3 * Copyright (C) 1991, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999,
4 * 2000, 2001, 2002, 2003, 2004, 2005, 2006, by Larry Wall and others
6 * You may distribute under the terms of either the GNU General Public
7 * License or the Artistic License, as specified in the README file.
12 * "It all comes from here, the stench and the peril." --Frodo
16 * This file is the lexer for Perl. It's closely linked to the
19 * The main routine is yylex(), which returns the next token.
23 #define PERL_IN_TOKE_C
26 #define yychar (*PL_yycharp)
27 #define yylval (*PL_yylvalp)
29 static const char ident_too_long[] = "Identifier too long";
30 static const char commaless_variable_list[] = "comma-less variable list";
32 static void restore_rsfp(pTHX_ void *f);
33 #ifndef PERL_NO_UTF16_FILTER
34 static I32 utf16_textfilter(pTHX_ int idx, SV *sv, int maxlen);
35 static I32 utf16rev_textfilter(pTHX_ int idx, SV *sv, int maxlen);
39 # define CURMAD(slot,sv) if (PL_madskills) { curmad(slot,sv); sv = 0; }
40 # define NEXTVAL_NEXTTOKE PL_nexttoke[PL_curforce].next_val
42 # define CURMAD(slot,sv)
43 # define NEXTVAL_NEXTTOKE PL_nextval[PL_nexttoke]
46 #define XFAKEBRACK 128
49 #ifdef USE_UTF8_SCRIPTS
50 # define UTF (!IN_BYTES)
52 # define UTF ((PL_linestr && DO_UTF8(PL_linestr)) || (PL_hints & HINT_UTF8))
55 /* In variables named $^X, these are the legal values for X.
56 * 1999-02-27 mjd-perl-patch@plover.com */
57 #define isCONTROLVAR(x) (isUPPER(x) || strchr("[\\]^_?", (x)))
59 /* On MacOS, respect nonbreaking spaces */
60 #ifdef MACOS_TRADITIONAL
61 #define SPACE_OR_TAB(c) ((c)==' '||(c)=='\312'||(c)=='\t')
63 #define SPACE_OR_TAB(c) ((c)==' '||(c)=='\t')
66 /* LEX_* are values for PL_lex_state, the state of the lexer.
67 * They are arranged oddly so that the guard on the switch statement
68 * can get by with a single comparison (if the compiler is smart enough).
71 /* #define LEX_NOTPARSING 11 is done in perl.h. */
73 #define LEX_NORMAL 10 /* normal code (ie not within "...") */
74 #define LEX_INTERPNORMAL 9 /* code within a string, eg "$foo[$x+1]" */
75 #define LEX_INTERPCASEMOD 8 /* expecting a \U, \Q or \E etc */
76 #define LEX_INTERPPUSH 7 /* starting a new sublex parse level */
77 #define LEX_INTERPSTART 6 /* expecting the start of a $var */
79 /* at end of code, eg "$x" followed by: */
80 #define LEX_INTERPEND 5 /* ... eg not one of [, { or -> */
81 #define LEX_INTERPENDMAYBE 4 /* ... eg one of [, { or -> */
83 #define LEX_INTERPCONCAT 3 /* expecting anything, eg at start of
84 string or after \E, $foo, etc */
85 #define LEX_INTERPCONST 2 /* NOT USED */
86 #define LEX_FORMLINE 1 /* expecting a format line */
87 #define LEX_KNOWNEXT 0 /* next token known; just return it */
91 static const char* const lex_state_names[] = {
110 #include "keywords.h"
112 /* CLINE is a macro that ensures PL_copline has a sane value */
117 #define CLINE (PL_copline = (CopLINE(PL_curcop) < PL_copline ? CopLINE(PL_curcop) : PL_copline))
120 # define SKIPSPACE0(s) skipspace0(s)
121 # define SKIPSPACE1(s) skipspace1(s)
122 # define SKIPSPACE2(s,tsv) skipspace2(s,&tsv)
123 # define PEEKSPACE(s) skipspace2(s,0)
125 # define SKIPSPACE0(s) skipspace(s)
126 # define SKIPSPACE1(s) skipspace(s)
127 # define SKIPSPACE2(s,tsv) skipspace(s)
128 # define PEEKSPACE(s) skipspace(s)
132 * Convenience functions to return different tokens and prime the
133 * lexer for the next token. They all take an argument.
135 * TOKEN : generic token (used for '(', DOLSHARP, etc)
136 * OPERATOR : generic operator
137 * AOPERATOR : assignment operator
138 * PREBLOCK : beginning the block after an if, while, foreach, ...
139 * PRETERMBLOCK : beginning a non-code-defining {} block (eg, hash ref)
140 * PREREF : *EXPR where EXPR is not a simple identifier
141 * TERM : expression term
142 * LOOPX : loop exiting command (goto, last, dump, etc)
143 * FTST : file test operator
144 * FUN0 : zero-argument function
145 * FUN1 : not used, except for not, which isn't a UNIOP
146 * BOop : bitwise or or xor
148 * SHop : shift operator
149 * PWop : power operator
150 * PMop : pattern-matching operator
151 * Aop : addition-level operator
152 * Mop : multiplication-level operator
153 * Eop : equality-testing operator
154 * Rop : relational operator <= != gt
156 * Also see LOP and lop() below.
159 #ifdef DEBUGGING /* Serve -DT. */
160 # define REPORT(retval) tokereport((I32)retval)
162 # define REPORT(retval) (retval)
165 #define TOKEN(retval) return ( PL_bufptr = s, REPORT(retval))
166 #define OPERATOR(retval) return (PL_expect = XTERM, PL_bufptr = s, REPORT(retval))
167 #define AOPERATOR(retval) return ao((PL_expect = XTERM, PL_bufptr = s, REPORT(retval)))
168 #define PREBLOCK(retval) return (PL_expect = XBLOCK,PL_bufptr = s, REPORT(retval))
169 #define PRETERMBLOCK(retval) return (PL_expect = XTERMBLOCK,PL_bufptr = s, REPORT(retval))
170 #define PREREF(retval) return (PL_expect = XREF,PL_bufptr = s, REPORT(retval))
171 #define TERM(retval) return (CLINE, PL_expect = XOPERATOR, PL_bufptr = s, REPORT(retval))
172 #define LOOPX(f) return (yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)LOOPEX))
173 #define FTST(f) return (yylval.ival=f, PL_expect=XTERMORDORDOR, PL_bufptr=s, REPORT((int)UNIOP))
174 #define FUN0(f) return (yylval.ival=f, PL_expect=XOPERATOR, PL_bufptr=s, REPORT((int)FUNC0))
175 #define FUN1(f) return (yylval.ival=f, PL_expect=XOPERATOR, PL_bufptr=s, REPORT((int)FUNC1))
176 #define BOop(f) return ao((yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)BITOROP)))
177 #define BAop(f) return ao((yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)BITANDOP)))
178 #define SHop(f) return ao((yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)SHIFTOP)))
179 #define PWop(f) return ao((yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)POWOP)))
180 #define PMop(f) return(yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)MATCHOP))
181 #define Aop(f) return ao((yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)ADDOP)))
182 #define Mop(f) return ao((yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)MULOP)))
183 #define Eop(f) return (yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)EQOP))
184 #define Rop(f) return (yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)RELOP))
186 /* This bit of chicanery makes a unary function followed by
187 * a parenthesis into a function with one argument, highest precedence.
188 * The UNIDOR macro is for unary functions that can be followed by the //
189 * operator (such as C<shift // 0>).
191 #define UNI2(f,x) { \
195 PL_last_uni = PL_oldbufptr; \
196 PL_last_lop_op = f; \
198 return REPORT( (int)FUNC1 ); \
200 return REPORT( *s=='(' ? (int)FUNC1 : (int)UNIOP ); \
202 #define UNI(f) UNI2(f,XTERM)
203 #define UNIDOR(f) UNI2(f,XTERMORDORDOR)
205 #define UNIBRACK(f) { \
208 PL_last_uni = PL_oldbufptr; \
210 return REPORT( (int)FUNC1 ); \
212 return REPORT( (*s == '(') ? (int)FUNC1 : (int)UNIOP ); \
215 /* grandfather return to old style */
216 #define OLDLOP(f) return(yylval.ival=f,PL_expect = XTERM,PL_bufptr = s,(int)LSTOP)
220 /* how to interpret the yylval associated with the token */
224 TOKENTYPE_OPNUM, /* yylval.ival contains an opcode number */
230 static struct debug_tokens {
232 enum token_type type;
234 } const debug_tokens[] =
236 { ADDOP, TOKENTYPE_OPNUM, "ADDOP" },
237 { ANDAND, TOKENTYPE_NONE, "ANDAND" },
238 { ANDOP, TOKENTYPE_NONE, "ANDOP" },
239 { ANONSUB, TOKENTYPE_IVAL, "ANONSUB" },
240 { ARROW, TOKENTYPE_NONE, "ARROW" },
241 { ASSIGNOP, TOKENTYPE_OPNUM, "ASSIGNOP" },
242 { BITANDOP, TOKENTYPE_OPNUM, "BITANDOP" },
243 { BITOROP, TOKENTYPE_OPNUM, "BITOROP" },
244 { COLONATTR, TOKENTYPE_NONE, "COLONATTR" },
245 { CONTINUE, TOKENTYPE_NONE, "CONTINUE" },
246 { DEFAULT, TOKENTYPE_NONE, "DEFAULT" },
247 { DO, TOKENTYPE_NONE, "DO" },
248 { DOLSHARP, TOKENTYPE_NONE, "DOLSHARP" },
249 { DORDOR, TOKENTYPE_NONE, "DORDOR" },
250 { DOROP, TOKENTYPE_OPNUM, "DOROP" },
251 { DOTDOT, TOKENTYPE_IVAL, "DOTDOT" },
252 { ELSE, TOKENTYPE_NONE, "ELSE" },
253 { ELSIF, TOKENTYPE_IVAL, "ELSIF" },
254 { EQOP, TOKENTYPE_OPNUM, "EQOP" },
255 { FOR, TOKENTYPE_IVAL, "FOR" },
256 { FORMAT, TOKENTYPE_NONE, "FORMAT" },
257 { FUNC, TOKENTYPE_OPNUM, "FUNC" },
258 { FUNC0, TOKENTYPE_OPNUM, "FUNC0" },
259 { FUNC0SUB, TOKENTYPE_OPVAL, "FUNC0SUB" },
260 { FUNC1, TOKENTYPE_OPNUM, "FUNC1" },
261 { FUNCMETH, TOKENTYPE_OPVAL, "FUNCMETH" },
262 { GIVEN, TOKENTYPE_IVAL, "GIVEN" },
263 { HASHBRACK, TOKENTYPE_NONE, "HASHBRACK" },
264 { IF, TOKENTYPE_IVAL, "IF" },
265 { LABEL, TOKENTYPE_PVAL, "LABEL" },
266 { LOCAL, TOKENTYPE_IVAL, "LOCAL" },
267 { LOOPEX, TOKENTYPE_OPNUM, "LOOPEX" },
268 { LSTOP, TOKENTYPE_OPNUM, "LSTOP" },
269 { LSTOPSUB, TOKENTYPE_OPVAL, "LSTOPSUB" },
270 { MATCHOP, TOKENTYPE_OPNUM, "MATCHOP" },
271 { METHOD, TOKENTYPE_OPVAL, "METHOD" },
272 { MULOP, TOKENTYPE_OPNUM, "MULOP" },
273 { MY, TOKENTYPE_IVAL, "MY" },
274 { MYSUB, TOKENTYPE_NONE, "MYSUB" },
275 { NOAMP, TOKENTYPE_NONE, "NOAMP" },
276 { NOTOP, TOKENTYPE_NONE, "NOTOP" },
277 { OROP, TOKENTYPE_IVAL, "OROP" },
278 { OROR, TOKENTYPE_NONE, "OROR" },
279 { PACKAGE, TOKENTYPE_NONE, "PACKAGE" },
280 { PMFUNC, TOKENTYPE_OPVAL, "PMFUNC" },
281 { POSTDEC, TOKENTYPE_NONE, "POSTDEC" },
282 { POSTINC, TOKENTYPE_NONE, "POSTINC" },
283 { POWOP, TOKENTYPE_OPNUM, "POWOP" },
284 { PREDEC, TOKENTYPE_NONE, "PREDEC" },
285 { PREINC, TOKENTYPE_NONE, "PREINC" },
286 { PRIVATEREF, TOKENTYPE_OPVAL, "PRIVATEREF" },
287 { REFGEN, TOKENTYPE_NONE, "REFGEN" },
288 { RELOP, TOKENTYPE_OPNUM, "RELOP" },
289 { SHIFTOP, TOKENTYPE_OPNUM, "SHIFTOP" },
290 { SUB, TOKENTYPE_NONE, "SUB" },
291 { THING, TOKENTYPE_OPVAL, "THING" },
292 { UMINUS, TOKENTYPE_NONE, "UMINUS" },
293 { UNIOP, TOKENTYPE_OPNUM, "UNIOP" },
294 { UNIOPSUB, TOKENTYPE_OPVAL, "UNIOPSUB" },
295 { UNLESS, TOKENTYPE_IVAL, "UNLESS" },
296 { UNTIL, TOKENTYPE_IVAL, "UNTIL" },
297 { USE, TOKENTYPE_IVAL, "USE" },
298 { WHEN, TOKENTYPE_IVAL, "WHEN" },
299 { WHILE, TOKENTYPE_IVAL, "WHILE" },
300 { WORD, TOKENTYPE_OPVAL, "WORD" },
301 { 0, TOKENTYPE_NONE, 0 }
304 /* dump the returned token in rv, plus any optional arg in yylval */
307 S_tokereport(pTHX_ I32 rv)
311 const char *name = NULL;
312 enum token_type type = TOKENTYPE_NONE;
313 const struct debug_tokens *p;
314 SV* const report = newSVpvs("<== ");
316 for (p = debug_tokens; p->token; p++) {
317 if (p->token == (int)rv) {
324 Perl_sv_catpv(aTHX_ report, name);
325 else if ((char)rv > ' ' && (char)rv < '~')
326 Perl_sv_catpvf(aTHX_ report, "'%c'", (char)rv);
328 sv_catpvs(report, "EOF");
330 Perl_sv_catpvf(aTHX_ report, "?? %"IVdf, (IV)rv);
333 case TOKENTYPE_GVVAL: /* doesn't appear to be used */
336 Perl_sv_catpvf(aTHX_ report, "(ival=%"IVdf")", (IV)yylval.ival);
338 case TOKENTYPE_OPNUM:
339 Perl_sv_catpvf(aTHX_ report, "(ival=op_%s)",
340 PL_op_name[yylval.ival]);
343 Perl_sv_catpvf(aTHX_ report, "(pval=\"%s\")", yylval.pval);
345 case TOKENTYPE_OPVAL:
347 Perl_sv_catpvf(aTHX_ report, "(opval=op_%s)",
348 PL_op_name[yylval.opval->op_type]);
349 if (yylval.opval->op_type == OP_CONST) {
350 Perl_sv_catpvf(aTHX_ report, " %s",
351 SvPEEK(cSVOPx_sv(yylval.opval)));
356 sv_catpvs(report, "(opval=null)");
359 PerlIO_printf(Perl_debug_log, "### %s\n\n", SvPV_nolen_const(report));
365 /* print the buffer with suitable escapes */
368 S_printbuf(pTHX_ const char* fmt, const char* s)
370 SV* const tmp = newSVpvs("");
371 PerlIO_printf(Perl_debug_log, fmt, pv_display(tmp, s, strlen(s), 0, 60));
380 * This subroutine detects &&=, ||=, and //= and turns an ANDAND, OROR or DORDOR
381 * into an OP_ANDASSIGN, OP_ORASSIGN, or OP_DORASSIGN
385 S_ao(pTHX_ int toketype)
388 if (*PL_bufptr == '=') {
390 if (toketype == ANDAND)
391 yylval.ival = OP_ANDASSIGN;
392 else if (toketype == OROR)
393 yylval.ival = OP_ORASSIGN;
394 else if (toketype == DORDOR)
395 yylval.ival = OP_DORASSIGN;
403 * When Perl expects an operator and finds something else, no_op
404 * prints the warning. It always prints "<something> found where
405 * operator expected. It prints "Missing semicolon on previous line?"
406 * if the surprise occurs at the start of the line. "do you need to
407 * predeclare ..." is printed out for code like "sub bar; foo bar $x"
408 * where the compiler doesn't know if foo is a method call or a function.
409 * It prints "Missing operator before end of line" if there's nothing
410 * after the missing operator, or "... before <...>" if there is something
411 * after the missing operator.
415 S_no_op(pTHX_ const char *what, char *s)
418 char * const oldbp = PL_bufptr;
419 const bool is_first = (PL_oldbufptr == PL_linestart);
425 yywarn(Perl_form(aTHX_ "%s found where operator expected", what));
426 if (ckWARN_d(WARN_SYNTAX)) {
428 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
429 "\t(Missing semicolon on previous line?)\n");
430 else if (PL_oldoldbufptr && isIDFIRST_lazy_if(PL_oldoldbufptr,UTF)) {
432 for (t = PL_oldoldbufptr; *t && (isALNUM_lazy_if(t,UTF) || *t == ':'); t++)
434 if (t < PL_bufptr && isSPACE(*t))
435 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
436 "\t(Do you need to predeclare %.*s?)\n",
437 (int)(t - PL_oldoldbufptr), PL_oldoldbufptr);
441 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
442 "\t(Missing operator before %.*s?)\n", (int)(s - oldbp), oldbp);
450 * Complain about missing quote/regexp/heredoc terminator.
451 * If it's called with NULL then it cauterizes the line buffer.
452 * If we're in a delimited string and the delimiter is a control
453 * character, it's reformatted into a two-char sequence like ^C.
458 S_missingterm(pTHX_ char *s)
464 char * const nl = strrchr(s,'\n');
470 iscntrl(PL_multi_close)
472 PL_multi_close < 32 || PL_multi_close == 127
476 tmpbuf[1] = (char)toCTRL(PL_multi_close);
481 *tmpbuf = (char)PL_multi_close;
485 q = strchr(s,'"') ? '\'' : '"';
486 Perl_croak(aTHX_ "Can't find string terminator %c%s%c anywhere before EOF",q,s,q);
489 #define FEATURE_IS_ENABLED(name) \
490 ((0 != (PL_hints & HINT_LOCALIZE_HH)) \
491 && S_feature_is_enabled(aTHX_ STR_WITH_LEN(name)))
493 * S_feature_is_enabled
494 * Check whether the named feature is enabled.
497 S_feature_is_enabled(pTHX_ const char *name, STRLEN namelen)
500 HV * const hinthv = GvHV(PL_hintgv);
501 char he_name[32] = "feature_";
502 (void) strncpy(&he_name[8], name, 24);
504 return (hinthv && hv_exists(hinthv, he_name, 8 + namelen));
512 Perl_deprecate(pTHX_ const char *s)
514 if (ckWARN(WARN_DEPRECATED))
515 Perl_warner(aTHX_ packWARN(WARN_DEPRECATED), "Use of %s is deprecated", s);
519 Perl_deprecate_old(pTHX_ const char *s)
521 /* This function should NOT be called for any new deprecated warnings */
522 /* Use Perl_deprecate instead */
524 /* It is here to maintain backward compatibility with the pre-5.8 */
525 /* warnings category hierarchy. The "deprecated" category used to */
526 /* live under the "syntax" category. It is now a top-level category */
527 /* in its own right. */
529 if (ckWARN2(WARN_DEPRECATED, WARN_SYNTAX))
530 Perl_warner(aTHX_ packWARN2(WARN_DEPRECATED, WARN_SYNTAX),
531 "Use of %s is deprecated", s);
535 * experimental text filters for win32 carriage-returns, utf16-to-utf8 and
536 * utf16-to-utf8-reversed.
539 #ifdef PERL_CR_FILTER
543 register const char *s = SvPVX_const(sv);
544 register const char * const e = s + SvCUR(sv);
545 /* outer loop optimized to do nothing if there are no CR-LFs */
547 if (*s++ == '\r' && *s == '\n') {
548 /* hit a CR-LF, need to copy the rest */
549 register char *d = s - 1;
552 if (*s == '\r' && s[1] == '\n')
563 S_cr_textfilter(pTHX_ int idx, SV *sv, int maxlen)
565 const I32 count = FILTER_READ(idx+1, sv, maxlen);
566 if (count > 0 && !maxlen)
574 * Initialize variables. Uses the Perl save_stack to save its state (for
575 * recursive calls to the parser).
579 Perl_lex_start(pTHX_ SV *line)
585 SAVEI32(PL_lex_dojoin);
586 SAVEI32(PL_lex_brackets);
587 SAVEI32(PL_lex_casemods);
588 SAVEI32(PL_lex_starts);
589 SAVEI32(PL_lex_state);
590 SAVEVPTR(PL_lex_inpat);
591 SAVEI32(PL_lex_inwhat);
593 if (PL_lex_state == LEX_KNOWNEXT) {
594 I32 toke = PL_lasttoke;
595 while (--toke >= 0) {
596 SAVEI32(PL_nexttoke[toke].next_type);
597 SAVEVPTR(PL_nexttoke[toke].next_val);
599 SAVEVPTR(PL_nexttoke[toke].next_mad);
601 SAVEI32(PL_lasttoke);
604 SAVESPTR(PL_thistoken);
605 SAVESPTR(PL_thiswhite);
606 SAVESPTR(PL_nextwhite);
607 SAVESPTR(PL_thisopen);
608 SAVESPTR(PL_thisclose);
609 SAVESPTR(PL_thisstuff);
610 SAVEVPTR(PL_thismad);
611 SAVEI32(PL_realtokenstart);
612 SAVEI32(PL_faketokens);
614 SAVEI32(PL_curforce);
616 if (PL_lex_state == LEX_KNOWNEXT) {
617 I32 toke = PL_nexttoke;
618 while (--toke >= 0) {
619 SAVEI32(PL_nexttype[toke]);
620 SAVEVPTR(PL_nextval[toke]);
622 SAVEI32(PL_nexttoke);
625 SAVECOPLINE(PL_curcop);
628 SAVEPPTR(PL_oldbufptr);
629 SAVEPPTR(PL_oldoldbufptr);
630 SAVEPPTR(PL_last_lop);
631 SAVEPPTR(PL_last_uni);
632 SAVEPPTR(PL_linestart);
633 SAVESPTR(PL_linestr);
634 SAVEGENERICPV(PL_lex_brackstack);
635 SAVEGENERICPV(PL_lex_casestack);
636 SAVEDESTRUCTOR_X(restore_rsfp, PL_rsfp);
637 SAVESPTR(PL_lex_stuff);
638 SAVEI32(PL_lex_defer);
639 SAVEI32(PL_sublex_info.sub_inwhat);
640 SAVESPTR(PL_lex_repl);
642 SAVEINT(PL_lex_expect);
644 PL_lex_state = LEX_NORMAL;
648 Newx(PL_lex_brackstack, 120, char);
649 Newx(PL_lex_casestack, 12, char);
651 *PL_lex_casestack = '\0';
663 PL_sublex_info.sub_inwhat = 0;
665 if (SvREADONLY(PL_linestr))
666 PL_linestr = sv_2mortal(newSVsv(PL_linestr));
667 s = SvPV_const(PL_linestr, len);
668 if (!len || s[len-1] != ';') {
669 if (!(SvFLAGS(PL_linestr) & SVs_TEMP))
670 PL_linestr = sv_2mortal(newSVsv(PL_linestr));
671 sv_catpvs(PL_linestr, "\n;");
673 SvTEMP_off(PL_linestr);
674 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = SvPVX(PL_linestr);
675 PL_bufend = PL_bufptr + SvCUR(PL_linestr);
676 PL_last_lop = PL_last_uni = NULL;
682 * Finalizer for lexing operations. Must be called when the parser is
683 * done with the lexer.
690 PL_doextract = FALSE;
695 * This subroutine has nothing to do with tilting, whether at windmills
696 * or pinball tables. Its name is short for "increment line". It
697 * increments the current line number in CopLINE(PL_curcop) and checks
698 * to see whether the line starts with a comment of the form
699 * # line 500 "foo.pm"
700 * If so, it sets the current line number and file to the values in the comment.
704 S_incline(pTHX_ char *s)
712 CopLINE_inc(PL_curcop);
715 while (SPACE_OR_TAB(*s))
717 if (strnEQ(s, "line", 4))
721 if (SPACE_OR_TAB(*s))
725 while (SPACE_OR_TAB(*s))
733 while (SPACE_OR_TAB(*s))
735 if (*s == '"' && (t = strchr(s+1, '"'))) {
740 for (t = s; !isSPACE(*t); t++) ;
743 while (SPACE_OR_TAB(*e) || *e == '\r' || *e == '\f')
745 if (*e != '\n' && *e != '\0')
746 return; /* false alarm */
752 const char * const cf = CopFILE(PL_curcop);
753 STRLEN tmplen = cf ? strlen(cf) : 0;
754 if (tmplen > 7 && strnEQ(cf, "(eval ", 6)) {
755 /* must copy *{"::_<(eval N)[oldfilename:L]"}
756 * to *{"::_<newfilename"} */
757 char smallbuf[256], smallbuf2[256];
758 char *tmpbuf, *tmpbuf2;
760 STRLEN tmplen2 = strlen(s);
761 if (tmplen + 3 < sizeof smallbuf)
764 Newx(tmpbuf, tmplen + 3, char);
765 if (tmplen2 + 3 < sizeof smallbuf2)
768 Newx(tmpbuf2, tmplen2 + 3, char);
769 tmpbuf[0] = tmpbuf2[0] = '_';
770 tmpbuf[1] = tmpbuf2[1] = '<';
771 memcpy(tmpbuf + 2, cf, ++tmplen);
772 memcpy(tmpbuf2 + 2, s, ++tmplen2);
774 gvp = (GV**)hv_fetch(PL_defstash, tmpbuf, tmplen, FALSE);
776 gv2 = *(GV**)hv_fetch(PL_defstash, tmpbuf2, tmplen2, TRUE);
778 gv_init(gv2, PL_defstash, tmpbuf2, tmplen2, FALSE);
779 /* adjust ${"::_<newfilename"} to store the new file name */
780 GvSV(gv2) = newSVpvn(tmpbuf2 + 2, tmplen2 - 2);
781 GvHV(gv2) = (HV*)SvREFCNT_inc(GvHV(*gvp));
782 GvAV(gv2) = (AV*)SvREFCNT_inc(GvAV(*gvp));
784 if (tmpbuf != smallbuf) Safefree(tmpbuf);
785 if (tmpbuf2 != smallbuf2) Safefree(tmpbuf2);
788 CopFILE_free(PL_curcop);
789 CopFILE_set(PL_curcop, s);
792 CopLINE_set(PL_curcop, atoi(n)-1);
796 /* skip space before PL_thistoken */
799 S_skipspace0(pTHX_ register char *s)
806 PL_thiswhite = newSVpvn("",0);
807 sv_catsv(PL_thiswhite, PL_skipwhite);
808 sv_free(PL_skipwhite);
811 PL_realtokenstart = s - SvPVX(PL_linestr);
815 /* skip space after PL_thistoken */
818 S_skipspace1(pTHX_ register char *s)
820 const char *start = s;
821 I32 startoff = start - SvPVX(PL_linestr);
826 start = SvPVX(PL_linestr) + startoff;
827 if (!PL_thistoken && PL_realtokenstart >= 0) {
828 const char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
829 PL_thistoken = newSVpvn(tstart, start - tstart);
831 PL_realtokenstart = -1;
834 PL_nextwhite = newSVpvn("",0);
835 sv_catsv(PL_nextwhite, PL_skipwhite);
836 sv_free(PL_skipwhite);
843 S_skipspace2(pTHX_ register char *s, SV **svp)
846 I32 bufptroff = PL_bufptr - SvPVX(PL_linestr);
847 I32 startoff = start - SvPVX(PL_linestr);
849 PL_bufptr = SvPVX(PL_linestr) + bufptroff;
850 if (!PL_madskills || !svp)
852 start = SvPVX(PL_linestr) + startoff;
853 if (!PL_thistoken && PL_realtokenstart >= 0) {
854 char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
855 PL_thistoken = newSVpvn(tstart, start - tstart);
856 PL_realtokenstart = -1;
860 *svp = newSVpvn("",0);
861 sv_setsv(*svp, PL_skipwhite);
862 sv_free(PL_skipwhite);
872 * Called to gobble the appropriate amount and type of whitespace.
873 * Skips comments as well.
877 S_skipspace(pTHX_ register char *s)
882 int startoff = s - SvPVX(PL_linestr);
885 sv_free(PL_skipwhite);
890 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
891 while (s < PL_bufend && SPACE_OR_TAB(*s))
901 SSize_t oldprevlen, oldoldprevlen;
902 SSize_t oldloplen = 0, oldunilen = 0;
903 while (s < PL_bufend && isSPACE(*s)) {
904 if (*s++ == '\n' && PL_in_eval && !PL_rsfp)
909 if (s < PL_bufend && *s == '#') {
910 while (s < PL_bufend && *s != '\n')
914 if (PL_in_eval && !PL_rsfp) {
921 /* only continue to recharge the buffer if we're at the end
922 * of the buffer, we're not reading from a source filter, and
923 * we're in normal lexing mode
925 if (s < PL_bufend || !PL_rsfp || PL_sublex_info.sub_inwhat ||
926 PL_lex_state == LEX_FORMLINE)
933 /* try to recharge the buffer */
935 curoff = s - SvPVX(PL_linestr);
938 if ((s = filter_gets(PL_linestr, PL_rsfp,
939 (prevlen = SvCUR(PL_linestr)))) == NULL)
942 if (PL_madskills && curoff != startoff) {
944 PL_skipwhite = newSVpvn("",0);
945 sv_catpvn(PL_skipwhite, SvPVX(PL_linestr) + startoff,
949 /* mustn't throw out old stuff yet if madpropping */
950 SvCUR(PL_linestr) = curoff;
951 s = SvPVX(PL_linestr) + curoff;
953 if (curoff && s[-1] == '\n')
957 /* end of file. Add on the -p or -n magic */
958 /* XXX these shouldn't really be added here, can't set PL_faketokens */
962 ";}continue{print or die qq(-p destination: $!\\n);}");
965 ";}continue{print or die qq(-p destination: $!\\n);}");
967 PL_minus_n = PL_minus_p = 0;
969 else if (PL_minus_n) {
971 sv_catpvn(PL_linestr, ";}", 2);
973 sv_setpvn(PL_linestr, ";}", 2);
979 sv_catpvn(PL_linestr,";", 1);
981 sv_setpvn(PL_linestr,";", 1);
984 /* reset variables for next time we lex */
985 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = s = PL_linestart
991 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
992 PL_last_lop = PL_last_uni = NULL;
994 /* Close the filehandle. Could be from -P preprocessor,
995 * STDIN, or a regular file. If we were reading code from
996 * STDIN (because the commandline held no -e or filename)
997 * then we don't close it, we reset it so the code can
998 * read from STDIN too.
1001 if (PL_preprocess && !PL_in_eval)
1002 (void)PerlProc_pclose(PL_rsfp);
1003 else if ((PerlIO*)PL_rsfp == PerlIO_stdin())
1004 PerlIO_clearerr(PL_rsfp);
1006 (void)PerlIO_close(PL_rsfp);
1011 /* not at end of file, so we only read another line */
1012 /* make corresponding updates to old pointers, for yyerror() */
1013 oldprevlen = PL_oldbufptr - PL_bufend;
1014 oldoldprevlen = PL_oldoldbufptr - PL_bufend;
1016 oldunilen = PL_last_uni - PL_bufend;
1018 oldloplen = PL_last_lop - PL_bufend;
1019 PL_linestart = PL_bufptr = s + prevlen;
1020 PL_bufend = s + SvCUR(PL_linestr);
1022 PL_oldbufptr = s + oldprevlen;
1023 PL_oldoldbufptr = s + oldoldprevlen;
1025 PL_last_uni = s + oldunilen;
1027 PL_last_lop = s + oldloplen;
1030 /* debugger active and we're not compiling the debugger code,
1031 * so store the line into the debugger's array of lines
1033 if (PERLDB_LINE && PL_curstash != PL_debstash) {
1034 SV * const sv = newSV(0);
1036 sv_upgrade(sv, SVt_PVMG);
1037 sv_setpvn(sv,PL_bufptr,PL_bufend-PL_bufptr);
1040 av_store(CopFILEAVx(PL_curcop),(I32)CopLINE(PL_curcop),sv);
1048 PL_skipwhite = newSVpvn("",0);
1049 curoff = s - SvPVX(PL_linestr);
1050 if (curoff - startoff)
1051 sv_catpvn(PL_skipwhite, SvPVX(PL_linestr) + startoff,
1060 * Check the unary operators to ensure there's no ambiguity in how they're
1061 * used. An ambiguous piece of code would be:
1063 * This doesn't mean rand() + 5. Because rand() is a unary operator,
1064 * the +5 is its argument.
1074 if (PL_oldoldbufptr != PL_last_uni)
1076 while (isSPACE(*PL_last_uni))
1078 for (s = PL_last_uni; isALNUM_lazy_if(s,UTF) || *s == '-'; s++)
1080 if ((t = strchr(s, '(')) && t < PL_bufptr)
1083 if (ckWARN_d(WARN_AMBIGUOUS)){
1084 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
1085 "Warning: Use of \"%.*s\" without parentheses is ambiguous",
1086 (int)(s - PL_last_uni), PL_last_uni);
1091 * LOP : macro to build a list operator. Its behaviour has been replaced
1092 * with a subroutine, S_lop() for which LOP is just another name.
1095 #define LOP(f,x) return lop(f,x,s)
1099 * Build a list operator (or something that might be one). The rules:
1100 * - if we have a next token, then it's a list operator [why?]
1101 * - if the next thing is an opening paren, then it's a function
1102 * - else it's a list operator
1106 S_lop(pTHX_ I32 f, int x, char *s)
1113 PL_last_lop = PL_oldbufptr;
1114 PL_last_lop_op = (OPCODE)f;
1117 return REPORT(LSTOP);
1120 return REPORT(LSTOP);
1123 return REPORT(FUNC);
1126 return REPORT(FUNC);
1128 return REPORT(LSTOP);
1134 * Sets up for an eventual force_next(). start_force(0) basically does
1135 * an unshift, while start_force(-1) does a push. yylex removes items
1140 S_start_force(pTHX_ int where)
1144 if (where < 0) /* so people can duplicate start_force(PL_curforce) */
1145 where = PL_lasttoke;
1146 assert(PL_curforce < 0 || PL_curforce == where);
1147 if (PL_curforce != where) {
1148 for (i = PL_lasttoke; i > where; --i) {
1149 PL_nexttoke[i] = PL_nexttoke[i-1];
1153 if (PL_curforce < 0) /* in case of duplicate start_force() */
1154 Zero(&PL_nexttoke[where], 1, NEXTTOKE);
1155 PL_curforce = where;
1158 curmad('^', newSVpvn("",0));
1159 CURMAD('_', PL_nextwhite);
1164 S_curmad(pTHX_ char slot, SV *sv)
1170 if (PL_curforce < 0)
1171 where = &PL_thismad;
1173 where = &PL_nexttoke[PL_curforce].next_mad;
1176 sv_setpvn(sv, "", 0);
1179 if (UTF && is_utf8_string((U8*)SvPVX(sv), SvCUR(sv)))
1181 else if (PL_encoding) {
1182 sv_recode_to_utf8(sv, PL_encoding);
1187 /* keep a slot open for the head of the list? */
1188 if (slot != '_' && *where && (*where)->mad_key == '^') {
1189 (*where)->mad_key = slot;
1190 sv_free((*where)->mad_val);
1191 (*where)->mad_val = (void*)sv;
1194 addmad(newMADsv(slot, sv), where, 0);
1197 # define start_force(where) NOOP
1198 # define curmad(slot, sv) NOOP
1203 * When the lexer realizes it knows the next token (for instance,
1204 * it is reordering tokens for the parser) then it can call S_force_next
1205 * to know what token to return the next time the lexer is called. Caller
1206 * will need to set PL_nextval[] (or PL_nexttoke[].next_val with PERL_MAD),
1207 * and possibly PL_expect to ensure the lexer handles the token correctly.
1211 S_force_next(pTHX_ I32 type)
1215 if (PL_curforce < 0)
1216 start_force(PL_lasttoke);
1217 PL_nexttoke[PL_curforce].next_type = type;
1218 if (PL_lex_state != LEX_KNOWNEXT)
1219 PL_lex_defer = PL_lex_state;
1220 PL_lex_state = LEX_KNOWNEXT;
1221 PL_lex_expect = PL_expect;
1224 PL_nexttype[PL_nexttoke] = type;
1226 if (PL_lex_state != LEX_KNOWNEXT) {
1227 PL_lex_defer = PL_lex_state;
1228 PL_lex_expect = PL_expect;
1229 PL_lex_state = LEX_KNOWNEXT;
1235 S_newSV_maybe_utf8(pTHX_ const char *start, STRLEN len)
1238 SV * const sv = newSVpvn(start,len);
1239 if (UTF && !IN_BYTES && is_utf8_string((const U8*)start, len))
1246 * When the lexer knows the next thing is a word (for instance, it has
1247 * just seen -> and it knows that the next char is a word char, then
1248 * it calls S_force_word to stick the next word into the PL_next lookahead.
1251 * char *start : buffer position (must be within PL_linestr)
1252 * int token : PL_next will be this type of bare word (e.g., METHOD,WORD)
1253 * int check_keyword : if true, Perl checks to make sure the word isn't
1254 * a keyword (do this if the word is a label, e.g. goto FOO)
1255 * int allow_pack : if true, : characters will also be allowed (require,
1256 * use, etc. do this)
1257 * int allow_initial_tick : used by the "sub" lexer only.
1261 S_force_word(pTHX_ register char *start, int token, int check_keyword, int allow_pack, int allow_initial_tick)
1267 start = SKIPSPACE1(start);
1269 if (isIDFIRST_lazy_if(s,UTF) ||
1270 (allow_pack && *s == ':') ||
1271 (allow_initial_tick && *s == '\'') )
1273 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, allow_pack, &len);
1274 if (check_keyword && keyword(PL_tokenbuf, len))
1276 start_force(PL_curforce);
1278 curmad('X', newSVpvn(start,s-start));
1279 if (token == METHOD) {
1284 PL_expect = XOPERATOR;
1287 NEXTVAL_NEXTTOKE.opval
1288 = (OP*)newSVOP(OP_CONST,0,
1289 S_newSV_maybe_utf8(aTHX_ PL_tokenbuf, len));
1290 NEXTVAL_NEXTTOKE.opval->op_private |= OPpCONST_BARE;
1298 * Called when the lexer wants $foo *foo &foo etc, but the program
1299 * text only contains the "foo" portion. The first argument is a pointer
1300 * to the "foo", and the second argument is the type symbol to prefix.
1301 * Forces the next token to be a "WORD".
1302 * Creates the symbol if it didn't already exist (via gv_fetchpv()).
1306 S_force_ident(pTHX_ register const char *s, int kind)
1310 const STRLEN len = strlen(s);
1311 OP* const o = (OP*)newSVOP(OP_CONST, 0, newSVpvn(s, len));
1312 start_force(PL_curforce);
1313 NEXTVAL_NEXTTOKE.opval = o;
1316 o->op_private = OPpCONST_ENTERED;
1317 /* XXX see note in pp_entereval() for why we forgo typo
1318 warnings if the symbol must be introduced in an eval.
1320 gv_fetchpvn_flags(s, len,
1321 PL_in_eval ? (GV_ADDMULTI | GV_ADDINEVAL)
1323 kind == '$' ? SVt_PV :
1324 kind == '@' ? SVt_PVAV :
1325 kind == '%' ? SVt_PVHV :
1333 Perl_str_to_version(pTHX_ SV *sv)
1338 const char *start = SvPV_const(sv,len);
1339 const char * const end = start + len;
1340 const bool utf = SvUTF8(sv) ? TRUE : FALSE;
1341 while (start < end) {
1345 n = utf8n_to_uvchr((U8*)start, len, &skip, 0);
1350 retval += ((NV)n)/nshift;
1359 * Forces the next token to be a version number.
1360 * If the next token appears to be an invalid version number, (e.g. "v2b"),
1361 * and if "guessing" is TRUE, then no new token is created (and the caller
1362 * must use an alternative parsing method).
1366 S_force_version(pTHX_ char *s, int guessing)
1372 I32 startoff = s - SvPVX(PL_linestr);
1381 while (isDIGIT(*d) || *d == '_' || *d == '.')
1385 start_force(PL_curforce);
1386 curmad('X', newSVpvn(s,d-s));
1389 if (*d == ';' || isSPACE(*d) || *d == '}' || !*d) {
1391 s = scan_num(s, &yylval);
1392 version = yylval.opval;
1393 ver = cSVOPx(version)->op_sv;
1394 if (SvPOK(ver) && !SvNIOK(ver)) {
1395 SvUPGRADE(ver, SVt_PVNV);
1396 SvNV_set(ver, str_to_version(ver));
1397 SvNOK_on(ver); /* hint that it is a version */
1400 else if (guessing) {
1403 sv_free(PL_nextwhite); /* let next token collect whitespace */
1405 s = SvPVX(PL_linestr) + startoff;
1413 if (PL_madskills && !version) {
1414 sv_free(PL_nextwhite); /* let next token collect whitespace */
1416 s = SvPVX(PL_linestr) + startoff;
1419 /* NOTE: The parser sees the package name and the VERSION swapped */
1420 start_force(PL_curforce);
1421 NEXTVAL_NEXTTOKE.opval = version;
1429 * Tokenize a quoted string passed in as an SV. It finds the next
1430 * chunk, up to end of string or a backslash. It may make a new
1431 * SV containing that chunk (if HINT_NEW_STRING is on). It also
1436 S_tokeq(pTHX_ SV *sv)
1440 register char *send;
1448 s = SvPV_force(sv, len);
1449 if (SvTYPE(sv) >= SVt_PVIV && SvIVX(sv) == -1)
1452 while (s < send && *s != '\\')
1457 if ( PL_hints & HINT_NEW_STRING ) {
1458 pv = sv_2mortal(newSVpvn(SvPVX_const(pv), len));
1464 if (s + 1 < send && (s[1] == '\\'))
1465 s++; /* all that, just for this */
1470 SvCUR_set(sv, d - SvPVX_const(sv));
1472 if ( PL_hints & HINT_NEW_STRING )
1473 return new_constant(NULL, 0, "q", sv, pv, "q");
1478 * Now come three functions related to double-quote context,
1479 * S_sublex_start, S_sublex_push, and S_sublex_done. They're used when
1480 * converting things like "\u\Lgnat" into ucfirst(lc("gnat")). They
1481 * interact with PL_lex_state, and create fake ( ... ) argument lists
1482 * to handle functions and concatenation.
1483 * They assume that whoever calls them will be setting up a fake
1484 * join call, because each subthing puts a ',' after it. This lets
1487 * join($, , 'lower ', lcfirst( 'uPpEr', ) ,)
1489 * (I'm not sure whether the spurious commas at the end of lcfirst's
1490 * arguments and join's arguments are created or not).
1495 * Assumes that yylval.ival is the op we're creating (e.g. OP_LCFIRST).
1497 * Pattern matching will set PL_lex_op to the pattern-matching op to
1498 * make (we return THING if yylval.ival is OP_NULL, PMFUNC otherwise).
1500 * OP_CONST and OP_READLINE are easy--just make the new op and return.
1502 * Everything else becomes a FUNC.
1504 * Sets PL_lex_state to LEX_INTERPPUSH unless (ival was OP_NULL or we
1505 * had an OP_CONST or OP_READLINE). This just sets us up for a
1506 * call to S_sublex_push().
1510 S_sublex_start(pTHX)
1513 register const I32 op_type = yylval.ival;
1515 if (op_type == OP_NULL) {
1516 yylval.opval = PL_lex_op;
1520 if (op_type == OP_CONST || op_type == OP_READLINE) {
1521 SV *sv = tokeq(PL_lex_stuff);
1523 if (SvTYPE(sv) == SVt_PVIV) {
1524 /* Overloaded constants, nothing fancy: Convert to SVt_PV: */
1526 const char * const p = SvPV_const(sv, len);
1527 SV * const nsv = newSVpvn(p, len);
1533 yylval.opval = (OP*)newSVOP(op_type, 0, sv);
1534 PL_lex_stuff = NULL;
1535 /* Allow <FH> // "foo" */
1536 if (op_type == OP_READLINE)
1537 PL_expect = XTERMORDORDOR;
1541 PL_sublex_info.super_state = PL_lex_state;
1542 PL_sublex_info.sub_inwhat = op_type;
1543 PL_sublex_info.sub_op = PL_lex_op;
1544 PL_lex_state = LEX_INTERPPUSH;
1548 yylval.opval = PL_lex_op;
1558 * Create a new scope to save the lexing state. The scope will be
1559 * ended in S_sublex_done. Returns a '(', starting the function arguments
1560 * to the uc, lc, etc. found before.
1561 * Sets PL_lex_state to LEX_INTERPCONCAT.
1570 PL_lex_state = PL_sublex_info.super_state;
1571 SAVEI32(PL_lex_dojoin);
1572 SAVEI32(PL_lex_brackets);
1573 SAVEI32(PL_lex_casemods);
1574 SAVEI32(PL_lex_starts);
1575 SAVEI32(PL_lex_state);
1576 SAVEVPTR(PL_lex_inpat);
1577 SAVEI32(PL_lex_inwhat);
1578 SAVECOPLINE(PL_curcop);
1579 SAVEPPTR(PL_bufptr);
1580 SAVEPPTR(PL_bufend);
1581 SAVEPPTR(PL_oldbufptr);
1582 SAVEPPTR(PL_oldoldbufptr);
1583 SAVEPPTR(PL_last_lop);
1584 SAVEPPTR(PL_last_uni);
1585 SAVEPPTR(PL_linestart);
1586 SAVESPTR(PL_linestr);
1587 SAVEGENERICPV(PL_lex_brackstack);
1588 SAVEGENERICPV(PL_lex_casestack);
1590 PL_linestr = PL_lex_stuff;
1591 PL_lex_stuff = NULL;
1593 PL_bufend = PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart
1594 = SvPVX(PL_linestr);
1595 PL_bufend += SvCUR(PL_linestr);
1596 PL_last_lop = PL_last_uni = NULL;
1597 SAVEFREESV(PL_linestr);
1599 PL_lex_dojoin = FALSE;
1600 PL_lex_brackets = 0;
1601 Newx(PL_lex_brackstack, 120, char);
1602 Newx(PL_lex_casestack, 12, char);
1603 PL_lex_casemods = 0;
1604 *PL_lex_casestack = '\0';
1606 PL_lex_state = LEX_INTERPCONCAT;
1607 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
1609 PL_lex_inwhat = PL_sublex_info.sub_inwhat;
1610 if (PL_lex_inwhat == OP_MATCH || PL_lex_inwhat == OP_QR || PL_lex_inwhat == OP_SUBST)
1611 PL_lex_inpat = PL_sublex_info.sub_op;
1613 PL_lex_inpat = NULL;
1620 * Restores lexer state after a S_sublex_push.
1627 if (!PL_lex_starts++) {
1628 SV * const sv = newSVpvs("");
1629 if (SvUTF8(PL_linestr))
1631 PL_expect = XOPERATOR;
1632 yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
1636 if (PL_lex_casemods) { /* oops, we've got some unbalanced parens */
1637 PL_lex_state = LEX_INTERPCASEMOD;
1641 /* Is there a right-hand side to take care of? (s//RHS/ or tr//RHS/) */
1642 if (PL_lex_repl && (PL_lex_inwhat == OP_SUBST || PL_lex_inwhat == OP_TRANS)) {
1643 PL_linestr = PL_lex_repl;
1645 PL_bufend = PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart = SvPVX(PL_linestr);
1646 PL_bufend += SvCUR(PL_linestr);
1647 PL_last_lop = PL_last_uni = NULL;
1648 SAVEFREESV(PL_linestr);
1649 PL_lex_dojoin = FALSE;
1650 PL_lex_brackets = 0;
1651 PL_lex_casemods = 0;
1652 *PL_lex_casestack = '\0';
1654 if (SvEVALED(PL_lex_repl)) {
1655 PL_lex_state = LEX_INTERPNORMAL;
1657 /* we don't clear PL_lex_repl here, so that we can check later
1658 whether this is an evalled subst; that means we rely on the
1659 logic to ensure sublex_done() is called again only via the
1660 branch (in yylex()) that clears PL_lex_repl, else we'll loop */
1663 PL_lex_state = LEX_INTERPCONCAT;
1673 PL_endwhite = newSVpvn("",0);
1674 sv_catsv(PL_endwhite, PL_thiswhite);
1678 sv_setpvn(PL_thistoken,"",0);
1680 PL_realtokenstart = -1;
1684 PL_bufend = SvPVX(PL_linestr);
1685 PL_bufend += SvCUR(PL_linestr);
1686 PL_expect = XOPERATOR;
1687 PL_sublex_info.sub_inwhat = 0;
1695 Extracts a pattern, double-quoted string, or transliteration. This
1698 It looks at lex_inwhat and PL_lex_inpat to find out whether it's
1699 processing a pattern (PL_lex_inpat is true), a transliteration
1700 (lex_inwhat & OP_TRANS is true), or a double-quoted string.
1702 Returns a pointer to the character scanned up to. Iff this is
1703 advanced from the start pointer supplied (ie if anything was
1704 successfully parsed), will leave an OP for the substring scanned
1705 in yylval. Caller must intuit reason for not parsing further
1706 by looking at the next characters herself.
1710 double-quoted style: \r and \n
1711 regexp special ones: \D \s
1713 backrefs: \1 (deprecated in substitution replacements)
1714 case and quoting: \U \Q \E
1715 stops on @ and $, but not for $ as tail anchor
1717 In transliterations:
1718 characters are VERY literal, except for - not at the start or end
1719 of the string, which indicates a range. scan_const expands the
1720 range to the full set of intermediate characters.
1722 In double-quoted strings:
1724 double-quoted style: \r and \n
1726 backrefs: \1 (deprecated)
1727 case and quoting: \U \Q \E
1730 scan_const does *not* construct ops to handle interpolated strings.
1731 It stops processing as soon as it finds an embedded $ or @ variable
1732 and leaves it to the caller to work out what's going on.
1734 @ in pattern could be: @foo, @{foo}, @$foo, @'foo, @::foo.
1736 $ in pattern could be $foo or could be tail anchor. Assumption:
1737 it's a tail anchor if $ is the last thing in the string, or if it's
1738 followed by one of ")| \n\t"
1740 \1 (backreferences) are turned into $1
1742 The structure of the code is
1743 while (there's a character to process) {
1744 handle transliteration ranges
1745 skip regexp comments
1746 skip # initiated comments in //x patterns
1747 check for embedded @foo
1748 check for embedded scalars
1750 leave intact backslashes from leave (below)
1751 deprecate \1 in strings and sub replacements
1752 handle string-changing backslashes \l \U \Q \E, etc.
1753 switch (what was escaped) {
1754 handle - in a transliteration (becomes a literal -)
1755 handle \132 octal characters
1756 handle 0x15 hex characters
1757 handle \cV (control V)
1758 handle printf backslashes (\f, \r, \n, etc)
1760 } (end if backslash)
1761 } (end while character to read)
1766 S_scan_const(pTHX_ char *start)
1769 register char *send = PL_bufend; /* end of the constant */
1770 SV *sv = newSV(send - start); /* sv for the constant */
1771 register char *s = start; /* start of the constant */
1772 register char *d = SvPVX(sv); /* destination for copies */
1773 bool dorange = FALSE; /* are we in a translit range? */
1774 bool didrange = FALSE; /* did we just finish a range? */
1775 I32 has_utf8 = FALSE; /* Output constant is UTF8 */
1776 I32 this_utf8 = UTF; /* The source string is assumed to be UTF8 */
1779 UV literal_endpoint = 0;
1782 const char * const leaveit = /* set of acceptably-backslashed characters */
1784 ? "\\.^$@AGZdDwWsSbBpPXC+*?|()-nrtfeaxz0123456789[{]} \t\n\r\f\v#"
1787 if (PL_lex_inwhat == OP_TRANS && PL_sublex_info.sub_op) {
1788 /* If we are doing a trans and we know we want UTF8 set expectation */
1789 has_utf8 = PL_sublex_info.sub_op->op_private & (OPpTRANS_FROM_UTF|OPpTRANS_TO_UTF);
1790 this_utf8 = PL_sublex_info.sub_op->op_private & (PL_lex_repl ? OPpTRANS_FROM_UTF : OPpTRANS_TO_UTF);
1794 while (s < send || dorange) {
1795 /* get transliterations out of the way (they're most literal) */
1796 if (PL_lex_inwhat == OP_TRANS) {
1797 /* expand a range A-Z to the full set of characters. AIE! */
1799 I32 i; /* current expanded character */
1800 I32 min; /* first character in range */
1801 I32 max; /* last character in range */
1804 char * const c = (char*)utf8_hop((U8*)d, -1);
1808 *c = (char)UTF_TO_NATIVE(0xff);
1809 /* mark the range as done, and continue */
1815 i = d - SvPVX_const(sv); /* remember current offset */
1816 SvGROW(sv, SvLEN(sv) + 256); /* never more than 256 chars in a range */
1817 d = SvPVX(sv) + i; /* refresh d after realloc */
1818 d -= 2; /* eat the first char and the - */
1820 min = (U8)*d; /* first char in range */
1821 max = (U8)d[1]; /* last char in range */
1825 "Invalid range \"%c-%c\" in transliteration operator",
1826 (char)min, (char)max);
1830 if (literal_endpoint == 2 &&
1831 ((isLOWER(min) && isLOWER(max)) ||
1832 (isUPPER(min) && isUPPER(max)))) {
1834 for (i = min; i <= max; i++)
1836 *d++ = NATIVE_TO_NEED(has_utf8,i);
1838 for (i = min; i <= max; i++)
1840 *d++ = NATIVE_TO_NEED(has_utf8,i);
1845 for (i = min; i <= max; i++)
1848 /* mark the range as done, and continue */
1852 literal_endpoint = 0;
1857 /* range begins (ignore - as first or last char) */
1858 else if (*s == '-' && s+1 < send && s != start) {
1860 Perl_croak(aTHX_ "Ambiguous range in transliteration operator");
1863 *d++ = (char)UTF_TO_NATIVE(0xff); /* use illegal utf8 byte--see pmtrans */
1873 literal_endpoint = 0;
1878 /* if we get here, we're not doing a transliteration */
1880 /* skip for regexp comments /(?#comment)/ and code /(?{code})/,
1881 except for the last char, which will be done separately. */
1882 else if (*s == '(' && PL_lex_inpat && s[1] == '?') {
1884 while (s+1 < send && *s != ')')
1885 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
1887 else if (s[2] == '{' /* This should match regcomp.c */
1888 || ((s[2] == 'p' || s[2] == '?') && s[3] == '{'))
1891 char *regparse = s + (s[2] == '{' ? 3 : 4);
1894 while (count && (c = *regparse)) {
1895 if (c == '\\' && regparse[1])
1903 if (*regparse != ')')
1904 regparse--; /* Leave one char for continuation. */
1905 while (s < regparse)
1906 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
1910 /* likewise skip #-initiated comments in //x patterns */
1911 else if (*s == '#' && PL_lex_inpat &&
1912 ((PMOP*)PL_lex_inpat)->op_pmflags & PMf_EXTENDED) {
1913 while (s+1 < send && *s != '\n')
1914 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
1917 /* check for embedded arrays
1918 (@foo, @::foo, @'foo, @{foo}, @$foo, @+, @-)
1920 else if (*s == '@' && s[1]
1921 && (isALNUM_lazy_if(s+1,UTF) || strchr(":'{$+-", s[1])))
1924 /* check for embedded scalars. only stop if we're sure it's a
1927 else if (*s == '$') {
1928 if (!PL_lex_inpat) /* not a regexp, so $ must be var */
1930 if (s + 1 < send && !strchr("()| \r\n\t", s[1]))
1931 break; /* in regexp, $ might be tail anchor */
1934 /* End of else if chain - OP_TRANS rejoin rest */
1937 if (*s == '\\' && s+1 < send) {
1940 /* some backslashes we leave behind */
1941 if (*leaveit && *s && strchr(leaveit, *s)) {
1942 *d++ = NATIVE_TO_NEED(has_utf8,'\\');
1943 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
1947 /* deprecate \1 in strings and substitution replacements */
1948 if (PL_lex_inwhat == OP_SUBST && !PL_lex_inpat &&
1949 isDIGIT(*s) && *s != '0' && !isDIGIT(s[1]))
1951 if (ckWARN(WARN_SYNTAX))
1952 Perl_warner(aTHX_ packWARN(WARN_SYNTAX), "\\%c better written as $%c", *s, *s);
1957 /* string-change backslash escapes */
1958 if (PL_lex_inwhat != OP_TRANS && *s && strchr("lLuUEQ", *s)) {
1963 /* if we get here, it's either a quoted -, or a digit */
1966 /* quoted - in transliterations */
1968 if (PL_lex_inwhat == OP_TRANS) {
1978 Perl_warner(aTHX_ packWARN(WARN_MISC),
1979 "Unrecognized escape \\%c passed through",
1981 /* default action is to copy the quoted character */
1982 goto default_action;
1985 /* \132 indicates an octal constant */
1986 case '0': case '1': case '2': case '3':
1987 case '4': case '5': case '6': case '7':
1991 uv = grok_oct(s, &len, &flags, NULL);
1994 goto NUM_ESCAPE_INSERT;
1996 /* \x24 indicates a hex constant */
2000 char* const e = strchr(s, '}');
2001 I32 flags = PERL_SCAN_ALLOW_UNDERSCORES |
2002 PERL_SCAN_DISALLOW_PREFIX;
2007 yyerror("Missing right brace on \\x{}");
2011 uv = grok_hex(s, &len, &flags, NULL);
2017 I32 flags = PERL_SCAN_DISALLOW_PREFIX;
2018 uv = grok_hex(s, &len, &flags, NULL);
2024 /* Insert oct or hex escaped character.
2025 * There will always enough room in sv since such
2026 * escapes will be longer than any UTF-8 sequence
2027 * they can end up as. */
2029 /* We need to map to chars to ASCII before doing the tests
2032 if (!UNI_IS_INVARIANT(NATIVE_TO_UNI(uv))) {
2033 if (!has_utf8 && uv > 255) {
2034 /* Might need to recode whatever we have
2035 * accumulated so far if it contains any
2038 * (Can't we keep track of that and avoid
2039 * this rescan? --jhi)
2043 for (c = (U8 *) SvPVX(sv); c < (U8 *)d; c++) {
2044 if (!NATIVE_IS_INVARIANT(*c)) {
2049 const STRLEN offset = d - SvPVX_const(sv);
2051 d = SvGROW(sv, SvLEN(sv) + hicount + 1) + offset;
2055 while (src >= (const U8 *)SvPVX_const(sv)) {
2056 if (!NATIVE_IS_INVARIANT(*src)) {
2057 const U8 ch = NATIVE_TO_ASCII(*src);
2058 *dst-- = (U8)UTF8_EIGHT_BIT_LO(ch);
2059 *dst-- = (U8)UTF8_EIGHT_BIT_HI(ch);
2069 if (has_utf8 || uv > 255) {
2070 d = (char*)uvchr_to_utf8((U8*)d, uv);
2072 if (PL_lex_inwhat == OP_TRANS &&
2073 PL_sublex_info.sub_op) {
2074 PL_sublex_info.sub_op->op_private |=
2075 (PL_lex_repl ? OPpTRANS_FROM_UTF
2088 /* \N{LATIN SMALL LETTER A} is a named character */
2092 char* e = strchr(s, '}');
2098 yyerror("Missing right brace on \\N{}");
2102 if (e > s + 2 && s[1] == 'U' && s[2] == '+') {
2104 I32 flags = PERL_SCAN_ALLOW_UNDERSCORES |
2105 PERL_SCAN_DISALLOW_PREFIX;
2108 uv = grok_hex(s, &len, &flags, NULL);
2110 goto NUM_ESCAPE_INSERT;
2112 res = newSVpvn(s + 1, e - s - 1);
2113 res = new_constant( NULL, 0, "charnames",
2114 res, NULL, "\\N{...}" );
2116 sv_utf8_upgrade(res);
2117 str = SvPV_const(res,len);
2118 #ifdef EBCDIC_NEVER_MIND
2119 /* charnames uses pack U and that has been
2120 * recently changed to do the below uni->native
2121 * mapping, so this would be redundant (and wrong,
2122 * the code point would be doubly converted).
2123 * But leave this in just in case the pack U change
2124 * gets revoked, but the semantics is still
2125 * desireable for charnames. --jhi */
2127 UV uv = utf8_to_uvchr((const U8*)str, 0);
2130 U8 tmpbuf[UTF8_MAXBYTES+1], *d;
2132 d = uvchr_to_utf8(tmpbuf, UNI_TO_NATIVE(uv));
2133 sv_setpvn(res, (char *)tmpbuf, d - tmpbuf);
2134 str = SvPV_const(res, len);
2138 if (!has_utf8 && SvUTF8(res)) {
2139 const char * const ostart = SvPVX_const(sv);
2140 SvCUR_set(sv, d - ostart);
2143 sv_utf8_upgrade(sv);
2144 /* this just broke our allocation above... */
2145 SvGROW(sv, (STRLEN)(send - start));
2146 d = SvPVX(sv) + SvCUR(sv);
2149 if (len > (STRLEN)(e - s + 4)) { /* I _guess_ 4 is \N{} --jhi */
2150 const char * const odest = SvPVX_const(sv);
2152 SvGROW(sv, (SvLEN(sv) + len - (e - s + 4)));
2153 d = SvPVX(sv) + (d - odest);
2155 Copy(str, d, len, char);
2162 yyerror("Missing braces on \\N{}");
2165 /* \c is a control character */
2174 *d++ = NATIVE_TO_NEED(has_utf8,toCTRL(c));
2177 yyerror("Missing control char name in \\c");
2181 /* printf-style backslashes, formfeeds, newlines, etc */
2183 *d++ = NATIVE_TO_NEED(has_utf8,'\b');
2186 *d++ = NATIVE_TO_NEED(has_utf8,'\n');
2189 *d++ = NATIVE_TO_NEED(has_utf8,'\r');
2192 *d++ = NATIVE_TO_NEED(has_utf8,'\f');
2195 *d++ = NATIVE_TO_NEED(has_utf8,'\t');
2198 *d++ = ASCII_TO_NEED(has_utf8,'\033');
2201 *d++ = ASCII_TO_NEED(has_utf8,'\007');
2207 } /* end if (backslash) */
2214 /* If we started with encoded form, or already know we want it
2215 and then encode the next character */
2216 if ((has_utf8 || this_utf8) && !NATIVE_IS_INVARIANT((U8)(*s))) {
2218 const UV nextuv = (this_utf8) ? utf8n_to_uvchr((U8*)s, send - s, &len, 0) : (UV) ((U8) *s);
2219 const STRLEN need = UNISKIP(NATIVE_TO_UNI(nextuv));
2222 /* encoded value larger than old, need extra space (NOTE: SvCUR() not set here) */
2223 const STRLEN off = d - SvPVX_const(sv);
2224 d = SvGROW(sv, SvLEN(sv) + (need-len)) + off;
2226 d = (char*)uvchr_to_utf8((U8*)d, nextuv);
2230 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2232 } /* while loop to process each character */
2234 /* terminate the string and set up the sv */
2236 SvCUR_set(sv, d - SvPVX_const(sv));
2237 if (SvCUR(sv) >= SvLEN(sv))
2238 Perl_croak(aTHX_ "panic: constant overflowed allocated space");
2241 if (PL_encoding && !has_utf8) {
2242 sv_recode_to_utf8(sv, PL_encoding);
2248 if (PL_lex_inwhat == OP_TRANS && PL_sublex_info.sub_op) {
2249 PL_sublex_info.sub_op->op_private |=
2250 (PL_lex_repl ? OPpTRANS_FROM_UTF : OPpTRANS_TO_UTF);
2254 /* shrink the sv if we allocated more than we used */
2255 if (SvCUR(sv) + 5 < SvLEN(sv)) {
2256 SvPV_shrink_to_cur(sv);
2259 /* return the substring (via yylval) only if we parsed anything */
2260 if (s > PL_bufptr) {
2261 if ( PL_hints & ( PL_lex_inpat ? HINT_NEW_RE : HINT_NEW_STRING ) )
2262 sv = new_constant(start, s - start, (PL_lex_inpat ? "qr" : "q"),
2264 ( PL_lex_inwhat == OP_TRANS
2266 : ( (PL_lex_inwhat == OP_SUBST && !PL_lex_inpat)
2269 yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
2276 * Returns TRUE if there's more to the expression (e.g., a subscript),
2279 * It deals with "$foo[3]" and /$foo[3]/ and /$foo[0123456789$]+/
2281 * ->[ and ->{ return TRUE
2282 * { and [ outside a pattern are always subscripts, so return TRUE
2283 * if we're outside a pattern and it's not { or [, then return FALSE
2284 * if we're in a pattern and the first char is a {
2285 * {4,5} (any digits around the comma) returns FALSE
2286 * if we're in a pattern and the first char is a [
2288 * [SOMETHING] has a funky algorithm to decide whether it's a
2289 * character class or not. It has to deal with things like
2290 * /$foo[-3]/ and /$foo[$bar]/ as well as /$foo[$\d]+/
2291 * anything else returns TRUE
2294 /* This is the one truly awful dwimmer necessary to conflate C and sed. */
2297 S_intuit_more(pTHX_ register char *s)
2300 if (PL_lex_brackets)
2302 if (*s == '-' && s[1] == '>' && (s[2] == '[' || s[2] == '{'))
2304 if (*s != '{' && *s != '[')
2309 /* In a pattern, so maybe we have {n,m}. */
2326 /* On the other hand, maybe we have a character class */
2329 if (*s == ']' || *s == '^')
2332 /* this is terrifying, and it works */
2333 int weight = 2; /* let's weigh the evidence */
2335 unsigned char un_char = 255, last_un_char;
2336 const char * const send = strchr(s,']');
2337 char tmpbuf[sizeof PL_tokenbuf * 4];
2339 if (!send) /* has to be an expression */
2342 Zero(seen,256,char);
2345 else if (isDIGIT(*s)) {
2347 if (isDIGIT(s[1]) && s[2] == ']')
2353 for (; s < send; s++) {
2354 last_un_char = un_char;
2355 un_char = (unsigned char)*s;
2360 weight -= seen[un_char] * 10;
2361 if (isALNUM_lazy_if(s+1,UTF)) {
2363 scan_ident(s, send, tmpbuf, sizeof tmpbuf, FALSE);
2364 len = (int)strlen(tmpbuf);
2365 if (len > 1 && gv_fetchpvn_flags(tmpbuf, len, 0, SVt_PV))
2370 else if (*s == '$' && s[1] &&
2371 strchr("[#!%*<>()-=",s[1])) {
2372 if (/*{*/ strchr("])} =",s[2]))
2381 if (strchr("wds]",s[1]))
2383 else if (seen['\''] || seen['"'])
2385 else if (strchr("rnftbxcav",s[1]))
2387 else if (isDIGIT(s[1])) {
2389 while (s[1] && isDIGIT(s[1]))
2399 if (strchr("aA01! ",last_un_char))
2401 if (strchr("zZ79~",s[1]))
2403 if (last_un_char == 255 && (isDIGIT(s[1]) || s[1] == '$'))
2404 weight -= 5; /* cope with negative subscript */
2407 if (!isALNUM(last_un_char)
2408 && !(last_un_char == '$' || last_un_char == '@'
2409 || last_un_char == '&')
2410 && isALPHA(*s) && s[1] && isALPHA(s[1])) {
2415 if (keyword(tmpbuf, d - tmpbuf))
2418 if (un_char == last_un_char + 1)
2420 weight -= seen[un_char];
2425 if (weight >= 0) /* probably a character class */
2435 * Does all the checking to disambiguate
2437 * between foo(bar) and bar->foo. Returns 0 if not a method, otherwise
2438 * FUNCMETH (bar->foo(args)) or METHOD (bar->foo args).
2440 * First argument is the stuff after the first token, e.g. "bar".
2442 * Not a method if bar is a filehandle.
2443 * Not a method if foo is a subroutine prototyped to take a filehandle.
2444 * Not a method if it's really "Foo $bar"
2445 * Method if it's "foo $bar"
2446 * Not a method if it's really "print foo $bar"
2447 * Method if it's really "foo package::" (interpreted as package->foo)
2448 * Not a method if bar is known to be a subroutine ("sub bar; foo bar")
2449 * Not a method if bar is a filehandle or package, but is quoted with
2454 S_intuit_method(pTHX_ char *start, GV *gv, CV *cv)
2457 char *s = start + (*start == '$');
2458 char tmpbuf[sizeof PL_tokenbuf];
2466 if (SvTYPE(gv) == SVt_PVGV && GvIO(gv))
2470 const char *proto = SvPVX_const(cv);
2481 s = scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
2482 /* start is the beginning of the possible filehandle/object,
2483 * and s is the end of it
2484 * tmpbuf is a copy of it
2487 if (*start == '$') {
2488 if (gv || PL_last_lop_op == OP_PRINT || isUPPER(*PL_tokenbuf))
2491 len = start - SvPVX(PL_linestr);
2495 start = SvPVX(PL_linestr) + len;
2499 return *s == '(' ? FUNCMETH : METHOD;
2501 if (!keyword(tmpbuf, len)) {
2502 if (len > 2 && tmpbuf[len - 2] == ':' && tmpbuf[len - 1] == ':') {
2506 soff = s - SvPVX(PL_linestr);
2510 indirgv = gv_fetchpvn_flags(tmpbuf, len, 0, SVt_PVCV);
2511 if (indirgv && GvCVu(indirgv))
2513 /* filehandle or package name makes it a method */
2514 if (!gv || GvIO(indirgv) || gv_stashpvn(tmpbuf, len, FALSE)) {
2516 soff = s - SvPVX(PL_linestr);
2519 if ((PL_bufend - s) >= 2 && *s == '=' && *(s+1) == '>')
2520 return 0; /* no assumptions -- "=>" quotes bearword */
2522 start_force(PL_curforce);
2523 NEXTVAL_NEXTTOKE.opval = (OP*)newSVOP(OP_CONST, 0,
2524 newSVpvn(tmpbuf,len));
2525 NEXTVAL_NEXTTOKE.opval->op_private = OPpCONST_BARE;
2527 curmad('X', newSVpvn(start,SvPVX(PL_linestr) + soff - start));
2532 PL_bufptr = SvPVX(PL_linestr) + soff; /* restart before space */
2534 return *s == '(' ? FUNCMETH : METHOD;
2542 * Return a string of Perl code to load the debugger. If PERL5DB
2543 * is set, it will return the contents of that, otherwise a
2544 * compile-time require of perl5db.pl.
2552 const char * const pdb = PerlEnv_getenv("PERL5DB");
2556 SETERRNO(0,SS_NORMAL);
2557 return "BEGIN { require 'perl5db.pl' }";
2563 /* Encoded script support. filter_add() effectively inserts a
2564 * 'pre-processing' function into the current source input stream.
2565 * Note that the filter function only applies to the current source file
2566 * (e.g., it will not affect files 'require'd or 'use'd by this one).
2568 * The datasv parameter (which may be NULL) can be used to pass
2569 * private data to this instance of the filter. The filter function
2570 * can recover the SV using the FILTER_DATA macro and use it to
2571 * store private buffers and state information.
2573 * The supplied datasv parameter is upgraded to a PVIO type
2574 * and the IoDIRP/IoANY field is used to store the function pointer,
2575 * and IOf_FAKE_DIRP is enabled on datasv to mark this as such.
2576 * Note that IoTOP_NAME, IoFMT_NAME, IoBOTTOM_NAME, if set for
2577 * private use must be set using malloc'd pointers.
2581 Perl_filter_add(pTHX_ filter_t funcp, SV *datasv)
2587 if (!PL_rsfp_filters)
2588 PL_rsfp_filters = newAV();
2591 SvUPGRADE(datasv, SVt_PVIO);
2592 IoANY(datasv) = FPTR2DPTR(void *, funcp); /* stash funcp into spare field */
2593 IoFLAGS(datasv) |= IOf_FAKE_DIRP;
2594 DEBUG_P(PerlIO_printf(Perl_debug_log, "filter_add func %p (%s)\n",
2595 IoANY(datasv), SvPV_nolen(datasv)));
2596 av_unshift(PL_rsfp_filters, 1);
2597 av_store(PL_rsfp_filters, 0, datasv) ;
2602 /* Delete most recently added instance of this filter function. */
2604 Perl_filter_del(pTHX_ filter_t funcp)
2610 DEBUG_P(PerlIO_printf(Perl_debug_log, "filter_del func %p", FPTR2DPTR(XPVIO *, funcp)));
2612 if (!PL_rsfp_filters || AvFILLp(PL_rsfp_filters)<0)
2614 /* if filter is on top of stack (usual case) just pop it off */
2615 datasv = FILTER_DATA(AvFILLp(PL_rsfp_filters));
2616 if (IoANY(datasv) == FPTR2DPTR(void *, funcp)) {
2617 IoFLAGS(datasv) &= ~IOf_FAKE_DIRP;
2618 IoANY(datasv) = (void *)NULL;
2619 sv_free(av_pop(PL_rsfp_filters));
2623 /* we need to search for the correct entry and clear it */
2624 Perl_die(aTHX_ "filter_del can only delete in reverse order (currently)");
2628 /* Invoke the idxth filter function for the current rsfp. */
2629 /* maxlen 0 = read one text line */
2631 Perl_filter_read(pTHX_ int idx, SV *buf_sv, int maxlen)
2636 /* This API is bad. It should have been using unsigned int for maxlen.
2637 Not sure if we want to change the API, but if not we should sanity
2638 check the value here. */
2639 const unsigned int correct_length
2648 if (!PL_rsfp_filters)
2650 if (idx > AvFILLp(PL_rsfp_filters)) { /* Any more filters? */
2651 /* Provide a default input filter to make life easy. */
2652 /* Note that we append to the line. This is handy. */
2653 DEBUG_P(PerlIO_printf(Perl_debug_log,
2654 "filter_read %d: from rsfp\n", idx));
2655 if (correct_length) {
2658 const int old_len = SvCUR(buf_sv);
2660 /* ensure buf_sv is large enough */
2661 SvGROW(buf_sv, (STRLEN)(old_len + correct_length)) ;
2662 if ((len = PerlIO_read(PL_rsfp, SvPVX(buf_sv) + old_len,
2663 correct_length)) <= 0) {
2664 if (PerlIO_error(PL_rsfp))
2665 return -1; /* error */
2667 return 0 ; /* end of file */
2669 SvCUR_set(buf_sv, old_len + len) ;
2672 if (sv_gets(buf_sv, PL_rsfp, SvCUR(buf_sv)) == NULL) {
2673 if (PerlIO_error(PL_rsfp))
2674 return -1; /* error */
2676 return 0 ; /* end of file */
2679 return SvCUR(buf_sv);
2681 /* Skip this filter slot if filter has been deleted */
2682 if ( (datasv = FILTER_DATA(idx)) == &PL_sv_undef) {
2683 DEBUG_P(PerlIO_printf(Perl_debug_log,
2684 "filter_read %d: skipped (filter deleted)\n",
2686 return FILTER_READ(idx+1, buf_sv, correct_length); /* recurse */
2688 /* Get function pointer hidden within datasv */
2689 funcp = DPTR2FPTR(filter_t, IoANY(datasv));
2690 DEBUG_P(PerlIO_printf(Perl_debug_log,
2691 "filter_read %d: via function %p (%s)\n",
2692 idx, datasv, SvPV_nolen_const(datasv)));
2693 /* Call function. The function is expected to */
2694 /* call "FILTER_READ(idx+1, buf_sv)" first. */
2695 /* Return: <0:error, =0:eof, >0:not eof */
2696 return (*funcp)(aTHX_ idx, buf_sv, correct_length);
2700 S_filter_gets(pTHX_ register SV *sv, register PerlIO *fp, STRLEN append)
2703 #ifdef PERL_CR_FILTER
2704 if (!PL_rsfp_filters) {
2705 filter_add(S_cr_textfilter,NULL);
2708 if (PL_rsfp_filters) {
2710 SvCUR_set(sv, 0); /* start with empty line */
2711 if (FILTER_READ(0, sv, 0) > 0)
2712 return ( SvPVX(sv) ) ;
2717 return (sv_gets(sv, fp, append));
2721 S_find_in_my_stash(pTHX_ const char *pkgname, I32 len)
2726 if (len == 11 && *pkgname == '_' && strEQ(pkgname, "__PACKAGE__"))
2730 (pkgname[len - 2] == ':' && pkgname[len - 1] == ':') &&
2731 (gv = gv_fetchpvn_flags(pkgname, len, 0, SVt_PVHV)))
2733 return GvHV(gv); /* Foo:: */
2736 /* use constant CLASS => 'MyClass' */
2737 if ((gv = gv_fetchpvn_flags(pkgname, len, 0, SVt_PVCV))) {
2739 if (GvCV(gv) && (sv = cv_const_sv(GvCV(gv)))) {
2740 pkgname = SvPV_nolen_const(sv);
2744 return gv_stashpv(pkgname, FALSE);
2750 * The intent of this yylex wrapper is to minimize the changes to the
2751 * tokener when we aren't interested in collecting madprops. It remains
2752 * to be seen how successful this strategy will be...
2759 char *s = PL_bufptr;
2761 /* make sure PL_thiswhite is initialized */
2765 /* just do what yylex would do on pending identifier; leave PL_thiswhite alone */
2766 if (PL_pending_ident)
2767 return S_pending_ident(aTHX);
2769 /* previous token ate up our whitespace? */
2770 if (!PL_lasttoke && PL_nextwhite) {
2771 PL_thiswhite = PL_nextwhite;
2775 /* isolate the token, and figure out where it is without whitespace */
2776 PL_realtokenstart = -1;
2780 assert(PL_curforce < 0);
2782 if (!PL_thismad || PL_thismad->mad_key == '^') { /* not forced already? */
2783 if (!PL_thistoken) {
2784 if (PL_realtokenstart < 0 || !CopLINE(PL_curcop))
2785 PL_thistoken = newSVpvn("",0);
2787 char *tstart = SvPVX(PL_linestr) + PL_realtokenstart;
2788 PL_thistoken = newSVpvn(tstart, s - tstart);
2791 if (PL_thismad) /* install head */
2792 CURMAD('X', PL_thistoken);
2795 /* last whitespace of a sublex? */
2796 if (optype == ')' && PL_endwhite) {
2797 CURMAD('X', PL_endwhite);
2802 /* if no whitespace and we're at EOF, bail. Otherwise fake EOF below. */
2803 if (!PL_thiswhite && !PL_endwhite && !optype) {
2804 sv_free(PL_thistoken);
2809 /* put off final whitespace till peg */
2810 if (optype == ';' && !PL_rsfp) {
2811 PL_nextwhite = PL_thiswhite;
2814 else if (PL_thisopen) {
2815 CURMAD('q', PL_thisopen);
2817 sv_free(PL_thistoken);
2821 /* Store actual token text as madprop X */
2822 CURMAD('X', PL_thistoken);
2826 /* add preceding whitespace as madprop _ */
2827 CURMAD('_', PL_thiswhite);
2831 /* add quoted material as madprop = */
2832 CURMAD('=', PL_thisstuff);
2836 /* add terminating quote as madprop Q */
2837 CURMAD('Q', PL_thisclose);
2841 /* special processing based on optype */
2845 /* opval doesn't need a TOKEN since it can already store mp */
2856 append_madprops(PL_thismad, yylval.opval, 0);
2864 addmad(newMADsv('p', PL_endwhite), &PL_thismad, 0);
2873 /* remember any fake bracket that lexer is about to discard */
2874 if (PL_lex_brackets == 1 &&
2875 ((expectation)PL_lex_brackstack[0] & XFAKEBRACK))
2878 while (s < PL_bufend && (*s == ' ' || *s == '\t'))
2881 PL_thiswhite = newSVpvn(PL_bufptr, ++s - PL_bufptr);
2882 addmad(newMADsv('#', PL_thiswhite), &PL_thismad, 0);
2885 break; /* don't bother looking for trailing comment */
2894 /* attach a trailing comment to its statement instead of next token */
2898 if (PL_bufptr > PL_oldbufptr && PL_bufptr[-1] == optype) {
2900 while (s < PL_bufend && (*s == ' ' || *s == '\t'))
2902 if (*s == '\n' || *s == '#') {
2903 while (s < PL_bufend && *s != '\n')
2907 PL_thiswhite = newSVpvn(PL_bufptr, s - PL_bufptr);
2908 addmad(newMADsv('#', PL_thiswhite), &PL_thismad, 0);
2925 /* Create new token struct. Note: opvals return early above. */
2926 yylval.tkval = newTOKEN(optype, yylval, PL_thismad);
2933 S_tokenize_use(pTHX_ int is_use, char *s) {
2935 if (PL_expect != XSTATE)
2936 yyerror(Perl_form(aTHX_ "\"%s\" not allowed in expression",
2937 is_use ? "use" : "no"));
2939 if (isDIGIT(*s) || (*s == 'v' && isDIGIT(s[1]))) {
2940 s = force_version(s, TRUE);
2941 if (*s == ';' || (s = SKIPSPACE1(s), *s == ';')) {
2942 start_force(PL_curforce);
2943 NEXTVAL_NEXTTOKE.opval = NULL;
2946 else if (*s == 'v') {
2947 s = force_word(s,WORD,FALSE,TRUE,FALSE);
2948 s = force_version(s, FALSE);
2952 s = force_word(s,WORD,FALSE,TRUE,FALSE);
2953 s = force_version(s, FALSE);
2955 yylval.ival = is_use;
2959 static const char* const exp_name[] =
2960 { "OPERATOR", "TERM", "REF", "STATE", "BLOCK", "ATTRBLOCK",
2961 "ATTRTERM", "TERMBLOCK", "TERMORDORDOR"
2968 Works out what to call the token just pulled out of the input
2969 stream. The yacc parser takes care of taking the ops we return and
2970 stitching them into a tree.
2976 if read an identifier
2977 if we're in a my declaration
2978 croak if they tried to say my($foo::bar)
2979 build the ops for a my() declaration
2980 if it's an access to a my() variable
2981 are we in a sort block?
2982 croak if my($a); $a <=> $b
2983 build ops for access to a my() variable
2984 if in a dq string, and they've said @foo and we can't find @foo
2986 build ops for a bareword
2987 if we already built the token before, use it.
2992 #pragma segment Perl_yylex
2998 register char *s = PL_bufptr;
3004 SV* tmp = newSVpvs("");
3005 PerlIO_printf(Perl_debug_log, "### %"IVdf":LEX_%s/X%s %s\n",
3006 (IV)CopLINE(PL_curcop),
3007 lex_state_names[PL_lex_state],
3008 exp_name[PL_expect],
3009 pv_display(tmp, s, strlen(s), 0, 60));
3012 /* check if there's an identifier for us to look at */
3013 if (PL_pending_ident)
3014 return REPORT(S_pending_ident(aTHX));
3016 /* no identifier pending identification */
3018 switch (PL_lex_state) {
3020 case LEX_NORMAL: /* Some compilers will produce faster */
3021 case LEX_INTERPNORMAL: /* code if we comment these out. */
3025 /* when we've already built the next token, just pull it out of the queue */
3029 yylval = PL_nexttoke[PL_lasttoke].next_val;
3031 PL_thismad = PL_nexttoke[PL_lasttoke].next_mad;
3032 PL_nexttoke[PL_lasttoke].next_mad = 0;
3033 if (PL_thismad && PL_thismad->mad_key == '_') {
3034 PL_thiswhite = (SV*)PL_thismad->mad_val;
3035 PL_thismad->mad_val = 0;
3036 mad_free(PL_thismad);
3041 PL_lex_state = PL_lex_defer;
3042 PL_expect = PL_lex_expect;
3043 PL_lex_defer = LEX_NORMAL;
3044 if (!PL_nexttoke[PL_lasttoke].next_type)
3049 yylval = PL_nextval[PL_nexttoke];
3051 PL_lex_state = PL_lex_defer;
3052 PL_expect = PL_lex_expect;
3053 PL_lex_defer = LEX_NORMAL;
3057 /* FIXME - can these be merged? */
3058 return(PL_nexttoke[PL_lasttoke].next_type);
3060 return REPORT(PL_nexttype[PL_nexttoke]);
3063 /* interpolated case modifiers like \L \U, including \Q and \E.
3064 when we get here, PL_bufptr is at the \
3066 case LEX_INTERPCASEMOD:
3068 if (PL_bufptr != PL_bufend && *PL_bufptr != '\\')
3069 Perl_croak(aTHX_ "panic: INTERPCASEMOD");
3071 /* handle \E or end of string */
3072 if (PL_bufptr == PL_bufend || PL_bufptr[1] == 'E') {
3074 if (PL_lex_casemods) {
3075 const char oldmod = PL_lex_casestack[--PL_lex_casemods];
3076 PL_lex_casestack[PL_lex_casemods] = '\0';
3078 if (PL_bufptr != PL_bufend
3079 && (oldmod == 'L' || oldmod == 'U' || oldmod == 'Q')) {
3081 PL_lex_state = LEX_INTERPCONCAT;
3084 PL_thistoken = newSVpvn("\\E",2);
3090 while (PL_bufptr != PL_bufend &&
3091 PL_bufptr[0] == '\\' && PL_bufptr[1] == 'E') {
3093 PL_thiswhite = newSVpvn("",0);
3094 sv_catpvn(PL_thiswhite, PL_bufptr, 2);
3098 if (PL_bufptr != PL_bufend)
3101 PL_lex_state = LEX_INTERPCONCAT;
3105 DEBUG_T({ PerlIO_printf(Perl_debug_log,
3106 "### Saw case modifier\n"); });
3108 if (s[1] == '\\' && s[2] == 'E') {
3111 PL_thiswhite = newSVpvn("",0);
3112 sv_catpvn(PL_thiswhite, PL_bufptr, 4);
3115 PL_lex_state = LEX_INTERPCONCAT;
3120 if (!PL_madskills) /* when just compiling don't need correct */
3121 if (strnEQ(s, "L\\u", 3) || strnEQ(s, "U\\l", 3))
3122 tmp = *s, *s = s[2], s[2] = (char)tmp; /* misordered... */
3123 if ((*s == 'L' || *s == 'U') &&
3124 (strchr(PL_lex_casestack, 'L') || strchr(PL_lex_casestack, 'U'))) {
3125 PL_lex_casestack[--PL_lex_casemods] = '\0';
3128 if (PL_lex_casemods > 10)
3129 Renew(PL_lex_casestack, PL_lex_casemods + 2, char);
3130 PL_lex_casestack[PL_lex_casemods++] = *s;
3131 PL_lex_casestack[PL_lex_casemods] = '\0';
3132 PL_lex_state = LEX_INTERPCONCAT;
3133 start_force(PL_curforce);
3134 NEXTVAL_NEXTTOKE.ival = 0;
3136 start_force(PL_curforce);
3138 NEXTVAL_NEXTTOKE.ival = OP_LCFIRST;
3140 NEXTVAL_NEXTTOKE.ival = OP_UCFIRST;
3142 NEXTVAL_NEXTTOKE.ival = OP_LC;
3144 NEXTVAL_NEXTTOKE.ival = OP_UC;
3146 NEXTVAL_NEXTTOKE.ival = OP_QUOTEMETA;
3148 Perl_croak(aTHX_ "panic: yylex");
3150 SV* const tmpsv = newSVpvn("",0);
3151 Perl_sv_catpvf(aTHX_ tmpsv, "\\%c", *s);
3157 if (PL_lex_starts) {
3163 sv_free(PL_thistoken);
3164 PL_thistoken = newSVpvn("",0);
3167 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
3168 if (PL_lex_casemods == 1 && PL_lex_inpat)
3177 case LEX_INTERPPUSH:
3178 return REPORT(sublex_push());
3180 case LEX_INTERPSTART:
3181 if (PL_bufptr == PL_bufend)
3182 return REPORT(sublex_done());
3183 DEBUG_T({ PerlIO_printf(Perl_debug_log,
3184 "### Interpolated variable\n"); });
3186 PL_lex_dojoin = (*PL_bufptr == '@');
3187 PL_lex_state = LEX_INTERPNORMAL;
3188 if (PL_lex_dojoin) {
3189 start_force(PL_curforce);
3190 NEXTVAL_NEXTTOKE.ival = 0;
3192 start_force(PL_curforce);
3193 force_ident("\"", '$');
3194 start_force(PL_curforce);
3195 NEXTVAL_NEXTTOKE.ival = 0;
3197 start_force(PL_curforce);
3198 NEXTVAL_NEXTTOKE.ival = 0;
3200 start_force(PL_curforce);
3201 NEXTVAL_NEXTTOKE.ival = OP_JOIN; /* emulate join($", ...) */
3204 if (PL_lex_starts++) {
3209 sv_free(PL_thistoken);
3210 PL_thistoken = newSVpvn("",0);
3213 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
3214 if (!PL_lex_casemods && PL_lex_inpat)
3221 case LEX_INTERPENDMAYBE:
3222 if (intuit_more(PL_bufptr)) {
3223 PL_lex_state = LEX_INTERPNORMAL; /* false alarm, more expr */
3229 if (PL_lex_dojoin) {
3230 PL_lex_dojoin = FALSE;
3231 PL_lex_state = LEX_INTERPCONCAT;
3235 sv_free(PL_thistoken);
3236 PL_thistoken = newSVpvn("",0);
3241 if (PL_lex_inwhat == OP_SUBST && PL_linestr == PL_lex_repl
3242 && SvEVALED(PL_lex_repl))
3244 if (PL_bufptr != PL_bufend)
3245 Perl_croak(aTHX_ "Bad evalled substitution pattern");
3249 case LEX_INTERPCONCAT:
3251 if (PL_lex_brackets)
3252 Perl_croak(aTHX_ "panic: INTERPCONCAT");
3254 if (PL_bufptr == PL_bufend)
3255 return REPORT(sublex_done());
3257 if (SvIVX(PL_linestr) == '\'') {
3258 SV *sv = newSVsv(PL_linestr);
3261 else if ( PL_hints & HINT_NEW_RE )
3262 sv = new_constant(NULL, 0, "qr", sv, sv, "q");
3263 yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
3267 s = scan_const(PL_bufptr);
3269 PL_lex_state = LEX_INTERPCASEMOD;
3271 PL_lex_state = LEX_INTERPSTART;
3274 if (s != PL_bufptr) {
3275 start_force(PL_curforce);
3277 curmad('X', newSVpvn(PL_bufptr,s-PL_bufptr));
3279 NEXTVAL_NEXTTOKE = yylval;
3282 if (PL_lex_starts++) {
3286 sv_free(PL_thistoken);
3287 PL_thistoken = newSVpvn("",0);
3290 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
3291 if (!PL_lex_casemods && PL_lex_inpat)
3304 PL_lex_state = LEX_NORMAL;
3305 s = scan_formline(PL_bufptr);
3306 if (!PL_lex_formbrack)
3312 PL_oldoldbufptr = PL_oldbufptr;
3318 sv_free(PL_thistoken);
3321 PL_realtokenstart = s - SvPVX(PL_linestr); /* assume but undo on ws */
3325 if (isIDFIRST_lazy_if(s,UTF))
3327 Perl_croak(aTHX_ "Unrecognized character \\x%02X", *s & 255);
3330 goto fake_eof; /* emulate EOF on ^D or ^Z */
3339 if (PL_lex_brackets) {
3340 yyerror(PL_lex_formbrack
3341 ? "Format not terminated"
3342 : "Missing right curly or square bracket");
3344 DEBUG_T( { PerlIO_printf(Perl_debug_log,
3345 "### Tokener got EOF\n");
3349 if (s++ < PL_bufend)
3350 goto retry; /* ignore stray nulls */
3353 if (!PL_in_eval && !PL_preambled) {
3354 PL_preambled = TRUE;
3359 sv_setpv(PL_linestr,incl_perldb());
3360 if (SvCUR(PL_linestr))
3361 sv_catpvs(PL_linestr,";");
3363 while(AvFILLp(PL_preambleav) >= 0) {
3364 SV *tmpsv = av_shift(PL_preambleav);
3365 sv_catsv(PL_linestr, tmpsv);
3366 sv_catpvs(PL_linestr, ";");
3369 sv_free((SV*)PL_preambleav);
3370 PL_preambleav = NULL;
3372 if (PL_minus_n || PL_minus_p) {
3373 sv_catpvs(PL_linestr, "LINE: while (<>) {");
3375 sv_catpvs(PL_linestr,"chomp;");
3378 if ((*PL_splitstr == '/' || *PL_splitstr == '\''
3379 || *PL_splitstr == '"')
3380 && strchr(PL_splitstr + 1, *PL_splitstr))
3381 Perl_sv_catpvf(aTHX_ PL_linestr, "our @F=split(%s);", PL_splitstr);
3383 /* "q\0${splitstr}\0" is legal perl. Yes, even NUL
3384 bytes can be used as quoting characters. :-) */
3385 const char *splits = PL_splitstr;
3386 sv_catpvs(PL_linestr, "our @F=split(q\0");
3389 if (*splits == '\\')
3390 sv_catpvn(PL_linestr, splits, 1);
3391 sv_catpvn(PL_linestr, splits, 1);
3392 } while (*splits++);
3393 /* This loop will embed the trailing NUL of
3394 PL_linestr as the last thing it does before
3396 sv_catpvs(PL_linestr, ");");
3400 sv_catpvs(PL_linestr,"our @F=split(' ');");
3404 sv_catpvs(PL_linestr,"use feature ':5.10';");
3405 sv_catpvs(PL_linestr, "\n");
3406 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3407 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3408 PL_last_lop = PL_last_uni = NULL;
3409 if (PERLDB_LINE && PL_curstash != PL_debstash) {
3410 SV * const sv = newSV(0);
3412 sv_upgrade(sv, SVt_PVMG);
3413 sv_setsv(sv,PL_linestr);
3416 av_store(CopFILEAVx(PL_curcop),(I32)CopLINE(PL_curcop),sv);
3421 bof = PL_rsfp ? TRUE : FALSE;
3422 if ((s = filter_gets(PL_linestr, PL_rsfp, 0)) == NULL) {
3425 PL_realtokenstart = -1;
3428 if (PL_preprocess && !PL_in_eval)
3429 (void)PerlProc_pclose(PL_rsfp);
3430 else if ((PerlIO *)PL_rsfp == PerlIO_stdin())
3431 PerlIO_clearerr(PL_rsfp);
3433 (void)PerlIO_close(PL_rsfp);
3435 PL_doextract = FALSE;
3437 if (!PL_in_eval && (PL_minus_n || PL_minus_p)) {
3442 sv_setpv(PL_linestr,PL_minus_p
3443 ? ";}continue{print;}" : ";}");
3444 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3445 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3446 PL_last_lop = PL_last_uni = NULL;
3447 PL_minus_n = PL_minus_p = 0;
3450 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3451 PL_last_lop = PL_last_uni = NULL;
3452 sv_setpvn(PL_linestr,"",0);
3453 TOKEN(';'); /* not infinite loop because rsfp is NULL now */
3455 /* If it looks like the start of a BOM or raw UTF-16,
3456 * check if it in fact is. */
3462 #ifdef PERLIO_IS_STDIO
3463 # ifdef __GNU_LIBRARY__
3464 # if __GNU_LIBRARY__ == 1 /* Linux glibc5 */
3465 # define FTELL_FOR_PIPE_IS_BROKEN
3469 # if __GLIBC__ == 1 /* maybe some glibc5 release had it like this? */
3470 # define FTELL_FOR_PIPE_IS_BROKEN
3475 #ifdef FTELL_FOR_PIPE_IS_BROKEN
3476 /* This loses the possibility to detect the bof
3477 * situation on perl -P when the libc5 is being used.
3478 * Workaround? Maybe attach some extra state to PL_rsfp?
3481 bof = PerlIO_tell(PL_rsfp) == SvCUR(PL_linestr);
3483 bof = PerlIO_tell(PL_rsfp) == (Off_t)SvCUR(PL_linestr);
3486 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3487 s = swallow_bom((U8*)s);
3491 /* Incest with pod. */
3494 sv_catsv(PL_thiswhite, PL_linestr);
3496 if (*s == '=' && strnEQ(s, "=cut", 4)) {
3497 sv_setpvn(PL_linestr, "", 0);
3498 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3499 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3500 PL_last_lop = PL_last_uni = NULL;
3501 PL_doextract = FALSE;
3505 } while (PL_doextract);
3506 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = s;
3507 if (PERLDB_LINE && PL_curstash != PL_debstash) {
3508 SV * const sv = newSV(0);
3510 sv_upgrade(sv, SVt_PVMG);
3511 sv_setsv(sv,PL_linestr);
3514 av_store(CopFILEAVx(PL_curcop),(I32)CopLINE(PL_curcop),sv);
3516 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3517 PL_last_lop = PL_last_uni = NULL;
3518 if (CopLINE(PL_curcop) == 1) {
3519 while (s < PL_bufend && isSPACE(*s))
3521 if (*s == ':' && s[1] != ':') /* for csh execing sh scripts */
3525 PL_thiswhite = newSVpvn(PL_linestart, s - PL_linestart);
3529 if (*s == '#' && *(s+1) == '!')
3531 #ifdef ALTERNATE_SHEBANG
3533 static char const as[] = ALTERNATE_SHEBANG;
3534 if (*s == as[0] && strnEQ(s, as, sizeof(as) - 1))
3535 d = s + (sizeof(as) - 1);
3537 #endif /* ALTERNATE_SHEBANG */
3546 while (*d && !isSPACE(*d))
3550 #ifdef ARG_ZERO_IS_SCRIPT
3551 if (ipathend > ipath) {
3553 * HP-UX (at least) sets argv[0] to the script name,
3554 * which makes $^X incorrect. And Digital UNIX and Linux,
3555 * at least, set argv[0] to the basename of the Perl
3556 * interpreter. So, having found "#!", we'll set it right.
3558 SV * const x = GvSV(gv_fetchpvs("\030", GV_ADD|GV_NOTQUAL,
3560 assert(SvPOK(x) || SvGMAGICAL(x));
3561 if (sv_eq(x, CopFILESV(PL_curcop))) {
3562 sv_setpvn(x, ipath, ipathend - ipath);
3568 const char *bstart = SvPV_const(CopFILESV(PL_curcop),blen);
3569 const char * const lstart = SvPV_const(x,llen);
3571 bstart += blen - llen;
3572 if (strnEQ(bstart, lstart, llen) && bstart[-1] == '/') {
3573 sv_setpvn(x, ipath, ipathend - ipath);
3578 TAINT_NOT; /* $^X is always tainted, but that's OK */
3580 #endif /* ARG_ZERO_IS_SCRIPT */
3585 d = instr(s,"perl -");
3587 d = instr(s,"perl");
3589 /* avoid getting into infinite loops when shebang
3590 * line contains "Perl" rather than "perl" */
3592 for (d = ipathend-4; d >= ipath; --d) {
3593 if ((*d == 'p' || *d == 'P')
3594 && !ibcmp(d, "perl", 4))
3604 #ifdef ALTERNATE_SHEBANG
3606 * If the ALTERNATE_SHEBANG on this system starts with a
3607 * character that can be part of a Perl expression, then if
3608 * we see it but not "perl", we're probably looking at the
3609 * start of Perl code, not a request to hand off to some
3610 * other interpreter. Similarly, if "perl" is there, but
3611 * not in the first 'word' of the line, we assume the line
3612 * contains the start of the Perl program.
3614 if (d && *s != '#') {
3615 const char *c = ipath;
3616 while (*c && !strchr("; \t\r\n\f\v#", *c))
3619 d = NULL; /* "perl" not in first word; ignore */
3621 *s = '#'; /* Don't try to parse shebang line */
3623 #endif /* ALTERNATE_SHEBANG */
3624 #ifndef MACOS_TRADITIONAL
3629 !instr(s,"indir") &&
3630 instr(PL_origargv[0],"perl"))
3637 while (s < PL_bufend && isSPACE(*s))
3639 if (s < PL_bufend) {
3640 Newxz(newargv,PL_origargc+3,char*);
3642 while (s < PL_bufend && !isSPACE(*s))
3645 Copy(PL_origargv+1, newargv+2, PL_origargc+1, char*);
3648 newargv = PL_origargv;
3651 PerlProc_execv(ipath, EXEC_ARGV_CAST(newargv));
3653 Perl_croak(aTHX_ "Can't exec %s", ipath);
3657 while (*d && !isSPACE(*d)) d++;
3658 while (SPACE_OR_TAB(*d)) d++;
3661 const bool switches_done = PL_doswitches;
3662 const U32 oldpdb = PL_perldb;
3663 const bool oldn = PL_minus_n;
3664 const bool oldp = PL_minus_p;
3667 if (*d == 'M' || *d == 'm' || *d == 'C') {
3668 const char * const m = d;
3669 while (*d && !isSPACE(*d))
3671 Perl_croak(aTHX_ "Too late for \"-%.*s\" option",
3674 d = moreswitches(d);
3676 if (PL_doswitches && !switches_done) {
3677 int argc = PL_origargc;
3678 char **argv = PL_origargv;
3681 } while (argc && argv[0][0] == '-' && argv[0][1]);
3682 init_argv_symbols(argc,argv);
3684 if ((PERLDB_LINE && !oldpdb) ||
3685 ((PL_minus_n || PL_minus_p) && !(oldn || oldp)))
3686 /* if we have already added "LINE: while (<>) {",
3687 we must not do it again */
3689 sv_setpvn(PL_linestr, "", 0);
3690 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3691 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3692 PL_last_lop = PL_last_uni = NULL;
3693 PL_preambled = FALSE;
3695 (void)gv_fetchfile(PL_origfilename);
3702 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
3704 PL_lex_state = LEX_FORMLINE;
3709 #ifdef PERL_STRICT_CR
3710 Perl_warn(aTHX_ "Illegal character \\%03o (carriage return)", '\r');
3712 "\t(Maybe you didn't strip carriage returns after a network transfer?)\n");
3714 case ' ': case '\t': case '\f': case 013:
3715 #ifdef MACOS_TRADITIONAL
3719 PL_realtokenstart = -1;
3728 PL_realtokenstart = -1;
3732 if (PL_lex_state != LEX_NORMAL || (PL_in_eval && !PL_rsfp)) {
3733 if (*s == '#' && s == PL_linestart && PL_in_eval && !PL_rsfp) {
3734 /* handle eval qq[#line 1 "foo"\n ...] */
3735 CopLINE_dec(PL_curcop);
3738 if (PL_madskills && !PL_lex_formbrack && !PL_in_eval) {
3740 if (!PL_in_eval || PL_rsfp)
3745 while (d < PL_bufend && *d != '\n')
3749 else if (d > PL_bufend) /* Found by Ilya: feed random input to Perl. */
3750 Perl_croak(aTHX_ "panic: input overflow");
3753 PL_thiswhite = newSVpvn(s, d - s);
3758 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
3760 PL_lex_state = LEX_FORMLINE;
3766 if (PL_madskills && CopLINE(PL_curcop) >= 1 && !PL_lex_formbrack) {
3767 if (CopLINE(PL_curcop) == 1 && s[0] == '#' && s[1] == '!') {
3770 TOKEN(PEG); /* make sure any #! line is accessible */
3775 /* if (PL_madskills && PL_lex_formbrack) { */
3777 while (d < PL_bufend && *d != '\n')
3781 else if (d > PL_bufend) /* Found by Ilya: feed random input to Perl. */
3782 Perl_croak(aTHX_ "panic: input overflow");
3783 if (PL_madskills && CopLINE(PL_curcop) >= 1) {
3785 PL_thiswhite = newSVpvn("",0);
3786 if (CopLINE(PL_curcop) == 1) {
3787 sv_setpvn(PL_thiswhite, "", 0);
3790 sv_catpvn(PL_thiswhite, s, d - s);
3804 if (s[1] && isALPHA(s[1]) && !isALNUM(s[2])) {
3812 while (s < PL_bufend && SPACE_OR_TAB(*s))
3815 if (strnEQ(s,"=>",2)) {
3816 s = force_word(PL_bufptr,WORD,FALSE,FALSE,FALSE);
3817 DEBUG_T( { S_printbuf(aTHX_
3818 "### Saw unary minus before =>, forcing word %s\n", s);
3820 OPERATOR('-'); /* unary minus */
3822 PL_last_uni = PL_oldbufptr;
3824 case 'r': ftst = OP_FTEREAD; break;
3825 case 'w': ftst = OP_FTEWRITE; break;
3826 case 'x': ftst = OP_FTEEXEC; break;
3827 case 'o': ftst = OP_FTEOWNED; break;
3828 case 'R': ftst = OP_FTRREAD; break;
3829 case 'W': ftst = OP_FTRWRITE; break;
3830 case 'X': ftst = OP_FTREXEC; break;
3831 case 'O': ftst = OP_FTROWNED; break;
3832 case 'e': ftst = OP_FTIS; break;
3833 case 'z': ftst = OP_FTZERO; break;
3834 case 's': ftst = OP_FTSIZE; break;
3835 case 'f': ftst = OP_FTFILE; break;
3836 case 'd': ftst = OP_FTDIR; break;
3837 case 'l': ftst = OP_FTLINK; break;
3838 case 'p': ftst = OP_FTPIPE; break;
3839 case 'S': ftst = OP_FTSOCK; break;
3840 case 'u': ftst = OP_FTSUID; break;
3841 case 'g': ftst = OP_FTSGID; break;
3842 case 'k': ftst = OP_FTSVTX; break;
3843 case 'b': ftst = OP_FTBLK; break;
3844 case 'c': ftst = OP_FTCHR; break;
3845 case 't': ftst = OP_FTTTY; break;
3846 case 'T': ftst = OP_FTTEXT; break;
3847 case 'B': ftst = OP_FTBINARY; break;
3848 case 'M': case 'A': case 'C':
3849 gv_fetchpvs("\024", GV_ADD|GV_NOTQUAL, SVt_PV);
3851 case 'M': ftst = OP_FTMTIME; break;
3852 case 'A': ftst = OP_FTATIME; break;
3853 case 'C': ftst = OP_FTCTIME; break;
3861 PL_last_lop_op = (OPCODE)ftst;
3862 DEBUG_T( { PerlIO_printf(Perl_debug_log,
3863 "### Saw file test %c\n", (int)tmp);
3868 /* Assume it was a minus followed by a one-letter named
3869 * subroutine call (or a -bareword), then. */
3870 DEBUG_T( { PerlIO_printf(Perl_debug_log,
3871 "### '-%c' looked like a file test but was not\n",
3878 const char tmp = *s++;
3881 if (PL_expect == XOPERATOR)
3886 else if (*s == '>') {
3889 if (isIDFIRST_lazy_if(s,UTF)) {
3890 s = force_word(s,METHOD,FALSE,TRUE,FALSE);
3898 if (PL_expect == XOPERATOR)
3901 if (isSPACE(*s) || !isSPACE(*PL_bufptr))
3903 OPERATOR('-'); /* unary minus */
3909 const char tmp = *s++;
3912 if (PL_expect == XOPERATOR)
3917 if (PL_expect == XOPERATOR)
3920 if (isSPACE(*s) || !isSPACE(*PL_bufptr))
3927 if (PL_expect != XOPERATOR) {
3928 s = scan_ident(s, PL_bufend, PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
3929 PL_expect = XOPERATOR;
3930 force_ident(PL_tokenbuf, '*');
3943 if (PL_expect == XOPERATOR) {
3947 PL_tokenbuf[0] = '%';
3948 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1, TRUE);
3949 if (!PL_tokenbuf[1]) {
3952 PL_pending_ident = '%';
3963 && (PL_expect == XOPERATOR || PL_expect == XTERMORDORDOR)
3964 && FEATURE_IS_ENABLED("~~"))
3971 const char tmp = *s++;
3977 goto just_a_word_zero_gv;
3980 switch (PL_expect) {
3986 if (!PL_in_my || PL_lex_state != LEX_NORMAL)
3988 PL_bufptr = s; /* update in case we back off */
3994 PL_expect = XTERMBLOCK;
3997 stuffstart = s - SvPVX(PL_linestr) - 1;
4001 while (isIDFIRST_lazy_if(s,UTF)) {
4004 d = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
4005 if (isLOWER(*s) && (tmp = keyword(PL_tokenbuf, len))) {
4006 if (tmp < 0) tmp = -tmp;
4021 sv = newSVpvn(s, len);
4023 d = scan_str(d,TRUE,TRUE);
4025 /* MUST advance bufptr here to avoid bogus
4026 "at end of line" context messages from yyerror().
4028 PL_bufptr = s + len;
4029 yyerror("Unterminated attribute parameter in attribute list");
4033 return REPORT(0); /* EOF indicator */
4037 sv_catsv(sv, PL_lex_stuff);
4038 attrs = append_elem(OP_LIST, attrs,
4039 newSVOP(OP_CONST, 0, sv));
4040 SvREFCNT_dec(PL_lex_stuff);
4041 PL_lex_stuff = NULL;
4044 if (len == 6 && strnEQ(SvPVX(sv), "unique", len)) {
4046 if (PL_in_my == KEY_our) {
4048 GvUNIQUE_on(cGVOPx_gv(yylval.opval));
4050 /* skip to avoid loading attributes.pm */
4052 deprecate(":unique");
4055 Perl_croak(aTHX_ "The 'unique' attribute may only be applied to 'our' variables");
4058 /* NOTE: any CV attrs applied here need to be part of
4059 the CVf_BUILTIN_ATTRS define in cv.h! */
4060 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "lvalue", len)) {
4062 CvLVALUE_on(PL_compcv);
4064 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "locked", len)) {
4066 CvLOCKED_on(PL_compcv);
4068 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "method", len)) {
4070 CvMETHOD_on(PL_compcv);
4072 else if (!PL_in_my && len == 9 && strnEQ(SvPVX(sv), "assertion", len)) {
4074 CvASSERTION_on(PL_compcv);
4076 /* After we've set the flags, it could be argued that
4077 we don't need to do the attributes.pm-based setting
4078 process, and shouldn't bother appending recognized
4079 flags. To experiment with that, uncomment the
4080 following "else". (Note that's already been
4081 uncommented. That keeps the above-applied built-in
4082 attributes from being intercepted (and possibly
4083 rejected) by a package's attribute routines, but is
4084 justified by the performance win for the common case
4085 of applying only built-in attributes.) */
4087 attrs = append_elem(OP_LIST, attrs,
4088 newSVOP(OP_CONST, 0,
4092 if (*s == ':' && s[1] != ':')
4095 break; /* require real whitespace or :'s */
4096 /* XXX losing whitespace on sequential attributes here */
4100 = (PL_expect == XOPERATOR ? '=' : '{'); /*'}(' for vi */
4101 if (*s != ';' && *s != '}' && *s != tmp
4102 && (tmp != '=' || *s != ')')) {
4103 const char q = ((*s == '\'') ? '"' : '\'');
4104 /* If here for an expression, and parsed no attrs, back
4106 if (tmp == '=' && !attrs) {
4110 /* MUST advance bufptr here to avoid bogus "at end of line"
4111 context messages from yyerror().
4115 ? Perl_form(aTHX_ "Invalid separator character "
4116 "%c%c%c in attribute list", q, *s, q)
4117 : "Unterminated attribute list" );
4125 start_force(PL_curforce);
4126 NEXTVAL_NEXTTOKE.opval = attrs;
4127 CURMAD('_', PL_nextwhite);
4132 PL_thistoken = newSVpvn(SvPVX(PL_linestr) + stuffstart,
4133 (s - SvPVX(PL_linestr)) - stuffstart);
4141 if (PL_last_lop == PL_oldoldbufptr || PL_last_uni == PL_oldoldbufptr)
4142 PL_oldbufptr = PL_oldoldbufptr; /* allow print(STDOUT 123) */
4150 const char tmp = *s++;
4155 const char tmp = *s++;
4163 if (PL_lex_brackets <= 0)
4164 yyerror("Unmatched right square bracket");
4167 if (PL_lex_state == LEX_INTERPNORMAL) {
4168 if (PL_lex_brackets == 0) {
4169 if (*s != '[' && *s != '{' && (*s != '-' || s[1] != '>'))
4170 PL_lex_state = LEX_INTERPEND;
4177 if (PL_lex_brackets > 100) {
4178 Renew(PL_lex_brackstack, PL_lex_brackets + 10, char);
4180 switch (PL_expect) {
4182 if (PL_lex_formbrack) {
4186 if (PL_oldoldbufptr == PL_last_lop)
4187 PL_lex_brackstack[PL_lex_brackets++] = XTERM;
4189 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
4190 OPERATOR(HASHBRACK);
4192 while (s < PL_bufend && SPACE_OR_TAB(*s))
4195 PL_tokenbuf[0] = '\0';
4196 if (d < PL_bufend && *d == '-') {
4197 PL_tokenbuf[0] = '-';
4199 while (d < PL_bufend && SPACE_OR_TAB(*d))
4202 if (d < PL_bufend && isIDFIRST_lazy_if(d,UTF)) {
4203 d = scan_word(d, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1,
4205 while (d < PL_bufend && SPACE_OR_TAB(*d))
4208 const char minus = (PL_tokenbuf[0] == '-');
4209 s = force_word(s + minus, WORD, FALSE, TRUE, FALSE);
4217 PL_lex_brackstack[PL_lex_brackets++] = XSTATE;
4222 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
4227 if (PL_oldoldbufptr == PL_last_lop)
4228 PL_lex_brackstack[PL_lex_brackets++] = XTERM;
4230 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
4233 if (PL_expect == XREF && PL_lex_state == LEX_INTERPNORMAL) {
4235 /* This hack is to get the ${} in the message. */
4237 yyerror("syntax error");
4240 OPERATOR(HASHBRACK);
4242 /* This hack serves to disambiguate a pair of curlies
4243 * as being a block or an anon hash. Normally, expectation
4244 * determines that, but in cases where we're not in a
4245 * position to expect anything in particular (like inside
4246 * eval"") we have to resolve the ambiguity. This code
4247 * covers the case where the first term in the curlies is a
4248 * quoted string. Most other cases need to be explicitly
4249 * disambiguated by prepending a "+" before the opening
4250 * curly in order to force resolution as an anon hash.
4252 * XXX should probably propagate the outer expectation
4253 * into eval"" to rely less on this hack, but that could
4254 * potentially break current behavior of eval"".
4258 if (*s == '\'' || *s == '"' || *s == '`') {
4259 /* common case: get past first string, handling escapes */
4260 for (t++; t < PL_bufend && *t != *s;)
4261 if (*t++ == '\\' && (*t == '\\' || *t == *s))
4265 else if (*s == 'q') {
4268 || ((*t == 'q' || *t == 'x') && ++t < PL_bufend
4271 /* skip q//-like construct */
4273 char open, close, term;
4276 while (t < PL_bufend && isSPACE(*t))
4278 /* check for q => */
4279 if (t+1 < PL_bufend && t[0] == '=' && t[1] == '>') {
4280 OPERATOR(HASHBRACK);
4284 if (term && (tmps = strchr("([{< )]}> )]}>",term)))
4288 for (t++; t < PL_bufend; t++) {
4289 if (*t == '\\' && t+1 < PL_bufend && open != '\\')
4291 else if (*t == open)
4295 for (t++; t < PL_bufend; t++) {
4296 if (*t == '\\' && t+1 < PL_bufend)
4298 else if (*t == close && --brackets <= 0)
4300 else if (*t == open)
4307 /* skip plain q word */
4308 while (t < PL_bufend && isALNUM_lazy_if(t,UTF))
4311 else if (isALNUM_lazy_if(t,UTF)) {
4313 while (t < PL_bufend && isALNUM_lazy_if(t,UTF))
4316 while (t < PL_bufend && isSPACE(*t))
4318 /* if comma follows first term, call it an anon hash */
4319 /* XXX it could be a comma expression with loop modifiers */
4320 if (t < PL_bufend && ((*t == ',' && (*s == 'q' || !isLOWER(*s)))
4321 || (*t == '=' && t[1] == '>')))
4322 OPERATOR(HASHBRACK);
4323 if (PL_expect == XREF)
4326 PL_lex_brackstack[PL_lex_brackets-1] = XSTATE;
4332 yylval.ival = CopLINE(PL_curcop);
4333 if (isSPACE(*s) || *s == '#')
4334 PL_copline = NOLINE; /* invalidate current command line number */
4339 if (PL_lex_brackets <= 0)
4340 yyerror("Unmatched right curly bracket");
4342 PL_expect = (expectation)PL_lex_brackstack[--PL_lex_brackets];
4343 if (PL_lex_brackets < PL_lex_formbrack && PL_lex_state != LEX_INTERPNORMAL)
4344 PL_lex_formbrack = 0;
4345 if (PL_lex_state == LEX_INTERPNORMAL) {
4346 if (PL_lex_brackets == 0) {
4347 if (PL_expect & XFAKEBRACK) {
4348 PL_expect &= XENUMMASK;
4349 PL_lex_state = LEX_INTERPEND;
4354 PL_thiswhite = newSVpvn("",0);
4355 sv_catpvn(PL_thiswhite,"}",1);
4358 return yylex(); /* ignore fake brackets */
4360 if (*s == '-' && s[1] == '>')
4361 PL_lex_state = LEX_INTERPENDMAYBE;
4362 else if (*s != '[' && *s != '{')
4363 PL_lex_state = LEX_INTERPEND;
4366 if (PL_expect & XFAKEBRACK) {
4367 PL_expect &= XENUMMASK;
4369 return yylex(); /* ignore fake brackets */
4371 start_force(PL_curforce);
4373 curmad('X', newSVpvn(s-1,1));
4374 CURMAD('_', PL_thiswhite);
4379 PL_thistoken = newSVpvn("",0);
4387 if (PL_expect == XOPERATOR) {
4388 if (PL_bufptr == PL_linestart && ckWARN(WARN_SEMICOLON)
4389 && isIDFIRST_lazy_if(s,UTF))
4391 CopLINE_dec(PL_curcop);
4392 Perl_warner(aTHX_ packWARN(WARN_SEMICOLON), PL_warn_nosemi);
4393 CopLINE_inc(PL_curcop);
4398 s = scan_ident(s - 1, PL_bufend, PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
4400 PL_expect = XOPERATOR;
4401 force_ident(PL_tokenbuf, '&');
4405 yylval.ival = (OPpENTERSUB_AMPER<<8);
4417 const char tmp = *s++;
4424 if (tmp && isSPACE(*s) && ckWARN(WARN_SYNTAX)
4425 && strchr("+-*/%.^&|<",tmp))
4426 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
4427 "Reversed %c= operator",(int)tmp);
4429 if (PL_expect == XSTATE && isALPHA(tmp) &&
4430 (s == PL_linestart+1 || s[-2] == '\n') )
4432 if (PL_in_eval && !PL_rsfp) {
4437 if (strnEQ(s,"=cut",4)) {
4453 PL_thiswhite = newSVpvn("",0);
4454 sv_catpvn(PL_thiswhite, PL_linestart,
4455 PL_bufend - PL_linestart);
4459 PL_doextract = TRUE;
4463 if (PL_lex_brackets < PL_lex_formbrack) {
4465 #ifdef PERL_STRICT_CR
4466 for (t = s; SPACE_OR_TAB(*t); t++) ;
4468 for (t = s; SPACE_OR_TAB(*t) || *t == '\r'; t++) ;
4470 if (*t == '\n' || *t == '#') {
4481 const char tmp = *s++;
4483 /* was this !=~ where !~ was meant?
4484 * warn on m:!=~\s+([/?]|[msy]\W|tr\W): */
4486 if (*s == '~' && ckWARN(WARN_SYNTAX)) {
4487 const char *t = s+1;
4489 while (t < PL_bufend && isSPACE(*t))
4492 if (*t == '/' || *t == '?' ||
4493 ((*t == 'm' || *t == 's' || *t == 'y')
4494 && !isALNUM(t[1])) ||
4495 (*t == 't' && t[1] == 'r' && !isALNUM(t[2])))
4496 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
4497 "!=~ should be !~");
4507 if (PL_expect != XOPERATOR) {
4508 if (s[1] != '<' && !strchr(s,'>'))
4511 s = scan_heredoc(s);
4513 s = scan_inputsymbol(s);
4514 TERM(sublex_start());
4520 SHop(OP_LEFT_SHIFT);
4534 const char tmp = *s++;
4536 SHop(OP_RIGHT_SHIFT);
4537 else if (tmp == '=')
4546 if (PL_expect == XOPERATOR) {
4547 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
4549 deprecate_old(commaless_variable_list);
4550 return REPORT(','); /* grandfather non-comma-format format */
4554 if (s[1] == '#' && (isIDFIRST_lazy_if(s+2,UTF) || strchr("{$:+-", s[2]))) {
4555 PL_tokenbuf[0] = '@';
4556 s = scan_ident(s + 1, PL_bufend, PL_tokenbuf + 1,
4557 sizeof PL_tokenbuf - 1, FALSE);
4558 if (PL_expect == XOPERATOR)
4559 no_op("Array length", s);
4560 if (!PL_tokenbuf[1])
4562 PL_expect = XOPERATOR;
4563 PL_pending_ident = '#';
4567 PL_tokenbuf[0] = '$';
4568 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1,
4569 sizeof PL_tokenbuf - 1, FALSE);
4570 if (PL_expect == XOPERATOR)
4572 if (!PL_tokenbuf[1]) {
4574 yyerror("Final $ should be \\$ or $name");
4578 /* This kludge not intended to be bulletproof. */
4579 if (PL_tokenbuf[1] == '[' && !PL_tokenbuf[2]) {
4580 yylval.opval = newSVOP(OP_CONST, 0,
4581 newSViv(CopARYBASE_get(&PL_compiling)));
4582 yylval.opval->op_private = OPpCONST_ARYBASE;
4588 const char tmp = *s;
4589 if (PL_lex_state == LEX_NORMAL)
4592 if ((PL_expect != XREF || PL_oldoldbufptr == PL_last_lop)
4593 && intuit_more(s)) {
4595 PL_tokenbuf[0] = '@';
4596 if (ckWARN(WARN_SYNTAX)) {
4599 isSPACE(*t) || isALNUM_lazy_if(t,UTF) || *t == '$';
4602 PL_bufptr = PEEKSPACE(PL_bufptr); /* XXX can realloc */
4603 while (t < PL_bufend && *t != ']')
4605 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
4606 "Multidimensional syntax %.*s not supported",
4607 (int)((t - PL_bufptr) + 1), PL_bufptr);
4611 else if (*s == '{') {
4613 PL_tokenbuf[0] = '%';
4614 if (strEQ(PL_tokenbuf+1, "SIG") && ckWARN(WARN_SYNTAX)
4615 && (t = strchr(s, '}')) && (t = strchr(t, '=')))
4617 char tmpbuf[sizeof PL_tokenbuf];
4618 for (t++; isSPACE(*t); t++) ;
4619 if (isIDFIRST_lazy_if(t,UTF)) {
4621 t = scan_word(t, tmpbuf, sizeof tmpbuf, TRUE,
4623 for (; isSPACE(*t); t++) ;
4624 if (*t == ';' && get_cv(tmpbuf, FALSE))
4625 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
4626 "You need to quote \"%s\"",
4633 PL_expect = XOPERATOR;
4634 if (PL_lex_state == LEX_NORMAL && isSPACE((char)tmp)) {
4635 const bool islop = (PL_last_lop == PL_oldoldbufptr);
4636 if (!islop || PL_last_lop_op == OP_GREPSTART)
4637 PL_expect = XOPERATOR;
4638 else if (strchr("$@\"'`q", *s))
4639 PL_expect = XTERM; /* e.g. print $fh "foo" */
4640 else if (strchr("&*<%", *s) && isIDFIRST_lazy_if(s+1,UTF))
4641 PL_expect = XTERM; /* e.g. print $fh &sub */
4642 else if (isIDFIRST_lazy_if(s,UTF)) {
4643 char tmpbuf[sizeof PL_tokenbuf];
4645 scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
4646 if ((t2 = keyword(tmpbuf, len))) {
4647 /* binary operators exclude handle interpretations */
4659 PL_expect = XTERM; /* e.g. print $fh length() */
4664 PL_expect = XTERM; /* e.g. print $fh subr() */
4667 else if (isDIGIT(*s))
4668 PL_expect = XTERM; /* e.g. print $fh 3 */
4669 else if (*s == '.' && isDIGIT(s[1]))
4670 PL_expect = XTERM; /* e.g. print $fh .3 */
4671 else if ((*s == '?' || *s == '-' || *s == '+')
4672 && !isSPACE(s[1]) && s[1] != '=')
4673 PL_expect = XTERM; /* e.g. print $fh -1 */
4674 else if (*s == '/' && !isSPACE(s[1]) && s[1] != '='
4676 PL_expect = XTERM; /* e.g. print $fh /.../
4677 XXX except DORDOR operator
4679 else if (*s == '<' && s[1] == '<' && !isSPACE(s[2])
4681 PL_expect = XTERM; /* print $fh <<"EOF" */
4684 PL_pending_ident = '$';
4688 if (PL_expect == XOPERATOR)
4690 PL_tokenbuf[0] = '@';
4691 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1, FALSE);
4692 if (!PL_tokenbuf[1]) {
4695 if (PL_lex_state == LEX_NORMAL)
4697 if ((PL_expect != XREF || PL_oldoldbufptr == PL_last_lop) && intuit_more(s)) {
4699 PL_tokenbuf[0] = '%';
4701 /* Warn about @ where they meant $. */
4702 if (*s == '[' || *s == '{') {
4703 if (ckWARN(WARN_SYNTAX)) {
4704 const char *t = s + 1;
4705 while (*t && (isALNUM_lazy_if(t,UTF) || strchr(" \t$#+-'\"", *t)))
4707 if (*t == '}' || *t == ']') {
4709 PL_bufptr = PEEKSPACE(PL_bufptr); /* XXX can realloc */
4710 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
4711 "Scalar value %.*s better written as $%.*s",
4712 (int)(t-PL_bufptr), PL_bufptr,
4713 (int)(t-PL_bufptr-1), PL_bufptr+1);
4718 PL_pending_ident = '@';
4721 case '/': /* may be division, defined-or, or pattern */
4722 if (PL_expect == XTERMORDORDOR && s[1] == '/') {
4726 case '?': /* may either be conditional or pattern */
4727 if(PL_expect == XOPERATOR) {
4735 /* A // operator. */
4745 /* Disable warning on "study /blah/" */
4746 if (PL_oldoldbufptr == PL_last_uni
4747 && (*PL_last_uni != 's' || s - PL_last_uni < 5
4748 || memNE(PL_last_uni, "study", 5)
4749 || isALNUM_lazy_if(PL_last_uni+5,UTF)
4752 s = scan_pat(s,OP_MATCH);
4753 TERM(sublex_start());
4757 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack
4758 #ifdef PERL_STRICT_CR
4761 && (s[1] == '\n' || (s[1] == '\r' && s[2] == '\n'))
4763 && (s == PL_linestart || s[-1] == '\n') )
4765 PL_lex_formbrack = 0;
4769 if (PL_expect == XOPERATOR || !isDIGIT(s[1])) {
4775 yylval.ival = OPf_SPECIAL;
4781 if (PL_expect != XOPERATOR)
4786 case '0': case '1': case '2': case '3': case '4':
4787 case '5': case '6': case '7': case '8': case '9':
4788 s = scan_num(s, &yylval);
4789 DEBUG_T( { S_printbuf(aTHX_ "### Saw number in %s\n", s); } );
4790 if (PL_expect == XOPERATOR)
4795 s = scan_str(s,!!PL_madskills,FALSE);
4796 DEBUG_T( { S_printbuf(aTHX_ "### Saw string before %s\n", s); } );
4797 if (PL_expect == XOPERATOR) {
4798 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
4800 deprecate_old(commaless_variable_list);
4801 return REPORT(','); /* grandfather non-comma-format format */
4808 yylval.ival = OP_CONST;
4809 TERM(sublex_start());
4812 s = scan_str(s,!!PL_madskills,FALSE);
4813 DEBUG_T( { S_printbuf(aTHX_ "### Saw string before %s\n", s); } );
4814 if (PL_expect == XOPERATOR) {
4815 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
4817 deprecate_old(commaless_variable_list);
4818 return REPORT(','); /* grandfather non-comma-format format */
4825 yylval.ival = OP_CONST;
4826 /* FIXME. I think that this can be const if char *d is replaced by
4827 more localised variables. */
4828 for (d = SvPV(PL_lex_stuff, len); len; len--, d++) {
4829 if (*d == '$' || *d == '@' || *d == '\\' || !UTF8_IS_INVARIANT((U8)*d)) {
4830 yylval.ival = OP_STRINGIFY;
4834 TERM(sublex_start());
4837 s = scan_str(s,!!PL_madskills,FALSE);
4838 DEBUG_T( { S_printbuf(aTHX_ "### Saw backtick string before %s\n", s); } );
4839 if (PL_expect == XOPERATOR)
4840 no_op("Backticks",s);
4843 yylval.ival = OP_BACKTICK;
4845 TERM(sublex_start());
4849 if (PL_lex_inwhat && isDIGIT(*s) && ckWARN(WARN_SYNTAX))
4850 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),"Can't use \\%c to mean $%c in expression",
4852 if (PL_expect == XOPERATOR)
4853 no_op("Backslash",s);
4857 if (isDIGIT(s[1]) && PL_expect != XOPERATOR) {
4858 char *start = s + 2;
4859 while (isDIGIT(*start) || *start == '_')
4861 if (*start == '.' && isDIGIT(start[1])) {
4862 s = scan_num(s, &yylval);
4865 /* avoid v123abc() or $h{v1}, allow C<print v10;> */
4866 else if (!isALPHA(*start) && (PL_expect == XTERM
4867 || PL_expect == XREF || PL_expect == XSTATE
4868 || PL_expect == XTERMORDORDOR)) {
4869 /* XXX Use gv_fetchpvn rather than stomping on a const string */
4870 const char c = *start;
4873 gv = gv_fetchpv(s, 0, SVt_PVCV);
4876 s = scan_num(s, &yylval);
4883 if (isDIGIT(s[1]) && PL_expect == XOPERATOR) {
4919 I32 orig_keyword = 0;
4924 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
4926 /* Some keywords can be followed by any delimiter, including ':' */
4927 tmp = ((len == 1 && strchr("msyq", PL_tokenbuf[0])) ||
4928 (len == 2 && ((PL_tokenbuf[0] == 't' && PL_tokenbuf[1] == 'r') ||
4929 (PL_tokenbuf[0] == 'q' &&
4930 strchr("qwxr", PL_tokenbuf[1])))));
4932 /* x::* is just a word, unless x is "CORE" */
4933 if (!tmp && *s == ':' && s[1] == ':' && strNE(PL_tokenbuf, "CORE"))
4937 while (d < PL_bufend && isSPACE(*d))
4938 d++; /* no comments skipped here, or s### is misparsed */
4940 /* Is this a label? */
4941 if (!tmp && PL_expect == XSTATE
4942 && d < PL_bufend && *d == ':' && *(d + 1) != ':') {
4944 yylval.pval = savepv(PL_tokenbuf);
4949 /* Check for keywords */
4950 tmp = keyword(PL_tokenbuf, len);
4952 /* Is this a word before a => operator? */
4953 if (*d == '=' && d[1] == '>') {
4956 = (OP*)newSVOP(OP_CONST, 0,
4957 S_newSV_maybe_utf8(aTHX_ PL_tokenbuf, len));
4958 yylval.opval->op_private = OPpCONST_BARE;
4962 if (tmp < 0) { /* second-class keyword? */
4963 GV *ogv = NULL; /* override (winner) */
4964 GV *hgv = NULL; /* hidden (loser) */
4965 if (PL_expect != XOPERATOR && (*s != ':' || s[1] != ':')) {
4967 if ((gv = gv_fetchpvn_flags(PL_tokenbuf, len, 0, SVt_PVCV)) &&
4970 if (GvIMPORTED_CV(gv))
4972 else if (! CvMETHOD(cv))
4976 (gvp = (GV**)hv_fetch(PL_globalstash,PL_tokenbuf,len,FALSE)) &&
4977 (gv = *gvp) != (GV*)&PL_sv_undef &&
4978 GvCVu(gv) && GvIMPORTED_CV(gv))
4985 tmp = 0; /* overridden by import or by GLOBAL */
4988 && -tmp==KEY_lock /* XXX generalizable kludge */
4990 && !hv_fetchs(GvHVn(PL_incgv), "Thread.pm", FALSE))
4992 tmp = 0; /* any sub overrides "weak" keyword */
4994 else { /* no override */
4996 if (tmp == KEY_dump && ckWARN(WARN_MISC)) {
4997 Perl_warner(aTHX_ packWARN(WARN_MISC),
4998 "dump() better written as CORE::dump()");
5002 if (hgv && tmp != KEY_x && tmp != KEY_CORE
5003 && ckWARN(WARN_AMBIGUOUS)) /* never ambiguous */
5004 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
5005 "Ambiguous call resolved as CORE::%s(), %s",
5006 GvENAME(hgv), "qualify as such or use &");
5013 default: /* not a keyword */
5014 /* Trade off - by using this evil construction we can pull the
5015 variable gv into the block labelled keylookup. If not, then
5016 we have to give it function scope so that the goto from the
5017 earlier ':' case doesn't bypass the initialisation. */
5019 just_a_word_zero_gv:
5027 const char lastchar = (PL_bufptr == PL_oldoldbufptr ? 0 : PL_bufptr[-1]);
5030 SV *nextPL_nextwhite = 0;
5034 /* Get the rest if it looks like a package qualifier */
5036 if (*s == '\'' || (*s == ':' && s[1] == ':')) {
5038 s = scan_word(s, PL_tokenbuf + len, sizeof PL_tokenbuf - len,
5041 Perl_croak(aTHX_ "Bad name after %s%s", PL_tokenbuf,
5042 *s == '\'' ? "'" : "::");
5047 if (PL_expect == XOPERATOR) {
5048 if (PL_bufptr == PL_linestart) {
5049 CopLINE_dec(PL_curcop);
5050 Perl_warner(aTHX_ packWARN(WARN_SEMICOLON), PL_warn_nosemi);
5051 CopLINE_inc(PL_curcop);
5054 no_op("Bareword",s);
5057 /* Look for a subroutine with this name in current package,
5058 unless name is "Foo::", in which case Foo is a bearword
5059 (and a package name). */
5061 if (len > 2 && !PL_madskills &&
5062 PL_tokenbuf[len - 2] == ':' && PL_tokenbuf[len - 1] == ':')
5064 if (ckWARN(WARN_BAREWORD)
5065 && ! gv_fetchpvn_flags(PL_tokenbuf, len, 0, SVt_PVHV))
5066 Perl_warner(aTHX_ packWARN(WARN_BAREWORD),
5067 "Bareword \"%s\" refers to nonexistent package",
5070 PL_tokenbuf[len] = '\0';
5076 /* Mustn't actually add anything to a symbol table.
5077 But also don't want to "initialise" any placeholder
5078 constants that might already be there into full
5079 blown PVGVs with attached PVCV. */
5080 gv = gv_fetchpvn_flags(PL_tokenbuf, len,
5081 GV_NOADD_NOINIT, SVt_PVCV);
5086 /* if we saw a global override before, get the right name */
5089 sv = newSVpvs("CORE::GLOBAL::");
5090 sv_catpv(sv,PL_tokenbuf);
5093 /* If len is 0, newSVpv does strlen(), which is correct.
5094 If len is non-zero, then it will be the true length,
5095 and so the scalar will be created correctly. */
5096 sv = newSVpv(PL_tokenbuf,len);
5099 if (PL_madskills && !PL_thistoken) {
5100 char *start = SvPVX(PL_linestr) + PL_realtokenstart;
5101 PL_thistoken = newSVpv(start,s - start);
5102 PL_realtokenstart = s - SvPVX(PL_linestr);
5106 /* Presume this is going to be a bareword of some sort. */
5109 yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
5110 yylval.opval->op_private = OPpCONST_BARE;
5111 /* UTF-8 package name? */
5112 if (UTF && !IN_BYTES &&
5113 is_utf8_string((U8*)SvPVX_const(sv), SvCUR(sv)))
5116 /* And if "Foo::", then that's what it certainly is. */
5121 /* Do the explicit type check so that we don't need to force
5122 the initialisation of the symbol table to have a real GV.
5123 Beware - gv may not really be a PVGV, cv may not really be
5124 a PVCV, (because of the space optimisations that gv_init
5125 understands) But they're true if for this symbol there is
5126 respectively a typeglob and a subroutine.
5128 cv = gv ? ((SvTYPE(gv) == SVt_PVGV)
5129 /* Real typeglob, so get the real subroutine: */
5131 /* A proxy for a subroutine in this package? */
5132 : SvOK(gv) ? (CV *) gv : NULL)
5135 /* See if it's the indirect object for a list operator. */
5137 if (PL_oldoldbufptr &&
5138 PL_oldoldbufptr < PL_bufptr &&
5139 (PL_oldoldbufptr == PL_last_lop
5140 || PL_oldoldbufptr == PL_last_uni) &&
5141 /* NO SKIPSPACE BEFORE HERE! */
5142 (PL_expect == XREF ||
5143 ((PL_opargs[PL_last_lop_op] >> OASHIFT)& 7) == OA_FILEREF))
5145 bool immediate_paren = *s == '(';
5147 /* (Now we can afford to cross potential line boundary.) */
5148 s = SKIPSPACE2(s,nextPL_nextwhite);
5150 PL_nextwhite = nextPL_nextwhite; /* assume no & deception */
5153 /* Two barewords in a row may indicate method call. */
5155 if ((isIDFIRST_lazy_if(s,UTF) || *s == '$') &&
5156 (tmp = intuit_method(s, gv, cv)))
5159 /* If not a declared subroutine, it's an indirect object. */
5160 /* (But it's an indir obj regardless for sort.) */
5161 /* Also, if "_" follows a filetest operator, it's a bareword */
5164 ( !immediate_paren && (PL_last_lop_op == OP_SORT ||
5166 (PL_last_lop_op != OP_MAPSTART &&
5167 PL_last_lop_op != OP_GREPSTART))))
5168 || (PL_tokenbuf[0] == '_' && PL_tokenbuf[1] == '\0'
5169 && ((PL_opargs[PL_last_lop_op] & OA_CLASS_MASK) == OA_FILESTATOP))
5172 PL_expect = (PL_last_lop == PL_oldoldbufptr) ? XTERM : XOPERATOR;
5177 PL_expect = XOPERATOR;
5180 s = SKIPSPACE2(s,nextPL_nextwhite);
5181 PL_nextwhite = nextPL_nextwhite;
5186 /* Is this a word before a => operator? */
5187 if (*s == '=' && s[1] == '>' && !pkgname) {
5189 sv_setpv(((SVOP*)yylval.opval)->op_sv, PL_tokenbuf);
5190 if (UTF && !IN_BYTES && is_utf8_string((U8*)PL_tokenbuf, len))
5191 SvUTF8_on(((SVOP*)yylval.opval)->op_sv);
5195 /* If followed by a paren, it's certainly a subroutine. */
5199 for (d = s + 1; SPACE_OR_TAB(*d); d++) ;
5200 if (*d == ')' && (sv = gv_const_sv(gv))) {
5204 char *par = SvPVX(PL_linestr) + PL_realtokenstart;
5205 sv_catpvn(PL_thistoken, par, s - par);
5207 sv_free(PL_nextwhite);
5217 PL_nextwhite = PL_thiswhite;
5220 start_force(PL_curforce);
5222 NEXTVAL_NEXTTOKE.opval = yylval.opval;
5223 PL_expect = XOPERATOR;
5226 PL_nextwhite = nextPL_nextwhite;
5227 curmad('X', PL_thistoken);
5228 PL_thistoken = newSVpvn("",0);
5236 /* If followed by var or block, call it a method (unless sub) */
5238 if ((*s == '$' || *s == '{') && (!gv || !cv)) {
5239 PL_last_lop = PL_oldbufptr;
5240 PL_last_lop_op = OP_METHOD;
5244 /* If followed by a bareword, see if it looks like indir obj. */
5247 && (isIDFIRST_lazy_if(s,UTF) || *s == '$')
5248 && (tmp = intuit_method(s, gv, cv)))
5251 /* Not a method, so call it a subroutine (if defined) */
5254 if (lastchar == '-' && ckWARN_d(WARN_AMBIGUOUS))
5255 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
5256 "Ambiguous use of -%s resolved as -&%s()",
5257 PL_tokenbuf, PL_tokenbuf);
5258 /* Check for a constant sub */
5259 if ((sv = gv_const_sv(gv))) {
5261 SvREFCNT_dec(((SVOP*)yylval.opval)->op_sv);
5262 ((SVOP*)yylval.opval)->op_sv = SvREFCNT_inc_simple(sv);
5263 yylval.opval->op_private = 0;
5267 /* Resolve to GV now. */
5268 if (SvTYPE(gv) != SVt_PVGV) {
5269 gv = gv_fetchpv(PL_tokenbuf, 0, SVt_PVCV);
5270 assert (SvTYPE(gv) == SVt_PVGV);
5271 /* cv must have been some sort of placeholder, so
5272 now needs replacing with a real code reference. */
5276 op_free(yylval.opval);
5277 yylval.opval = newCVREF(0, newGVOP(OP_GV, 0, gv));
5278 yylval.opval->op_private |= OPpENTERSUB_NOPAREN;
5279 PL_last_lop = PL_oldbufptr;
5280 PL_last_lop_op = OP_ENTERSUB;
5281 /* Is there a prototype? */
5288 const char *proto = SvPV_const((SV*)cv, protolen);
5291 if (*proto == '$' && proto[1] == '\0')
5293 while (*proto == ';')
5295 if (*proto == '&' && *s == '{') {
5296 sv_setpv(PL_subname, PL_curstash ?
5297 "__ANON__" : "__ANON__::__ANON__");
5304 PL_nextwhite = PL_thiswhite;
5307 start_force(PL_curforce);
5308 NEXTVAL_NEXTTOKE.opval = yylval.opval;
5311 PL_nextwhite = nextPL_nextwhite;
5312 curmad('X', PL_thistoken);
5313 PL_thistoken = newSVpvn("",0);
5320 /* Guess harder when madskills require "best effort". */
5321 if (PL_madskills && (!gv || !GvCVu(gv))) {
5322 int probable_sub = 0;
5323 if (strchr("\"'`$@%0123456789!*+{[<", *s))
5325 else if (isALPHA(*s)) {
5329 d = scan_word(d, tmpbuf, sizeof tmpbuf, TRUE, &tmplen);
5330 if (!keyword(tmpbuf,tmplen))
5333 while (d < PL_bufend && isSPACE(*d))
5335 if (*d == '=' && d[1] == '>')
5340 gv = gv_fetchpv(PL_tokenbuf, TRUE, SVt_PVCV);
5341 op_free(yylval.opval);
5342 yylval.opval = newCVREF(0, newGVOP(OP_GV, 0, gv));
5343 yylval.opval->op_private |= OPpENTERSUB_NOPAREN;
5344 PL_last_lop = PL_oldbufptr;
5345 PL_last_lop_op = OP_ENTERSUB;
5346 PL_nextwhite = PL_thiswhite;
5348 start_force(PL_curforce);
5349 NEXTVAL_NEXTTOKE.opval = yylval.opval;
5351 PL_nextwhite = nextPL_nextwhite;
5352 curmad('X', PL_thistoken);
5353 PL_thistoken = newSVpvn("",0);
5358 NEXTVAL_NEXTTOKE.opval = yylval.opval;
5365 /* Call it a bare word */
5367 if (PL_hints & HINT_STRICT_SUBS)
5368 yylval.opval->op_private |= OPpCONST_STRICT;
5371 if (lastchar != '-') {
5372 if (ckWARN(WARN_RESERVED)) {
5373 for (d = PL_tokenbuf; *d && isLOWER(*d); d++) ;
5374 if (!*d && !gv_stashpv(PL_tokenbuf,FALSE))
5375 Perl_warner(aTHX_ packWARN(WARN_RESERVED), PL_warn_reserved,
5382 if ((lastchar == '*' || lastchar == '%' || lastchar == '&')
5383 && ckWARN_d(WARN_AMBIGUOUS)) {
5384 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
5385 "Operator or semicolon missing before %c%s",
5386 lastchar, PL_tokenbuf);
5387 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
5388 "Ambiguous use of %c resolved as operator %c",
5389 lastchar, lastchar);
5395 yylval.opval = (OP*)newSVOP(OP_CONST, 0,
5396 newSVpv(CopFILE(PL_curcop),0));
5400 yylval.opval = (OP*)newSVOP(OP_CONST, 0,
5401 Perl_newSVpvf(aTHX_ "%"IVdf, (IV)CopLINE(PL_curcop)));
5404 case KEY___PACKAGE__:
5405 yylval.opval = (OP*)newSVOP(OP_CONST, 0,
5407 ? newSVhek(HvNAME_HEK(PL_curstash))
5414 if (PL_rsfp && (!PL_in_eval || PL_tokenbuf[2] == 'D')) {
5415 const char *pname = "main";
5416 if (PL_tokenbuf[2] == 'D')
5417 pname = HvNAME_get(PL_curstash ? PL_curstash : PL_defstash);
5418 gv = gv_fetchpv(Perl_form(aTHX_ "%s::DATA", pname), GV_ADD,
5422 GvIOp(gv) = newIO();
5423 IoIFP(GvIOp(gv)) = PL_rsfp;
5424 #if defined(HAS_FCNTL) && defined(F_SETFD)
5426 const int fd = PerlIO_fileno(PL_rsfp);
5427 fcntl(fd,F_SETFD,fd >= 3);
5430 /* Mark this internal pseudo-handle as clean */
5431 IoFLAGS(GvIOp(gv)) |= IOf_UNTAINT;
5433 IoTYPE(GvIOp(gv)) = IoTYPE_PIPE;
5434 else if ((PerlIO*)PL_rsfp == PerlIO_stdin())
5435 IoTYPE(GvIOp(gv)) = IoTYPE_STD;
5437 IoTYPE(GvIOp(gv)) = IoTYPE_RDONLY;
5438 #if defined(WIN32) && !defined(PERL_TEXTMODE_SCRIPTS)
5439 /* if the script was opened in binmode, we need to revert
5440 * it to text mode for compatibility; but only iff it has CRs
5441 * XXX this is a questionable hack at best. */
5442 if (PL_bufend-PL_bufptr > 2
5443 && PL_bufend[-1] == '\n' && PL_bufend[-2] == '\r')
5446 if (IoTYPE(GvIOp(gv)) == IoTYPE_RDONLY) {
5447 loc = PerlIO_tell(PL_rsfp);
5448 (void)PerlIO_seek(PL_rsfp, 0L, 0);
5451 if (PerlLIO_setmode(PL_rsfp, O_TEXT) != -1) {
5453 if (PerlLIO_setmode(PerlIO_fileno(PL_rsfp), O_TEXT) != -1) {
5454 #endif /* NETWARE */
5455 #ifdef PERLIO_IS_STDIO /* really? */
5456 # if defined(__BORLANDC__)
5457 /* XXX see note in do_binmode() */
5458 ((FILE*)PL_rsfp)->flags &= ~_F_BIN;
5462 PerlIO_seek(PL_rsfp, loc, 0);
5466 #ifdef PERLIO_LAYERS
5469 PerlIO_apply_layers(aTHX_ PL_rsfp, NULL, ":utf8");
5470 else if (PL_encoding) {
5477 XPUSHs(PL_encoding);
5479 call_method("name", G_SCALAR);
5483 PerlIO_apply_layers(aTHX_ PL_rsfp, NULL,
5484 Perl_form(aTHX_ ":encoding(%"SVf")",
5493 if (PL_realtokenstart >= 0) {
5494 char *tstart = SvPVX(PL_linestr) + PL_realtokenstart;
5496 PL_endwhite = newSVpvn("",0);
5497 sv_catsv(PL_endwhite, PL_thiswhite);
5499 sv_catpvn(PL_endwhite, tstart, PL_bufend - tstart);
5500 PL_realtokenstart = -1;
5502 while ((s = filter_gets(PL_endwhite, PL_rsfp,
5503 SvCUR(PL_endwhite))) != Nullch) ;
5517 if (PL_expect == XSTATE) {
5524 if (*s == ':' && s[1] == ':') {
5527 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
5528 if (!(tmp = keyword(PL_tokenbuf, len)))
5529 Perl_croak(aTHX_ "CORE::%s is not a keyword", PL_tokenbuf);
5532 else if (tmp == KEY_require || tmp == KEY_do)
5533 /* that's a way to remember we saw "CORE::" */
5546 LOP(OP_ACCEPT,XTERM);
5552 LOP(OP_ATAN2,XTERM);
5558 LOP(OP_BINMODE,XTERM);
5561 LOP(OP_BLESS,XTERM);
5570 /* When 'use switch' is in effect, continue has a dual
5571 life as a control operator. */
5573 if (!FEATURE_IS_ENABLED("switch"))
5576 /* We have to disambiguate the two senses of
5577 "continue". If the next token is a '{' then
5578 treat it as the start of a continue block;
5579 otherwise treat it as a control operator.
5591 (void)gv_fetchpvs("ENV", GV_ADD|GV_NOTQUAL, SVt_PVHV);
5608 if (!PL_cryptseen) {
5609 PL_cryptseen = TRUE;
5613 LOP(OP_CRYPT,XTERM);
5616 LOP(OP_CHMOD,XTERM);
5619 LOP(OP_CHOWN,XTERM);
5622 LOP(OP_CONNECT,XTERM);
5641 s = force_word(s,WORD,TRUE,TRUE,FALSE);
5642 if (orig_keyword == KEY_do) {
5651 PL_hints |= HINT_BLOCK_SCOPE;
5661 gv_fetchpvs("AnyDBM_File::ISA", GV_ADDMULTI, SVt_PVAV);
5662 LOP(OP_DBMOPEN,XTERM);
5668 s = force_word(s,WORD,TRUE,FALSE,FALSE);
5675 yylval.ival = CopLINE(PL_curcop);
5691 PL_expect = (*s == '{') ? XTERMBLOCK : XTERM;
5692 UNIBRACK(OP_ENTEREVAL);
5710 case KEY_endhostent:
5716 case KEY_endservent:
5719 case KEY_endprotoent:
5730 yylval.ival = CopLINE(PL_curcop);
5732 if (PL_expect == XSTATE && isIDFIRST_lazy_if(s,UTF)) {
5735 int soff = s - SvPVX(PL_linestr); /* for skipspace realloc */
5738 if ((PL_bufend - p) >= 3 &&
5739 strnEQ(p, "my", 2) && isSPACE(*(p + 2)))
5741 else if ((PL_bufend - p) >= 4 &&
5742 strnEQ(p, "our", 3) && isSPACE(*(p + 3)))
5745 if (isIDFIRST_lazy_if(p,UTF)) {
5746 p = scan_ident(p, PL_bufend,
5747 PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
5751 Perl_croak(aTHX_ "Missing $ on loop variable");
5753 s = SvPVX(PL_linestr) + soff;
5759 LOP(OP_FORMLINE,XTERM);
5765 LOP(OP_FCNTL,XTERM);
5771 LOP(OP_FLOCK,XTERM);
5780 LOP(OP_GREPSTART, XREF);
5783 s = force_word(s,WORD,TRUE,FALSE,FALSE);
5798 case KEY_getpriority:
5799 LOP(OP_GETPRIORITY,XTERM);
5801 case KEY_getprotobyname:
5804 case KEY_getprotobynumber:
5805 LOP(OP_GPBYNUMBER,XTERM);
5807 case KEY_getprotoent:
5819 case KEY_getpeername:
5820 UNI(OP_GETPEERNAME);
5822 case KEY_gethostbyname:
5825 case KEY_gethostbyaddr:
5826 LOP(OP_GHBYADDR,XTERM);
5828 case KEY_gethostent:
5831 case KEY_getnetbyname:
5834 case KEY_getnetbyaddr:
5835 LOP(OP_GNBYADDR,XTERM);
5840 case KEY_getservbyname:
5841 LOP(OP_GSBYNAME,XTERM);
5843 case KEY_getservbyport:
5844 LOP(OP_GSBYPORT,XTERM);
5846 case KEY_getservent:
5849 case KEY_getsockname:
5850 UNI(OP_GETSOCKNAME);
5852 case KEY_getsockopt:
5853 LOP(OP_GSOCKOPT,XTERM);
5868 yylval.ival = CopLINE(PL_curcop);
5879 yylval.ival = CopLINE(PL_curcop);
5883 LOP(OP_INDEX,XTERM);
5889 LOP(OP_IOCTL,XTERM);
5901 s = force_word(s,WORD,TRUE,FALSE,FALSE);
5933 LOP(OP_LISTEN,XTERM);
5942 s = scan_pat(s,OP_MATCH);
5943 TERM(sublex_start());
5946 LOP(OP_MAPSTART, XREF);
5949 LOP(OP_MKDIR,XTERM);
5952 LOP(OP_MSGCTL,XTERM);
5955 LOP(OP_MSGGET,XTERM);
5958 LOP(OP_MSGRCV,XTERM);
5961 LOP(OP_MSGSND,XTERM);
5967 if (isIDFIRST_lazy_if(s,UTF)) {
5971 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, TRUE, &len);
5972 if (len == 3 && strnEQ(PL_tokenbuf, "sub", 3))
5974 PL_in_my_stash = find_in_my_stash(PL_tokenbuf, len);
5975 if (!PL_in_my_stash) {
5978 sprintf(tmpbuf, "No such class %.1000s", PL_tokenbuf);
5982 if (PL_madskills) { /* just add type to declarator token */
5983 sv_catsv(PL_thistoken, PL_nextwhite);
5985 sv_catpvn(PL_thistoken, start, s - start);
5993 s = force_word(s,WORD,TRUE,FALSE,FALSE);
6000 s = tokenize_use(0, s);
6004 if (*s == '(' || (s = SKIPSPACE1(s), *s == '('))
6011 if (isIDFIRST_lazy_if(s,UTF)) {
6013 for (d = s; isALNUM_lazy_if(d,UTF); d++) ;
6014 for (t=d; *t && isSPACE(*t); t++) ;
6015 if ( *t && strchr("|&*+-=!?:.", *t) && ckWARN_d(WARN_PRECEDENCE)
6017 && !(t[0] == '=' && t[1] == '>')
6019 int parms_len = (int)(d-s);
6020 Perl_warner(aTHX_ packWARN(WARN_PRECEDENCE),
6021 "Precedence problem: open %.*s should be open(%.*s)",
6022 parms_len, s, parms_len, s);
6028 yylval.ival = OP_OR;
6038 LOP(OP_OPEN_DIR,XTERM);
6041 checkcomma(s,PL_tokenbuf,"filehandle");
6045 checkcomma(s,PL_tokenbuf,"filehandle");
6064 s = force_word(s,WORD,FALSE,TRUE,FALSE);
6068 LOP(OP_PIPE_OP,XTERM);
6071 s = scan_str(s,!!PL_madskills,FALSE);
6074 yylval.ival = OP_CONST;
6075 TERM(sublex_start());
6081 s = scan_str(s,!!PL_madskills,FALSE);
6084 PL_expect = XOPERATOR;
6086 if (SvCUR(PL_lex_stuff)) {
6089 d = SvPV_force(PL_lex_stuff, len);
6091 for (; isSPACE(*d) && len; --len, ++d)
6096 if (!warned && ckWARN(WARN_QW)) {
6097 for (; !isSPACE(*d) && len; --len, ++d) {
6099 Perl_warner(aTHX_ packWARN(WARN_QW),
6100 "Possible attempt to separate words with commas");
6103 else if (*d == '#') {
6104 Perl_warner(aTHX_ packWARN(WARN_QW),
6105 "Possible attempt to put comments in qw() list");
6111 for (; !isSPACE(*d) && len; --len, ++d)
6114 sv = newSVpvn(b, d-b);
6115 if (DO_UTF8(PL_lex_stuff))
6117 words = append_elem(OP_LIST, words,
6118 newSVOP(OP_CONST, 0, tokeq(sv)));
6122 start_force(PL_curforce);
6123 NEXTVAL_NEXTTOKE.opval = words;
6128 SvREFCNT_dec(PL_lex_stuff);
6129 PL_lex_stuff = NULL;
6135 s = scan_str(s,!!PL_madskills,FALSE);
6138 yylval.ival = OP_STRINGIFY;
6139 if (SvIVX(PL_lex_stuff) == '\'')
6140 SvIV_set(PL_lex_stuff, 0); /* qq'$foo' should intepolate */
6141 TERM(sublex_start());
6144 s = scan_pat(s,OP_QR);
6145 TERM(sublex_start());
6148 s = scan_str(s,!!PL_madskills,FALSE);
6151 yylval.ival = OP_BACKTICK;
6153 TERM(sublex_start());
6161 s = force_version(s, FALSE);
6163 else if (*s != 'v' || !isDIGIT(s[1])
6164 || (s = force_version(s, TRUE), *s == 'v'))
6166 *PL_tokenbuf = '\0';
6167 s = force_word(s,WORD,TRUE,TRUE,FALSE);
6168 if (isIDFIRST_lazy_if(PL_tokenbuf,UTF))
6169 gv_stashpvn(PL_tokenbuf, strlen(PL_tokenbuf), TRUE);
6171 yyerror("<> should be quotes");
6173 if (orig_keyword == KEY_require) {
6181 PL_last_uni = PL_oldbufptr;
6182 PL_last_lop_op = OP_REQUIRE;
6184 return REPORT( (int)REQUIRE );
6190 s = force_word(s,WORD,TRUE,FALSE,FALSE);
6194 LOP(OP_RENAME,XTERM);
6203 LOP(OP_RINDEX,XTERM);
6213 UNIDOR(OP_READLINE);
6226 LOP(OP_REVERSE,XTERM);
6229 UNIDOR(OP_READLINK);
6237 TERM(sublex_start());
6239 TOKEN(1); /* force error */
6242 checkcomma(s,PL_tokenbuf,"filehandle");
6252 LOP(OP_SELECT,XTERM);
6258 LOP(OP_SEMCTL,XTERM);
6261 LOP(OP_SEMGET,XTERM);
6264 LOP(OP_SEMOP,XTERM);
6270 LOP(OP_SETPGRP,XTERM);
6272 case KEY_setpriority:
6273 LOP(OP_SETPRIORITY,XTERM);
6275 case KEY_sethostent:
6281 case KEY_setservent:
6284 case KEY_setprotoent:
6294 LOP(OP_SEEKDIR,XTERM);
6296 case KEY_setsockopt:
6297 LOP(OP_SSOCKOPT,XTERM);
6303 LOP(OP_SHMCTL,XTERM);
6306 LOP(OP_SHMGET,XTERM);
6309 LOP(OP_SHMREAD,XTERM);
6312 LOP(OP_SHMWRITE,XTERM);
6315 LOP(OP_SHUTDOWN,XTERM);
6324 LOP(OP_SOCKET,XTERM);
6326 case KEY_socketpair:
6327 LOP(OP_SOCKPAIR,XTERM);
6330 checkcomma(s,PL_tokenbuf,"subroutine name");
6332 if (*s == ';' || *s == ')') /* probably a close */
6333 Perl_croak(aTHX_ "sort is now a reserved word");
6335 s = force_word(s,WORD,TRUE,TRUE,FALSE);
6339 LOP(OP_SPLIT,XTERM);
6342 LOP(OP_SPRINTF,XTERM);
6345 LOP(OP_SPLICE,XTERM);
6360 LOP(OP_SUBSTR,XTERM);
6366 char tmpbuf[sizeof PL_tokenbuf];
6367 SSize_t tboffset = 0;
6368 expectation attrful;
6369 bool have_name, have_proto, bad_proto;
6370 const int key = tmp;
6375 char *tstart = SvPVX(PL_linestr) + PL_realtokenstart;
6376 SV *subtoken = newSVpvn(tstart, s - tstart);
6380 s = SKIPSPACE2(s,tmpwhite);
6385 if (isIDFIRST_lazy_if(s,UTF) || *s == '\'' ||
6386 (*s == ':' && s[1] == ':'))
6393 attrful = XATTRBLOCK;
6394 /* remember buffer pos'n for later force_word */
6395 tboffset = s - PL_oldbufptr;
6396 d = scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
6399 nametoke = newSVpvn(s, d - s);
6401 if (strchr(tmpbuf, ':'))
6402 sv_setpv(PL_subname, tmpbuf);
6404 sv_setsv(PL_subname,PL_curstname);
6405 sv_catpvs(PL_subname,"::");
6406 sv_catpvn(PL_subname,tmpbuf,len);
6413 CURMAD('X', nametoke);
6414 CURMAD('_', tmpwhite);
6415 (void) force_word(PL_oldbufptr + tboffset, WORD,
6418 s = SKIPSPACE2(d,tmpwhite);
6425 Perl_croak(aTHX_ "Missing name in \"my sub\"");
6426 PL_expect = XTERMBLOCK;
6427 attrful = XATTRTERM;
6428 sv_setpvn(PL_subname,"?",1);
6432 if (key == KEY_format) {
6434 PL_lex_formbrack = PL_lex_brackets + 1;
6436 PL_thistoken = subtoken;
6440 (void) force_word(PL_oldbufptr + tboffset, WORD,
6446 /* Look for a prototype */
6450 s = scan_str(s,!!PL_madskills,FALSE);
6452 Perl_croak(aTHX_ "Prototype not terminated");
6453 /* strip spaces and check for bad characters */
6454 d = SvPVX(PL_lex_stuff);
6457 for (p = d; *p; ++p) {
6460 if (!strchr("$@%*;[]&\\", *p))
6465 if (bad_proto && ckWARN(WARN_SYNTAX))
6466 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
6467 "Illegal character in prototype for %"SVf" : %s",
6469 SvCUR_set(PL_lex_stuff, tmp);
6474 CURMAD('q', PL_thisopen);
6475 CURMAD('_', tmpwhite);
6476 CURMAD('=', PL_thisstuff);
6477 CURMAD('Q', PL_thisclose);
6478 NEXTVAL_NEXTTOKE.opval =
6479 (OP*)newSVOP(OP_CONST, 0, PL_lex_stuff);
6480 PL_lex_stuff = Nullsv;
6483 s = SKIPSPACE2(s,tmpwhite);
6491 if (*s == ':' && s[1] != ':')
6492 PL_expect = attrful;
6493 else if (*s != '{' && key == KEY_sub) {
6495 Perl_croak(aTHX_ "Illegal declaration of anonymous subroutine");
6497 Perl_croak(aTHX_ "Illegal declaration of subroutine %"SVf, PL_subname);
6504 curmad('^', newSVpvn("",0));
6505 CURMAD('_', tmpwhite);
6509 PL_thistoken = subtoken;
6512 NEXTVAL_NEXTTOKE.opval =
6513 (OP*)newSVOP(OP_CONST, 0, PL_lex_stuff);
6514 PL_lex_stuff = NULL;
6519 sv_setpv(PL_subname,
6520 PL_curstash ? "__ANON__" : "__ANON__::__ANON__");
6524 (void) force_word(PL_oldbufptr + tboffset, WORD,
6534 LOP(OP_SYSTEM,XREF);
6537 LOP(OP_SYMLINK,XTERM);
6540 LOP(OP_SYSCALL,XTERM);
6543 LOP(OP_SYSOPEN,XTERM);
6546 LOP(OP_SYSSEEK,XTERM);
6549 LOP(OP_SYSREAD,XTERM);
6552 LOP(OP_SYSWRITE,XTERM);
6556 TERM(sublex_start());
6577 LOP(OP_TRUNCATE,XTERM);
6589 yylval.ival = CopLINE(PL_curcop);
6593 yylval.ival = CopLINE(PL_curcop);
6597 LOP(OP_UNLINK,XTERM);
6603 LOP(OP_UNPACK,XTERM);
6606 LOP(OP_UTIME,XTERM);
6612 LOP(OP_UNSHIFT,XTERM);
6615 s = tokenize_use(1, s);
6625 yylval.ival = CopLINE(PL_curcop);
6629 yylval.ival = CopLINE(PL_curcop);
6633 PL_hints |= HINT_BLOCK_SCOPE;
6640 LOP(OP_WAITPID,XTERM);
6649 ctl_l[0] = toCTRL('L');
6651 gv_fetchpvn_flags(ctl_l, 1, GV_ADD|GV_NOTQUAL, SVt_PV);
6654 /* Make sure $^L is defined */
6655 gv_fetchpvs("\f", GV_ADD|GV_NOTQUAL, SVt_PV);
6660 if (PL_expect == XOPERATOR)
6666 yylval.ival = OP_XOR;
6671 TERM(sublex_start());
6676 #pragma segment Main
6680 S_pending_ident(pTHX)
6684 register PADOFFSET tmp = 0;
6685 /* pit holds the identifier we read and pending_ident is reset */
6686 char pit = PL_pending_ident;
6687 PL_pending_ident = 0;
6689 /* PL_realtokenstart = realtokenend = PL_bufptr - SvPVX(PL_linestr); */
6690 DEBUG_T({ PerlIO_printf(Perl_debug_log,
6691 "### Pending identifier '%s'\n", PL_tokenbuf); });
6693 /* if we're in a my(), we can't allow dynamics here.
6694 $foo'bar has already been turned into $foo::bar, so
6695 just check for colons.
6697 if it's a legal name, the OP is a PADANY.
6700 if (PL_in_my == KEY_our) { /* "our" is merely analogous to "my" */
6701 if (strchr(PL_tokenbuf,':'))
6702 yyerror(Perl_form(aTHX_ "No package name allowed for "
6703 "variable %s in \"our\"",
6705 tmp = allocmy(PL_tokenbuf);
6708 if (strchr(PL_tokenbuf,':'))
6709 yyerror(Perl_form(aTHX_ PL_no_myglob,PL_tokenbuf));
6711 yylval.opval = newOP(OP_PADANY, 0);
6712 yylval.opval->op_targ = allocmy(PL_tokenbuf);
6718 build the ops for accesses to a my() variable.
6720 Deny my($a) or my($b) in a sort block, *if* $a or $b is
6721 then used in a comparison. This catches most, but not
6722 all cases. For instance, it catches
6723 sort { my($a); $a <=> $b }
6725 sort { my($a); $a < $b ? -1 : $a == $b ? 0 : 1; }
6726 (although why you'd do that is anyone's guess).
6729 if (!strchr(PL_tokenbuf,':')) {
6731 tmp = pad_findmy(PL_tokenbuf);
6732 if (tmp != NOT_IN_PAD) {
6733 /* might be an "our" variable" */
6734 if (PAD_COMPNAME_FLAGS_isOUR(tmp)) {
6735 /* build ops for a bareword */
6736 HV * const stash = PAD_COMPNAME_OURSTASH(tmp);
6737 HEK * const stashname = HvNAME_HEK(stash);
6738 SV * const sym = newSVhek(stashname);
6739 sv_catpvs(sym, "::");
6740 sv_catpv(sym, PL_tokenbuf+1);
6741 yylval.opval = (OP*)newSVOP(OP_CONST, 0, sym);
6742 yylval.opval->op_private = OPpCONST_ENTERED;
6745 ? (GV_ADDMULTI | GV_ADDINEVAL)
6748 ((PL_tokenbuf[0] == '$') ? SVt_PV
6749 : (PL_tokenbuf[0] == '@') ? SVt_PVAV
6754 /* if it's a sort block and they're naming $a or $b */
6755 if (PL_last_lop_op == OP_SORT &&
6756 PL_tokenbuf[0] == '$' &&
6757 (PL_tokenbuf[1] == 'a' || PL_tokenbuf[1] == 'b')
6760 for (d = PL_in_eval ? PL_oldoldbufptr : PL_linestart;
6761 d < PL_bufend && *d != '\n';
6764 if (strnEQ(d,"<=>",3) || strnEQ(d,"cmp",3)) {
6765 Perl_croak(aTHX_ "Can't use \"my %s\" in sort comparison",
6771 yylval.opval = newOP(OP_PADANY, 0);
6772 yylval.opval->op_targ = tmp;
6778 Whine if they've said @foo in a doublequoted string,
6779 and @foo isn't a variable we can find in the symbol
6782 if (pit == '@' && PL_lex_state != LEX_NORMAL && !PL_lex_brackets) {
6783 GV *gv = gv_fetchpv(PL_tokenbuf+1, 0, SVt_PVAV);
6784 if ((!gv || ((PL_tokenbuf[0] == '@') ? !GvAV(gv) : !GvHV(gv)))
6785 && ckWARN(WARN_AMBIGUOUS))
6787 /* Downgraded from fatal to warning 20000522 mjd */
6788 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
6789 "Possible unintended interpolation of %s in string",
6794 /* build ops for a bareword */
6795 yylval.opval = (OP*)newSVOP(OP_CONST, 0, newSVpv(PL_tokenbuf+1, 0));
6796 yylval.opval->op_private = OPpCONST_ENTERED;
6799 /* If the identifier refers to a stash, don't autovivify it.
6800 * Change 24660 had the side effect of causing symbol table
6801 * hashes to always be defined, even if they were freshly
6802 * created and the only reference in the entire program was
6803 * the single statement with the defined %foo::bar:: test.
6804 * It appears that all code in the wild doing this actually
6805 * wants to know whether sub-packages have been loaded, so
6806 * by avoiding auto-vivifying symbol tables, we ensure that
6807 * defined %foo::bar:: continues to be false, and the existing
6808 * tests still give the expected answers, even though what
6809 * they're actually testing has now changed subtly.
6811 (*PL_tokenbuf == '%' && *(d = PL_tokenbuf + strlen(PL_tokenbuf) - 1) == ':' && d[-1] == ':'
6813 : PL_in_eval ? (GV_ADDMULTI | GV_ADDINEVAL) : GV_ADD),
6814 ((PL_tokenbuf[0] == '$') ? SVt_PV
6815 : (PL_tokenbuf[0] == '@') ? SVt_PVAV
6821 * The following code was generated by perl_keyword.pl.
6825 Perl_keyword (pTHX_ const char *name, I32 len)
6830 case 1: /* 5 tokens of length 1 */
6862 case 2: /* 18 tokens of length 2 */
7008 case 3: /* 29 tokens of length 3 */
7012 if (name[1] == 'N' &&
7075 if (name[1] == 'i' &&
7097 return (FEATURE_IS_ENABLED("err") ? -KEY_err : 0);
7115 if (name[1] == 'o' &&
7124 if (name[1] == 'e' &&
7133 if (name[1] == 'n' &&
7142 if (name[1] == 'o' &&
7151 if (name[1] == 'a' &&
7160 if (name[1] == 'o' &&
7222 if (name[1] == 'e' &&
7236 return (FEATURE_IS_ENABLED("say") ? -KEY_say : 0);
7262 if (name[1] == 'i' &&
7271 if (name[1] == 's' &&
7280 if (name[1] == 'e' &&
7289 if (name[1] == 'o' &&
7301 case 4: /* 41 tokens of length 4 */
7305 if (name[1] == 'O' &&
7315 if (name[1] == 'N' &&
7325 if (name[1] == 'i' &&
7335 if (name[1] == 'h' &&
7345 if (name[1] == 'u' &&
7358 if (name[2] == 'c' &&
7367 if (name[2] == 's' &&
7376 if (name[2] == 'a' &&
7412 if (name[1] == 'o' &&
7425 if (name[2] == 't' &&
7434 if (name[2] == 'o' &&
7443 if (name[2] == 't' &&
7452 if (name[2] == 'e' &&
7465 if (name[1] == 'o' &&
7478 if (name[2] == 'y' &&
7487 if (name[2] == 'l' &&
7503 if (name[2] == 's' &&
7512 if (name[2] == 'n' &&
7521 if (name[2] == 'c' &&
7534 if (name[1] == 'e' &&
7544 if (name[1] == 'p' &&
7557 if (name[2] == 'c' &&
7566 if (name[2] == 'p' &&
7575 if (name[2] == 's' &&
7591 if (name[2] == 'n' &&
7661 if (name[2] == 'r' &&
7670 if (name[2] == 'r' &&
7679 if (name[2] == 'a' &&
7695 if (name[2] == 'l' &&
7757 if (name[2] == 'e' &&
7760 return (FEATURE_IS_ENABLED("switch") ? KEY_when : 0);
7773 case 5: /* 38 tokens of length 5 */
7777 if (name[1] == 'E' &&
7788 if (name[1] == 'H' &&
7802 if (name[2] == 'a' &&
7812 if (name[2] == 'a' &&
7829 if (name[2] == 'e' &&
7839 if (name[2] == 'e' &&
7843 return (FEATURE_IS_ENABLED("switch") ? -KEY_break : 0);
7859 if (name[3] == 'i' &&
7868 if (name[3] == 'o' &&
7904 if (name[2] == 'o' &&
7914 if (name[2] == 'y' &&
7928 if (name[1] == 'l' &&
7942 if (name[2] == 'n' &&
7952 if (name[2] == 'o' &&
7966 if (name[1] == 'i' &&
7971 return (FEATURE_IS_ENABLED("switch") ? KEY_given : 0);
7980 if (name[2] == 'd' &&
7990 if (name[2] == 'c' &&
8007 if (name[2] == 'c' &&
8017 if (name[2] == 't' &&
8031 if (name[1] == 'k' &&
8042 if (name[1] == 'r' &&
8056 if (name[2] == 's' &&
8066 if (name[2] == 'd' &&
8083 if (name[2] == 'm' &&
8093 if (name[2] == 'i' &&
8103 if (name[2] == 'e' &&
8113 if (name[2] == 'l' &&
8123 if (name[2] == 'a' &&
8133 if (name[2] == 'u' &&
8147 if (name[1] == 'i' &&
8161 if (name[2] == 'a' &&
8174 if (name[3] == 'e' &&
8209 if (name[2] == 'i' &&
8226 if (name[2] == 'i' &&
8236 if (name[2] == 'i' &&
8253 case 6: /* 33 tokens of length 6 */
8257 if (name[1] == 'c' &&
8272 if (name[2] == 'l' &&
8283 if (name[2] == 'r' &&
8298 if (name[1] == 'e' &&
8313 if (name[2] == 's' &&
8318 if(ckWARN_d(WARN_SYNTAX))
8319 Perl_warner(aTHX_ packWARN(WARN_SYNTAX), "elseif should be elsif");
8325 if (name[2] == 'i' &&
8343 if (name[2] == 'l' &&
8354 if (name[2] == 'r' &&
8369 if (name[1] == 'm' &&
8384 if (name[2] == 'n' &&
8395 if (name[2] == 's' &&
8410 if (name[1] == 's' &&
8416 if (name[4] == 't' &&
8425 if (name[4] == 'e' &&
8434 if (name[4] == 'c' &&
8443 if (name[4] == 'n' &&
8459 if (name[1] == 'r' &&
8477 if (name[3] == 'a' &&
8487 if (name[3] == 'u' &&
8501 if (name[2] == 'n' &&
8519 if (name[2] == 'a' &&
8533 if (name[3] == 'e' &&
8546 if (name[4] == 't' &&
8555 if (name[4] == 'e' &&
8577 if (name[4] == 't' &&
8586 if (name[4] == 'e' &&
8602 if (name[2] == 'c' &&
8613 if (name[2] == 'l' &&
8624 if (name[2] == 'b' &&
8635 if (name[2] == 's' &&
8658 if (name[4] == 's' &&
8667 if (name[4] == 'n' &&
8680 if (name[3] == 'a' &&
8697 if (name[1] == 'a' &&
8712 case 7: /* 29 tokens of length 7 */
8716 if (name[1] == 'E' &&
8729 if (name[1] == '_' &&
8742 if (name[1] == 'i' &&
8749 return -KEY_binmode;
8755 if (name[1] == 'o' &&
8762 return -KEY_connect;
8771 if (name[2] == 'm' &&
8777 return -KEY_dbmopen;
8788 if (name[4] == 'u' &&
8792 return (FEATURE_IS_ENABLED("switch") ? KEY_default : 0);
8798 if (name[4] == 'n' &&
8819 if (name[1] == 'o' &&
8832 if (name[1] == 'e' &&
8839 if (name[5] == 'r' &&
8842 return -KEY_getpgrp;
8848 if (name[5] == 'i' &&
8851 return -KEY_getppid;
8864 if (name[1] == 'c' &&
8871 return -KEY_lcfirst;
8877 if (name[1] == 'p' &&
8884 return -KEY_opendir;
8890 if (name[1] == 'a' &&
8908 if (name[3] == 'd' &&
8913 return -KEY_readdir;
8919 if (name[3] == 'u' &&
8930 if (name[3] == 'e' &&
8935 return -KEY_reverse;
8954 if (name[3] == 'k' &&
8959 return -KEY_seekdir;
8965 if (name[3] == 'p' &&
8970 return -KEY_setpgrp;
8980 if (name[2] == 'm' &&
8986 return -KEY_shmread;
8992 if (name[2] == 'r' &&
8998 return -KEY_sprintf;
9007 if (name[3] == 'l' &&
9012 return -KEY_symlink;
9021 if (name[4] == 'a' &&
9025 return -KEY_syscall;
9031 if (name[4] == 'p' &&
9035 return -KEY_sysopen;
9041 if (name[4] == 'e' &&
9045 return -KEY_sysread;
9051 if (name[4] == 'e' &&
9055 return -KEY_sysseek;
9073 if (name[1] == 'e' &&
9080 return -KEY_telldir;
9089 if (name[2] == 'f' &&
9095 return -KEY_ucfirst;
9101 if (name[2] == 's' &&
9107 return -KEY_unshift;
9117 if (name[1] == 'a' &&
9124 return -KEY_waitpid;
9133 case 8: /* 26 tokens of length 8 */
9137 if (name[1] == 'U' &&
9145 return KEY_AUTOLOAD;
9156 if (name[3] == 'A' &&
9162 return KEY___DATA__;
9168 if (name[3] == 'I' &&
9174 return -KEY___FILE__;
9180 if (name[3] == 'I' &&
9186 return -KEY___LINE__;
9202 if (name[2] == 'o' &&
9209 return -KEY_closedir;
9215 if (name[2] == 'n' &&
9222 return -KEY_continue;
9232 if (name[1] == 'b' &&
9240 return -KEY_dbmclose;
9246 if (name[1] == 'n' &&
9252 if (name[4] == 'r' &&
9257 return -KEY_endgrent;
9263 if (name[4] == 'w' &&
9268 return -KEY_endpwent;
9281 if (name[1] == 'o' &&
9289 return -KEY_formline;
9295 if (name[1] == 'e' &&
9306 if (name[6] == 'n' &&
9309 return -KEY_getgrent;
9315 if (name[6] == 'i' &&
9318 return -KEY_getgrgid;
9324 if (name[6] == 'a' &&
9327 return -KEY_getgrnam;
9340 if (name[4] == 'o' &&
9345 return -KEY_getlogin;
9356 if (name[6] == 'n' &&
9359 return -KEY_getpwent;
9365 if (name[6] == 'a' &&
9368 return -KEY_getpwnam;
9374 if (name[6] == 'i' &&
9377 return -KEY_getpwuid;
9397 if (name[1] == 'e' &&
9404 if (name[5] == 'i' &&
9411 return -KEY_readline;
9416 return -KEY_readlink;
9427 if (name[5] == 'i' &&
9431 return -KEY_readpipe;
9452 if (name[4] == 'r' &&
9457 return -KEY_setgrent;
9463 if (name[4] == 'w' &&
9468 return -KEY_setpwent;
9484 if (name[3] == 'w' &&
9490 return -KEY_shmwrite;
9496 if (name[3] == 't' &&
9502 return -KEY_shutdown;
9512 if (name[2] == 's' &&
9519 return -KEY_syswrite;
9529 if (name[1] == 'r' &&
9537 return -KEY_truncate;
9546 case 9: /* 8 tokens of length 9 */
9550 if (name[1] == 'n' &&
9559 return -KEY_endnetent;
9565 if (name[1] == 'e' &&
9574 return -KEY_getnetent;
9580 if (name[1] == 'o' &&
9589 return -KEY_localtime;
9595 if (name[1] == 'r' &&
9604 return KEY_prototype;
9610 if (name[1] == 'u' &&
9619 return -KEY_quotemeta;
9625 if (name[1] == 'e' &&
9634 return -KEY_rewinddir;
9640 if (name[1] == 'e' &&
9649 return -KEY_setnetent;
9655 if (name[1] == 'a' &&
9664 return -KEY_wantarray;
9673 case 10: /* 9 tokens of length 10 */
9677 if (name[1] == 'n' &&
9683 if (name[4] == 'o' &&
9690 return -KEY_endhostent;
9696 if (name[4] == 'e' &&
9703 return -KEY_endservent;
9716 if (name[1] == 'e' &&
9722 if (name[4] == 'o' &&
9729 return -KEY_gethostent;
9738 if (name[5] == 'r' &&
9744 return -KEY_getservent;
9750 if (name[5] == 'c' &&
9756 return -KEY_getsockopt;
9781 if (name[4] == 'o' &&
9788 return -KEY_sethostent;
9797 if (name[5] == 'r' &&
9803 return -KEY_setservent;
9809 if (name[5] == 'c' &&
9815 return -KEY_setsockopt;
9832 if (name[2] == 'c' &&
9841 return -KEY_socketpair;
9854 case 11: /* 8 tokens of length 11 */
9858 if (name[1] == '_' &&
9869 return -KEY___PACKAGE__;
9875 if (name[1] == 'n' &&
9886 return -KEY_endprotoent;
9892 if (name[1] == 'e' &&
9901 if (name[5] == 'e' &&
9908 return -KEY_getpeername;
9917 if (name[6] == 'o' &&
9923 return -KEY_getpriority;
9929 if (name[6] == 't' &&
9935 return -KEY_getprotoent;
9949 if (name[4] == 'o' &&
9957 return -KEY_getsockname;
9970 if (name[1] == 'e' &&
9978 if (name[6] == 'o' &&
9984 return -KEY_setpriority;
9990 if (name[6] == 't' &&
9996 return -KEY_setprotoent;
10012 case 12: /* 2 tokens of length 12 */
10013 if (name[0] == 'g' &&
10025 if (name[9] == 'd' &&
10028 { /* getnetbyaddr */
10029 return -KEY_getnetbyaddr;
10035 if (name[9] == 'a' &&
10038 { /* getnetbyname */
10039 return -KEY_getnetbyname;
10051 case 13: /* 4 tokens of length 13 */
10052 if (name[0] == 'g' &&
10059 if (name[4] == 'o' &&
10068 if (name[10] == 'd' &&
10071 { /* gethostbyaddr */
10072 return -KEY_gethostbyaddr;
10078 if (name[10] == 'a' &&
10081 { /* gethostbyname */
10082 return -KEY_gethostbyname;
10095 if (name[4] == 'e' &&
10104 if (name[10] == 'a' &&
10107 { /* getservbyname */
10108 return -KEY_getservbyname;
10114 if (name[10] == 'o' &&
10117 { /* getservbyport */
10118 return -KEY_getservbyport;
10137 case 14: /* 1 tokens of length 14 */
10138 if (name[0] == 'g' &&
10152 { /* getprotobyname */
10153 return -KEY_getprotobyname;
10158 case 16: /* 1 tokens of length 16 */
10159 if (name[0] == 'g' &&
10175 { /* getprotobynumber */
10176 return -KEY_getprotobynumber;
10190 S_checkcomma(pTHX_ const char *s, const char *name, const char *what)
10194 if (*s == ' ' && s[1] == '(') { /* XXX gotta be a better way */
10195 if (ckWARN(WARN_SYNTAX)) {
10198 for (w = s+2; *w && level; w++) {
10201 else if (*w == ')')
10204 while (isSPACE(*w))
10206 if (!*w || !strchr(";|})]oaiuw!=", *w)) /* an advisory hack only... */
10207 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
10208 "%s (...) interpreted as function",name);
10211 while (s < PL_bufend && isSPACE(*s))
10215 while (s < PL_bufend && isSPACE(*s))
10217 if (isIDFIRST_lazy_if(s,UTF)) {
10218 const char * const w = s++;
10219 while (isALNUM_lazy_if(s,UTF))
10221 while (s < PL_bufend && isSPACE(*s))
10225 if (keyword(w, s - w))
10228 gv = gv_fetchpvn_flags(w, s - w, 0, SVt_PVCV);
10229 if (gv && GvCVu(gv))
10231 Perl_croak(aTHX_ "No comma allowed after %s", what);
10236 /* Either returns sv, or mortalizes sv and returns a new SV*.
10237 Best used as sv=new_constant(..., sv, ...).
10238 If s, pv are NULL, calls subroutine with one argument,
10239 and type is used with error messages only. */
10242 S_new_constant(pTHX_ const char *s, STRLEN len, const char *key, SV *sv, SV *pv,
10246 HV * const table = GvHV(PL_hintgv); /* ^H */
10250 const char *why1 = "", *why2 = "", *why3 = "";
10252 if (!table || !(PL_hints & HINT_LOCALIZE_HH)) {
10255 why2 = strEQ(key,"charnames")
10256 ? "(possibly a missing \"use charnames ...\")"
10258 msg = Perl_newSVpvf(aTHX_ "Constant(%s) unknown: %s",
10259 (type ? type: "undef"), why2);
10261 /* This is convoluted and evil ("goto considered harmful")
10262 * but I do not understand the intricacies of all the different
10263 * failure modes of %^H in here. The goal here is to make
10264 * the most probable error message user-friendly. --jhi */
10269 msg = Perl_newSVpvf(aTHX_ "Constant(%s): %s%s%s",
10270 (type ? type: "undef"), why1, why2, why3);
10272 yyerror(SvPVX_const(msg));
10276 cvp = hv_fetch(table, key, strlen(key), FALSE);
10277 if (!cvp || !SvOK(*cvp)) {
10280 why3 = "} is not defined";
10283 sv_2mortal(sv); /* Parent created it permanently */
10286 pv = sv_2mortal(newSVpvn(s, len));
10288 typesv = sv_2mortal(newSVpv(type, 0));
10290 typesv = &PL_sv_undef;
10292 PUSHSTACKi(PERLSI_OVERLOAD);
10304 call_sv(cv, G_SCALAR | ( PL_in_eval ? 0 : G_EVAL));
10308 /* Check the eval first */
10309 if (!PL_in_eval && SvTRUE(ERRSV)) {
10310 sv_catpvs(ERRSV, "Propagated");
10311 yyerror(SvPV_nolen_const(ERRSV)); /* Duplicates the message inside eval */
10313 res = SvREFCNT_inc_simple(sv);
10317 SvREFCNT_inc_simple_void(res);
10326 why1 = "Call to &{$^H{";
10328 why3 = "}} did not return a defined value";
10336 /* Returns a NUL terminated string, with the length of the string written to
10340 S_scan_word(pTHX_ register char *s, char *dest, STRLEN destlen, int allow_package, STRLEN *slp)
10343 register char *d = dest;
10344 register char * const e = d + destlen - 3; /* two-character token, ending NUL */
10347 Perl_croak(aTHX_ ident_too_long);
10348 if (isALNUM(*s)) /* UTF handled below */
10350 else if (*s == '\'' && allow_package && isIDFIRST_lazy_if(s+1,UTF)) {
10355 else if (*s == ':' && s[1] == ':' && allow_package && s[2] != '$') {
10359 else if (UTF && UTF8_IS_START(*s) && isALNUM_utf8((U8*)s)) {
10360 char *t = s + UTF8SKIP(s);
10361 while (UTF8_IS_CONTINUED(*t) && is_utf8_mark((U8*)t))
10363 if (d + (t - s) > e)
10364 Perl_croak(aTHX_ ident_too_long);
10365 Copy(s, d, t - s, char);
10378 S_scan_ident(pTHX_ register char *s, register const char *send, char *dest, STRLEN destlen, I32 ck_uni)
10381 char *bracket = NULL;
10383 register char *d = dest;
10384 register char * const e = d + destlen + 3; /* two-character token, ending NUL */
10389 while (isDIGIT(*s)) {
10391 Perl_croak(aTHX_ ident_too_long);
10398 Perl_croak(aTHX_ ident_too_long);
10399 if (isALNUM(*s)) /* UTF handled below */
10401 else if (*s == '\'' && isIDFIRST_lazy_if(s+1,UTF)) {
10406 else if (*s == ':' && s[1] == ':') {
10410 else if (UTF && UTF8_IS_START(*s) && isALNUM_utf8((U8*)s)) {
10411 char *t = s + UTF8SKIP(s);
10412 while (UTF8_IS_CONTINUED(*t) && is_utf8_mark((U8*)t))
10414 if (d + (t - s) > e)
10415 Perl_croak(aTHX_ ident_too_long);
10416 Copy(s, d, t - s, char);
10427 if (PL_lex_state != LEX_NORMAL)
10428 PL_lex_state = LEX_INTERPENDMAYBE;
10431 if (*s == '$' && s[1] &&
10432 (isALNUM_lazy_if(s+1,UTF) || s[1] == '$' || s[1] == '{' || strnEQ(s+1,"::",2)) )
10445 if (*d == '^' && *s && isCONTROLVAR(*s)) {
10450 if (isSPACE(s[-1])) {
10452 const char ch = *s++;
10453 if (!SPACE_OR_TAB(ch)) {
10459 if (isIDFIRST_lazy_if(d,UTF)) {
10463 while ((end < send && isALNUM_lazy_if(end,UTF)) || *end == ':') {
10464 end += UTF8SKIP(end);
10465 while (end < send && UTF8_IS_CONTINUED(*end) && is_utf8_mark((U8*)end))
10466 end += UTF8SKIP(end);
10468 Copy(s, d, end - s, char);
10473 while ((isALNUM(*s) || *s == ':') && d < e)
10476 Perl_croak(aTHX_ ident_too_long);
10479 while (s < send && SPACE_OR_TAB(*s)) s++;
10480 if ((*s == '[' || (*s == '{' && strNE(dest, "sub")))) {
10481 if (ckWARN(WARN_AMBIGUOUS) && keyword(dest, d - dest)) {
10482 const char *brack = *s == '[' ? "[...]" : "{...}";
10483 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
10484 "Ambiguous use of %c{%s%s} resolved to %c%s%s",
10485 funny, dest, brack, funny, dest, brack);
10488 PL_lex_brackstack[PL_lex_brackets++] = (char)(XOPERATOR | XFAKEBRACK);
10492 /* Handle extended ${^Foo} variables
10493 * 1999-02-27 mjd-perl-patch@plover.com */
10494 else if (!isALNUM(*d) && !isPRINT(*d) /* isCTRL(d) */
10498 while (isALNUM(*s) && d < e) {
10502 Perl_croak(aTHX_ ident_too_long);
10507 if (PL_lex_state == LEX_INTERPNORMAL && !PL_lex_brackets) {
10508 PL_lex_state = LEX_INTERPEND;
10513 if (PL_lex_state == LEX_NORMAL) {
10514 if (ckWARN(WARN_AMBIGUOUS) &&
10515 (keyword(dest, d - dest) || get_cv(dest, FALSE)))
10517 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
10518 "Ambiguous use of %c{%s} resolved to %c%s",
10519 funny, dest, funny, dest);
10524 s = bracket; /* let the parser handle it */
10528 else if (PL_lex_state == LEX_INTERPNORMAL && !PL_lex_brackets && !intuit_more(s))
10529 PL_lex_state = LEX_INTERPEND;
10534 Perl_pmflag(pTHX_ U32* pmfl, int ch)
10536 PERL_UNUSED_CONTEXT;
10539 else if (ch == 'g')
10540 *pmfl |= PMf_GLOBAL;
10541 else if (ch == 'c')
10542 *pmfl |= PMf_CONTINUE;
10543 else if (ch == 'o')
10545 else if (ch == 'm')
10546 *pmfl |= PMf_MULTILINE;
10547 else if (ch == 's')
10548 *pmfl |= PMf_SINGLELINE;
10549 else if (ch == 'x')
10550 *pmfl |= PMf_EXTENDED;
10554 S_scan_pat(pTHX_ char *start, I32 type)
10558 char *s = scan_str(start,!!PL_madskills,FALSE);
10559 const char * const valid_flags = (type == OP_QR) ? "iomsx" : "iogcmsx";
10566 const char * const delimiter = skipspace(start);
10567 Perl_croak(aTHX_ *delimiter == '?'
10568 ? "Search pattern not terminated or ternary operator parsed as search pattern"
10569 : "Search pattern not terminated" );
10572 pm = (PMOP*)newPMOP(type, 0);
10573 if (PL_multi_open == '?')
10574 pm->op_pmflags |= PMf_ONCE;
10578 while (*s && strchr(valid_flags, *s))
10579 pmflag(&pm->op_pmflags,*s++);
10581 if (PL_madskills && modstart != s) {
10582 SV* tmptoken = newSVpvn(modstart, s - modstart);
10583 append_madprops(newMADPROP('m', MAD_SV, tmptoken, 0), (OP*)pm, 0);
10586 /* issue a warning if /c is specified,but /g is not */
10587 if ((pm->op_pmflags & PMf_CONTINUE) && !(pm->op_pmflags & PMf_GLOBAL)
10588 && ckWARN(WARN_REGEXP))
10590 Perl_warner(aTHX_ packWARN(WARN_REGEXP), "Use of /c modifier is meaningless without /g" );
10593 pm->op_pmpermflags = pm->op_pmflags;
10595 PL_lex_op = (OP*)pm;
10596 yylval.ival = OP_MATCH;
10601 S_scan_subst(pTHX_ char *start)
10612 yylval.ival = OP_NULL;
10614 s = scan_str(start,!!PL_madskills,FALSE);
10617 Perl_croak(aTHX_ "Substitution pattern not terminated");
10619 if (s[-1] == PL_multi_open)
10622 if (PL_madskills) {
10623 CURMAD('q', PL_thisopen);
10624 CURMAD('_', PL_thiswhite);
10625 CURMAD('E', PL_thisstuff);
10626 CURMAD('Q', PL_thisclose);
10627 PL_realtokenstart = s - SvPVX(PL_linestr);
10631 first_start = PL_multi_start;
10632 s = scan_str(s,!!PL_madskills,FALSE);
10634 if (PL_lex_stuff) {
10635 SvREFCNT_dec(PL_lex_stuff);
10636 PL_lex_stuff = NULL;
10638 Perl_croak(aTHX_ "Substitution replacement not terminated");
10640 PL_multi_start = first_start; /* so whole substitution is taken together */
10642 pm = (PMOP*)newPMOP(OP_SUBST, 0);
10645 if (PL_madskills) {
10646 CURMAD('z', PL_thisopen);
10647 CURMAD('R', PL_thisstuff);
10648 CURMAD('Z', PL_thisclose);
10658 else if (strchr("iogcmsx", *s))
10659 pmflag(&pm->op_pmflags,*s++);
10665 if (PL_madskills) {
10667 curmad('m', newSVpvn(modstart, s - modstart));
10668 append_madprops(PL_thismad, (OP*)pm, 0);
10672 if ((pm->op_pmflags & PMf_CONTINUE) && ckWARN(WARN_REGEXP)) {
10673 Perl_warner(aTHX_ packWARN(WARN_REGEXP), "Use of /c modifier is meaningless in s///" );
10677 SV * const repl = newSVpvs("");
10679 PL_sublex_info.super_bufptr = s;
10680 PL_sublex_info.super_bufend = PL_bufend;
10682 pm->op_pmflags |= PMf_EVAL;
10684 sv_catpv(repl, es ? "eval " : "do ");
10685 sv_catpvs(repl, "{");
10686 sv_catsv(repl, PL_lex_repl);
10687 sv_catpvs(repl, "}");
10689 SvREFCNT_dec(PL_lex_repl);
10690 PL_lex_repl = repl;
10693 pm->op_pmpermflags = pm->op_pmflags;
10694 PL_lex_op = (OP*)pm;
10695 yylval.ival = OP_SUBST;
10700 S_scan_trans(pTHX_ char *start)
10713 yylval.ival = OP_NULL;
10715 s = scan_str(start,!!PL_madskills,FALSE);
10717 Perl_croak(aTHX_ "Transliteration pattern not terminated");
10719 if (s[-1] == PL_multi_open)
10722 if (PL_madskills) {
10723 CURMAD('q', PL_thisopen);
10724 CURMAD('_', PL_thiswhite);
10725 CURMAD('E', PL_thisstuff);
10726 CURMAD('Q', PL_thisclose);
10727 PL_realtokenstart = s - SvPVX(PL_linestr);
10731 s = scan_str(s,!!PL_madskills,FALSE);
10733 if (PL_lex_stuff) {
10734 SvREFCNT_dec(PL_lex_stuff);
10735 PL_lex_stuff = NULL;
10737 Perl_croak(aTHX_ "Transliteration replacement not terminated");
10739 if (PL_madskills) {
10740 CURMAD('z', PL_thisopen);
10741 CURMAD('R', PL_thisstuff);
10742 CURMAD('Z', PL_thisclose);
10745 complement = del = squash = 0;
10752 complement = OPpTRANS_COMPLEMENT;
10755 del = OPpTRANS_DELETE;
10758 squash = OPpTRANS_SQUASH;
10767 Newx(tbl, complement&&!del?258:256, short);
10768 o = newPVOP(OP_TRANS, 0, (char*)tbl);
10769 o->op_private &= ~OPpTRANS_ALL;
10770 o->op_private |= del|squash|complement|
10771 (DO_UTF8(PL_lex_stuff)? OPpTRANS_FROM_UTF : 0)|
10772 (DO_UTF8(PL_lex_repl) ? OPpTRANS_TO_UTF : 0);
10775 yylval.ival = OP_TRANS;
10778 if (PL_madskills) {
10780 curmad('m', newSVpvn(modstart, s - modstart));
10781 append_madprops(PL_thismad, o, 0);
10790 S_scan_heredoc(pTHX_ register char *s)
10794 I32 op_type = OP_SCALAR;
10798 const char *found_newline;
10802 const int outer = (PL_rsfp && !(PL_lex_inwhat == OP_SCALAR));
10804 I32 stuffstart = s - SvPVX(PL_linestr);
10807 PL_realtokenstart = -1;
10812 e = PL_tokenbuf + sizeof PL_tokenbuf - 1;
10815 for (peek = s; SPACE_OR_TAB(*peek); peek++) ;
10816 if (*peek == '`' || *peek == '\'' || *peek =='"') {
10819 s = delimcpy(d, e, s, PL_bufend, term, &len);
10829 if (!isALNUM_lazy_if(s,UTF))
10830 deprecate_old("bare << to mean <<\"\"");
10831 for (; isALNUM_lazy_if(s,UTF); s++) {
10836 if (d >= PL_tokenbuf + sizeof PL_tokenbuf - 1)
10837 Perl_croak(aTHX_ "Delimiter for here document is too long");
10840 len = d - PL_tokenbuf;
10843 if (PL_madskills) {
10844 tstart = PL_tokenbuf + !outer;
10845 PL_thisclose = newSVpvn(tstart, len - !outer);
10846 tstart = SvPVX(PL_linestr) + stuffstart;
10847 PL_thisopen = newSVpvn(tstart, s - tstart);
10848 stuffstart = s - SvPVX(PL_linestr);
10851 #ifndef PERL_STRICT_CR
10852 d = strchr(s, '\r');
10854 char * const olds = s;
10856 while (s < PL_bufend) {
10862 else if (*s == '\n' && s[1] == '\r') { /* \015\013 on a mac? */
10871 SvCUR_set(PL_linestr, PL_bufend - SvPVX_const(PL_linestr));
10878 if ( outer || !(found_newline = memchr(s, '\n', PL_bufend - s)) ) {
10879 herewas = newSVpvn(s,PL_bufend-s);
10883 herewas = newSVpvn(s-1,found_newline-s+1);
10886 herewas = newSVpvn(s,found_newline-s);
10890 if (PL_madskills) {
10891 tstart = SvPVX(PL_linestr) + stuffstart;
10893 sv_catpvn(PL_thisstuff, tstart, s - tstart);
10895 PL_thisstuff = newSVpvn(tstart, s - tstart);
10898 s += SvCUR(herewas);
10901 stuffstart = s - SvPVX(PL_linestr);
10907 tmpstr = newSV(79);
10908 sv_upgrade(tmpstr, SVt_PVIV);
10909 if (term == '\'') {
10910 op_type = OP_CONST;
10911 SvIV_set(tmpstr, -1);
10913 else if (term == '`') {
10914 op_type = OP_BACKTICK;
10915 SvIV_set(tmpstr, '\\');
10919 PL_multi_start = CopLINE(PL_curcop);
10920 PL_multi_open = PL_multi_close = '<';
10921 term = *PL_tokenbuf;
10922 if (PL_lex_inwhat == OP_SUBST && PL_in_eval && !PL_rsfp) {
10923 char * const bufptr = PL_sublex_info.super_bufptr;
10924 char * const bufend = PL_sublex_info.super_bufend;
10925 char * const olds = s - SvCUR(herewas);
10926 s = strchr(bufptr, '\n');
10930 while (s < bufend &&
10931 (*s != term || memNE(s,PL_tokenbuf,len)) ) {
10933 CopLINE_inc(PL_curcop);
10936 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
10937 missingterm(PL_tokenbuf);
10939 sv_setpvn(herewas,bufptr,d-bufptr+1);
10940 sv_setpvn(tmpstr,d+1,s-d);
10942 sv_catpvn(herewas,s,bufend-s);
10943 Copy(SvPVX_const(herewas),bufptr,SvCUR(herewas) + 1,char);
10950 while (s < PL_bufend &&
10951 (*s != term || memNE(s,PL_tokenbuf,len)) ) {
10953 CopLINE_inc(PL_curcop);
10955 if (s >= PL_bufend) {
10956 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
10957 missingterm(PL_tokenbuf);
10959 sv_setpvn(tmpstr,d+1,s-d);
10961 if (PL_madskills) {
10963 sv_catpvn(PL_thisstuff, d + 1, s - d);
10965 PL_thisstuff = newSVpvn(d + 1, s - d);
10966 stuffstart = s - SvPVX(PL_linestr);
10970 CopLINE_inc(PL_curcop); /* the preceding stmt passes a newline */
10972 sv_catpvn(herewas,s,PL_bufend-s);
10973 sv_setsv(PL_linestr,herewas);
10974 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = s = PL_linestart = SvPVX(PL_linestr);
10975 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
10976 PL_last_lop = PL_last_uni = NULL;
10979 sv_setpvn(tmpstr,"",0); /* avoid "uninitialized" warning */
10980 while (s >= PL_bufend) { /* multiple line string? */
10982 if (PL_madskills) {
10983 tstart = SvPVX(PL_linestr) + stuffstart;
10985 sv_catpvn(PL_thisstuff, tstart, PL_bufend - tstart);
10987 PL_thisstuff = newSVpvn(tstart, PL_bufend - tstart);
10991 !(PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = filter_gets(PL_linestr, PL_rsfp, 0))) {
10992 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
10993 missingterm(PL_tokenbuf);
10996 stuffstart = s - SvPVX(PL_linestr);
10998 CopLINE_inc(PL_curcop);
10999 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
11000 PL_last_lop = PL_last_uni = NULL;
11001 #ifndef PERL_STRICT_CR
11002 if (PL_bufend - PL_linestart >= 2) {
11003 if ((PL_bufend[-2] == '\r' && PL_bufend[-1] == '\n') ||
11004 (PL_bufend[-2] == '\n' && PL_bufend[-1] == '\r'))
11006 PL_bufend[-2] = '\n';
11008 SvCUR_set(PL_linestr, PL_bufend - SvPVX_const(PL_linestr));
11010 else if (PL_bufend[-1] == '\r')
11011 PL_bufend[-1] = '\n';
11013 else if (PL_bufend - PL_linestart == 1 && PL_bufend[-1] == '\r')
11014 PL_bufend[-1] = '\n';
11016 if (PERLDB_LINE && PL_curstash != PL_debstash) {
11017 SV * const sv = newSV(0);
11019 sv_upgrade(sv, SVt_PVMG);
11020 sv_setsv(sv,PL_linestr);
11021 (void)SvIOK_on(sv);
11023 av_store(CopFILEAVx(PL_curcop), (I32)CopLINE(PL_curcop),sv);
11025 if (*s == term && memEQ(s,PL_tokenbuf,len)) {
11026 STRLEN off = PL_bufend - 1 - SvPVX_const(PL_linestr);
11027 *(SvPVX(PL_linestr) + off ) = ' ';
11028 sv_catsv(PL_linestr,herewas);
11029 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
11030 s = SvPVX(PL_linestr) + off; /* In case PV of PL_linestr moved. */
11034 sv_catsv(tmpstr,PL_linestr);
11039 PL_multi_end = CopLINE(PL_curcop);
11040 if (SvCUR(tmpstr) + 5 < SvLEN(tmpstr)) {
11041 SvPV_shrink_to_cur(tmpstr);
11043 SvREFCNT_dec(herewas);
11045 if (UTF && is_utf8_string((U8*)SvPVX_const(tmpstr), SvCUR(tmpstr)))
11047 else if (PL_encoding)
11048 sv_recode_to_utf8(tmpstr, PL_encoding);
11050 PL_lex_stuff = tmpstr;
11051 yylval.ival = op_type;
11055 /* scan_inputsymbol
11056 takes: current position in input buffer
11057 returns: new position in input buffer
11058 side-effects: yylval and lex_op are set.
11063 <FH> read from filehandle
11064 <pkg::FH> read from package qualified filehandle
11065 <pkg'FH> read from package qualified filehandle
11066 <$fh> read from filehandle in $fh
11067 <*.h> filename glob
11072 S_scan_inputsymbol(pTHX_ char *start)
11075 register char *s = start; /* current position in buffer */
11079 char *d = PL_tokenbuf; /* start of temp holding space */
11080 const char * const e = PL_tokenbuf + sizeof PL_tokenbuf; /* end of temp holding space */
11082 end = strchr(s, '\n');
11085 s = delimcpy(d, e, s + 1, end, '>', &len); /* extract until > */
11087 /* die if we didn't have space for the contents of the <>,
11088 or if it didn't end, or if we see a newline
11091 if (len >= (I32)sizeof PL_tokenbuf)
11092 Perl_croak(aTHX_ "Excessively long <> operator");
11094 Perl_croak(aTHX_ "Unterminated <> operator");
11099 Remember, only scalar variables are interpreted as filehandles by
11100 this code. Anything more complex (e.g., <$fh{$num}>) will be
11101 treated as a glob() call.
11102 This code makes use of the fact that except for the $ at the front,
11103 a scalar variable and a filehandle look the same.
11105 if (*d == '$' && d[1]) d++;
11107 /* allow <Pkg'VALUE> or <Pkg::VALUE> */
11108 while (*d && (isALNUM_lazy_if(d,UTF) || *d == '\'' || *d == ':'))
11111 /* If we've tried to read what we allow filehandles to look like, and
11112 there's still text left, then it must be a glob() and not a getline.
11113 Use scan_str to pull out the stuff between the <> and treat it
11114 as nothing more than a string.
11117 if (d - PL_tokenbuf != len) {
11118 yylval.ival = OP_GLOB;
11120 s = scan_str(start,!!PL_madskills,FALSE);
11122 Perl_croak(aTHX_ "Glob not terminated");
11126 bool readline_overriden = FALSE;
11129 /* we're in a filehandle read situation */
11132 /* turn <> into <ARGV> */
11134 Copy("ARGV",d,5,char);
11136 /* Check whether readline() is overriden */
11137 gv_readline = gv_fetchpvs("readline", GV_NOTQUAL, SVt_PVCV);
11139 && GvCVu(gv_readline) && GvIMPORTED_CV(gv_readline))
11141 ((gvp = (GV**)hv_fetchs(PL_globalstash, "readline", FALSE))
11142 && (gv_readline = *gvp) != (GV*)&PL_sv_undef
11143 && GvCVu(gv_readline) && GvIMPORTED_CV(gv_readline)))
11144 readline_overriden = TRUE;
11146 /* if <$fh>, create the ops to turn the variable into a
11152 /* try to find it in the pad for this block, otherwise find
11153 add symbol table ops
11155 if ((tmp = pad_findmy(d)) != NOT_IN_PAD) {
11156 if (PAD_COMPNAME_FLAGS_isOUR(tmp)) {
11157 HV * const stash = PAD_COMPNAME_OURSTASH(tmp);
11158 HEK * const stashname = HvNAME_HEK(stash);
11159 SV * const sym = sv_2mortal(newSVhek(stashname));
11160 sv_catpvs(sym, "::");
11161 sv_catpv(sym, d+1);
11166 OP * const o = newOP(OP_PADSV, 0);
11168 PL_lex_op = readline_overriden
11169 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
11170 append_elem(OP_LIST, o,
11171 newCVREF(0, newGVOP(OP_GV,0,gv_readline))))
11172 : (OP*)newUNOP(OP_READLINE, 0, o);
11181 ? (GV_ADDMULTI | GV_ADDINEVAL)
11184 PL_lex_op = readline_overriden
11185 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
11186 append_elem(OP_LIST,
11187 newUNOP(OP_RV2SV, 0, newGVOP(OP_GV, 0, gv)),
11188 newCVREF(0, newGVOP(OP_GV, 0, gv_readline))))
11189 : (OP*)newUNOP(OP_READLINE, 0,
11190 newUNOP(OP_RV2SV, 0,
11191 newGVOP(OP_GV, 0, gv)));
11193 if (!readline_overriden)
11194 PL_lex_op->op_flags |= OPf_SPECIAL;
11195 /* we created the ops in PL_lex_op, so make yylval.ival a null op */
11196 yylval.ival = OP_NULL;
11199 /* If it's none of the above, it must be a literal filehandle
11200 (<Foo::BAR> or <FOO>) so build a simple readline OP */
11202 GV * const gv = gv_fetchpv(d, GV_ADD, SVt_PVIO);
11203 PL_lex_op = readline_overriden
11204 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
11205 append_elem(OP_LIST,
11206 newGVOP(OP_GV, 0, gv),
11207 newCVREF(0, newGVOP(OP_GV, 0, gv_readline))))
11208 : (OP*)newUNOP(OP_READLINE, 0, newGVOP(OP_GV, 0, gv));
11209 yylval.ival = OP_NULL;
11218 takes: start position in buffer
11219 keep_quoted preserve \ on the embedded delimiter(s)
11220 keep_delims preserve the delimiters around the string
11221 returns: position to continue reading from buffer
11222 side-effects: multi_start, multi_close, lex_repl or lex_stuff, and
11223 updates the read buffer.
11225 This subroutine pulls a string out of the input. It is called for:
11226 q single quotes q(literal text)
11227 ' single quotes 'literal text'
11228 qq double quotes qq(interpolate $here please)
11229 " double quotes "interpolate $here please"
11230 qx backticks qx(/bin/ls -l)
11231 ` backticks `/bin/ls -l`
11232 qw quote words @EXPORT_OK = qw( func() $spam )
11233 m// regexp match m/this/
11234 s/// regexp substitute s/this/that/
11235 tr/// string transliterate tr/this/that/
11236 y/// string transliterate y/this/that/
11237 ($*@) sub prototypes sub foo ($)
11238 (stuff) sub attr parameters sub foo : attr(stuff)
11239 <> readline or globs <FOO>, <>, <$fh>, or <*.c>
11241 In most of these cases (all but <>, patterns and transliterate)
11242 yylex() calls scan_str(). m// makes yylex() call scan_pat() which
11243 calls scan_str(). s/// makes yylex() call scan_subst() which calls
11244 scan_str(). tr/// and y/// make yylex() call scan_trans() which
11247 It skips whitespace before the string starts, and treats the first
11248 character as the delimiter. If the delimiter is one of ([{< then
11249 the corresponding "close" character )]}> is used as the closing
11250 delimiter. It allows quoting of delimiters, and if the string has
11251 balanced delimiters ([{<>}]) it allows nesting.
11253 On success, the SV with the resulting string is put into lex_stuff or,
11254 if that is already non-NULL, into lex_repl. The second case occurs only
11255 when parsing the RHS of the special constructs s/// and tr/// (y///).
11256 For convenience, the terminating delimiter character is stuffed into
11261 S_scan_str(pTHX_ char *start, int keep_quoted, int keep_delims)
11264 SV *sv; /* scalar value: string */
11265 char *tmps; /* temp string, used for delimiter matching */
11266 register char *s = start; /* current position in the buffer */
11267 register char term; /* terminating character */
11268 register char *to; /* current position in the sv's data */
11269 I32 brackets = 1; /* bracket nesting level */
11270 bool has_utf8 = FALSE; /* is there any utf8 content? */
11271 I32 termcode; /* terminating char. code */
11272 U8 termstr[UTF8_MAXBYTES]; /* terminating string */
11273 STRLEN termlen; /* length of terminating string */
11274 char *last = NULL; /* last position for nesting bracket */
11280 /* skip space before the delimiter */
11286 if (PL_realtokenstart >= 0) {
11287 stuffstart = PL_realtokenstart;
11288 PL_realtokenstart = -1;
11291 stuffstart = start - SvPVX(PL_linestr);
11293 /* mark where we are, in case we need to report errors */
11296 /* after skipping whitespace, the next character is the terminator */
11299 termcode = termstr[0] = term;
11303 termcode = utf8_to_uvchr((U8*)s, &termlen);
11304 Copy(s, termstr, termlen, U8);
11305 if (!UTF8_IS_INVARIANT(term))
11309 /* mark where we are */
11310 PL_multi_start = CopLINE(PL_curcop);
11311 PL_multi_open = term;
11313 /* find corresponding closing delimiter */
11314 if (term && (tmps = strchr("([{< )]}> )]}>",term)))
11315 termcode = termstr[0] = term = tmps[5];
11317 PL_multi_close = term;
11319 /* create a new SV to hold the contents. 79 is the SV's initial length.
11320 What a random number. */
11322 sv_upgrade(sv, SVt_PVIV);
11323 SvIV_set(sv, termcode);
11324 (void)SvPOK_only(sv); /* validate pointer */
11326 /* move past delimiter and try to read a complete string */
11328 sv_catpvn(sv, s, termlen);
11331 tstart = SvPVX(PL_linestr) + stuffstart;
11332 if (!PL_thisopen && !keep_delims) {
11333 PL_thisopen = newSVpvn(tstart, s - tstart);
11334 stuffstart = s - SvPVX(PL_linestr);
11338 if (PL_encoding && !UTF) {
11342 int offset = s - SvPVX_const(PL_linestr);
11343 const bool found = sv_cat_decode(sv, PL_encoding, PL_linestr,
11344 &offset, (char*)termstr, termlen);
11345 const char * const ns = SvPVX_const(PL_linestr) + offset;
11346 char * const svlast = SvEND(sv) - 1;
11348 for (; s < ns; s++) {
11349 if (*s == '\n' && !PL_rsfp)
11350 CopLINE_inc(PL_curcop);
11353 goto read_more_line;
11355 /* handle quoted delimiters */
11356 if (SvCUR(sv) > 1 && *(svlast-1) == '\\') {
11358 for (t = svlast-2; t >= SvPVX_const(sv) && *t == '\\';)
11360 if ((svlast-1 - t) % 2) {
11361 if (!keep_quoted) {
11362 *(svlast-1) = term;
11364 SvCUR_set(sv, SvCUR(sv) - 1);
11369 if (PL_multi_open == PL_multi_close) {
11377 for (t = w = last; t < svlast; w++, t++) {
11378 /* At here, all closes are "was quoted" one,
11379 so we don't check PL_multi_close. */
11381 if (!keep_quoted && *(t+1) == PL_multi_open)
11386 else if (*t == PL_multi_open)
11394 SvCUR_set(sv, w - SvPVX_const(sv));
11397 if (--brackets <= 0)
11402 if (!keep_delims) {
11403 SvCUR_set(sv, SvCUR(sv) - 1);
11409 /* extend sv if need be */
11410 SvGROW(sv, SvCUR(sv) + (PL_bufend - s) + 1);
11411 /* set 'to' to the next character in the sv's string */
11412 to = SvPVX(sv)+SvCUR(sv);
11414 /* if open delimiter is the close delimiter read unbridle */
11415 if (PL_multi_open == PL_multi_close) {
11416 for (; s < PL_bufend; s++,to++) {
11417 /* embedded newlines increment the current line number */
11418 if (*s == '\n' && !PL_rsfp)
11419 CopLINE_inc(PL_curcop);
11420 /* handle quoted delimiters */
11421 if (*s == '\\' && s+1 < PL_bufend && term != '\\') {
11422 if (!keep_quoted && s[1] == term)
11424 /* any other quotes are simply copied straight through */
11428 /* terminate when run out of buffer (the for() condition), or
11429 have found the terminator */
11430 else if (*s == term) {
11433 if (s+termlen <= PL_bufend && memEQ(s, (char*)termstr, termlen))
11436 else if (!has_utf8 && !UTF8_IS_INVARIANT((U8)*s) && UTF)
11442 /* if the terminator isn't the same as the start character (e.g.,
11443 matched brackets), we have to allow more in the quoting, and
11444 be prepared for nested brackets.
11447 /* read until we run out of string, or we find the terminator */
11448 for (; s < PL_bufend; s++,to++) {
11449 /* embedded newlines increment the line count */
11450 if (*s == '\n' && !PL_rsfp)
11451 CopLINE_inc(PL_curcop);
11452 /* backslashes can escape the open or closing characters */
11453 if (*s == '\\' && s+1 < PL_bufend) {
11454 if (!keep_quoted &&
11455 ((s[1] == PL_multi_open) || (s[1] == PL_multi_close)))
11460 /* allow nested opens and closes */
11461 else if (*s == PL_multi_close && --brackets <= 0)
11463 else if (*s == PL_multi_open)
11465 else if (!has_utf8 && !UTF8_IS_INVARIANT((U8)*s) && UTF)
11470 /* terminate the copied string and update the sv's end-of-string */
11472 SvCUR_set(sv, to - SvPVX_const(sv));
11475 * this next chunk reads more into the buffer if we're not done yet
11479 break; /* handle case where we are done yet :-) */
11481 #ifndef PERL_STRICT_CR
11482 if (to - SvPVX_const(sv) >= 2) {
11483 if ((to[-2] == '\r' && to[-1] == '\n') ||
11484 (to[-2] == '\n' && to[-1] == '\r'))
11488 SvCUR_set(sv, to - SvPVX_const(sv));
11490 else if (to[-1] == '\r')
11493 else if (to - SvPVX_const(sv) == 1 && to[-1] == '\r')
11498 /* if we're out of file, or a read fails, bail and reset the current
11499 line marker so we can report where the unterminated string began
11502 if (PL_madskills) {
11503 char *tstart = SvPVX(PL_linestr) + stuffstart;
11505 sv_catpvn(PL_thisstuff, tstart, PL_bufend - tstart);
11507 PL_thisstuff = newSVpvn(tstart, PL_bufend - tstart);
11511 !(PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = filter_gets(PL_linestr, PL_rsfp, 0))) {
11513 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
11519 /* we read a line, so increment our line counter */
11520 CopLINE_inc(PL_curcop);
11522 /* update debugger info */
11523 if (PERLDB_LINE && PL_curstash != PL_debstash) {
11524 SV * const line_sv = newSV(0);
11526 sv_upgrade(line_sv, SVt_PVMG);
11527 sv_setsv(line_sv,PL_linestr);
11528 (void)SvIOK_on(line_sv);
11529 SvIV_set(line_sv, 0);
11530 av_store(CopFILEAVx(PL_curcop), (I32)CopLINE(PL_curcop), line_sv);
11533 /* having changed the buffer, we must update PL_bufend */
11534 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
11535 PL_last_lop = PL_last_uni = NULL;
11538 /* at this point, we have successfully read the delimited string */
11540 if (!PL_encoding || UTF) {
11542 if (PL_madskills) {
11543 char *tstart = SvPVX(PL_linestr) + stuffstart;
11545 sv_catpvn(PL_thisstuff, tstart, s - tstart);
11547 PL_thisstuff = newSVpvn(tstart, s - tstart);
11548 if (!PL_thisclose && !keep_delims)
11549 PL_thisclose = newSVpvn(s,termlen);
11554 sv_catpvn(sv, s, termlen);
11559 if (PL_madskills) {
11560 char *tstart = SvPVX(PL_linestr) + stuffstart;
11562 sv_catpvn(PL_thisstuff, tstart, s - tstart - termlen);
11564 PL_thisstuff = newSVpvn(tstart, s - tstart - termlen);
11565 if (!PL_thisclose && !keep_delims)
11566 PL_thisclose = newSVpvn(s - termlen,termlen);
11570 if (has_utf8 || PL_encoding)
11573 PL_multi_end = CopLINE(PL_curcop);
11575 /* if we allocated too much space, give some back */
11576 if (SvCUR(sv) + 5 < SvLEN(sv)) {
11577 SvLEN_set(sv, SvCUR(sv) + 1);
11578 SvPV_renew(sv, SvLEN(sv));
11581 /* decide whether this is the first or second quoted string we've read
11594 takes: pointer to position in buffer
11595 returns: pointer to new position in buffer
11596 side-effects: builds ops for the constant in yylval.op
11598 Read a number in any of the formats that Perl accepts:
11600 \d(_?\d)*(\.(\d(_?\d)*)?)?[Ee][\+\-]?(\d(_?\d)*) 12 12.34 12.
11601 \.\d(_?\d)*[Ee][\+\-]?(\d(_?\d)*) .34
11604 0x[0-9A-Fa-f](_?[0-9A-Fa-f])*
11606 Like most scan_ routines, it uses the PL_tokenbuf buffer to hold the
11609 If it reads a number without a decimal point or an exponent, it will
11610 try converting the number to an integer and see if it can do so
11611 without loss of precision.
11615 Perl_scan_num(pTHX_ const char *start, YYSTYPE* lvalp)
11618 register const char *s = start; /* current position in buffer */
11619 register char *d; /* destination in temp buffer */
11620 register char *e; /* end of temp buffer */
11621 NV nv; /* number read, as a double */
11622 SV *sv = NULL; /* place to put the converted number */
11623 bool floatit; /* boolean: int or float? */
11624 const char *lastub = NULL; /* position of last underbar */
11625 static char const number_too_long[] = "Number too long";
11627 /* We use the first character to decide what type of number this is */
11631 Perl_croak(aTHX_ "panic: scan_num");
11633 /* if it starts with a 0, it could be an octal number, a decimal in
11634 0.13 disguise, or a hexadecimal number, or a binary number. */
11638 u holds the "number so far"
11639 shift the power of 2 of the base
11640 (hex == 4, octal == 3, binary == 1)
11641 overflowed was the number more than we can hold?
11643 Shift is used when we add a digit. It also serves as an "are
11644 we in octal/hex/binary?" indicator to disallow hex characters
11645 when in octal mode.
11650 bool overflowed = FALSE;
11651 bool just_zero = TRUE; /* just plain 0 or binary number? */
11652 static const NV nvshift[5] = { 1.0, 2.0, 4.0, 8.0, 16.0 };
11653 static const char* const bases[5] =
11654 { "", "binary", "", "octal", "hexadecimal" };
11655 static const char* const Bases[5] =
11656 { "", "Binary", "", "Octal", "Hexadecimal" };
11657 static const char* const maxima[5] =
11659 "0b11111111111111111111111111111111",
11663 const char *base, *Base, *max;
11665 /* check for hex */
11670 } else if (s[1] == 'b') {
11675 /* check for a decimal in disguise */
11676 else if (s[1] == '.' || s[1] == 'e' || s[1] == 'E')
11678 /* so it must be octal */
11685 if (ckWARN(WARN_SYNTAX))
11686 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
11687 "Misplaced _ in number");
11691 base = bases[shift];
11692 Base = Bases[shift];
11693 max = maxima[shift];
11695 /* read the rest of the number */
11697 /* x is used in the overflow test,
11698 b is the digit we're adding on. */
11703 /* if we don't mention it, we're done */
11707 /* _ are ignored -- but warned about if consecutive */
11709 if (lastub && s == lastub + 1 && ckWARN(WARN_SYNTAX))
11710 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
11711 "Misplaced _ in number");
11715 /* 8 and 9 are not octal */
11716 case '8': case '9':
11718 yyerror(Perl_form(aTHX_ "Illegal octal digit '%c'", *s));
11722 case '2': case '3': case '4':
11723 case '5': case '6': case '7':
11725 yyerror(Perl_form(aTHX_ "Illegal binary digit '%c'", *s));
11728 case '0': case '1':
11729 b = *s++ & 15; /* ASCII digit -> value of digit */
11733 case 'a': case 'b': case 'c': case 'd': case 'e': case 'f':
11734 case 'A': case 'B': case 'C': case 'D': case 'E': case 'F':
11735 /* make sure they said 0x */
11738 b = (*s++ & 7) + 9;
11740 /* Prepare to put the digit we have onto the end
11741 of the number so far. We check for overflows.
11747 x = u << shift; /* make room for the digit */
11749 if ((x >> shift) != u
11750 && !(PL_hints & HINT_NEW_BINARY)) {
11753 if (ckWARN_d(WARN_OVERFLOW))
11754 Perl_warner(aTHX_ packWARN(WARN_OVERFLOW),
11755 "Integer overflow in %s number",
11758 u = x | b; /* add the digit to the end */
11761 n *= nvshift[shift];
11762 /* If an NV has not enough bits in its
11763 * mantissa to represent an UV this summing of
11764 * small low-order numbers is a waste of time
11765 * (because the NV cannot preserve the
11766 * low-order bits anyway): we could just
11767 * remember when did we overflow and in the
11768 * end just multiply n by the right
11776 /* if we get here, we had success: make a scalar value from
11781 /* final misplaced underbar check */
11782 if (s[-1] == '_') {
11783 if (ckWARN(WARN_SYNTAX))
11784 Perl_warner(aTHX_ packWARN(WARN_SYNTAX), "Misplaced _ in number");
11789 if (n > 4294967295.0 && ckWARN(WARN_PORTABLE))
11790 Perl_warner(aTHX_ packWARN(WARN_PORTABLE),
11791 "%s number > %s non-portable",
11797 if (u > 0xffffffff && ckWARN(WARN_PORTABLE))
11798 Perl_warner(aTHX_ packWARN(WARN_PORTABLE),
11799 "%s number > %s non-portable",
11804 if (just_zero && (PL_hints & HINT_NEW_INTEGER))
11805 sv = new_constant(start, s - start, "integer",
11807 else if (PL_hints & HINT_NEW_BINARY)
11808 sv = new_constant(start, s - start, "binary", sv, NULL, NULL);
11813 handle decimal numbers.
11814 we're also sent here when we read a 0 as the first digit
11816 case '1': case '2': case '3': case '4': case '5':
11817 case '6': case '7': case '8': case '9': case '.':
11820 e = PL_tokenbuf + sizeof PL_tokenbuf - 6; /* room for various punctuation */
11823 /* read next group of digits and _ and copy into d */
11824 while (isDIGIT(*s) || *s == '_') {
11825 /* skip underscores, checking for misplaced ones
11829 if (lastub && s == lastub + 1 && ckWARN(WARN_SYNTAX))
11830 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
11831 "Misplaced _ in number");
11835 /* check for end of fixed-length buffer */
11837 Perl_croak(aTHX_ number_too_long);
11838 /* if we're ok, copy the character */
11843 /* final misplaced underbar check */
11844 if (lastub && s == lastub + 1) {
11845 if (ckWARN(WARN_SYNTAX))
11846 Perl_warner(aTHX_ packWARN(WARN_SYNTAX), "Misplaced _ in number");
11849 /* read a decimal portion if there is one. avoid
11850 3..5 being interpreted as the number 3. followed
11853 if (*s == '.' && s[1] != '.') {
11858 if (ckWARN(WARN_SYNTAX))
11859 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
11860 "Misplaced _ in number");
11864 /* copy, ignoring underbars, until we run out of digits.
11866 for (; isDIGIT(*s) || *s == '_'; s++) {
11867 /* fixed length buffer check */
11869 Perl_croak(aTHX_ number_too_long);
11871 if (lastub && s == lastub + 1 && ckWARN(WARN_SYNTAX))
11872 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
11873 "Misplaced _ in number");
11879 /* fractional part ending in underbar? */
11880 if (s[-1] == '_') {
11881 if (ckWARN(WARN_SYNTAX))
11882 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
11883 "Misplaced _ in number");
11885 if (*s == '.' && isDIGIT(s[1])) {
11886 /* oops, it's really a v-string, but without the "v" */
11892 /* read exponent part, if present */
11893 if ((*s == 'e' || *s == 'E') && strchr("+-0123456789_", s[1])) {
11897 /* regardless of whether user said 3E5 or 3e5, use lower 'e' */
11898 *d++ = 'e'; /* At least some Mach atof()s don't grok 'E' */
11900 /* stray preinitial _ */
11902 if (ckWARN(WARN_SYNTAX))
11903 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
11904 "Misplaced _ in number");
11908 /* allow positive or negative exponent */
11909 if (*s == '+' || *s == '-')
11912 /* stray initial _ */
11914 if (ckWARN(WARN_SYNTAX))
11915 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
11916 "Misplaced _ in number");
11920 /* read digits of exponent */
11921 while (isDIGIT(*s) || *s == '_') {
11924 Perl_croak(aTHX_ number_too_long);
11928 if (((lastub && s == lastub + 1) ||
11929 (!isDIGIT(s[1]) && s[1] != '_'))
11930 && ckWARN(WARN_SYNTAX))
11931 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
11932 "Misplaced _ in number");
11939 /* make an sv from the string */
11943 We try to do an integer conversion first if no characters
11944 indicating "float" have been found.
11949 const int flags = grok_number (PL_tokenbuf, d - PL_tokenbuf, &uv);
11951 if (flags == IS_NUMBER_IN_UV) {
11953 sv_setiv(sv, uv); /* Prefer IVs over UVs. */
11956 } else if (flags == (IS_NUMBER_IN_UV | IS_NUMBER_NEG)) {
11957 if (uv <= (UV) IV_MIN)
11958 sv_setiv(sv, -(IV)uv);
11965 /* terminate the string */
11967 nv = Atof(PL_tokenbuf);
11971 if ( floatit ? (PL_hints & HINT_NEW_FLOAT) :
11972 (PL_hints & HINT_NEW_INTEGER) )
11973 sv = new_constant(PL_tokenbuf, d - PL_tokenbuf,
11974 (floatit ? "float" : "integer"),
11978 /* if it starts with a v, it could be a v-string */
11981 sv = newSV(5); /* preallocate storage space */
11982 s = scan_vstring(s,sv);
11986 /* make the op for the constant and return */
11989 lvalp->opval = newSVOP(OP_CONST, 0, sv);
11991 lvalp->opval = NULL;
11997 S_scan_formline(pTHX_ register char *s)
12000 register char *eol;
12002 SV * const stuff = newSVpvs("");
12003 bool needargs = FALSE;
12004 bool eofmt = FALSE;
12006 char *tokenstart = s;
12009 if (PL_madskills) {
12010 savewhite = PL_thiswhite;
12015 while (!needargs) {
12017 #ifdef PERL_STRICT_CR
12018 for (t = s+1;SPACE_OR_TAB(*t); t++) ;
12020 for (t = s+1;SPACE_OR_TAB(*t) || *t == '\r'; t++) ;
12022 if (*t == '\n' || t == PL_bufend) {
12027 if (PL_in_eval && !PL_rsfp) {
12028 eol = (char *) memchr(s,'\n',PL_bufend-s);
12033 eol = PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
12035 for (t = s; t < eol; t++) {
12036 if (*t == '~' && t[1] == '~' && SvCUR(stuff)) {
12038 goto enough; /* ~~ must be first line in formline */
12040 if (*t == '@' || *t == '^')
12044 sv_catpvn(stuff, s, eol-s);
12045 #ifndef PERL_STRICT_CR
12046 if (eol-s > 1 && eol[-2] == '\r' && eol[-1] == '\n') {
12047 char *end = SvPVX(stuff) + SvCUR(stuff);
12050 SvCUR_set(stuff, SvCUR(stuff) - 1);
12060 if (PL_madskills) {
12062 sv_catpvn(PL_thistoken, tokenstart, PL_bufend - tokenstart);
12064 PL_thistoken = newSVpvn(tokenstart, PL_bufend - tokenstart);
12067 s = filter_gets(PL_linestr, PL_rsfp, 0);
12069 tokenstart = PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = SvPVX(PL_linestr);
12071 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = SvPVX(PL_linestr);
12073 PL_bufend = PL_bufptr + SvCUR(PL_linestr);
12074 PL_last_lop = PL_last_uni = NULL;
12083 if (SvCUR(stuff)) {
12086 PL_lex_state = LEX_NORMAL;
12087 start_force(PL_curforce);
12088 NEXTVAL_NEXTTOKE.ival = 0;
12092 PL_lex_state = LEX_FORMLINE;
12094 if (UTF && is_utf8_string((U8*)SvPVX_const(stuff), SvCUR(stuff)))
12096 else if (PL_encoding)
12097 sv_recode_to_utf8(stuff, PL_encoding);
12099 start_force(PL_curforce);
12100 NEXTVAL_NEXTTOKE.opval = (OP*)newSVOP(OP_CONST, 0, stuff);
12102 start_force(PL_curforce);
12103 NEXTVAL_NEXTTOKE.ival = OP_FORMLINE;
12107 SvREFCNT_dec(stuff);
12109 PL_lex_formbrack = 0;
12113 if (PL_madskills) {
12115 sv_catpvn(PL_thistoken, tokenstart, s - tokenstart);
12117 PL_thistoken = newSVpvn(tokenstart, s - tokenstart);
12118 PL_thiswhite = savewhite;
12130 PL_cshlen = strlen(PL_cshname);
12132 #if defined(USE_ITHREADS)
12133 PERL_UNUSED_CONTEXT;
12139 Perl_start_subparse(pTHX_ I32 is_format, U32 flags)
12142 const I32 oldsavestack_ix = PL_savestack_ix;
12143 CV* const outsidecv = PL_compcv;
12146 assert(SvTYPE(PL_compcv) == SVt_PVCV);
12148 SAVEI32(PL_subline);
12149 save_item(PL_subname);
12150 SAVESPTR(PL_compcv);
12152 PL_compcv = (CV*)newSV(0);
12153 sv_upgrade((SV *)PL_compcv, is_format ? SVt_PVFM : SVt_PVCV);
12154 CvFLAGS(PL_compcv) |= flags;
12156 PL_subline = CopLINE(PL_curcop);
12157 CvPADLIST(PL_compcv) = pad_new(padnew_SAVE|padnew_SAVESUB);
12158 CvOUTSIDE(PL_compcv) = (CV*)SvREFCNT_inc_simple(outsidecv);
12159 CvOUTSIDE_SEQ(PL_compcv) = PL_cop_seqmax;
12161 return oldsavestack_ix;
12165 #pragma segment Perl_yylex
12168 Perl_yywarn(pTHX_ const char *s)
12171 PL_in_eval |= EVAL_WARNONLY;
12173 PL_in_eval &= ~EVAL_WARNONLY;
12178 Perl_yyerror(pTHX_ const char *s)
12181 const char *where = NULL;
12182 const char *context = NULL;
12186 if (!yychar || (yychar == ';' && !PL_rsfp))
12188 else if (PL_oldoldbufptr && PL_bufptr > PL_oldoldbufptr &&
12189 PL_bufptr - PL_oldoldbufptr < 200 && PL_oldoldbufptr != PL_oldbufptr &&
12190 PL_oldbufptr != PL_bufptr) {
12193 The code below is removed for NetWare because it abends/crashes on NetWare
12194 when the script has error such as not having the closing quotes like:
12195 if ($var eq "value)
12196 Checking of white spaces is anyway done in NetWare code.
12199 while (isSPACE(*PL_oldoldbufptr))
12202 context = PL_oldoldbufptr;
12203 contlen = PL_bufptr - PL_oldoldbufptr;
12205 else if (PL_oldbufptr && PL_bufptr > PL_oldbufptr &&
12206 PL_bufptr - PL_oldbufptr < 200 && PL_oldbufptr != PL_bufptr) {
12209 The code below is removed for NetWare because it abends/crashes on NetWare
12210 when the script has error such as not having the closing quotes like:
12211 if ($var eq "value)
12212 Checking of white spaces is anyway done in NetWare code.
12215 while (isSPACE(*PL_oldbufptr))
12218 context = PL_oldbufptr;
12219 contlen = PL_bufptr - PL_oldbufptr;
12221 else if (yychar > 255)
12222 where = "next token ???";
12223 else if (yychar == -2) { /* YYEMPTY */
12224 if (PL_lex_state == LEX_NORMAL ||
12225 (PL_lex_state == LEX_KNOWNEXT && PL_lex_defer == LEX_NORMAL))
12226 where = "at end of line";
12227 else if (PL_lex_inpat)
12228 where = "within pattern";
12230 where = "within string";
12233 SV * const where_sv = sv_2mortal(newSVpvs("next char "));
12235 Perl_sv_catpvf(aTHX_ where_sv, "^%c", toCTRL(yychar));
12236 else if (isPRINT_LC(yychar))
12237 Perl_sv_catpvf(aTHX_ where_sv, "%c", yychar);
12239 Perl_sv_catpvf(aTHX_ where_sv, "\\%03o", yychar & 255);
12240 where = SvPVX_const(where_sv);
12242 msg = sv_2mortal(newSVpv(s, 0));
12243 Perl_sv_catpvf(aTHX_ msg, " at %s line %"IVdf", ",
12244 OutCopFILE(PL_curcop), (IV)CopLINE(PL_curcop));
12246 Perl_sv_catpvf(aTHX_ msg, "near \"%.*s\"\n", contlen, context);
12248 Perl_sv_catpvf(aTHX_ msg, "%s\n", where);
12249 if (PL_multi_start < PL_multi_end && (U32)(CopLINE(PL_curcop) - PL_multi_end) <= 1) {
12250 Perl_sv_catpvf(aTHX_ msg,
12251 " (Might be a runaway multi-line %c%c string starting on line %"IVdf")\n",
12252 (int)PL_multi_open,(int)PL_multi_close,(IV)PL_multi_start);
12255 if (PL_in_eval & EVAL_WARNONLY && ckWARN_d(WARN_SYNTAX))
12256 Perl_warner(aTHX_ packWARN(WARN_SYNTAX), "%"SVf, msg);
12259 if (PL_error_count >= 10) {
12260 if (PL_in_eval && SvCUR(ERRSV))
12261 Perl_croak(aTHX_ "%"SVf"%s has too many errors.\n",
12262 ERRSV, OutCopFILE(PL_curcop));
12264 Perl_croak(aTHX_ "%s has too many errors.\n",
12265 OutCopFILE(PL_curcop));
12268 PL_in_my_stash = NULL;
12272 #pragma segment Main
12276 S_swallow_bom(pTHX_ U8 *s)
12279 const STRLEN slen = SvCUR(PL_linestr);
12282 if (s[1] == 0xFE) {
12283 /* UTF-16 little-endian? (or UTF32-LE?) */
12284 if (s[2] == 0 && s[3] == 0) /* UTF-32 little-endian */
12285 Perl_croak(aTHX_ "Unsupported script encoding UTF32-LE");
12286 #ifndef PERL_NO_UTF16_FILTER
12287 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF16-LE script encoding (BOM)\n");
12290 if (PL_bufend > (char*)s) {
12294 filter_add(utf16rev_textfilter, NULL);
12295 Newx(news, (PL_bufend - (char*)s) * 3 / 2 + 1, U8);
12296 utf16_to_utf8_reversed(s, news,
12297 PL_bufend - (char*)s - 1,
12299 sv_setpvn(PL_linestr, (const char*)news, newlen);
12301 s = (U8*)SvPVX(PL_linestr);
12302 Copy(news, s, newlen, U8);
12306 SvUTF8_on(PL_linestr);
12307 s = (U8*)SvPVX(PL_linestr);
12309 /* FIXME - is this a general bug fix? */
12312 PL_bufend = SvPVX(PL_linestr) + newlen;
12315 Perl_croak(aTHX_ "Unsupported script encoding UTF16-LE");
12320 if (s[1] == 0xFF) { /* UTF-16 big-endian? */
12321 #ifndef PERL_NO_UTF16_FILTER
12322 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16BE script encoding (BOM)\n");
12325 if (PL_bufend > (char *)s) {
12329 filter_add(utf16_textfilter, NULL);
12330 Newx(news, (PL_bufend - (char*)s) * 3 / 2 + 1, U8);
12331 utf16_to_utf8(s, news,
12332 PL_bufend - (char*)s,
12334 sv_setpvn(PL_linestr, (const char*)news, newlen);
12336 SvUTF8_on(PL_linestr);
12337 s = (U8*)SvPVX(PL_linestr);
12338 PL_bufend = SvPVX(PL_linestr) + newlen;
12341 Perl_croak(aTHX_ "Unsupported script encoding UTF16-BE");
12346 if (slen > 2 && s[1] == 0xBB && s[2] == 0xBF) {
12347 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-8 script encoding (BOM)\n");
12348 s += 3; /* UTF-8 */
12354 if (s[2] == 0xFE && s[3] == 0xFF) {
12355 /* UTF-32 big-endian */
12356 Perl_croak(aTHX_ "Unsupported script encoding UTF32-BE");
12359 else if (s[2] == 0 && s[3] != 0) {
12362 * are a good indicator of UTF-16BE. */
12363 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16BE script encoding (no BOM)\n");
12368 if (slen > 3 && s[1] == 0 && s[2] != 0 && s[3] == 0) {
12371 * are a good indicator of UTF-16LE. */
12372 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16LE script encoding (no BOM)\n");
12381 * Restore a source filter.
12385 restore_rsfp(pTHX_ void *f)
12388 PerlIO * const fp = (PerlIO*)f;
12390 if (PL_rsfp == PerlIO_stdin())
12391 PerlIO_clearerr(PL_rsfp);
12392 else if (PL_rsfp && (PL_rsfp != fp))
12393 PerlIO_close(PL_rsfp);
12397 #ifndef PERL_NO_UTF16_FILTER
12399 utf16_textfilter(pTHX_ int idx, SV *sv, int maxlen)
12402 const STRLEN old = SvCUR(sv);
12403 const I32 count = FILTER_READ(idx+1, sv, maxlen);
12404 DEBUG_P(PerlIO_printf(Perl_debug_log,
12405 "utf16_textfilter(%p): %d %d (%d)\n",
12406 utf16_textfilter, idx, maxlen, (int) count));
12410 Newx(tmps, SvCUR(sv) * 3 / 2 + 1, U8);
12411 Copy(SvPVX_const(sv), tmps, old, char);
12412 utf16_to_utf8((U8*)SvPVX_const(sv) + old, tmps + old,
12413 SvCUR(sv) - old, &newlen);
12414 sv_usepvn(sv, (char*)tmps, (STRLEN)newlen + old);
12416 DEBUG_P({sv_dump(sv);});
12421 utf16rev_textfilter(pTHX_ int idx, SV *sv, int maxlen)
12424 const STRLEN old = SvCUR(sv);
12425 const I32 count = FILTER_READ(idx+1, sv, maxlen);
12426 DEBUG_P(PerlIO_printf(Perl_debug_log,
12427 "utf16rev_textfilter(%p): %d %d (%d)\n",
12428 utf16rev_textfilter, idx, maxlen, (int) count));
12432 Newx(tmps, SvCUR(sv) * 3 / 2 + 1, U8);
12433 Copy(SvPVX_const(sv), tmps, old, char);
12434 utf16_to_utf8((U8*)SvPVX_const(sv) + old, tmps + old,
12435 SvCUR(sv) - old, &newlen);
12436 sv_usepvn(sv, (char*)tmps, (STRLEN)newlen + old);
12438 DEBUG_P({ sv_dump(sv); });
12444 Returns a pointer to the next character after the parsed
12445 vstring, as well as updating the passed in sv.
12447 Function must be called like
12450 s = scan_vstring(s,sv);
12452 The sv should already be large enough to store the vstring
12453 passed in, for performance reasons.
12458 Perl_scan_vstring(pTHX_ const char *s, SV *sv)
12461 const char *pos = s;
12462 const char *start = s;
12463 if (*pos == 'v') pos++; /* get past 'v' */
12464 while (pos < PL_bufend && (isDIGIT(*pos) || *pos == '_'))
12466 if ( *pos != '.') {
12467 /* this may not be a v-string if followed by => */
12468 const char *next = pos;
12469 while (next < PL_bufend && isSPACE(*next))
12471 if ((PL_bufend - next) >= 2 && *next == '=' && next[1] == '>' ) {
12472 /* return string not v-string */
12473 sv_setpvn(sv,(char *)s,pos-s);
12474 return (char *)pos;
12478 if (!isALPHA(*pos)) {
12479 U8 tmpbuf[UTF8_MAXBYTES+1];
12482 s++; /* get past 'v' */
12484 sv_setpvn(sv, "", 0);
12487 /* this is atoi() that tolerates underscores */
12490 const char *end = pos;
12492 while (--end >= s) {
12494 const UV orev = rev;
12495 rev += (*end - '0') * mult;
12497 if (orev > rev && ckWARN_d(WARN_OVERFLOW))
12498 Perl_warner(aTHX_ packWARN(WARN_OVERFLOW),
12499 "Integer overflow in decimal number");
12503 if (rev > 0x7FFFFFFF)
12504 Perl_croak(aTHX_ "In EBCDIC the v-string components cannot exceed 2147483647");
12506 /* Append native character for the rev point */
12507 tmpend = uvchr_to_utf8(tmpbuf, rev);
12508 sv_catpvn(sv, (const char*)tmpbuf, tmpend - tmpbuf);
12509 if (!UNI_IS_INVARIANT(NATIVE_TO_UNI(rev)))
12511 if (pos + 1 < PL_bufend && *pos == '.' && isDIGIT(pos[1]))
12517 while (pos < PL_bufend && (isDIGIT(*pos) || *pos == '_'))
12521 sv_magic(sv,NULL,PERL_MAGIC_vstring,(const char*)start, pos-start);
12529 * c-indentation-style: bsd
12530 * c-basic-offset: 4
12531 * indent-tabs-mode: t
12534 * ex: set ts=8 sts=4 sw=4 noet: