2453 * Not a method if it's really "Foo $bar"
2454 * Method if it's "foo $bar"
2455 * Not a method if it's really "print foo $bar"
2456 * Method if it's really "foo package::" (interpreted as package->foo)
2457 * Not a method if bar is known to be a subroutine ("sub bar; foo bar")
2458 * Not a method if bar is a filehandle or package, but is quoted with
2463 S_intuit_method(pTHX_ char *start, GV *gv, CV *cv)
2466 char *s = start + (*start == '$');
2467 char tmpbuf[sizeof PL_tokenbuf];
2475 if (SvTYPE(gv) == SVt_PVGV && GvIO(gv))
2479 const char *proto = SvPVX_const(cv);
2490 s = scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
2491 /* start is the beginning of the possible filehandle/object,
2492 * and s is the end of it
2493 * tmpbuf is a copy of it
2496 if (*start == '$') {
2497 if (gv || PL_last_lop_op == OP_PRINT || isUPPER(*PL_tokenbuf))
2500 len = start - SvPVX(PL_linestr);
2504 start = SvPVX(PL_linestr) + len;
2508 return *s == '(' ? FUNCMETH : METHOD;
2510 if (!keyword(tmpbuf, len)) {
2511 if (len > 2 && tmpbuf[len - 2] == ':' && tmpbuf[len - 1] == ':') {
2515 soff = s - SvPVX(PL_linestr);
2519 indirgv = gv_fetchpvn_flags(tmpbuf, len, 0, SVt_PVCV);
2520 if (indirgv && GvCVu(indirgv))
2522 /* filehandle or package name makes it a method */
2523 if (!gv || GvIO(indirgv) || gv_stashpvn(tmpbuf, len, FALSE)) {
2525 soff = s - SvPVX(PL_linestr);
2528 if ((PL_bufend - s) >= 2 && *s == '=' && *(s+1) == '>')
2529 return 0; /* no assumptions -- "=>" quotes bearword */
2531 start_force(PL_curforce);
2532 NEXTVAL_NEXTTOKE.opval = (OP*)newSVOP(OP_CONST, 0,
2533 newSVpvn(tmpbuf,len));
2534 NEXTVAL_NEXTTOKE.opval->op_private = OPpCONST_BARE;
2536 curmad('X', newSVpvn(start,SvPVX(PL_linestr) + soff - start));
2541 PL_bufptr = SvPVX(PL_linestr) + soff; /* restart before space */
2543 return *s == '(' ? FUNCMETH : METHOD;
2551 * Return a string of Perl code to load the debugger. If PERL5DB
2552 * is set, it will return the contents of that, otherwise a
2553 * compile-time require of perl5db.pl.
2561 const char * const pdb = PerlEnv_getenv("PERL5DB");
2565 SETERRNO(0,SS_NORMAL);
2566 return "BEGIN { require 'perl5db.pl' }";
2572 /* Encoded script support. filter_add() effectively inserts a
2573 * 'pre-processing' function into the current source input stream.
2574 * Note that the filter function only applies to the current source file
2575 * (e.g., it will not affect files 'require'd or 'use'd by this one).
2577 * The datasv parameter (which may be NULL) can be used to pass
2578 * private data to this instance of the filter. The filter function
2579 * can recover the SV using the FILTER_DATA macro and use it to
2580 * store private buffers and state information.
2582 * The supplied datasv parameter is upgraded to a PVIO type
2583 * and the IoDIRP/IoANY field is used to store the function pointer,
2584 * and IOf_FAKE_DIRP is enabled on datasv to mark this as such.
2585 * Note that IoTOP_NAME, IoFMT_NAME, IoBOTTOM_NAME, if set for
2586 * private use must be set using malloc'd pointers.
2590 Perl_filter_add(pTHX_ filter_t funcp, SV *datasv)
2596 if (!PL_rsfp_filters)
2597 PL_rsfp_filters = newAV();
2600 SvUPGRADE(datasv, SVt_PVIO);
2601 IoANY(datasv) = FPTR2DPTR(void *, funcp); /* stash funcp into spare field */
2602 IoFLAGS(datasv) |= IOf_FAKE_DIRP;
2603 DEBUG_P(PerlIO_printf(Perl_debug_log, "filter_add func %p (%s)\n",
2604 FPTR2DPTR(void *, IoANY(datasv)),
2605 SvPV_nolen(datasv)));
2606 av_unshift(PL_rsfp_filters, 1);
2607 av_store(PL_rsfp_filters, 0, datasv) ;
2612 /* Delete most recently added instance of this filter function. */
2614 Perl_filter_del(pTHX_ filter_t funcp)
2620 DEBUG_P(PerlIO_printf(Perl_debug_log, "filter_del func %p",
2621 FPTR2DPTR(void*, funcp)));
2623 if (!PL_rsfp_filters || AvFILLp(PL_rsfp_filters)<0)
2625 /* if filter is on top of stack (usual case) just pop it off */
2626 datasv = FILTER_DATA(AvFILLp(PL_rsfp_filters));
2627 if (IoANY(datasv) == FPTR2DPTR(void *, funcp)) {
2628 IoFLAGS(datasv) &= ~IOf_FAKE_DIRP;
2629 IoANY(datasv) = (void *)NULL;
2630 sv_free(av_pop(PL_rsfp_filters));
2634 /* we need to search for the correct entry and clear it */
2635 Perl_die(aTHX_ "filter_del can only delete in reverse order (currently)");
2639 /* Invoke the idxth filter function for the current rsfp. */
2640 /* maxlen 0 = read one text line */
2642 Perl_filter_read(pTHX_ int idx, SV *buf_sv, int maxlen)
2647 /* This API is bad. It should have been using unsigned int for maxlen.
2648 Not sure if we want to change the API, but if not we should sanity
2649 check the value here. */
2650 const unsigned int correct_length
2659 if (!PL_rsfp_filters)
2661 if (idx > AvFILLp(PL_rsfp_filters)) { /* Any more filters? */
2662 /* Provide a default input filter to make life easy. */
2663 /* Note that we append to the line. This is handy. */
2664 DEBUG_P(PerlIO_printf(Perl_debug_log,
2665 "filter_read %d: from rsfp\n", idx));
2666 if (correct_length) {
2669 const int old_len = SvCUR(buf_sv);
2671 /* ensure buf_sv is large enough */
2672 SvGROW(buf_sv, (STRLEN)(old_len + correct_length)) ;
2673 if ((len = PerlIO_read(PL_rsfp, SvPVX(buf_sv) + old_len,
2674 correct_length)) <= 0) {
2675 if (PerlIO_error(PL_rsfp))
2676 return -1; /* error */
2678 return 0 ; /* end of file */
2680 SvCUR_set(buf_sv, old_len + len) ;
2683 if (sv_gets(buf_sv, PL_rsfp, SvCUR(buf_sv)) == NULL) {
2684 if (PerlIO_error(PL_rsfp))
2685 return -1; /* error */
2687 return 0 ; /* end of file */
2690 return SvCUR(buf_sv);
2692 /* Skip this filter slot if filter has been deleted */
2693 if ( (datasv = FILTER_DATA(idx)) == &PL_sv_undef) {
2694 DEBUG_P(PerlIO_printf(Perl_debug_log,
2695 "filter_read %d: skipped (filter deleted)\n",
2697 return FILTER_READ(idx+1, buf_sv, correct_length); /* recurse */
2699 /* Get function pointer hidden within datasv */
2700 funcp = DPTR2FPTR(filter_t, IoANY(datasv));
2701 DEBUG_P(PerlIO_printf(Perl_debug_log,
2702 "filter_read %d: via function %p (%s)\n",
2703 idx, (void*)datasv, SvPV_nolen_const(datasv)));
2704 /* Call function. The function is expected to */
2705 /* call "FILTER_READ(idx+1, buf_sv)" first. */
2706 /* Return: <0:error, =0:eof, >0:not eof */
2707 return (*funcp)(aTHX_ idx, buf_sv, correct_length);
2711 S_filter_gets(pTHX_ register SV *sv, register PerlIO *fp, STRLEN append)
2714 #ifdef PERL_CR_FILTER
2715 if (!PL_rsfp_filters) {
2716 filter_add(S_cr_textfilter,NULL);
2719 if (PL_rsfp_filters) {
2721 SvCUR_set(sv, 0); /* start with empty line */
2722 if (FILTER_READ(0, sv, 0) > 0)
2723 return ( SvPVX(sv) ) ;
2728 return (sv_gets(sv, fp, append));
2732 S_find_in_my_stash(pTHX_ const char *pkgname, I32 len)
2737 if (len == 11 && *pkgname == '_' && strEQ(pkgname, "__PACKAGE__"))
2741 (pkgname[len - 2] == ':' && pkgname[len - 1] == ':') &&
2742 (gv = gv_fetchpvn_flags(pkgname, len, 0, SVt_PVHV)))
2744 return GvHV(gv); /* Foo:: */
2747 /* use constant CLASS => 'MyClass' */
2748 gv = gv_fetchpvn_flags(pkgname, len, 0, SVt_PVCV);
2749 if (gv && GvCV(gv)) {
2750 SV * const sv = cv_const_sv(GvCV(gv));
2752 pkgname = SvPV_nolen_const(sv);
2755 return gv_stashpv(pkgname, FALSE);
2761 * The intent of this yylex wrapper is to minimize the changes to the
2762 * tokener when we aren't interested in collecting madprops. It remains
2763 * to be seen how successful this strategy will be...
2770 char *s = PL_bufptr;
2772 /* make sure PL_thiswhite is initialized */
2776 /* just do what yylex would do on pending identifier; leave PL_thiswhite alone */
2777 if (PL_pending_ident)
2778 return S_pending_ident(aTHX);
2780 /* previous token ate up our whitespace? */
2781 if (!PL_lasttoke && PL_nextwhite) {
2782 PL_thiswhite = PL_nextwhite;
2786 /* isolate the token, and figure out where it is without whitespace */
2787 PL_realtokenstart = -1;
2791 assert(PL_curforce < 0);
2793 if (!PL_thismad || PL_thismad->mad_key == '^') { /* not forced already? */
2794 if (!PL_thistoken) {
2795 if (PL_realtokenstart < 0 || !CopLINE(PL_curcop))
2796 PL_thistoken = newSVpvn("",0);
2798 char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
2799 PL_thistoken = newSVpvn(tstart, s - tstart);
2802 if (PL_thismad) /* install head */
2803 CURMAD('X', PL_thistoken);
2806 /* last whitespace of a sublex? */
2807 if (optype == ')' && PL_endwhite) {
2808 CURMAD('X', PL_endwhite);
2813 /* if no whitespace and we're at EOF, bail. Otherwise fake EOF below. */
2814 if (!PL_thiswhite && !PL_endwhite && !optype) {
2815 sv_free(PL_thistoken);
2820 /* put off final whitespace till peg */
2821 if (optype == ';' && !PL_rsfp) {
2822 PL_nextwhite = PL_thiswhite;
2825 else if (PL_thisopen) {
2826 CURMAD('q', PL_thisopen);
2828 sv_free(PL_thistoken);
2832 /* Store actual token text as madprop X */
2833 CURMAD('X', PL_thistoken);
2837 /* add preceding whitespace as madprop _ */
2838 CURMAD('_', PL_thiswhite);
2842 /* add quoted material as madprop = */
2843 CURMAD('=', PL_thisstuff);
2847 /* add terminating quote as madprop Q */
2848 CURMAD('Q', PL_thisclose);
2852 /* special processing based on optype */
2856 /* opval doesn't need a TOKEN since it can already store mp */
2867 append_madprops(PL_thismad, yylval.opval, 0);
2875 addmad(newMADsv('p', PL_endwhite), &PL_thismad, 0);
2884 /* remember any fake bracket that lexer is about to discard */
2885 if (PL_lex_brackets == 1 &&
2886 ((expectation)PL_lex_brackstack[0] & XFAKEBRACK))
2889 while (s < PL_bufend && (*s == ' ' || *s == '\t'))
2892 PL_thiswhite = newSVpvn(PL_bufptr, ++s - PL_bufptr);
2893 addmad(newMADsv('#', PL_thiswhite), &PL_thismad, 0);
2896 break; /* don't bother looking for trailing comment */
2905 /* attach a trailing comment to its statement instead of next token */
2909 if (PL_bufptr > PL_oldbufptr && PL_bufptr[-1] == optype) {
2911 while (s < PL_bufend && (*s == ' ' || *s == '\t'))
2913 if (*s == '\n' || *s == '#') {
2914 while (s < PL_bufend && *s != '\n')
2918 PL_thiswhite = newSVpvn(PL_bufptr, s - PL_bufptr);
2919 addmad(newMADsv('#', PL_thiswhite), &PL_thismad, 0);
2936 /* Create new token struct. Note: opvals return early above. */
2937 yylval.tkval = newTOKEN(optype, yylval, PL_thismad);
2944 S_tokenize_use(pTHX_ int is_use, char *s) {
2946 if (PL_expect != XSTATE)
2947 yyerror(Perl_form(aTHX_ "\"%s\" not allowed in expression",
2948 is_use ? "use" : "no"));
2950 if (isDIGIT(*s) || (*s == 'v' && isDIGIT(s[1]))) {
2951 s = force_version(s, TRUE);
2952 if (*s == ';' || (s = SKIPSPACE1(s), *s == ';')) {
2953 start_force(PL_curforce);
2954 NEXTVAL_NEXTTOKE.opval = NULL;
2957 else if (*s == 'v') {
2958 s = force_word(s,WORD,FALSE,TRUE,FALSE);
2959 s = force_version(s, FALSE);
2963 s = force_word(s,WORD,FALSE,TRUE,FALSE);
2964 s = force_version(s, FALSE);
2966 yylval.ival = is_use;
2970 static const char* const exp_name[] =
2971 { "OPERATOR", "TERM", "REF", "STATE", "BLOCK", "ATTRBLOCK",
2972 "ATTRTERM", "TERMBLOCK", "TERMORDORDOR"
2979 Works out what to call the token just pulled out of the input
2980 stream. The yacc parser takes care of taking the ops we return and
2981 stitching them into a tree.
2987 if read an identifier
2988 if we're in a my declaration
2989 croak if they tried to say my($foo::bar)
2990 build the ops for a my() declaration
2991 if it's an access to a my() variable
2992 are we in a sort block?
2993 croak if my($a); $a <=> $b
2994 build ops for access to a my() variable
2995 if in a dq string, and they've said @foo and we can't find @foo
2997 build ops for a bareword
2998 if we already built the token before, use it.
3003 #pragma segment Perl_yylex
3009 register char *s = PL_bufptr;
3015 SV* tmp = newSVpvs("");
3016 PerlIO_printf(Perl_debug_log, "### %"IVdf":LEX_%s/X%s %s\n",
3017 (IV)CopLINE(PL_curcop),
3018 lex_state_names[PL_lex_state],
3019 exp_name[PL_expect],
3020 pv_display(tmp, s, strlen(s), 0, 60));
3023 /* check if there's an identifier for us to look at */
3024 if (PL_pending_ident)
3025 return REPORT(S_pending_ident(aTHX));
3027 /* no identifier pending identification */
3029 switch (PL_lex_state) {
3031 case LEX_NORMAL: /* Some compilers will produce faster */
3032 case LEX_INTERPNORMAL: /* code if we comment these out. */
3036 /* when we've already built the next token, just pull it out of the queue */
3040 yylval = PL_nexttoke[PL_lasttoke].next_val;
3042 PL_thismad = PL_nexttoke[PL_lasttoke].next_mad;
3043 PL_nexttoke[PL_lasttoke].next_mad = 0;
3044 if (PL_thismad && PL_thismad->mad_key == '_') {
3045 PL_thiswhite = (SV*)PL_thismad->mad_val;
3046 PL_thismad->mad_val = 0;
3047 mad_free(PL_thismad);
3052 PL_lex_state = PL_lex_defer;
3053 PL_expect = PL_lex_expect;
3054 PL_lex_defer = LEX_NORMAL;
3055 if (!PL_nexttoke[PL_lasttoke].next_type)
3060 yylval = PL_nextval[PL_nexttoke];
3062 PL_lex_state = PL_lex_defer;
3063 PL_expect = PL_lex_expect;
3064 PL_lex_defer = LEX_NORMAL;
3068 /* FIXME - can these be merged? */
3069 return(PL_nexttoke[PL_lasttoke].next_type);
3071 return REPORT(PL_nexttype[PL_nexttoke]);
3074 /* interpolated case modifiers like \L \U, including \Q and \E.
3075 when we get here, PL_bufptr is at the \
3077 case LEX_INTERPCASEMOD:
3079 if (PL_bufptr != PL_bufend && *PL_bufptr != '\\')
3080 Perl_croak(aTHX_ "panic: INTERPCASEMOD");
3082 /* handle \E or end of string */
3083 if (PL_bufptr == PL_bufend || PL_bufptr[1] == 'E') {
3085 if (PL_lex_casemods) {
3086 const char oldmod = PL_lex_casestack[--PL_lex_casemods];
3087 PL_lex_casestack[PL_lex_casemods] = '\0';
3089 if (PL_bufptr != PL_bufend
3090 && (oldmod == 'L' || oldmod == 'U' || oldmod == 'Q')) {
3092 PL_lex_state = LEX_INTERPCONCAT;
3095 PL_thistoken = newSVpvn("\\E",2);
3101 while (PL_bufptr != PL_bufend &&
3102 PL_bufptr[0] == '\\' && PL_bufptr[1] == 'E') {
3104 PL_thiswhite = newSVpvn("",0);
3105 sv_catpvn(PL_thiswhite, PL_bufptr, 2);
3109 if (PL_bufptr != PL_bufend)
3112 PL_lex_state = LEX_INTERPCONCAT;
3116 DEBUG_T({ PerlIO_printf(Perl_debug_log,
3117 "### Saw case modifier\n"); });
3119 if (s[1] == '\\' && s[2] == 'E') {
3122 PL_thiswhite = newSVpvn("",0);
3123 sv_catpvn(PL_thiswhite, PL_bufptr, 4);
3126 PL_lex_state = LEX_INTERPCONCAT;
3131 if (!PL_madskills) /* when just compiling don't need correct */
3132 if (strnEQ(s, "L\\u", 3) || strnEQ(s, "U\\l", 3))
3133 tmp = *s, *s = s[2], s[2] = (char)tmp; /* misordered... */
3134 if ((*s == 'L' || *s == 'U') &&
3135 (strchr(PL_lex_casestack, 'L') || strchr(PL_lex_casestack, 'U'))) {
3136 PL_lex_casestack[--PL_lex_casemods] = '\0';
3139 if (PL_lex_casemods > 10)
3140 Renew(PL_lex_casestack, PL_lex_casemods + 2, char);
3141 PL_lex_casestack[PL_lex_casemods++] = *s;
3142 PL_lex_casestack[PL_lex_casemods] = '\0';
3143 PL_lex_state = LEX_INTERPCONCAT;
3144 start_force(PL_curforce);
3145 NEXTVAL_NEXTTOKE.ival = 0;
3147 start_force(PL_curforce);
3149 NEXTVAL_NEXTTOKE.ival = OP_LCFIRST;
3151 NEXTVAL_NEXTTOKE.ival = OP_UCFIRST;
3153 NEXTVAL_NEXTTOKE.ival = OP_LC;
3155 NEXTVAL_NEXTTOKE.ival = OP_UC;
3157 NEXTVAL_NEXTTOKE.ival = OP_QUOTEMETA;
3159 Perl_croak(aTHX_ "panic: yylex");
3161 SV* const tmpsv = newSVpvn("",0);
3162 Perl_sv_catpvf(aTHX_ tmpsv, "\\%c", *s);
3168 if (PL_lex_starts) {
3174 sv_free(PL_thistoken);
3175 PL_thistoken = newSVpvn("",0);
3178 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
3179 if (PL_lex_casemods == 1 && PL_lex_inpat)
3188 case LEX_INTERPPUSH:
3189 return REPORT(sublex_push());
3191 case LEX_INTERPSTART:
3192 if (PL_bufptr == PL_bufend)
3193 return REPORT(sublex_done());
3194 DEBUG_T({ PerlIO_printf(Perl_debug_log,
3195 "### Interpolated variable\n"); });
3197 PL_lex_dojoin = (*PL_bufptr == '@');
3198 PL_lex_state = LEX_INTERPNORMAL;
3199 if (PL_lex_dojoin) {
3200 start_force(PL_curforce);
3201 NEXTVAL_NEXTTOKE.ival = 0;
3203 start_force(PL_curforce);
3204 force_ident("\"", '$');
3205 start_force(PL_curforce);
3206 NEXTVAL_NEXTTOKE.ival = 0;
3208 start_force(PL_curforce);
3209 NEXTVAL_NEXTTOKE.ival = 0;
3211 start_force(PL_curforce);
3212 NEXTVAL_NEXTTOKE.ival = OP_JOIN; /* emulate join($", ...) */
3215 if (PL_lex_starts++) {
3220 sv_free(PL_thistoken);
3221 PL_thistoken = newSVpvn("",0);
3224 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
3225 if (!PL_lex_casemods && PL_lex_inpat)
3232 case LEX_INTERPENDMAYBE:
3233 if (intuit_more(PL_bufptr)) {
3234 PL_lex_state = LEX_INTERPNORMAL; /* false alarm, more expr */
3240 if (PL_lex_dojoin) {
3241 PL_lex_dojoin = FALSE;
3242 PL_lex_state = LEX_INTERPCONCAT;
3246 sv_free(PL_thistoken);
3247 PL_thistoken = newSVpvn("",0);
3252 if (PL_lex_inwhat == OP_SUBST && PL_linestr == PL_lex_repl
3253 && SvEVALED(PL_lex_repl))
3255 if (PL_bufptr != PL_bufend)
3256 Perl_croak(aTHX_ "Bad evalled substitution pattern");
3260 case LEX_INTERPCONCAT:
3262 if (PL_lex_brackets)
3263 Perl_croak(aTHX_ "panic: INTERPCONCAT");
3265 if (PL_bufptr == PL_bufend)
3266 return REPORT(sublex_done());
3268 if (SvIVX(PL_linestr) == '\'') {
3269 SV *sv = newSVsv(PL_linestr);
3272 else if ( PL_hints & HINT_NEW_RE )
3273 sv = new_constant(NULL, 0, "qr", sv, sv, "q");
3274 yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
3278 s = scan_const(PL_bufptr);
3280 PL_lex_state = LEX_INTERPCASEMOD;
3282 PL_lex_state = LEX_INTERPSTART;
3285 if (s != PL_bufptr) {
3286 start_force(PL_curforce);
3288 curmad('X', newSVpvn(PL_bufptr,s-PL_bufptr));
3290 NEXTVAL_NEXTTOKE = yylval;
3293 if (PL_lex_starts++) {
3297 sv_free(PL_thistoken);
3298 PL_thistoken = newSVpvn("",0);
3301 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
3302 if (!PL_lex_casemods && PL_lex_inpat)
3315 PL_lex_state = LEX_NORMAL;
3316 s = scan_formline(PL_bufptr);
3317 if (!PL_lex_formbrack)
3323 PL_oldoldbufptr = PL_oldbufptr;
3329 sv_free(PL_thistoken);
3332 PL_realtokenstart = s - SvPVX(PL_linestr); /* assume but undo on ws */
3336 if (isIDFIRST_lazy_if(s,UTF))
3338 Perl_croak(aTHX_ "Unrecognized character \\x%02X", *s & 255);
3341 goto fake_eof; /* emulate EOF on ^D or ^Z */
3350 if (PL_lex_brackets) {
3351 yyerror(PL_lex_formbrack
3352 ? "Format not terminated"
3353 : "Missing right curly or square bracket");
3355 DEBUG_T( { PerlIO_printf(Perl_debug_log,
3356 "### Tokener got EOF\n");
3360 if (s++ < PL_bufend)
3361 goto retry; /* ignore stray nulls */
3364 if (!PL_in_eval && !PL_preambled) {
3365 PL_preambled = TRUE;
3370 sv_setpv(PL_linestr,incl_perldb());
3371 if (SvCUR(PL_linestr))
3372 sv_catpvs(PL_linestr,";");
3374 while(AvFILLp(PL_preambleav) >= 0) {
3375 SV *tmpsv = av_shift(PL_preambleav);
3376 sv_catsv(PL_linestr, tmpsv);
3377 sv_catpvs(PL_linestr, ";");
3380 sv_free((SV*)PL_preambleav);
3381 PL_preambleav = NULL;
3383 if (PL_minus_n || PL_minus_p) {
3384 sv_catpvs(PL_linestr, "LINE: while (<>) {");
3386 sv_catpvs(PL_linestr,"chomp;");
3389 if ((*PL_splitstr == '/' || *PL_splitstr == '\''
3390 || *PL_splitstr == '"')
3391 && strchr(PL_splitstr + 1, *PL_splitstr))
3392 Perl_sv_catpvf(aTHX_ PL_linestr, "our @F=split(%s);", PL_splitstr);
3394 /* "q\0${splitstr}\0" is legal perl. Yes, even NUL
3395 bytes can be used as quoting characters. :-) */
3396 const char *splits = PL_splitstr;
3397 sv_catpvs(PL_linestr, "our @F=split(q\0");
3400 if (*splits == '\\')
3401 sv_catpvn(PL_linestr, splits, 1);
3402 sv_catpvn(PL_linestr, splits, 1);
3403 } while (*splits++);
3404 /* This loop will embed the trailing NUL of
3405 PL_linestr as the last thing it does before
3407 sv_catpvs(PL_linestr, ");");
3411 sv_catpvs(PL_linestr,"our @F=split(' ');");
3415 sv_catpvs(PL_linestr,"use feature ':5.10';");
3416 sv_catpvs(PL_linestr, "\n");
3417 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3418 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3419 PL_last_lop = PL_last_uni = NULL;
3420 if (PERLDB_LINE && PL_curstash != PL_debstash) {
3421 SV * const sv = newSV(0);
3423 sv_upgrade(sv, SVt_PVMG);
3424 sv_setsv(sv,PL_linestr);
3427 av_store(CopFILEAVx(PL_curcop),(I32)CopLINE(PL_curcop),sv);
3432 bof = PL_rsfp ? TRUE : FALSE;
3433 if ((s = filter_gets(PL_linestr, PL_rsfp, 0)) == NULL) {
3436 PL_realtokenstart = -1;
3439 if (PL_preprocess && !PL_in_eval)
3440 (void)PerlProc_pclose(PL_rsfp);
3441 else if ((PerlIO *)PL_rsfp == PerlIO_stdin())
3442 PerlIO_clearerr(PL_rsfp);
3444 (void)PerlIO_close(PL_rsfp);
3446 PL_doextract = FALSE;
3448 if (!PL_in_eval && (PL_minus_n || PL_minus_p)) {
3453 sv_setpv(PL_linestr,PL_minus_p
3454 ? ";}continue{print;}" : ";}");
3455 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3456 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3457 PL_last_lop = PL_last_uni = NULL;
3458 PL_minus_n = PL_minus_p = 0;
3461 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3462 PL_last_lop = PL_last_uni = NULL;
3463 sv_setpvn(PL_linestr,"",0);
3464 TOKEN(';'); /* not infinite loop because rsfp is NULL now */
3466 /* If it looks like the start of a BOM or raw UTF-16,
3467 * check if it in fact is. */
3473 #ifdef PERLIO_IS_STDIO
3474 # ifdef __GNU_LIBRARY__
3475 # if __GNU_LIBRARY__ == 1 /* Linux glibc5 */
3476 # define FTELL_FOR_PIPE_IS_BROKEN
3480 # if __GLIBC__ == 1 /* maybe some glibc5 release had it like this? */
3481 # define FTELL_FOR_PIPE_IS_BROKEN
3486 #ifdef FTELL_FOR_PIPE_IS_BROKEN
3487 /* This loses the possibility to detect the bof
3488 * situation on perl -P when the libc5 is being used.
3489 * Workaround? Maybe attach some extra state to PL_rsfp?
3492 bof = PerlIO_tell(PL_rsfp) == SvCUR(PL_linestr);
3494 bof = PerlIO_tell(PL_rsfp) == (Off_t)SvCUR(PL_linestr);
3497 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3498 s = swallow_bom((U8*)s);
3502 /* Incest with pod. */
3505 sv_catsv(PL_thiswhite, PL_linestr);
3507 if (*s == '=' && strnEQ(s, "=cut", 4)) {
3508 sv_setpvn(PL_linestr, "", 0);
3509 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3510 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3511 PL_last_lop = PL_last_uni = NULL;
3512 PL_doextract = FALSE;
3516 } while (PL_doextract);
3517 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = s;
3518 if (PERLDB_LINE && PL_curstash != PL_debstash) {
3519 SV * const sv = newSV(0);
3521 sv_upgrade(sv, SVt_PVMG);
3522 sv_setsv(sv,PL_linestr);
3525 av_store(CopFILEAVx(PL_curcop),(I32)CopLINE(PL_curcop),sv);
3527 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3528 PL_last_lop = PL_last_uni = NULL;
3529 if (CopLINE(PL_curcop) == 1) {
3530 while (s < PL_bufend && isSPACE(*s))
3532 if (*s == ':' && s[1] != ':') /* for csh execing sh scripts */
3536 PL_thiswhite = newSVpvn(PL_linestart, s - PL_linestart);
3540 if (*s == '#' && *(s+1) == '!')
3542 #ifdef ALTERNATE_SHEBANG
3544 static char const as[] = ALTERNATE_SHEBANG;
3545 if (*s == as[0] && strnEQ(s, as, sizeof(as) - 1))
3546 d = s + (sizeof(as) - 1);
3548 #endif /* ALTERNATE_SHEBANG */
3557 while (*d && !isSPACE(*d))
3561 #ifdef ARG_ZERO_IS_SCRIPT
3562 if (ipathend > ipath) {
3564 * HP-UX (at least) sets argv[0] to the script name,
3565 * which makes $^X incorrect. And Digital UNIX and Linux,
3566 * at least, set argv[0] to the basename of the Perl
3567 * interpreter. So, having found "#!", we'll set it right.
3569 SV * const x = GvSV(gv_fetchpvs("\030", GV_ADD|GV_NOTQUAL,
3571 assert(SvPOK(x) || SvGMAGICAL(x));
3572 if (sv_eq(x, CopFILESV(PL_curcop))) {
3573 sv_setpvn(x, ipath, ipathend - ipath);
3579 const char *bstart = SvPV_const(CopFILESV(PL_curcop),blen);
3580 const char * const lstart = SvPV_const(x,llen);
3582 bstart += blen - llen;
3583 if (strnEQ(bstart, lstart, llen) && bstart[-1] == '/') {
3584 sv_setpvn(x, ipath, ipathend - ipath);
3589 TAINT_NOT; /* $^X is always tainted, but that's OK */
3591 #endif /* ARG_ZERO_IS_SCRIPT */
3596 d = instr(s,"perl -");
3598 d = instr(s,"perl");
3600 /* avoid getting into infinite loops when shebang
3601 * line contains "Perl" rather than "perl" */
3603 for (d = ipathend-4; d >= ipath; --d) {
3604 if ((*d == 'p' || *d == 'P')
3605 && !ibcmp(d, "perl", 4))
3615 #ifdef ALTERNATE_SHEBANG
3617 * If the ALTERNATE_SHEBANG on this system starts with a
3618 * character that can be part of a Perl expression, then if
3619 * we see it but not "perl", we're probably looking at the
3620 * start of Perl code, not a request to hand off to some
3621 * other interpreter. Similarly, if "perl" is there, but
3622 * not in the first 'word' of the line, we assume the line
3623 * contains the start of the Perl program.
3625 if (d && *s != '#') {
3626 const char *c = ipath;
3627 while (*c && !strchr("; \t\r\n\f\v#", *c))
3630 d = NULL; /* "perl" not in first word; ignore */
3632 *s = '#'; /* Don't try to parse shebang line */
3634 #endif /* ALTERNATE_SHEBANG */
3635 #ifndef MACOS_TRADITIONAL
3640 !instr(s,"indir") &&
3641 instr(PL_origargv[0],"perl"))
3648 while (s < PL_bufend && isSPACE(*s))
3650 if (s < PL_bufend) {
3651 Newxz(newargv,PL_origargc+3,char*);
3653 while (s < PL_bufend && !isSPACE(*s))
3656 Copy(PL_origargv+1, newargv+2, PL_origargc+1, char*);
3659 newargv = PL_origargv;
3662 PerlProc_execv(ipath, EXEC_ARGV_CAST(newargv));
3664 Perl_croak(aTHX_ "Can't exec %s", ipath);
3668 while (*d && !isSPACE(*d))
3670 while (SPACE_OR_TAB(*d))
3674 const bool switches_done = PL_doswitches;
3675 const U32 oldpdb = PL_perldb;
3676 const bool oldn = PL_minus_n;
3677 const bool oldp = PL_minus_p;
3680 if (*d == 'M' || *d == 'm' || *d == 'C') {
3681 const char * const m = d;
3682 while (*d && !isSPACE(*d))
3684 Perl_croak(aTHX_ "Too late for \"-%.*s\" option",
3687 d = moreswitches(d);
3689 if (PL_doswitches && !switches_done) {
3690 int argc = PL_origargc;
3691 char **argv = PL_origargv;
3694 } while (argc && argv[0][0] == '-' && argv[0][1]);
3695 init_argv_symbols(argc,argv);
3697 if ((PERLDB_LINE && !oldpdb) ||
3698 ((PL_minus_n || PL_minus_p) && !(oldn || oldp)))
3699 /* if we have already added "LINE: while (<>) {",
3700 we must not do it again */
3702 sv_setpvn(PL_linestr, "", 0);
3703 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3704 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3705 PL_last_lop = PL_last_uni = NULL;
3706 PL_preambled = FALSE;
3708 (void)gv_fetchfile(PL_origfilename);
3715 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
3717 PL_lex_state = LEX_FORMLINE;
3722 #ifdef PERL_STRICT_CR
3723 Perl_warn(aTHX_ "Illegal character \\%03o (carriage return)", '\r');
3725 "\t(Maybe you didn't strip carriage returns after a network transfer?)\n");
3727 case ' ': case '\t': case '\f': case 013:
3728 #ifdef MACOS_TRADITIONAL
3732 PL_realtokenstart = -1;
3741 PL_realtokenstart = -1;
3745 if (PL_lex_state != LEX_NORMAL || (PL_in_eval && !PL_rsfp)) {
3746 if (*s == '#' && s == PL_linestart && PL_in_eval && !PL_rsfp) {
3747 /* handle eval qq[#line 1 "foo"\n ...] */
3748 CopLINE_dec(PL_curcop);
3751 if (PL_madskills && !PL_lex_formbrack && !PL_in_eval) {
3753 if (!PL_in_eval || PL_rsfp)
3758 while (d < PL_bufend && *d != '\n')
3762 else if (d > PL_bufend) /* Found by Ilya: feed random input to Perl. */
3763 Perl_croak(aTHX_ "panic: input overflow");
3766 PL_thiswhite = newSVpvn(s, d - s);
3771 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
3773 PL_lex_state = LEX_FORMLINE;
3779 if (PL_madskills && CopLINE(PL_curcop) >= 1 && !PL_lex_formbrack) {
3780 if (CopLINE(PL_curcop) == 1 && s[0] == '#' && s[1] == '!') {
3783 TOKEN(PEG); /* make sure any #! line is accessible */
3788 /* if (PL_madskills && PL_lex_formbrack) { */
3790 while (d < PL_bufend && *d != '\n')
3794 else if (d > PL_bufend) /* Found by Ilya: feed random input to Perl. */
3795 Perl_croak(aTHX_ "panic: input overflow");
3796 if (PL_madskills && CopLINE(PL_curcop) >= 1) {
3798 PL_thiswhite = newSVpvn("",0);
3799 if (CopLINE(PL_curcop) == 1) {
3800 sv_setpvn(PL_thiswhite, "", 0);
3803 sv_catpvn(PL_thiswhite, s, d - s);
3817 if (s[1] && isALPHA(s[1]) && !isALNUM(s[2])) {
3825 while (s < PL_bufend && SPACE_OR_TAB(*s))
3828 if (strnEQ(s,"=>",2)) {
3829 s = force_word(PL_bufptr,WORD,FALSE,FALSE,FALSE);
3830 DEBUG_T( { printbuf("### Saw unary minus before =>, forcing word %s\n", s); } );
3831 OPERATOR('-'); /* unary minus */
3833 PL_last_uni = PL_oldbufptr;
3835 case 'r': ftst = OP_FTEREAD; break;
3836 case 'w': ftst = OP_FTEWRITE; break;
3837 case 'x': ftst = OP_FTEEXEC; break;
3838 case 'o': ftst = OP_FTEOWNED; break;
3839 case 'R': ftst = OP_FTRREAD; break;
3840 case 'W': ftst = OP_FTRWRITE; break;
3841 case 'X': ftst = OP_FTREXEC; break;
3842 case 'O': ftst = OP_FTROWNED; break;
3843 case 'e': ftst = OP_FTIS; break;
3844 case 'z': ftst = OP_FTZERO; break;
3845 case 's': ftst = OP_FTSIZE; break;
3846 case 'f': ftst = OP_FTFILE; break;
3847 case 'd': ftst = OP_FTDIR; break;
3848 case 'l': ftst = OP_FTLINK; break;
3849 case 'p': ftst = OP_FTPIPE; break;
3850 case 'S': ftst = OP_FTSOCK; break;
3851 case 'u': ftst = OP_FTSUID; break;
3852 case 'g': ftst = OP_FTSGID; break;
3853 case 'k': ftst = OP_FTSVTX; break;
3854 case 'b': ftst = OP_FTBLK; break;
3855 case 'c': ftst = OP_FTCHR; break;
3856 case 't': ftst = OP_FTTTY; break;
3857 case 'T': ftst = OP_FTTEXT; break;
3858 case 'B': ftst = OP_FTBINARY; break;
3859 case 'M': case 'A': case 'C':
3860 gv_fetchpvs("\024", GV_ADD|GV_NOTQUAL, SVt_PV);
3862 case 'M': ftst = OP_FTMTIME; break;
3863 case 'A': ftst = OP_FTATIME; break;
3864 case 'C': ftst = OP_FTCTIME; break;
3872 PL_last_lop_op = (OPCODE)ftst;
3873 DEBUG_T( { PerlIO_printf(Perl_debug_log,
3874 "### Saw file test %c\n", (int)tmp);
3879 /* Assume it was a minus followed by a one-letter named
3880 * subroutine call (or a -bareword), then. */
3881 DEBUG_T( { PerlIO_printf(Perl_debug_log,
3882 "### '-%c' looked like a file test but was not\n",
3889 const char tmp = *s++;
3892 if (PL_expect == XOPERATOR)
3897 else if (*s == '>') {
3900 if (isIDFIRST_lazy_if(s,UTF)) {
3901 s = force_word(s,METHOD,FALSE,TRUE,FALSE);
3909 if (PL_expect == XOPERATOR)
3912 if (isSPACE(*s) || !isSPACE(*PL_bufptr))
3914 OPERATOR('-'); /* unary minus */
3920 const char tmp = *s++;
3923 if (PL_expect == XOPERATOR)
3928 if (PL_expect == XOPERATOR)
3931 if (isSPACE(*s) || !isSPACE(*PL_bufptr))
3938 if (PL_expect != XOPERATOR) {
3939 s = scan_ident(s, PL_bufend, PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
3940 PL_expect = XOPERATOR;
3941 force_ident(PL_tokenbuf, '*');
3954 if (PL_expect == XOPERATOR) {
3958 PL_tokenbuf[0] = '%';
3959 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1, TRUE);
3960 if (!PL_tokenbuf[1]) {
3963 PL_pending_ident = '%';
3974 && (PL_expect == XOPERATOR || PL_expect == XTERMORDORDOR)
3975 && FEATURE_IS_ENABLED("~~"))
3982 const char tmp = *s++;
3988 goto just_a_word_zero_gv;
3991 switch (PL_expect) {
3997 if (!PL_in_my || PL_lex_state != LEX_NORMAL)
3999 PL_bufptr = s; /* update in case we back off */
4005 PL_expect = XTERMBLOCK;
4008 stuffstart = s - SvPVX(PL_linestr) - 1;
4012 while (isIDFIRST_lazy_if(s,UTF)) {
4015 d = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
4016 if (isLOWER(*s) && (tmp = keyword(PL_tokenbuf, len))) {
4017 if (tmp < 0) tmp = -tmp;
4032 sv = newSVpvn(s, len);
4034 d = scan_str(d,TRUE,TRUE);
4036 /* MUST advance bufptr here to avoid bogus
4037 "at end of line" context messages from yyerror().
4039 PL_bufptr = s + len;
4040 yyerror("Unterminated attribute parameter in attribute list");
4044 return REPORT(0); /* EOF indicator */
4048 sv_catsv(sv, PL_lex_stuff);
4049 attrs = append_elem(OP_LIST, attrs,
4050 newSVOP(OP_CONST, 0, sv));
4051 SvREFCNT_dec(PL_lex_stuff);
4052 PL_lex_stuff = NULL;
4055 if (len == 6 && strnEQ(SvPVX(sv), "unique", len)) {
4057 if (PL_in_my == KEY_our) {
4059 GvUNIQUE_on(cGVOPx_gv(yylval.opval));
4061 /* skip to avoid loading attributes.pm */
4063 deprecate(":unique");
4066 Perl_croak(aTHX_ "The 'unique' attribute may only be applied to 'our' variables");
4069 /* NOTE: any CV attrs applied here need to be part of
4070 the CVf_BUILTIN_ATTRS define in cv.h! */
4071 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "lvalue", len)) {
4073 CvLVALUE_on(PL_compcv);
4075 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "locked", len)) {
4077 CvLOCKED_on(PL_compcv);
4079 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "method", len)) {
4081 CvMETHOD_on(PL_compcv);
4083 else if (!PL_in_my && len == 9 && strnEQ(SvPVX(sv), "assertion", len)) {
4085 CvASSERTION_on(PL_compcv);
4087 /* After we've set the flags, it could be argued that
4088 we don't need to do the attributes.pm-based setting
4089 process, and shouldn't bother appending recognized
4090 flags. To experiment with that, uncomment the
4091 following "else". (Note that's already been
4092 uncommented. That keeps the above-applied built-in
4093 attributes from being intercepted (and possibly
4094 rejected) by a package's attribute routines, but is
4095 justified by the performance win for the common case
4096 of applying only built-in attributes.) */
4098 attrs = append_elem(OP_LIST, attrs,
4099 newSVOP(OP_CONST, 0,
4103 if (*s == ':' && s[1] != ':')
4106 break; /* require real whitespace or :'s */
4107 /* XXX losing whitespace on sequential attributes here */
4111 = (PL_expect == XOPERATOR ? '=' : '{'); /*'}(' for vi */
4112 if (*s != ';' && *s != '}' && *s != tmp
4113 && (tmp != '=' || *s != ')')) {
4114 const char q = ((*s == '\'') ? '"' : '\'');
4115 /* If here for an expression, and parsed no attrs, back
4117 if (tmp == '=' && !attrs) {
4121 /* MUST advance bufptr here to avoid bogus "at end of line"
4122 context messages from yyerror().
4126 ? Perl_form(aTHX_ "Invalid separator character "
4127 "%c%c%c in attribute list", q, *s, q)
4128 : "Unterminated attribute list" );
4136 start_force(PL_curforce);
4137 NEXTVAL_NEXTTOKE.opval = attrs;
4138 CURMAD('_', PL_nextwhite);
4143 PL_thistoken = newSVpvn(SvPVX(PL_linestr) + stuffstart,
4144 (s - SvPVX(PL_linestr)) - stuffstart);
4152 if (PL_last_lop == PL_oldoldbufptr || PL_last_uni == PL_oldoldbufptr)
4153 PL_oldbufptr = PL_oldoldbufptr; /* allow print(STDOUT 123) */
4161 const char tmp = *s++;
4166 const char tmp = *s++;
4174 if (PL_lex_brackets <= 0)
4175 yyerror("Unmatched right square bracket");
4178 if (PL_lex_state == LEX_INTERPNORMAL) {
4179 if (PL_lex_brackets == 0) {
4180 if (*s != '[' && *s != '{' && (*s != '-' || s[1] != '>'))
4181 PL_lex_state = LEX_INTERPEND;
4188 if (PL_lex_brackets > 100) {
4189 Renew(PL_lex_brackstack, PL_lex_brackets + 10, char);
4191 switch (PL_expect) {
4193 if (PL_lex_formbrack) {
4197 if (PL_oldoldbufptr == PL_last_lop)
4198 PL_lex_brackstack[PL_lex_brackets++] = XTERM;
4200 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
4201 OPERATOR(HASHBRACK);
4203 while (s < PL_bufend && SPACE_OR_TAB(*s))
4206 PL_tokenbuf[0] = '\0';
4207 if (d < PL_bufend && *d == '-') {
4208 PL_tokenbuf[0] = '-';
4210 while (d < PL_bufend && SPACE_OR_TAB(*d))
4213 if (d < PL_bufend && isIDFIRST_lazy_if(d,UTF)) {
4214 d = scan_word(d, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1,
4216 while (d < PL_bufend && SPACE_OR_TAB(*d))
4219 const char minus = (PL_tokenbuf[0] == '-');
4220 s = force_word(s + minus, WORD, FALSE, TRUE, FALSE);
4228 PL_lex_brackstack[PL_lex_brackets++] = XSTATE;
4233 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
4238 if (PL_oldoldbufptr == PL_last_lop)
4239 PL_lex_brackstack[PL_lex_brackets++] = XTERM;
4241 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
4244 if (PL_expect == XREF && PL_lex_state == LEX_INTERPNORMAL) {
4246 /* This hack is to get the ${} in the message. */
4248 yyerror("syntax error");
4251 OPERATOR(HASHBRACK);
4253 /* This hack serves to disambiguate a pair of curlies
4254 * as being a block or an anon hash. Normally, expectation
4255 * determines that, but in cases where we're not in a
4256 * position to expect anything in particular (like inside
4257 * eval"") we have to resolve the ambiguity. This code
4258 * covers the case where the first term in the curlies is a
4259 * quoted string. Most other cases need to be explicitly
4260 * disambiguated by prepending a "+" before the opening
4261 * curly in order to force resolution as an anon hash.
4263 * XXX should probably propagate the outer expectation
4264 * into eval"" to rely less on this hack, but that could
4265 * potentially break current behavior of eval"".
4269 if (*s == '\'' || *s == '"' || *s == '`') {
4270 /* common case: get past first string, handling escapes */
4271 for (t++; t < PL_bufend && *t != *s;)
4272 if (*t++ == '\\' && (*t == '\\' || *t == *s))
4276 else if (*s == 'q') {
4279 || ((*t == 'q' || *t == 'x') && ++t < PL_bufend
4282 /* skip q//-like construct */
4284 char open, close, term;
4287 while (t < PL_bufend && isSPACE(*t))
4289 /* check for q => */
4290 if (t+1 < PL_bufend && t[0] == '=' && t[1] == '>') {
4291 OPERATOR(HASHBRACK);
4295 if (term && (tmps = strchr("([{< )]}> )]}>",term)))
4299 for (t++; t < PL_bufend; t++) {
4300 if (*t == '\\' && t+1 < PL_bufend && open != '\\')
4302 else if (*t == open)
4306 for (t++; t < PL_bufend; t++) {
4307 if (*t == '\\' && t+1 < PL_bufend)
4309 else if (*t == close && --brackets <= 0)
4311 else if (*t == open)
4318 /* skip plain q word */
4319 while (t < PL_bufend && isALNUM_lazy_if(t,UTF))
4322 else if (isALNUM_lazy_if(t,UTF)) {
4324 while (t < PL_bufend && isALNUM_lazy_if(t,UTF))
4327 while (t < PL_bufend && isSPACE(*t))
4329 /* if comma follows first term, call it an anon hash */
4330 /* XXX it could be a comma expression with loop modifiers */
4331 if (t < PL_bufend && ((*t == ',' && (*s == 'q' || !isLOWER(*s)))
4332 || (*t == '=' && t[1] == '>')))
4333 OPERATOR(HASHBRACK);
4334 if (PL_expect == XREF)
4337 PL_lex_brackstack[PL_lex_brackets-1] = XSTATE;
4343 yylval.ival = CopLINE(PL_curcop);
4344 if (isSPACE(*s) || *s == '#')
4345 PL_copline = NOLINE; /* invalidate current command line number */
4350 if (PL_lex_brackets <= 0)
4351 yyerror("Unmatched right curly bracket");
4353 PL_expect = (expectation)PL_lex_brackstack[--PL_lex_brackets];
4354 if (PL_lex_brackets < PL_lex_formbrack && PL_lex_state != LEX_INTERPNORMAL)
4355 PL_lex_formbrack = 0;
4356 if (PL_lex_state == LEX_INTERPNORMAL) {
4357 if (PL_lex_brackets == 0) {
4358 if (PL_expect & XFAKEBRACK) {
4359 PL_expect &= XENUMMASK;
4360 PL_lex_state = LEX_INTERPEND;
4365 PL_thiswhite = newSVpvn("",0);
4366 sv_catpvn(PL_thiswhite,"}",1);
4369 return yylex(); /* ignore fake brackets */
4371 if (*s == '-' && s[1] == '>')
4372 PL_lex_state = LEX_INTERPENDMAYBE;
4373 else if (*s != '[' && *s != '{')
4374 PL_lex_state = LEX_INTERPEND;
4377 if (PL_expect & XFAKEBRACK) {
4378 PL_expect &= XENUMMASK;
4380 return yylex(); /* ignore fake brackets */
4382 start_force(PL_curforce);
4384 curmad('X', newSVpvn(s-1,1));
4385 CURMAD('_', PL_thiswhite);
4390 PL_thistoken = newSVpvn("",0);
4398 if (PL_expect == XOPERATOR) {
4399 if (PL_bufptr == PL_linestart && ckWARN(WARN_SEMICOLON)
4400 && isIDFIRST_lazy_if(s,UTF))
4402 CopLINE_dec(PL_curcop);
4403 Perl_warner(aTHX_ packWARN(WARN_SEMICOLON), PL_warn_nosemi);
4404 CopLINE_inc(PL_curcop);
4409 s = scan_ident(s - 1, PL_bufend, PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
4411 PL_expect = XOPERATOR;
4412 force_ident(PL_tokenbuf, '&');
4416 yylval.ival = (OPpENTERSUB_AMPER<<8);
4428 const char tmp = *s++;
4435 if (tmp && isSPACE(*s) && ckWARN(WARN_SYNTAX)
4436 && strchr("+-*/%.^&|<",tmp))
4437 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
4438 "Reversed %c= operator",(int)tmp);
4440 if (PL_expect == XSTATE && isALPHA(tmp) &&
4441 (s == PL_linestart+1 || s[-2] == '\n') )
4443 if (PL_in_eval && !PL_rsfp) {
4448 if (strnEQ(s,"=cut",4)) {
4464 PL_thiswhite = newSVpvn("",0);
4465 sv_catpvn(PL_thiswhite, PL_linestart,
4466 PL_bufend - PL_linestart);
4470 PL_doextract = TRUE;
4474 if (PL_lex_brackets < PL_lex_formbrack) {
4476 #ifdef PERL_STRICT_CR
4477 while (SPACE_OR_TAB(*t))
4479 while (SPACE_OR_TAB(*t) || *t == '\r')
4482 if (*t == '\n' || *t == '#') {
4493 const char tmp = *s++;
4495 /* was this !=~ where !~ was meant?
4496 * warn on m:!=~\s+([/?]|[msy]\W|tr\W): */
4498 if (*s == '~' && ckWARN(WARN_SYNTAX)) {
4499 const char *t = s+1;
4501 while (t < PL_bufend && isSPACE(*t))
4504 if (*t == '/' || *t == '?' ||
4505 ((*t == 'm' || *t == 's' || *t == 'y')
4506 && !isALNUM(t[1])) ||
4507 (*t == 't' && t[1] == 'r' && !isALNUM(t[2])))
4508 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
4509 "!=~ should be !~");
4519 if (PL_expect != XOPERATOR) {
4520 if (s[1] != '<' && !strchr(s,'>'))
4523 s = scan_heredoc(s);
4525 s = scan_inputsymbol(s);
4526 TERM(sublex_start());
4532 SHop(OP_LEFT_SHIFT);
4546 const char tmp = *s++;
4548 SHop(OP_RIGHT_SHIFT);
4549 else if (tmp == '=')
4558 if (PL_expect == XOPERATOR) {
4559 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
4561 deprecate_old(commaless_variable_list);
4562 return REPORT(','); /* grandfather non-comma-format format */
4566 if (s[1] == '#' && (isIDFIRST_lazy_if(s+2,UTF) || strchr("{$:+-", s[2]))) {
4567 PL_tokenbuf[0] = '@';
4568 s = scan_ident(s + 1, PL_bufend, PL_tokenbuf + 1,
4569 sizeof PL_tokenbuf - 1, FALSE);
4570 if (PL_expect == XOPERATOR)
4571 no_op("Array length", s);
4572 if (!PL_tokenbuf[1])
4574 PL_expect = XOPERATOR;
4575 PL_pending_ident = '#';
4579 PL_tokenbuf[0] = '$';
4580 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1,
4581 sizeof PL_tokenbuf - 1, FALSE);
4582 if (PL_expect == XOPERATOR)
4584 if (!PL_tokenbuf[1]) {
4586 yyerror("Final $ should be \\$ or $name");
4590 /* This kludge not intended to be bulletproof. */
4591 if (PL_tokenbuf[1] == '[' && !PL_tokenbuf[2]) {
4592 yylval.opval = newSVOP(OP_CONST, 0,
4593 newSViv(CopARYBASE_get(&PL_compiling)));
4594 yylval.opval->op_private = OPpCONST_ARYBASE;
4600 const char tmp = *s;
4601 if (PL_lex_state == LEX_NORMAL)
4604 if ((PL_expect != XREF || PL_oldoldbufptr == PL_last_lop)
4605 && intuit_more(s)) {
4607 PL_tokenbuf[0] = '@';
4608 if (ckWARN(WARN_SYNTAX)) {
4611 while (isSPACE(*t) || isALNUM_lazy_if(t,UTF) || *t == '$')
4614 PL_bufptr = PEEKSPACE(PL_bufptr); /* XXX can realloc */
4615 while (t < PL_bufend && *t != ']')
4617 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
4618 "Multidimensional syntax %.*s not supported",
4619 (int)((t - PL_bufptr) + 1), PL_bufptr);
4623 else if (*s == '{') {
4625 PL_tokenbuf[0] = '%';
4626 if (strEQ(PL_tokenbuf+1, "SIG") && ckWARN(WARN_SYNTAX)
4627 && (t = strchr(s, '}')) && (t = strchr(t, '=')))
4629 char tmpbuf[sizeof PL_tokenbuf];
4632 } while (isSPACE(*t));
4633 if (isIDFIRST_lazy_if(t,UTF)) {
4635 t = scan_word(t, tmpbuf, sizeof tmpbuf, TRUE,
4639 if (*t == ';' && get_cv(tmpbuf, FALSE))
4640 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
4641 "You need to quote \"%s\"",
4648 PL_expect = XOPERATOR;
4649 if (PL_lex_state == LEX_NORMAL && isSPACE((char)tmp)) {
4650 const bool islop = (PL_last_lop == PL_oldoldbufptr);
4651 if (!islop || PL_last_lop_op == OP_GREPSTART)
4652 PL_expect = XOPERATOR;
4653 else if (strchr("$@\"'`q", *s))
4654 PL_expect = XTERM; /* e.g. print $fh "foo" */
4655 else if (strchr("&*<%", *s) && isIDFIRST_lazy_if(s+1,UTF))
4656 PL_expect = XTERM; /* e.g. print $fh &sub */
4657 else if (isIDFIRST_lazy_if(s,UTF)) {
4658 char tmpbuf[sizeof PL_tokenbuf];
4660 scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
4661 if ((t2 = keyword(tmpbuf, len))) {
4662 /* binary operators exclude handle interpretations */
4674 PL_expect = XTERM; /* e.g. print $fh length() */
4679 PL_expect = XTERM; /* e.g. print $fh subr() */
4682 else if (isDIGIT(*s))
4683 PL_expect = XTERM; /* e.g. print $fh 3 */
4684 else if (*s == '.' && isDIGIT(s[1]))
4685 PL_expect = XTERM; /* e.g. print $fh .3 */
4686 else if ((*s == '?' || *s == '-' || *s == '+')
4687 && !isSPACE(s[1]) && s[1] != '=')
4688 PL_expect = XTERM; /* e.g. print $fh -1 */
4689 else if (*s == '/' && !isSPACE(s[1]) && s[1] != '='
4691 PL_expect = XTERM; /* e.g. print $fh /.../
4692 XXX except DORDOR operator
4694 else if (*s == '<' && s[1] == '<' && !isSPACE(s[2])
4696 PL_expect = XTERM; /* print $fh <<"EOF" */
4699 PL_pending_ident = '$';
4703 if (PL_expect == XOPERATOR)
4705 PL_tokenbuf[0] = '@';
4706 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1, FALSE);
4707 if (!PL_tokenbuf[1]) {
4710 if (PL_lex_state == LEX_NORMAL)
4712 if ((PL_expect != XREF || PL_oldoldbufptr == PL_last_lop) && intuit_more(s)) {
4714 PL_tokenbuf[0] = '%';
4716 /* Warn about @ where they meant $. */
4717 if (*s == '[' || *s == '{') {
4718 if (ckWARN(WARN_SYNTAX)) {
4719 const char *t = s + 1;
4720 while (*t && (isALNUM_lazy_if(t,UTF) || strchr(" \t$#+-'\"", *t)))
4722 if (*t == '}' || *t == ']') {
4724 PL_bufptr = PEEKSPACE(PL_bufptr); /* XXX can realloc */
4725 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
4726 "Scalar value %.*s better written as $%.*s",
4727 (int)(t-PL_bufptr), PL_bufptr,
4728 (int)(t-PL_bufptr-1), PL_bufptr+1);
4733 PL_pending_ident = '@';
4736 case '/': /* may be division, defined-or, or pattern */
4737 if (PL_expect == XTERMORDORDOR && s[1] == '/') {
4741 case '?': /* may either be conditional or pattern */
4742 if(PL_expect == XOPERATOR) {
4750 /* A // operator. */
4760 /* Disable warning on "study /blah/" */
4761 if (PL_oldoldbufptr == PL_last_uni
4762 && (*PL_last_uni != 's' || s - PL_last_uni < 5
4763 || memNE(PL_last_uni, "study", 5)
4764 || isALNUM_lazy_if(PL_last_uni+5,UTF)
4767 s = scan_pat(s,OP_MATCH);
4768 TERM(sublex_start());
4772 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack
4773 #ifdef PERL_STRICT_CR
4776 && (s[1] == '\n' || (s[1] == '\r' && s[2] == '\n'))
4778 && (s == PL_linestart || s[-1] == '\n') )
4780 PL_lex_formbrack = 0;
4784 if (PL_expect == XOPERATOR || !isDIGIT(s[1])) {
4790 yylval.ival = OPf_SPECIAL;
4796 if (PL_expect != XOPERATOR)
4801 case '0': case '1': case '2': case '3': case '4':
4802 case '5': case '6': case '7': case '8': case '9':
4803 s = scan_num(s, &yylval);
4804 DEBUG_T( { printbuf("### Saw number in %s\n", s); } );
4805 if (PL_expect == XOPERATOR)
4810 s = scan_str(s,!!PL_madskills,FALSE);
4811 DEBUG_T( { printbuf("### Saw string before %s\n", s); } );
4812 if (PL_expect == XOPERATOR) {
4813 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
4815 deprecate_old(commaless_variable_list);
4816 return REPORT(','); /* grandfather non-comma-format format */
4823 yylval.ival = OP_CONST;
4824 TERM(sublex_start());
4827 s = scan_str(s,!!PL_madskills,FALSE);
4828 DEBUG_T( { printbuf("### Saw string before %s\n", s); } );
4829 if (PL_expect == XOPERATOR) {
4830 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
4832 deprecate_old(commaless_variable_list);
4833 return REPORT(','); /* grandfather non-comma-format format */
4840 yylval.ival = OP_CONST;
4841 /* FIXME. I think that this can be const if char *d is replaced by
4842 more localised variables. */
4843 for (d = SvPV(PL_lex_stuff, len); len; len--, d++) {
4844 if (*d == '$' || *d == '@' || *d == '\\' || !UTF8_IS_INVARIANT((U8)*d)) {
4845 yylval.ival = OP_STRINGIFY;
4849 TERM(sublex_start());
4852 s = scan_str(s,!!PL_madskills,FALSE);
4853 DEBUG_T( { printbuf("### Saw backtick string before %s\n", s); } );
4854 if (PL_expect == XOPERATOR)
4855 no_op("Backticks",s);
4858 yylval.ival = OP_BACKTICK;
4860 TERM(sublex_start());
4864 if (PL_lex_inwhat && isDIGIT(*s) && ckWARN(WARN_SYNTAX))
4865 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),"Can't use \\%c to mean $%c in expression",
4867 if (PL_expect == XOPERATOR)
4868 no_op("Backslash",s);
4872 if (isDIGIT(s[1]) && PL_expect != XOPERATOR) {
4873 char *start = s + 2;
4874 while (isDIGIT(*start) || *start == '_')
4876 if (*start == '.' && isDIGIT(start[1])) {
4877 s = scan_num(s, &yylval);
4880 /* avoid v123abc() or $h{v1}, allow C<print v10;> */
4881 else if (!isALPHA(*start) && (PL_expect == XTERM
4882 || PL_expect == XREF || PL_expect == XSTATE
4883 || PL_expect == XTERMORDORDOR)) {
4884 /* XXX Use gv_fetchpvn rather than stomping on a const string */
4885 const char c = *start;
4888 gv = gv_fetchpv(s, 0, SVt_PVCV);
4891 s = scan_num(s, &yylval);
4898 if (isDIGIT(s[1]) && PL_expect == XOPERATOR) {
4934 I32 orig_keyword = 0;
4939 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
4941 /* Some keywords can be followed by any delimiter, including ':' */
4942 tmp = ((len == 1 && strchr("msyq", PL_tokenbuf[0])) ||
4943 (len == 2 && ((PL_tokenbuf[0] == 't' && PL_tokenbuf[1] == 'r') ||
4944 (PL_tokenbuf[0] == 'q' &&
4945 strchr("qwxr", PL_tokenbuf[1])))));
4947 /* x::* is just a word, unless x is "CORE" */
4948 if (!tmp && *s == ':' && s[1] == ':' && strNE(PL_tokenbuf, "CORE"))
4952 while (d < PL_bufend && isSPACE(*d))
4953 d++; /* no comments skipped here, or s### is misparsed */
4955 /* Is this a label? */
4956 if (!tmp && PL_expect == XSTATE
4957 && d < PL_bufend && *d == ':' && *(d + 1) != ':') {
4959 yylval.pval = savepv(PL_tokenbuf);
4964 /* Check for keywords */
4965 tmp = keyword(PL_tokenbuf, len);
4967 /* Is this a word before a => operator? */
4968 if (*d == '=' && d[1] == '>') {
4971 = (OP*)newSVOP(OP_CONST, 0,
4972 S_newSV_maybe_utf8(aTHX_ PL_tokenbuf, len));
4973 yylval.opval->op_private = OPpCONST_BARE;
4977 if (tmp < 0) { /* second-class keyword? */
4978 GV *ogv = NULL; /* override (winner) */
4979 GV *hgv = NULL; /* hidden (loser) */
4980 if (PL_expect != XOPERATOR && (*s != ':' || s[1] != ':')) {
4982 if ((gv = gv_fetchpvn_flags(PL_tokenbuf, len, 0, SVt_PVCV)) &&
4985 if (GvIMPORTED_CV(gv))
4987 else if (! CvMETHOD(cv))
4991 (gvp = (GV**)hv_fetch(PL_globalstash,PL_tokenbuf,len,FALSE)) &&
4992 (gv = *gvp) != (GV*)&PL_sv_undef &&
4993 GvCVu(gv) && GvIMPORTED_CV(gv))
5000 tmp = 0; /* overridden by import or by GLOBAL */
5003 && -tmp==KEY_lock /* XXX generalizable kludge */
5005 && !hv_fetchs(GvHVn(PL_incgv), "Thread.pm", FALSE))
5007 tmp = 0; /* any sub overrides "weak" keyword */
5009 else { /* no override */
5011 if (tmp == KEY_dump && ckWARN(WARN_MISC)) {
5012 Perl_warner(aTHX_ packWARN(WARN_MISC),
5013 "dump() better written as CORE::dump()");
5017 if (hgv && tmp != KEY_x && tmp != KEY_CORE
5018 && ckWARN(WARN_AMBIGUOUS)) /* never ambiguous */
5019 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
5020 "Ambiguous call resolved as CORE::%s(), %s",
5021 GvENAME(hgv), "qualify as such or use &");
5028 default: /* not a keyword */
5029 /* Trade off - by using this evil construction we can pull the
5030 variable gv into the block labelled keylookup. If not, then
5031 we have to give it function scope so that the goto from the
5032 earlier ':' case doesn't bypass the initialisation. */
5034 just_a_word_zero_gv:
5042 const char lastchar = (PL_bufptr == PL_oldoldbufptr ? 0 : PL_bufptr[-1]);
5045 SV *nextPL_nextwhite = 0;
5049 /* Get the rest if it looks like a package qualifier */
5051 if (*s == '\'' || (*s == ':' && s[1] == ':')) {
5053 s = scan_word(s, PL_tokenbuf + len, sizeof PL_tokenbuf - len,
5056 Perl_croak(aTHX_ "Bad name after %s%s", PL_tokenbuf,
5057 *s == '\'' ? "'" : "::");
5062 if (PL_expect == XOPERATOR) {
5063 if (PL_bufptr == PL_linestart) {
5064 CopLINE_dec(PL_curcop);
5065 Perl_warner(aTHX_ packWARN(WARN_SEMICOLON), PL_warn_nosemi);
5066 CopLINE_inc(PL_curcop);
5069 no_op("Bareword",s);
5072 /* Look for a subroutine with this name in current package,
5073 unless name is "Foo::", in which case Foo is a bearword
5074 (and a package name). */
5076 if (len > 2 && !PL_madskills &&
5077 PL_tokenbuf[len - 2] == ':' && PL_tokenbuf[len - 1] == ':')
5079 if (ckWARN(WARN_BAREWORD)
5080 && ! gv_fetchpvn_flags(PL_tokenbuf, len, 0, SVt_PVHV))
5081 Perl_warner(aTHX_ packWARN(WARN_BAREWORD),
5082 "Bareword \"%s\" refers to nonexistent package",
5085 PL_tokenbuf[len] = '\0';
5091 /* Mustn't actually add anything to a symbol table.
5092 But also don't want to "initialise" any placeholder
5093 constants that might already be there into full
5094 blown PVGVs with attached PVCV. */
5095 gv = gv_fetchpvn_flags(PL_tokenbuf, len,
5096 GV_NOADD_NOINIT, SVt_PVCV);
5101 /* if we saw a global override before, get the right name */
5104 sv = newSVpvs("CORE::GLOBAL::");
5105 sv_catpv(sv,PL_tokenbuf);
5108 /* If len is 0, newSVpv does strlen(), which is correct.
5109 If len is non-zero, then it will be the true length,
5110 and so the scalar will be created correctly. */
5111 sv = newSVpv(PL_tokenbuf,len);
5114 if (PL_madskills && !PL_thistoken) {
5115 char *start = SvPVX(PL_linestr) + PL_realtokenstart;
5116 PL_thistoken = newSVpv(start,s - start);
5117 PL_realtokenstart = s - SvPVX(PL_linestr);
5121 /* Presume this is going to be a bareword of some sort. */
5124 yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
5125 yylval.opval->op_private = OPpCONST_BARE;
5126 /* UTF-8 package name? */
5127 if (UTF && !IN_BYTES &&
5128 is_utf8_string((U8*)SvPVX_const(sv), SvCUR(sv)))
5131 /* And if "Foo::", then that's what it certainly is. */
5136 /* Do the explicit type check so that we don't need to force
5137 the initialisation of the symbol table to have a real GV.
5138 Beware - gv may not really be a PVGV, cv may not really be
5139 a PVCV, (because of the space optimisations that gv_init
5140 understands) But they're true if for this symbol there is
5141 respectively a typeglob and a subroutine.
5143 cv = gv ? ((SvTYPE(gv) == SVt_PVGV)
5144 /* Real typeglob, so get the real subroutine: */
5146 /* A proxy for a subroutine in this package? */
5147 : SvOK(gv) ? (CV *) gv : NULL)
5150 /* See if it's the indirect object for a list operator. */
5152 if (PL_oldoldbufptr &&
5153 PL_oldoldbufptr < PL_bufptr &&
5154 (PL_oldoldbufptr == PL_last_lop
5155 || PL_oldoldbufptr == PL_last_uni) &&
5156 /* NO SKIPSPACE BEFORE HERE! */
5157 (PL_expect == XREF ||
5158 ((PL_opargs[PL_last_lop_op] >> OASHIFT)& 7) == OA_FILEREF))
5160 bool immediate_paren = *s == '(';
5162 /* (Now we can afford to cross potential line boundary.) */
5163 s = SKIPSPACE2(s,nextPL_nextwhite);
5165 PL_nextwhite = nextPL_nextwhite; /* assume no & deception */
5168 /* Two barewords in a row may indicate method call. */
5170 if ((isIDFIRST_lazy_if(s,UTF) || *s == '$') &&
5171 (tmp = intuit_method(s, gv, cv)))
5174 /* If not a declared subroutine, it's an indirect object. */
5175 /* (But it's an indir obj regardless for sort.) */
5176 /* Also, if "_" follows a filetest operator, it's a bareword */
5179 ( !immediate_paren && (PL_last_lop_op == OP_SORT ||
5181 (PL_last_lop_op != OP_MAPSTART &&
5182 PL_last_lop_op != OP_GREPSTART))))
5183 || (PL_tokenbuf[0] == '_' && PL_tokenbuf[1] == '\0'
5184 && ((PL_opargs[PL_last_lop_op] & OA_CLASS_MASK) == OA_FILESTATOP))
5187 PL_expect = (PL_last_lop == PL_oldoldbufptr) ? XTERM : XOPERATOR;
5192 PL_expect = XOPERATOR;
5195 s = SKIPSPACE2(s,nextPL_nextwhite);
5196 PL_nextwhite = nextPL_nextwhite;
5201 /* Is this a word before a => operator? */
5202 if (*s == '=' && s[1] == '>' && !pkgname) {
5204 sv_setpv(((SVOP*)yylval.opval)->op_sv, PL_tokenbuf);
5205 if (UTF && !IN_BYTES && is_utf8_string((U8*)PL_tokenbuf, len))
5206 SvUTF8_on(((SVOP*)yylval.opval)->op_sv);
5210 /* If followed by a paren, it's certainly a subroutine. */
5215 while (SPACE_OR_TAB(*d))
5217 if (*d == ')' && (sv = gv_const_sv(gv))) {
5221 char *par = SvPVX(PL_linestr) + PL_realtokenstart;
5222 sv_catpvn(PL_thistoken, par, s - par);
5224 sv_free(PL_nextwhite);
5234 PL_nextwhite = PL_thiswhite;
5237 start_force(PL_curforce);
5239 NEXTVAL_NEXTTOKE.opval = yylval.opval;
5240 PL_expect = XOPERATOR;
5243 PL_nextwhite = nextPL_nextwhite;
5244 curmad('X', PL_thistoken);
5245 PL_thistoken = newSVpvn("",0);
5253 /* If followed by var or block, call it a method (unless sub) */
5255 if ((*s == '$' || *s == '{') && (!gv || !cv)) {
5256 PL_last_lop = PL_oldbufptr;
5257 PL_last_lop_op = OP_METHOD;
5261 /* If followed by a bareword, see if it looks like indir obj. */
5264 && (isIDFIRST_lazy_if(s,UTF) || *s == '$')
5265 && (tmp = intuit_method(s, gv, cv)))
5268 /* Not a method, so call it a subroutine (if defined) */
5271 if (lastchar == '-' && ckWARN_d(WARN_AMBIGUOUS))
5272 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
5273 "Ambiguous use of -%s resolved as -&%s()",
5274 PL_tokenbuf, PL_tokenbuf);
5275 /* Check for a constant sub */
5276 if ((sv = gv_const_sv(gv))) {
5278 SvREFCNT_dec(((SVOP*)yylval.opval)->op_sv);
5279 ((SVOP*)yylval.opval)->op_sv = SvREFCNT_inc_simple(sv);
5280 yylval.opval->op_private = 0;
5284 /* Resolve to GV now. */
5285 if (SvTYPE(gv) != SVt_PVGV) {
5286 gv = gv_fetchpv(PL_tokenbuf, 0, SVt_PVCV);
5287 assert (SvTYPE(gv) == SVt_PVGV);
5288 /* cv must have been some sort of placeholder, so
5289 now needs replacing with a real code reference. */
5293 op_free(yylval.opval);
5294 yylval.opval = newCVREF(0, newGVOP(OP_GV, 0, gv));
5295 yylval.opval->op_private |= OPpENTERSUB_NOPAREN;
5296 PL_last_lop = PL_oldbufptr;
5297 PL_last_lop_op = OP_ENTERSUB;
5298 /* Is there a prototype? */
5305 const char *proto = SvPV_const((SV*)cv, protolen);
5308 if (*proto == '$' && proto[1] == '\0')
5310 while (*proto == ';')
5312 if (*proto == '&' && *s == '{') {
5313 sv_setpv(PL_subname, PL_curstash ?
5314 "__ANON__" : "__ANON__::__ANON__");
5321 PL_nextwhite = PL_thiswhite;
5324 start_force(PL_curforce);
5325 NEXTVAL_NEXTTOKE.opval = yylval.opval;
5328 PL_nextwhite = nextPL_nextwhite;
5329 curmad('X', PL_thistoken);
5330 PL_thistoken = newSVpvn("",0);
5337 /* Guess harder when madskills require "best effort". */
5338 if (PL_madskills && (!gv || !GvCVu(gv))) {
5339 int probable_sub = 0;
5340 if (strchr("\"'`$@%0123456789!*+{[<", *s))
5342 else if (isALPHA(*s)) {
5346 d = scan_word(d, tmpbuf, sizeof tmpbuf, TRUE, &tmplen);
5347 if (!keyword(tmpbuf,tmplen))
5350 while (d < PL_bufend && isSPACE(*d))
5352 if (*d == '=' && d[1] == '>')
5357 gv = gv_fetchpv(PL_tokenbuf, TRUE, SVt_PVCV);
5358 op_free(yylval.opval);
5359 yylval.opval = newCVREF(0, newGVOP(OP_GV, 0, gv));
5360 yylval.opval->op_private |= OPpENTERSUB_NOPAREN;
5361 PL_last_lop = PL_oldbufptr;
5362 PL_last_lop_op = OP_ENTERSUB;
5363 PL_nextwhite = PL_thiswhite;
5365 start_force(PL_curforce);
5366 NEXTVAL_NEXTTOKE.opval = yylval.opval;
5368 PL_nextwhite = nextPL_nextwhite;
5369 curmad('X', PL_thistoken);
5370 PL_thistoken = newSVpvn("",0);
5375 NEXTVAL_NEXTTOKE.opval = yylval.opval;
5382 /* Call it a bare word */
5384 if (PL_hints & HINT_STRICT_SUBS)
5385 yylval.opval->op_private |= OPpCONST_STRICT;
5388 if (lastchar != '-') {
5389 if (ckWARN(WARN_RESERVED)) {
5393 if (!*d && !gv_stashpv(PL_tokenbuf,FALSE))
5394 Perl_warner(aTHX_ packWARN(WARN_RESERVED), PL_warn_reserved,
5401 if ((lastchar == '*' || lastchar == '%' || lastchar == '&')
5402 && ckWARN_d(WARN_AMBIGUOUS)) {
5403 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
5404 "Operator or semicolon missing before %c%s",
5405 lastchar, PL_tokenbuf);
5406 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
5407 "Ambiguous use of %c resolved as operator %c",
5408 lastchar, lastchar);
5414 yylval.opval = (OP*)newSVOP(OP_CONST, 0,
5415 newSVpv(CopFILE(PL_curcop),0));
5419 yylval.opval = (OP*)newSVOP(OP_CONST, 0,
5420 Perl_newSVpvf(aTHX_ "%"IVdf, (IV)CopLINE(PL_curcop)));
5423 case KEY___PACKAGE__:
5424 yylval.opval = (OP*)newSVOP(OP_CONST, 0,
5426 ? newSVhek(HvNAME_HEK(PL_curstash))
5433 if (PL_rsfp && (!PL_in_eval || PL_tokenbuf[2] == 'D')) {
5434 const char *pname = "main";
5435 if (PL_tokenbuf[2] == 'D')
5436 pname = HvNAME_get(PL_curstash ? PL_curstash : PL_defstash);
5437 gv = gv_fetchpv(Perl_form(aTHX_ "%s::DATA", pname), GV_ADD,
5441 GvIOp(gv) = newIO();
5442 IoIFP(GvIOp(gv)) = PL_rsfp;
5443 #if defined(HAS_FCNTL) && defined(F_SETFD)
5445 const int fd = PerlIO_fileno(PL_rsfp);
5446 fcntl(fd,F_SETFD,fd >= 3);
5449 /* Mark this internal pseudo-handle as clean */
5450 IoFLAGS(GvIOp(gv)) |= IOf_UNTAINT;
5452 IoTYPE(GvIOp(gv)) = IoTYPE_PIPE;
5453 else if ((PerlIO*)PL_rsfp == PerlIO_stdin())
5454 IoTYPE(GvIOp(gv)) = IoTYPE_STD;
5456 IoTYPE(GvIOp(gv)) = IoTYPE_RDONLY;
5457 #if defined(WIN32) && !defined(PERL_TEXTMODE_SCRIPTS)
5458 /* if the script was opened in binmode, we need to revert
5459 * it to text mode for compatibility; but only iff it has CRs
5460 * XXX this is a questionable hack at best. */
5461 if (PL_bufend-PL_bufptr > 2
5462 && PL_bufend[-1] == '\n' && PL_bufend[-2] == '\r')
5465 if (IoTYPE(GvIOp(gv)) == IoTYPE_RDONLY) {
5466 loc = PerlIO_tell(PL_rsfp);
5467 (void)PerlIO_seek(PL_rsfp, 0L, 0);
5470 if (PerlLIO_setmode(PL_rsfp, O_TEXT) != -1) {
5472 if (PerlLIO_setmode(PerlIO_fileno(PL_rsfp), O_TEXT) != -1) {
5473 #endif /* NETWARE */
5474 #ifdef PERLIO_IS_STDIO /* really? */
5475 # if defined(__BORLANDC__)
5476 /* XXX see note in do_binmode() */
5477 ((FILE*)PL_rsfp)->flags &= ~_F_BIN;
5481 PerlIO_seek(PL_rsfp, loc, 0);
5485 #ifdef PERLIO_LAYERS
5488 PerlIO_apply_layers(aTHX_ PL_rsfp, NULL, ":utf8");
5489 else if (PL_encoding) {
5496 XPUSHs(PL_encoding);
5498 call_method("name", G_SCALAR);
5502 PerlIO_apply_layers(aTHX_ PL_rsfp, NULL,
5503 Perl_form(aTHX_ ":encoding(%"SVf")",
5512 if (PL_realtokenstart >= 0) {
5513 char *tstart = SvPVX(PL_linestr) + PL_realtokenstart;
5515 PL_endwhite = newSVpvn("",0);
5516 sv_catsv(PL_endwhite, PL_thiswhite);
5518 sv_catpvn(PL_endwhite, tstart, PL_bufend - tstart);
5519 PL_realtokenstart = -1;
5521 while ((s = filter_gets(PL_endwhite, PL_rsfp,
5522 SvCUR(PL_endwhite))) != Nullch) ;
5536 if (PL_expect == XSTATE) {
5543 if (*s == ':' && s[1] == ':') {
5546 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
5547 if (!(tmp = keyword(PL_tokenbuf, len)))
5548 Perl_croak(aTHX_ "CORE::%s is not a keyword", PL_tokenbuf);
5551 else if (tmp == KEY_require || tmp == KEY_do)
5552 /* that's a way to remember we saw "CORE::" */
5565 LOP(OP_ACCEPT,XTERM);
5571 LOP(OP_ATAN2,XTERM);
5577 LOP(OP_BINMODE,XTERM);
5580 LOP(OP_BLESS,XTERM);
5589 /* When 'use switch' is in effect, continue has a dual
5590 life as a control operator. */
5592 if (!FEATURE_IS_ENABLED("switch"))
5595 /* We have to disambiguate the two senses of
5596 "continue". If the next token is a '{' then
5597 treat it as the start of a continue block;
5598 otherwise treat it as a control operator.
5610 (void)gv_fetchpvs("ENV", GV_ADD|GV_NOTQUAL, SVt_PVHV);
5627 if (!PL_cryptseen) {
5628 PL_cryptseen = TRUE;
5632 LOP(OP_CRYPT,XTERM);
5635 LOP(OP_CHMOD,XTERM);
5638 LOP(OP_CHOWN,XTERM);
5641 LOP(OP_CONNECT,XTERM);
5660 s = force_word(s,WORD,TRUE,TRUE,FALSE);
5661 if (orig_keyword == KEY_do) {
5670 PL_hints |= HINT_BLOCK_SCOPE;
5680 gv_fetchpvs("AnyDBM_File::ISA", GV_ADDMULTI, SVt_PVAV);
5681 LOP(OP_DBMOPEN,XTERM);
5687 s = force_word(s,WORD,TRUE,FALSE,FALSE);
5694 yylval.ival = CopLINE(PL_curcop);
5710 PL_expect = (*s == '{') ? XTERMBLOCK : XTERM;
5711 UNIBRACK(OP_ENTEREVAL);
5729 case KEY_endhostent:
5735 case KEY_endservent:
5738 case KEY_endprotoent:
5749 yylval.ival = CopLINE(PL_curcop);
5751 if (PL_expect == XSTATE && isIDFIRST_lazy_if(s,UTF)) {
5754 int soff = s - SvPVX(PL_linestr); /* for skipspace realloc */
5757 if ((PL_bufend - p) >= 3 &&
5758 strnEQ(p, "my", 2) && isSPACE(*(p + 2)))
5760 else if ((PL_bufend - p) >= 4 &&
5761 strnEQ(p, "our", 3) && isSPACE(*(p + 3)))
5764 if (isIDFIRST_lazy_if(p,UTF)) {
5765 p = scan_ident(p, PL_bufend,
5766 PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
5770 Perl_croak(aTHX_ "Missing $ on loop variable");
5772 s = SvPVX(PL_linestr) + soff;
5778 LOP(OP_FORMLINE,XTERM);
5784 LOP(OP_FCNTL,XTERM);
5790 LOP(OP_FLOCK,XTERM);
5799 LOP(OP_GREPSTART, XREF);
5802 s = force_word(s,WORD,TRUE,FALSE,FALSE);
5817 case KEY_getpriority:
5818 LOP(OP_GETPRIORITY,XTERM);
5820 case KEY_getprotobyname:
5823 case KEY_getprotobynumber:
5824 LOP(OP_GPBYNUMBER,XTERM);
5826 case KEY_getprotoent:
5838 case KEY_getpeername:
5839 UNI(OP_GETPEERNAME);
5841 case KEY_gethostbyname:
5844 case KEY_gethostbyaddr:
5845 LOP(OP_GHBYADDR,XTERM);
5847 case KEY_gethostent:
5850 case KEY_getnetbyname:
5853 case KEY_getnetbyaddr:
5854 LOP(OP_GNBYADDR,XTERM);
5859 case KEY_getservbyname:
5860 LOP(OP_GSBYNAME,XTERM);
5862 case KEY_getservbyport:
5863 LOP(OP_GSBYPORT,XTERM);
5865 case KEY_getservent:
5868 case KEY_getsockname:
5869 UNI(OP_GETSOCKNAME);
5871 case KEY_getsockopt:
5872 LOP(OP_GSOCKOPT,XTERM);
5887 yylval.ival = CopLINE(PL_curcop);
5898 yylval.ival = CopLINE(PL_curcop);
5902 LOP(OP_INDEX,XTERM);
5908 LOP(OP_IOCTL,XTERM);
5920 s = force_word(s,WORD,TRUE,FALSE,FALSE);
5952 LOP(OP_LISTEN,XTERM);
5961 s = scan_pat(s,OP_MATCH);
5962 TERM(sublex_start());
5965 LOP(OP_MAPSTART, XREF);
5968 LOP(OP_MKDIR,XTERM);
5971 LOP(OP_MSGCTL,XTERM);
5974 LOP(OP_MSGGET,XTERM);
5977 LOP(OP_MSGRCV,XTERM);
5980 LOP(OP_MSGSND,XTERM);
5987 if (isIDFIRST_lazy_if(s,UTF)) {
5991 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, TRUE, &len);
5992 if (len == 3 && strnEQ(PL_tokenbuf, "sub", 3))
5994 PL_in_my_stash = find_in_my_stash(PL_tokenbuf, len);
5995 if (!PL_in_my_stash) {
5998 my_snprintf(tmpbuf, sizeof(tmpbuf), "No such class %.1000s", PL_tokenbuf);
6002 if (PL_madskills) { /* just add type to declarator token */
6003 sv_catsv(PL_thistoken, PL_nextwhite);
6005 sv_catpvn(PL_thistoken, start, s - start);
6013 s = force_word(s,WORD,TRUE,FALSE,FALSE);
6020 s = tokenize_use(0, s);
6024 if (*s == '(' || (s = SKIPSPACE1(s), *s == '('))
6031 if (isIDFIRST_lazy_if(s,UTF)) {
6033 for (d = s; isALNUM_lazy_if(d,UTF);)
6035 for (t=d; isSPACE(*t);)
6037 if ( *t && strchr("|&*+-=!?:.", *t) && ckWARN_d(WARN_PRECEDENCE)
6039 && !(t[0] == '=' && t[1] == '>')
6041 int parms_len = (int)(d-s);
6042 Perl_warner(aTHX_ packWARN(WARN_PRECEDENCE),
6043 "Precedence problem: open %.*s should be open(%.*s)",
6044 parms_len, s, parms_len, s);
6050 yylval.ival = OP_OR;
6060 LOP(OP_OPEN_DIR,XTERM);
6063 checkcomma(s,PL_tokenbuf,"filehandle");
6067 checkcomma(s,PL_tokenbuf,"filehandle");
6086 s = force_word(s,WORD,FALSE,TRUE,FALSE);
6090 LOP(OP_PIPE_OP,XTERM);
6093 s = scan_str(s,!!PL_madskills,FALSE);
6096 yylval.ival = OP_CONST;
6097 TERM(sublex_start());
6103 s = scan_str(s,!!PL_madskills,FALSE);
6106 PL_expect = XOPERATOR;
6108 if (SvCUR(PL_lex_stuff)) {
6111 d = SvPV_force(PL_lex_stuff, len);
6113 for (; isSPACE(*d) && len; --len, ++d)
6118 if (!warned && ckWARN(WARN_QW)) {
6119 for (; !isSPACE(*d) && len; --len, ++d) {
6121 Perl_warner(aTHX_ packWARN(WARN_QW),
6122 "Possible attempt to separate words with commas");
6125 else if (*d == '#') {
6126 Perl_warner(aTHX_ packWARN(WARN_QW),
6127 "Possible attempt to put comments in qw() list");
6133 for (; !isSPACE(*d) && len; --len, ++d)
6136 sv = newSVpvn(b, d-b);
6137 if (DO_UTF8(PL_lex_stuff))
6139 words = append_elem(OP_LIST, words,
6140 newSVOP(OP_CONST, 0, tokeq(sv)));
6144 start_force(PL_curforce);
6145 NEXTVAL_NEXTTOKE.opval = words;
6150 SvREFCNT_dec(PL_lex_stuff);
6151 PL_lex_stuff = NULL;
6157 s = scan_str(s,!!PL_madskills,FALSE);
6160 yylval.ival = OP_STRINGIFY;
6161 if (SvIVX(PL_lex_stuff) == '\'')
6162 SvIV_set(PL_lex_stuff, 0); /* qq'$foo' should intepolate */
6163 TERM(sublex_start());
6166 s = scan_pat(s,OP_QR);
6167 TERM(sublex_start());
6170 s = scan_str(s,!!PL_madskills,FALSE);
6173 yylval.ival = OP_BACKTICK;
6175 TERM(sublex_start());
6183 s = force_version(s, FALSE);
6185 else if (*s != 'v' || !isDIGIT(s[1])
6186 || (s = force_version(s, TRUE), *s == 'v'))
6188 *PL_tokenbuf = '\0';
6189 s = force_word(s,WORD,TRUE,TRUE,FALSE);
6190 if (isIDFIRST_lazy_if(PL_tokenbuf,UTF))
6191 gv_stashpvn(PL_tokenbuf, strlen(PL_tokenbuf), TRUE);
6193 yyerror("<> should be quotes");
6195 if (orig_keyword == KEY_require) {
6203 PL_last_uni = PL_oldbufptr;
6204 PL_last_lop_op = OP_REQUIRE;
6206 return REPORT( (int)REQUIRE );
6212 s = force_word(s,WORD,TRUE,FALSE,FALSE);
6216 LOP(OP_RENAME,XTERM);
6225 LOP(OP_RINDEX,XTERM);
6235 UNIDOR(OP_READLINE);
6248 LOP(OP_REVERSE,XTERM);
6251 UNIDOR(OP_READLINK);
6259 TERM(sublex_start());
6261 TOKEN(1); /* force error */
6264 checkcomma(s,PL_tokenbuf,"filehandle");
6274 LOP(OP_SELECT,XTERM);
6280 LOP(OP_SEMCTL,XTERM);
6283 LOP(OP_SEMGET,XTERM);
6286 LOP(OP_SEMOP,XTERM);
6292 LOP(OP_SETPGRP,XTERM);
6294 case KEY_setpriority:
6295 LOP(OP_SETPRIORITY,XTERM);
6297 case KEY_sethostent:
6303 case KEY_setservent:
6306 case KEY_setprotoent:
6316 LOP(OP_SEEKDIR,XTERM);
6318 case KEY_setsockopt:
6319 LOP(OP_SSOCKOPT,XTERM);
6325 LOP(OP_SHMCTL,XTERM);
6328 LOP(OP_SHMGET,XTERM);
6331 LOP(OP_SHMREAD,XTERM);
6334 LOP(OP_SHMWRITE,XTERM);
6337 LOP(OP_SHUTDOWN,XTERM);
6346 LOP(OP_SOCKET,XTERM);
6348 case KEY_socketpair:
6349 LOP(OP_SOCKPAIR,XTERM);
6352 checkcomma(s,PL_tokenbuf,"subroutine name");
6354 if (*s == ';' || *s == ')') /* probably a close */
6355 Perl_croak(aTHX_ "sort is now a reserved word");
6357 s = force_word(s,WORD,TRUE,TRUE,FALSE);
6361 LOP(OP_SPLIT,XTERM);
6364 LOP(OP_SPRINTF,XTERM);
6367 LOP(OP_SPLICE,XTERM);
6382 LOP(OP_SUBSTR,XTERM);
6388 char tmpbuf[sizeof PL_tokenbuf];
6389 SSize_t tboffset = 0;
6390 expectation attrful;
6391 bool have_name, have_proto, bad_proto;
6392 const int key = tmp;
6397 char *tstart = SvPVX(PL_linestr) + PL_realtokenstart;
6398 SV *subtoken = newSVpvn(tstart, s - tstart);
6402 s = SKIPSPACE2(s,tmpwhite);
6407 if (isIDFIRST_lazy_if(s,UTF) || *s == '\'' ||
6408 (*s == ':' && s[1] == ':'))
6415 attrful = XATTRBLOCK;
6416 /* remember buffer pos'n for later force_word */
6417 tboffset = s - PL_oldbufptr;
6418 d = scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
6421 nametoke = newSVpvn(s, d - s);
6423 if (strchr(tmpbuf, ':'))
6424 sv_setpv(PL_subname, tmpbuf);
6426 sv_setsv(PL_subname,PL_curstname);
6427 sv_catpvs(PL_subname,"::");
6428 sv_catpvn(PL_subname,tmpbuf,len);
6435 CURMAD('X', nametoke);
6436 CURMAD('_', tmpwhite);
6437 (void) force_word(PL_oldbufptr + tboffset, WORD,
6440 s = SKIPSPACE2(d,tmpwhite);
6447 Perl_croak(aTHX_ "Missing name in \"my sub\"");
6448 PL_expect = XTERMBLOCK;
6449 attrful = XATTRTERM;
6450 sv_setpvn(PL_subname,"?",1);
6454 if (key == KEY_format) {
6456 PL_lex_formbrack = PL_lex_brackets + 1;
6458 PL_thistoken = subtoken;
6462 (void) force_word(PL_oldbufptr + tboffset, WORD,
6468 /* Look for a prototype */
6472 s = scan_str(s,!!PL_madskills,FALSE);
6474 Perl_croak(aTHX_ "Prototype not terminated");
6475 /* strip spaces and check for bad characters */
6476 d = SvPVX(PL_lex_stuff);
6479 for (p = d; *p; ++p) {
6482 if (!strchr("$@%*;[]&\\", *p))
6487 if (bad_proto && ckWARN(WARN_SYNTAX))
6488 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
6489 "Illegal character in prototype for %"SVf" : %s",
6490 (void*)PL_subname, d);
6491 SvCUR_set(PL_lex_stuff, tmp);
6496 CURMAD('q', PL_thisopen);
6497 CURMAD('_', tmpwhite);
6498 CURMAD('=', PL_thisstuff);
6499 CURMAD('Q', PL_thisclose);
6500 NEXTVAL_NEXTTOKE.opval =
6501 (OP*)newSVOP(OP_CONST, 0, PL_lex_stuff);
6502 PL_lex_stuff = Nullsv;
6505 s = SKIPSPACE2(s,tmpwhite);
6513 if (*s == ':' && s[1] != ':')
6514 PL_expect = attrful;
6515 else if (*s != '{' && key == KEY_sub) {
6517 Perl_croak(aTHX_ "Illegal declaration of anonymous subroutine");
6519 Perl_croak(aTHX_ "Illegal declaration of subroutine %"SVf, (void*)PL_subname);
6526 curmad('^', newSVpvn("",0));
6527 CURMAD('_', tmpwhite);
6531 PL_thistoken = subtoken;
6534 NEXTVAL_NEXTTOKE.opval =
6535 (OP*)newSVOP(OP_CONST, 0, PL_lex_stuff);
6536 PL_lex_stuff = NULL;
6541 sv_setpv(PL_subname,
6542 PL_curstash ? "__ANON__" : "__ANON__::__ANON__");
6546 (void) force_word(PL_oldbufptr + tboffset, WORD,
6556 LOP(OP_SYSTEM,XREF);
6559 LOP(OP_SYMLINK,XTERM);
6562 LOP(OP_SYSCALL,XTERM);
6565 LOP(OP_SYSOPEN,XTERM);
6568 LOP(OP_SYSSEEK,XTERM);
6571 LOP(OP_SYSREAD,XTERM);
6574 LOP(OP_SYSWRITE,XTERM);
6578 TERM(sublex_start());
6599 LOP(OP_TRUNCATE,XTERM);
6611 yylval.ival = CopLINE(PL_curcop);
6615 yylval.ival = CopLINE(PL_curcop);
6619 LOP(OP_UNLINK,XTERM);
6625 LOP(OP_UNPACK,XTERM);
6628 LOP(OP_UTIME,XTERM);
6634 LOP(OP_UNSHIFT,XTERM);
6637 s = tokenize_use(1, s);
6647 yylval.ival = CopLINE(PL_curcop);
6651 yylval.ival = CopLINE(PL_curcop);
6655 PL_hints |= HINT_BLOCK_SCOPE;
6662 LOP(OP_WAITPID,XTERM);
6671 ctl_l[0] = toCTRL('L');
6673 gv_fetchpvn_flags(ctl_l, 1, GV_ADD|GV_NOTQUAL, SVt_PV);
6676 /* Make sure $^L is defined */
6677 gv_fetchpvs("\f", GV_ADD|GV_NOTQUAL, SVt_PV);
6682 if (PL_expect == XOPERATOR)
6688 yylval.ival = OP_XOR;
6693 TERM(sublex_start());
6698 #pragma segment Main
6702 S_pending_ident(pTHX)
6707 /* pit holds the identifier we read and pending_ident is reset */
6708 char pit = PL_pending_ident;
6709 PL_pending_ident = 0;
6711 /* PL_realtokenstart = realtokenend = PL_bufptr - SvPVX(PL_linestr); */
6712 DEBUG_T({ PerlIO_printf(Perl_debug_log,
6713 "### Pending identifier '%s'\n", PL_tokenbuf); });
6715 /* if we're in a my(), we can't allow dynamics here.
6716 $foo'bar has already been turned into $foo::bar, so
6717 just check for colons.
6719 if it's a legal name, the OP is a PADANY.
6722 if (PL_in_my == KEY_our) { /* "our" is merely analogous to "my" */
6723 if (strchr(PL_tokenbuf,':'))
6724 yyerror(Perl_form(aTHX_ "No package name allowed for "
6725 "variable %s in \"our\"",
6727 tmp = allocmy(PL_tokenbuf);
6730 if (strchr(PL_tokenbuf,':'))
6731 yyerror(Perl_form(aTHX_ PL_no_myglob,
6732 PL_in_my == KEY_my ? "my" : "state", PL_tokenbuf));
6734 yylval.opval = newOP(OP_PADANY, 0);
6735 yylval.opval->op_targ = allocmy(PL_tokenbuf);
6741 build the ops for accesses to a my() variable.
6743 Deny my($a) or my($b) in a sort block, *if* $a or $b is
6744 then used in a comparison. This catches most, but not
6745 all cases. For instance, it catches
6746 sort { my($a); $a <=> $b }
6748 sort { my($a); $a < $b ? -1 : $a == $b ? 0 : 1; }
6749 (although why you'd do that is anyone's guess).
6752 if (!strchr(PL_tokenbuf,':')) {
6754 tmp = pad_findmy(PL_tokenbuf);
6755 if (tmp != NOT_IN_PAD) {
6756 /* might be an "our" variable" */
6757 if (PAD_COMPNAME_FLAGS_isOUR(tmp)) {
6758 /* build ops for a bareword */
6759 HV * const stash = PAD_COMPNAME_OURSTASH(tmp);
6760 HEK * const stashname = HvNAME_HEK(stash);
6761 SV * const sym = newSVhek(stashname);
6762 sv_catpvs(sym, "::");
6763 sv_catpv(sym, PL_tokenbuf+1);
6764 yylval.opval = (OP*)newSVOP(OP_CONST, 0, sym);
6765 yylval.opval->op_private = OPpCONST_ENTERED;
6768 ? (GV_ADDMULTI | GV_ADDINEVAL)
6771 ((PL_tokenbuf[0] == '$') ? SVt_PV
6772 : (PL_tokenbuf[0] == '@') ? SVt_PVAV
6777 /* if it's a sort block and they're naming $a or $b */
6778 if (PL_last_lop_op == OP_SORT &&
6779 PL_tokenbuf[0] == '$' &&
6780 (PL_tokenbuf[1] == 'a' || PL_tokenbuf[1] == 'b')
6783 for (d = PL_in_eval ? PL_oldoldbufptr : PL_linestart;
6784 d < PL_bufend && *d != '\n';
6787 if (strnEQ(d,"<=>",3) || strnEQ(d,"cmp",3)) {
6788 Perl_croak(aTHX_ "Can't use \"my %s\" in sort comparison",
6794 yylval.opval = newOP(OP_PADANY, 0);
6795 yylval.opval->op_targ = tmp;
6801 Whine if they've said @foo in a doublequoted string,
6802 and @foo isn't a variable we can find in the symbol
6805 if (pit == '@' && PL_lex_state != LEX_NORMAL && !PL_lex_brackets) {
6806 GV *gv = gv_fetchpv(PL_tokenbuf+1, 0, SVt_PVAV);
6807 if ((!gv || ((PL_tokenbuf[0] == '@') ? !GvAV(gv) : !GvHV(gv)))
6808 && ckWARN(WARN_AMBIGUOUS))
6810 /* Downgraded from fatal to warning 20000522 mjd */
6811 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
6812 "Possible unintended interpolation of %s in string",
6817 /* build ops for a bareword */
6818 yylval.opval = (OP*)newSVOP(OP_CONST, 0, newSVpv(PL_tokenbuf+1, 0));
6819 yylval.opval->op_private = OPpCONST_ENTERED;
6822 /* If the identifier refers to a stash, don't autovivify it.
6823 * Change 24660 had the side effect of causing symbol table
6824 * hashes to always be defined, even if they were freshly
6825 * created and the only reference in the entire program was
6826 * the single statement with the defined %foo::bar:: test.
6827 * It appears that all code in the wild doing this actually
6828 * wants to know whether sub-packages have been loaded, so
6829 * by avoiding auto-vivifying symbol tables, we ensure that
6830 * defined %foo::bar:: continues to be false, and the existing
6831 * tests still give the expected answers, even though what
6832 * they're actually testing has now changed subtly.
6834 (*PL_tokenbuf == '%' && *(d = PL_tokenbuf + strlen(PL_tokenbuf) - 1) == ':' && d[-1] == ':'
6836 : PL_in_eval ? (GV_ADDMULTI | GV_ADDINEVAL) : GV_ADD),
6837 ((PL_tokenbuf[0] == '$') ? SVt_PV
6838 : (PL_tokenbuf[0] == '@') ? SVt_PVAV
6844 * The following code was generated by perl_keyword.pl.
6848 Perl_keyword (pTHX_ const char *name, I32 len)
6853 case 1: /* 5 tokens of length 1 */
6885 case 2: /* 18 tokens of length 2 */
7031 case 3: /* 29 tokens of length 3 */
7035 if (name[1] == 'N' &&
7098 if (name[1] == 'i' &&
7120 return (FEATURE_IS_ENABLED("err") ? -KEY_err : 0);
7138 if (name[1] == 'o' &&
7147 if (name[1] == 'e' &&
7156 if (name[1] == 'n' &&
7165 if (name[1] == 'o' &&
7174 if (name[1] == 'a' &&
7183 if (name[1] == 'o' &&
7245 if (name[1] == 'e' &&
7259 return (FEATURE_IS_ENABLED("say") ? -KEY_say : 0);
7285 if (name[1] == 'i' &&
7294 if (name[1] == 's' &&
7303 if (name[1] == 'e' &&
7312 if (name[1] == 'o' &&
7324 case 4: /* 41 tokens of length 4 */
7328 if (name[1] == 'O' &&
7338 if (name[1] == 'N' &&
7348 if (name[1] == 'i' &&
7358 if (name[1] == 'h' &&
7368 if (name[1] == 'u' &&
7381 if (name[2] == 'c' &&
7390 if (name[2] == 's' &&
7399 if (name[2] == 'a' &&
7435 if (name[1] == 'o' &&
7448 if (name[2] == 't' &&
7457 if (name[2] == 'o' &&
7466 if (name[2] == 't' &&
7475 if (name[2] == 'e' &&
7488 if (name[1] == 'o' &&
7501 if (name[2] == 'y' &&
7510 if (name[2] == 'l' &&
7526 if (name[2] == 's' &&
7535 if (name[2] == 'n' &&
7544 if (name[2] == 'c' &&
7557 if (name[1] == 'e' &&
7567 if (name[1] == 'p' &&
7580 if (name[2] == 'c' &&
7589 if (name[2] == 'p' &&
7598 if (name[2] == 's' &&
7614 if (name[2] == 'n' &&
7684 if (name[2] == 'r' &&
7693 if (name[2] == 'r' &&
7702 if (name[2] == 'a' &&
7718 if (name[2] == 'l' &&
7780 if (name[2] == 'e' &&
7783 return (FEATURE_IS_ENABLED("switch") ? KEY_when : 0);
7796 case 5: /* 39 tokens of length 5 */
7800 if (name[1] == 'E' &&
7811 if (name[1] == 'H' &&
7825 if (name[2] == 'a' &&
7835 if (name[2] == 'a' &&
7852 if (name[2] == 'e' &&
7862 if (name[2] == 'e' &&
7866 return (FEATURE_IS_ENABLED("switch") ? -KEY_break : 0);
7882 if (name[3] == 'i' &&
7891 if (name[3] == 'o' &&
7927 if (name[2] == 'o' &&
7937 if (name[2] == 'y' &&
7951 if (name[1] == 'l' &&
7965 if (name[2] == 'n' &&
7975 if (name[2] == 'o' &&
7989 if (name[1] == 'i' &&
7994 return (FEATURE_IS_ENABLED("switch") ? KEY_given : 0);
8003 if (name[2] == 'd' &&
8013 if (name[2] == 'c' &&
8030 if (name[2] == 'c' &&
8040 if (name[2] == 't' &&
8054 if (name[1] == 'k' &&
8065 if (name[1] == 'r' &&
8079 if (name[2] == 's' &&
8089 if (name[2] == 'd' &&
8106 if (name[2] == 'm' &&
8116 if (name[2] == 'i' &&
8126 if (name[2] == 'e' &&
8136 if (name[2] == 'l' &&
8146 if (name[2] == 'a' &&
8159 if (name[3] == 't' &&
8162 return (FEATURE_IS_ENABLED("state") ? KEY_state : 0);
8168 if (name[3] == 'd' &&
8185 if (name[1] == 'i' &&
8199 if (name[2] == 'a' &&
8212 if (name[3] == 'e' &&
8247 if (name[2] == 'i' &&
8264 if (name[2] == 'i' &&
8274 if (name[2] == 'i' &&
8291 case 6: /* 33 tokens of length 6 */
8295 if (name[1] == 'c' &&
8310 if (name[2] == 'l' &&
8321 if (name[2] == 'r' &&
8336 if (name[1] == 'e' &&
8351 if (name[2] == 's' &&
8356 if(ckWARN_d(WARN_SYNTAX))
8357 Perl_warner(aTHX_ packWARN(WARN_SYNTAX), "elseif should be elsif");
8363 if (name[2] == 'i' &&
8381 if (name[2] == 'l' &&
8392 if (name[2] == 'r' &&
8407 if (name[1] == 'm' &&
8422 if (name[2] == 'n' &&
8433 if (name[2] == 's' &&
8448 if (name[1] == 's' &&
8454 if (name[4] == 't' &&
8463 if (name[4] == 'e' &&
8472 if (name[4] == 'c' &&
8481 if (name[4] == 'n' &&
8497 if (name[1] == 'r' &&
8515 if (name[3] == 'a' &&
8525 if (name[3] == 'u' &&
8539 if (name[2] == 'n' &&
8557 if (name[2] == 'a' &&
8571 if (name[3] == 'e' &&
8584 if (name[4] == 't' &&
8593 if (name[4] == 'e' &&
8615 if (name[4] == 't' &&
8624 if (name[4] == 'e' &&
8640 if (name[2] == 'c' &&
8651 if (name[2] == 'l' &&
8662 if (name[2] == 'b' &&
8673 if (name[2] == 's' &&
8696 if (name[4] == 's' &&
8705 if (name[4] == 'n' &&
8718 if (name[3] == 'a' &&
8735 if (name[1] == 'a' &&
8750 case 7: /* 29 tokens of length 7 */
8754 if (name[1] == 'E' &&
8767 if (name[1] == '_' &&
8780 if (name[1] == 'i' &&
8787 return -KEY_binmode;
8793 if (name[1] == 'o' &&
8800 return -KEY_connect;
8809 if (name[2] == 'm' &&
8815 return -KEY_dbmopen;
8826 if (name[4] == 'u' &&
8830 return (FEATURE_IS_ENABLED("switch") ? KEY_default : 0);
8836 if (name[4] == 'n' &&
8857 if (name[1] == 'o' &&
8870 if (name[1] == 'e' &&
8877 if (name[5] == 'r' &&
8880 return -KEY_getpgrp;
8886 if (name[5] == 'i' &&
8889 return -KEY_getppid;
8902 if (name[1] == 'c' &&
8909 return -KEY_lcfirst;
8915 if (name[1] == 'p' &&
8922 return -KEY_opendir;
8928 if (name[1] == 'a' &&
8946 if (name[3] == 'd' &&
8951 return -KEY_readdir;
8957 if (name[3] == 'u' &&
8968 if (name[3] == 'e' &&
8973 return -KEY_reverse;
8992 if (name[3] == 'k' &&
8997 return -KEY_seekdir;
9003 if (name[3] == 'p' &&
9008 return -KEY_setpgrp;
9018 if (name[2] == 'm' &&
9024 return -KEY_shmread;
9030 if (name[2] == 'r' &&
9036 return -KEY_sprintf;
9045 if (name[3] == 'l' &&
9050 return -KEY_symlink;
9059 if (name[4] == 'a' &&
9063 return -KEY_syscall;
9069 if (name[4] == 'p' &&
9073 return -KEY_sysopen;
9079 if (name[4] == 'e' &&
9083 return -KEY_sysread;
9089 if (name[4] == 'e' &&
9093 return -KEY_sysseek;
9111 if (name[1] == 'e' &&
9118 return -KEY_telldir;
9127 if (name[2] == 'f' &&
9133 return -KEY_ucfirst;
9139 if (name[2] == 's' &&
9145 return -KEY_unshift;
9155 if (name[1] == 'a' &&
9162 return -KEY_waitpid;
9171 case 8: /* 26 tokens of length 8 */
9175 if (name[1] == 'U' &&
9183 return KEY_AUTOLOAD;
9194 if (name[3] == 'A' &&
9200 return KEY___DATA__;
9206 if (name[3] == 'I' &&
9212 return -KEY___FILE__;
9218 if (name[3] == 'I' &&
9224 return -KEY___LINE__;
9240 if (name[2] == 'o' &&
9247 return -KEY_closedir;
9253 if (name[2] == 'n' &&
9260 return -KEY_continue;
9270 if (name[1] == 'b' &&
9278 return -KEY_dbmclose;
9284 if (name[1] == 'n' &&
9290 if (name[4] == 'r' &&
9295 return -KEY_endgrent;
9301 if (name[4] == 'w' &&
9306 return -KEY_endpwent;
9319 if (name[1] == 'o' &&
9327 return -KEY_formline;
9333 if (name[1] == 'e' &&
9344 if (name[6] == 'n' &&
9347 return -KEY_getgrent;
9353 if (name[6] == 'i' &&
9356 return -KEY_getgrgid;
9362 if (name[6] == 'a' &&
9365 return -KEY_getgrnam;
9378 if (name[4] == 'o' &&
9383 return -KEY_getlogin;
9394 if (name[6] == 'n' &&
9397 return -KEY_getpwent;
9403 if (name[6] == 'a' &&
9406 return -KEY_getpwnam;
9412 if (name[6] == 'i' &&
9415 return -KEY_getpwuid;
9435 if (name[1] == 'e' &&
9442 if (name[5] == 'i' &&
9449 return -KEY_readline;
9454 return -KEY_readlink;
9465 if (name[5] == 'i' &&
9469 return -KEY_readpipe;
9490 if (name[4] == 'r' &&
9495 return -KEY_setgrent;
9501 if (name[4] == 'w' &&
9506 return -KEY_setpwent;
9522 if (name[3] == 'w' &&
9528 return -KEY_shmwrite;
9534 if (name[3] == 't' &&
9540 return -KEY_shutdown;
9550 if (name[2] == 's' &&
9557 return -KEY_syswrite;
9567 if (name[1] == 'r' &&
9575 return -KEY_truncate;
9584 case 9: /* 8 tokens of length 9 */
9588 if (name[1] == 'n' &&
9597 return -KEY_endnetent;
9603 if (name[1] == 'e' &&
9612 return -KEY_getnetent;
9618 if (name[1] == 'o' &&
9627 return -KEY_localtime;
9633 if (name[1] == 'r' &&
9642 return KEY_prototype;
9648 if (name[1] == 'u' &&
9657 return -KEY_quotemeta;
9663 if (name[1] == 'e' &&
9672 return -KEY_rewinddir;
9678 if (name[1] == 'e' &&
9687 return -KEY_setnetent;
9693 if (name[1] == 'a' &&
9702 return -KEY_wantarray;
9711 case 10: /* 9 tokens of length 10 */
9715 if (name[1] == 'n' &&
9721 if (name[4] == 'o' &&
9728 return -KEY_endhostent;
9734 if (name[4] == 'e' &&
9741 return -KEY_endservent;
9754 if (name[1] == 'e' &&
9760 if (name[4] == 'o' &&
9767 return -KEY_gethostent;
9776 if (name[5] == 'r' &&
9782 return -KEY_getservent;
9788 if (name[5] == 'c' &&
9794 return -KEY_getsockopt;
9819 if (name[4] == 'o' &&
9826 return -KEY_sethostent;
9835 if (name[5] == 'r' &&
9841 return -KEY_setservent;
9847 if (name[5] == 'c' &&
9853 return -KEY_setsockopt;
9870 if (name[2] == 'c' &&
9879 return -KEY_socketpair;
9892 case 11: /* 8 tokens of length 11 */
9896 if (name[1] == '_' &&
9907 return -KEY___PACKAGE__;
9913 if (name[1] == 'n' &&
9924 return -KEY_endprotoent;
9930 if (name[1] == 'e' &&
9939 if (name[5] == 'e' &&
9946 return -KEY_getpeername;
9955 if (name[6] == 'o' &&
9961 return -KEY_getpriority;
9967 if (name[6] == 't' &&
9973 return -KEY_getprotoent;
9987 if (name[4] == 'o' &&
9995 return -KEY_getsockname;
10008 if (name[1] == 'e' &&
10016 if (name[6] == 'o' &&
10021 { /* setpriority */
10022 return -KEY_setpriority;
10028 if (name[6] == 't' &&
10033 { /* setprotoent */
10034 return -KEY_setprotoent;
10050 case 12: /* 2 tokens of length 12 */
10051 if (name[0] == 'g' &&
10063 if (name[9] == 'd' &&
10066 { /* getnetbyaddr */
10067 return -KEY_getnetbyaddr;
10073 if (name[9] == 'a' &&
10076 { /* getnetbyname */
10077 return -KEY_getnetbyname;
10089 case 13: /* 4 tokens of length 13 */
10090 if (name[0] == 'g' &&
10097 if (name[4] == 'o' &&
10106 if (name[10] == 'd' &&
10109 { /* gethostbyaddr */
10110 return -KEY_gethostbyaddr;
10116 if (name[10] == 'a' &&
10119 { /* gethostbyname */
10120 return -KEY_gethostbyname;
10133 if (name[4] == 'e' &&
10142 if (name[10] == 'a' &&
10145 { /* getservbyname */
10146 return -KEY_getservbyname;
10152 if (name[10] == 'o' &&
10155 { /* getservbyport */
10156 return -KEY_getservbyport;
10175 case 14: /* 1 tokens of length 14 */
10176 if (name[0] == 'g' &&
10190 { /* getprotobyname */
10191 return -KEY_getprotobyname;
10196 case 16: /* 1 tokens of length 16 */
10197 if (name[0] == 'g' &&
10213 { /* getprotobynumber */
10214 return -KEY_getprotobynumber;
10228 S_checkcomma(pTHX_ const char *s, const char *name, const char *what)
10232 if (*s == ' ' && s[1] == '(') { /* XXX gotta be a better way */
10233 if (ckWARN(WARN_SYNTAX)) {
10236 for (w = s+2; *w && level; w++) {
10239 else if (*w == ')')
10242 while (isSPACE(*w))
10244 if (!*w || !strchr(";|})]oaiuw!=", *w)) /* an advisory hack only... */
10245 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
10246 "%s (...) interpreted as function",name);
10249 while (s < PL_bufend && isSPACE(*s))
10253 while (s < PL_bufend && isSPACE(*s))
10255 if (isIDFIRST_lazy_if(s,UTF)) {
10256 const char * const w = s++;
10257 while (isALNUM_lazy_if(s,UTF))
10259 while (s < PL_bufend && isSPACE(*s))
10263 if (keyword(w, s - w))
10266 gv = gv_fetchpvn_flags(w, s - w, 0, SVt_PVCV);
10267 if (gv && GvCVu(gv))
10269 Perl_croak(aTHX_ "No comma allowed after %s", what);
10274 /* Either returns sv, or mortalizes sv and returns a new SV*.
10275 Best used as sv=new_constant(..., sv, ...).
10276 If s, pv are NULL, calls subroutine with one argument,
10277 and type is used with error messages only. */
10280 S_new_constant(pTHX_ const char *s, STRLEN len, const char *key, SV *sv, SV *pv,
10284 HV * const table = GvHV(PL_hintgv); /* ^H */
10288 const char *why1 = "", *why2 = "", *why3 = "";
10290 if (!table || !(PL_hints & HINT_LOCALIZE_HH)) {
10293 why2 = strEQ(key,"charnames")
10294 ? "(possibly a missing \"use charnames ...\")"
10296 msg = Perl_newSVpvf(aTHX_ "Constant(%s) unknown: %s",
10297 (type ? type: "undef"), why2);
10299 /* This is convoluted and evil ("goto considered harmful")
10300 * but I do not understand the intricacies of all the different
10301 * failure modes of %^H in here. The goal here is to make
10302 * the most probable error message user-friendly. --jhi */
10307 msg = Perl_newSVpvf(aTHX_ "Constant(%s): %s%s%s",
10308 (type ? type: "undef"), why1, why2, why3);
10310 yyerror(SvPVX_const(msg));
10314 cvp = hv_fetch(table, key, strlen(key), FALSE);
10315 if (!cvp || !SvOK(*cvp)) {
10318 why3 = "} is not defined";
10321 sv_2mortal(sv); /* Parent created it permanently */
10324 pv = sv_2mortal(newSVpvn(s, len));
10326 typesv = sv_2mortal(newSVpv(type, 0));
10328 typesv = &PL_sv_undef;
10330 PUSHSTACKi(PERLSI_OVERLOAD);
10342 call_sv(cv, G_SCALAR | ( PL_in_eval ? 0 : G_EVAL));
10346 /* Check the eval first */
10347 if (!PL_in_eval && SvTRUE(ERRSV)) {
10348 sv_catpvs(ERRSV, "Propagated");
10349 yyerror(SvPV_nolen_const(ERRSV)); /* Duplicates the message inside eval */
10351 res = SvREFCNT_inc_simple(sv);
10355 SvREFCNT_inc_simple_void(res);
10364 why1 = "Call to &{$^H{";
10366 why3 = "}} did not return a defined value";
10374 /* Returns a NUL terminated string, with the length of the string written to
10378 S_scan_word(pTHX_ register char *s, char *dest, STRLEN destlen, int allow_package, STRLEN *slp)
10381 register char *d = dest;
10382 register char * const e = d + destlen - 3; /* two-character token, ending NUL */
10385 Perl_croak(aTHX_ ident_too_long);
10386 if (isALNUM(*s)) /* UTF handled below */
10388 else if (allow_package && (*s == '\'') && isIDFIRST_lazy_if(s+1,UTF)) {
10393 else if (allow_package && (s[0] == ':') && (s[1] == ':') && (s[2] != '$')) {
10397 else if (UTF && UTF8_IS_START(*s) && isALNUM_utf8((U8*)s)) {
10398 char *t = s + UTF8SKIP(s);
10400 while (UTF8_IS_CONTINUED(*t) && is_utf8_mark((U8*)t))
10404 Perl_croak(aTHX_ ident_too_long);
10405 Copy(s, d, len, char);
10418 S_scan_ident(pTHX_ register char *s, register const char *send, char *dest, STRLEN destlen, I32 ck_uni)
10421 char *bracket = NULL;
10423 register char *d = dest;
10424 register char * const e = d + destlen + 3; /* two-character token, ending NUL */
10429 while (isDIGIT(*s)) {
10431 Perl_croak(aTHX_ ident_too_long);
10438 Perl_croak(aTHX_ ident_too_long);
10439 if (isALNUM(*s)) /* UTF handled below */
10441 else if (*s == '\'' && isIDFIRST_lazy_if(s+1,UTF)) {
10446 else if (*s == ':' && s[1] == ':') {
10450 else if (UTF && UTF8_IS_START(*s) && isALNUM_utf8((U8*)s)) {
10451 char *t = s + UTF8SKIP(s);
10452 while (UTF8_IS_CONTINUED(*t) && is_utf8_mark((U8*)t))
10454 if (d + (t - s) > e)
10455 Perl_croak(aTHX_ ident_too_long);
10456 Copy(s, d, t - s, char);
10467 if (PL_lex_state != LEX_NORMAL)
10468 PL_lex_state = LEX_INTERPENDMAYBE;
10471 if (*s == '$' && s[1] &&
10472 (isALNUM_lazy_if(s+1,UTF) || s[1] == '$' || s[1] == '{' || strnEQ(s+1,"::",2)) )
10485 if (*d == '^' && *s && isCONTROLVAR(*s)) {
10490 if (isSPACE(s[-1])) {
10492 const char ch = *s++;
10493 if (!SPACE_OR_TAB(ch)) {
10499 if (isIDFIRST_lazy_if(d,UTF)) {
10503 while ((end < send && isALNUM_lazy_if(end,UTF)) || *end == ':') {
10504 end += UTF8SKIP(end);
10505 while (end < send && UTF8_IS_CONTINUED(*end) && is_utf8_mark((U8*)end))
10506 end += UTF8SKIP(end);
10508 Copy(s, d, end - s, char);
10513 while ((isALNUM(*s) || *s == ':') && d < e)
10516 Perl_croak(aTHX_ ident_too_long);
10519 while (s < send && SPACE_OR_TAB(*s))
10521 if ((*s == '[' || (*s == '{' && strNE(dest, "sub")))) {
10522 if (ckWARN(WARN_AMBIGUOUS) && keyword(dest, d - dest)) {
10523 const char * const brack = (*s == '[') ? "[...]" : "{...}";
10524 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
10525 "Ambiguous use of %c{%s%s} resolved to %c%s%s",
10526 funny, dest, brack, funny, dest, brack);
10529 PL_lex_brackstack[PL_lex_brackets++] = (char)(XOPERATOR | XFAKEBRACK);
10533 /* Handle extended ${^Foo} variables
10534 * 1999-02-27 mjd-perl-patch@plover.com */
10535 else if (!isALNUM(*d) && !isPRINT(*d) /* isCTRL(d) */
10539 while (isALNUM(*s) && d < e) {
10543 Perl_croak(aTHX_ ident_too_long);
10548 if (PL_lex_state == LEX_INTERPNORMAL && !PL_lex_brackets) {
10549 PL_lex_state = LEX_INTERPEND;
10552 if (PL_lex_state == LEX_NORMAL) {
10553 if (ckWARN(WARN_AMBIGUOUS) &&
10554 (keyword(dest, d - dest) || get_cv(dest, FALSE)))
10558 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
10559 "Ambiguous use of %c{%s} resolved to %c%s",
10560 funny, dest, funny, dest);
10565 s = bracket; /* let the parser handle it */
10569 else if (PL_lex_state == LEX_INTERPNORMAL && !PL_lex_brackets && !intuit_more(s))
10570 PL_lex_state = LEX_INTERPEND;
10575 Perl_pmflag(pTHX_ U32* pmfl, int ch)
10577 PERL_UNUSED_CONTEXT;
10580 else if (ch == 'g')
10581 *pmfl |= PMf_GLOBAL;
10582 else if (ch == 'c')
10583 *pmfl |= PMf_CONTINUE;
10584 else if (ch == 'o')
10586 else if (ch == 'm')
10587 *pmfl |= PMf_MULTILINE;
10588 else if (ch == 's')
10589 *pmfl |= PMf_SINGLELINE;
10590 else if (ch == 'x')
10591 *pmfl |= PMf_EXTENDED;
10595 S_scan_pat(pTHX_ char *start, I32 type)
10599 char *s = scan_str(start,!!PL_madskills,FALSE);
10600 const char * const valid_flags = (type == OP_QR) ? "iomsx" : "iogcmsx";
10607 const char * const delimiter = skipspace(start);
10608 Perl_croak(aTHX_ *delimiter == '?'
10609 ? "Search pattern not terminated or ternary operator parsed as search pattern"
10610 : "Search pattern not terminated" );
10613 pm = (PMOP*)newPMOP(type, 0);
10614 if (PL_multi_open == '?')
10615 pm->op_pmflags |= PMf_ONCE;
10619 while (*s && strchr(valid_flags, *s))
10620 pmflag(&pm->op_pmflags,*s++);
10622 if (PL_madskills && modstart != s) {
10623 SV* tmptoken = newSVpvn(modstart, s - modstart);
10624 append_madprops(newMADPROP('m', MAD_SV, tmptoken, 0), (OP*)pm, 0);
10627 /* issue a warning if /c is specified,but /g is not */
10628 if ((pm->op_pmflags & PMf_CONTINUE) && !(pm->op_pmflags & PMf_GLOBAL)
10629 && ckWARN(WARN_REGEXP))
10631 Perl_warner(aTHX_ packWARN(WARN_REGEXP), "Use of /c modifier is meaningless without /g" );
10634 pm->op_pmpermflags = pm->op_pmflags;
10636 PL_lex_op = (OP*)pm;
10637 yylval.ival = OP_MATCH;
10642 S_scan_subst(pTHX_ char *start)
10653 yylval.ival = OP_NULL;
10655 s = scan_str(start,!!PL_madskills,FALSE);
10658 Perl_croak(aTHX_ "Substitution pattern not terminated");
10660 if (s[-1] == PL_multi_open)
10663 if (PL_madskills) {
10664 CURMAD('q', PL_thisopen);
10665 CURMAD('_', PL_thiswhite);
10666 CURMAD('E', PL_thisstuff);
10667 CURMAD('Q', PL_thisclose);
10668 PL_realtokenstart = s - SvPVX(PL_linestr);
10672 first_start = PL_multi_start;
10673 s = scan_str(s,!!PL_madskills,FALSE);
10675 if (PL_lex_stuff) {
10676 SvREFCNT_dec(PL_lex_stuff);
10677 PL_lex_stuff = NULL;
10679 Perl_croak(aTHX_ "Substitution replacement not terminated");
10681 PL_multi_start = first_start; /* so whole substitution is taken together */
10683 pm = (PMOP*)newPMOP(OP_SUBST, 0);
10686 if (PL_madskills) {
10687 CURMAD('z', PL_thisopen);
10688 CURMAD('R', PL_thisstuff);
10689 CURMAD('Z', PL_thisclose);
10699 else if (strchr("iogcmsx", *s))
10700 pmflag(&pm->op_pmflags,*s++);
10706 if (PL_madskills) {
10708 curmad('m', newSVpvn(modstart, s - modstart));
10709 append_madprops(PL_thismad, (OP*)pm, 0);
10713 if ((pm->op_pmflags & PMf_CONTINUE) && ckWARN(WARN_REGEXP)) {
10714 Perl_warner(aTHX_ packWARN(WARN_REGEXP), "Use of /c modifier is meaningless in s///" );
10718 SV * const repl = newSVpvs("");
10720 PL_sublex_info.super_bufptr = s;
10721 PL_sublex_info.super_bufend = PL_bufend;
10723 pm->op_pmflags |= PMf_EVAL;
10725 sv_catpv(repl, es ? "eval " : "do ");
10726 sv_catpvs(repl, "{");
10727 sv_catsv(repl, PL_lex_repl);
10728 if (strchr(SvPVX(PL_lex_repl), '#'))
10729 sv_catpvs(repl, "\n");
10730 sv_catpvs(repl, "}");
10732 SvREFCNT_dec(PL_lex_repl);
10733 PL_lex_repl = repl;
10736 pm->op_pmpermflags = pm->op_pmflags;
10737 PL_lex_op = (OP*)pm;
10738 yylval.ival = OP_SUBST;
10743 S_scan_trans(pTHX_ char *start)
10756 yylval.ival = OP_NULL;
10758 s = scan_str(start,!!PL_madskills,FALSE);
10760 Perl_croak(aTHX_ "Transliteration pattern not terminated");
10762 if (s[-1] == PL_multi_open)
10765 if (PL_madskills) {
10766 CURMAD('q', PL_thisopen);
10767 CURMAD('_', PL_thiswhite);
10768 CURMAD('E', PL_thisstuff);
10769 CURMAD('Q', PL_thisclose);
10770 PL_realtokenstart = s - SvPVX(PL_linestr);
10774 s = scan_str(s,!!PL_madskills,FALSE);
10776 if (PL_lex_stuff) {
10777 SvREFCNT_dec(PL_lex_stuff);
10778 PL_lex_stuff = NULL;
10780 Perl_croak(aTHX_ "Transliteration replacement not terminated");
10782 if (PL_madskills) {
10783 CURMAD('z', PL_thisopen);
10784 CURMAD('R', PL_thisstuff);
10785 CURMAD('Z', PL_thisclose);
10788 complement = del = squash = 0;
10795 complement = OPpTRANS_COMPLEMENT;
10798 del = OPpTRANS_DELETE;
10801 squash = OPpTRANS_SQUASH;
10810 Newx(tbl, complement&&!del?258:256, short);
10811 o = newPVOP(OP_TRANS, 0, (char*)tbl);
10812 o->op_private &= ~OPpTRANS_ALL;
10813 o->op_private |= del|squash|complement|
10814 (DO_UTF8(PL_lex_stuff)? OPpTRANS_FROM_UTF : 0)|
10815 (DO_UTF8(PL_lex_repl) ? OPpTRANS_TO_UTF : 0);
10818 yylval.ival = OP_TRANS;
10821 if (PL_madskills) {
10823 curmad('m', newSVpvn(modstart, s - modstart));
10824 append_madprops(PL_thismad, o, 0);
10833 S_scan_heredoc(pTHX_ register char *s)
10837 I32 op_type = OP_SCALAR;
10841 const char *found_newline;
10845 const int outer = (PL_rsfp && !(PL_lex_inwhat == OP_SCALAR));
10847 I32 stuffstart = s - SvPVX(PL_linestr);
10850 PL_realtokenstart = -1;
10855 e = PL_tokenbuf + sizeof PL_tokenbuf - 1;
10859 while (SPACE_OR_TAB(*peek))
10861 if (*peek == '`' || *peek == '\'' || *peek =='"') {
10864 s = delimcpy(d, e, s, PL_bufend, term, &len);
10874 if (!isALNUM_lazy_if(s,UTF))
10875 deprecate_old("bare << to mean <<\"\"");
10876 for (; isALNUM_lazy_if(s,UTF); s++) {
10881 if (d >= PL_tokenbuf + sizeof PL_tokenbuf - 1)
10882 Perl_croak(aTHX_ "Delimiter for here document is too long");
10885 len = d - PL_tokenbuf;
10888 if (PL_madskills) {
10889 tstart = PL_tokenbuf + !outer;
10890 PL_thisclose = newSVpvn(tstart, len - !outer);
10891 tstart = SvPVX(PL_linestr) + stuffstart;
10892 PL_thisopen = newSVpvn(tstart, s - tstart);
10893 stuffstart = s - SvPVX(PL_linestr);
10896 #ifndef PERL_STRICT_CR
10897 d = strchr(s, '\r');
10899 char * const olds = s;
10901 while (s < PL_bufend) {
10907 else if (*s == '\n' && s[1] == '\r') { /* \015\013 on a mac? */
10916 SvCUR_set(PL_linestr, PL_bufend - SvPVX_const(PL_linestr));
10923 if ( outer || !(found_newline = memchr(s, '\n', PL_bufend - s)) ) {
10924 herewas = newSVpvn(s,PL_bufend-s);
10928 herewas = newSVpvn(s-1,found_newline-s+1);
10931 herewas = newSVpvn(s,found_newline-s);
10935 if (PL_madskills) {
10936 tstart = SvPVX(PL_linestr) + stuffstart;
10938 sv_catpvn(PL_thisstuff, tstart, s - tstart);
10940 PL_thisstuff = newSVpvn(tstart, s - tstart);
10943 s += SvCUR(herewas);
10946 stuffstart = s - SvPVX(PL_linestr);
10952 tmpstr = newSV(79);
10953 sv_upgrade(tmpstr, SVt_PVIV);
10954 if (term == '\'') {
10955 op_type = OP_CONST;
10956 SvIV_set(tmpstr, -1);
10958 else if (term == '`') {
10959 op_type = OP_BACKTICK;
10960 SvIV_set(tmpstr, '\\');
10964 PL_multi_start = CopLINE(PL_curcop);
10965 PL_multi_open = PL_multi_close = '<';
10966 term = *PL_tokenbuf;
10967 if (PL_lex_inwhat == OP_SUBST && PL_in_eval && !PL_rsfp) {
10968 char * const bufptr = PL_sublex_info.super_bufptr;
10969 char * const bufend = PL_sublex_info.super_bufend;
10970 char * const olds = s - SvCUR(herewas);
10971 s = strchr(bufptr, '\n');
10975 while (s < bufend &&
10976 (*s != term || memNE(s,PL_tokenbuf,len)) ) {
10978 CopLINE_inc(PL_curcop);
10981 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
10982 missingterm(PL_tokenbuf);
10984 sv_setpvn(herewas,bufptr,d-bufptr+1);
10985 sv_setpvn(tmpstr,d+1,s-d);
10987 sv_catpvn(herewas,s,bufend-s);
10988 Copy(SvPVX_const(herewas),bufptr,SvCUR(herewas) + 1,char);
10995 while (s < PL_bufend &&
10996 (*s != term || memNE(s,PL_tokenbuf,len)) ) {
10998 CopLINE_inc(PL_curcop);
11000 if (s >= PL_bufend) {
11001 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
11002 missingterm(PL_tokenbuf);
11004 sv_setpvn(tmpstr,d+1,s-d);
11006 if (PL_madskills) {
11008 sv_catpvn(PL_thisstuff, d + 1, s - d);
11010 PL_thisstuff = newSVpvn(d + 1, s - d);
11011 stuffstart = s - SvPVX(PL_linestr);
11015 CopLINE_inc(PL_curcop); /* the preceding stmt passes a newline */
11017 sv_catpvn(herewas,s,PL_bufend-s);
11018 sv_setsv(PL_linestr,herewas);
11019 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = s = PL_linestart = SvPVX(PL_linestr);
11020 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
11021 PL_last_lop = PL_last_uni = NULL;
11024 sv_setpvn(tmpstr,"",0); /* avoid "uninitialized" warning */
11025 while (s >= PL_bufend) { /* multiple line string? */
11027 if (PL_madskills) {
11028 tstart = SvPVX(PL_linestr) + stuffstart;
11030 sv_catpvn(PL_thisstuff, tstart, PL_bufend - tstart);
11032 PL_thisstuff = newSVpvn(tstart, PL_bufend - tstart);
11036 !(PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = filter_gets(PL_linestr, PL_rsfp, 0))) {
11037 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
11038 missingterm(PL_tokenbuf);
11041 stuffstart = s - SvPVX(PL_linestr);
11043 CopLINE_inc(PL_curcop);
11044 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
11045 PL_last_lop = PL_last_uni = NULL;
11046 #ifndef PERL_STRICT_CR
11047 if (PL_bufend - PL_linestart >= 2) {
11048 if ((PL_bufend[-2] == '\r' && PL_bufend[-1] == '\n') ||
11049 (PL_bufend[-2] == '\n' && PL_bufend[-1] == '\r'))
11051 PL_bufend[-2] = '\n';
11053 SvCUR_set(PL_linestr, PL_bufend - SvPVX_const(PL_linestr));
11055 else if (PL_bufend[-1] == '\r')
11056 PL_bufend[-1] = '\n';
11058 else if (PL_bufend - PL_linestart == 1 && PL_bufend[-1] == '\r')
11059 PL_bufend[-1] = '\n';
11061 if (PERLDB_LINE && PL_curstash != PL_debstash) {
11062 SV * const sv = newSV(0);
11064 sv_upgrade(sv, SVt_PVMG);
11065 sv_setsv(sv,PL_linestr);
11066 (void)SvIOK_on(sv);
11068 av_store(CopFILEAVx(PL_curcop), (I32)CopLINE(PL_curcop),sv);
11070 if (*s == term && memEQ(s,PL_tokenbuf,len)) {
11071 STRLEN off = PL_bufend - 1 - SvPVX_const(PL_linestr);
11072 *(SvPVX(PL_linestr) + off ) = ' ';
11073 sv_catsv(PL_linestr,herewas);
11074 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
11075 s = SvPVX(PL_linestr) + off; /* In case PV of PL_linestr moved. */
11079 sv_catsv(tmpstr,PL_linestr);
11084 PL_multi_end = CopLINE(PL_curcop);
11085 if (SvCUR(tmpstr) + 5 < SvLEN(tmpstr)) {
11086 SvPV_shrink_to_cur(tmpstr);
11088 SvREFCNT_dec(herewas);
11090 if (UTF && is_utf8_string((U8*)SvPVX_const(tmpstr), SvCUR(tmpstr)))
11092 else if (PL_encoding)
11093 sv_recode_to_utf8(tmpstr, PL_encoding);
11095 PL_lex_stuff = tmpstr;
11096 yylval.ival = op_type;
11100 /* scan_inputsymbol
11101 takes: current position in input buffer
11102 returns: new position in input buffer
11103 side-effects: yylval and lex_op are set.
11108 <FH> read from filehandle
11109 <pkg::FH> read from package qualified filehandle
11110 <pkg'FH> read from package qualified filehandle
11111 <$fh> read from filehandle in $fh
11112 <*.h> filename glob
11117 S_scan_inputsymbol(pTHX_ char *start)
11120 register char *s = start; /* current position in buffer */
11124 char *d = PL_tokenbuf; /* start of temp holding space */
11125 const char * const e = PL_tokenbuf + sizeof PL_tokenbuf; /* end of temp holding space */
11127 end = strchr(s, '\n');
11130 s = delimcpy(d, e, s + 1, end, '>', &len); /* extract until > */
11132 /* die if we didn't have space for the contents of the <>,
11133 or if it didn't end, or if we see a newline
11136 if (len >= (I32)sizeof PL_tokenbuf)
11137 Perl_croak(aTHX_ "Excessively long <> operator");
11139 Perl_croak(aTHX_ "Unterminated <> operator");
11144 Remember, only scalar variables are interpreted as filehandles by
11145 this code. Anything more complex (e.g., <$fh{$num}>) will be
11146 treated as a glob() call.
11147 This code makes use of the fact that except for the $ at the front,
11148 a scalar variable and a filehandle look the same.
11150 if (*d == '$' && d[1]) d++;
11152 /* allow <Pkg'VALUE> or <Pkg::VALUE> */
11153 while (*d && (isALNUM_lazy_if(d,UTF) || *d == '\'' || *d == ':'))
11156 /* If we've tried to read what we allow filehandles to look like, and
11157 there's still text left, then it must be a glob() and not a getline.
11158 Use scan_str to pull out the stuff between the <> and treat it
11159 as nothing more than a string.
11162 if (d - PL_tokenbuf != len) {
11163 yylval.ival = OP_GLOB;
11165 s = scan_str(start,!!PL_madskills,FALSE);
11167 Perl_croak(aTHX_ "Glob not terminated");
11171 bool readline_overriden = FALSE;
11174 /* we're in a filehandle read situation */
11177 /* turn <> into <ARGV> */
11179 Copy("ARGV",d,5,char);
11181 /* Check whether readline() is overriden */
11182 gv_readline = gv_fetchpvs("readline", GV_NOTQUAL, SVt_PVCV);
11184 && GvCVu(gv_readline) && GvIMPORTED_CV(gv_readline))
11186 ((gvp = (GV**)hv_fetchs(PL_globalstash, "readline", FALSE))
11187 && (gv_readline = *gvp) != (GV*)&PL_sv_undef
11188 && GvCVu(gv_readline) && GvIMPORTED_CV(gv_readline)))
11189 readline_overriden = TRUE;
11191 /* if <$fh>, create the ops to turn the variable into a
11195 /* try to find it in the pad for this block, otherwise find
11196 add symbol table ops
11198 const PADOFFSET tmp = pad_findmy(d);
11199 if (tmp != NOT_IN_PAD) {
11200 if (PAD_COMPNAME_FLAGS_isOUR(tmp)) {
11201 HV * const stash = PAD_COMPNAME_OURSTASH(tmp);
11202 HEK * const stashname = HvNAME_HEK(stash);
11203 SV * const sym = sv_2mortal(newSVhek(stashname));
11204 sv_catpvs(sym, "::");
11205 sv_catpv(sym, d+1);
11210 OP * const o = newOP(OP_PADSV, 0);
11212 PL_lex_op = readline_overriden
11213 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
11214 append_elem(OP_LIST, o,
11215 newCVREF(0, newGVOP(OP_GV,0,gv_readline))))
11216 : (OP*)newUNOP(OP_READLINE, 0, o);
11225 ? (GV_ADDMULTI | GV_ADDINEVAL)
11228 PL_lex_op = readline_overriden
11229 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
11230 append_elem(OP_LIST,
11231 newUNOP(OP_RV2SV, 0, newGVOP(OP_GV, 0, gv)),
11232 newCVREF(0, newGVOP(OP_GV, 0, gv_readline))))
11233 : (OP*)newUNOP(OP_READLINE, 0,
11234 newUNOP(OP_RV2SV, 0,
11235 newGVOP(OP_GV, 0, gv)));
11237 if (!readline_overriden)
11238 PL_lex_op->op_flags |= OPf_SPECIAL;
11239 /* we created the ops in PL_lex_op, so make yylval.ival a null op */
11240 yylval.ival = OP_NULL;
11243 /* If it's none of the above, it must be a literal filehandle
11244 (<Foo::BAR> or <FOO>) so build a simple readline OP */
11246 GV * const gv = gv_fetchpv(d, GV_ADD, SVt_PVIO);
11247 PL_lex_op = readline_overriden
11248 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
11249 append_elem(OP_LIST,
11250 newGVOP(OP_GV, 0, gv),
11251 newCVREF(0, newGVOP(OP_GV, 0, gv_readline))))
11252 : (OP*)newUNOP(OP_READLINE, 0, newGVOP(OP_GV, 0, gv));
11253 yylval.ival = OP_NULL;
11262 takes: start position in buffer
11263 keep_quoted preserve \ on the embedded delimiter(s)
11264 keep_delims preserve the delimiters around the string
11265 returns: position to continue reading from buffer
11266 side-effects: multi_start, multi_close, lex_repl or lex_stuff, and
11267 updates the read buffer.
11269 This subroutine pulls a string out of the input. It is called for:
11270 q single quotes q(literal text)
11271 ' single quotes 'literal text'
11272 qq double quotes qq(interpolate $here please)
11273 " double quotes "interpolate $here please"
11274 qx backticks qx(/bin/ls -l)
11275 ` backticks `/bin/ls -l`
11276 qw quote words @EXPORT_OK = qw( func() $spam )
11277 m// regexp match m/this/
11278 s/// regexp substitute s/this/that/
11279 tr/// string transliterate tr/this/that/
11280 y/// string transliterate y/this/that/
11281 ($*@) sub prototypes sub foo ($)
11282 (stuff) sub attr parameters sub foo : attr(stuff)
11283 <> readline or globs <FOO>, <>, <$fh>, or <*.c>
11285 In most of these cases (all but <>, patterns and transliterate)
11286 yylex() calls scan_str(). m// makes yylex() call scan_pat() which
11287 calls scan_str(). s/// makes yylex() call scan_subst() which calls
11288 scan_str(). tr/// and y/// make yylex() call scan_trans() which
11291 It skips whitespace before the string starts, and treats the first
11292 character as the delimiter. If the delimiter is one of ([{< then
11293 the corresponding "close" character )]}> is used as the closing
11294 delimiter. It allows quoting of delimiters, and if the string has
11295 balanced delimiters ([{<>}]) it allows nesting.
11297 On success, the SV with the resulting string is put into lex_stuff or,
11298 if that is already non-NULL, into lex_repl. The second case occurs only
11299 when parsing the RHS of the special constructs s/// and tr/// (y///).
11300 For convenience, the terminating delimiter character is stuffed into
11305 S_scan_str(pTHX_ char *start, int keep_quoted, int keep_delims)
11308 SV *sv; /* scalar value: string */
11309 char *tmps; /* temp string, used for delimiter matching */
11310 register char *s = start; /* current position in the buffer */
11311 register char term; /* terminating character */
11312 register char *to; /* current position in the sv's data */
11313 I32 brackets = 1; /* bracket nesting level */
11314 bool has_utf8 = FALSE; /* is there any utf8 content? */
11315 I32 termcode; /* terminating char. code */
11316 U8 termstr[UTF8_MAXBYTES]; /* terminating string */
11317 STRLEN termlen; /* length of terminating string */
11318 char *last = NULL; /* last position for nesting bracket */
11324 /* skip space before the delimiter */
11330 if (PL_realtokenstart >= 0) {
11331 stuffstart = PL_realtokenstart;
11332 PL_realtokenstart = -1;
11335 stuffstart = start - SvPVX(PL_linestr);
11337 /* mark where we are, in case we need to report errors */
11340 /* after skipping whitespace, the next character is the terminator */
11343 termcode = termstr[0] = term;
11347 termcode = utf8_to_uvchr((U8*)s, &termlen);
11348 Copy(s, termstr, termlen, U8);
11349 if (!UTF8_IS_INVARIANT(term))
11353 /* mark where we are */
11354 PL_multi_start = CopLINE(PL_curcop);
11355 PL_multi_open = term;
11357 /* find corresponding closing delimiter */
11358 if (term && (tmps = strchr("([{< )]}> )]}>",term)))
11359 termcode = termstr[0] = term = tmps[5];
11361 PL_multi_close = term;
11363 /* create a new SV to hold the contents. 79 is the SV's initial length.
11364 What a random number. */
11366 sv_upgrade(sv, SVt_PVIV);
11367 SvIV_set(sv, termcode);
11368 (void)SvPOK_only(sv); /* validate pointer */
11370 /* move past delimiter and try to read a complete string */
11372 sv_catpvn(sv, s, termlen);
11375 tstart = SvPVX(PL_linestr) + stuffstart;
11376 if (!PL_thisopen && !keep_delims) {
11377 PL_thisopen = newSVpvn(tstart, s - tstart);
11378 stuffstart = s - SvPVX(PL_linestr);
11382 if (PL_encoding && !UTF) {
11386 int offset = s - SvPVX_const(PL_linestr);
11387 const bool found = sv_cat_decode(sv, PL_encoding, PL_linestr,
11388 &offset, (char*)termstr, termlen);
11389 const char * const ns = SvPVX_const(PL_linestr) + offset;
11390 char * const svlast = SvEND(sv) - 1;
11392 for (; s < ns; s++) {
11393 if (*s == '\n' && !PL_rsfp)
11394 CopLINE_inc(PL_curcop);
11397 goto read_more_line;
11399 /* handle quoted delimiters */
11400 if (SvCUR(sv) > 1 && *(svlast-1) == '\\') {
11402 for (t = svlast-2; t >= SvPVX_const(sv) && *t == '\\';)
11404 if ((svlast-1 - t) % 2) {
11405 if (!keep_quoted) {
11406 *(svlast-1) = term;
11408 SvCUR_set(sv, SvCUR(sv) - 1);
11413 if (PL_multi_open == PL_multi_close) {
11421 for (t = w = last; t < svlast; w++, t++) {
11422 /* At here, all closes are "was quoted" one,
11423 so we don't check PL_multi_close. */
11425 if (!keep_quoted && *(t+1) == PL_multi_open)
11430 else if (*t == PL_multi_open)
11438 SvCUR_set(sv, w - SvPVX_const(sv));
11441 if (--brackets <= 0)
11446 if (!keep_delims) {
11447 SvCUR_set(sv, SvCUR(sv) - 1);
11453 /* extend sv if need be */
11454 SvGROW(sv, SvCUR(sv) + (PL_bufend - s) + 1);
11455 /* set 'to' to the next character in the sv's string */
11456 to = SvPVX(sv)+SvCUR(sv);
11458 /* if open delimiter is the close delimiter read unbridle */
11459 if (PL_multi_open == PL_multi_close) {
11460 for (; s < PL_bufend; s++,to++) {
11461 /* embedded newlines increment the current line number */
11462 if (*s == '\n' && !PL_rsfp)
11463 CopLINE_inc(PL_curcop);
11464 /* handle quoted delimiters */
11465 if (*s == '\\' && s+1 < PL_bufend && term != '\\') {
11466 if (!keep_quoted && s[1] == term)
11468 /* any other quotes are simply copied straight through */
11472 /* terminate when run out of buffer (the for() condition), or
11473 have found the terminator */
11474 else if (*s == term) {
11477 if (s+termlen <= PL_bufend && memEQ(s, (char*)termstr, termlen))
11480 else if (!has_utf8 && !UTF8_IS_INVARIANT((U8)*s) && UTF)
11486 /* if the terminator isn't the same as the start character (e.g.,
11487 matched brackets), we have to allow more in the quoting, and
11488 be prepared for nested brackets.
11491 /* read until we run out of string, or we find the terminator */
11492 for (; s < PL_bufend; s++,to++) {
11493 /* embedded newlines increment the line count */
11494 if (*s == '\n' && !PL_rsfp)
11495 CopLINE_inc(PL_curcop);
11496 /* backslashes can escape the open or closing characters */
11497 if (*s == '\\' && s+1 < PL_bufend) {
11498 if (!keep_quoted &&
11499 ((s[1] == PL_multi_open) || (s[1] == PL_multi_close)))
11504 /* allow nested opens and closes */
11505 else if (*s == PL_multi_close && --brackets <= 0)
11507 else if (*s == PL_multi_open)
11509 else if (!has_utf8 && !UTF8_IS_INVARIANT((U8)*s) && UTF)
11514 /* terminate the copied string and update the sv's end-of-string */
11516 SvCUR_set(sv, to - SvPVX_const(sv));
11519 * this next chunk reads more into the buffer if we're not done yet
11523 break; /* handle case where we are done yet :-) */
11525 #ifndef PERL_STRICT_CR
11526 if (to - SvPVX_const(sv) >= 2) {
11527 if ((to[-2] == '\r' && to[-1] == '\n') ||
11528 (to[-2] == '\n' && to[-1] == '\r'))
11532 SvCUR_set(sv, to - SvPVX_const(sv));
11534 else if (to[-1] == '\r')
11537 else if (to - SvPVX_const(sv) == 1 && to[-1] == '\r')
11542 /* if we're out of file, or a read fails, bail and reset the current
11543 line marker so we can report where the unterminated string began
11546 if (PL_madskills) {
11547 char * const tstart = SvPVX(PL_linestr) + stuffstart;
11549 sv_catpvn(PL_thisstuff, tstart, PL_bufend - tstart);
11551 PL_thisstuff = newSVpvn(tstart, PL_bufend - tstart);
11555 !(PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = filter_gets(PL_linestr, PL_rsfp, 0))) {
11557 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
11563 /* we read a line, so increment our line counter */
11564 CopLINE_inc(PL_curcop);
11566 /* update debugger info */
11567 if (PERLDB_LINE && PL_curstash != PL_debstash) {
11568 SV * const line_sv = newSV(0);
11570 sv_upgrade(line_sv, SVt_PVMG);
11571 sv_setsv(line_sv,PL_linestr);
11572 (void)SvIOK_on(line_sv);
11573 SvIV_set(line_sv, 0);
11574 av_store(CopFILEAVx(PL_curcop), (I32)CopLINE(PL_curcop), line_sv);
11577 /* having changed the buffer, we must update PL_bufend */
11578 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
11579 PL_last_lop = PL_last_uni = NULL;
11582 /* at this point, we have successfully read the delimited string */
11584 if (!PL_encoding || UTF) {
11586 if (PL_madskills) {
11587 char * const tstart = SvPVX(PL_linestr) + stuffstart;
11588 const int len = s - start;
11590 sv_catpvn(PL_thisstuff, tstart, len);
11592 PL_thisstuff = newSVpvn(tstart, len);
11593 if (!PL_thisclose && !keep_delims)
11594 PL_thisclose = newSVpvn(s,termlen);
11599 sv_catpvn(sv, s, termlen);
11604 if (PL_madskills) {
11605 char * const tstart = SvPVX(PL_linestr) + stuffstart;
11606 const int len = s - tstart - termlen;
11608 sv_catpvn(PL_thisstuff, tstart, len);
11610 PL_thisstuff = newSVpvn(tstart, len);
11611 if (!PL_thisclose && !keep_delims)
11612 PL_thisclose = newSVpvn(s - termlen,termlen);
11616 if (has_utf8 || PL_encoding)
11619 PL_multi_end = CopLINE(PL_curcop);
11621 /* if we allocated too much space, give some back */
11622 if (SvCUR(sv) + 5 < SvLEN(sv)) {
11623 SvLEN_set(sv, SvCUR(sv) + 1);
11624 SvPV_renew(sv, SvLEN(sv));
11627 /* decide whether this is the first or second quoted string we've read
11640 takes: pointer to position in buffer
11641 returns: pointer to new position in buffer
11642 side-effects: builds ops for the constant in yylval.op
11644 Read a number in any of the formats that Perl accepts:
11646 \d(_?\d)*(\.(\d(_?\d)*)?)?[Ee][\+\-]?(\d(_?\d)*) 12 12.34 12.
11647 \.\d(_?\d)*[Ee][\+\-]?(\d(_?\d)*) .34
11650 0x[0-9A-Fa-f](_?[0-9A-Fa-f])*
11652 Like most scan_ routines, it uses the PL_tokenbuf buffer to hold the
11655 If it reads a number without a decimal point or an exponent, it will
11656 try converting the number to an integer and see if it can do so
11657 without loss of precision.
11661 Perl_scan_num(pTHX_ const char *start, YYSTYPE* lvalp)
11664 register const char *s = start; /* current position in buffer */
11665 register char *d; /* destination in temp buffer */
11666 register char *e; /* end of temp buffer */
11667 NV nv; /* number read, as a double */
11668 SV *sv = NULL; /* place to put the converted number */
11669 bool floatit; /* boolean: int or float? */
11670 const char *lastub = NULL; /* position of last underbar */
11671 static char const number_too_long[] = "Number too long";
11673 /* We use the first character to decide what type of number this is */
11677 Perl_croak(aTHX_ "panic: scan_num");
11679 /* if it starts with a 0, it could be an octal number, a decimal in
11680 0.13 disguise, or a hexadecimal number, or a binary number. */
11684 u holds the "number so far"
11685 shift the power of 2 of the base
11686 (hex == 4, octal == 3, binary == 1)
11687 overflowed was the number more than we can hold?
11689 Shift is used when we add a digit. It also serves as an "are
11690 we in octal/hex/binary?" indicator to disallow hex characters
11691 when in octal mode.
11696 bool overflowed = FALSE;
11697 bool just_zero = TRUE; /* just plain 0 or binary number? */
11698 static const NV nvshift[5] = { 1.0, 2.0, 4.0, 8.0, 16.0 };
11699 static const char* const bases[5] =
11700 { "", "binary", "", "octal", "hexadecimal" };
11701 static const char* const Bases[5] =
11702 { "", "Binary", "", "Octal", "Hexadecimal" };
11703 static const char* const maxima[5] =
11705 "0b11111111111111111111111111111111",
11709 const char *base, *Base, *max;
11711 /* check for hex */
11716 } else if (s[1] == 'b') {
11721 /* check for a decimal in disguise */
11722 else if (s[1] == '.' || s[1] == 'e' || s[1] == 'E')
11724 /* so it must be octal */
11731 if (ckWARN(WARN_SYNTAX))
11732 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
11733 "Misplaced _ in number");
11737 base = bases[shift];
11738 Base = Bases[shift];
11739 max = maxima[shift];
11741 /* read the rest of the number */
11743 /* x is used in the overflow test,
11744 b is the digit we're adding on. */
11749 /* if we don't mention it, we're done */
11753 /* _ are ignored -- but warned about if consecutive */
11755 if (lastub && s == lastub + 1 && ckWARN(WARN_SYNTAX))
11756 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
11757 "Misplaced _ in number");
11761 /* 8 and 9 are not octal */
11762 case '8': case '9':
11764 yyerror(Perl_form(aTHX_ "Illegal octal digit '%c'", *s));
11768 case '2': case '3': case '4':
11769 case '5': case '6': case '7':
11771 yyerror(Perl_form(aTHX_ "Illegal binary digit '%c'", *s));
11774 case '0': case '1':
11775 b = *s++ & 15; /* ASCII digit -> value of digit */
11779 case 'a': case 'b': case 'c': case 'd': case 'e': case 'f':
11780 case 'A': case 'B': case 'C': case 'D': case 'E': case 'F':
11781 /* make sure they said 0x */
11784 b = (*s++ & 7) + 9;
11786 /* Prepare to put the digit we have onto the end
11787 of the number so far. We check for overflows.
11793 x = u << shift; /* make room for the digit */
11795 if ((x >> shift) != u
11796 && !(PL_hints & HINT_NEW_BINARY)) {
11799 if (ckWARN_d(WARN_OVERFLOW))
11800 Perl_warner(aTHX_ packWARN(WARN_OVERFLOW),
11801 "Integer overflow in %s number",
11804 u = x | b; /* add the digit to the end */
11807 n *= nvshift[shift];
11808 /* If an NV has not enough bits in its
11809 * mantissa to represent an UV this summing of
11810 * small low-order numbers is a waste of time
11811 * (because the NV cannot preserve the
11812 * low-order bits anyway): we could just
11813 * remember when did we overflow and in the
11814 * end just multiply n by the right
11822 /* if we get here, we had success: make a scalar value from
11827 /* final misplaced underbar check */
11828 if (s[-1] == '_') {
11829 if (ckWARN(WARN_SYNTAX))
11830 Perl_warner(aTHX_ packWARN(WARN_SYNTAX), "Misplaced _ in number");
11835 if (n > 4294967295.0 && ckWARN(WARN_PORTABLE))
11836 Perl_warner(aTHX_ packWARN(WARN_PORTABLE),
11837 "%s number > %s non-portable",
11843 if (u > 0xffffffff && ckWARN(WARN_PORTABLE))
11844 Perl_warner(aTHX_ packWARN(WARN_PORTABLE),
11845 "%s number > %s non-portable",
11850 if (just_zero && (PL_hints & HINT_NEW_INTEGER))
11851 sv = new_constant(start, s - start, "integer",
11853 else if (PL_hints & HINT_NEW_BINARY)
11854 sv = new_constant(start, s - start, "binary", sv, NULL, NULL);
11859 handle decimal numbers.
11860 we're also sent here when we read a 0 as the first digit
11862 case '1': case '2': case '3': case '4': case '5':
11863 case '6': case '7': case '8': case '9': case '.':
11866 e = PL_tokenbuf + sizeof PL_tokenbuf - 6; /* room for various punctuation */
11869 /* read next group of digits and _ and copy into d */
11870 while (isDIGIT(*s) || *s == '_') {
11871 /* skip underscores, checking for misplaced ones
11875 if (lastub && s == lastub + 1 && ckWARN(WARN_SYNTAX))
11876 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
11877 "Misplaced _ in number");
11881 /* check for end of fixed-length buffer */
11883 Perl_croak(aTHX_ number_too_long);
11884 /* if we're ok, copy the character */
11889 /* final misplaced underbar check */
11890 if (lastub && s == lastub + 1) {
11891 if (ckWARN(WARN_SYNTAX))
11892 Perl_warner(aTHX_ packWARN(WARN_SYNTAX), "Misplaced _ in number");
11895 /* read a decimal portion if there is one. avoid
11896 3..5 being interpreted as the number 3. followed
11899 if (*s == '.' && s[1] != '.') {
11904 if (ckWARN(WARN_SYNTAX))
11905 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
11906 "Misplaced _ in number");
11910 /* copy, ignoring underbars, until we run out of digits.
11912 for (; isDIGIT(*s) || *s == '_'; s++) {
11913 /* fixed length buffer check */
11915 Perl_croak(aTHX_ number_too_long);
11917 if (lastub && s == lastub + 1 && ckWARN(WARN_SYNTAX))
11918 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
11919 "Misplaced _ in number");
11925 /* fractional part ending in underbar? */
11926 if (s[-1] == '_') {
11927 if (ckWARN(WARN_SYNTAX))
11928 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
11929 "Misplaced _ in number");
11931 if (*s == '.' && isDIGIT(s[1])) {
11932 /* oops, it's really a v-string, but without the "v" */
11938 /* read exponent part, if present */
11939 if ((*s == 'e' || *s == 'E') && strchr("+-0123456789_", s[1])) {
11943 /* regardless of whether user said 3E5 or 3e5, use lower 'e' */
11944 *d++ = 'e'; /* At least some Mach atof()s don't grok 'E' */
11946 /* stray preinitial _ */
11948 if (ckWARN(WARN_SYNTAX))
11949 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
11950 "Misplaced _ in number");
11954 /* allow positive or negative exponent */
11955 if (*s == '+' || *s == '-')
11958 /* stray initial _ */
11960 if (ckWARN(WARN_SYNTAX))
11961 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
11962 "Misplaced _ in number");
11966 /* read digits of exponent */
11967 while (isDIGIT(*s) || *s == '_') {
11970 Perl_croak(aTHX_ number_too_long);
11974 if (((lastub && s == lastub + 1) ||
11975 (!isDIGIT(s[1]) && s[1] != '_'))
11976 && ckWARN(WARN_SYNTAX))
11977 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
11978 "Misplaced _ in number");
11985 /* make an sv from the string */
11989 We try to do an integer conversion first if no characters
11990 indicating "float" have been found.
11995 const int flags = grok_number (PL_tokenbuf, d - PL_tokenbuf, &uv);
11997 if (flags == IS_NUMBER_IN_UV) {
11999 sv_setiv(sv, uv); /* Prefer IVs over UVs. */
12002 } else if (flags == (IS_NUMBER_IN_UV | IS_NUMBER_NEG)) {
12003 if (uv <= (UV) IV_MIN)
12004 sv_setiv(sv, -(IV)uv);
12011 /* terminate the string */
12013 nv = Atof(PL_tokenbuf);
12017 if ( floatit ? (PL_hints & HINT_NEW_FLOAT) :
12018 (PL_hints & HINT_NEW_INTEGER) )
12019 sv = new_constant(PL_tokenbuf, d - PL_tokenbuf,
12020 (floatit ? "float" : "integer"),
12024 /* if it starts with a v, it could be a v-string */
12027 sv = newSV(5); /* preallocate storage space */
12028 s = scan_vstring(s,sv);
12032 /* make the op for the constant and return */
12035 lvalp->opval = newSVOP(OP_CONST, 0, sv);
12037 lvalp->opval = NULL;
12043 S_scan_formline(pTHX_ register char *s)
12046 register char *eol;
12048 SV * const stuff = newSVpvs("");
12049 bool needargs = FALSE;
12050 bool eofmt = FALSE;
12052 char *tokenstart = s;
12055 if (PL_madskills) {
12056 savewhite = PL_thiswhite;
12061 while (!needargs) {
12064 #ifdef PERL_STRICT_CR
12065 while (SPACE_OR_TAB(*t))
12068 while (SPACE_OR_TAB(*t) || *t == '\r')
12071 if (*t == '\n' || t == PL_bufend) {
12076 if (PL_in_eval && !PL_rsfp) {
12077 eol = (char *) memchr(s,'\n',PL_bufend-s);
12082 eol = PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
12084 for (t = s; t < eol; t++) {
12085 if (*t == '~' && t[1] == '~' && SvCUR(stuff)) {
12087 goto enough; /* ~~ must be first line in formline */
12089 if (*t == '@' || *t == '^')
12093 sv_catpvn(stuff, s, eol-s);
12094 #ifndef PERL_STRICT_CR
12095 if (eol-s > 1 && eol[-2] == '\r' && eol[-1] == '\n') {
12096 char *end = SvPVX(stuff) + SvCUR(stuff);
12099 SvCUR_set(stuff, SvCUR(stuff) - 1);
12109 if (PL_madskills) {
12111 sv_catpvn(PL_thistoken, tokenstart, PL_bufend - tokenstart);
12113 PL_thistoken = newSVpvn(tokenstart, PL_bufend - tokenstart);
12116 s = filter_gets(PL_linestr, PL_rsfp, 0);
12118 tokenstart = PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = SvPVX(PL_linestr);
12120 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = SvPVX(PL_linestr);
12122 PL_bufend = PL_bufptr + SvCUR(PL_linestr);
12123 PL_last_lop = PL_last_uni = NULL;
12132 if (SvCUR(stuff)) {
12135 PL_lex_state = LEX_NORMAL;
12136 start_force(PL_curforce);
12137 NEXTVAL_NEXTTOKE.ival = 0;
12141 PL_lex_state = LEX_FORMLINE;
12143 if (UTF && is_utf8_string((U8*)SvPVX_const(stuff), SvCUR(stuff)))
12145 else if (PL_encoding)
12146 sv_recode_to_utf8(stuff, PL_encoding);
12148 start_force(PL_curforce);
12149 NEXTVAL_NEXTTOKE.opval = (OP*)newSVOP(OP_CONST, 0, stuff);
12151 start_force(PL_curforce);
12152 NEXTVAL_NEXTTOKE.ival = OP_FORMLINE;
12156 SvREFCNT_dec(stuff);
12158 PL_lex_formbrack = 0;
12162 if (PL_madskills) {
12164 sv_catpvn(PL_thistoken, tokenstart, s - tokenstart);
12166 PL_thistoken = newSVpvn(tokenstart, s - tokenstart);
12167 PL_thiswhite = savewhite;
12179 PL_cshlen = strlen(PL_cshname);
12181 #if defined(USE_ITHREADS)
12182 PERL_UNUSED_CONTEXT;
12188 Perl_start_subparse(pTHX_ I32 is_format, U32 flags)
12191 const I32 oldsavestack_ix = PL_savestack_ix;
12192 CV* const outsidecv = PL_compcv;
12195 assert(SvTYPE(PL_compcv) == SVt_PVCV);
12197 SAVEI32(PL_subline);
12198 save_item(PL_subname);
12199 SAVESPTR(PL_compcv);
12201 PL_compcv = (CV*)newSV(0);
12202 sv_upgrade((SV *)PL_compcv, is_format ? SVt_PVFM : SVt_PVCV);
12203 CvFLAGS(PL_compcv) |= flags;
12205 PL_subline = CopLINE(PL_curcop);
12206 CvPADLIST(PL_compcv) = pad_new(padnew_SAVE|padnew_SAVESUB);
12207 CvOUTSIDE(PL_compcv) = (CV*)SvREFCNT_inc_simple(outsidecv);
12208 CvOUTSIDE_SEQ(PL_compcv) = PL_cop_seqmax;
12210 return oldsavestack_ix;
12214 #pragma segment Perl_yylex
12217 Perl_yywarn(pTHX_ const char *s)
12220 PL_in_eval |= EVAL_WARNONLY;
12222 PL_in_eval &= ~EVAL_WARNONLY;
12227 Perl_yyerror(pTHX_ const char *s)
12230 const char *where = NULL;
12231 const char *context = NULL;
12235 if (!yychar || (yychar == ';' && !PL_rsfp))
12237 else if (PL_oldoldbufptr && PL_bufptr > PL_oldoldbufptr &&
12238 PL_bufptr - PL_oldoldbufptr < 200 && PL_oldoldbufptr != PL_oldbufptr &&
12239 PL_oldbufptr != PL_bufptr) {
12242 The code below is removed for NetWare because it abends/crashes on NetWare
12243 when the script has error such as not having the closing quotes like:
12244 if ($var eq "value)
12245 Checking of white spaces is anyway done in NetWare code.
12248 while (isSPACE(*PL_oldoldbufptr))
12251 context = PL_oldoldbufptr;
12252 contlen = PL_bufptr - PL_oldoldbufptr;
12254 else if (PL_oldbufptr && PL_bufptr > PL_oldbufptr &&
12255 PL_bufptr - PL_oldbufptr < 200 && PL_oldbufptr != PL_bufptr) {
12258 The code below is removed for NetWare because it abends/crashes on NetWare
12259 when the script has error such as not having the closing quotes like:
12260 if ($var eq "value)
12261 Checking of white spaces is anyway done in NetWare code.
12264 while (isSPACE(*PL_oldbufptr))
12267 context = PL_oldbufptr;
12268 contlen = PL_bufptr - PL_oldbufptr;
12270 else if (yychar > 255)
12271 where = "next token ???";
12272 else if (yychar == -2) { /* YYEMPTY */
12273 if (PL_lex_state == LEX_NORMAL ||
12274 (PL_lex_state == LEX_KNOWNEXT && PL_lex_defer == LEX_NORMAL))
12275 where = "at end of line";
12276 else if (PL_lex_inpat)
12277 where = "within pattern";
12279 where = "within string";
12282 SV * const where_sv = sv_2mortal(newSVpvs("next char "));
12284 Perl_sv_catpvf(aTHX_ where_sv, "^%c", toCTRL(yychar));
12285 else if (isPRINT_LC(yychar))
12286 Perl_sv_catpvf(aTHX_ where_sv, "%c", yychar);
12288 Perl_sv_catpvf(aTHX_ where_sv, "\\%03o", yychar & 255);
12289 where = SvPVX_const(where_sv);
12291 msg = sv_2mortal(newSVpv(s, 0));
12292 Perl_sv_catpvf(aTHX_ msg, " at %s line %"IVdf", ",
12293 OutCopFILE(PL_curcop), (IV)CopLINE(PL_curcop));
12295 Perl_sv_catpvf(aTHX_ msg, "near \"%.*s\"\n", contlen, context);
12297 Perl_sv_catpvf(aTHX_ msg, "%s\n", where);
12298 if (PL_multi_start < PL_multi_end && (U32)(CopLINE(PL_curcop) - PL_multi_end) <= 1) {
12299 Perl_sv_catpvf(aTHX_ msg,
12300 " (Might be a runaway multi-line %c%c string starting on line %"IVdf")\n",
12301 (int)PL_multi_open,(int)PL_multi_close,(IV)PL_multi_start);
12304 if (PL_in_eval & EVAL_WARNONLY && ckWARN_d(WARN_SYNTAX))
12305 Perl_warner(aTHX_ packWARN(WARN_SYNTAX), "%"SVf, (void*)msg);
12308 if (PL_error_count >= 10) {
12309 if (PL_in_eval && SvCUR(ERRSV))
12310 Perl_croak(aTHX_ "%"SVf"%s has too many errors.\n",
12311 (void*)ERRSV, OutCopFILE(PL_curcop));
12313 Perl_croak(aTHX_ "%s has too many errors.\n",
12314 OutCopFILE(PL_curcop));
12317 PL_in_my_stash = NULL;
12321 #pragma segment Main
12325 S_swallow_bom(pTHX_ U8 *s)
12328 const STRLEN slen = SvCUR(PL_linestr);
12331 if (s[1] == 0xFE) {
12332 /* UTF-16 little-endian? (or UTF32-LE?) */
12333 if (s[2] == 0 && s[3] == 0) /* UTF-32 little-endian */
12334 Perl_croak(aTHX_ "Unsupported script encoding UTF32-LE");
12335 #ifndef PERL_NO_UTF16_FILTER
12336 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF16-LE script encoding (BOM)\n");
12339 if (PL_bufend > (char*)s) {
12343 filter_add(utf16rev_textfilter, NULL);
12344 Newx(news, (PL_bufend - (char*)s) * 3 / 2 + 1, U8);
12345 utf16_to_utf8_reversed(s, news,
12346 PL_bufend - (char*)s - 1,
12348 sv_setpvn(PL_linestr, (const char*)news, newlen);
12350 s = (U8*)SvPVX(PL_linestr);
12351 Copy(news, s, newlen, U8);
12355 SvUTF8_on(PL_linestr);
12356 s = (U8*)SvPVX(PL_linestr);
12358 /* FIXME - is this a general bug fix? */
12361 PL_bufend = SvPVX(PL_linestr) + newlen;
12364 Perl_croak(aTHX_ "Unsupported script encoding UTF16-LE");
12369 if (s[1] == 0xFF) { /* UTF-16 big-endian? */
12370 #ifndef PERL_NO_UTF16_FILTER
12371 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16BE script encoding (BOM)\n");
12374 if (PL_bufend > (char *)s) {
12378 filter_add(utf16_textfilter, NULL);
12379 Newx(news, (PL_bufend - (char*)s) * 3 / 2 + 1, U8);
12380 utf16_to_utf8(s, news,
12381 PL_bufend - (char*)s,
12383 sv_setpvn(PL_linestr, (const char*)news, newlen);
12385 SvUTF8_on(PL_linestr);
12386 s = (U8*)SvPVX(PL_linestr);
12387 PL_bufend = SvPVX(PL_linestr) + newlen;
12390 Perl_croak(aTHX_ "Unsupported script encoding UTF16-BE");
12395 if (slen > 2 && s[1] == 0xBB && s[2] == 0xBF) {
12396 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-8 script encoding (BOM)\n");
12397 s += 3; /* UTF-8 */
12403 if (s[2] == 0xFE && s[3] == 0xFF) {
12404 /* UTF-32 big-endian */
12405 Perl_croak(aTHX_ "Unsupported script encoding UTF32-BE");
12408 else if (s[2] == 0 && s[3] != 0) {
12411 * are a good indicator of UTF-16BE. */
12412 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16BE script encoding (no BOM)\n");
12417 if (slen > 3 && s[1] == 0 && s[2] != 0 && s[3] == 0) {
12420 * are a good indicator of UTF-16LE. */
12421 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16LE script encoding (no BOM)\n");
12430 * Restore a source filter.
12434 restore_rsfp(pTHX_ void *f)
12437 PerlIO * const fp = (PerlIO*)f;
12439 if (PL_rsfp == PerlIO_stdin())
12440 PerlIO_clearerr(PL_rsfp);
12441 else if (PL_rsfp && (PL_rsfp != fp))
12442 PerlIO_close(PL_rsfp);
12446 #ifndef PERL_NO_UTF16_FILTER
12448 utf16_textfilter(pTHX_ int idx, SV *sv, int maxlen)
12451 const STRLEN old = SvCUR(sv);
12452 const I32 count = FILTER_READ(idx+1, sv, maxlen);
12453 DEBUG_P(PerlIO_printf(Perl_debug_log,
12454 "utf16_textfilter(%p): %d %d (%d)\n",
12455 FPTR2DPTR(void *, utf16_textfilter),
12456 idx, maxlen, (int) count));
12460 Newx(tmps, SvCUR(sv) * 3 / 2 + 1, U8);
12461 Copy(SvPVX_const(sv), tmps, old, char);
12462 utf16_to_utf8((U8*)SvPVX_const(sv) + old, tmps + old,
12463 SvCUR(sv) - old, &newlen);
12464 sv_usepvn(sv, (char*)tmps, (STRLEN)newlen + old);
12466 DEBUG_P({sv_dump(sv);});
12471 utf16rev_textfilter(pTHX_ int idx, SV *sv, int maxlen)
12474 const STRLEN old = SvCUR(sv);
12475 const I32 count = FILTER_READ(idx+1, sv, maxlen);
12476 DEBUG_P(PerlIO_printf(Perl_debug_log,
12477 "utf16rev_textfilter(%p): %d %d (%d)\n",
12478 FPTR2DPTR(void *, utf16rev_textfilter),
12479 idx, maxlen, (int) count));
12483 Newx(tmps, SvCUR(sv) * 3 / 2 + 1, U8);
12484 Copy(SvPVX_const(sv), tmps, old, char);
12485 utf16_to_utf8((U8*)SvPVX_const(sv) + old, tmps + old,
12486 SvCUR(sv) - old, &newlen);
12487 sv_usepvn(sv, (char*)tmps, (STRLEN)newlen + old);
12489 DEBUG_P({ sv_dump(sv); });
12495 Returns a pointer to the next character after the parsed
12496 vstring, as well as updating the passed in sv.
12498 Function must be called like
12501 s = scan_vstring(s,sv);
12503 The sv should already be large enough to store the vstring
12504 passed in, for performance reasons.
12509 Perl_scan_vstring(pTHX_ const char *s, SV *sv)
12512 const char *pos = s;
12513 const char *start = s;
12514 if (*pos == 'v') pos++; /* get past 'v' */
12515 while (pos < PL_bufend && (isDIGIT(*pos) || *pos == '_'))
12517 if ( *pos != '.') {
12518 /* this may not be a v-string if followed by => */
12519 const char *next = pos;
12520 while (next < PL_bufend && isSPACE(*next))
12522 if ((PL_bufend - next) >= 2 && *next == '=' && next[1] == '>' ) {
12523 /* return string not v-string */
12524 sv_setpvn(sv,(char *)s,pos-s);
12525 return (char *)pos;
12529 if (!isALPHA(*pos)) {
12530 U8 tmpbuf[UTF8_MAXBYTES+1];
12533 s++; /* get past 'v' */
12535 sv_setpvn(sv, "", 0);
12538 /* this is atoi() that tolerates underscores */
12541 const char *end = pos;
12543 while (--end >= s) {
12545 const UV orev = rev;
12546 rev += (*end - '0') * mult;
12548 if (orev > rev && ckWARN_d(WARN_OVERFLOW))
12549 Perl_warner(aTHX_ packWARN(WARN_OVERFLOW),
12550 "Integer overflow in decimal number");
12554 if (rev > 0x7FFFFFFF)
12555 Perl_croak(aTHX_ "In EBCDIC the v-string components cannot exceed 2147483647");
12557 /* Append native character for the rev point */
12558 tmpend = uvchr_to_utf8(tmpbuf, rev);
12559 sv_catpvn(sv, (const char*)tmpbuf, tmpend - tmpbuf);
12560 if (!UNI_IS_INVARIANT(NATIVE_TO_UNI(rev)))
12562 if (pos + 1 < PL_bufend && *pos == '.' && isDIGIT(pos[1]))
12568 while (pos < PL_bufend && (isDIGIT(*pos) || *pos == '_'))
12572 sv_magic(sv,NULL,PERL_MAGIC_vstring,(const char*)start, pos-start);
12580 * c-indentation-style: bsd
12581 * c-basic-offset: 4
12582 * indent-tabs-mode: t
12585 * ex: set ts=8 sts=4 sw=4 noet: