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) /*EMPTY*/
1198 # define curmad(slot, sv) /*EMPTY*/
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)
2637 if (!PL_rsfp_filters)
2639 if (idx > AvFILLp(PL_rsfp_filters)) { /* Any more filters? */
2640 /* Provide a default input filter to make life easy. */
2641 /* Note that we append to the line. This is handy. */
2642 DEBUG_P(PerlIO_printf(Perl_debug_log,
2643 "filter_read %d: from rsfp\n", idx));
2647 const int old_len = SvCUR(buf_sv);
2649 /* ensure buf_sv is large enough */
2650 SvGROW(buf_sv, (STRLEN)(old_len + maxlen)) ;
2651 if ((len = PerlIO_read(PL_rsfp, SvPVX(buf_sv) + old_len, maxlen)) <= 0){
2652 if (PerlIO_error(PL_rsfp))
2653 return -1; /* error */
2655 return 0 ; /* end of file */
2657 SvCUR_set(buf_sv, old_len + len) ;
2660 if (sv_gets(buf_sv, PL_rsfp, SvCUR(buf_sv)) == NULL) {
2661 if (PerlIO_error(PL_rsfp))
2662 return -1; /* error */
2664 return 0 ; /* end of file */
2667 return SvCUR(buf_sv);
2669 /* Skip this filter slot if filter has been deleted */
2670 if ( (datasv = FILTER_DATA(idx)) == &PL_sv_undef) {
2671 DEBUG_P(PerlIO_printf(Perl_debug_log,
2672 "filter_read %d: skipped (filter deleted)\n",
2674 return FILTER_READ(idx+1, buf_sv, maxlen); /* recurse */
2676 /* Get function pointer hidden within datasv */
2677 funcp = DPTR2FPTR(filter_t, IoANY(datasv));
2678 DEBUG_P(PerlIO_printf(Perl_debug_log,
2679 "filter_read %d: via function %p (%s)\n",
2680 idx, datasv, SvPV_nolen_const(datasv)));
2681 /* Call function. The function is expected to */
2682 /* call "FILTER_READ(idx+1, buf_sv)" first. */
2683 /* Return: <0:error, =0:eof, >0:not eof */
2684 return (*funcp)(aTHX_ idx, buf_sv, maxlen);
2688 S_filter_gets(pTHX_ register SV *sv, register PerlIO *fp, STRLEN append)
2691 #ifdef PERL_CR_FILTER
2692 if (!PL_rsfp_filters) {
2693 filter_add(S_cr_textfilter,NULL);
2696 if (PL_rsfp_filters) {
2698 SvCUR_set(sv, 0); /* start with empty line */
2699 if (FILTER_READ(0, sv, 0) > 0)
2700 return ( SvPVX(sv) ) ;
2705 return (sv_gets(sv, fp, append));
2709 S_find_in_my_stash(pTHX_ const char *pkgname, I32 len)
2714 if (len == 11 && *pkgname == '_' && strEQ(pkgname, "__PACKAGE__"))
2718 (pkgname[len - 2] == ':' && pkgname[len - 1] == ':') &&
2719 (gv = gv_fetchpvn_flags(pkgname, len, 0, SVt_PVHV)))
2721 return GvHV(gv); /* Foo:: */
2724 /* use constant CLASS => 'MyClass' */
2725 if ((gv = gv_fetchpvn_flags(pkgname, len, 0, SVt_PVCV))) {
2727 if (GvCV(gv) && (sv = cv_const_sv(GvCV(gv)))) {
2728 pkgname = SvPV_nolen_const(sv);
2732 return gv_stashpv(pkgname, FALSE);
2738 * The intent of this yylex wrapper is to minimize the changes to the
2739 * tokener when we aren't interested in collecting madprops. It remains
2740 * to be seen how successful this strategy will be...
2747 char *s = PL_bufptr;
2749 /* make sure PL_thiswhite is initialized */
2753 /* just do what yylex would do on pending identifier; leave PL_thiswhite alone */
2754 if (PL_pending_ident)
2755 return S_pending_ident(aTHX);
2757 /* previous token ate up our whitespace? */
2758 if (!PL_lasttoke && PL_nextwhite) {
2759 PL_thiswhite = PL_nextwhite;
2763 /* isolate the token, and figure out where it is without whitespace */
2764 PL_realtokenstart = -1;
2768 assert(PL_curforce < 0);
2770 if (!PL_thismad || PL_thismad->mad_key == '^') { /* not forced already? */
2771 if (!PL_thistoken) {
2772 if (PL_realtokenstart < 0 || !CopLINE(PL_curcop))
2773 PL_thistoken = newSVpvn("",0);
2775 char *tstart = SvPVX(PL_linestr) + PL_realtokenstart;
2776 PL_thistoken = newSVpvn(tstart, s - tstart);
2779 if (PL_thismad) /* install head */
2780 CURMAD('X', PL_thistoken);
2783 /* last whitespace of a sublex? */
2784 if (optype == ')' && PL_endwhite) {
2785 CURMAD('X', PL_endwhite);
2790 /* if no whitespace and we're at EOF, bail. Otherwise fake EOF below. */
2791 if (!PL_thiswhite && !PL_endwhite && !optype) {
2792 sv_free(PL_thistoken);
2797 /* put off final whitespace till peg */
2798 if (optype == ';' && !PL_rsfp) {
2799 PL_nextwhite = PL_thiswhite;
2802 else if (PL_thisopen) {
2803 CURMAD('q', PL_thisopen);
2805 sv_free(PL_thistoken);
2809 /* Store actual token text as madprop X */
2810 CURMAD('X', PL_thistoken);
2814 /* add preceding whitespace as madprop _ */
2815 CURMAD('_', PL_thiswhite);
2819 /* add quoted material as madprop = */
2820 CURMAD('=', PL_thisstuff);
2824 /* add terminating quote as madprop Q */
2825 CURMAD('Q', PL_thisclose);
2829 /* special processing based on optype */
2833 /* opval doesn't need a TOKEN since it can already store mp */
2844 append_madprops(PL_thismad, yylval.opval, 0);
2852 addmad(newMADsv('p', PL_endwhite), &PL_thismad, 0);
2861 /* remember any fake bracket that lexer is about to discard */
2862 if (PL_lex_brackets == 1 &&
2863 ((expectation)PL_lex_brackstack[0] & XFAKEBRACK))
2866 while (s < PL_bufend && (*s == ' ' || *s == '\t'))
2869 PL_thiswhite = newSVpvn(PL_bufptr, ++s - PL_bufptr);
2870 addmad(newMADsv('#', PL_thiswhite), &PL_thismad, 0);
2873 break; /* don't bother looking for trailing comment */
2882 /* attach a trailing comment to its statement instead of next token */
2886 if (PL_bufptr > PL_oldbufptr && PL_bufptr[-1] == optype) {
2888 while (s < PL_bufend && (*s == ' ' || *s == '\t'))
2890 if (*s == '\n' || *s == '#') {
2891 while (s < PL_bufend && *s != '\n')
2895 PL_thiswhite = newSVpvn(PL_bufptr, s - PL_bufptr);
2896 addmad(newMADsv('#', PL_thiswhite), &PL_thismad, 0);
2913 /* Create new token struct. Note: opvals return early above. */
2914 yylval.tkval = newTOKEN(optype, yylval, PL_thismad);
2921 S_tokenize_use(pTHX_ int is_use, char *s) {
2923 if (PL_expect != XSTATE)
2924 yyerror(Perl_form(aTHX_ "\"%s\" not allowed in expression",
2925 is_use ? "use" : "no"));
2927 if (isDIGIT(*s) || (*s == 'v' && isDIGIT(s[1]))) {
2928 s = force_version(s, TRUE);
2929 if (*s == ';' || (s = SKIPSPACE1(s), *s == ';')) {
2930 start_force(PL_curforce);
2931 NEXTVAL_NEXTTOKE.opval = NULL;
2934 else if (*s == 'v') {
2935 s = force_word(s,WORD,FALSE,TRUE,FALSE);
2936 s = force_version(s, FALSE);
2940 s = force_word(s,WORD,FALSE,TRUE,FALSE);
2941 s = force_version(s, FALSE);
2943 yylval.ival = is_use;
2947 static const char* const exp_name[] =
2948 { "OPERATOR", "TERM", "REF", "STATE", "BLOCK", "ATTRBLOCK",
2949 "ATTRTERM", "TERMBLOCK", "TERMORDORDOR"
2956 Works out what to call the token just pulled out of the input
2957 stream. The yacc parser takes care of taking the ops we return and
2958 stitching them into a tree.
2964 if read an identifier
2965 if we're in a my declaration
2966 croak if they tried to say my($foo::bar)
2967 build the ops for a my() declaration
2968 if it's an access to a my() variable
2969 are we in a sort block?
2970 croak if my($a); $a <=> $b
2971 build ops for access to a my() variable
2972 if in a dq string, and they've said @foo and we can't find @foo
2974 build ops for a bareword
2975 if we already built the token before, use it.
2980 #pragma segment Perl_yylex
2986 register char *s = PL_bufptr;
2992 SV* tmp = newSVpvs("");
2993 PerlIO_printf(Perl_debug_log, "### %"IVdf":LEX_%s/X%s %s\n",
2994 (IV)CopLINE(PL_curcop),
2995 lex_state_names[PL_lex_state],
2996 exp_name[PL_expect],
2997 pv_display(tmp, s, strlen(s), 0, 60));
3000 /* check if there's an identifier for us to look at */
3001 if (PL_pending_ident)
3002 return REPORT(S_pending_ident(aTHX));
3004 /* no identifier pending identification */
3006 switch (PL_lex_state) {
3008 case LEX_NORMAL: /* Some compilers will produce faster */
3009 case LEX_INTERPNORMAL: /* code if we comment these out. */
3013 /* when we've already built the next token, just pull it out of the queue */
3017 yylval = PL_nexttoke[PL_lasttoke].next_val;
3019 PL_thismad = PL_nexttoke[PL_lasttoke].next_mad;
3020 PL_nexttoke[PL_lasttoke].next_mad = 0;
3021 if (PL_thismad && PL_thismad->mad_key == '_') {
3022 PL_thiswhite = (SV*)PL_thismad->mad_val;
3023 PL_thismad->mad_val = 0;
3024 mad_free(PL_thismad);
3029 PL_lex_state = PL_lex_defer;
3030 PL_expect = PL_lex_expect;
3031 PL_lex_defer = LEX_NORMAL;
3032 if (!PL_nexttoke[PL_lasttoke].next_type)
3037 yylval = PL_nextval[PL_nexttoke];
3039 PL_lex_state = PL_lex_defer;
3040 PL_expect = PL_lex_expect;
3041 PL_lex_defer = LEX_NORMAL;
3045 /* FIXME - can these be merged? */
3046 return(PL_nexttoke[PL_lasttoke].next_type);
3048 return REPORT(PL_nexttype[PL_nexttoke]);
3051 /* interpolated case modifiers like \L \U, including \Q and \E.
3052 when we get here, PL_bufptr is at the \
3054 case LEX_INTERPCASEMOD:
3056 if (PL_bufptr != PL_bufend && *PL_bufptr != '\\')
3057 Perl_croak(aTHX_ "panic: INTERPCASEMOD");
3059 /* handle \E or end of string */
3060 if (PL_bufptr == PL_bufend || PL_bufptr[1] == 'E') {
3062 if (PL_lex_casemods) {
3063 const char oldmod = PL_lex_casestack[--PL_lex_casemods];
3064 PL_lex_casestack[PL_lex_casemods] = '\0';
3066 if (PL_bufptr != PL_bufend
3067 && (oldmod == 'L' || oldmod == 'U' || oldmod == 'Q')) {
3069 PL_lex_state = LEX_INTERPCONCAT;
3072 PL_thistoken = newSVpvn("\\E",2);
3078 while (PL_bufptr != PL_bufend &&
3079 PL_bufptr[0] == '\\' && PL_bufptr[1] == 'E') {
3081 PL_thiswhite = newSVpvn("",0);
3082 sv_catpvn(PL_thiswhite, PL_bufptr, 2);
3086 if (PL_bufptr != PL_bufend)
3089 PL_lex_state = LEX_INTERPCONCAT;
3093 DEBUG_T({ PerlIO_printf(Perl_debug_log,
3094 "### Saw case modifier\n"); });
3096 if (s[1] == '\\' && s[2] == 'E') {
3099 PL_thiswhite = newSVpvn("",0);
3100 sv_catpvn(PL_thiswhite, PL_bufptr, 4);
3103 PL_lex_state = LEX_INTERPCONCAT;
3108 if (!PL_madskills) /* when just compiling don't need correct */
3109 if (strnEQ(s, "L\\u", 3) || strnEQ(s, "U\\l", 3))
3110 tmp = *s, *s = s[2], s[2] = (char)tmp; /* misordered... */
3111 if ((*s == 'L' || *s == 'U') &&
3112 (strchr(PL_lex_casestack, 'L') || strchr(PL_lex_casestack, 'U'))) {
3113 PL_lex_casestack[--PL_lex_casemods] = '\0';
3116 if (PL_lex_casemods > 10)
3117 Renew(PL_lex_casestack, PL_lex_casemods + 2, char);
3118 PL_lex_casestack[PL_lex_casemods++] = *s;
3119 PL_lex_casestack[PL_lex_casemods] = '\0';
3120 PL_lex_state = LEX_INTERPCONCAT;
3121 start_force(PL_curforce);
3122 NEXTVAL_NEXTTOKE.ival = 0;
3124 start_force(PL_curforce);
3126 NEXTVAL_NEXTTOKE.ival = OP_LCFIRST;
3128 NEXTVAL_NEXTTOKE.ival = OP_UCFIRST;
3130 NEXTVAL_NEXTTOKE.ival = OP_LC;
3132 NEXTVAL_NEXTTOKE.ival = OP_UC;
3134 NEXTVAL_NEXTTOKE.ival = OP_QUOTEMETA;
3136 Perl_croak(aTHX_ "panic: yylex");
3138 SV* const tmpsv = newSVpvn("",0);
3139 Perl_sv_catpvf(aTHX_ tmpsv, "\\%c", *s);
3145 if (PL_lex_starts) {
3151 sv_free(PL_thistoken);
3152 PL_thistoken = newSVpvn("",0);
3155 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
3156 if (PL_lex_casemods == 1 && PL_lex_inpat)
3165 case LEX_INTERPPUSH:
3166 return REPORT(sublex_push());
3168 case LEX_INTERPSTART:
3169 if (PL_bufptr == PL_bufend)
3170 return REPORT(sublex_done());
3171 DEBUG_T({ PerlIO_printf(Perl_debug_log,
3172 "### Interpolated variable\n"); });
3174 PL_lex_dojoin = (*PL_bufptr == '@');
3175 PL_lex_state = LEX_INTERPNORMAL;
3176 if (PL_lex_dojoin) {
3177 start_force(PL_curforce);
3178 NEXTVAL_NEXTTOKE.ival = 0;
3180 start_force(PL_curforce);
3181 force_ident("\"", '$');
3182 start_force(PL_curforce);
3183 NEXTVAL_NEXTTOKE.ival = 0;
3185 start_force(PL_curforce);
3186 NEXTVAL_NEXTTOKE.ival = 0;
3188 start_force(PL_curforce);
3189 NEXTVAL_NEXTTOKE.ival = OP_JOIN; /* emulate join($", ...) */
3192 if (PL_lex_starts++) {
3197 sv_free(PL_thistoken);
3198 PL_thistoken = newSVpvn("",0);
3201 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
3202 if (!PL_lex_casemods && PL_lex_inpat)
3209 case LEX_INTERPENDMAYBE:
3210 if (intuit_more(PL_bufptr)) {
3211 PL_lex_state = LEX_INTERPNORMAL; /* false alarm, more expr */
3217 if (PL_lex_dojoin) {
3218 PL_lex_dojoin = FALSE;
3219 PL_lex_state = LEX_INTERPCONCAT;
3223 sv_free(PL_thistoken);
3224 PL_thistoken = newSVpvn("",0);
3229 if (PL_lex_inwhat == OP_SUBST && PL_linestr == PL_lex_repl
3230 && SvEVALED(PL_lex_repl))
3232 if (PL_bufptr != PL_bufend)
3233 Perl_croak(aTHX_ "Bad evalled substitution pattern");
3237 case LEX_INTERPCONCAT:
3239 if (PL_lex_brackets)
3240 Perl_croak(aTHX_ "panic: INTERPCONCAT");
3242 if (PL_bufptr == PL_bufend)
3243 return REPORT(sublex_done());
3245 if (SvIVX(PL_linestr) == '\'') {
3246 SV *sv = newSVsv(PL_linestr);
3249 else if ( PL_hints & HINT_NEW_RE )
3250 sv = new_constant(NULL, 0, "qr", sv, sv, "q");
3251 yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
3255 s = scan_const(PL_bufptr);
3257 PL_lex_state = LEX_INTERPCASEMOD;
3259 PL_lex_state = LEX_INTERPSTART;
3262 if (s != PL_bufptr) {
3263 start_force(PL_curforce);
3265 curmad('X', newSVpvn(PL_bufptr,s-PL_bufptr));
3267 NEXTVAL_NEXTTOKE = yylval;
3270 if (PL_lex_starts++) {
3274 sv_free(PL_thistoken);
3275 PL_thistoken = newSVpvn("",0);
3278 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
3279 if (!PL_lex_casemods && PL_lex_inpat)
3292 PL_lex_state = LEX_NORMAL;
3293 s = scan_formline(PL_bufptr);
3294 if (!PL_lex_formbrack)
3300 PL_oldoldbufptr = PL_oldbufptr;
3306 sv_free(PL_thistoken);
3309 PL_realtokenstart = s - SvPVX(PL_linestr); /* assume but undo on ws */
3313 if (isIDFIRST_lazy_if(s,UTF))
3315 Perl_croak(aTHX_ "Unrecognized character \\x%02X", *s & 255);
3318 goto fake_eof; /* emulate EOF on ^D or ^Z */
3327 if (PL_lex_brackets) {
3328 yyerror(PL_lex_formbrack
3329 ? "Format not terminated"
3330 : "Missing right curly or square bracket");
3332 DEBUG_T( { PerlIO_printf(Perl_debug_log,
3333 "### Tokener got EOF\n");
3337 if (s++ < PL_bufend)
3338 goto retry; /* ignore stray nulls */
3341 if (!PL_in_eval && !PL_preambled) {
3342 PL_preambled = TRUE;
3347 sv_setpv(PL_linestr,incl_perldb());
3348 if (SvCUR(PL_linestr))
3349 sv_catpvs(PL_linestr,";");
3351 while(AvFILLp(PL_preambleav) >= 0) {
3352 SV *tmpsv = av_shift(PL_preambleav);
3353 sv_catsv(PL_linestr, tmpsv);
3354 sv_catpvs(PL_linestr, ";");
3357 sv_free((SV*)PL_preambleav);
3358 PL_preambleav = NULL;
3360 if (PL_minus_n || PL_minus_p) {
3361 sv_catpvs(PL_linestr, "LINE: while (<>) {");
3363 sv_catpvs(PL_linestr,"chomp;");
3366 if ((*PL_splitstr == '/' || *PL_splitstr == '\''
3367 || *PL_splitstr == '"')
3368 && strchr(PL_splitstr + 1, *PL_splitstr))
3369 Perl_sv_catpvf(aTHX_ PL_linestr, "our @F=split(%s);", PL_splitstr);
3371 /* "q\0${splitstr}\0" is legal perl. Yes, even NUL
3372 bytes can be used as quoting characters. :-) */
3373 const char *splits = PL_splitstr;
3374 sv_catpvs(PL_linestr, "our @F=split(q\0");
3377 if (*splits == '\\')
3378 sv_catpvn(PL_linestr, splits, 1);
3379 sv_catpvn(PL_linestr, splits, 1);
3380 } while (*splits++);
3381 /* This loop will embed the trailing NUL of
3382 PL_linestr as the last thing it does before
3384 sv_catpvs(PL_linestr, ");");
3388 sv_catpvs(PL_linestr,"our @F=split(' ');");
3392 sv_catpvs(PL_linestr,"use feature ':5.10';");
3393 sv_catpvs(PL_linestr, "\n");
3394 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3395 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3396 PL_last_lop = PL_last_uni = NULL;
3397 if (PERLDB_LINE && PL_curstash != PL_debstash) {
3398 SV * const sv = newSV(0);
3400 sv_upgrade(sv, SVt_PVMG);
3401 sv_setsv(sv,PL_linestr);
3404 av_store(CopFILEAVx(PL_curcop),(I32)CopLINE(PL_curcop),sv);
3409 bof = PL_rsfp ? TRUE : FALSE;
3410 if ((s = filter_gets(PL_linestr, PL_rsfp, 0)) == NULL) {
3413 PL_realtokenstart = -1;
3416 if (PL_preprocess && !PL_in_eval)
3417 (void)PerlProc_pclose(PL_rsfp);
3418 else if ((PerlIO *)PL_rsfp == PerlIO_stdin())
3419 PerlIO_clearerr(PL_rsfp);
3421 (void)PerlIO_close(PL_rsfp);
3423 PL_doextract = FALSE;
3425 if (!PL_in_eval && (PL_minus_n || PL_minus_p)) {
3430 sv_setpv(PL_linestr,PL_minus_p
3431 ? ";}continue{print;}" : ";}");
3432 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3433 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3434 PL_last_lop = PL_last_uni = NULL;
3435 PL_minus_n = PL_minus_p = 0;
3438 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3439 PL_last_lop = PL_last_uni = NULL;
3440 sv_setpvn(PL_linestr,"",0);
3441 TOKEN(';'); /* not infinite loop because rsfp is NULL now */
3443 /* If it looks like the start of a BOM or raw UTF-16,
3444 * check if it in fact is. */
3450 #ifdef PERLIO_IS_STDIO
3451 # ifdef __GNU_LIBRARY__
3452 # if __GNU_LIBRARY__ == 1 /* Linux glibc5 */
3453 # define FTELL_FOR_PIPE_IS_BROKEN
3457 # if __GLIBC__ == 1 /* maybe some glibc5 release had it like this? */
3458 # define FTELL_FOR_PIPE_IS_BROKEN
3463 #ifdef FTELL_FOR_PIPE_IS_BROKEN
3464 /* This loses the possibility to detect the bof
3465 * situation on perl -P when the libc5 is being used.
3466 * Workaround? Maybe attach some extra state to PL_rsfp?
3469 bof = PerlIO_tell(PL_rsfp) == SvCUR(PL_linestr);
3471 bof = PerlIO_tell(PL_rsfp) == (Off_t)SvCUR(PL_linestr);
3474 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3475 s = swallow_bom((U8*)s);
3479 /* Incest with pod. */
3482 sv_catsv(PL_thiswhite, PL_linestr);
3484 if (*s == '=' && strnEQ(s, "=cut", 4)) {
3485 sv_setpvn(PL_linestr, "", 0);
3486 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3487 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3488 PL_last_lop = PL_last_uni = NULL;
3489 PL_doextract = FALSE;
3493 } while (PL_doextract);
3494 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = s;
3495 if (PERLDB_LINE && PL_curstash != PL_debstash) {
3496 SV * const sv = newSV(0);
3498 sv_upgrade(sv, SVt_PVMG);
3499 sv_setsv(sv,PL_linestr);
3502 av_store(CopFILEAVx(PL_curcop),(I32)CopLINE(PL_curcop),sv);
3504 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3505 PL_last_lop = PL_last_uni = NULL;
3506 if (CopLINE(PL_curcop) == 1) {
3507 while (s < PL_bufend && isSPACE(*s))
3509 if (*s == ':' && s[1] != ':') /* for csh execing sh scripts */
3513 PL_thiswhite = newSVpvn(PL_linestart, s - PL_linestart);
3517 if (*s == '#' && *(s+1) == '!')
3519 #ifdef ALTERNATE_SHEBANG
3521 static char const as[] = ALTERNATE_SHEBANG;
3522 if (*s == as[0] && strnEQ(s, as, sizeof(as) - 1))
3523 d = s + (sizeof(as) - 1);
3525 #endif /* ALTERNATE_SHEBANG */
3534 while (*d && !isSPACE(*d))
3538 #ifdef ARG_ZERO_IS_SCRIPT
3539 if (ipathend > ipath) {
3541 * HP-UX (at least) sets argv[0] to the script name,
3542 * which makes $^X incorrect. And Digital UNIX and Linux,
3543 * at least, set argv[0] to the basename of the Perl
3544 * interpreter. So, having found "#!", we'll set it right.
3546 SV * const x = GvSV(gv_fetchpvs("\030", GV_ADD|GV_NOTQUAL,
3548 assert(SvPOK(x) || SvGMAGICAL(x));
3549 if (sv_eq(x, CopFILESV(PL_curcop))) {
3550 sv_setpvn(x, ipath, ipathend - ipath);
3556 const char *bstart = SvPV_const(CopFILESV(PL_curcop),blen);
3557 const char * const lstart = SvPV_const(x,llen);
3559 bstart += blen - llen;
3560 if (strnEQ(bstart, lstart, llen) && bstart[-1] == '/') {
3561 sv_setpvn(x, ipath, ipathend - ipath);
3566 TAINT_NOT; /* $^X is always tainted, but that's OK */
3568 #endif /* ARG_ZERO_IS_SCRIPT */
3573 d = instr(s,"perl -");
3575 d = instr(s,"perl");
3577 /* avoid getting into infinite loops when shebang
3578 * line contains "Perl" rather than "perl" */
3580 for (d = ipathend-4; d >= ipath; --d) {
3581 if ((*d == 'p' || *d == 'P')
3582 && !ibcmp(d, "perl", 4))
3592 #ifdef ALTERNATE_SHEBANG
3594 * If the ALTERNATE_SHEBANG on this system starts with a
3595 * character that can be part of a Perl expression, then if
3596 * we see it but not "perl", we're probably looking at the
3597 * start of Perl code, not a request to hand off to some
3598 * other interpreter. Similarly, if "perl" is there, but
3599 * not in the first 'word' of the line, we assume the line
3600 * contains the start of the Perl program.
3602 if (d && *s != '#') {
3603 const char *c = ipath;
3604 while (*c && !strchr("; \t\r\n\f\v#", *c))
3607 d = NULL; /* "perl" not in first word; ignore */
3609 *s = '#'; /* Don't try to parse shebang line */
3611 #endif /* ALTERNATE_SHEBANG */
3612 #ifndef MACOS_TRADITIONAL
3617 !instr(s,"indir") &&
3618 instr(PL_origargv[0],"perl"))
3625 while (s < PL_bufend && isSPACE(*s))
3627 if (s < PL_bufend) {
3628 Newxz(newargv,PL_origargc+3,char*);
3630 while (s < PL_bufend && !isSPACE(*s))
3633 Copy(PL_origargv+1, newargv+2, PL_origargc+1, char*);
3636 newargv = PL_origargv;
3639 PerlProc_execv(ipath, EXEC_ARGV_CAST(newargv));
3641 Perl_croak(aTHX_ "Can't exec %s", ipath);
3645 while (*d && !isSPACE(*d)) d++;
3646 while (SPACE_OR_TAB(*d)) d++;
3649 const bool switches_done = PL_doswitches;
3650 const U32 oldpdb = PL_perldb;
3651 const bool oldn = PL_minus_n;
3652 const bool oldp = PL_minus_p;
3655 if (*d == 'M' || *d == 'm' || *d == 'C') {
3656 const char * const m = d;
3657 while (*d && !isSPACE(*d))
3659 Perl_croak(aTHX_ "Too late for \"-%.*s\" option",
3662 d = moreswitches(d);
3664 if (PL_doswitches && !switches_done) {
3665 int argc = PL_origargc;
3666 char **argv = PL_origargv;
3669 } while (argc && argv[0][0] == '-' && argv[0][1]);
3670 init_argv_symbols(argc,argv);
3672 if ((PERLDB_LINE && !oldpdb) ||
3673 ((PL_minus_n || PL_minus_p) && !(oldn || oldp)))
3674 /* if we have already added "LINE: while (<>) {",
3675 we must not do it again */
3677 sv_setpvn(PL_linestr, "", 0);
3678 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3679 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3680 PL_last_lop = PL_last_uni = NULL;
3681 PL_preambled = FALSE;
3683 (void)gv_fetchfile(PL_origfilename);
3690 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
3692 PL_lex_state = LEX_FORMLINE;
3697 #ifdef PERL_STRICT_CR
3698 Perl_warn(aTHX_ "Illegal character \\%03o (carriage return)", '\r');
3700 "\t(Maybe you didn't strip carriage returns after a network transfer?)\n");
3702 case ' ': case '\t': case '\f': case 013:
3703 #ifdef MACOS_TRADITIONAL
3707 PL_realtokenstart = -1;
3716 PL_realtokenstart = -1;
3720 if (PL_lex_state != LEX_NORMAL || (PL_in_eval && !PL_rsfp)) {
3721 if (*s == '#' && s == PL_linestart && PL_in_eval && !PL_rsfp) {
3722 /* handle eval qq[#line 1 "foo"\n ...] */
3723 CopLINE_dec(PL_curcop);
3726 if (PL_madskills && !PL_lex_formbrack && !PL_in_eval) {
3728 if (!PL_in_eval || PL_rsfp)
3733 while (d < PL_bufend && *d != '\n')
3737 else if (d > PL_bufend) /* Found by Ilya: feed random input to Perl. */
3738 Perl_croak(aTHX_ "panic: input overflow");
3741 PL_thiswhite = newSVpvn(s, d - s);
3746 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
3748 PL_lex_state = LEX_FORMLINE;
3754 if (PL_madskills && CopLINE(PL_curcop) >= 1 && !PL_lex_formbrack) {
3755 if (CopLINE(PL_curcop) == 1 && s[0] == '#' && s[1] == '!') {
3758 TOKEN(PEG); /* make sure any #! line is accessible */
3763 /* if (PL_madskills && PL_lex_formbrack) { */
3765 while (d < PL_bufend && *d != '\n')
3769 else if (d > PL_bufend) /* Found by Ilya: feed random input to Perl. */
3770 Perl_croak(aTHX_ "panic: input overflow");
3771 if (PL_madskills && CopLINE(PL_curcop) >= 1) {
3773 PL_thiswhite = newSVpvn("",0);
3774 if (CopLINE(PL_curcop) == 1) {
3775 sv_setpvn(PL_thiswhite, "", 0);
3778 sv_catpvn(PL_thiswhite, s, d - s);
3792 if (s[1] && isALPHA(s[1]) && !isALNUM(s[2])) {
3800 while (s < PL_bufend && SPACE_OR_TAB(*s))
3803 if (strnEQ(s,"=>",2)) {
3804 s = force_word(PL_bufptr,WORD,FALSE,FALSE,FALSE);
3805 DEBUG_T( { S_printbuf(aTHX_
3806 "### Saw unary minus before =>, forcing word %s\n", s);
3808 OPERATOR('-'); /* unary minus */
3810 PL_last_uni = PL_oldbufptr;
3812 case 'r': ftst = OP_FTEREAD; break;
3813 case 'w': ftst = OP_FTEWRITE; break;
3814 case 'x': ftst = OP_FTEEXEC; break;
3815 case 'o': ftst = OP_FTEOWNED; break;
3816 case 'R': ftst = OP_FTRREAD; break;
3817 case 'W': ftst = OP_FTRWRITE; break;
3818 case 'X': ftst = OP_FTREXEC; break;
3819 case 'O': ftst = OP_FTROWNED; break;
3820 case 'e': ftst = OP_FTIS; break;
3821 case 'z': ftst = OP_FTZERO; break;
3822 case 's': ftst = OP_FTSIZE; break;
3823 case 'f': ftst = OP_FTFILE; break;
3824 case 'd': ftst = OP_FTDIR; break;
3825 case 'l': ftst = OP_FTLINK; break;
3826 case 'p': ftst = OP_FTPIPE; break;
3827 case 'S': ftst = OP_FTSOCK; break;
3828 case 'u': ftst = OP_FTSUID; break;
3829 case 'g': ftst = OP_FTSGID; break;
3830 case 'k': ftst = OP_FTSVTX; break;
3831 case 'b': ftst = OP_FTBLK; break;
3832 case 'c': ftst = OP_FTCHR; break;
3833 case 't': ftst = OP_FTTTY; break;
3834 case 'T': ftst = OP_FTTEXT; break;
3835 case 'B': ftst = OP_FTBINARY; break;
3836 case 'M': case 'A': case 'C':
3837 gv_fetchpvs("\024", GV_ADD|GV_NOTQUAL, SVt_PV);
3839 case 'M': ftst = OP_FTMTIME; break;
3840 case 'A': ftst = OP_FTATIME; break;
3841 case 'C': ftst = OP_FTCTIME; break;
3849 PL_last_lop_op = (OPCODE)ftst;
3850 DEBUG_T( { PerlIO_printf(Perl_debug_log,
3851 "### Saw file test %c\n", (int)tmp);
3856 /* Assume it was a minus followed by a one-letter named
3857 * subroutine call (or a -bareword), then. */
3858 DEBUG_T( { PerlIO_printf(Perl_debug_log,
3859 "### '-%c' looked like a file test but was not\n",
3866 const char tmp = *s++;
3869 if (PL_expect == XOPERATOR)
3874 else if (*s == '>') {
3877 if (isIDFIRST_lazy_if(s,UTF)) {
3878 s = force_word(s,METHOD,FALSE,TRUE,FALSE);
3886 if (PL_expect == XOPERATOR)
3889 if (isSPACE(*s) || !isSPACE(*PL_bufptr))
3891 OPERATOR('-'); /* unary minus */
3897 const char tmp = *s++;
3900 if (PL_expect == XOPERATOR)
3905 if (PL_expect == XOPERATOR)
3908 if (isSPACE(*s) || !isSPACE(*PL_bufptr))
3915 if (PL_expect != XOPERATOR) {
3916 s = scan_ident(s, PL_bufend, PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
3917 PL_expect = XOPERATOR;
3918 force_ident(PL_tokenbuf, '*');
3931 if (PL_expect == XOPERATOR) {
3935 PL_tokenbuf[0] = '%';
3936 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1, TRUE);
3937 if (!PL_tokenbuf[1]) {
3940 PL_pending_ident = '%';
3951 && (PL_expect == XOPERATOR || PL_expect == XTERMORDORDOR)
3952 && FEATURE_IS_ENABLED("~~"))
3959 const char tmp = *s++;
3965 goto just_a_word_zero_gv;
3968 switch (PL_expect) {
3974 if (!PL_in_my || PL_lex_state != LEX_NORMAL)
3976 PL_bufptr = s; /* update in case we back off */
3982 PL_expect = XTERMBLOCK;
3985 stuffstart = s - SvPVX(PL_linestr) - 1;
3989 while (isIDFIRST_lazy_if(s,UTF)) {
3991 d = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
3992 if (isLOWER(*s) && (tmp = keyword(PL_tokenbuf, len))) {
3993 if (tmp < 0) tmp = -tmp;
4009 d = scan_str(d,TRUE,TRUE);
4011 /* MUST advance bufptr here to avoid bogus
4012 "at end of line" context messages from yyerror().
4014 PL_bufptr = s + len;
4015 yyerror("Unterminated attribute parameter in attribute list");
4018 return REPORT(0); /* EOF indicator */
4022 SV *sv = newSVpvn(s, len);
4023 sv_catsv(sv, PL_lex_stuff);
4024 attrs = append_elem(OP_LIST, attrs,
4025 newSVOP(OP_CONST, 0, sv));
4026 SvREFCNT_dec(PL_lex_stuff);
4027 PL_lex_stuff = NULL;
4030 if (len == 6 && strnEQ(s, "unique", len)) {
4031 if (PL_in_my == KEY_our) {
4033 GvUNIQUE_on(cGVOPx_gv(yylval.opval));
4035 /* skip to avoid loading attributes.pm */
4037 deprecate(":unique");
4040 Perl_croak(aTHX_ "The 'unique' attribute may only be applied to 'our' variables");
4043 /* NOTE: any CV attrs applied here need to be part of
4044 the CVf_BUILTIN_ATTRS define in cv.h! */
4045 else if (!PL_in_my && len == 6 && strnEQ(s, "lvalue", len))
4046 CvLVALUE_on(PL_compcv);
4047 else if (!PL_in_my && len == 6 && strnEQ(s, "locked", len))
4048 CvLOCKED_on(PL_compcv);
4049 else if (!PL_in_my && len == 6 && strnEQ(s, "method", len))
4050 CvMETHOD_on(PL_compcv);
4051 else if (!PL_in_my && len == 9 && strnEQ(s, "assertion", len))
4052 CvASSERTION_on(PL_compcv);
4053 /* After we've set the flags, it could be argued that
4054 we don't need to do the attributes.pm-based setting
4055 process, and shouldn't bother appending recognized
4056 flags. To experiment with that, uncomment the
4057 following "else". (Note that's already been
4058 uncommented. That keeps the above-applied built-in
4059 attributes from being intercepted (and possibly
4060 rejected) by a package's attribute routines, but is
4061 justified by the performance win for the common case
4062 of applying only built-in attributes.) */
4064 attrs = append_elem(OP_LIST, attrs,
4065 newSVOP(OP_CONST, 0,
4069 if (*s == ':' && s[1] != ':')
4072 break; /* require real whitespace or :'s */
4073 /* XXX losing whitespace on sequential attributes here */
4077 = (PL_expect == XOPERATOR ? '=' : '{'); /*'}(' for vi */
4078 if (*s != ';' && *s != '}' && *s != tmp
4079 && (tmp != '=' || *s != ')')) {
4080 const char q = ((*s == '\'') ? '"' : '\'');
4081 /* If here for an expression, and parsed no attrs, back
4083 if (tmp == '=' && !attrs) {
4087 /* MUST advance bufptr here to avoid bogus "at end of line"
4088 context messages from yyerror().
4092 ? Perl_form(aTHX_ "Invalid separator character "
4093 "%c%c%c in attribute list", q, *s, q)
4094 : "Unterminated attribute list" );
4102 start_force(PL_curforce);
4103 NEXTVAL_NEXTTOKE.opval = attrs;
4104 CURMAD('_', PL_nextwhite);
4109 PL_thistoken = newSVpvn(SvPVX(PL_linestr) + stuffstart,
4110 (s - SvPVX(PL_linestr)) - stuffstart);
4118 if (PL_last_lop == PL_oldoldbufptr || PL_last_uni == PL_oldoldbufptr)
4119 PL_oldbufptr = PL_oldoldbufptr; /* allow print(STDOUT 123) */
4127 const char tmp = *s++;
4132 const char tmp = *s++;
4140 if (PL_lex_brackets <= 0)
4141 yyerror("Unmatched right square bracket");
4144 if (PL_lex_state == LEX_INTERPNORMAL) {
4145 if (PL_lex_brackets == 0) {
4146 if (*s != '[' && *s != '{' && (*s != '-' || s[1] != '>'))
4147 PL_lex_state = LEX_INTERPEND;
4154 if (PL_lex_brackets > 100) {
4155 Renew(PL_lex_brackstack, PL_lex_brackets + 10, char);
4157 switch (PL_expect) {
4159 if (PL_lex_formbrack) {
4163 if (PL_oldoldbufptr == PL_last_lop)
4164 PL_lex_brackstack[PL_lex_brackets++] = XTERM;
4166 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
4167 OPERATOR(HASHBRACK);
4169 while (s < PL_bufend && SPACE_OR_TAB(*s))
4172 PL_tokenbuf[0] = '\0';
4173 if (d < PL_bufend && *d == '-') {
4174 PL_tokenbuf[0] = '-';
4176 while (d < PL_bufend && SPACE_OR_TAB(*d))
4179 if (d < PL_bufend && isIDFIRST_lazy_if(d,UTF)) {
4180 d = scan_word(d, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1,
4182 while (d < PL_bufend && SPACE_OR_TAB(*d))
4185 const char minus = (PL_tokenbuf[0] == '-');
4186 s = force_word(s + minus, WORD, FALSE, TRUE, FALSE);
4194 PL_lex_brackstack[PL_lex_brackets++] = XSTATE;
4199 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
4204 if (PL_oldoldbufptr == PL_last_lop)
4205 PL_lex_brackstack[PL_lex_brackets++] = XTERM;
4207 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
4210 if (PL_expect == XREF && PL_lex_state == LEX_INTERPNORMAL) {
4212 /* This hack is to get the ${} in the message. */
4214 yyerror("syntax error");
4217 OPERATOR(HASHBRACK);
4219 /* This hack serves to disambiguate a pair of curlies
4220 * as being a block or an anon hash. Normally, expectation
4221 * determines that, but in cases where we're not in a
4222 * position to expect anything in particular (like inside
4223 * eval"") we have to resolve the ambiguity. This code
4224 * covers the case where the first term in the curlies is a
4225 * quoted string. Most other cases need to be explicitly
4226 * disambiguated by prepending a "+" before the opening
4227 * curly in order to force resolution as an anon hash.
4229 * XXX should probably propagate the outer expectation
4230 * into eval"" to rely less on this hack, but that could
4231 * potentially break current behavior of eval"".
4235 if (*s == '\'' || *s == '"' || *s == '`') {
4236 /* common case: get past first string, handling escapes */
4237 for (t++; t < PL_bufend && *t != *s;)
4238 if (*t++ == '\\' && (*t == '\\' || *t == *s))
4242 else if (*s == 'q') {
4245 || ((*t == 'q' || *t == 'x') && ++t < PL_bufend
4248 /* skip q//-like construct */
4250 char open, close, term;
4253 while (t < PL_bufend && isSPACE(*t))
4255 /* check for q => */
4256 if (t+1 < PL_bufend && t[0] == '=' && t[1] == '>') {
4257 OPERATOR(HASHBRACK);
4261 if (term && (tmps = strchr("([{< )]}> )]}>",term)))
4265 for (t++; t < PL_bufend; t++) {
4266 if (*t == '\\' && t+1 < PL_bufend && open != '\\')
4268 else if (*t == open)
4272 for (t++; t < PL_bufend; t++) {
4273 if (*t == '\\' && t+1 < PL_bufend)
4275 else if (*t == close && --brackets <= 0)
4277 else if (*t == open)
4284 /* skip plain q word */
4285 while (t < PL_bufend && isALNUM_lazy_if(t,UTF))
4288 else if (isALNUM_lazy_if(t,UTF)) {
4290 while (t < PL_bufend && isALNUM_lazy_if(t,UTF))
4293 while (t < PL_bufend && isSPACE(*t))
4295 /* if comma follows first term, call it an anon hash */
4296 /* XXX it could be a comma expression with loop modifiers */
4297 if (t < PL_bufend && ((*t == ',' && (*s == 'q' || !isLOWER(*s)))
4298 || (*t == '=' && t[1] == '>')))
4299 OPERATOR(HASHBRACK);
4300 if (PL_expect == XREF)
4303 PL_lex_brackstack[PL_lex_brackets-1] = XSTATE;
4309 yylval.ival = CopLINE(PL_curcop);
4310 if (isSPACE(*s) || *s == '#')
4311 PL_copline = NOLINE; /* invalidate current command line number */
4316 if (PL_lex_brackets <= 0)
4317 yyerror("Unmatched right curly bracket");
4319 PL_expect = (expectation)PL_lex_brackstack[--PL_lex_brackets];
4320 if (PL_lex_brackets < PL_lex_formbrack && PL_lex_state != LEX_INTERPNORMAL)
4321 PL_lex_formbrack = 0;
4322 if (PL_lex_state == LEX_INTERPNORMAL) {
4323 if (PL_lex_brackets == 0) {
4324 if (PL_expect & XFAKEBRACK) {
4325 PL_expect &= XENUMMASK;
4326 PL_lex_state = LEX_INTERPEND;
4331 PL_thiswhite = newSVpvn("",0);
4332 sv_catpvn(PL_thiswhite,"}",1);
4335 return yylex(); /* ignore fake brackets */
4337 if (*s == '-' && s[1] == '>')
4338 PL_lex_state = LEX_INTERPENDMAYBE;
4339 else if (*s != '[' && *s != '{')
4340 PL_lex_state = LEX_INTERPEND;
4343 if (PL_expect & XFAKEBRACK) {
4344 PL_expect &= XENUMMASK;
4346 return yylex(); /* ignore fake brackets */
4348 start_force(PL_curforce);
4350 curmad('X', newSVpvn(s-1,1));
4351 CURMAD('_', PL_thiswhite);
4356 PL_thistoken = newSVpvn("",0);
4364 if (PL_expect == XOPERATOR) {
4365 if (PL_bufptr == PL_linestart && ckWARN(WARN_SEMICOLON)
4366 && isIDFIRST_lazy_if(s,UTF))
4368 CopLINE_dec(PL_curcop);
4369 Perl_warner(aTHX_ packWARN(WARN_SEMICOLON), PL_warn_nosemi);
4370 CopLINE_inc(PL_curcop);
4375 s = scan_ident(s - 1, PL_bufend, PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
4377 PL_expect = XOPERATOR;
4378 force_ident(PL_tokenbuf, '&');
4382 yylval.ival = (OPpENTERSUB_AMPER<<8);
4394 const char tmp = *s++;
4401 if (tmp && isSPACE(*s) && ckWARN(WARN_SYNTAX)
4402 && strchr("+-*/%.^&|<",tmp))
4403 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
4404 "Reversed %c= operator",(int)tmp);
4406 if (PL_expect == XSTATE && isALPHA(tmp) &&
4407 (s == PL_linestart+1 || s[-2] == '\n') )
4409 if (PL_in_eval && !PL_rsfp) {
4414 if (strnEQ(s,"=cut",4)) {
4430 PL_thiswhite = newSVpvn("",0);
4431 sv_catpvn(PL_thiswhite, PL_linestart,
4432 PL_bufend - PL_linestart);
4436 PL_doextract = TRUE;
4440 if (PL_lex_brackets < PL_lex_formbrack) {
4442 #ifdef PERL_STRICT_CR
4443 for (t = s; SPACE_OR_TAB(*t); t++) ;
4445 for (t = s; SPACE_OR_TAB(*t) || *t == '\r'; t++) ;
4447 if (*t == '\n' || *t == '#') {
4458 const char tmp = *s++;
4460 /* was this !=~ where !~ was meant?
4461 * warn on m:!=~\s+([/?]|[msy]\W|tr\W): */
4463 if (*s == '~' && ckWARN(WARN_SYNTAX)) {
4464 const char *t = s+1;
4466 while (t < PL_bufend && isSPACE(*t))
4469 if (*t == '/' || *t == '?' ||
4470 ((*t == 'm' || *t == 's' || *t == 'y')
4471 && !isALNUM(t[1])) ||
4472 (*t == 't' && t[1] == 'r' && !isALNUM(t[2])))
4473 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
4474 "!=~ should be !~");
4484 if (PL_expect != XOPERATOR) {
4485 if (s[1] != '<' && !strchr(s,'>'))
4488 s = scan_heredoc(s);
4490 s = scan_inputsymbol(s);
4491 TERM(sublex_start());
4497 SHop(OP_LEFT_SHIFT);
4511 const char tmp = *s++;
4513 SHop(OP_RIGHT_SHIFT);
4514 else if (tmp == '=')
4523 if (PL_expect == XOPERATOR) {
4524 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
4526 deprecate_old(commaless_variable_list);
4527 return REPORT(','); /* grandfather non-comma-format format */
4531 if (s[1] == '#' && (isIDFIRST_lazy_if(s+2,UTF) || strchr("{$:+-", s[2]))) {
4532 PL_tokenbuf[0] = '@';
4533 s = scan_ident(s + 1, PL_bufend, PL_tokenbuf + 1,
4534 sizeof PL_tokenbuf - 1, FALSE);
4535 if (PL_expect == XOPERATOR)
4536 no_op("Array length", s);
4537 if (!PL_tokenbuf[1])
4539 PL_expect = XOPERATOR;
4540 PL_pending_ident = '#';
4544 PL_tokenbuf[0] = '$';
4545 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1,
4546 sizeof PL_tokenbuf - 1, FALSE);
4547 if (PL_expect == XOPERATOR)
4549 if (!PL_tokenbuf[1]) {
4551 yyerror("Final $ should be \\$ or $name");
4555 /* This kludge not intended to be bulletproof. */
4556 if (PL_tokenbuf[1] == '[' && !PL_tokenbuf[2]) {
4557 yylval.opval = newSVOP(OP_CONST, 0,
4558 newSViv(CopARYBASE_get(&PL_compiling)));
4559 yylval.opval->op_private = OPpCONST_ARYBASE;
4565 const char tmp = *s;
4566 if (PL_lex_state == LEX_NORMAL)
4569 if ((PL_expect != XREF || PL_oldoldbufptr == PL_last_lop)
4570 && intuit_more(s)) {
4572 PL_tokenbuf[0] = '@';
4573 if (ckWARN(WARN_SYNTAX)) {
4576 isSPACE(*t) || isALNUM_lazy_if(t,UTF) || *t == '$';
4579 PL_bufptr = PEEKSPACE(PL_bufptr); /* XXX can realloc */
4580 while (t < PL_bufend && *t != ']')
4582 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
4583 "Multidimensional syntax %.*s not supported",
4584 (int)((t - PL_bufptr) + 1), PL_bufptr);
4588 else if (*s == '{') {
4590 PL_tokenbuf[0] = '%';
4591 if (strEQ(PL_tokenbuf+1, "SIG") && ckWARN(WARN_SYNTAX)
4592 && (t = strchr(s, '}')) && (t = strchr(t, '=')))
4594 char tmpbuf[sizeof PL_tokenbuf];
4595 for (t++; isSPACE(*t); t++) ;
4596 if (isIDFIRST_lazy_if(t,UTF)) {
4598 t = scan_word(t, tmpbuf, sizeof tmpbuf, TRUE,
4600 for (; isSPACE(*t); t++) ;
4601 if (*t == ';' && get_cv(tmpbuf, FALSE))
4602 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
4603 "You need to quote \"%s\"",
4610 PL_expect = XOPERATOR;
4611 if (PL_lex_state == LEX_NORMAL && isSPACE((char)tmp)) {
4612 const bool islop = (PL_last_lop == PL_oldoldbufptr);
4613 if (!islop || PL_last_lop_op == OP_GREPSTART)
4614 PL_expect = XOPERATOR;
4615 else if (strchr("$@\"'`q", *s))
4616 PL_expect = XTERM; /* e.g. print $fh "foo" */
4617 else if (strchr("&*<%", *s) && isIDFIRST_lazy_if(s+1,UTF))
4618 PL_expect = XTERM; /* e.g. print $fh &sub */
4619 else if (isIDFIRST_lazy_if(s,UTF)) {
4620 char tmpbuf[sizeof PL_tokenbuf];
4622 scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
4623 if ((t2 = keyword(tmpbuf, len))) {
4624 /* binary operators exclude handle interpretations */
4636 PL_expect = XTERM; /* e.g. print $fh length() */
4641 PL_expect = XTERM; /* e.g. print $fh subr() */
4644 else if (isDIGIT(*s))
4645 PL_expect = XTERM; /* e.g. print $fh 3 */
4646 else if (*s == '.' && isDIGIT(s[1]))
4647 PL_expect = XTERM; /* e.g. print $fh .3 */
4648 else if ((*s == '?' || *s == '-' || *s == '+')
4649 && !isSPACE(s[1]) && s[1] != '=')
4650 PL_expect = XTERM; /* e.g. print $fh -1 */
4651 else if (*s == '/' && !isSPACE(s[1]) && s[1] != '='
4653 PL_expect = XTERM; /* e.g. print $fh /.../
4654 XXX except DORDOR operator
4656 else if (*s == '<' && s[1] == '<' && !isSPACE(s[2])
4658 PL_expect = XTERM; /* print $fh <<"EOF" */
4661 PL_pending_ident = '$';
4665 if (PL_expect == XOPERATOR)
4667 PL_tokenbuf[0] = '@';
4668 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1, FALSE);
4669 if (!PL_tokenbuf[1]) {
4672 if (PL_lex_state == LEX_NORMAL)
4674 if ((PL_expect != XREF || PL_oldoldbufptr == PL_last_lop) && intuit_more(s)) {
4676 PL_tokenbuf[0] = '%';
4678 /* Warn about @ where they meant $. */
4679 if (*s == '[' || *s == '{') {
4680 if (ckWARN(WARN_SYNTAX)) {
4681 const char *t = s + 1;
4682 while (*t && (isALNUM_lazy_if(t,UTF) || strchr(" \t$#+-'\"", *t)))
4684 if (*t == '}' || *t == ']') {
4686 PL_bufptr = PEEKSPACE(PL_bufptr); /* XXX can realloc */
4687 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
4688 "Scalar value %.*s better written as $%.*s",
4689 (int)(t-PL_bufptr), PL_bufptr,
4690 (int)(t-PL_bufptr-1), PL_bufptr+1);
4695 PL_pending_ident = '@';
4698 case '/': /* may be division, defined-or, or pattern */
4699 if (PL_expect == XTERMORDORDOR && s[1] == '/') {
4703 case '?': /* may either be conditional or pattern */
4704 if(PL_expect == XOPERATOR) {
4712 /* A // operator. */
4722 /* Disable warning on "study /blah/" */
4723 if (PL_oldoldbufptr == PL_last_uni
4724 && (*PL_last_uni != 's' || s - PL_last_uni < 5
4725 || memNE(PL_last_uni, "study", 5)
4726 || isALNUM_lazy_if(PL_last_uni+5,UTF)
4729 s = scan_pat(s,OP_MATCH);
4730 TERM(sublex_start());
4734 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack
4735 #ifdef PERL_STRICT_CR
4738 && (s[1] == '\n' || (s[1] == '\r' && s[2] == '\n'))
4740 && (s == PL_linestart || s[-1] == '\n') )
4742 PL_lex_formbrack = 0;
4746 if (PL_expect == XOPERATOR || !isDIGIT(s[1])) {
4752 yylval.ival = OPf_SPECIAL;
4758 if (PL_expect != XOPERATOR)
4763 case '0': case '1': case '2': case '3': case '4':
4764 case '5': case '6': case '7': case '8': case '9':
4765 s = scan_num(s, &yylval);
4766 DEBUG_T( { S_printbuf(aTHX_ "### Saw number in %s\n", s); } );
4767 if (PL_expect == XOPERATOR)
4772 s = scan_str(s,!!PL_madskills,FALSE);
4773 DEBUG_T( { S_printbuf(aTHX_ "### Saw string before %s\n", s); } );
4774 if (PL_expect == XOPERATOR) {
4775 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
4777 deprecate_old(commaless_variable_list);
4778 return REPORT(','); /* grandfather non-comma-format format */
4785 yylval.ival = OP_CONST;
4786 TERM(sublex_start());
4789 s = scan_str(s,!!PL_madskills,FALSE);
4790 DEBUG_T( { S_printbuf(aTHX_ "### Saw string before %s\n", s); } );
4791 if (PL_expect == XOPERATOR) {
4792 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
4794 deprecate_old(commaless_variable_list);
4795 return REPORT(','); /* grandfather non-comma-format format */
4802 yylval.ival = OP_CONST;
4803 /* FIXME. I think that this can be const if char *d is replaced by
4804 more localised variables. */
4805 for (d = SvPV(PL_lex_stuff, len); len; len--, d++) {
4806 if (*d == '$' || *d == '@' || *d == '\\' || !UTF8_IS_INVARIANT((U8)*d)) {
4807 yylval.ival = OP_STRINGIFY;
4811 TERM(sublex_start());
4814 s = scan_str(s,!!PL_madskills,FALSE);
4815 DEBUG_T( { S_printbuf(aTHX_ "### Saw backtick string before %s\n", s); } );
4816 if (PL_expect == XOPERATOR)
4817 no_op("Backticks",s);
4820 yylval.ival = OP_BACKTICK;
4822 TERM(sublex_start());
4826 if (PL_lex_inwhat && isDIGIT(*s) && ckWARN(WARN_SYNTAX))
4827 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),"Can't use \\%c to mean $%c in expression",
4829 if (PL_expect == XOPERATOR)
4830 no_op("Backslash",s);
4834 if (isDIGIT(s[1]) && PL_expect != XOPERATOR) {
4835 char *start = s + 2;
4836 while (isDIGIT(*start) || *start == '_')
4838 if (*start == '.' && isDIGIT(start[1])) {
4839 s = scan_num(s, &yylval);
4842 /* avoid v123abc() or $h{v1}, allow C<print v10;> */
4843 else if (!isALPHA(*start) && (PL_expect == XTERM
4844 || PL_expect == XREF || PL_expect == XSTATE
4845 || PL_expect == XTERMORDORDOR)) {
4846 /* XXX Use gv_fetchpvn rather than stomping on a const string */
4847 const char c = *start;
4850 gv = gv_fetchpv(s, 0, SVt_PVCV);
4853 s = scan_num(s, &yylval);
4860 if (isDIGIT(s[1]) && PL_expect == XOPERATOR) {
4896 I32 orig_keyword = 0;
4901 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
4903 /* Some keywords can be followed by any delimiter, including ':' */
4904 tmp = ((len == 1 && strchr("msyq", PL_tokenbuf[0])) ||
4905 (len == 2 && ((PL_tokenbuf[0] == 't' && PL_tokenbuf[1] == 'r') ||
4906 (PL_tokenbuf[0] == 'q' &&
4907 strchr("qwxr", PL_tokenbuf[1])))));
4909 /* x::* is just a word, unless x is "CORE" */
4910 if (!tmp && *s == ':' && s[1] == ':' && strNE(PL_tokenbuf, "CORE"))
4914 while (d < PL_bufend && isSPACE(*d))
4915 d++; /* no comments skipped here, or s### is misparsed */
4917 /* Is this a label? */
4918 if (!tmp && PL_expect == XSTATE
4919 && d < PL_bufend && *d == ':' && *(d + 1) != ':') {
4921 yylval.pval = savepv(PL_tokenbuf);
4926 /* Check for keywords */
4927 tmp = keyword(PL_tokenbuf, len);
4929 /* Is this a word before a => operator? */
4930 if (*d == '=' && d[1] == '>') {
4933 = (OP*)newSVOP(OP_CONST, 0,
4934 S_newSV_maybe_utf8(aTHX_ PL_tokenbuf, len));
4935 yylval.opval->op_private = OPpCONST_BARE;
4939 if (tmp < 0) { /* second-class keyword? */
4940 GV *ogv = NULL; /* override (winner) */
4941 GV *hgv = NULL; /* hidden (loser) */
4942 if (PL_expect != XOPERATOR && (*s != ':' || s[1] != ':')) {
4944 if ((gv = gv_fetchpvn_flags(PL_tokenbuf, len, 0, SVt_PVCV)) &&
4947 if (GvIMPORTED_CV(gv))
4949 else if (! CvMETHOD(cv))
4953 (gvp = (GV**)hv_fetch(PL_globalstash,PL_tokenbuf,len,FALSE)) &&
4954 (gv = *gvp) != (GV*)&PL_sv_undef &&
4955 GvCVu(gv) && GvIMPORTED_CV(gv))
4962 tmp = 0; /* overridden by import or by GLOBAL */
4965 && -tmp==KEY_lock /* XXX generalizable kludge */
4967 && !hv_fetchs(GvHVn(PL_incgv), "Thread.pm", FALSE))
4969 tmp = 0; /* any sub overrides "weak" keyword */
4971 else { /* no override */
4973 if (tmp == KEY_dump && ckWARN(WARN_MISC)) {
4974 Perl_warner(aTHX_ packWARN(WARN_MISC),
4975 "dump() better written as CORE::dump()");
4979 if (hgv && tmp != KEY_x && tmp != KEY_CORE
4980 && ckWARN(WARN_AMBIGUOUS)) /* never ambiguous */
4981 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
4982 "Ambiguous call resolved as CORE::%s(), %s",
4983 GvENAME(hgv), "qualify as such or use &");
4990 default: /* not a keyword */
4991 /* Trade off - by using this evil construction we can pull the
4992 variable gv into the block labelled keylookup. If not, then
4993 we have to give it function scope so that the goto from the
4994 earlier ':' case doesn't bypass the initialisation. */
4996 just_a_word_zero_gv:
5004 const char lastchar = (PL_bufptr == PL_oldoldbufptr ? 0 : PL_bufptr[-1]);
5007 SV *nextPL_nextwhite = 0;
5011 /* Get the rest if it looks like a package qualifier */
5013 if (*s == '\'' || (*s == ':' && s[1] == ':')) {
5015 s = scan_word(s, PL_tokenbuf + len, sizeof PL_tokenbuf - len,
5018 Perl_croak(aTHX_ "Bad name after %s%s", PL_tokenbuf,
5019 *s == '\'' ? "'" : "::");
5024 if (PL_expect == XOPERATOR) {
5025 if (PL_bufptr == PL_linestart) {
5026 CopLINE_dec(PL_curcop);
5027 Perl_warner(aTHX_ packWARN(WARN_SEMICOLON), PL_warn_nosemi);
5028 CopLINE_inc(PL_curcop);
5031 no_op("Bareword",s);
5034 /* Look for a subroutine with this name in current package,
5035 unless name is "Foo::", in which case Foo is a bearword
5036 (and a package name). */
5038 if (len > 2 && !PL_madskills &&
5039 PL_tokenbuf[len - 2] == ':' && PL_tokenbuf[len - 1] == ':')
5041 if (ckWARN(WARN_BAREWORD)
5042 && ! gv_fetchpvn_flags(PL_tokenbuf, len, 0, SVt_PVHV))
5043 Perl_warner(aTHX_ packWARN(WARN_BAREWORD),
5044 "Bareword \"%s\" refers to nonexistent package",
5047 PL_tokenbuf[len] = '\0';
5053 /* Mustn't actually add anything to a symbol table.
5054 But also don't want to "initialise" any placeholder
5055 constants that might already be there into full
5056 blown PVGVs with attached PVCV. */
5057 gv = gv_fetchpvn_flags(PL_tokenbuf, len,
5058 GV_NOADD_NOINIT, SVt_PVCV);
5063 /* if we saw a global override before, get the right name */
5066 sv = newSVpvs("CORE::GLOBAL::");
5067 sv_catpv(sv,PL_tokenbuf);
5070 /* If len is 0, newSVpv does strlen(), which is correct.
5071 If len is non-zero, then it will be the true length,
5072 and so the scalar will be created correctly. */
5073 sv = newSVpv(PL_tokenbuf,len);
5076 if (PL_madskills && !PL_thistoken) {
5077 char *start = SvPVX(PL_linestr) + PL_realtokenstart;
5078 PL_thistoken = newSVpv(start,s - start);
5079 PL_realtokenstart = s - SvPVX(PL_linestr);
5083 /* Presume this is going to be a bareword of some sort. */
5086 yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
5087 yylval.opval->op_private = OPpCONST_BARE;
5088 /* UTF-8 package name? */
5089 if (UTF && !IN_BYTES &&
5090 is_utf8_string((U8*)SvPVX_const(sv), SvCUR(sv)))
5093 /* And if "Foo::", then that's what it certainly is. */
5098 /* Do the explicit type check so that we don't need to force
5099 the initialisation of the symbol table to have a real GV.
5100 Beware - gv may not really be a PVGV, cv may not really be
5101 a PVCV, (because of the space optimisations that gv_init
5102 understands) But they're true if for this symbol there is
5103 respectively a typeglob and a subroutine.
5105 cv = gv ? ((SvTYPE(gv) == SVt_PVGV)
5106 /* Real typeglob, so get the real subroutine: */
5108 /* A proxy for a subroutine in this package? */
5109 : SvOK(gv) ? (CV *) gv : NULL)
5112 /* See if it's the indirect object for a list operator. */
5114 if (PL_oldoldbufptr &&
5115 PL_oldoldbufptr < PL_bufptr &&
5116 (PL_oldoldbufptr == PL_last_lop
5117 || PL_oldoldbufptr == PL_last_uni) &&
5118 /* NO SKIPSPACE BEFORE HERE! */
5119 (PL_expect == XREF ||
5120 ((PL_opargs[PL_last_lop_op] >> OASHIFT)& 7) == OA_FILEREF))
5122 bool immediate_paren = *s == '(';
5124 /* (Now we can afford to cross potential line boundary.) */
5125 s = SKIPSPACE2(s,nextPL_nextwhite);
5127 PL_nextwhite = nextPL_nextwhite; /* assume no & deception */
5130 /* Two barewords in a row may indicate method call. */
5132 if ((isIDFIRST_lazy_if(s,UTF) || *s == '$') &&
5133 (tmp = intuit_method(s, gv, cv)))
5136 /* If not a declared subroutine, it's an indirect object. */
5137 /* (But it's an indir obj regardless for sort.) */
5138 /* Also, if "_" follows a filetest operator, it's a bareword */
5141 ( !immediate_paren && (PL_last_lop_op == OP_SORT ||
5143 (PL_last_lop_op != OP_MAPSTART &&
5144 PL_last_lop_op != OP_GREPSTART))))
5145 || (PL_tokenbuf[0] == '_' && PL_tokenbuf[1] == '\0'
5146 && ((PL_opargs[PL_last_lop_op] & OA_CLASS_MASK) == OA_FILESTATOP))
5149 PL_expect = (PL_last_lop == PL_oldoldbufptr) ? XTERM : XOPERATOR;
5154 PL_expect = XOPERATOR;
5157 s = SKIPSPACE2(s,nextPL_nextwhite);
5158 PL_nextwhite = nextPL_nextwhite;
5163 /* Is this a word before a => operator? */
5164 if (*s == '=' && s[1] == '>' && !pkgname) {
5166 sv_setpv(((SVOP*)yylval.opval)->op_sv, PL_tokenbuf);
5167 if (UTF && !IN_BYTES && is_utf8_string((U8*)PL_tokenbuf, len))
5168 SvUTF8_on(((SVOP*)yylval.opval)->op_sv);
5172 /* If followed by a paren, it's certainly a subroutine. */
5176 for (d = s + 1; SPACE_OR_TAB(*d); d++) ;
5177 if (*d == ')' && (sv = gv_const_sv(gv))) {
5181 char *par = SvPVX(PL_linestr) + PL_realtokenstart;
5182 sv_catpvn(PL_thistoken, par, s - par);
5184 sv_free(PL_nextwhite);
5194 PL_nextwhite = PL_thiswhite;
5197 start_force(PL_curforce);
5199 NEXTVAL_NEXTTOKE.opval = yylval.opval;
5200 PL_expect = XOPERATOR;
5203 PL_nextwhite = nextPL_nextwhite;
5204 curmad('X', PL_thistoken);
5205 PL_thistoken = newSVpvn("",0);
5213 /* If followed by var or block, call it a method (unless sub) */
5215 if ((*s == '$' || *s == '{') && (!gv || !cv)) {
5216 PL_last_lop = PL_oldbufptr;
5217 PL_last_lop_op = OP_METHOD;
5221 /* If followed by a bareword, see if it looks like indir obj. */
5224 && (isIDFIRST_lazy_if(s,UTF) || *s == '$')
5225 && (tmp = intuit_method(s, gv, cv)))
5228 /* Not a method, so call it a subroutine (if defined) */
5231 if (lastchar == '-' && ckWARN_d(WARN_AMBIGUOUS))
5232 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
5233 "Ambiguous use of -%s resolved as -&%s()",
5234 PL_tokenbuf, PL_tokenbuf);
5235 /* Check for a constant sub */
5236 if ((sv = gv_const_sv(gv))) {
5238 SvREFCNT_dec(((SVOP*)yylval.opval)->op_sv);
5239 ((SVOP*)yylval.opval)->op_sv = SvREFCNT_inc_simple(sv);
5240 yylval.opval->op_private = 0;
5244 /* Resolve to GV now. */
5245 if (SvTYPE(gv) != SVt_PVGV) {
5246 gv = gv_fetchpv(PL_tokenbuf, 0, SVt_PVCV);
5247 assert (SvTYPE(gv) == SVt_PVGV);
5248 /* cv must have been some sort of placeholder, so
5249 now needs replacing with a real code reference. */
5253 op_free(yylval.opval);
5254 yylval.opval = newCVREF(0, newGVOP(OP_GV, 0, gv));
5255 yylval.opval->op_private |= OPpENTERSUB_NOPAREN;
5256 PL_last_lop = PL_oldbufptr;
5257 PL_last_lop_op = OP_ENTERSUB;
5258 /* Is there a prototype? */
5265 const char *proto = SvPV_const((SV*)cv, protolen);
5268 if (*proto == '$' && proto[1] == '\0')
5270 while (*proto == ';')
5272 if (*proto == '&' && *s == '{') {
5273 sv_setpv(PL_subname, PL_curstash ?
5274 "__ANON__" : "__ANON__::__ANON__");
5281 PL_nextwhite = PL_thiswhite;
5284 start_force(PL_curforce);
5285 NEXTVAL_NEXTTOKE.opval = yylval.opval;
5288 PL_nextwhite = nextPL_nextwhite;
5289 curmad('X', PL_thistoken);
5290 PL_thistoken = newSVpvn("",0);
5297 /* Guess harder when madskills require "best effort". */
5298 if (PL_madskills && (!gv || !GvCVu(gv))) {
5299 int probable_sub = 0;
5300 if (strchr("\"'`$@%0123456789!*+{[<", *s))
5302 else if (isALPHA(*s)) {
5306 d = scan_word(d, tmpbuf, sizeof tmpbuf, TRUE, &tmplen);
5307 if (!keyword(tmpbuf,tmplen))
5310 while (d < PL_bufend && isSPACE(*d))
5312 if (*d == '=' && d[1] == '>')
5317 gv = gv_fetchpv(PL_tokenbuf, TRUE, SVt_PVCV);
5318 op_free(yylval.opval);
5319 yylval.opval = newCVREF(0, newGVOP(OP_GV, 0, gv));
5320 yylval.opval->op_private |= OPpENTERSUB_NOPAREN;
5321 PL_last_lop = PL_oldbufptr;
5322 PL_last_lop_op = OP_ENTERSUB;
5323 PL_nextwhite = PL_thiswhite;
5325 start_force(PL_curforce);
5326 NEXTVAL_NEXTTOKE.opval = yylval.opval;
5328 PL_nextwhite = nextPL_nextwhite;
5329 curmad('X', PL_thistoken);
5330 PL_thistoken = newSVpvn("",0);
5335 NEXTVAL_NEXTTOKE.opval = yylval.opval;
5342 /* Call it a bare word */
5344 if (PL_hints & HINT_STRICT_SUBS)
5345 yylval.opval->op_private |= OPpCONST_STRICT;
5348 if (lastchar != '-') {
5349 if (ckWARN(WARN_RESERVED)) {
5350 for (d = PL_tokenbuf; *d && isLOWER(*d); d++) ;
5351 if (!*d && !gv_stashpv(PL_tokenbuf,FALSE))
5352 Perl_warner(aTHX_ packWARN(WARN_RESERVED), PL_warn_reserved,
5359 if ((lastchar == '*' || lastchar == '%' || lastchar == '&')
5360 && ckWARN_d(WARN_AMBIGUOUS)) {
5361 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
5362 "Operator or semicolon missing before %c%s",
5363 lastchar, PL_tokenbuf);
5364 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
5365 "Ambiguous use of %c resolved as operator %c",
5366 lastchar, lastchar);
5372 yylval.opval = (OP*)newSVOP(OP_CONST, 0,
5373 newSVpv(CopFILE(PL_curcop),0));
5377 yylval.opval = (OP*)newSVOP(OP_CONST, 0,
5378 Perl_newSVpvf(aTHX_ "%"IVdf, (IV)CopLINE(PL_curcop)));
5381 case KEY___PACKAGE__:
5382 yylval.opval = (OP*)newSVOP(OP_CONST, 0,
5384 ? newSVhek(HvNAME_HEK(PL_curstash))
5391 if (PL_rsfp && (!PL_in_eval || PL_tokenbuf[2] == 'D')) {
5392 const char *pname = "main";
5393 if (PL_tokenbuf[2] == 'D')
5394 pname = HvNAME_get(PL_curstash ? PL_curstash : PL_defstash);
5395 gv = gv_fetchpv(Perl_form(aTHX_ "%s::DATA", pname), GV_ADD,
5399 GvIOp(gv) = newIO();
5400 IoIFP(GvIOp(gv)) = PL_rsfp;
5401 #if defined(HAS_FCNTL) && defined(F_SETFD)
5403 const int fd = PerlIO_fileno(PL_rsfp);
5404 fcntl(fd,F_SETFD,fd >= 3);
5407 /* Mark this internal pseudo-handle as clean */
5408 IoFLAGS(GvIOp(gv)) |= IOf_UNTAINT;
5410 IoTYPE(GvIOp(gv)) = IoTYPE_PIPE;
5411 else if ((PerlIO*)PL_rsfp == PerlIO_stdin())
5412 IoTYPE(GvIOp(gv)) = IoTYPE_STD;
5414 IoTYPE(GvIOp(gv)) = IoTYPE_RDONLY;
5415 #if defined(WIN32) && !defined(PERL_TEXTMODE_SCRIPTS)
5416 /* if the script was opened in binmode, we need to revert
5417 * it to text mode for compatibility; but only iff it has CRs
5418 * XXX this is a questionable hack at best. */
5419 if (PL_bufend-PL_bufptr > 2
5420 && PL_bufend[-1] == '\n' && PL_bufend[-2] == '\r')
5423 if (IoTYPE(GvIOp(gv)) == IoTYPE_RDONLY) {
5424 loc = PerlIO_tell(PL_rsfp);
5425 (void)PerlIO_seek(PL_rsfp, 0L, 0);
5428 if (PerlLIO_setmode(PL_rsfp, O_TEXT) != -1) {
5430 if (PerlLIO_setmode(PerlIO_fileno(PL_rsfp), O_TEXT) != -1) {
5431 #endif /* NETWARE */
5432 #ifdef PERLIO_IS_STDIO /* really? */
5433 # if defined(__BORLANDC__)
5434 /* XXX see note in do_binmode() */
5435 ((FILE*)PL_rsfp)->flags &= ~_F_BIN;
5439 PerlIO_seek(PL_rsfp, loc, 0);
5443 #ifdef PERLIO_LAYERS
5446 PerlIO_apply_layers(aTHX_ PL_rsfp, NULL, ":utf8");
5447 else if (PL_encoding) {
5454 XPUSHs(PL_encoding);
5456 call_method("name", G_SCALAR);
5460 PerlIO_apply_layers(aTHX_ PL_rsfp, NULL,
5461 Perl_form(aTHX_ ":encoding(%"SVf")",
5470 if (PL_realtokenstart >= 0) {
5471 char *tstart = SvPVX(PL_linestr) + PL_realtokenstart;
5473 PL_endwhite = newSVpvn("",0);
5474 sv_catsv(PL_endwhite, PL_thiswhite);
5476 sv_catpvn(PL_endwhite, tstart, PL_bufend - tstart);
5477 PL_realtokenstart = -1;
5479 while ((s = filter_gets(PL_endwhite, PL_rsfp,
5480 SvCUR(PL_endwhite))) != Nullch) ;
5494 if (PL_expect == XSTATE) {
5501 if (*s == ':' && s[1] == ':') {
5504 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
5505 if (!(tmp = keyword(PL_tokenbuf, len)))
5506 Perl_croak(aTHX_ "CORE::%s is not a keyword", PL_tokenbuf);
5509 else if (tmp == KEY_require || tmp == KEY_do)
5510 /* that's a way to remember we saw "CORE::" */
5523 LOP(OP_ACCEPT,XTERM);
5529 LOP(OP_ATAN2,XTERM);
5535 LOP(OP_BINMODE,XTERM);
5538 LOP(OP_BLESS,XTERM);
5547 /* When 'use switch' is in effect, continue has a dual
5548 life as a control operator. */
5550 if (!FEATURE_IS_ENABLED("switch"))
5553 /* We have to disambiguate the two senses of
5554 "continue". If the next token is a '{' then
5555 treat it as the start of a continue block;
5556 otherwise treat it as a control operator.
5568 (void)gv_fetchpvs("ENV", GV_ADD|GV_NOTQUAL, SVt_PVHV);
5585 if (!PL_cryptseen) {
5586 PL_cryptseen = TRUE;
5590 LOP(OP_CRYPT,XTERM);
5593 LOP(OP_CHMOD,XTERM);
5596 LOP(OP_CHOWN,XTERM);
5599 LOP(OP_CONNECT,XTERM);
5618 s = force_word(s,WORD,TRUE,TRUE,FALSE);
5619 if (orig_keyword == KEY_do) {
5628 PL_hints |= HINT_BLOCK_SCOPE;
5638 gv_fetchpvs("AnyDBM_File::ISA", GV_ADDMULTI, SVt_PVAV);
5639 LOP(OP_DBMOPEN,XTERM);
5645 s = force_word(s,WORD,TRUE,FALSE,FALSE);
5652 yylval.ival = CopLINE(PL_curcop);
5668 PL_expect = (*s == '{') ? XTERMBLOCK : XTERM;
5669 UNIBRACK(OP_ENTEREVAL);
5687 case KEY_endhostent:
5693 case KEY_endservent:
5696 case KEY_endprotoent:
5707 yylval.ival = CopLINE(PL_curcop);
5709 if (PL_expect == XSTATE && isIDFIRST_lazy_if(s,UTF)) {
5712 int soff = s - SvPVX(PL_linestr); /* for skipspace realloc */
5715 if ((PL_bufend - p) >= 3 &&
5716 strnEQ(p, "my", 2) && isSPACE(*(p + 2)))
5718 else if ((PL_bufend - p) >= 4 &&
5719 strnEQ(p, "our", 3) && isSPACE(*(p + 3)))
5722 if (isIDFIRST_lazy_if(p,UTF)) {
5723 p = scan_ident(p, PL_bufend,
5724 PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
5728 Perl_croak(aTHX_ "Missing $ on loop variable");
5730 s = SvPVX(PL_linestr) + soff;
5736 LOP(OP_FORMLINE,XTERM);
5742 LOP(OP_FCNTL,XTERM);
5748 LOP(OP_FLOCK,XTERM);
5757 LOP(OP_GREPSTART, XREF);
5760 s = force_word(s,WORD,TRUE,FALSE,FALSE);
5775 case KEY_getpriority:
5776 LOP(OP_GETPRIORITY,XTERM);
5778 case KEY_getprotobyname:
5781 case KEY_getprotobynumber:
5782 LOP(OP_GPBYNUMBER,XTERM);
5784 case KEY_getprotoent:
5796 case KEY_getpeername:
5797 UNI(OP_GETPEERNAME);
5799 case KEY_gethostbyname:
5802 case KEY_gethostbyaddr:
5803 LOP(OP_GHBYADDR,XTERM);
5805 case KEY_gethostent:
5808 case KEY_getnetbyname:
5811 case KEY_getnetbyaddr:
5812 LOP(OP_GNBYADDR,XTERM);
5817 case KEY_getservbyname:
5818 LOP(OP_GSBYNAME,XTERM);
5820 case KEY_getservbyport:
5821 LOP(OP_GSBYPORT,XTERM);
5823 case KEY_getservent:
5826 case KEY_getsockname:
5827 UNI(OP_GETSOCKNAME);
5829 case KEY_getsockopt:
5830 LOP(OP_GSOCKOPT,XTERM);
5845 yylval.ival = CopLINE(PL_curcop);
5856 yylval.ival = CopLINE(PL_curcop);
5860 LOP(OP_INDEX,XTERM);
5866 LOP(OP_IOCTL,XTERM);
5878 s = force_word(s,WORD,TRUE,FALSE,FALSE);
5910 LOP(OP_LISTEN,XTERM);
5919 s = scan_pat(s,OP_MATCH);
5920 TERM(sublex_start());
5923 LOP(OP_MAPSTART, XREF);
5926 LOP(OP_MKDIR,XTERM);
5929 LOP(OP_MSGCTL,XTERM);
5932 LOP(OP_MSGGET,XTERM);
5935 LOP(OP_MSGRCV,XTERM);
5938 LOP(OP_MSGSND,XTERM);
5944 if (isIDFIRST_lazy_if(s,UTF)) {
5948 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, TRUE, &len);
5949 if (len == 3 && strnEQ(PL_tokenbuf, "sub", 3))
5951 PL_in_my_stash = find_in_my_stash(PL_tokenbuf, len);
5952 if (!PL_in_my_stash) {
5955 sprintf(tmpbuf, "No such class %.1000s", PL_tokenbuf);
5959 if (PL_madskills) { /* just add type to declarator token */
5960 sv_catsv(PL_thistoken, PL_nextwhite);
5962 sv_catpvn(PL_thistoken, start, s - start);
5970 s = force_word(s,WORD,TRUE,FALSE,FALSE);
5977 s = tokenize_use(0, s);
5981 if (*s == '(' || (s = SKIPSPACE1(s), *s == '('))
5988 if (isIDFIRST_lazy_if(s,UTF)) {
5990 for (d = s; isALNUM_lazy_if(d,UTF); d++) ;
5991 for (t=d; *t && isSPACE(*t); t++) ;
5992 if ( *t && strchr("|&*+-=!?:.", *t) && ckWARN_d(WARN_PRECEDENCE)
5994 && !(t[0] == '=' && t[1] == '>')
5996 int parms_len = (int)(d-s);
5997 Perl_warner(aTHX_ packWARN(WARN_PRECEDENCE),
5998 "Precedence problem: open %.*s should be open(%.*s)",
5999 parms_len, s, parms_len, s);
6005 yylval.ival = OP_OR;
6015 LOP(OP_OPEN_DIR,XTERM);
6018 checkcomma(s,PL_tokenbuf,"filehandle");
6022 checkcomma(s,PL_tokenbuf,"filehandle");
6041 s = force_word(s,WORD,FALSE,TRUE,FALSE);
6045 LOP(OP_PIPE_OP,XTERM);
6048 s = scan_str(s,!!PL_madskills,FALSE);
6051 yylval.ival = OP_CONST;
6052 TERM(sublex_start());
6058 s = scan_str(s,!!PL_madskills,FALSE);
6061 PL_expect = XOPERATOR;
6063 if (SvCUR(PL_lex_stuff)) {
6066 d = SvPV_force(PL_lex_stuff, len);
6068 for (; isSPACE(*d) && len; --len, ++d)
6073 if (!warned && ckWARN(WARN_QW)) {
6074 for (; !isSPACE(*d) && len; --len, ++d) {
6076 Perl_warner(aTHX_ packWARN(WARN_QW),
6077 "Possible attempt to separate words with commas");
6080 else if (*d == '#') {
6081 Perl_warner(aTHX_ packWARN(WARN_QW),
6082 "Possible attempt to put comments in qw() list");
6088 for (; !isSPACE(*d) && len; --len, ++d)
6091 sv = newSVpvn(b, d-b);
6092 if (DO_UTF8(PL_lex_stuff))
6094 words = append_elem(OP_LIST, words,
6095 newSVOP(OP_CONST, 0, tokeq(sv)));
6099 start_force(PL_curforce);
6100 NEXTVAL_NEXTTOKE.opval = words;
6105 SvREFCNT_dec(PL_lex_stuff);
6106 PL_lex_stuff = NULL;
6112 s = scan_str(s,!!PL_madskills,FALSE);
6115 yylval.ival = OP_STRINGIFY;
6116 if (SvIVX(PL_lex_stuff) == '\'')
6117 SvIV_set(PL_lex_stuff, 0); /* qq'$foo' should intepolate */
6118 TERM(sublex_start());
6121 s = scan_pat(s,OP_QR);
6122 TERM(sublex_start());
6125 s = scan_str(s,!!PL_madskills,FALSE);
6128 yylval.ival = OP_BACKTICK;
6130 TERM(sublex_start());
6138 s = force_version(s, FALSE);
6140 else if (*s != 'v' || !isDIGIT(s[1])
6141 || (s = force_version(s, TRUE), *s == 'v'))
6143 *PL_tokenbuf = '\0';
6144 s = force_word(s,WORD,TRUE,TRUE,FALSE);
6145 if (isIDFIRST_lazy_if(PL_tokenbuf,UTF))
6146 gv_stashpvn(PL_tokenbuf, strlen(PL_tokenbuf), TRUE);
6148 yyerror("<> should be quotes");
6150 if (orig_keyword == KEY_require) {
6158 PL_last_uni = PL_oldbufptr;
6159 PL_last_lop_op = OP_REQUIRE;
6161 return REPORT( (int)REQUIRE );
6167 s = force_word(s,WORD,TRUE,FALSE,FALSE);
6171 LOP(OP_RENAME,XTERM);
6180 LOP(OP_RINDEX,XTERM);
6190 UNIDOR(OP_READLINE);
6203 LOP(OP_REVERSE,XTERM);
6206 UNIDOR(OP_READLINK);
6214 TERM(sublex_start());
6216 TOKEN(1); /* force error */
6219 checkcomma(s,PL_tokenbuf,"filehandle");
6229 LOP(OP_SELECT,XTERM);
6235 LOP(OP_SEMCTL,XTERM);
6238 LOP(OP_SEMGET,XTERM);
6241 LOP(OP_SEMOP,XTERM);
6247 LOP(OP_SETPGRP,XTERM);
6249 case KEY_setpriority:
6250 LOP(OP_SETPRIORITY,XTERM);
6252 case KEY_sethostent:
6258 case KEY_setservent:
6261 case KEY_setprotoent:
6271 LOP(OP_SEEKDIR,XTERM);
6273 case KEY_setsockopt:
6274 LOP(OP_SSOCKOPT,XTERM);
6280 LOP(OP_SHMCTL,XTERM);
6283 LOP(OP_SHMGET,XTERM);
6286 LOP(OP_SHMREAD,XTERM);
6289 LOP(OP_SHMWRITE,XTERM);
6292 LOP(OP_SHUTDOWN,XTERM);
6301 LOP(OP_SOCKET,XTERM);
6303 case KEY_socketpair:
6304 LOP(OP_SOCKPAIR,XTERM);
6307 checkcomma(s,PL_tokenbuf,"subroutine name");
6309 if (*s == ';' || *s == ')') /* probably a close */
6310 Perl_croak(aTHX_ "sort is now a reserved word");
6312 s = force_word(s,WORD,TRUE,TRUE,FALSE);
6316 LOP(OP_SPLIT,XTERM);
6319 LOP(OP_SPRINTF,XTERM);
6322 LOP(OP_SPLICE,XTERM);
6337 LOP(OP_SUBSTR,XTERM);
6343 char tmpbuf[sizeof PL_tokenbuf];
6344 SSize_t tboffset = 0;
6345 expectation attrful;
6346 bool have_name, have_proto, bad_proto;
6347 const int key = tmp;
6352 char *tstart = SvPVX(PL_linestr) + PL_realtokenstart;
6353 SV *subtoken = newSVpvn(tstart, s - tstart);
6357 s = SKIPSPACE2(s,tmpwhite);
6362 if (isIDFIRST_lazy_if(s,UTF) || *s == '\'' ||
6363 (*s == ':' && s[1] == ':'))
6370 attrful = XATTRBLOCK;
6371 /* remember buffer pos'n for later force_word */
6372 tboffset = s - PL_oldbufptr;
6373 d = scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
6376 nametoke = newSVpvn(s, d - s);
6378 if (strchr(tmpbuf, ':'))
6379 sv_setpv(PL_subname, tmpbuf);
6381 sv_setsv(PL_subname,PL_curstname);
6382 sv_catpvs(PL_subname,"::");
6383 sv_catpvn(PL_subname,tmpbuf,len);
6390 CURMAD('X', nametoke);
6391 CURMAD('_', tmpwhite);
6392 (void) force_word(PL_oldbufptr + tboffset, WORD,
6395 s = SKIPSPACE2(d,tmpwhite);
6402 Perl_croak(aTHX_ "Missing name in \"my sub\"");
6403 PL_expect = XTERMBLOCK;
6404 attrful = XATTRTERM;
6405 sv_setpvn(PL_subname,"?",1);
6409 if (key == KEY_format) {
6411 PL_lex_formbrack = PL_lex_brackets + 1;
6413 PL_thistoken = subtoken;
6417 (void) force_word(PL_oldbufptr + tboffset, WORD,
6423 /* Look for a prototype */
6427 s = scan_str(s,!!PL_madskills,FALSE);
6429 Perl_croak(aTHX_ "Prototype not terminated");
6430 /* strip spaces and check for bad characters */
6431 d = SvPVX(PL_lex_stuff);
6434 for (p = d; *p; ++p) {
6437 if (!strchr("$@%*;[]&\\", *p))
6442 if (bad_proto && ckWARN(WARN_SYNTAX))
6443 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
6444 "Illegal character in prototype for %"SVf" : %s",
6446 SvCUR_set(PL_lex_stuff, tmp);
6451 CURMAD('q', PL_thisopen);
6452 CURMAD('_', tmpwhite);
6453 CURMAD('=', PL_thisstuff);
6454 CURMAD('Q', PL_thisclose);
6455 NEXTVAL_NEXTTOKE.opval =
6456 (OP*)newSVOP(OP_CONST, 0, PL_lex_stuff);
6457 PL_lex_stuff = Nullsv;
6460 s = SKIPSPACE2(s,tmpwhite);
6468 if (*s == ':' && s[1] != ':')
6469 PL_expect = attrful;
6470 else if (*s != '{' && key == KEY_sub) {
6472 Perl_croak(aTHX_ "Illegal declaration of anonymous subroutine");
6474 Perl_croak(aTHX_ "Illegal declaration of subroutine %"SVf, PL_subname);
6481 curmad('^', newSVpvn("",0));
6482 CURMAD('_', tmpwhite);
6486 PL_thistoken = subtoken;
6489 NEXTVAL_NEXTTOKE.opval =
6490 (OP*)newSVOP(OP_CONST, 0, PL_lex_stuff);
6491 PL_lex_stuff = NULL;
6496 sv_setpv(PL_subname,
6497 PL_curstash ? "__ANON__" : "__ANON__::__ANON__");
6501 (void) force_word(PL_oldbufptr + tboffset, WORD,
6511 LOP(OP_SYSTEM,XREF);
6514 LOP(OP_SYMLINK,XTERM);
6517 LOP(OP_SYSCALL,XTERM);
6520 LOP(OP_SYSOPEN,XTERM);
6523 LOP(OP_SYSSEEK,XTERM);
6526 LOP(OP_SYSREAD,XTERM);
6529 LOP(OP_SYSWRITE,XTERM);
6533 TERM(sublex_start());
6554 LOP(OP_TRUNCATE,XTERM);
6566 yylval.ival = CopLINE(PL_curcop);
6570 yylval.ival = CopLINE(PL_curcop);
6574 LOP(OP_UNLINK,XTERM);
6580 LOP(OP_UNPACK,XTERM);
6583 LOP(OP_UTIME,XTERM);
6589 LOP(OP_UNSHIFT,XTERM);
6592 s = tokenize_use(1, s);
6602 yylval.ival = CopLINE(PL_curcop);
6606 yylval.ival = CopLINE(PL_curcop);
6610 PL_hints |= HINT_BLOCK_SCOPE;
6617 LOP(OP_WAITPID,XTERM);
6626 ctl_l[0] = toCTRL('L');
6628 gv_fetchpvn_flags(ctl_l, 1, GV_ADD|GV_NOTQUAL, SVt_PV);
6631 /* Make sure $^L is defined */
6632 gv_fetchpvs("\f", GV_ADD|GV_NOTQUAL, SVt_PV);
6637 if (PL_expect == XOPERATOR)
6643 yylval.ival = OP_XOR;
6648 TERM(sublex_start());
6653 #pragma segment Main
6657 S_pending_ident(pTHX)
6661 register I32 tmp = 0;
6662 /* pit holds the identifier we read and pending_ident is reset */
6663 char pit = PL_pending_ident;
6664 PL_pending_ident = 0;
6666 /* PL_realtokenstart = realtokenend = PL_bufptr - SvPVX(PL_linestr); */
6667 DEBUG_T({ PerlIO_printf(Perl_debug_log,
6668 "### Pending identifier '%s'\n", PL_tokenbuf); });
6670 /* if we're in a my(), we can't allow dynamics here.
6671 $foo'bar has already been turned into $foo::bar, so
6672 just check for colons.
6674 if it's a legal name, the OP is a PADANY.
6677 if (PL_in_my == KEY_our) { /* "our" is merely analogous to "my" */
6678 if (strchr(PL_tokenbuf,':'))
6679 yyerror(Perl_form(aTHX_ "No package name allowed for "
6680 "variable %s in \"our\"",
6682 tmp = allocmy(PL_tokenbuf);
6685 if (strchr(PL_tokenbuf,':'))
6686 yyerror(Perl_form(aTHX_ PL_no_myglob,PL_tokenbuf));
6688 yylval.opval = newOP(OP_PADANY, 0);
6689 yylval.opval->op_targ = allocmy(PL_tokenbuf);
6695 build the ops for accesses to a my() variable.
6697 Deny my($a) or my($b) in a sort block, *if* $a or $b is
6698 then used in a comparison. This catches most, but not
6699 all cases. For instance, it catches
6700 sort { my($a); $a <=> $b }
6702 sort { my($a); $a < $b ? -1 : $a == $b ? 0 : 1; }
6703 (although why you'd do that is anyone's guess).
6706 if (!strchr(PL_tokenbuf,':')) {
6708 tmp = pad_findmy(PL_tokenbuf);
6709 if (tmp != NOT_IN_PAD) {
6710 /* might be an "our" variable" */
6711 if (PAD_COMPNAME_FLAGS_isOUR(tmp)) {
6712 /* build ops for a bareword */
6713 HV * const stash = PAD_COMPNAME_OURSTASH(tmp);
6714 HEK * const stashname = HvNAME_HEK(stash);
6715 SV * const sym = newSVhek(stashname);
6716 sv_catpvs(sym, "::");
6717 sv_catpv(sym, PL_tokenbuf+1);
6718 yylval.opval = (OP*)newSVOP(OP_CONST, 0, sym);
6719 yylval.opval->op_private = OPpCONST_ENTERED;
6722 ? (GV_ADDMULTI | GV_ADDINEVAL)
6725 ((PL_tokenbuf[0] == '$') ? SVt_PV
6726 : (PL_tokenbuf[0] == '@') ? SVt_PVAV
6731 /* if it's a sort block and they're naming $a or $b */
6732 if (PL_last_lop_op == OP_SORT &&
6733 PL_tokenbuf[0] == '$' &&
6734 (PL_tokenbuf[1] == 'a' || PL_tokenbuf[1] == 'b')
6737 for (d = PL_in_eval ? PL_oldoldbufptr : PL_linestart;
6738 d < PL_bufend && *d != '\n';
6741 if (strnEQ(d,"<=>",3) || strnEQ(d,"cmp",3)) {
6742 Perl_croak(aTHX_ "Can't use \"my %s\" in sort comparison",
6748 yylval.opval = newOP(OP_PADANY, 0);
6749 yylval.opval->op_targ = tmp;
6755 Whine if they've said @foo in a doublequoted string,
6756 and @foo isn't a variable we can find in the symbol
6759 if (pit == '@' && PL_lex_state != LEX_NORMAL && !PL_lex_brackets) {
6760 GV *gv = gv_fetchpv(PL_tokenbuf+1, 0, SVt_PVAV);
6761 if ((!gv || ((PL_tokenbuf[0] == '@') ? !GvAV(gv) : !GvHV(gv)))
6762 && ckWARN(WARN_AMBIGUOUS))
6764 /* Downgraded from fatal to warning 20000522 mjd */
6765 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
6766 "Possible unintended interpolation of %s in string",
6771 /* build ops for a bareword */
6772 yylval.opval = (OP*)newSVOP(OP_CONST, 0, newSVpv(PL_tokenbuf+1, 0));
6773 yylval.opval->op_private = OPpCONST_ENTERED;
6776 /* If the identifier refers to a stash, don't autovivify it.
6777 * Change 24660 had the side effect of causing symbol table
6778 * hashes to always be defined, even if they were freshly
6779 * created and the only reference in the entire program was
6780 * the single statement with the defined %foo::bar:: test.
6781 * It appears that all code in the wild doing this actually
6782 * wants to know whether sub-packages have been loaded, so
6783 * by avoiding auto-vivifying symbol tables, we ensure that
6784 * defined %foo::bar:: continues to be false, and the existing
6785 * tests still give the expected answers, even though what
6786 * they're actually testing has now changed subtly.
6788 (*PL_tokenbuf == '%' && *(d = PL_tokenbuf + strlen(PL_tokenbuf) - 1) == ':' && d[-1] == ':'
6790 : PL_in_eval ? (GV_ADDMULTI | GV_ADDINEVAL) : GV_ADD),
6791 ((PL_tokenbuf[0] == '$') ? SVt_PV
6792 : (PL_tokenbuf[0] == '@') ? SVt_PVAV
6798 * The following code was generated by perl_keyword.pl.
6802 Perl_keyword (pTHX_ const char *name, I32 len)
6807 case 1: /* 5 tokens of length 1 */
6839 case 2: /* 18 tokens of length 2 */
6985 case 3: /* 29 tokens of length 3 */
6989 if (name[1] == 'N' &&
7052 if (name[1] == 'i' &&
7074 return (FEATURE_IS_ENABLED("err") ? -KEY_err : 0);
7092 if (name[1] == 'o' &&
7101 if (name[1] == 'e' &&
7110 if (name[1] == 'n' &&
7119 if (name[1] == 'o' &&
7128 if (name[1] == 'a' &&
7137 if (name[1] == 'o' &&
7199 if (name[1] == 'e' &&
7213 return (FEATURE_IS_ENABLED("say") ? -KEY_say : 0);
7239 if (name[1] == 'i' &&
7248 if (name[1] == 's' &&
7257 if (name[1] == 'e' &&
7266 if (name[1] == 'o' &&
7278 case 4: /* 41 tokens of length 4 */
7282 if (name[1] == 'O' &&
7292 if (name[1] == 'N' &&
7302 if (name[1] == 'i' &&
7312 if (name[1] == 'h' &&
7322 if (name[1] == 'u' &&
7335 if (name[2] == 'c' &&
7344 if (name[2] == 's' &&
7353 if (name[2] == 'a' &&
7389 if (name[1] == 'o' &&
7402 if (name[2] == 't' &&
7411 if (name[2] == 'o' &&
7420 if (name[2] == 't' &&
7429 if (name[2] == 'e' &&
7442 if (name[1] == 'o' &&
7455 if (name[2] == 'y' &&
7464 if (name[2] == 'l' &&
7480 if (name[2] == 's' &&
7489 if (name[2] == 'n' &&
7498 if (name[2] == 'c' &&
7511 if (name[1] == 'e' &&
7521 if (name[1] == 'p' &&
7534 if (name[2] == 'c' &&
7543 if (name[2] == 'p' &&
7552 if (name[2] == 's' &&
7568 if (name[2] == 'n' &&
7638 if (name[2] == 'r' &&
7647 if (name[2] == 'r' &&
7656 if (name[2] == 'a' &&
7672 if (name[2] == 'l' &&
7734 if (name[2] == 'e' &&
7737 return (FEATURE_IS_ENABLED("switch") ? KEY_when : 0);
7750 case 5: /* 38 tokens of length 5 */
7754 if (name[1] == 'E' &&
7765 if (name[1] == 'H' &&
7779 if (name[2] == 'a' &&
7789 if (name[2] == 'a' &&
7806 if (name[2] == 'e' &&
7816 if (name[2] == 'e' &&
7820 return (FEATURE_IS_ENABLED("switch") ? -KEY_break : 0);
7836 if (name[3] == 'i' &&
7845 if (name[3] == 'o' &&
7881 if (name[2] == 'o' &&
7891 if (name[2] == 'y' &&
7905 if (name[1] == 'l' &&
7919 if (name[2] == 'n' &&
7929 if (name[2] == 'o' &&
7943 if (name[1] == 'i' &&
7948 return (FEATURE_IS_ENABLED("switch") ? KEY_given : 0);
7957 if (name[2] == 'd' &&
7967 if (name[2] == 'c' &&
7984 if (name[2] == 'c' &&
7994 if (name[2] == 't' &&
8008 if (name[1] == 'k' &&
8019 if (name[1] == 'r' &&
8033 if (name[2] == 's' &&
8043 if (name[2] == 'd' &&
8060 if (name[2] == 'm' &&
8070 if (name[2] == 'i' &&
8080 if (name[2] == 'e' &&
8090 if (name[2] == 'l' &&
8100 if (name[2] == 'a' &&
8110 if (name[2] == 'u' &&
8124 if (name[1] == 'i' &&
8138 if (name[2] == 'a' &&
8151 if (name[3] == 'e' &&
8186 if (name[2] == 'i' &&
8203 if (name[2] == 'i' &&
8213 if (name[2] == 'i' &&
8230 case 6: /* 33 tokens of length 6 */
8234 if (name[1] == 'c' &&
8249 if (name[2] == 'l' &&
8260 if (name[2] == 'r' &&
8275 if (name[1] == 'e' &&
8290 if (name[2] == 's' &&
8295 if(ckWARN_d(WARN_SYNTAX))
8296 Perl_warner(aTHX_ packWARN(WARN_SYNTAX), "elseif should be elsif");
8302 if (name[2] == 'i' &&
8320 if (name[2] == 'l' &&
8331 if (name[2] == 'r' &&
8346 if (name[1] == 'm' &&
8361 if (name[2] == 'n' &&
8372 if (name[2] == 's' &&
8387 if (name[1] == 's' &&
8393 if (name[4] == 't' &&
8402 if (name[4] == 'e' &&
8411 if (name[4] == 'c' &&
8420 if (name[4] == 'n' &&
8436 if (name[1] == 'r' &&
8454 if (name[3] == 'a' &&
8464 if (name[3] == 'u' &&
8478 if (name[2] == 'n' &&
8496 if (name[2] == 'a' &&
8510 if (name[3] == 'e' &&
8523 if (name[4] == 't' &&
8532 if (name[4] == 'e' &&
8554 if (name[4] == 't' &&
8563 if (name[4] == 'e' &&
8579 if (name[2] == 'c' &&
8590 if (name[2] == 'l' &&
8601 if (name[2] == 'b' &&
8612 if (name[2] == 's' &&
8635 if (name[4] == 's' &&
8644 if (name[4] == 'n' &&
8657 if (name[3] == 'a' &&
8674 if (name[1] == 'a' &&
8689 case 7: /* 29 tokens of length 7 */
8693 if (name[1] == 'E' &&
8706 if (name[1] == '_' &&
8719 if (name[1] == 'i' &&
8726 return -KEY_binmode;
8732 if (name[1] == 'o' &&
8739 return -KEY_connect;
8748 if (name[2] == 'm' &&
8754 return -KEY_dbmopen;
8765 if (name[4] == 'u' &&
8769 return (FEATURE_IS_ENABLED("switch") ? KEY_default : 0);
8775 if (name[4] == 'n' &&
8796 if (name[1] == 'o' &&
8809 if (name[1] == 'e' &&
8816 if (name[5] == 'r' &&
8819 return -KEY_getpgrp;
8825 if (name[5] == 'i' &&
8828 return -KEY_getppid;
8841 if (name[1] == 'c' &&
8848 return -KEY_lcfirst;
8854 if (name[1] == 'p' &&
8861 return -KEY_opendir;
8867 if (name[1] == 'a' &&
8885 if (name[3] == 'd' &&
8890 return -KEY_readdir;
8896 if (name[3] == 'u' &&
8907 if (name[3] == 'e' &&
8912 return -KEY_reverse;
8931 if (name[3] == 'k' &&
8936 return -KEY_seekdir;
8942 if (name[3] == 'p' &&
8947 return -KEY_setpgrp;
8957 if (name[2] == 'm' &&
8963 return -KEY_shmread;
8969 if (name[2] == 'r' &&
8975 return -KEY_sprintf;
8984 if (name[3] == 'l' &&
8989 return -KEY_symlink;
8998 if (name[4] == 'a' &&
9002 return -KEY_syscall;
9008 if (name[4] == 'p' &&
9012 return -KEY_sysopen;
9018 if (name[4] == 'e' &&
9022 return -KEY_sysread;
9028 if (name[4] == 'e' &&
9032 return -KEY_sysseek;
9050 if (name[1] == 'e' &&
9057 return -KEY_telldir;
9066 if (name[2] == 'f' &&
9072 return -KEY_ucfirst;
9078 if (name[2] == 's' &&
9084 return -KEY_unshift;
9094 if (name[1] == 'a' &&
9101 return -KEY_waitpid;
9110 case 8: /* 26 tokens of length 8 */
9114 if (name[1] == 'U' &&
9122 return KEY_AUTOLOAD;
9133 if (name[3] == 'A' &&
9139 return KEY___DATA__;
9145 if (name[3] == 'I' &&
9151 return -KEY___FILE__;
9157 if (name[3] == 'I' &&
9163 return -KEY___LINE__;
9179 if (name[2] == 'o' &&
9186 return -KEY_closedir;
9192 if (name[2] == 'n' &&
9199 return -KEY_continue;
9209 if (name[1] == 'b' &&
9217 return -KEY_dbmclose;
9223 if (name[1] == 'n' &&
9229 if (name[4] == 'r' &&
9234 return -KEY_endgrent;
9240 if (name[4] == 'w' &&
9245 return -KEY_endpwent;
9258 if (name[1] == 'o' &&
9266 return -KEY_formline;
9272 if (name[1] == 'e' &&
9283 if (name[6] == 'n' &&
9286 return -KEY_getgrent;
9292 if (name[6] == 'i' &&
9295 return -KEY_getgrgid;
9301 if (name[6] == 'a' &&
9304 return -KEY_getgrnam;
9317 if (name[4] == 'o' &&
9322 return -KEY_getlogin;
9333 if (name[6] == 'n' &&
9336 return -KEY_getpwent;
9342 if (name[6] == 'a' &&
9345 return -KEY_getpwnam;
9351 if (name[6] == 'i' &&
9354 return -KEY_getpwuid;
9374 if (name[1] == 'e' &&
9381 if (name[5] == 'i' &&
9388 return -KEY_readline;
9393 return -KEY_readlink;
9404 if (name[5] == 'i' &&
9408 return -KEY_readpipe;
9429 if (name[4] == 'r' &&
9434 return -KEY_setgrent;
9440 if (name[4] == 'w' &&
9445 return -KEY_setpwent;
9461 if (name[3] == 'w' &&
9467 return -KEY_shmwrite;
9473 if (name[3] == 't' &&
9479 return -KEY_shutdown;
9489 if (name[2] == 's' &&
9496 return -KEY_syswrite;
9506 if (name[1] == 'r' &&
9514 return -KEY_truncate;
9523 case 9: /* 8 tokens of length 9 */
9527 if (name[1] == 'n' &&
9536 return -KEY_endnetent;
9542 if (name[1] == 'e' &&
9551 return -KEY_getnetent;
9557 if (name[1] == 'o' &&
9566 return -KEY_localtime;
9572 if (name[1] == 'r' &&
9581 return KEY_prototype;
9587 if (name[1] == 'u' &&
9596 return -KEY_quotemeta;
9602 if (name[1] == 'e' &&
9611 return -KEY_rewinddir;
9617 if (name[1] == 'e' &&
9626 return -KEY_setnetent;
9632 if (name[1] == 'a' &&
9641 return -KEY_wantarray;
9650 case 10: /* 9 tokens of length 10 */
9654 if (name[1] == 'n' &&
9660 if (name[4] == 'o' &&
9667 return -KEY_endhostent;
9673 if (name[4] == 'e' &&
9680 return -KEY_endservent;
9693 if (name[1] == 'e' &&
9699 if (name[4] == 'o' &&
9706 return -KEY_gethostent;
9715 if (name[5] == 'r' &&
9721 return -KEY_getservent;
9727 if (name[5] == 'c' &&
9733 return -KEY_getsockopt;
9758 if (name[4] == 'o' &&
9765 return -KEY_sethostent;
9774 if (name[5] == 'r' &&
9780 return -KEY_setservent;
9786 if (name[5] == 'c' &&
9792 return -KEY_setsockopt;
9809 if (name[2] == 'c' &&
9818 return -KEY_socketpair;
9831 case 11: /* 8 tokens of length 11 */
9835 if (name[1] == '_' &&
9846 return -KEY___PACKAGE__;
9852 if (name[1] == 'n' &&
9863 return -KEY_endprotoent;
9869 if (name[1] == 'e' &&
9878 if (name[5] == 'e' &&
9885 return -KEY_getpeername;
9894 if (name[6] == 'o' &&
9900 return -KEY_getpriority;
9906 if (name[6] == 't' &&
9912 return -KEY_getprotoent;
9926 if (name[4] == 'o' &&
9934 return -KEY_getsockname;
9947 if (name[1] == 'e' &&
9955 if (name[6] == 'o' &&
9961 return -KEY_setpriority;
9967 if (name[6] == 't' &&
9973 return -KEY_setprotoent;
9989 case 12: /* 2 tokens of length 12 */
9990 if (name[0] == 'g' &&
10002 if (name[9] == 'd' &&
10005 { /* getnetbyaddr */
10006 return -KEY_getnetbyaddr;
10012 if (name[9] == 'a' &&
10015 { /* getnetbyname */
10016 return -KEY_getnetbyname;
10028 case 13: /* 4 tokens of length 13 */
10029 if (name[0] == 'g' &&
10036 if (name[4] == 'o' &&
10045 if (name[10] == 'd' &&
10048 { /* gethostbyaddr */
10049 return -KEY_gethostbyaddr;
10055 if (name[10] == 'a' &&
10058 { /* gethostbyname */
10059 return -KEY_gethostbyname;
10072 if (name[4] == 'e' &&
10081 if (name[10] == 'a' &&
10084 { /* getservbyname */
10085 return -KEY_getservbyname;
10091 if (name[10] == 'o' &&
10094 { /* getservbyport */
10095 return -KEY_getservbyport;
10114 case 14: /* 1 tokens of length 14 */
10115 if (name[0] == 'g' &&
10129 { /* getprotobyname */
10130 return -KEY_getprotobyname;
10135 case 16: /* 1 tokens of length 16 */
10136 if (name[0] == 'g' &&
10152 { /* getprotobynumber */
10153 return -KEY_getprotobynumber;
10167 S_checkcomma(pTHX_ const char *s, const char *name, const char *what)
10171 if (*s == ' ' && s[1] == '(') { /* XXX gotta be a better way */
10172 if (ckWARN(WARN_SYNTAX)) {
10175 for (w = s+2; *w && level; w++) {
10178 else if (*w == ')')
10181 while (isSPACE(*w))
10183 if (!*w || !strchr(";|})]oaiuw!=", *w)) /* an advisory hack only... */
10184 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
10185 "%s (...) interpreted as function",name);
10188 while (s < PL_bufend && isSPACE(*s))
10192 while (s < PL_bufend && isSPACE(*s))
10194 if (isIDFIRST_lazy_if(s,UTF)) {
10195 const char * const w = s++;
10196 while (isALNUM_lazy_if(s,UTF))
10198 while (s < PL_bufend && isSPACE(*s))
10202 if (keyword(w, s - w))
10205 gv = gv_fetchpvn_flags(w, s - w, 0, SVt_PVCV);
10206 if (gv && GvCVu(gv))
10208 Perl_croak(aTHX_ "No comma allowed after %s", what);
10213 /* Either returns sv, or mortalizes sv and returns a new SV*.
10214 Best used as sv=new_constant(..., sv, ...).
10215 If s, pv are NULL, calls subroutine with one argument,
10216 and type is used with error messages only. */
10219 S_new_constant(pTHX_ const char *s, STRLEN len, const char *key, SV *sv, SV *pv,
10223 HV * const table = GvHV(PL_hintgv); /* ^H */
10227 const char *why1 = "", *why2 = "", *why3 = "";
10229 if (!table || !(PL_hints & HINT_LOCALIZE_HH)) {
10232 why2 = strEQ(key,"charnames")
10233 ? "(possibly a missing \"use charnames ...\")"
10235 msg = Perl_newSVpvf(aTHX_ "Constant(%s) unknown: %s",
10236 (type ? type: "undef"), why2);
10238 /* This is convoluted and evil ("goto considered harmful")
10239 * but I do not understand the intricacies of all the different
10240 * failure modes of %^H in here. The goal here is to make
10241 * the most probable error message user-friendly. --jhi */
10246 msg = Perl_newSVpvf(aTHX_ "Constant(%s): %s%s%s",
10247 (type ? type: "undef"), why1, why2, why3);
10249 yyerror(SvPVX_const(msg));
10253 cvp = hv_fetch(table, key, strlen(key), FALSE);
10254 if (!cvp || !SvOK(*cvp)) {
10257 why3 = "} is not defined";
10260 sv_2mortal(sv); /* Parent created it permanently */
10263 pv = sv_2mortal(newSVpvn(s, len));
10265 typesv = sv_2mortal(newSVpv(type, 0));
10267 typesv = &PL_sv_undef;
10269 PUSHSTACKi(PERLSI_OVERLOAD);
10281 call_sv(cv, G_SCALAR | ( PL_in_eval ? 0 : G_EVAL));
10285 /* Check the eval first */
10286 if (!PL_in_eval && SvTRUE(ERRSV)) {
10287 sv_catpvs(ERRSV, "Propagated");
10288 yyerror(SvPV_nolen_const(ERRSV)); /* Duplicates the message inside eval */
10290 res = SvREFCNT_inc_simple(sv);
10294 SvREFCNT_inc_simple_void(res);
10303 why1 = "Call to &{$^H{";
10305 why3 = "}} did not return a defined value";
10313 /* Returns a NUL terminated string, with the length of the string written to
10317 S_scan_word(pTHX_ register char *s, char *dest, STRLEN destlen, int allow_package, STRLEN *slp)
10320 register char *d = dest;
10321 register char * const e = d + destlen - 3; /* two-character token, ending NUL */
10324 Perl_croak(aTHX_ ident_too_long);
10325 if (isALNUM(*s)) /* UTF handled below */
10327 else if (*s == '\'' && allow_package && isIDFIRST_lazy_if(s+1,UTF)) {
10332 else if (*s == ':' && s[1] == ':' && allow_package && s[2] != '$') {
10336 else if (UTF && UTF8_IS_START(*s) && isALNUM_utf8((U8*)s)) {
10337 char *t = s + UTF8SKIP(s);
10338 while (UTF8_IS_CONTINUED(*t) && is_utf8_mark((U8*)t))
10340 if (d + (t - s) > e)
10341 Perl_croak(aTHX_ ident_too_long);
10342 Copy(s, d, t - s, char);
10355 S_scan_ident(pTHX_ register char *s, register const char *send, char *dest, STRLEN destlen, I32 ck_uni)
10358 char *bracket = NULL;
10360 register char *d = dest;
10361 register char * const e = d + destlen + 3; /* two-character token, ending NUL */
10366 while (isDIGIT(*s)) {
10368 Perl_croak(aTHX_ ident_too_long);
10375 Perl_croak(aTHX_ ident_too_long);
10376 if (isALNUM(*s)) /* UTF handled below */
10378 else if (*s == '\'' && isIDFIRST_lazy_if(s+1,UTF)) {
10383 else if (*s == ':' && s[1] == ':') {
10387 else if (UTF && UTF8_IS_START(*s) && isALNUM_utf8((U8*)s)) {
10388 char *t = s + UTF8SKIP(s);
10389 while (UTF8_IS_CONTINUED(*t) && is_utf8_mark((U8*)t))
10391 if (d + (t - s) > e)
10392 Perl_croak(aTHX_ ident_too_long);
10393 Copy(s, d, t - s, char);
10404 if (PL_lex_state != LEX_NORMAL)
10405 PL_lex_state = LEX_INTERPENDMAYBE;
10408 if (*s == '$' && s[1] &&
10409 (isALNUM_lazy_if(s+1,UTF) || s[1] == '$' || s[1] == '{' || strnEQ(s+1,"::",2)) )
10422 if (*d == '^' && *s && isCONTROLVAR(*s)) {
10427 if (isSPACE(s[-1])) {
10429 const char ch = *s++;
10430 if (!SPACE_OR_TAB(ch)) {
10436 if (isIDFIRST_lazy_if(d,UTF)) {
10440 while ((end < send && isALNUM_lazy_if(end,UTF)) || *end == ':') {
10441 end += UTF8SKIP(end);
10442 while (end < send && UTF8_IS_CONTINUED(*end) && is_utf8_mark((U8*)end))
10443 end += UTF8SKIP(end);
10445 Copy(s, d, end - s, char);
10450 while ((isALNUM(*s) || *s == ':') && d < e)
10453 Perl_croak(aTHX_ ident_too_long);
10456 while (s < send && SPACE_OR_TAB(*s)) s++;
10457 if ((*s == '[' || (*s == '{' && strNE(dest, "sub")))) {
10458 if (ckWARN(WARN_AMBIGUOUS) && keyword(dest, d - dest)) {
10459 const char *brack = *s == '[' ? "[...]" : "{...}";
10460 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
10461 "Ambiguous use of %c{%s%s} resolved to %c%s%s",
10462 funny, dest, brack, funny, dest, brack);
10465 PL_lex_brackstack[PL_lex_brackets++] = (char)(XOPERATOR | XFAKEBRACK);
10469 /* Handle extended ${^Foo} variables
10470 * 1999-02-27 mjd-perl-patch@plover.com */
10471 else if (!isALNUM(*d) && !isPRINT(*d) /* isCTRL(d) */
10475 while (isALNUM(*s) && d < e) {
10479 Perl_croak(aTHX_ ident_too_long);
10484 if (PL_lex_state == LEX_INTERPNORMAL && !PL_lex_brackets) {
10485 PL_lex_state = LEX_INTERPEND;
10490 if (PL_lex_state == LEX_NORMAL) {
10491 if (ckWARN(WARN_AMBIGUOUS) &&
10492 (keyword(dest, d - dest) || get_cv(dest, FALSE)))
10494 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
10495 "Ambiguous use of %c{%s} resolved to %c%s",
10496 funny, dest, funny, dest);
10501 s = bracket; /* let the parser handle it */
10505 else if (PL_lex_state == LEX_INTERPNORMAL && !PL_lex_brackets && !intuit_more(s))
10506 PL_lex_state = LEX_INTERPEND;
10511 Perl_pmflag(pTHX_ U32* pmfl, int ch)
10513 PERL_UNUSED_CONTEXT;
10516 else if (ch == 'g')
10517 *pmfl |= PMf_GLOBAL;
10518 else if (ch == 'c')
10519 *pmfl |= PMf_CONTINUE;
10520 else if (ch == 'o')
10522 else if (ch == 'm')
10523 *pmfl |= PMf_MULTILINE;
10524 else if (ch == 's')
10525 *pmfl |= PMf_SINGLELINE;
10526 else if (ch == 'x')
10527 *pmfl |= PMf_EXTENDED;
10531 S_scan_pat(pTHX_ char *start, I32 type)
10535 char *s = scan_str(start,!!PL_madskills,FALSE);
10536 const char * const valid_flags = (type == OP_QR) ? "iomsx" : "iogcmsx";
10543 const char * const delimiter = skipspace(start);
10544 Perl_croak(aTHX_ *delimiter == '?'
10545 ? "Search pattern not terminated or ternary operator parsed as search pattern"
10546 : "Search pattern not terminated" );
10549 pm = (PMOP*)newPMOP(type, 0);
10550 if (PL_multi_open == '?')
10551 pm->op_pmflags |= PMf_ONCE;
10555 while (*s && strchr(valid_flags, *s))
10556 pmflag(&pm->op_pmflags,*s++);
10558 if (PL_madskills && modstart != s) {
10559 SV* tmptoken = newSVpvn(modstart, s - modstart);
10560 append_madprops(newMADPROP('m', MAD_SV, tmptoken, 0), (OP*)pm, 0);
10563 /* issue a warning if /c is specified,but /g is not */
10564 if ((pm->op_pmflags & PMf_CONTINUE) && !(pm->op_pmflags & PMf_GLOBAL)
10565 && ckWARN(WARN_REGEXP))
10567 Perl_warner(aTHX_ packWARN(WARN_REGEXP), "Use of /c modifier is meaningless without /g" );
10570 pm->op_pmpermflags = pm->op_pmflags;
10572 PL_lex_op = (OP*)pm;
10573 yylval.ival = OP_MATCH;
10578 S_scan_subst(pTHX_ char *start)
10589 yylval.ival = OP_NULL;
10591 s = scan_str(start,!!PL_madskills,FALSE);
10594 Perl_croak(aTHX_ "Substitution pattern not terminated");
10596 if (s[-1] == PL_multi_open)
10599 if (PL_madskills) {
10600 CURMAD('q', PL_thisopen);
10601 CURMAD('_', PL_thiswhite);
10602 CURMAD('E', PL_thisstuff);
10603 CURMAD('Q', PL_thisclose);
10604 PL_realtokenstart = s - SvPVX(PL_linestr);
10608 first_start = PL_multi_start;
10609 s = scan_str(s,!!PL_madskills,FALSE);
10611 if (PL_lex_stuff) {
10612 SvREFCNT_dec(PL_lex_stuff);
10613 PL_lex_stuff = NULL;
10615 Perl_croak(aTHX_ "Substitution replacement not terminated");
10617 PL_multi_start = first_start; /* so whole substitution is taken together */
10619 pm = (PMOP*)newPMOP(OP_SUBST, 0);
10622 if (PL_madskills) {
10623 CURMAD('z', PL_thisopen);
10624 CURMAD('R', PL_thisstuff);
10625 CURMAD('Z', PL_thisclose);
10635 else if (strchr("iogcmsx", *s))
10636 pmflag(&pm->op_pmflags,*s++);
10642 if (PL_madskills) {
10644 curmad('m', newSVpvn(modstart, s - modstart));
10645 append_madprops(PL_thismad, (OP*)pm, 0);
10649 if ((pm->op_pmflags & PMf_CONTINUE) && ckWARN(WARN_REGEXP)) {
10650 Perl_warner(aTHX_ packWARN(WARN_REGEXP), "Use of /c modifier is meaningless in s///" );
10654 SV * const repl = newSVpvs("");
10656 PL_sublex_info.super_bufptr = s;
10657 PL_sublex_info.super_bufend = PL_bufend;
10659 pm->op_pmflags |= PMf_EVAL;
10661 sv_catpv(repl, es ? "eval " : "do ");
10662 sv_catpvs(repl, "{");
10663 sv_catsv(repl, PL_lex_repl);
10664 sv_catpvs(repl, "}");
10666 SvREFCNT_dec(PL_lex_repl);
10667 PL_lex_repl = repl;
10670 pm->op_pmpermflags = pm->op_pmflags;
10671 PL_lex_op = (OP*)pm;
10672 yylval.ival = OP_SUBST;
10677 S_scan_trans(pTHX_ char *start)
10690 yylval.ival = OP_NULL;
10692 s = scan_str(start,!!PL_madskills,FALSE);
10694 Perl_croak(aTHX_ "Transliteration pattern not terminated");
10696 if (s[-1] == PL_multi_open)
10699 if (PL_madskills) {
10700 CURMAD('q', PL_thisopen);
10701 CURMAD('_', PL_thiswhite);
10702 CURMAD('E', PL_thisstuff);
10703 CURMAD('Q', PL_thisclose);
10704 PL_realtokenstart = s - SvPVX(PL_linestr);
10708 s = scan_str(s,!!PL_madskills,FALSE);
10710 if (PL_lex_stuff) {
10711 SvREFCNT_dec(PL_lex_stuff);
10712 PL_lex_stuff = NULL;
10714 Perl_croak(aTHX_ "Transliteration replacement not terminated");
10716 if (PL_madskills) {
10717 CURMAD('z', PL_thisopen);
10718 CURMAD('R', PL_thisstuff);
10719 CURMAD('Z', PL_thisclose);
10722 complement = del = squash = 0;
10729 complement = OPpTRANS_COMPLEMENT;
10732 del = OPpTRANS_DELETE;
10735 squash = OPpTRANS_SQUASH;
10744 Newx(tbl, complement&&!del?258:256, short);
10745 o = newPVOP(OP_TRANS, 0, (char*)tbl);
10746 o->op_private &= ~OPpTRANS_ALL;
10747 o->op_private |= del|squash|complement|
10748 (DO_UTF8(PL_lex_stuff)? OPpTRANS_FROM_UTF : 0)|
10749 (DO_UTF8(PL_lex_repl) ? OPpTRANS_TO_UTF : 0);
10752 yylval.ival = OP_TRANS;
10755 if (PL_madskills) {
10757 curmad('m', newSVpvn(modstart, s - modstart));
10758 append_madprops(PL_thismad, o, 0);
10767 S_scan_heredoc(pTHX_ register char *s)
10771 I32 op_type = OP_SCALAR;
10775 const char *found_newline;
10779 const int outer = (PL_rsfp && !(PL_lex_inwhat == OP_SCALAR));
10781 I32 stuffstart = s - SvPVX(PL_linestr);
10784 PL_realtokenstart = -1;
10789 e = PL_tokenbuf + sizeof PL_tokenbuf - 1;
10792 for (peek = s; SPACE_OR_TAB(*peek); peek++) ;
10793 if (*peek == '`' || *peek == '\'' || *peek =='"') {
10796 s = delimcpy(d, e, s, PL_bufend, term, &len);
10806 if (!isALNUM_lazy_if(s,UTF))
10807 deprecate_old("bare << to mean <<\"\"");
10808 for (; isALNUM_lazy_if(s,UTF); s++) {
10813 if (d >= PL_tokenbuf + sizeof PL_tokenbuf - 1)
10814 Perl_croak(aTHX_ "Delimiter for here document is too long");
10817 len = d - PL_tokenbuf;
10820 if (PL_madskills) {
10821 tstart = PL_tokenbuf + !outer;
10822 PL_thisclose = newSVpvn(tstart, len - !outer);
10823 tstart = SvPVX(PL_linestr) + stuffstart;
10824 PL_thisopen = newSVpvn(tstart, s - tstart);
10825 stuffstart = s - SvPVX(PL_linestr);
10828 #ifndef PERL_STRICT_CR
10829 d = strchr(s, '\r');
10831 char * const olds = s;
10833 while (s < PL_bufend) {
10839 else if (*s == '\n' && s[1] == '\r') { /* \015\013 on a mac? */
10848 SvCUR_set(PL_linestr, PL_bufend - SvPVX_const(PL_linestr));
10855 if ( outer || !(found_newline = memchr(s, '\n', PL_bufend - s)) ) {
10856 herewas = newSVpvn(s,PL_bufend-s);
10860 herewas = newSVpvn(s-1,found_newline-s+1);
10863 herewas = newSVpvn(s,found_newline-s);
10867 if (PL_madskills) {
10868 tstart = SvPVX(PL_linestr) + stuffstart;
10870 sv_catpvn(PL_thisstuff, tstart, s - tstart);
10872 PL_thisstuff = newSVpvn(tstart, s - tstart);
10875 s += SvCUR(herewas);
10878 stuffstart = s - SvPVX(PL_linestr);
10884 tmpstr = newSV(79);
10885 sv_upgrade(tmpstr, SVt_PVIV);
10886 if (term == '\'') {
10887 op_type = OP_CONST;
10888 SvIV_set(tmpstr, -1);
10890 else if (term == '`') {
10891 op_type = OP_BACKTICK;
10892 SvIV_set(tmpstr, '\\');
10896 PL_multi_start = CopLINE(PL_curcop);
10897 PL_multi_open = PL_multi_close = '<';
10898 term = *PL_tokenbuf;
10899 if (PL_lex_inwhat == OP_SUBST && PL_in_eval && !PL_rsfp) {
10900 char * const bufptr = PL_sublex_info.super_bufptr;
10901 char * const bufend = PL_sublex_info.super_bufend;
10902 char * const olds = s - SvCUR(herewas);
10903 s = strchr(bufptr, '\n');
10907 while (s < bufend &&
10908 (*s != term || memNE(s,PL_tokenbuf,len)) ) {
10910 CopLINE_inc(PL_curcop);
10913 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
10914 missingterm(PL_tokenbuf);
10916 sv_setpvn(herewas,bufptr,d-bufptr+1);
10917 sv_setpvn(tmpstr,d+1,s-d);
10919 sv_catpvn(herewas,s,bufend-s);
10920 Copy(SvPVX_const(herewas),bufptr,SvCUR(herewas) + 1,char);
10927 while (s < PL_bufend &&
10928 (*s != term || memNE(s,PL_tokenbuf,len)) ) {
10930 CopLINE_inc(PL_curcop);
10932 if (s >= PL_bufend) {
10933 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
10934 missingterm(PL_tokenbuf);
10936 sv_setpvn(tmpstr,d+1,s-d);
10938 if (PL_madskills) {
10940 sv_catpvn(PL_thisstuff, d + 1, s - d);
10942 PL_thisstuff = newSVpvn(d + 1, s - d);
10943 stuffstart = s - SvPVX(PL_linestr);
10947 CopLINE_inc(PL_curcop); /* the preceding stmt passes a newline */
10949 sv_catpvn(herewas,s,PL_bufend-s);
10950 sv_setsv(PL_linestr,herewas);
10951 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = s = PL_linestart = SvPVX(PL_linestr);
10952 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
10953 PL_last_lop = PL_last_uni = NULL;
10956 sv_setpvn(tmpstr,"",0); /* avoid "uninitialized" warning */
10957 while (s >= PL_bufend) { /* multiple line string? */
10959 if (PL_madskills) {
10960 tstart = SvPVX(PL_linestr) + stuffstart;
10962 sv_catpvn(PL_thisstuff, tstart, PL_bufend - tstart);
10964 PL_thisstuff = newSVpvn(tstart, PL_bufend - tstart);
10968 !(PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = filter_gets(PL_linestr, PL_rsfp, 0))) {
10969 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
10970 missingterm(PL_tokenbuf);
10973 stuffstart = s - SvPVX(PL_linestr);
10975 CopLINE_inc(PL_curcop);
10976 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
10977 PL_last_lop = PL_last_uni = NULL;
10978 #ifndef PERL_STRICT_CR
10979 if (PL_bufend - PL_linestart >= 2) {
10980 if ((PL_bufend[-2] == '\r' && PL_bufend[-1] == '\n') ||
10981 (PL_bufend[-2] == '\n' && PL_bufend[-1] == '\r'))
10983 PL_bufend[-2] = '\n';
10985 SvCUR_set(PL_linestr, PL_bufend - SvPVX_const(PL_linestr));
10987 else if (PL_bufend[-1] == '\r')
10988 PL_bufend[-1] = '\n';
10990 else if (PL_bufend - PL_linestart == 1 && PL_bufend[-1] == '\r')
10991 PL_bufend[-1] = '\n';
10993 if (PERLDB_LINE && PL_curstash != PL_debstash) {
10994 SV * const sv = newSV(0);
10996 sv_upgrade(sv, SVt_PVMG);
10997 sv_setsv(sv,PL_linestr);
10998 (void)SvIOK_on(sv);
11000 av_store(CopFILEAVx(PL_curcop), (I32)CopLINE(PL_curcop),sv);
11002 if (*s == term && memEQ(s,PL_tokenbuf,len)) {
11003 STRLEN off = PL_bufend - 1 - SvPVX_const(PL_linestr);
11004 *(SvPVX(PL_linestr) + off ) = ' ';
11005 sv_catsv(PL_linestr,herewas);
11006 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
11007 s = SvPVX(PL_linestr) + off; /* In case PV of PL_linestr moved. */
11011 sv_catsv(tmpstr,PL_linestr);
11016 PL_multi_end = CopLINE(PL_curcop);
11017 if (SvCUR(tmpstr) + 5 < SvLEN(tmpstr)) {
11018 SvPV_shrink_to_cur(tmpstr);
11020 SvREFCNT_dec(herewas);
11022 if (UTF && is_utf8_string((U8*)SvPVX_const(tmpstr), SvCUR(tmpstr)))
11024 else if (PL_encoding)
11025 sv_recode_to_utf8(tmpstr, PL_encoding);
11027 PL_lex_stuff = tmpstr;
11028 yylval.ival = op_type;
11032 /* scan_inputsymbol
11033 takes: current position in input buffer
11034 returns: new position in input buffer
11035 side-effects: yylval and lex_op are set.
11040 <FH> read from filehandle
11041 <pkg::FH> read from package qualified filehandle
11042 <pkg'FH> read from package qualified filehandle
11043 <$fh> read from filehandle in $fh
11044 <*.h> filename glob
11049 S_scan_inputsymbol(pTHX_ char *start)
11052 register char *s = start; /* current position in buffer */
11056 char *d = PL_tokenbuf; /* start of temp holding space */
11057 const char * const e = PL_tokenbuf + sizeof PL_tokenbuf; /* end of temp holding space */
11059 end = strchr(s, '\n');
11062 s = delimcpy(d, e, s + 1, end, '>', &len); /* extract until > */
11064 /* die if we didn't have space for the contents of the <>,
11065 or if it didn't end, or if we see a newline
11068 if (len >= sizeof PL_tokenbuf)
11069 Perl_croak(aTHX_ "Excessively long <> operator");
11071 Perl_croak(aTHX_ "Unterminated <> operator");
11076 Remember, only scalar variables are interpreted as filehandles by
11077 this code. Anything more complex (e.g., <$fh{$num}>) will be
11078 treated as a glob() call.
11079 This code makes use of the fact that except for the $ at the front,
11080 a scalar variable and a filehandle look the same.
11082 if (*d == '$' && d[1]) d++;
11084 /* allow <Pkg'VALUE> or <Pkg::VALUE> */
11085 while (*d && (isALNUM_lazy_if(d,UTF) || *d == '\'' || *d == ':'))
11088 /* If we've tried to read what we allow filehandles to look like, and
11089 there's still text left, then it must be a glob() and not a getline.
11090 Use scan_str to pull out the stuff between the <> and treat it
11091 as nothing more than a string.
11094 if (d - PL_tokenbuf != len) {
11095 yylval.ival = OP_GLOB;
11097 s = scan_str(start,!!PL_madskills,FALSE);
11099 Perl_croak(aTHX_ "Glob not terminated");
11103 bool readline_overriden = FALSE;
11106 /* we're in a filehandle read situation */
11109 /* turn <> into <ARGV> */
11111 Copy("ARGV",d,5,char);
11113 /* Check whether readline() is overriden */
11114 gv_readline = gv_fetchpvs("readline", GV_NOTQUAL, SVt_PVCV);
11116 && GvCVu(gv_readline) && GvIMPORTED_CV(gv_readline))
11118 ((gvp = (GV**)hv_fetchs(PL_globalstash, "readline", FALSE))
11119 && (gv_readline = *gvp) != (GV*)&PL_sv_undef
11120 && GvCVu(gv_readline) && GvIMPORTED_CV(gv_readline)))
11121 readline_overriden = TRUE;
11123 /* if <$fh>, create the ops to turn the variable into a
11129 /* try to find it in the pad for this block, otherwise find
11130 add symbol table ops
11132 if ((tmp = pad_findmy(d)) != NOT_IN_PAD) {
11133 if (PAD_COMPNAME_FLAGS_isOUR(tmp)) {
11134 HV * const stash = PAD_COMPNAME_OURSTASH(tmp);
11135 HEK * const stashname = HvNAME_HEK(stash);
11136 SV * const sym = sv_2mortal(newSVhek(stashname));
11137 sv_catpvs(sym, "::");
11138 sv_catpv(sym, d+1);
11143 OP * const o = newOP(OP_PADSV, 0);
11145 PL_lex_op = readline_overriden
11146 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
11147 append_elem(OP_LIST, o,
11148 newCVREF(0, newGVOP(OP_GV,0,gv_readline))))
11149 : (OP*)newUNOP(OP_READLINE, 0, o);
11158 ? (GV_ADDMULTI | GV_ADDINEVAL)
11161 PL_lex_op = readline_overriden
11162 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
11163 append_elem(OP_LIST,
11164 newUNOP(OP_RV2SV, 0, newGVOP(OP_GV, 0, gv)),
11165 newCVREF(0, newGVOP(OP_GV, 0, gv_readline))))
11166 : (OP*)newUNOP(OP_READLINE, 0,
11167 newUNOP(OP_RV2SV, 0,
11168 newGVOP(OP_GV, 0, gv)));
11170 if (!readline_overriden)
11171 PL_lex_op->op_flags |= OPf_SPECIAL;
11172 /* we created the ops in PL_lex_op, so make yylval.ival a null op */
11173 yylval.ival = OP_NULL;
11176 /* If it's none of the above, it must be a literal filehandle
11177 (<Foo::BAR> or <FOO>) so build a simple readline OP */
11179 GV * const gv = gv_fetchpv(d, GV_ADD, SVt_PVIO);
11180 PL_lex_op = readline_overriden
11181 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
11182 append_elem(OP_LIST,
11183 newGVOP(OP_GV, 0, gv),
11184 newCVREF(0, newGVOP(OP_GV, 0, gv_readline))))
11185 : (OP*)newUNOP(OP_READLINE, 0, newGVOP(OP_GV, 0, gv));
11186 yylval.ival = OP_NULL;
11195 takes: start position in buffer
11196 keep_quoted preserve \ on the embedded delimiter(s)
11197 keep_delims preserve the delimiters around the string
11198 returns: position to continue reading from buffer
11199 side-effects: multi_start, multi_close, lex_repl or lex_stuff, and
11200 updates the read buffer.
11202 This subroutine pulls a string out of the input. It is called for:
11203 q single quotes q(literal text)
11204 ' single quotes 'literal text'
11205 qq double quotes qq(interpolate $here please)
11206 " double quotes "interpolate $here please"
11207 qx backticks qx(/bin/ls -l)
11208 ` backticks `/bin/ls -l`
11209 qw quote words @EXPORT_OK = qw( func() $spam )
11210 m// regexp match m/this/
11211 s/// regexp substitute s/this/that/
11212 tr/// string transliterate tr/this/that/
11213 y/// string transliterate y/this/that/
11214 ($*@) sub prototypes sub foo ($)
11215 (stuff) sub attr parameters sub foo : attr(stuff)
11216 <> readline or globs <FOO>, <>, <$fh>, or <*.c>
11218 In most of these cases (all but <>, patterns and transliterate)
11219 yylex() calls scan_str(). m// makes yylex() call scan_pat() which
11220 calls scan_str(). s/// makes yylex() call scan_subst() which calls
11221 scan_str(). tr/// and y/// make yylex() call scan_trans() which
11224 It skips whitespace before the string starts, and treats the first
11225 character as the delimiter. If the delimiter is one of ([{< then
11226 the corresponding "close" character )]}> is used as the closing
11227 delimiter. It allows quoting of delimiters, and if the string has
11228 balanced delimiters ([{<>}]) it allows nesting.
11230 On success, the SV with the resulting string is put into lex_stuff or,
11231 if that is already non-NULL, into lex_repl. The second case occurs only
11232 when parsing the RHS of the special constructs s/// and tr/// (y///).
11233 For convenience, the terminating delimiter character is stuffed into
11238 S_scan_str(pTHX_ char *start, int keep_quoted, int keep_delims)
11241 SV *sv; /* scalar value: string */
11242 char *tmps; /* temp string, used for delimiter matching */
11243 register char *s = start; /* current position in the buffer */
11244 register char term; /* terminating character */
11245 register char *to; /* current position in the sv's data */
11246 I32 brackets = 1; /* bracket nesting level */
11247 bool has_utf8 = FALSE; /* is there any utf8 content? */
11248 I32 termcode; /* terminating char. code */
11249 U8 termstr[UTF8_MAXBYTES]; /* terminating string */
11250 STRLEN termlen; /* length of terminating string */
11251 char *last = NULL; /* last position for nesting bracket */
11257 /* skip space before the delimiter */
11263 if (PL_realtokenstart >= 0) {
11264 stuffstart = PL_realtokenstart;
11265 PL_realtokenstart = -1;
11268 stuffstart = start - SvPVX(PL_linestr);
11270 /* mark where we are, in case we need to report errors */
11273 /* after skipping whitespace, the next character is the terminator */
11276 termcode = termstr[0] = term;
11280 termcode = utf8_to_uvchr((U8*)s, &termlen);
11281 Copy(s, termstr, termlen, U8);
11282 if (!UTF8_IS_INVARIANT(term))
11286 /* mark where we are */
11287 PL_multi_start = CopLINE(PL_curcop);
11288 PL_multi_open = term;
11290 /* find corresponding closing delimiter */
11291 if (term && (tmps = strchr("([{< )]}> )]}>",term)))
11292 termcode = termstr[0] = term = tmps[5];
11294 PL_multi_close = term;
11296 /* create a new SV to hold the contents. 79 is the SV's initial length.
11297 What a random number. */
11299 sv_upgrade(sv, SVt_PVIV);
11300 SvIV_set(sv, termcode);
11301 (void)SvPOK_only(sv); /* validate pointer */
11303 /* move past delimiter and try to read a complete string */
11305 sv_catpvn(sv, s, termlen);
11308 tstart = SvPVX(PL_linestr) + stuffstart;
11309 if (!PL_thisopen && !keep_delims) {
11310 PL_thisopen = newSVpvn(tstart, s - tstart);
11311 stuffstart = s - SvPVX(PL_linestr);
11315 if (PL_encoding && !UTF) {
11319 int offset = s - SvPVX_const(PL_linestr);
11320 const bool found = sv_cat_decode(sv, PL_encoding, PL_linestr,
11321 &offset, (char*)termstr, termlen);
11322 const char * const ns = SvPVX_const(PL_linestr) + offset;
11323 char * const svlast = SvEND(sv) - 1;
11325 for (; s < ns; s++) {
11326 if (*s == '\n' && !PL_rsfp)
11327 CopLINE_inc(PL_curcop);
11330 goto read_more_line;
11332 /* handle quoted delimiters */
11333 if (SvCUR(sv) > 1 && *(svlast-1) == '\\') {
11335 for (t = svlast-2; t >= SvPVX_const(sv) && *t == '\\';)
11337 if ((svlast-1 - t) % 2) {
11338 if (!keep_quoted) {
11339 *(svlast-1) = term;
11341 SvCUR_set(sv, SvCUR(sv) - 1);
11346 if (PL_multi_open == PL_multi_close) {
11354 for (t = w = last; t < svlast; w++, t++) {
11355 /* At here, all closes are "was quoted" one,
11356 so we don't check PL_multi_close. */
11358 if (!keep_quoted && *(t+1) == PL_multi_open)
11363 else if (*t == PL_multi_open)
11371 SvCUR_set(sv, w - SvPVX_const(sv));
11374 if (--brackets <= 0)
11379 if (!keep_delims) {
11380 SvCUR_set(sv, SvCUR(sv) - 1);
11386 /* extend sv if need be */
11387 SvGROW(sv, SvCUR(sv) + (PL_bufend - s) + 1);
11388 /* set 'to' to the next character in the sv's string */
11389 to = SvPVX(sv)+SvCUR(sv);
11391 /* if open delimiter is the close delimiter read unbridle */
11392 if (PL_multi_open == PL_multi_close) {
11393 for (; s < PL_bufend; s++,to++) {
11394 /* embedded newlines increment the current line number */
11395 if (*s == '\n' && !PL_rsfp)
11396 CopLINE_inc(PL_curcop);
11397 /* handle quoted delimiters */
11398 if (*s == '\\' && s+1 < PL_bufend && term != '\\') {
11399 if (!keep_quoted && s[1] == term)
11401 /* any other quotes are simply copied straight through */
11405 /* terminate when run out of buffer (the for() condition), or
11406 have found the terminator */
11407 else if (*s == term) {
11410 if (s+termlen <= PL_bufend && memEQ(s, (char*)termstr, termlen))
11413 else if (!has_utf8 && !UTF8_IS_INVARIANT((U8)*s) && UTF)
11419 /* if the terminator isn't the same as the start character (e.g.,
11420 matched brackets), we have to allow more in the quoting, and
11421 be prepared for nested brackets.
11424 /* read until we run out of string, or we find the terminator */
11425 for (; s < PL_bufend; s++,to++) {
11426 /* embedded newlines increment the line count */
11427 if (*s == '\n' && !PL_rsfp)
11428 CopLINE_inc(PL_curcop);
11429 /* backslashes can escape the open or closing characters */
11430 if (*s == '\\' && s+1 < PL_bufend) {
11431 if (!keep_quoted &&
11432 ((s[1] == PL_multi_open) || (s[1] == PL_multi_close)))
11437 /* allow nested opens and closes */
11438 else if (*s == PL_multi_close && --brackets <= 0)
11440 else if (*s == PL_multi_open)
11442 else if (!has_utf8 && !UTF8_IS_INVARIANT((U8)*s) && UTF)
11447 /* terminate the copied string and update the sv's end-of-string */
11449 SvCUR_set(sv, to - SvPVX_const(sv));
11452 * this next chunk reads more into the buffer if we're not done yet
11456 break; /* handle case where we are done yet :-) */
11458 #ifndef PERL_STRICT_CR
11459 if (to - SvPVX_const(sv) >= 2) {
11460 if ((to[-2] == '\r' && to[-1] == '\n') ||
11461 (to[-2] == '\n' && to[-1] == '\r'))
11465 SvCUR_set(sv, to - SvPVX_const(sv));
11467 else if (to[-1] == '\r')
11470 else if (to - SvPVX_const(sv) == 1 && to[-1] == '\r')
11475 /* if we're out of file, or a read fails, bail and reset the current
11476 line marker so we can report where the unterminated string began
11479 if (PL_madskills) {
11480 char *tstart = SvPVX(PL_linestr) + stuffstart;
11482 sv_catpvn(PL_thisstuff, tstart, PL_bufend - tstart);
11484 PL_thisstuff = newSVpvn(tstart, PL_bufend - tstart);
11488 !(PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = filter_gets(PL_linestr, PL_rsfp, 0))) {
11490 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
11496 /* we read a line, so increment our line counter */
11497 CopLINE_inc(PL_curcop);
11499 /* update debugger info */
11500 if (PERLDB_LINE && PL_curstash != PL_debstash) {
11501 SV * const line_sv = newSV(0);
11503 sv_upgrade(line_sv, SVt_PVMG);
11504 sv_setsv(line_sv,PL_linestr);
11505 (void)SvIOK_on(line_sv);
11506 SvIV_set(line_sv, 0);
11507 av_store(CopFILEAVx(PL_curcop), (I32)CopLINE(PL_curcop), line_sv);
11510 /* having changed the buffer, we must update PL_bufend */
11511 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
11512 PL_last_lop = PL_last_uni = NULL;
11515 /* at this point, we have successfully read the delimited string */
11517 if (!PL_encoding || UTF) {
11519 if (PL_madskills) {
11520 char *tstart = SvPVX(PL_linestr) + stuffstart;
11522 sv_catpvn(PL_thisstuff, tstart, s - tstart);
11524 PL_thisstuff = newSVpvn(tstart, s - tstart);
11525 if (!PL_thisclose && !keep_delims)
11526 PL_thisclose = newSVpvn(s,termlen);
11531 sv_catpvn(sv, s, termlen);
11536 if (PL_madskills) {
11537 char *tstart = SvPVX(PL_linestr) + stuffstart;
11539 sv_catpvn(PL_thisstuff, tstart, s - tstart - termlen);
11541 PL_thisstuff = newSVpvn(tstart, s - tstart - termlen);
11542 if (!PL_thisclose && !keep_delims)
11543 PL_thisclose = newSVpvn(s - termlen,termlen);
11547 if (has_utf8 || PL_encoding)
11550 PL_multi_end = CopLINE(PL_curcop);
11552 /* if we allocated too much space, give some back */
11553 if (SvCUR(sv) + 5 < SvLEN(sv)) {
11554 SvLEN_set(sv, SvCUR(sv) + 1);
11555 SvPV_renew(sv, SvLEN(sv));
11558 /* decide whether this is the first or second quoted string we've read
11571 takes: pointer to position in buffer
11572 returns: pointer to new position in buffer
11573 side-effects: builds ops for the constant in yylval.op
11575 Read a number in any of the formats that Perl accepts:
11577 \d(_?\d)*(\.(\d(_?\d)*)?)?[Ee][\+\-]?(\d(_?\d)*) 12 12.34 12.
11578 \.\d(_?\d)*[Ee][\+\-]?(\d(_?\d)*) .34
11581 0x[0-9A-Fa-f](_?[0-9A-Fa-f])*
11583 Like most scan_ routines, it uses the PL_tokenbuf buffer to hold the
11586 If it reads a number without a decimal point or an exponent, it will
11587 try converting the number to an integer and see if it can do so
11588 without loss of precision.
11592 Perl_scan_num(pTHX_ const char *start, YYSTYPE* lvalp)
11595 register const char *s = start; /* current position in buffer */
11596 register char *d; /* destination in temp buffer */
11597 register char *e; /* end of temp buffer */
11598 NV nv; /* number read, as a double */
11599 SV *sv = NULL; /* place to put the converted number */
11600 bool floatit; /* boolean: int or float? */
11601 const char *lastub = NULL; /* position of last underbar */
11602 static char const number_too_long[] = "Number too long";
11604 /* We use the first character to decide what type of number this is */
11608 Perl_croak(aTHX_ "panic: scan_num");
11610 /* if it starts with a 0, it could be an octal number, a decimal in
11611 0.13 disguise, or a hexadecimal number, or a binary number. */
11615 u holds the "number so far"
11616 shift the power of 2 of the base
11617 (hex == 4, octal == 3, binary == 1)
11618 overflowed was the number more than we can hold?
11620 Shift is used when we add a digit. It also serves as an "are
11621 we in octal/hex/binary?" indicator to disallow hex characters
11622 when in octal mode.
11627 bool overflowed = FALSE;
11628 bool just_zero = TRUE; /* just plain 0 or binary number? */
11629 static const NV nvshift[5] = { 1.0, 2.0, 4.0, 8.0, 16.0 };
11630 static const char* const bases[5] =
11631 { "", "binary", "", "octal", "hexadecimal" };
11632 static const char* const Bases[5] =
11633 { "", "Binary", "", "Octal", "Hexadecimal" };
11634 static const char* const maxima[5] =
11636 "0b11111111111111111111111111111111",
11640 const char *base, *Base, *max;
11642 /* check for hex */
11647 } else if (s[1] == 'b') {
11652 /* check for a decimal in disguise */
11653 else if (s[1] == '.' || s[1] == 'e' || s[1] == 'E')
11655 /* so it must be octal */
11662 if (ckWARN(WARN_SYNTAX))
11663 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
11664 "Misplaced _ in number");
11668 base = bases[shift];
11669 Base = Bases[shift];
11670 max = maxima[shift];
11672 /* read the rest of the number */
11674 /* x is used in the overflow test,
11675 b is the digit we're adding on. */
11680 /* if we don't mention it, we're done */
11684 /* _ are ignored -- but warned about if consecutive */
11686 if (lastub && s == lastub + 1 && ckWARN(WARN_SYNTAX))
11687 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
11688 "Misplaced _ in number");
11692 /* 8 and 9 are not octal */
11693 case '8': case '9':
11695 yyerror(Perl_form(aTHX_ "Illegal octal digit '%c'", *s));
11699 case '2': case '3': case '4':
11700 case '5': case '6': case '7':
11702 yyerror(Perl_form(aTHX_ "Illegal binary digit '%c'", *s));
11705 case '0': case '1':
11706 b = *s++ & 15; /* ASCII digit -> value of digit */
11710 case 'a': case 'b': case 'c': case 'd': case 'e': case 'f':
11711 case 'A': case 'B': case 'C': case 'D': case 'E': case 'F':
11712 /* make sure they said 0x */
11715 b = (*s++ & 7) + 9;
11717 /* Prepare to put the digit we have onto the end
11718 of the number so far. We check for overflows.
11724 x = u << shift; /* make room for the digit */
11726 if ((x >> shift) != u
11727 && !(PL_hints & HINT_NEW_BINARY)) {
11730 if (ckWARN_d(WARN_OVERFLOW))
11731 Perl_warner(aTHX_ packWARN(WARN_OVERFLOW),
11732 "Integer overflow in %s number",
11735 u = x | b; /* add the digit to the end */
11738 n *= nvshift[shift];
11739 /* If an NV has not enough bits in its
11740 * mantissa to represent an UV this summing of
11741 * small low-order numbers is a waste of time
11742 * (because the NV cannot preserve the
11743 * low-order bits anyway): we could just
11744 * remember when did we overflow and in the
11745 * end just multiply n by the right
11753 /* if we get here, we had success: make a scalar value from
11758 /* final misplaced underbar check */
11759 if (s[-1] == '_') {
11760 if (ckWARN(WARN_SYNTAX))
11761 Perl_warner(aTHX_ packWARN(WARN_SYNTAX), "Misplaced _ in number");
11766 if (n > 4294967295.0 && ckWARN(WARN_PORTABLE))
11767 Perl_warner(aTHX_ packWARN(WARN_PORTABLE),
11768 "%s number > %s non-portable",
11774 if (u > 0xffffffff && ckWARN(WARN_PORTABLE))
11775 Perl_warner(aTHX_ packWARN(WARN_PORTABLE),
11776 "%s number > %s non-portable",
11781 if (just_zero && (PL_hints & HINT_NEW_INTEGER))
11782 sv = new_constant(start, s - start, "integer",
11784 else if (PL_hints & HINT_NEW_BINARY)
11785 sv = new_constant(start, s - start, "binary", sv, NULL, NULL);
11790 handle decimal numbers.
11791 we're also sent here when we read a 0 as the first digit
11793 case '1': case '2': case '3': case '4': case '5':
11794 case '6': case '7': case '8': case '9': case '.':
11797 e = PL_tokenbuf + sizeof PL_tokenbuf - 6; /* room for various punctuation */
11800 /* read next group of digits and _ and copy into d */
11801 while (isDIGIT(*s) || *s == '_') {
11802 /* skip underscores, checking for misplaced ones
11806 if (lastub && s == lastub + 1 && ckWARN(WARN_SYNTAX))
11807 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
11808 "Misplaced _ in number");
11812 /* check for end of fixed-length buffer */
11814 Perl_croak(aTHX_ number_too_long);
11815 /* if we're ok, copy the character */
11820 /* final misplaced underbar check */
11821 if (lastub && s == lastub + 1) {
11822 if (ckWARN(WARN_SYNTAX))
11823 Perl_warner(aTHX_ packWARN(WARN_SYNTAX), "Misplaced _ in number");
11826 /* read a decimal portion if there is one. avoid
11827 3..5 being interpreted as the number 3. followed
11830 if (*s == '.' && s[1] != '.') {
11835 if (ckWARN(WARN_SYNTAX))
11836 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
11837 "Misplaced _ in number");
11841 /* copy, ignoring underbars, until we run out of digits.
11843 for (; isDIGIT(*s) || *s == '_'; s++) {
11844 /* fixed length buffer check */
11846 Perl_croak(aTHX_ number_too_long);
11848 if (lastub && s == lastub + 1 && ckWARN(WARN_SYNTAX))
11849 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
11850 "Misplaced _ in number");
11856 /* fractional part ending in underbar? */
11857 if (s[-1] == '_') {
11858 if (ckWARN(WARN_SYNTAX))
11859 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
11860 "Misplaced _ in number");
11862 if (*s == '.' && isDIGIT(s[1])) {
11863 /* oops, it's really a v-string, but without the "v" */
11869 /* read exponent part, if present */
11870 if ((*s == 'e' || *s == 'E') && strchr("+-0123456789_", s[1])) {
11874 /* regardless of whether user said 3E5 or 3e5, use lower 'e' */
11875 *d++ = 'e'; /* At least some Mach atof()s don't grok 'E' */
11877 /* stray preinitial _ */
11879 if (ckWARN(WARN_SYNTAX))
11880 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
11881 "Misplaced _ in number");
11885 /* allow positive or negative exponent */
11886 if (*s == '+' || *s == '-')
11889 /* stray initial _ */
11891 if (ckWARN(WARN_SYNTAX))
11892 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
11893 "Misplaced _ in number");
11897 /* read digits of exponent */
11898 while (isDIGIT(*s) || *s == '_') {
11901 Perl_croak(aTHX_ number_too_long);
11905 if (((lastub && s == lastub + 1) ||
11906 (!isDIGIT(s[1]) && s[1] != '_'))
11907 && ckWARN(WARN_SYNTAX))
11908 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
11909 "Misplaced _ in number");
11916 /* make an sv from the string */
11920 We try to do an integer conversion first if no characters
11921 indicating "float" have been found.
11926 const int flags = grok_number (PL_tokenbuf, d - PL_tokenbuf, &uv);
11928 if (flags == IS_NUMBER_IN_UV) {
11930 sv_setiv(sv, uv); /* Prefer IVs over UVs. */
11933 } else if (flags == (IS_NUMBER_IN_UV | IS_NUMBER_NEG)) {
11934 if (uv <= (UV) IV_MIN)
11935 sv_setiv(sv, -(IV)uv);
11942 /* terminate the string */
11944 nv = Atof(PL_tokenbuf);
11948 if ( floatit ? (PL_hints & HINT_NEW_FLOAT) :
11949 (PL_hints & HINT_NEW_INTEGER) )
11950 sv = new_constant(PL_tokenbuf, d - PL_tokenbuf,
11951 (floatit ? "float" : "integer"),
11955 /* if it starts with a v, it could be a v-string */
11958 sv = newSV(5); /* preallocate storage space */
11959 s = scan_vstring(s,sv);
11963 /* make the op for the constant and return */
11966 lvalp->opval = newSVOP(OP_CONST, 0, sv);
11968 lvalp->opval = NULL;
11974 S_scan_formline(pTHX_ register char *s)
11977 register char *eol;
11979 SV * const stuff = newSVpvs("");
11980 bool needargs = FALSE;
11981 bool eofmt = FALSE;
11983 char *tokenstart = s;
11986 if (PL_madskills) {
11987 savewhite = PL_thiswhite;
11992 while (!needargs) {
11994 #ifdef PERL_STRICT_CR
11995 for (t = s+1;SPACE_OR_TAB(*t); t++) ;
11997 for (t = s+1;SPACE_OR_TAB(*t) || *t == '\r'; t++) ;
11999 if (*t == '\n' || t == PL_bufend) {
12004 if (PL_in_eval && !PL_rsfp) {
12005 eol = (char *) memchr(s,'\n',PL_bufend-s);
12010 eol = PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
12012 for (t = s; t < eol; t++) {
12013 if (*t == '~' && t[1] == '~' && SvCUR(stuff)) {
12015 goto enough; /* ~~ must be first line in formline */
12017 if (*t == '@' || *t == '^')
12021 sv_catpvn(stuff, s, eol-s);
12022 #ifndef PERL_STRICT_CR
12023 if (eol-s > 1 && eol[-2] == '\r' && eol[-1] == '\n') {
12024 char *end = SvPVX(stuff) + SvCUR(stuff);
12027 SvCUR_set(stuff, SvCUR(stuff) - 1);
12037 if (PL_madskills) {
12039 sv_catpvn(PL_thistoken, tokenstart, PL_bufend - tokenstart);
12041 PL_thistoken = newSVpvn(tokenstart, PL_bufend - tokenstart);
12044 s = filter_gets(PL_linestr, PL_rsfp, 0);
12046 tokenstart = PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = SvPVX(PL_linestr);
12048 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = SvPVX(PL_linestr);
12050 PL_bufend = PL_bufptr + SvCUR(PL_linestr);
12051 PL_last_lop = PL_last_uni = NULL;
12060 if (SvCUR(stuff)) {
12063 PL_lex_state = LEX_NORMAL;
12064 start_force(PL_curforce);
12065 NEXTVAL_NEXTTOKE.ival = 0;
12069 PL_lex_state = LEX_FORMLINE;
12071 if (UTF && is_utf8_string((U8*)SvPVX_const(stuff), SvCUR(stuff)))
12073 else if (PL_encoding)
12074 sv_recode_to_utf8(stuff, PL_encoding);
12076 start_force(PL_curforce);
12077 NEXTVAL_NEXTTOKE.opval = (OP*)newSVOP(OP_CONST, 0, stuff);
12079 start_force(PL_curforce);
12080 NEXTVAL_NEXTTOKE.ival = OP_FORMLINE;
12084 SvREFCNT_dec(stuff);
12086 PL_lex_formbrack = 0;
12090 if (PL_madskills) {
12092 sv_catpvn(PL_thistoken, tokenstart, s - tokenstart);
12094 PL_thistoken = newSVpvn(tokenstart, s - tokenstart);
12095 PL_thiswhite = savewhite;
12107 PL_cshlen = strlen(PL_cshname);
12109 #if defined(USE_ITHREADS)
12110 PERL_UNUSED_CONTEXT;
12116 Perl_start_subparse(pTHX_ I32 is_format, U32 flags)
12119 const I32 oldsavestack_ix = PL_savestack_ix;
12120 CV* const outsidecv = PL_compcv;
12123 assert(SvTYPE(PL_compcv) == SVt_PVCV);
12125 SAVEI32(PL_subline);
12126 save_item(PL_subname);
12127 SAVESPTR(PL_compcv);
12129 PL_compcv = (CV*)newSV(0);
12130 sv_upgrade((SV *)PL_compcv, is_format ? SVt_PVFM : SVt_PVCV);
12131 CvFLAGS(PL_compcv) |= flags;
12133 PL_subline = CopLINE(PL_curcop);
12134 CvPADLIST(PL_compcv) = pad_new(padnew_SAVE|padnew_SAVESUB);
12135 CvOUTSIDE(PL_compcv) = (CV*)SvREFCNT_inc_simple(outsidecv);
12136 CvOUTSIDE_SEQ(PL_compcv) = PL_cop_seqmax;
12138 return oldsavestack_ix;
12142 #pragma segment Perl_yylex
12145 Perl_yywarn(pTHX_ const char *s)
12148 PL_in_eval |= EVAL_WARNONLY;
12150 PL_in_eval &= ~EVAL_WARNONLY;
12155 Perl_yyerror(pTHX_ const char *s)
12158 const char *where = NULL;
12159 const char *context = NULL;
12163 if (!yychar || (yychar == ';' && !PL_rsfp))
12165 else if (PL_oldoldbufptr && PL_bufptr > PL_oldoldbufptr &&
12166 PL_bufptr - PL_oldoldbufptr < 200 && PL_oldoldbufptr != PL_oldbufptr &&
12167 PL_oldbufptr != PL_bufptr) {
12170 The code below is removed for NetWare because it abends/crashes on NetWare
12171 when the script has error such as not having the closing quotes like:
12172 if ($var eq "value)
12173 Checking of white spaces is anyway done in NetWare code.
12176 while (isSPACE(*PL_oldoldbufptr))
12179 context = PL_oldoldbufptr;
12180 contlen = PL_bufptr - PL_oldoldbufptr;
12182 else if (PL_oldbufptr && PL_bufptr > PL_oldbufptr &&
12183 PL_bufptr - PL_oldbufptr < 200 && PL_oldbufptr != PL_bufptr) {
12186 The code below is removed for NetWare because it abends/crashes on NetWare
12187 when the script has error such as not having the closing quotes like:
12188 if ($var eq "value)
12189 Checking of white spaces is anyway done in NetWare code.
12192 while (isSPACE(*PL_oldbufptr))
12195 context = PL_oldbufptr;
12196 contlen = PL_bufptr - PL_oldbufptr;
12198 else if (yychar > 255)
12199 where = "next token ???";
12200 else if (yychar == -2) { /* YYEMPTY */
12201 if (PL_lex_state == LEX_NORMAL ||
12202 (PL_lex_state == LEX_KNOWNEXT && PL_lex_defer == LEX_NORMAL))
12203 where = "at end of line";
12204 else if (PL_lex_inpat)
12205 where = "within pattern";
12207 where = "within string";
12210 SV * const where_sv = sv_2mortal(newSVpvs("next char "));
12212 Perl_sv_catpvf(aTHX_ where_sv, "^%c", toCTRL(yychar));
12213 else if (isPRINT_LC(yychar))
12214 Perl_sv_catpvf(aTHX_ where_sv, "%c", yychar);
12216 Perl_sv_catpvf(aTHX_ where_sv, "\\%03o", yychar & 255);
12217 where = SvPVX_const(where_sv);
12219 msg = sv_2mortal(newSVpv(s, 0));
12220 Perl_sv_catpvf(aTHX_ msg, " at %s line %"IVdf", ",
12221 OutCopFILE(PL_curcop), (IV)CopLINE(PL_curcop));
12223 Perl_sv_catpvf(aTHX_ msg, "near \"%.*s\"\n", contlen, context);
12225 Perl_sv_catpvf(aTHX_ msg, "%s\n", where);
12226 if (PL_multi_start < PL_multi_end && (U32)(CopLINE(PL_curcop) - PL_multi_end) <= 1) {
12227 Perl_sv_catpvf(aTHX_ msg,
12228 " (Might be a runaway multi-line %c%c string starting on line %"IVdf")\n",
12229 (int)PL_multi_open,(int)PL_multi_close,(IV)PL_multi_start);
12232 if (PL_in_eval & EVAL_WARNONLY && ckWARN_d(WARN_SYNTAX))
12233 Perl_warner(aTHX_ packWARN(WARN_SYNTAX), "%"SVf, msg);
12236 if (PL_error_count >= 10) {
12237 if (PL_in_eval && SvCUR(ERRSV))
12238 Perl_croak(aTHX_ "%"SVf"%s has too many errors.\n",
12239 ERRSV, OutCopFILE(PL_curcop));
12241 Perl_croak(aTHX_ "%s has too many errors.\n",
12242 OutCopFILE(PL_curcop));
12245 PL_in_my_stash = NULL;
12249 #pragma segment Main
12253 S_swallow_bom(pTHX_ U8 *s)
12256 const STRLEN slen = SvCUR(PL_linestr);
12259 if (s[1] == 0xFE) {
12260 /* UTF-16 little-endian? (or UTF32-LE?) */
12261 if (s[2] == 0 && s[3] == 0) /* UTF-32 little-endian */
12262 Perl_croak(aTHX_ "Unsupported script encoding UTF32-LE");
12263 #ifndef PERL_NO_UTF16_FILTER
12264 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF16-LE script encoding (BOM)\n");
12267 if (PL_bufend > (char*)s) {
12271 filter_add(utf16rev_textfilter, NULL);
12272 Newx(news, (PL_bufend - (char*)s) * 3 / 2 + 1, U8);
12273 utf16_to_utf8_reversed(s, news,
12274 PL_bufend - (char*)s - 1,
12276 sv_setpvn(PL_linestr, (const char*)news, newlen);
12278 s = (U8*)SvPVX(PL_linestr);
12279 Copy(news, s, newlen, U8);
12283 SvUTF8_on(PL_linestr);
12284 s = (U8*)SvPVX(PL_linestr);
12286 /* FIXME - is this a general bug fix? */
12289 PL_bufend = SvPVX(PL_linestr) + newlen;
12292 Perl_croak(aTHX_ "Unsupported script encoding UTF16-LE");
12297 if (s[1] == 0xFF) { /* UTF-16 big-endian? */
12298 #ifndef PERL_NO_UTF16_FILTER
12299 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16BE script encoding (BOM)\n");
12302 if (PL_bufend > (char *)s) {
12306 filter_add(utf16_textfilter, NULL);
12307 Newx(news, (PL_bufend - (char*)s) * 3 / 2 + 1, U8);
12308 utf16_to_utf8(s, news,
12309 PL_bufend - (char*)s,
12311 sv_setpvn(PL_linestr, (const char*)news, newlen);
12313 SvUTF8_on(PL_linestr);
12314 s = (U8*)SvPVX(PL_linestr);
12315 PL_bufend = SvPVX(PL_linestr) + newlen;
12318 Perl_croak(aTHX_ "Unsupported script encoding UTF16-BE");
12323 if (slen > 2 && s[1] == 0xBB && s[2] == 0xBF) {
12324 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-8 script encoding (BOM)\n");
12325 s += 3; /* UTF-8 */
12331 if (s[2] == 0xFE && s[3] == 0xFF) {
12332 /* UTF-32 big-endian */
12333 Perl_croak(aTHX_ "Unsupported script encoding UTF32-BE");
12336 else if (s[2] == 0 && s[3] != 0) {
12339 * are a good indicator of UTF-16BE. */
12340 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16BE script encoding (no BOM)\n");
12345 if (slen > 3 && s[1] == 0 && s[2] != 0 && s[3] == 0) {
12348 * are a good indicator of UTF-16LE. */
12349 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16LE script encoding (no BOM)\n");
12358 * Restore a source filter.
12362 restore_rsfp(pTHX_ void *f)
12365 PerlIO * const fp = (PerlIO*)f;
12367 if (PL_rsfp == PerlIO_stdin())
12368 PerlIO_clearerr(PL_rsfp);
12369 else if (PL_rsfp && (PL_rsfp != fp))
12370 PerlIO_close(PL_rsfp);
12374 #ifndef PERL_NO_UTF16_FILTER
12376 utf16_textfilter(pTHX_ int idx, SV *sv, int maxlen)
12379 const STRLEN old = SvCUR(sv);
12380 const I32 count = FILTER_READ(idx+1, sv, maxlen);
12381 DEBUG_P(PerlIO_printf(Perl_debug_log,
12382 "utf16_textfilter(%p): %d %d (%d)\n",
12383 utf16_textfilter, idx, maxlen, (int) count));
12387 Newx(tmps, SvCUR(sv) * 3 / 2 + 1, U8);
12388 Copy(SvPVX_const(sv), tmps, old, char);
12389 utf16_to_utf8((U8*)SvPVX_const(sv) + old, tmps + old,
12390 SvCUR(sv) - old, &newlen);
12391 sv_usepvn(sv, (char*)tmps, (STRLEN)newlen + old);
12393 DEBUG_P({sv_dump(sv);});
12398 utf16rev_textfilter(pTHX_ int idx, SV *sv, int maxlen)
12401 const STRLEN old = SvCUR(sv);
12402 const I32 count = FILTER_READ(idx+1, sv, maxlen);
12403 DEBUG_P(PerlIO_printf(Perl_debug_log,
12404 "utf16rev_textfilter(%p): %d %d (%d)\n",
12405 utf16rev_textfilter, idx, maxlen, (int) count));
12409 Newx(tmps, SvCUR(sv) * 3 / 2 + 1, U8);
12410 Copy(SvPVX_const(sv), tmps, old, char);
12411 utf16_to_utf8((U8*)SvPVX_const(sv) + old, tmps + old,
12412 SvCUR(sv) - old, &newlen);
12413 sv_usepvn(sv, (char*)tmps, (STRLEN)newlen + old);
12415 DEBUG_P({ sv_dump(sv); });
12421 Returns a pointer to the next character after the parsed
12422 vstring, as well as updating the passed in sv.
12424 Function must be called like
12427 s = scan_vstring(s,sv);
12429 The sv should already be large enough to store the vstring
12430 passed in, for performance reasons.
12435 Perl_scan_vstring(pTHX_ const char *s, SV *sv)
12438 const char *pos = s;
12439 const char *start = s;
12440 if (*pos == 'v') pos++; /* get past 'v' */
12441 while (pos < PL_bufend && (isDIGIT(*pos) || *pos == '_'))
12443 if ( *pos != '.') {
12444 /* this may not be a v-string if followed by => */
12445 const char *next = pos;
12446 while (next < PL_bufend && isSPACE(*next))
12448 if ((PL_bufend - next) >= 2 && *next == '=' && next[1] == '>' ) {
12449 /* return string not v-string */
12450 sv_setpvn(sv,(char *)s,pos-s);
12451 return (char *)pos;
12455 if (!isALPHA(*pos)) {
12456 U8 tmpbuf[UTF8_MAXBYTES+1];
12459 s++; /* get past 'v' */
12461 sv_setpvn(sv, "", 0);
12464 /* this is atoi() that tolerates underscores */
12467 const char *end = pos;
12469 while (--end >= s) {
12471 const UV orev = rev;
12472 rev += (*end - '0') * mult;
12474 if (orev > rev && ckWARN_d(WARN_OVERFLOW))
12475 Perl_warner(aTHX_ packWARN(WARN_OVERFLOW),
12476 "Integer overflow in decimal number");
12480 if (rev > 0x7FFFFFFF)
12481 Perl_croak(aTHX_ "In EBCDIC the v-string components cannot exceed 2147483647");
12483 /* Append native character for the rev point */
12484 tmpend = uvchr_to_utf8(tmpbuf, rev);
12485 sv_catpvn(sv, (const char*)tmpbuf, tmpend - tmpbuf);
12486 if (!UNI_IS_INVARIANT(NATIVE_TO_UNI(rev)))
12488 if (pos + 1 < PL_bufend && *pos == '.' && isDIGIT(pos[1]))
12494 while (pos < PL_bufend && (isDIGIT(*pos) || *pos == '_'))
12498 sv_magic(sv,NULL,PERL_MAGIC_vstring,(const char*)start, pos-start);
12506 * c-indentation-style: bsd
12507 * c-basic-offset: 4
12508 * indent-tabs-mode: t
12511 * ex: set ts=8 sts=4 sw=4 noet: