This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
toke.c: White-space only
[perl5.git] / toke.c
CommitLineData
a0d0e21e 1/* toke.c
a687059c 2 *
1129b882
NC
3 * Copyright (C) 1991, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999, 2000,
4 * 2001, 2002, 2003, 2004, 2005, 2006, 2007, 2008 by Larry Wall and others
a687059c 5 *
d48672a2
LW
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.
378cc40b 8 *
a0d0e21e
LW
9 */
10
11/*
4ac71550
TC
12 * 'It all comes from here, the stench and the peril.' --Frodo
13 *
14 * [p.719 of _The Lord of the Rings_, IV/ix: "Shelob's Lair"]
378cc40b
LW
15 */
16
9cbb5ea2
GS
17/*
18 * This file is the lexer for Perl. It's closely linked to the
4e553d73 19 * parser, perly.y.
ffb4593c
NT
20 *
21 * The main routine is yylex(), which returns the next token.
22 */
23
f0e67a1d
Z
24/*
25=head1 Lexer interface
f0e67a1d
Z
26This is the lower layer of the Perl parser, managing characters and tokens.
27
28=for apidoc AmU|yy_parser *|PL_parser
29
30Pointer to a structure encapsulating the state of the parsing operation
31currently in progress. The pointer can be locally changed to perform
32a nested parse without interfering with the state of an outer parse.
33Individual members of C<PL_parser> have their own documentation.
34
35=cut
36*/
37
378cc40b 38#include "EXTERN.h"
864dbfa3 39#define PERL_IN_TOKE_C
378cc40b 40#include "perl.h"
f7e03a10 41#include "dquote_inline.h"
378cc40b 42
eb0d8d16
NC
43#define new_constant(a,b,c,d,e,f,g) \
44 S_new_constant(aTHX_ a,b,STR_WITH_LEN(c),d,e,f, g)
45
6154021b 46#define pl_yylval (PL_parser->yylval)
d3b6f988 47
199e78b7
DM
48/* XXX temporary backwards compatibility */
49#define PL_lex_brackets (PL_parser->lex_brackets)
78cdf107
Z
50#define PL_lex_allbrackets (PL_parser->lex_allbrackets)
51#define PL_lex_fakeeof (PL_parser->lex_fakeeof)
199e78b7
DM
52#define PL_lex_brackstack (PL_parser->lex_brackstack)
53#define PL_lex_casemods (PL_parser->lex_casemods)
54#define PL_lex_casestack (PL_parser->lex_casestack)
199e78b7 55#define PL_lex_dojoin (PL_parser->lex_dojoin)
199e78b7
DM
56#define PL_lex_formbrack (PL_parser->lex_formbrack)
57#define PL_lex_inpat (PL_parser->lex_inpat)
58#define PL_lex_inwhat (PL_parser->lex_inwhat)
59#define PL_lex_op (PL_parser->lex_op)
60#define PL_lex_repl (PL_parser->lex_repl)
61#define PL_lex_starts (PL_parser->lex_starts)
62#define PL_lex_stuff (PL_parser->lex_stuff)
63#define PL_multi_start (PL_parser->multi_start)
64#define PL_multi_open (PL_parser->multi_open)
65#define PL_multi_close (PL_parser->multi_close)
199e78b7 66#define PL_preambled (PL_parser->preambled)
bdc0bf6f 67#define PL_linestr (PL_parser->linestr)
c2598295
DM
68#define PL_expect (PL_parser->expect)
69#define PL_copline (PL_parser->copline)
f06b5848
DM
70#define PL_bufptr (PL_parser->bufptr)
71#define PL_oldbufptr (PL_parser->oldbufptr)
72#define PL_oldoldbufptr (PL_parser->oldoldbufptr)
73#define PL_linestart (PL_parser->linestart)
74#define PL_bufend (PL_parser->bufend)
75#define PL_last_uni (PL_parser->last_uni)
76#define PL_last_lop (PL_parser->last_lop)
77#define PL_last_lop_op (PL_parser->last_lop_op)
bc177e6b 78#define PL_lex_state (PL_parser->lex_state)
2f9285f8 79#define PL_rsfp (PL_parser->rsfp)
5486870f 80#define PL_rsfp_filters (PL_parser->rsfp_filters)
12bd6ede
DM
81#define PL_in_my (PL_parser->in_my)
82#define PL_in_my_stash (PL_parser->in_my_stash)
14047fc9 83#define PL_tokenbuf (PL_parser->tokenbuf)
670a9cb2 84#define PL_multi_end (PL_parser->multi_end)
13765c85 85#define PL_error_count (PL_parser->error_count)
199e78b7 86
fb205e7a
DM
87# define PL_nexttoke (PL_parser->nexttoke)
88# define PL_nexttype (PL_parser->nexttype)
89# define PL_nextval (PL_parser->nextval)
199e78b7 90
6432a58a
DM
91
92#define SvEVALED(sv) \
93 (SvTYPE(sv) >= SVt_PVNV \
94 && ((XPVIV*)SvANY(sv))->xiv_u.xivu_eval_seen)
95
a1894d81 96static const char* const ident_too_long = "Identifier too long";
8903cb82 97
9ded7720 98# define NEXTVAL_NEXTTOKE PL_nextval[PL_nexttoke]
29595ff2 99
a7aaec61
Z
100#define XENUMMASK 0x3f
101#define XFAKEEOF 0x40
102#define XFAKEBRACK 0x80
9059aa12 103
39e02b42 104#ifdef USE_UTF8_SCRIPTS
b3041197 105# define UTF cBOOL(!IN_BYTES)
2b9d42f0 106#else
b3041197 107# define UTF cBOOL((PL_linestr && DO_UTF8(PL_linestr)) || ( !(PL_parser->lex_flags & LEX_IGNORE_UTF8_HINTS) && (PL_hints & HINT_UTF8)))
2b9d42f0 108#endif
a0ed51b3 109
b1fc3636
CJ
110/* The maximum number of characters preceding the unrecognized one to display */
111#define UNRECOGNIZED_PRECEDE_COUNT 10
112
61f0cdd9 113/* In variables named $^X, these are the legal values for X.
2b92dfce
GS
114 * 1999-02-27 mjd-perl-patch@plover.com */
115#define isCONTROLVAR(x) (isUPPER(x) || strchr("[\\]^_?", (x)))
116
14bd96d0 117#define SPACE_OR_TAB(c) isBLANK_A(c)
bf4acbe4 118
9ff909cf
JH
119#define HEXFP_PEEK(s) \
120 (((s[0] == '.') && \
121 (isXDIGIT(s[1]) || isALPHA_FOLD_EQ(s[1], 'p'))) || \
122 isALPHA_FOLD_EQ(s[0], 'p'))
123
ffb4593c
NT
124/* LEX_* are values for PL_lex_state, the state of the lexer.
125 * They are arranged oddly so that the guard on the switch statement
79072805 126 * can get by with a single comparison (if the compiler is smart enough).
9da1dd8f
DM
127 *
128 * These values refer to the various states within a sublex parse,
129 * i.e. within a double quotish string
79072805
LW
130 */
131
fb73857a
PP
132/* #define LEX_NOTPARSING 11 is done in perl.h. */
133
b6007c36
DM
134#define LEX_NORMAL 10 /* normal code (ie not within "...") */
135#define LEX_INTERPNORMAL 9 /* code within a string, eg "$foo[$x+1]" */
136#define LEX_INTERPCASEMOD 8 /* expecting a \U, \Q or \E etc */
137#define LEX_INTERPPUSH 7 /* starting a new sublex parse level */
138#define LEX_INTERPSTART 6 /* expecting the start of a $var */
139
140 /* at end of code, eg "$x" followed by: */
141#define LEX_INTERPEND 5 /* ... eg not one of [, { or -> */
142#define LEX_INTERPENDMAYBE 4 /* ... eg one of [, { or -> */
143
144#define LEX_INTERPCONCAT 3 /* expecting anything, eg at start of
145 string or after \E, $foo, etc */
146#define LEX_INTERPCONST 2 /* NOT USED */
147#define LEX_FORMLINE 1 /* expecting a format line */
b6007c36 148
79072805 149
bbf60fe6 150#ifdef DEBUGGING
27da23d5 151static const char* const lex_state_names[] = {
bbf60fe6
DM
152 "KNOWNEXT",
153 "FORMLINE",
154 "INTERPCONST",
155 "INTERPCONCAT",
156 "INTERPENDMAYBE",
157 "INTERPEND",
158 "INTERPSTART",
159 "INTERPPUSH",
160 "INTERPCASEMOD",
161 "INTERPNORMAL",
162 "NORMAL"
163};
164#endif
165
79072805 166#include "keywords.h"
fe14fcc3 167
ffb4593c
NT
168/* CLINE is a macro that ensures PL_copline has a sane value */
169
57843af0 170#define CLINE (PL_copline = (CopLINE(PL_curcop) < PL_copline ? CopLINE(PL_curcop) : PL_copline))
3280af22 171
ffb4593c
NT
172/*
173 * Convenience functions to return different tokens and prime the
9cbb5ea2 174 * lexer for the next token. They all take an argument.
ffb4593c
NT
175 *
176 * TOKEN : generic token (used for '(', DOLSHARP, etc)
177 * OPERATOR : generic operator
178 * AOPERATOR : assignment operator
179 * PREBLOCK : beginning the block after an if, while, foreach, ...
180 * PRETERMBLOCK : beginning a non-code-defining {} block (eg, hash ref)
181 * PREREF : *EXPR where EXPR is not a simple identifier
182 * TERM : expression term
89f35911 183 * POSTDEREF : postfix dereference (->$* ->@[...] etc.)
ffb4593c
NT
184 * LOOPX : loop exiting command (goto, last, dump, etc)
185 * FTST : file test operator
186 * FUN0 : zero-argument function
7eb971ee 187 * FUN0OP : zero-argument function, with its op created in this file
2d2e263d 188 * FUN1 : not used, except for not, which isn't a UNIOP
ffb4593c
NT
189 * BOop : bitwise or or xor
190 * BAop : bitwise and
8823cb89 191 * BCop : bitwise complement
ffb4593c
NT
192 * SHop : shift operator
193 * PWop : power operator
9cbb5ea2 194 * PMop : pattern-matching operator
ffb4593c 195 * Aop : addition-level operator
e4916dd1 196 * AopNOASSIGN : addition-level operator that is never part of .=
ffb4593c
NT
197 * Mop : multiplication-level operator
198 * Eop : equality-testing operator
e5edeb50 199 * Rop : relational operator <= != gt
ffb4593c
NT
200 *
201 * Also see LOP and lop() below.
202 */
203
998054bd 204#ifdef DEBUGGING /* Serve -DT. */
704d4215 205# define REPORT(retval) tokereport((I32)retval, &pl_yylval)
998054bd 206#else
bbf60fe6 207# define REPORT(retval) (retval)
998054bd
SC
208#endif
209
bbf60fe6
DM
210#define TOKEN(retval) return ( PL_bufptr = s, REPORT(retval))
211#define OPERATOR(retval) return (PL_expect = XTERM, PL_bufptr = s, REPORT(retval))
b1764551 212#define AOPERATOR(retval) return ao((PL_expect = XTERM, PL_bufptr = s, retval))
bbf60fe6
DM
213#define PREBLOCK(retval) return (PL_expect = XBLOCK,PL_bufptr = s, REPORT(retval))
214#define PRETERMBLOCK(retval) return (PL_expect = XTERMBLOCK,PL_bufptr = s, REPORT(retval))
215#define PREREF(retval) return (PL_expect = XREF,PL_bufptr = s, REPORT(retval))
216#define TERM(retval) return (CLINE, PL_expect = XOPERATOR, PL_bufptr = s, REPORT(retval))
89f35911 217#define POSTDEREF(f) return (PL_bufptr = s, S_postderef(aTHX_ REPORT(f),s[1]))
185c2e96 218#define LOOPX(f) return (PL_bufptr = force_word(s,BAREWORD,TRUE,FALSE), \
7a61bf3c 219 pl_yylval.ival=f, \
a49203fd 220 PL_expect = PL_nexttoke ? XOPERATOR : XTERM, \
7a61bf3c 221 REPORT((int)LOOPEX))
6154021b
RGS
222#define FTST(f) return (pl_yylval.ival=f, PL_expect=XTERMORDORDOR, PL_bufptr=s, REPORT((int)UNIOP))
223#define FUN0(f) return (pl_yylval.ival=f, PL_expect=XOPERATOR, PL_bufptr=s, REPORT((int)FUNC0))
7eb971ee 224#define FUN0OP(f) return (pl_yylval.opval=f, CLINE, PL_expect=XOPERATOR, PL_bufptr=s, REPORT((int)FUNC0OP))
6154021b 225#define FUN1(f) return (pl_yylval.ival=f, PL_expect=XOPERATOR, PL_bufptr=s, REPORT((int)FUNC1))
b1764551
FC
226#define BOop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, (int)BITOROP))
227#define BAop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, (int)BITANDOP))
8823cb89
FC
228#define BCop(f) return pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr = s, \
229 REPORT('~')
b1764551
FC
230#define SHop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, (int)SHIFTOP))
231#define PWop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, (int)POWOP))
6154021b 232#define PMop(f) return(pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)MATCHOP))
b1764551 233#define Aop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, (int)ADDOP))
e4916dd1 234#define AopNOASSIGN(f) return (pl_yylval.ival=f, PL_bufptr=s, REPORT((int)ADDOP))
b1764551 235#define Mop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, (int)MULOP))
6154021b
RGS
236#define Eop(f) return (pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)EQOP))
237#define Rop(f) return (pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)RELOP))
2f3197b3 238
a687059c
LW
239/* This bit of chicanery makes a unary function followed by
240 * a parenthesis into a function with one argument, highest precedence.
6f33ba73
RGS
241 * The UNIDOR macro is for unary functions that can be followed by the //
242 * operator (such as C<shift // 0>).
a687059c 243 */
d68ce4ac 244#define UNI3(f,x,have_x) { \
6154021b 245 pl_yylval.ival = f; \
d68ce4ac 246 if (have_x) PL_expect = x; \
376fcdbf
AL
247 PL_bufptr = s; \
248 PL_last_uni = PL_oldbufptr; \
0af40c75 249 PL_last_lop_op = (f) < 0 ? -(f) : (f); \
376fcdbf
AL
250 if (*s == '(') \
251 return REPORT( (int)FUNC1 ); \
294a536f 252 s = skipspace(s); \
376fcdbf
AL
253 return REPORT( *s=='(' ? (int)FUNC1 : (int)UNIOP ); \
254 }
d68ce4ac
FC
255#define UNI(f) UNI3(f,XTERM,1)
256#define UNIDOR(f) UNI3(f,XTERMORDORDOR,1)
b5fb7ce3
FC
257#define UNIPROTO(f,optional) { \
258 if (optional) PL_last_uni = PL_oldbufptr; \
22393538
MH
259 OPERATOR(f); \
260 }
a687059c 261
d68ce4ac 262#define UNIBRACK(f) UNI3(f,0,0)
79072805 263
9f68db38 264/* grandfather return to old style */
78cdf107
Z
265#define OLDLOP(f) \
266 do { \
267 if (!PL_lex_allbrackets && PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC) \
268 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC; \
269 pl_yylval.ival = (f); \
270 PL_expect = XTERM; \
271 PL_bufptr = s; \
272 return (int)LSTOP; \
273 } while(0)
79072805 274
83944c01
FC
275#define COPLINE_INC_WITH_HERELINES \
276 STMT_START { \
277 CopLINE_inc(PL_curcop); \
851b527a
FC
278 if (PL_parser->herelines) \
279 CopLINE(PL_curcop) += PL_parser->herelines, \
280 PL_parser->herelines = 0; \
83944c01 281 } STMT_END
ffdb8b16
FC
282/* Called after scan_str to update CopLINE(PL_curcop), but only when there
283 * is no sublex_push to follow. */
284#define COPLINE_SET_FROM_MULTI_END \
285 STMT_START { \
286 CopLINE_set(PL_curcop, PL_multi_end); \
287 if (PL_multi_end != PL_multi_start) \
851b527a 288 PL_parser->herelines = 0; \
ffdb8b16 289 } STMT_END
83944c01
FC
290
291
8fa7f367
JH
292#ifdef DEBUGGING
293
6154021b 294/* how to interpret the pl_yylval associated with the token */
bbf60fe6
DM
295enum token_type {
296 TOKENTYPE_NONE,
297 TOKENTYPE_IVAL,
6154021b 298 TOKENTYPE_OPNUM, /* pl_yylval.ival contains an opcode number */
bbf60fe6 299 TOKENTYPE_PVAL,
aeaef349 300 TOKENTYPE_OPVAL
bbf60fe6
DM
301};
302
6d4a66ac
NC
303static struct debug_tokens {
304 const int token;
305 enum token_type type;
306 const char *name;
307} const debug_tokens[] =
9041c2e3 308{
bbf60fe6
DM
309 { ADDOP, TOKENTYPE_OPNUM, "ADDOP" },
310 { ANDAND, TOKENTYPE_NONE, "ANDAND" },
311 { ANDOP, TOKENTYPE_NONE, "ANDOP" },
312 { ANONSUB, TOKENTYPE_IVAL, "ANONSUB" },
313 { ARROW, TOKENTYPE_NONE, "ARROW" },
314 { ASSIGNOP, TOKENTYPE_OPNUM, "ASSIGNOP" },
315 { BITANDOP, TOKENTYPE_OPNUM, "BITANDOP" },
316 { BITOROP, TOKENTYPE_OPNUM, "BITOROP" },
317 { COLONATTR, TOKENTYPE_NONE, "COLONATTR" },
318 { CONTINUE, TOKENTYPE_NONE, "CONTINUE" },
0d863452 319 { DEFAULT, TOKENTYPE_NONE, "DEFAULT" },
bbf60fe6
DM
320 { DO, TOKENTYPE_NONE, "DO" },
321 { DOLSHARP, TOKENTYPE_NONE, "DOLSHARP" },
322 { DORDOR, TOKENTYPE_NONE, "DORDOR" },
323 { DOROP, TOKENTYPE_OPNUM, "DOROP" },
324 { DOTDOT, TOKENTYPE_IVAL, "DOTDOT" },
325 { ELSE, TOKENTYPE_NONE, "ELSE" },
326 { ELSIF, TOKENTYPE_IVAL, "ELSIF" },
327 { EQOP, TOKENTYPE_OPNUM, "EQOP" },
328 { FOR, TOKENTYPE_IVAL, "FOR" },
329 { FORMAT, TOKENTYPE_NONE, "FORMAT" },
705fe0e5
FC
330 { FORMLBRACK, TOKENTYPE_NONE, "FORMLBRACK" },
331 { FORMRBRACK, TOKENTYPE_NONE, "FORMRBRACK" },
bbf60fe6
DM
332 { FUNC, TOKENTYPE_OPNUM, "FUNC" },
333 { FUNC0, TOKENTYPE_OPNUM, "FUNC0" },
7eb971ee 334 { FUNC0OP, TOKENTYPE_OPVAL, "FUNC0OP" },
bbf60fe6
DM
335 { FUNC0SUB, TOKENTYPE_OPVAL, "FUNC0SUB" },
336 { FUNC1, TOKENTYPE_OPNUM, "FUNC1" },
337 { FUNCMETH, TOKENTYPE_OPVAL, "FUNCMETH" },
0d863452 338 { GIVEN, TOKENTYPE_IVAL, "GIVEN" },
bbf60fe6
DM
339 { HASHBRACK, TOKENTYPE_NONE, "HASHBRACK" },
340 { IF, TOKENTYPE_IVAL, "IF" },
5504e6cf 341 { LABEL, TOKENTYPE_PVAL, "LABEL" },
bbf60fe6
DM
342 { LOCAL, TOKENTYPE_IVAL, "LOCAL" },
343 { LOOPEX, TOKENTYPE_OPNUM, "LOOPEX" },
344 { LSTOP, TOKENTYPE_OPNUM, "LSTOP" },
345 { LSTOPSUB, TOKENTYPE_OPVAL, "LSTOPSUB" },
346 { MATCHOP, TOKENTYPE_OPNUM, "MATCHOP" },
347 { METHOD, TOKENTYPE_OPVAL, "METHOD" },
348 { MULOP, TOKENTYPE_OPNUM, "MULOP" },
349 { MY, TOKENTYPE_IVAL, "MY" },
bbf60fe6
DM
350 { NOAMP, TOKENTYPE_NONE, "NOAMP" },
351 { NOTOP, TOKENTYPE_NONE, "NOTOP" },
352 { OROP, TOKENTYPE_IVAL, "OROP" },
353 { OROR, TOKENTYPE_NONE, "OROR" },
354 { PACKAGE, TOKENTYPE_NONE, "PACKAGE" },
88e1f1a2
JV
355 { PLUGEXPR, TOKENTYPE_OPVAL, "PLUGEXPR" },
356 { PLUGSTMT, TOKENTYPE_OPVAL, "PLUGSTMT" },
bbf60fe6 357 { PMFUNC, TOKENTYPE_OPVAL, "PMFUNC" },
cc624add 358 { POSTJOIN, TOKENTYPE_NONE, "POSTJOIN" },
bbf60fe6
DM
359 { POSTDEC, TOKENTYPE_NONE, "POSTDEC" },
360 { POSTINC, TOKENTYPE_NONE, "POSTINC" },
361 { POWOP, TOKENTYPE_OPNUM, "POWOP" },
362 { PREDEC, TOKENTYPE_NONE, "PREDEC" },
363 { PREINC, TOKENTYPE_NONE, "PREINC" },
364 { PRIVATEREF, TOKENTYPE_OPVAL, "PRIVATEREF" },
f3f204dc 365 { QWLIST, TOKENTYPE_OPVAL, "QWLIST" },
bbf60fe6
DM
366 { REFGEN, TOKENTYPE_NONE, "REFGEN" },
367 { RELOP, TOKENTYPE_OPNUM, "RELOP" },
f3f204dc 368 { REQUIRE, TOKENTYPE_NONE, "REQUIRE" },
bbf60fe6
DM
369 { SHIFTOP, TOKENTYPE_OPNUM, "SHIFTOP" },
370 { SUB, TOKENTYPE_NONE, "SUB" },
371 { THING, TOKENTYPE_OPVAL, "THING" },
372 { UMINUS, TOKENTYPE_NONE, "UMINUS" },
373 { UNIOP, TOKENTYPE_OPNUM, "UNIOP" },
374 { UNIOPSUB, TOKENTYPE_OPVAL, "UNIOPSUB" },
375 { UNLESS, TOKENTYPE_IVAL, "UNLESS" },
376 { UNTIL, TOKENTYPE_IVAL, "UNTIL" },
377 { USE, TOKENTYPE_IVAL, "USE" },
0d863452 378 { WHEN, TOKENTYPE_IVAL, "WHEN" },
bbf60fe6 379 { WHILE, TOKENTYPE_IVAL, "WHILE" },
185c2e96 380 { BAREWORD, TOKENTYPE_OPVAL, "BAREWORD" },
be25f609 381 { YADAYADA, TOKENTYPE_IVAL, "YADAYADA" },
c35e046a 382 { 0, TOKENTYPE_NONE, NULL }
bbf60fe6
DM
383};
384
6154021b 385/* dump the returned token in rv, plus any optional arg in pl_yylval */
998054bd 386
bbf60fe6 387STATIC int
704d4215 388S_tokereport(pTHX_ I32 rv, const YYSTYPE* lvalp)
bbf60fe6 389{
7918f24d
NC
390 PERL_ARGS_ASSERT_TOKEREPORT;
391
bbf60fe6 392 if (DEBUG_T_TEST) {
bd61b366 393 const char *name = NULL;
bbf60fe6 394 enum token_type type = TOKENTYPE_NONE;
f54cb97a 395 const struct debug_tokens *p;
396482e1 396 SV* const report = newSVpvs("<== ");
bbf60fe6 397
f54cb97a 398 for (p = debug_tokens; p->token; p++) {
bbf60fe6
DM
399 if (p->token == (int)rv) {
400 name = p->name;
401 type = p->type;
402 break;
403 }
404 }
405 if (name)
54667de8 406 Perl_sv_catpv(aTHX_ report, name);
239f83d5 407 else if (isGRAPH(rv))
4ebc7986 408 {
bbf60fe6 409 Perl_sv_catpvf(aTHX_ report, "'%c'", (char)rv);
4ebc7986
FC
410 if ((char)rv == 'p')
411 sv_catpvs(report, " (pending identifier)");
412 }
bbf60fe6 413 else if (!rv)
396482e1 414 sv_catpvs(report, "EOF");
bbf60fe6 415 else
147e3846 416 Perl_sv_catpvf(aTHX_ report, "?? %" IVdf, (IV)rv);
bbf60fe6
DM
417 switch (type) {
418 case TOKENTYPE_NONE:
bbf60fe6
DM
419 break;
420 case TOKENTYPE_IVAL:
147e3846 421 Perl_sv_catpvf(aTHX_ report, "(ival=%" IVdf ")", (IV)lvalp->ival);
bbf60fe6
DM
422 break;
423 case TOKENTYPE_OPNUM:
424 Perl_sv_catpvf(aTHX_ report, "(ival=op_%s)",
704d4215 425 PL_op_name[lvalp->ival]);
bbf60fe6
DM
426 break;
427 case TOKENTYPE_PVAL:
704d4215 428 Perl_sv_catpvf(aTHX_ report, "(pval=\"%s\")", lvalp->pval);
bbf60fe6
DM
429 break;
430 case TOKENTYPE_OPVAL:
704d4215 431 if (lvalp->opval) {
401441c0 432 Perl_sv_catpvf(aTHX_ report, "(opval=op_%s)",
704d4215
GG
433 PL_op_name[lvalp->opval->op_type]);
434 if (lvalp->opval->op_type == OP_CONST) {
b6007c36 435 Perl_sv_catpvf(aTHX_ report, " %s",
704d4215 436 SvPEEK(cSVOPx_sv(lvalp->opval)));
b6007c36
DM
437 }
438
439 }
401441c0 440 else
396482e1 441 sv_catpvs(report, "(opval=null)");
bbf60fe6
DM
442 break;
443 }
b6007c36 444 PerlIO_printf(Perl_debug_log, "### %s\n\n", SvPV_nolen_const(report));
bbf60fe6
DM
445 };
446 return (int)rv;
998054bd
SC
447}
448
b6007c36
DM
449
450/* print the buffer with suitable escapes */
451
452STATIC void
15f169a1 453S_printbuf(pTHX_ const char *const fmt, const char *const s)
b6007c36 454{
396482e1 455 SV* const tmp = newSVpvs("");
7918f24d
NC
456
457 PERL_ARGS_ASSERT_PRINTBUF;
458
5d37acd6 459 GCC_DIAG_IGNORE(-Wformat-nonliteral); /* fmt checked by caller */
b6007c36 460 PerlIO_printf(Perl_debug_log, fmt, pv_display(tmp, s, strlen(s), 0, 60));
5d37acd6 461 GCC_DIAG_RESTORE;
b6007c36
DM
462 SvREFCNT_dec(tmp);
463}
464
8fa7f367
JH
465#endif
466
8290c323
NC
467static int
468S_deprecate_commaless_var_list(pTHX) {
469 PL_expect = XTERM;
470 deprecate("comma-less variable list");
471 return REPORT(','); /* grandfather non-comma-format format */
472}
473
ffb4593c
NT
474/*
475 * S_ao
476 *
f393a21a
FC
477 * This subroutine looks for an '=' next to the operator that has just been
478 * parsed and turns it into an ASSIGNOP if it finds one.
ffb4593c
NT
479 */
480
76e3520e 481STATIC int
cea2e8a9 482S_ao(pTHX_ int toketype)
a0d0e21e 483{
3280af22
NIS
484 if (*PL_bufptr == '=') {
485 PL_bufptr++;
a0d0e21e 486 if (toketype == ANDAND)
6154021b 487 pl_yylval.ival = OP_ANDASSIGN;
a0d0e21e 488 else if (toketype == OROR)
6154021b 489 pl_yylval.ival = OP_ORASSIGN;
c963b151 490 else if (toketype == DORDOR)
6154021b 491 pl_yylval.ival = OP_DORASSIGN;
a0d0e21e
LW
492 toketype = ASSIGNOP;
493 }
b1764551 494 return REPORT(toketype);
a0d0e21e
LW
495}
496
ffb4593c
NT
497/*
498 * S_no_op
499 * When Perl expects an operator and finds something else, no_op
500 * prints the warning. It always prints "<something> found where
501 * operator expected. It prints "Missing semicolon on previous line?"
502 * if the surprise occurs at the start of the line. "do you need to
503 * predeclare ..." is printed out for code like "sub bar; foo bar $x"
504 * where the compiler doesn't know if foo is a method call or a function.
505 * It prints "Missing operator before end of line" if there's nothing
506 * after the missing operator, or "... before <...>" if there is something
507 * after the missing operator.
488bc579
FC
508 *
509 * PL_bufptr is expected to point to the start of the thing that was found,
510 * and s after the next token or partial token.
ffb4593c
NT
511 */
512
76e3520e 513STATIC void
15f169a1 514S_no_op(pTHX_ const char *const what, char *s)
463ee0b2 515{
9d4ba2ae
AL
516 char * const oldbp = PL_bufptr;
517 const bool is_first = (PL_oldbufptr == PL_linestart);
68dc0745 518
7918f24d
NC
519 PERL_ARGS_ASSERT_NO_OP;
520
1189a94a
GS
521 if (!s)
522 s = oldbp;
07c798fb 523 else
1189a94a 524 PL_bufptr = s;
734ab321 525 yywarn(Perl_form(aTHX_ "%s found where operator expected", what), UTF ? SVf_UTF8 : 0);
56da5a46
RGS
526 if (ckWARN_d(WARN_SYNTAX)) {
527 if (is_first)
528 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
529 "\t(Missing semicolon on previous line?)\n");
fac0f7a3
KW
530 else if (PL_oldoldbufptr && isIDFIRST_lazy_if_safe(PL_oldoldbufptr,
531 PL_bufend,
532 UTF))
533 {
f54cb97a 534 const char *t;
fac0f7a3
KW
535 for (t = PL_oldoldbufptr;
536 (isWORDCHAR_lazy_if_safe(t, PL_bufend, UTF) || *t == ':');
537 t += UTF ? UTF8SKIP(t) : 1)
538 {
c35e046a 539 NOOP;
fac0f7a3 540 }
56da5a46
RGS
541 if (t < PL_bufptr && isSPACE(*t))
542 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
147e3846 543 "\t(Do you need to predeclare %" UTF8f "?)\n",
b17a0679 544 UTF8fARG(UTF, t - PL_oldoldbufptr, PL_oldoldbufptr));
56da5a46
RGS
545 }
546 else {
547 assert(s >= oldbp);
548 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
147e3846 549 "\t(Missing operator before %" UTF8f "?)\n",
b17a0679 550 UTF8fARG(UTF, s - oldbp, oldbp));
56da5a46 551 }
07c798fb 552 }
3280af22 553 PL_bufptr = oldbp;
8990e307
LW
554}
555
ffb4593c
NT
556/*
557 * S_missingterm
558 * Complain about missing quote/regexp/heredoc terminator.
d4c19fe8 559 * If it's called with NULL then it cauterizes the line buffer.
ffb4593c
NT
560 * If we're in a delimited string and the delimiter is a control
561 * character, it's reformatted into a two-char sequence like ^C.
562 * This is fatal.
563 */
564
76e3520e 565STATIC void
cea2e8a9 566S_missingterm(pTHX_ char *s)
8990e307 567{
e487ff5e 568 char tmpbuf[UTF8_MAXBYTES + 1];
8990e307 569 char q;
cb650135
FC
570 bool uni = FALSE;
571 SV *sv;
8990e307 572 if (s) {
9d4ba2ae 573 char * const nl = strrchr(s,'\n');
d2719217 574 if (nl)
8990e307 575 *nl = '\0';
7f7f7d08 576 uni = UTF;
8990e307 577 }
cb650135 578 else if (PL_multi_close < 32) {
8990e307 579 *tmpbuf = '^';
585ec06d 580 tmpbuf[1] = (char)toCTRL(PL_multi_close);
8990e307
LW
581 tmpbuf[2] = '\0';
582 s = tmpbuf;
583 }
584 else {
cb650135
FC
585 if (LIKELY(PL_multi_close < 256)) {
586 *tmpbuf = (char)PL_multi_close;
587 tmpbuf[1] = '\0';
588 }
589 else {
590 uni = TRUE;
591 *uvchr_to_utf8((U8 *)tmpbuf, PL_multi_close) = 0;
592 }
8990e307
LW
593 s = tmpbuf;
594 }
595 q = strchr(s,'"') ? '\'' : '"';
cb650135
FC
596 sv = sv_2mortal(newSVpv(s,0));
597 if (uni)
598 SvUTF8_on(sv);
147e3846 599 Perl_croak(aTHX_ "Can't find string terminator %c%" SVf
cb650135 600 "%c anywhere before EOF",q,SVfARG(sv),q);
463ee0b2 601}
79072805 602
dd0ac2b9
FC
603#include "feature.h"
604
0d863452 605/*
0d863452
RH
606 * Check whether the named feature is enabled.
607 */
26ea9e12 608bool
3fff3427 609Perl_feature_is_enabled(pTHX_ const char *const name, STRLEN namelen)
0d863452 610{
4a731d7b 611 char he_name[8 + MAX_FEATURE_LEN] = "feature_";
7918f24d
NC
612
613 PERL_ARGS_ASSERT_FEATURE_IS_ENABLED;
ca4d40c4
FC
614
615 assert(CURRENT_FEATURE_BUNDLE == FEATURE_BUNDLE_CUSTOM);
7918f24d 616
26ea9e12
NC
617 if (namelen > MAX_FEATURE_LEN)
618 return FALSE;
3fff3427 619 memcpy(&he_name[8], name, namelen);
7d69d4a6 620
c8ca97b0
NC
621 return cBOOL(cop_hints_fetch_pvn(PL_curcop, he_name, 8 + namelen, 0,
622 REFCOUNTED_HE_EXISTS));
0d863452
RH
623}
624
ffb4593c 625/*
9cbb5ea2
GS
626 * experimental text filters for win32 carriage-returns, utf16-to-utf8 and
627 * utf16-to-utf8-reversed.
ffb4593c
NT
628 */
629
c39cd008
GS
630#ifdef PERL_CR_FILTER
631static void
632strip_return(SV *sv)
633{
eb578fdb
KW
634 const char *s = SvPVX_const(sv);
635 const char * const e = s + SvCUR(sv);
7918f24d
NC
636
637 PERL_ARGS_ASSERT_STRIP_RETURN;
638
c39cd008
GS
639 /* outer loop optimized to do nothing if there are no CR-LFs */
640 while (s < e) {
641 if (*s++ == '\r' && *s == '\n') {
642 /* hit a CR-LF, need to copy the rest */
eb578fdb 643 char *d = s - 1;
c39cd008
GS
644 *d++ = *s++;
645 while (s < e) {
646 if (*s == '\r' && s[1] == '\n')
647 s++;
648 *d++ = *s++;
649 }
650 SvCUR(sv) -= s - d;
651 return;
652 }
653 }
654}
a868473f 655
76e3520e 656STATIC I32
c39cd008 657S_cr_textfilter(pTHX_ int idx, SV *sv, int maxlen)
a868473f 658{
f54cb97a 659 const I32 count = FILTER_READ(idx+1, sv, maxlen);
c39cd008
GS
660 if (count > 0 && !maxlen)
661 strip_return(sv);
662 return count;
a868473f
NIS
663}
664#endif
665
ffb4593c 666/*
8eaa0acf
Z
667=for apidoc Amx|void|lex_start|SV *line|PerlIO *rsfp|U32 flags
668
669Creates and initialises a new lexer/parser state object, supplying
670a context in which to lex and parse from a new source of Perl code.
671A pointer to the new state object is placed in L</PL_parser>. An entry
672is made on the save stack so that upon unwinding the new state object
673will be destroyed and the former value of L</PL_parser> will be restored.
674Nothing else need be done to clean up the parsing context.
675
2d7f6611 676The code to be parsed comes from C<line> and C<rsfp>. C<line>, if
8eaa0acf 677non-null, provides a string (in SV form) containing code to be parsed.
2d7f6611
KW
678A copy of the string is made, so subsequent modification of C<line>
679does not affect parsing. C<rsfp>, if non-null, provides an input stream
8eaa0acf 680from which code will be read to be parsed. If both are non-null, the
2d7f6611
KW
681code in C<line> comes first and must consist of complete lines of input,
682and C<rsfp> supplies the remainder of the source.
8eaa0acf 683
2d7f6611 684The C<flags> parameter is reserved for future use. Currently it is only
e368b3bd 685used by perl internally, so extensions should always pass zero.
8eaa0acf
Z
686
687=cut
688*/
ffb4593c 689
27fcb6ee 690/* LEX_START_SAME_FILTER indicates that this is not a new file, so it
87606032
NC
691 can share filters with the current parser.
692 LEX_START_DONT_CLOSE indicates that the file handle wasn't opened by the
693 caller, hence isn't owned by the parser, so shouldn't be closed on parser
694 destruction. This is used to handle the case of defaulting to reading the
695 script from the standard input because no filename was given on the command
696 line (without getting confused by situation where STDIN has been closed, so
697 the script handle is opened on fd 0) */
27fcb6ee 698
a0d0e21e 699void
8eaa0acf 700Perl_lex_start(pTHX_ SV *line, PerlIO *rsfp, U32 flags)
79072805 701{
6ef55633 702 const char *s = NULL;
5486870f 703 yy_parser *parser, *oparser;
60d63348 704 if (flags && flags & ~LEX_START_FLAGS)
8eaa0acf 705 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_start");
acdf0a21
DM
706
707 /* create and initialise a parser */
708
199e78b7 709 Newxz(parser, 1, yy_parser);
5486870f 710 parser->old_parser = oparser = PL_parser;
acdf0a21
DM
711 PL_parser = parser;
712
28ac2b49 713 parser->stack = NULL;
df13534a 714 parser->stack_max1 = NULL;
28ac2b49 715 parser->ps = NULL;
acdf0a21 716
e3abe207
DM
717 /* on scope exit, free this parser and restore any outer one */
718 SAVEPARSER(parser);
7c4baf47 719 parser->saved_curcop = PL_curcop;
e3abe207 720
acdf0a21 721 /* initialise lexer state */
8990e307 722
fb205e7a 723 parser->nexttoke = 0;
ca4cfd28 724 parser->error_count = oparser ? oparser->error_count : 0;
7f1c3e8c 725 parser->copline = parser->preambling = NOLINE;
5afb0a62 726 parser->lex_state = LEX_NORMAL;
c2598295 727 parser->expect = XSTATE;
2f9285f8 728 parser->rsfp = rsfp;
27fcb6ee
FC
729 parser->rsfp_filters =
730 !(flags & LEX_START_SAME_FILTER) || !oparser
d3cd8e11
FC
731 ? NULL
732 : MUTABLE_AV(SvREFCNT_inc(
733 oparser->rsfp_filters
734 ? oparser->rsfp_filters
735 : (oparser->rsfp_filters = newAV())
736 ));
2f9285f8 737
199e78b7
DM
738 Newx(parser->lex_brackstack, 120, char);
739 Newx(parser->lex_casestack, 12, char);
740 *parser->lex_casestack = '\0';
d794b522 741 Newxz(parser->lex_shared, 1, LEXSHARED);
02b34bbe 742
10efb74f 743 if (line) {
0528fd32 744 STRLEN len;
10efb74f 745 s = SvPV_const(line, len);
0abcdfa4
FC
746 parser->linestr = flags & LEX_START_COPIED
747 ? SvREFCNT_inc_simple_NN(line)
748 : newSVpvn_flags(s, len, SvUTF8(line));
b3dd0aba
FC
749 if (!rsfp)
750 sv_catpvs(parser->linestr, "\n;");
0abcdfa4 751 } else {
bf1b738b 752 parser->linestr = newSVpvn("\n;", rsfp ? 1 : 2);
8990e307 753 }
f06b5848
DM
754 parser->oldoldbufptr =
755 parser->oldbufptr =
756 parser->bufptr =
757 parser->linestart = SvPVX(parser->linestr);
758 parser->bufend = parser->bufptr + SvCUR(parser->linestr);
759 parser->last_lop = parser->last_uni = NULL;
b54f893d 760
6d59e610 761 STATIC_ASSERT_STMT(FITS_IN_8_BITS(LEX_IGNORE_UTF8_HINTS|LEX_EVALBYTES
b54f893d
KW
762 |LEX_DONT_CLOSE_RSFP));
763 parser->lex_flags = (U8) (flags & (LEX_IGNORE_UTF8_HINTS|LEX_EVALBYTES
764 |LEX_DONT_CLOSE_RSFP));
737c24fc 765
60d63348 766 parser->in_pod = parser->filtered = 0;
79072805 767}
a687059c 768
e3abe207
DM
769
770/* delete a parser object */
771
772void
773Perl_parser_free(pTHX_ const yy_parser *parser)
774{
7918f24d
NC
775 PERL_ARGS_ASSERT_PARSER_FREE;
776
7c4baf47 777 PL_curcop = parser->saved_curcop;
bdc0bf6f
DM
778 SvREFCNT_dec(parser->linestr);
779
87606032 780 if (PL_parser->lex_flags & LEX_DONT_CLOSE_RSFP)
2f9285f8 781 PerlIO_clearerr(parser->rsfp);
407f8cf2
KW
782 else if (parser->rsfp && (!parser->old_parser
783 || (parser->old_parser && parser->rsfp != parser->old_parser->rsfp)))
2f9285f8 784 PerlIO_close(parser->rsfp);
5486870f 785 SvREFCNT_dec(parser->rsfp_filters);
10002bc1 786 SvREFCNT_dec(parser->lex_stuff);
7ef70b3d 787 SvREFCNT_dec(parser->lex_sub_repl);
3ac7ff8f
FC
788
789 Safefree(parser->lex_brackstack);
790 Safefree(parser->lex_casestack);
791 Safefree(parser->lex_shared);
792 PL_parser = parser->old_parser;
793 Safefree(parser);
794}
795
796void
797Perl_parser_free_nexttoke_ops(pTHX_ yy_parser *parser, OPSLAB *slab)
798{
3ac7ff8f 799 I32 nexttoke = parser->nexttoke;
3ac7ff8f 800 PERL_ARGS_ASSERT_PARSER_FREE_NEXTTOKE_OPS;
3ce3dcd9 801 while (nexttoke--) {
3ac7ff8f
FC
802 if (S_is_opval_token(parser->nexttype[nexttoke] & 0xffff)
803 && parser->nextval[nexttoke].opval
804 && parser->nextval[nexttoke].opval->op_slabbed
805 && OpSLAB(parser->nextval[nexttoke].opval) == slab) {
3ce3dcd9 806 op_free(parser->nextval[nexttoke].opval);
3ac7ff8f
FC
807 parser->nextval[nexttoke].opval = NULL;
808 }
3ce3dcd9 809 }
e3abe207
DM
810}
811
812
ffb4593c 813/*
f0e67a1d
Z
814=for apidoc AmxU|SV *|PL_parser-E<gt>linestr
815
816Buffer scalar containing the chunk currently under consideration of the
817text currently being lexed. This is always a plain string scalar (for
818which C<SvPOK> is true). It is not intended to be used as a scalar by
819normal scalar means; instead refer to the buffer directly by the pointer
820variables described below.
821
822The lexer maintains various C<char*> pointers to things in the
823C<PL_parser-E<gt>linestr> buffer. If C<PL_parser-E<gt>linestr> is ever
824reallocated, all of these pointers must be updated. Don't attempt to
825do this manually, but rather use L</lex_grow_linestr> if you need to
826reallocate the buffer.
827
828The content of the text chunk in the buffer is commonly exactly one
829complete line of input, up to and including a newline terminator,
830but there are situations where it is otherwise. The octets of the
831buffer may be intended to be interpreted as either UTF-8 or Latin-1.
832The function L</lex_bufutf8> tells you which. Do not use the C<SvUTF8>
833flag on this scalar, which may disagree with it.
834
835For direct examination of the buffer, the variable
836L</PL_parser-E<gt>bufend> points to the end of the buffer. The current
837lexing position is pointed to by L</PL_parser-E<gt>bufptr>. Direct use
838of these pointers is usually preferable to examination of the scalar
839through normal scalar means.
840
841=for apidoc AmxU|char *|PL_parser-E<gt>bufend
842
843Direct pointer to the end of the chunk of text currently being lexed, the
844end of the lexer buffer. This is equal to C<SvPVX(PL_parser-E<gt>linestr)
6602b933 845+ SvCUR(PL_parser-E<gt>linestr)>. A C<NUL> character (zero octet) is
f0e67a1d
Z
846always located at the end of the buffer, and does not count as part of
847the buffer's contents.
848
849=for apidoc AmxU|char *|PL_parser-E<gt>bufptr
850
851Points to the current position of lexing inside the lexer buffer.
852Characters around this point may be freely examined, within
853the range delimited by C<SvPVX(L</PL_parser-E<gt>linestr>)> and
854L</PL_parser-E<gt>bufend>. The octets of the buffer may be intended to be
855interpreted as either UTF-8 or Latin-1, as indicated by L</lex_bufutf8>.
856
857Lexing code (whether in the Perl core or not) moves this pointer past
858the characters that it consumes. It is also expected to perform some
859bookkeeping whenever a newline character is consumed. This movement
860can be more conveniently performed by the function L</lex_read_to>,
861which handles newlines appropriately.
862
863Interpretation of the buffer's octets can be abstracted out by
864using the slightly higher-level functions L</lex_peek_unichar> and
865L</lex_read_unichar>.
866
867=for apidoc AmxU|char *|PL_parser-E<gt>linestart
868
869Points to the start of the current line inside the lexer buffer.
870This is useful for indicating at which column an error occurred, and
871not much else. This must be updated by any lexing code that consumes
872a newline; the function L</lex_read_to> handles this detail.
873
874=cut
875*/
876
877/*
878=for apidoc Amx|bool|lex_bufutf8
879
880Indicates whether the octets in the lexer buffer
881(L</PL_parser-E<gt>linestr>) should be interpreted as the UTF-8 encoding
882of Unicode characters. If not, they should be interpreted as Latin-1
883characters. This is analogous to the C<SvUTF8> flag for scalars.
884
885In UTF-8 mode, it is not guaranteed that the lexer buffer actually
886contains valid UTF-8. Lexing code must be robust in the face of invalid
887encoding.
888
889The actual C<SvUTF8> flag of the L</PL_parser-E<gt>linestr> scalar
890is significant, but not the whole story regarding the input character
891encoding. Normally, when a file is being read, the scalar contains octets
892and its C<SvUTF8> flag is off, but the octets should be interpreted as
893UTF-8 if the C<use utf8> pragma is in effect. During a string eval,
894however, the scalar may have the C<SvUTF8> flag on, and in this case its
895octets should be interpreted as UTF-8 unless the C<use bytes> pragma
896is in effect. This logic may change in the future; use this function
897instead of implementing the logic yourself.
898
899=cut
900*/
901
902bool
903Perl_lex_bufutf8(pTHX)
904{
905 return UTF;
906}
907
908/*
909=for apidoc Amx|char *|lex_grow_linestr|STRLEN len
910
911Reallocates the lexer buffer (L</PL_parser-E<gt>linestr>) to accommodate
2d7f6611 912at least C<len> octets (including terminating C<NUL>). Returns a
f0e67a1d
Z
913pointer to the reallocated buffer. This is necessary before making
914any direct modification of the buffer that would increase its length.
915L</lex_stuff_pvn> provides a more convenient way to insert text into
916the buffer.
917
918Do not use C<SvGROW> or C<sv_grow> directly on C<PL_parser-E<gt>linestr>;
919this function updates all of the lexer's variables that point directly
920into the buffer.
921
922=cut
923*/
924
925char *
926Perl_lex_grow_linestr(pTHX_ STRLEN len)
927{
928 SV *linestr;
929 char *buf;
930 STRLEN bufend_pos, bufptr_pos, oldbufptr_pos, oldoldbufptr_pos;
c7641931 931 STRLEN linestart_pos, last_uni_pos, last_lop_pos, re_eval_start_pos;
98d5e3ef
DM
932 bool current;
933
f0e67a1d
Z
934 linestr = PL_parser->linestr;
935 buf = SvPVX(linestr);
936 if (len <= SvLEN(linestr))
937 return buf;
98d5e3ef
DM
938
939 /* Is the lex_shared linestr SV the same as the current linestr SV?
940 * Only in this case does re_eval_start need adjusting, since it
941 * points within lex_shared->ls_linestr's buffer */
b1b8fb6a
DM
942 current = ( !PL_parser->lex_shared->ls_linestr
943 || linestr == PL_parser->lex_shared->ls_linestr);
98d5e3ef 944
f0e67a1d
Z
945 bufend_pos = PL_parser->bufend - buf;
946 bufptr_pos = PL_parser->bufptr - buf;
947 oldbufptr_pos = PL_parser->oldbufptr - buf;
948 oldoldbufptr_pos = PL_parser->oldoldbufptr - buf;
949 linestart_pos = PL_parser->linestart - buf;
950 last_uni_pos = PL_parser->last_uni ? PL_parser->last_uni - buf : 0;
951 last_lop_pos = PL_parser->last_lop ? PL_parser->last_lop - buf : 0;
98d5e3ef 952 re_eval_start_pos = (current && PL_parser->lex_shared->re_eval_start) ?
3328ab5a 953 PL_parser->lex_shared->re_eval_start - buf : 0;
c7641931 954
f0e67a1d 955 buf = sv_grow(linestr, len);
c7641931 956
f0e67a1d
Z
957 PL_parser->bufend = buf + bufend_pos;
958 PL_parser->bufptr = buf + bufptr_pos;
959 PL_parser->oldbufptr = buf + oldbufptr_pos;
960 PL_parser->oldoldbufptr = buf + oldoldbufptr_pos;
961 PL_parser->linestart = buf + linestart_pos;
962 if (PL_parser->last_uni)
963 PL_parser->last_uni = buf + last_uni_pos;
964 if (PL_parser->last_lop)
965 PL_parser->last_lop = buf + last_lop_pos;
98d5e3ef 966 if (current && PL_parser->lex_shared->re_eval_start)
3328ab5a 967 PL_parser->lex_shared->re_eval_start = buf + re_eval_start_pos;
f0e67a1d
Z
968 return buf;
969}
970
971/*
83aa740e 972=for apidoc Amx|void|lex_stuff_pvn|const char *pv|STRLEN len|U32 flags
f0e67a1d
Z
973
974Insert characters into the lexer buffer (L</PL_parser-E<gt>linestr>),
975immediately after the current lexing point (L</PL_parser-E<gt>bufptr>),
976reallocating the buffer if necessary. This means that lexing code that
977runs later will see the characters as if they had appeared in the input.
978It is not recommended to do this as part of normal parsing, and most
979uses of this facility run the risk of the inserted characters being
980interpreted in an unintended manner.
981
2d7f6611
KW
982The string to be inserted is represented by C<len> octets starting
983at C<pv>. These octets are interpreted as either UTF-8 or Latin-1,
984according to whether the C<LEX_STUFF_UTF8> flag is set in C<flags>.
f0e67a1d
Z
985The characters are recoded for the lexer buffer, according to how the
986buffer is currently being interpreted (L</lex_bufutf8>). If a string
9dcc53ea 987to be inserted is available as a Perl scalar, the L</lex_stuff_sv>
f0e67a1d
Z
988function is more convenient.
989
990=cut
991*/
992
993void
83aa740e 994Perl_lex_stuff_pvn(pTHX_ const char *pv, STRLEN len, U32 flags)
f0e67a1d 995{
749123ff 996 dVAR;
f0e67a1d
Z
997 char *bufptr;
998 PERL_ARGS_ASSERT_LEX_STUFF_PVN;
999 if (flags & ~(LEX_STUFF_UTF8))
1000 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_stuff_pvn");
1001 if (UTF) {
1002 if (flags & LEX_STUFF_UTF8) {
1003 goto plain_copy;
1004 } else {
54d004e8 1005 STRLEN highhalf = 0; /* Count of variants */
83aa740e 1006 const char *p, *e = pv+len;
54d004e8
KW
1007 for (p = pv; p != e; p++) {
1008 if (! UTF8_IS_INVARIANT(*p)) {
1009 highhalf++;
1010 }
1011 }
f0e67a1d
Z
1012 if (!highhalf)
1013 goto plain_copy;
1014 lex_grow_linestr(SvCUR(PL_parser->linestr)+1+len+highhalf);
1015 bufptr = PL_parser->bufptr;
1016 Move(bufptr, bufptr+len+highhalf, PL_parser->bufend+1-bufptr, char);
255fdf19
Z
1017 SvCUR_set(PL_parser->linestr,
1018 SvCUR(PL_parser->linestr) + len+highhalf);
f0e67a1d
Z
1019 PL_parser->bufend += len+highhalf;
1020 for (p = pv; p != e; p++) {
1021 U8 c = (U8)*p;
54d004e8
KW
1022 if (! UTF8_IS_INVARIANT(c)) {
1023 *bufptr++ = UTF8_TWO_BYTE_HI(c);
1024 *bufptr++ = UTF8_TWO_BYTE_LO(c);
f0e67a1d
Z
1025 } else {
1026 *bufptr++ = (char)c;
1027 }
1028 }
1029 }
1030 } else {
1031 if (flags & LEX_STUFF_UTF8) {
1032 STRLEN highhalf = 0;
83aa740e 1033 const char *p, *e = pv+len;
f0e67a1d
Z
1034 for (p = pv; p != e; p++) {
1035 U8 c = (U8)*p;
54d004e8 1036 if (UTF8_IS_ABOVE_LATIN1(c)) {
f0e67a1d
Z
1037 Perl_croak(aTHX_ "Lexing code attempted to stuff "
1038 "non-Latin-1 character into Latin-1 input");
54d004e8 1039 } else if (UTF8_IS_NEXT_CHAR_DOWNGRADEABLE(p, e)) {
f0e67a1d
Z
1040 p++;
1041 highhalf++;
75219bac
KW
1042 } else if (! UTF8_IS_INVARIANT(c)) {
1043 _force_out_malformed_utf8_message((U8 *) p, (U8 *) e,
1044 0,
1045 1 /* 1 means die */ );
1046 NOT_REACHED; /* NOTREACHED */
f0e67a1d
Z
1047 }
1048 }
1049 if (!highhalf)
1050 goto plain_copy;
1051 lex_grow_linestr(SvCUR(PL_parser->linestr)+1+len-highhalf);
1052 bufptr = PL_parser->bufptr;
1053 Move(bufptr, bufptr+len-highhalf, PL_parser->bufend+1-bufptr, char);
255fdf19
Z
1054 SvCUR_set(PL_parser->linestr,
1055 SvCUR(PL_parser->linestr) + len-highhalf);
f0e67a1d 1056 PL_parser->bufend += len-highhalf;
54d004e8
KW
1057 p = pv;
1058 while (p < e) {
1059 if (UTF8_IS_INVARIANT(*p)) {
1060 *bufptr++ = *p;
1061 p++;
f0e67a1d 1062 }
54d004e8
KW
1063 else {
1064 assert(p < e -1 );
a62b247b 1065 *bufptr++ = EIGHT_BIT_UTF8_TO_NATIVE(*p, *(p+1));
54d004e8
KW
1066 p += 2;
1067 }
f0e67a1d
Z
1068 }
1069 } else {
54d004e8 1070 plain_copy:
f0e67a1d
Z
1071 lex_grow_linestr(SvCUR(PL_parser->linestr)+1+len);
1072 bufptr = PL_parser->bufptr;
1073 Move(bufptr, bufptr+len, PL_parser->bufend+1-bufptr, char);
255fdf19 1074 SvCUR_set(PL_parser->linestr, SvCUR(PL_parser->linestr) + len);
f0e67a1d
Z
1075 PL_parser->bufend += len;
1076 Copy(pv, bufptr, len, char);
1077 }
1078 }
1079}
1080
1081/*
9dcc53ea
Z
1082=for apidoc Amx|void|lex_stuff_pv|const char *pv|U32 flags
1083
1084Insert characters into the lexer buffer (L</PL_parser-E<gt>linestr>),
1085immediately after the current lexing point (L</PL_parser-E<gt>bufptr>),
1086reallocating the buffer if necessary. This means that lexing code that
1087runs later will see the characters as if they had appeared in the input.
1088It is not recommended to do this as part of normal parsing, and most
1089uses of this facility run the risk of the inserted characters being
1090interpreted in an unintended manner.
1091
2d7f6611 1092The string to be inserted is represented by octets starting at C<pv>
9dcc53ea
Z
1093and continuing to the first nul. These octets are interpreted as either
1094UTF-8 or Latin-1, according to whether the C<LEX_STUFF_UTF8> flag is set
2d7f6611 1095in C<flags>. The characters are recoded for the lexer buffer, according
9dcc53ea
Z
1096to how the buffer is currently being interpreted (L</lex_bufutf8>).
1097If it is not convenient to nul-terminate a string to be inserted, the
1098L</lex_stuff_pvn> function is more appropriate.
1099
1100=cut
1101*/
1102
1103void
1104Perl_lex_stuff_pv(pTHX_ const char *pv, U32 flags)
1105{
1106 PERL_ARGS_ASSERT_LEX_STUFF_PV;
1107 lex_stuff_pvn(pv, strlen(pv), flags);
1108}
1109
1110/*
f0e67a1d
Z
1111=for apidoc Amx|void|lex_stuff_sv|SV *sv|U32 flags
1112
1113Insert characters into the lexer buffer (L</PL_parser-E<gt>linestr>),
1114immediately after the current lexing point (L</PL_parser-E<gt>bufptr>),
1115reallocating the buffer if necessary. This means that lexing code that
1116runs later will see the characters as if they had appeared in the input.
1117It is not recommended to do this as part of normal parsing, and most
1118uses of this facility run the risk of the inserted characters being
1119interpreted in an unintended manner.
1120
2d7f6611 1121The string to be inserted is the string value of C<sv>. The characters
f0e67a1d 1122are recoded for the lexer buffer, according to how the buffer is currently
9dcc53ea 1123being interpreted (L</lex_bufutf8>). If a string to be inserted is
f0e67a1d
Z
1124not already a Perl scalar, the L</lex_stuff_pvn> function avoids the
1125need to construct a scalar.
1126
1127=cut
1128*/
1129
1130void
1131Perl_lex_stuff_sv(pTHX_ SV *sv, U32 flags)
1132{
1133 char *pv;
1134 STRLEN len;
1135 PERL_ARGS_ASSERT_LEX_STUFF_SV;
1136 if (flags)
1137 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_stuff_sv");
1138 pv = SvPV(sv, len);
1139 lex_stuff_pvn(pv, len, flags | (SvUTF8(sv) ? LEX_STUFF_UTF8 : 0));
1140}
1141
1142/*
1143=for apidoc Amx|void|lex_unstuff|char *ptr
1144
1145Discards text about to be lexed, from L</PL_parser-E<gt>bufptr> up to
2d7f6611 1146C<ptr>. Text following C<ptr> will be moved, and the buffer shortened.
f0e67a1d
Z
1147This hides the discarded text from any lexing code that runs later,
1148as if the text had never appeared.
1149
1150This is not the normal way to consume lexed text. For that, use
1151L</lex_read_to>.
1152
1153=cut
1154*/
1155
1156void
1157Perl_lex_unstuff(pTHX_ char *ptr)
1158{
1159 char *buf, *bufend;
1160 STRLEN unstuff_len;
1161 PERL_ARGS_ASSERT_LEX_UNSTUFF;
1162 buf = PL_parser->bufptr;
1163 if (ptr < buf)
1164 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_unstuff");
1165 if (ptr == buf)
1166 return;
1167 bufend = PL_parser->bufend;
1168 if (ptr > bufend)
1169 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_unstuff");
1170 unstuff_len = ptr - buf;
1171 Move(ptr, buf, bufend+1-ptr, char);
1172 SvCUR_set(PL_parser->linestr, SvCUR(PL_parser->linestr) - unstuff_len);
1173 PL_parser->bufend = bufend - unstuff_len;
1174}
1175
1176/*
1177=for apidoc Amx|void|lex_read_to|char *ptr
1178
1179Consume text in the lexer buffer, from L</PL_parser-E<gt>bufptr> up
2d7f6611 1180to C<ptr>. This advances L</PL_parser-E<gt>bufptr> to match C<ptr>,
f0e67a1d
Z
1181performing the correct bookkeeping whenever a newline character is passed.
1182This is the normal way to consume lexed text.
1183
1184Interpretation of the buffer's octets can be abstracted out by
1185using the slightly higher-level functions L</lex_peek_unichar> and
1186L</lex_read_unichar>.
1187
1188=cut
1189*/
1190
1191void
1192Perl_lex_read_to(pTHX_ char *ptr)
1193{
1194 char *s;
1195 PERL_ARGS_ASSERT_LEX_READ_TO;
1196 s = PL_parser->bufptr;
1197 if (ptr < s || ptr > PL_parser->bufend)
1198 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_read_to");
1199 for (; s != ptr; s++)
1200 if (*s == '\n') {
83944c01 1201 COPLINE_INC_WITH_HERELINES;
f0e67a1d
Z
1202 PL_parser->linestart = s+1;
1203 }
1204 PL_parser->bufptr = ptr;
1205}
1206
1207/*
1208=for apidoc Amx|void|lex_discard_to|char *ptr
1209
1210Discards the first part of the L</PL_parser-E<gt>linestr> buffer,
2d7f6611
KW
1211up to C<ptr>. The remaining content of the buffer will be moved, and
1212all pointers into the buffer updated appropriately. C<ptr> must not
f0e67a1d
Z
1213be later in the buffer than the position of L</PL_parser-E<gt>bufptr>:
1214it is not permitted to discard text that has yet to be lexed.
1215
1216Normally it is not necessarily to do this directly, because it suffices to
1217use the implicit discarding behaviour of L</lex_next_chunk> and things
1218based on it. However, if a token stretches across multiple lines,
1f317c95 1219and the lexing code has kept multiple lines of text in the buffer for
f0e67a1d
Z
1220that purpose, then after completion of the token it would be wise to
1221explicitly discard the now-unneeded earlier lines, to avoid future
1222multi-line tokens growing the buffer without bound.
1223
1224=cut
1225*/
1226
1227void
1228Perl_lex_discard_to(pTHX_ char *ptr)
1229{
1230 char *buf;
1231 STRLEN discard_len;
1232 PERL_ARGS_ASSERT_LEX_DISCARD_TO;
1233 buf = SvPVX(PL_parser->linestr);
1234 if (ptr < buf)
1235 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_discard_to");
1236 if (ptr == buf)
1237 return;
1238 if (ptr > PL_parser->bufptr)
1239 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_discard_to");
1240 discard_len = ptr - buf;
1241 if (PL_parser->oldbufptr < ptr)
1242 PL_parser->oldbufptr = ptr;
1243 if (PL_parser->oldoldbufptr < ptr)
1244 PL_parser->oldoldbufptr = ptr;
1245 if (PL_parser->last_uni && PL_parser->last_uni < ptr)
1246 PL_parser->last_uni = NULL;
1247 if (PL_parser->last_lop && PL_parser->last_lop < ptr)
1248 PL_parser->last_lop = NULL;
1249 Move(ptr, buf, PL_parser->bufend+1-ptr, char);
1250 SvCUR_set(PL_parser->linestr, SvCUR(PL_parser->linestr) - discard_len);
1251 PL_parser->bufend -= discard_len;
1252 PL_parser->bufptr -= discard_len;
1253 PL_parser->oldbufptr -= discard_len;
1254 PL_parser->oldoldbufptr -= discard_len;
1255 if (PL_parser->last_uni)
1256 PL_parser->last_uni -= discard_len;
1257 if (PL_parser->last_lop)
1258 PL_parser->last_lop -= discard_len;
1259}
1260
1261/*
1262=for apidoc Amx|bool|lex_next_chunk|U32 flags
1263
1264Reads in the next chunk of text to be lexed, appending it to
1265L</PL_parser-E<gt>linestr>. This should be called when lexing code has
1266looked to the end of the current chunk and wants to know more. It is
1267usual, but not necessary, for lexing to have consumed the entirety of
1268the current chunk at this time.
1269
1270If L</PL_parser-E<gt>bufptr> is pointing to the very end of the current
1271chunk (i.e., the current chunk has been entirely consumed), normally the
1272current chunk will be discarded at the same time that the new chunk is
c5608a1f 1273read in. If C<flags> has the C<LEX_KEEP_PREVIOUS> bit set, the current chunk
f0e67a1d
Z
1274will not be discarded. If the current chunk has not been entirely
1275consumed, then it will not be discarded regardless of the flag.
1276
1277Returns true if some new text was added to the buffer, or false if the
1278buffer has reached the end of the input text.
1279
1280=cut
1281*/
1282
1283#define LEX_FAKE_EOF 0x80000000
e47d32dc 1284#define LEX_NO_TERM 0x40000000 /* here-doc */
f0e67a1d
Z
1285
1286bool
1287Perl_lex_next_chunk(pTHX_ U32 flags)
1288{
1289 SV *linestr;
1290 char *buf;
1291 STRLEN old_bufend_pos, new_bufend_pos;
1292 STRLEN bufptr_pos, oldbufptr_pos, oldoldbufptr_pos;
1293 STRLEN linestart_pos, last_uni_pos, last_lop_pos;
17cc9359 1294 bool got_some_for_debugger = 0;
f0e67a1d 1295 bool got_some;
6cdc5cd8
KW
1296 const U8* first_bad_char_loc;
1297
112d1284 1298 if (flags & ~(LEX_KEEP_PREVIOUS|LEX_FAKE_EOF|LEX_NO_TERM))
f0e67a1d 1299 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_next_chunk");
d27f4b91 1300 if (!(flags & LEX_NO_TERM) && PL_lex_inwhat)
e47d32dc 1301 return FALSE;
f0e67a1d
Z
1302 linestr = PL_parser->linestr;
1303 buf = SvPVX(linestr);
407f8cf2
KW
1304 if (!(flags & LEX_KEEP_PREVIOUS)
1305 && PL_parser->bufptr == PL_parser->bufend)
1306 {
f0e67a1d
Z
1307 old_bufend_pos = bufptr_pos = oldbufptr_pos = oldoldbufptr_pos = 0;
1308 linestart_pos = 0;
1309 if (PL_parser->last_uni != PL_parser->bufend)
1310 PL_parser->last_uni = NULL;
1311 if (PL_parser->last_lop != PL_parser->bufend)
1312 PL_parser->last_lop = NULL;
1313 last_uni_pos = last_lop_pos = 0;
1314 *buf = 0;
1315 SvCUR(linestr) = 0;
1316 } else {
1317 old_bufend_pos = PL_parser->bufend - buf;
1318 bufptr_pos = PL_parser->bufptr - buf;
1319 oldbufptr_pos = PL_parser->oldbufptr - buf;
1320 oldoldbufptr_pos = PL_parser->oldoldbufptr - buf;
1321 linestart_pos = PL_parser->linestart - buf;
1322 last_uni_pos = PL_parser->last_uni ? PL_parser->last_uni - buf : 0;
1323 last_lop_pos = PL_parser->last_lop ? PL_parser->last_lop - buf : 0;
1324 }
1325 if (flags & LEX_FAKE_EOF) {
1326 goto eof;
60d63348 1327 } else if (!PL_parser->rsfp && !PL_parser->filtered) {
f0e67a1d
Z
1328 got_some = 0;
1329 } else if (filter_gets(linestr, old_bufend_pos)) {
1330 got_some = 1;
17cc9359 1331 got_some_for_debugger = 1;
112d1284
FC
1332 } else if (flags & LEX_NO_TERM) {
1333 got_some = 0;
f0e67a1d 1334 } else {
580561a3 1335 if (!SvPOK(linestr)) /* can get undefined by filter_gets */
847cc851 1336 SvPVCLEAR(linestr);
f0e67a1d
Z
1337 eof:
1338 /* End of real input. Close filehandle (unless it was STDIN),
1339 * then add implicit termination.
1340 */
87606032 1341 if (PL_parser->lex_flags & LEX_DONT_CLOSE_RSFP)
f0e67a1d
Z
1342 PerlIO_clearerr(PL_parser->rsfp);
1343 else if (PL_parser->rsfp)
1344 (void)PerlIO_close(PL_parser->rsfp);
1345 PL_parser->rsfp = NULL;
60d63348 1346 PL_parser->in_pod = PL_parser->filtered = 0;
f0e67a1d
Z
1347 if (!PL_in_eval && PL_minus_p) {
1348 sv_catpvs(linestr,
1349 /*{*/";}continue{print or die qq(-p destination: $!\\n);}");
1350 PL_minus_n = PL_minus_p = 0;
1351 } else if (!PL_in_eval && PL_minus_n) {
1352 sv_catpvs(linestr, /*{*/";}");
1353 PL_minus_n = 0;
1354 } else
1355 sv_catpvs(linestr, ";");
1356 got_some = 1;
1357 }
1358 buf = SvPVX(linestr);
1359 new_bufend_pos = SvCUR(linestr);
1360 PL_parser->bufend = buf + new_bufend_pos;
1361 PL_parser->bufptr = buf + bufptr_pos;
6cdc5cd8
KW
1362
1363 if (UTF && ! is_utf8_string_loc((U8 *) PL_parser->bufptr,
1364 PL_parser->bufend - PL_parser->bufptr,
1365 &first_bad_char_loc))
1366 {
6cdc5cd8
KW
1367 _force_out_malformed_utf8_message(first_bad_char_loc,
1368 (U8 *) PL_parser->bufend,
1369 0,
1370 1 /* 1 means die */ );
1371 NOT_REACHED; /* NOTREACHED */
1372 }
1373
f0e67a1d
Z
1374 PL_parser->oldbufptr = buf + oldbufptr_pos;
1375 PL_parser->oldoldbufptr = buf + oldoldbufptr_pos;
1376 PL_parser->linestart = buf + linestart_pos;
1377 if (PL_parser->last_uni)
1378 PL_parser->last_uni = buf + last_uni_pos;
1379 if (PL_parser->last_lop)
1380 PL_parser->last_lop = buf + last_lop_pos;
7f1c3e8c
FC
1381 if (PL_parser->preambling != NOLINE) {
1382 CopLINE_set(PL_curcop, PL_parser->preambling + 1);
1383 PL_parser->preambling = NOLINE;
1384 }
407f8cf2
KW
1385 if ( got_some_for_debugger
1386 && PERLDB_LINE_OR_SAVESRC
1387 && PL_curstash != PL_debstash)
1388 {
f0e67a1d
Z
1389 /* debugger active and we're not compiling the debugger code,
1390 * so store the line into the debugger's array of lines
1391 */
1392 update_debugger_info(NULL, buf+old_bufend_pos,
1393 new_bufend_pos-old_bufend_pos);
1394 }
1395 return got_some;
1396}
1397
1398/*
1399=for apidoc Amx|I32|lex_peek_unichar|U32 flags
1400
1401Looks ahead one (Unicode) character in the text currently being lexed.
1402Returns the codepoint (unsigned integer value) of the next character,
1403or -1 if lexing has reached the end of the input text. To consume the
1404peeked character, use L</lex_read_unichar>.
1405
1406If the next character is in (or extends into) the next chunk of input
1407text, the next chunk will be read in. Normally the current chunk will be
c5608a1f
KW
1408discarded at the same time, but if C<flags> has the C<LEX_KEEP_PREVIOUS>
1409bit set, then the current chunk will not be discarded.
f0e67a1d
Z
1410
1411If the input is being interpreted as UTF-8 and a UTF-8 encoding error
1412is encountered, an exception is generated.
1413
1414=cut
1415*/
1416
1417I32
1418Perl_lex_peek_unichar(pTHX_ U32 flags)
1419{
749123ff 1420 dVAR;
f0e67a1d
Z
1421 char *s, *bufend;
1422 if (flags & ~(LEX_KEEP_PREVIOUS))
1423 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_peek_unichar");
1424 s = PL_parser->bufptr;
1425 bufend = PL_parser->bufend;
1426 if (UTF) {
1427 U8 head;
1428 I32 unichar;
1429 STRLEN len, retlen;
1430 if (s == bufend) {
1431 if (!lex_next_chunk(flags))
1432 return -1;
1433 s = PL_parser->bufptr;
1434 bufend = PL_parser->bufend;
1435 }
1436 head = (U8)*s;
54d004e8 1437 if (UTF8_IS_INVARIANT(head))
f0e67a1d 1438 return head;
54d004e8
KW
1439 if (UTF8_IS_START(head)) {
1440 len = UTF8SKIP(&head);
f0e67a1d
Z
1441 while ((STRLEN)(bufend-s) < len) {
1442 if (!lex_next_chunk(flags | LEX_KEEP_PREVIOUS))
1443 break;
1444 s = PL_parser->bufptr;
1445 bufend = PL_parser->bufend;
1446 }
1447 }
c80e42f3 1448 unichar = utf8n_to_uvchr((U8*)s, bufend-s, &retlen, UTF8_CHECK_ONLY);
f0e67a1d 1449 if (retlen == (STRLEN)-1) {
75219bac
KW
1450 _force_out_malformed_utf8_message((U8 *) s,
1451 (U8 *) bufend,
1452 0,
1453 1 /* 1 means die */ );
1454 NOT_REACHED; /* NOTREACHED */
f0e67a1d
Z
1455 }
1456 return unichar;
1457 } else {
1458 if (s == bufend) {
1459 if (!lex_next_chunk(flags))
1460 return -1;
1461 s = PL_parser->bufptr;
1462 }
1463 return (U8)*s;
1464 }
1465}
1466
1467/*
1468=for apidoc Amx|I32|lex_read_unichar|U32 flags
1469
1470Reads the next (Unicode) character in the text currently being lexed.
1471Returns the codepoint (unsigned integer value) of the character read,
1472and moves L</PL_parser-E<gt>bufptr> past the character, or returns -1
1473if lexing has reached the end of the input text. To non-destructively
1474examine the next character, use L</lex_peek_unichar> instead.
1475
1476If the next character is in (or extends into) the next chunk of input
1477text, the next chunk will be read in. Normally the current chunk will be
c5608a1f
KW
1478discarded at the same time, but if C<flags> has the C<LEX_KEEP_PREVIOUS>
1479bit set, then the current chunk will not be discarded.
f0e67a1d
Z
1480
1481If the input is being interpreted as UTF-8 and a UTF-8 encoding error
1482is encountered, an exception is generated.
1483
1484=cut
1485*/
1486
1487I32
1488Perl_lex_read_unichar(pTHX_ U32 flags)
1489{
1490 I32 c;
1491 if (flags & ~(LEX_KEEP_PREVIOUS))
1492 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_read_unichar");
1493 c = lex_peek_unichar(flags);
1494 if (c != -1) {
1495 if (c == '\n')
83944c01 1496 COPLINE_INC_WITH_HERELINES;
d9018cbe
EB
1497 if (UTF)
1498 PL_parser->bufptr += UTF8SKIP(PL_parser->bufptr);
1499 else
1500 ++(PL_parser->bufptr);
f0e67a1d
Z
1501 }
1502 return c;
1503}
1504
1505/*
1506=for apidoc Amx|void|lex_read_space|U32 flags
1507
1508Reads optional spaces, in Perl style, in the text currently being
1509lexed. The spaces may include ordinary whitespace characters and
1510Perl-style comments. C<#line> directives are processed if encountered.
1511L</PL_parser-E<gt>bufptr> is moved past the spaces, so that it points
1512at a non-space character (or the end of the input text).
1513
1514If spaces extend into the next chunk of input text, the next chunk will
1515be read in. Normally the current chunk will be discarded at the same
c5608a1f 1516time, but if C<flags> has the C<LEX_KEEP_PREVIOUS> bit set, then the current
f0e67a1d
Z
1517chunk will not be discarded.
1518
1519=cut
1520*/
1521
21791330 1522#define LEX_NO_INCLINE 0x40000000
f0998909
Z
1523#define LEX_NO_NEXT_CHUNK 0x80000000
1524
f0e67a1d
Z
1525void
1526Perl_lex_read_space(pTHX_ U32 flags)
1527{
1528 char *s, *bufend;
21791330 1529 const bool can_incline = !(flags & LEX_NO_INCLINE);
f0e67a1d 1530 bool need_incline = 0;
21791330 1531 if (flags & ~(LEX_KEEP_PREVIOUS|LEX_NO_NEXT_CHUNK|LEX_NO_INCLINE))
f0e67a1d 1532 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_read_space");
f0e67a1d
Z
1533 s = PL_parser->bufptr;
1534 bufend = PL_parser->bufend;
1535 while (1) {
1536 char c = *s;
1537 if (c == '#') {
1538 do {
1539 c = *++s;
1540 } while (!(c == '\n' || (c == 0 && s == bufend)));
1541 } else if (c == '\n') {
1542 s++;
21791330
FC
1543 if (can_incline) {
1544 PL_parser->linestart = s;
1545 if (s == bufend)
1546 need_incline = 1;
1547 else
1548 incline(s);
1549 }
f0e67a1d
Z
1550 } else if (isSPACE(c)) {
1551 s++;
1552 } else if (c == 0 && s == bufend) {
1553 bool got_more;
65c68e17 1554 line_t l;
f0998909
Z
1555 if (flags & LEX_NO_NEXT_CHUNK)
1556 break;
f0e67a1d 1557 PL_parser->bufptr = s;
65c68e17 1558 l = CopLINE(PL_curcop);
851b527a 1559 CopLINE(PL_curcop) += PL_parser->herelines + 1;
f0e67a1d 1560 got_more = lex_next_chunk(flags);
65c68e17 1561 CopLINE_set(PL_curcop, l);
f0e67a1d
Z
1562 s = PL_parser->bufptr;
1563 bufend = PL_parser->bufend;
1564 if (!got_more)
1565 break;
21791330 1566 if (can_incline && need_incline && PL_parser->rsfp) {
f0e67a1d
Z
1567 incline(s);
1568 need_incline = 0;
1569 }
3c47da3c
FC
1570 } else if (!c) {
1571 s++;
f0e67a1d
Z
1572 } else {
1573 break;
1574 }
1575 }
f0e67a1d
Z
1576 PL_parser->bufptr = s;
1577}
1578
1579/*
fe788d6b
PM
1580
1581=for apidoc EXMp|bool|validate_proto|SV *name|SV *proto|bool warn
1582
1583This function performs syntax checking on a prototype, C<proto>.
1584If C<warn> is true, any illegal characters or mismatched brackets
1585will trigger illegalproto warnings, declaring that they were
1586detected in the prototype for C<name>.
1587
1588The return value is C<true> if this is a valid prototype, and
1589C<false> if it is not, regardless of whether C<warn> was C<true> or
1590C<false>.
1591
1592Note that C<NULL> is a valid C<proto> and will always return C<true>.
1593
1594=cut
1595
1596 */
1597
1598bool
1599Perl_validate_proto(pTHX_ SV *name, SV *proto, bool warn)
1600{
1601 STRLEN len, origlen;
11327fa1 1602 char *p;
fe788d6b
PM
1603 bool bad_proto = FALSE;
1604 bool in_brackets = FALSE;
1605 bool after_slash = FALSE;
1606 char greedy_proto = ' ';
1607 bool proto_after_greedy_proto = FALSE;
1608 bool must_be_last = FALSE;
1609 bool underscore = FALSE;
f791a21a 1610 bool bad_proto_after_underscore = FALSE;
fe788d6b
PM
1611
1612 PERL_ARGS_ASSERT_VALIDATE_PROTO;
1613
1614 if (!proto)
1615 return TRUE;
1616
11327fa1 1617 p = SvPV(proto, len);
fe788d6b
PM
1618 origlen = len;
1619 for (; len--; p++) {
1620 if (!isSPACE(*p)) {
1621 if (must_be_last)
1622 proto_after_greedy_proto = TRUE;
f791a21a
PM
1623 if (underscore) {
1624 if (!strchr(";@%", *p))
1625 bad_proto_after_underscore = TRUE;
1626 underscore = FALSE;
1627 }
fe788d6b
PM
1628 if (!strchr("$@%*;[]&\\_+", *p) || *p == '\0') {
1629 bad_proto = TRUE;
1630 }
1631 else {
fe788d6b
PM
1632 if (*p == '[')
1633 in_brackets = TRUE;
1634 else if (*p == ']')
1635 in_brackets = FALSE;
407f8cf2
KW
1636 else if ((*p == '@' || *p == '%')
1637 && !after_slash
1638 && !in_brackets )
1639 {
fe788d6b
PM
1640 must_be_last = TRUE;
1641 greedy_proto = *p;
1642 }
1643 else if (*p == '_')
f791a21a 1644 underscore = TRUE;
fe788d6b
PM
1645 }
1646 if (*p == '\\')
1647 after_slash = TRUE;
1648 else
1649 after_slash = FALSE;
1650 }
1651 }
1652
1653 if (warn) {
b54d603d 1654 SV *tmpsv = newSVpvs_flags("", SVs_TEMP);
fe788d6b 1655 p -= origlen;
b54d603d
PM
1656 p = SvUTF8(proto)
1657 ? sv_uni_display(tmpsv, newSVpvn_flags(p, origlen, SVs_TEMP | SVf_UTF8),
1658 origlen, UNI_DISPLAY_ISPRINT)
1659 : pv_pretty(tmpsv, p, origlen, 60, NULL, NULL, PERL_PV_ESCAPE_NONASCII);
1660
fe788d6b
PM
1661 if (proto_after_greedy_proto)
1662 Perl_warner(aTHX_ packWARN(WARN_ILLEGALPROTO),
147e3846 1663 "Prototype after '%c' for %" SVf " : %s",
fe788d6b 1664 greedy_proto, SVfARG(name), p);
50278ed0
PM
1665 if (in_brackets)
1666 Perl_warner(aTHX_ packWARN(WARN_ILLEGALPROTO),
147e3846 1667 "Missing ']' in prototype for %" SVf " : %s",
50278ed0 1668 SVfARG(name), p);
b54d603d 1669 if (bad_proto)
fe788d6b 1670 Perl_warner(aTHX_ packWARN(WARN_ILLEGALPROTO),
147e3846 1671 "Illegal character in prototype for %" SVf " : %s",
f791a21a
PM
1672 SVfARG(name), p);
1673 if (bad_proto_after_underscore)
1674 Perl_warner(aTHX_ packWARN(WARN_ILLEGALPROTO),
147e3846 1675 "Illegal character after '_' in prototype for %" SVf " : %s",
f791a21a 1676 SVfARG(name), p);
fe788d6b
PM
1677 }
1678
1679 return (! (proto_after_greedy_proto || bad_proto) );
1680}
1681
1682/*
ffb4593c
NT
1683 * S_incline
1684 * This subroutine has nothing to do with tilting, whether at windmills
1685 * or pinball tables. Its name is short for "increment line". It
57843af0 1686 * increments the current line number in CopLINE(PL_curcop) and checks
ffb4593c 1687 * to see whether the line starts with a comment of the form
9cbb5ea2
GS
1688 * # line 500 "foo.pm"
1689 * If so, it sets the current line number and file to the values in the comment.
ffb4593c
NT
1690 */
1691
76e3520e 1692STATIC void
d9095cec 1693S_incline(pTHX_ const char *s)
463ee0b2 1694{
d9095cec
NC
1695 const char *t;
1696 const char *n;
1697 const char *e;
8818d409 1698 line_t line_num;
22ff3130 1699 UV uv;
463ee0b2 1700
7918f24d
NC
1701 PERL_ARGS_ASSERT_INCLINE;
1702
83944c01 1703 COPLINE_INC_WITH_HERELINES;
451f421f
FC
1704 if (!PL_rsfp && !PL_parser->filtered && PL_lex_state == LEX_NORMAL
1705 && s+1 == PL_bufend && *s == ';') {
1706 /* fake newline in string eval */
1707 CopLINE_dec(PL_curcop);
1708 return;
1709 }
463ee0b2
LW
1710 if (*s++ != '#')
1711 return;
d4c19fe8
AL
1712 while (SPACE_OR_TAB(*s))
1713 s++;
847cc851 1714 if (strEQs(s, "line"))
73659bf1
GS
1715 s += 4;
1716 else
1717 return;
084592ab 1718 if (SPACE_OR_TAB(*s))
73659bf1 1719 s++;
4e553d73 1720 else
73659bf1 1721 return;
d4c19fe8
AL
1722 while (SPACE_OR_TAB(*s))
1723 s++;
463ee0b2
LW
1724 if (!isDIGIT(*s))
1725 return;
d4c19fe8 1726
463ee0b2
LW
1727 n = s;
1728 while (isDIGIT(*s))
1729 s++;
07714eb4 1730 if (!SPACE_OR_TAB(*s) && *s != '\r' && *s != '\n' && *s != '\0')
26b6dc3f 1731 return;
bf4acbe4 1732 while (SPACE_OR_TAB(*s))
463ee0b2 1733 s++;
73659bf1 1734 if (*s == '"' && (t = strchr(s+1, '"'))) {
463ee0b2 1735 s++;
73659bf1
GS
1736 e = t + 1;
1737 }
463ee0b2 1738 else {
c35e046a 1739 t = s;
1bb1a3d6 1740 while (*t && !isSPACE(*t))
c35e046a 1741 t++;
73659bf1 1742 e = t;
463ee0b2 1743 }
bf4acbe4 1744 while (SPACE_OR_TAB(*e) || *e == '\r' || *e == '\f')
73659bf1
GS
1745 e++;
1746 if (*e != '\n' && *e != '\0')
1747 return; /* false alarm */
1748
22ff3130
HS
1749 if (!grok_atoUV(n, &uv, &e))
1750 return;
1751 line_num = ((line_t)uv) - 1;
8818d409 1752
f4dd75d9 1753 if (t - s > 0) {
d9095cec 1754 const STRLEN len = t - s;
3df32bda 1755
d36ee5be 1756 if (!PL_rsfp && !PL_parser->filtered) {
e66cf94c
RGS
1757 /* must copy *{"::_<(eval N)[oldfilename:L]"}
1758 * to *{"::_<newfilename"} */
44867030
NC
1759 /* However, the long form of evals is only turned on by the
1760 debugger - usually they're "(eval %lu)" */
d36ee5be
FC
1761 GV * const cfgv = CopFILEGV(PL_curcop);
1762 if (cfgv) {
38bd7ad8
FC
1763 char smallbuf[128];
1764 STRLEN tmplen2 = len;
44867030 1765 char *tmpbuf2;
449dd039 1766 GV *gv2;
44867030
NC
1767
1768 if (tmplen2 + 2 <= sizeof smallbuf)
1769 tmpbuf2 = smallbuf;
1770 else
1771 Newx(tmpbuf2, tmplen2 + 2, char);
1772
38bd7ad8
FC
1773 tmpbuf2[0] = '_';
1774 tmpbuf2[1] = '<';
44867030
NC
1775
1776 memcpy(tmpbuf2 + 2, s, tmplen2);
1777 tmplen2 += 2;
1778
8a5ee598 1779 gv2 = *(GV**)hv_fetch(PL_defstash, tmpbuf2, tmplen2, TRUE);
e5527e4b 1780 if (!isGV(gv2)) {
8a5ee598 1781 gv_init(gv2, PL_defstash, tmpbuf2, tmplen2, FALSE);
e5527e4b
RGS
1782 /* adjust ${"::_<newfilename"} to store the new file name */
1783 GvSV(gv2) = newSVpvn(tmpbuf2 + 2, tmplen2 - 2);
8818d409
FC
1784 /* The line number may differ. If that is the case,
1785 alias the saved lines that are in the array.
1786 Otherwise alias the whole array. */
1787 if (CopLINE(PL_curcop) == line_num) {
38bd7ad8
FC
1788 GvHV(gv2) = MUTABLE_HV(SvREFCNT_inc(GvHV(cfgv)));
1789 GvAV(gv2) = MUTABLE_AV(SvREFCNT_inc(GvAV(cfgv)));
8818d409 1790 }
38bd7ad8
FC
1791 else if (GvAV(cfgv)) {
1792 AV * const av = GvAV(cfgv);
8818d409
FC
1793 const I32 start = CopLINE(PL_curcop)+1;
1794 I32 items = AvFILLp(av) - start;
1795 if (items > 0) {
1796 AV * const av2 = GvAVn(gv2);
1797 SV **svp = AvARRAY(av) + start;
1798 I32 l = (I32)line_num+1;
1799 while (items--)
1800 av_store(av2, l++, SvREFCNT_inc(*svp++));
1801 }
1802 }
e5527e4b 1803 }
44867030
NC
1804
1805 if (tmpbuf2 != smallbuf) Safefree(tmpbuf2);
d36ee5be 1806 }
e66cf94c 1807 }
05ec9bb3 1808 CopFILE_free(PL_curcop);
449dd039 1809 CopFILE_setn(PL_curcop, s, len);
f4dd75d9 1810 }
8818d409 1811 CopLINE_set(PL_curcop, line_num);
463ee0b2
LW
1812}
1813
80a702cd 1814STATIC void
15f169a1 1815S_update_debugger_info(pTHX_ SV *orig_sv, const char *const buf, STRLEN len)
80a702cd
RGS
1816{
1817 AV *av = CopFILEAVx(PL_curcop);
1818 if (av) {
7f1c3e8c
FC
1819 SV * sv;
1820 if (PL_parser->preambling == NOLINE) sv = newSV_type(SVt_PVMG);
1821 else {
1822 sv = *av_fetch(av, 0, 1);
1823 SvUPGRADE(sv, SVt_PVMG);
1824 }
847cc851 1825 if (!SvPOK(sv)) SvPVCLEAR(sv);
5fa550fb 1826 if (orig_sv)
7f1c3e8c 1827 sv_catsv(sv, orig_sv);
5fa550fb 1828 else
7f1c3e8c
FC
1829 sv_catpvn(sv, buf, len);
1830 if (!SvIOK(sv)) {
1831 (void)SvIOK_on(sv);
1832 SvIV_set(sv, 0);
1833 }
1834 if (PL_parser->preambling == NOLINE)
1835 av_store(av, CopLINE(PL_curcop), sv);
80a702cd
RGS
1836 }
1837}
1838
ffb4593c 1839/*
8c6b0c7d 1840 * skipspace
ffb4593c
NT
1841 * Called to gobble the appropriate amount and type of whitespace.
1842 * Skips comments as well.
71fff7cb 1843 * Returns the next character after the whitespace that is skipped.
8c6b0c7d
FC
1844 *
1845 * peekspace
1846 * Same thing, but look ahead without incrementing line numbers or
1847 * adjusting PL_linestart.
ffb4593c
NT
1848 */
1849
3218e223 1850#define skipspace(s) skipspace_flags(s, 0)
8c6b0c7d 1851#define peekspace(s) skipspace_flags(s, LEX_NO_INCLINE)
3218e223 1852
76e3520e 1853STATIC char *
21791330 1854S_skipspace_flags(pTHX_ char *s, U32 flags)
a687059c 1855{
21791330 1856 PERL_ARGS_ASSERT_SKIPSPACE_FLAGS;
3280af22 1857 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
3c47da3c 1858 while (s < PL_bufend && (SPACE_OR_TAB(*s) || !*s))
463ee0b2 1859 s++;
f0e67a1d
Z
1860 } else {
1861 STRLEN bufptr_pos = PL_bufptr - SvPVX(PL_linestr);
1862 PL_bufptr = s;
21791330 1863 lex_read_space(flags | LEX_KEEP_PREVIOUS |
d27f4b91 1864 (PL_lex_inwhat || PL_lex_state == LEX_FORMLINE ?
f0998909 1865 LEX_NO_NEXT_CHUNK : 0));
3280af22 1866 s = PL_bufptr;
f0e67a1d
Z
1867 PL_bufptr = SvPVX(PL_linestr) + bufptr_pos;
1868 if (PL_linestart > PL_bufptr)
1869 PL_bufptr = PL_linestart;
1870 return s;
463ee0b2 1871 }
5db06880 1872 return s;
a687059c 1873}
378cc40b 1874
ffb4593c
NT
1875/*
1876 * S_check_uni
1877 * Check the unary operators to ensure there's no ambiguity in how they're
1878 * used. An ambiguous piece of code would be:
1879 * rand + 5
1880 * This doesn't mean rand() + 5. Because rand() is a unary operator,
1881 * the +5 is its argument.
1882 */
1883
76e3520e 1884STATIC void
cea2e8a9 1885S_check_uni(pTHX)
ba106d47 1886{
d4c19fe8
AL
1887 const char *s;
1888 const char *t;
2f3197b3 1889
3280af22 1890 if (PL_oldoldbufptr != PL_last_uni)
2f3197b3 1891 return;
3280af22
NIS
1892 while (isSPACE(*PL_last_uni))
1893 PL_last_uni++;
c35e046a 1894 s = PL_last_uni;
fac0f7a3 1895 while (isWORDCHAR_lazy_if_safe(s, PL_bufend, UTF) || *s == '-')
8ce2ba82 1896 s += UTF ? UTF8SKIP(s) : 1;
3280af22 1897 if ((t = strchr(s, '(')) && t < PL_bufptr)
a0d0e21e 1898 return;
6136c704 1899
9b387841 1900 Perl_ck_warner_d(aTHX_ packWARN(WARN_AMBIGUOUS),
147e3846 1901 "Warning: Use of \"%" UTF8f "\" without parentheses is ambiguous",
b59c097b 1902 UTF8fARG(UTF, (int)(s - PL_last_uni), PL_last_uni));
2f3197b3
LW
1903}
1904
ffb4593c
NT
1905/*
1906 * LOP : macro to build a list operator. Its behaviour has been replaced
1907 * with a subroutine, S_lop() for which LOP is just another name.
1908 */
1909
a0d0e21e
LW
1910#define LOP(f,x) return lop(f,x,s)
1911
ffb4593c
NT
1912/*
1913 * S_lop
1914 * Build a list operator (or something that might be one). The rules:
41e8cbf4
FC
1915 * - if we have a next token, then it's a list operator (no parens) for
1916 * which the next token has already been parsed; e.g.,
1917 * sort foo @args
1918 * sort foo (@args)
ffb4593c
NT
1919 * - if the next thing is an opening paren, then it's a function
1920 * - else it's a list operator
1921 */
1922
76e3520e 1923STATIC I32
11288bb3 1924S_lop(pTHX_ I32 f, U8 x, char *s)
ffed7fef 1925{
7918f24d
NC
1926 PERL_ARGS_ASSERT_LOP;
1927
6154021b 1928 pl_yylval.ival = f;
35c8bce7 1929 CLINE;
3280af22
NIS
1930 PL_bufptr = s;
1931 PL_last_lop = PL_oldbufptr;
eb160463 1932 PL_last_lop_op = (OPCODE)f;
3280af22 1933 if (PL_nexttoke)
78cdf107 1934 goto lstop;
19f1898a 1935 PL_expect = x;
79072805 1936 if (*s == '(')
bbf60fe6 1937 return REPORT(FUNC);
294a536f 1938 s = skipspace(s);
79072805 1939 if (*s == '(')
bbf60fe6 1940 return REPORT(FUNC);
78cdf107
Z
1941 else {
1942 lstop:
1943 if (!PL_lex_allbrackets && PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC)
1944 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC;
bbf60fe6 1945 return REPORT(LSTOP);
78cdf107 1946 }
79072805
LW
1947}
1948
ffb4593c
NT
1949/*
1950 * S_force_next
9cbb5ea2 1951 * When the lexer realizes it knows the next token (for instance,
ffb4593c 1952 * it is reordering tokens for the parser) then it can call S_force_next
9cbb5ea2 1953 * to know what token to return the next time the lexer is called. Caller
b5bbe64a
JH
1954 * will need to set PL_nextval[] and possibly PL_expect to ensure
1955 * the lexer handles the token correctly.
ffb4593c
NT
1956 */
1957
4e553d73 1958STATIC void
cea2e8a9 1959S_force_next(pTHX_ I32 type)
79072805 1960{
704d4215
GG
1961#ifdef DEBUGGING
1962 if (DEBUG_T_TEST) {
1963 PerlIO_printf(Perl_debug_log, "### forced token:\n");
f05d7009 1964 tokereport(type, &NEXTVAL_NEXTTOKE);
704d4215
GG
1965 }
1966#endif
1f7c3e7c 1967 assert(PL_nexttoke < C_ARRAY_LENGTH(PL_nexttype));
3280af22
NIS
1968 PL_nexttype[PL_nexttoke] = type;
1969 PL_nexttoke++;
79072805
LW
1970}
1971
89f35911
FC
1972/*
1973 * S_postderef
1974 *
1975 * This subroutine handles postfix deref syntax after the arrow has already
1976 * been emitted. @* $* etc. are emitted as two separate token right here.
1977 * @[ @{ %[ %{ *{ are emitted also as two tokens, but this function emits
1978 * only the first, leaving yylex to find the next.
89f35911
FC
1979 */
1980
1981static int
ff25e5db 1982S_postderef(pTHX_ int const funny, char const next)
89f35911 1983{
ff25e5db 1984 assert(funny == DOLSHARP || strchr("$@%&*", funny));
89f35911
FC
1985 if (next == '*') {
1986 PL_expect = XOPERATOR;
cc624add 1987 if (PL_lex_state == LEX_INTERPNORMAL && !PL_lex_brackets) {
ff25e5db 1988 assert('@' == funny || '$' == funny || DOLSHARP == funny);
cc624add 1989 PL_lex_state = LEX_INTERPEND;
c3492809
FC
1990 if ('@' == funny)
1991 force_next(POSTJOIN);
cc624add 1992 }
89f35911
FC
1993 force_next(next);
1994 PL_bufptr+=2;
1995 }
1996 else {
760ca746
FC
1997 if ('@' == funny && PL_lex_state == LEX_INTERPNORMAL
1998 && !PL_lex_brackets)
1999 PL_lex_dojoin = 2;
89f35911
FC
2000 PL_expect = XOPERATOR;
2001 PL_bufptr++;
2002 }
2003 return funny;
2004}
2005
28ac2b49
Z
2006void
2007Perl_yyunlex(pTHX)
2008{
a7aaec61
Z
2009 int yyc = PL_parser->yychar;
2010 if (yyc != YYEMPTY) {
2011 if (yyc) {
a7aaec61
Z
2012 NEXTVAL_NEXTTOKE = PL_parser->yylval;
2013 if (yyc == '{'/*}*/ || yyc == HASHBRACK || yyc == '['/*]*/) {
78cdf107 2014 PL_lex_allbrackets--;
a7aaec61 2015 PL_lex_brackets--;
78cdf107
Z
2016 yyc |= (3<<24) | (PL_lex_brackstack[PL_lex_brackets] << 16);
2017 } else if (yyc == '('/*)*/) {
2018 PL_lex_allbrackets--;
2019 yyc |= (2<<24);
a7aaec61
Z
2020 }
2021 force_next(yyc);
2022 }
28ac2b49
Z
2023 PL_parser->yychar = YYEMPTY;
2024 }
2025}
2026
d0a148a6 2027STATIC SV *
15f169a1 2028S_newSV_maybe_utf8(pTHX_ const char *const start, STRLEN len)
d0a148a6 2029{
740cce10 2030 SV * const sv = newSVpvn_utf8(start, len,
c5f058df
KW
2031 !IN_BYTES
2032 && UTF
2033 && !is_utf8_invariant_string((const U8*)start, len)
2034 && is_utf8_string((const U8*)start, len));
d0a148a6
NC
2035 return sv;
2036}
2037
ffb4593c
NT
2038/*
2039 * S_force_word
2040 * When the lexer knows the next thing is a word (for instance, it has
2041 * just seen -> and it knows that the next char is a word char, then
02b34bbe
DM
2042 * it calls S_force_word to stick the next word into the PL_nexttoke/val
2043 * lookahead.
ffb4593c
NT
2044 *
2045 * Arguments:
b1b65b59 2046 * char *start : buffer position (must be within PL_linestr)
185c2e96
DM
2047 * int token : PL_next* will be this type of bare word
2048 * (e.g., METHOD,BAREWORD)
ffb4593c
NT
2049 * int check_keyword : if true, Perl checks to make sure the word isn't
2050 * a keyword (do this if the word is a label, e.g. goto FOO)
2051 * int allow_pack : if true, : characters will also be allowed (require,
2052 * use, etc. do this)
ffb4593c
NT
2053 */
2054
76e3520e 2055STATIC char *
345b3785 2056S_force_word(pTHX_ char *start, int token, int check_keyword, int allow_pack)
79072805 2057{
eb578fdb 2058 char *s;
463ee0b2 2059 STRLEN len;
4e553d73 2060
7918f24d
NC
2061 PERL_ARGS_ASSERT_FORCE_WORD;
2062
294a536f 2063 start = skipspace(start);
463ee0b2 2064 s = start;
fac0f7a3 2065 if ( isIDFIRST_lazy_if_safe(s, PL_bufend, UTF)
e7127e21 2066 || (allow_pack && *s == ':' && s[1] == ':') )
a0d0e21e 2067 {
3280af22 2068 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, allow_pack, &len);
01b5ef50
FC
2069 if (check_keyword) {
2070 char *s2 = PL_tokenbuf;
487e470d 2071 STRLEN len2 = len;
847cc851 2072 if (allow_pack && len > 6 && strEQs(s2, "CORE::"))
487e470d
FC
2073 s2 += 6, len2 -= 6;
2074 if (keyword(s2, len2, 0))
463ee0b2 2075 return start;
01b5ef50 2076 }
463ee0b2 2077 if (token == METHOD) {
294a536f 2078 s = skipspace(s);
463ee0b2 2079 if (*s == '(')
3280af22 2080 PL_expect = XTERM;
463ee0b2 2081 else {
3280af22 2082 PL_expect = XOPERATOR;
463ee0b2 2083 }
79072805 2084 }
9ded7720 2085 NEXTVAL_NEXTTOKE.opval
275103cd 2086 = newSVOP(OP_CONST,0,
d0a148a6 2087 S_newSV_maybe_utf8(aTHX_ PL_tokenbuf, len));
9ded7720 2088 NEXTVAL_NEXTTOKE.opval->op_private |= OPpCONST_BARE;
79072805
LW
2089 force_next(token);
2090 }
2091 return s;
2092}
2093
ffb4593c
NT
2094/*
2095 * S_force_ident
9cbb5ea2 2096 * Called when the lexer wants $foo *foo &foo etc, but the program
ffb4593c
NT
2097 * text only contains the "foo" portion. The first argument is a pointer
2098 * to the "foo", and the second argument is the type symbol to prefix.
185c2e96 2099 * Forces the next token to be a "BAREWORD".
9cbb5ea2 2100 * Creates the symbol if it didn't already exist (via gv_fetchpv()).
ffb4593c
NT
2101 */
2102
76e3520e 2103STATIC void
5aaab254 2104S_force_ident(pTHX_ const char *s, int kind)
79072805 2105{
7918f24d
NC
2106 PERL_ARGS_ASSERT_FORCE_IDENT;
2107
c9b48522
DD
2108 if (s[0]) {
2109 const STRLEN len = s[1] ? strlen(s) : 1; /* s = "\"" see yylex */
275103cd 2110 OP* const o = newSVOP(OP_CONST, 0, newSVpvn_flags(s, len,
728847b1 2111 UTF ? SVf_UTF8 : 0));
9ded7720 2112 NEXTVAL_NEXTTOKE.opval = o;
185c2e96 2113 force_next(BAREWORD);
748a9306 2114 if (kind) {
11343788 2115 o->op_private = OPpCONST_ENTERED;
55497cff
PP
2116 /* XXX see note in pp_entereval() for why we forgo typo
2117 warnings if the symbol must be introduced in an eval.
2118 GSAR 96-10-12 */
90e5519e 2119 gv_fetchpvn_flags(s, len,
4bff32c5 2120 (PL_in_eval ? GV_ADDMULTI
728847b1 2121 : GV_ADD) | ( UTF ? SVf_UTF8 : 0 ),
90e5519e
NC
2122 kind == '$' ? SVt_PV :
2123 kind == '@' ? SVt_PVAV :
2124 kind == '%' ? SVt_PVHV :
a0d0e21e 2125 SVt_PVGV
90e5519e 2126 );
748a9306 2127 }
79072805
LW
2128 }
2129}
2130
3f33d153
FC
2131static void
2132S_force_ident_maybe_lex(pTHX_ char pit)
2133{
3f33d153
FC
2134 NEXTVAL_NEXTTOKE.ival = pit;
2135 force_next('p');
2136}
2137
1571675a
GS
2138NV
2139Perl_str_to_version(pTHX_ SV *sv)
2140{
2141 NV retval = 0.0;
2142 NV nshift = 1.0;
2143 STRLEN len;
cfd0369c 2144 const char *start = SvPV_const(sv,len);
9d4ba2ae 2145 const char * const end = start + len;
504618e9 2146 const bool utf = SvUTF8(sv) ? TRUE : FALSE;
7918f24d
NC
2147
2148 PERL_ARGS_ASSERT_STR_TO_VERSION;
2149
1571675a 2150 while (start < end) {
ba210ebe 2151 STRLEN skip;
1571675a
GS
2152 UV n;
2153 if (utf)
9041c2e3 2154 n = utf8n_to_uvchr((U8*)start, len, &skip, 0);
1571675a
GS
2155 else {
2156 n = *(U8*)start;
2157 skip = 1;
2158 }
2159 retval += ((NV)n)/nshift;
2160 start += skip;
2161 nshift *= 1000;
2162 }
2163 return retval;
2164}
2165
4e553d73 2166/*
ffb4593c
NT
2167 * S_force_version
2168 * Forces the next token to be a version number.
e759cc13
RGS
2169 * If the next token appears to be an invalid version number, (e.g. "v2b"),
2170 * and if "guessing" is TRUE, then no new token is created (and the caller
2171 * must use an alternative parsing method).
ffb4593c
NT
2172 */
2173
76e3520e 2174STATIC char *
e759cc13 2175S_force_version(pTHX_ char *s, int guessing)
89bfa8cd 2176{
5f66b61c 2177 OP *version = NULL;
44dcb63b 2178 char *d;
89bfa8cd 2179
7918f24d
NC
2180 PERL_ARGS_ASSERT_FORCE_VERSION;
2181
294a536f 2182 s = skipspace(s);
89bfa8cd 2183
44dcb63b 2184 d = s;
dd629d5b 2185 if (*d == 'v')
44dcb63b 2186 d++;
44dcb63b 2187 if (isDIGIT(*d)) {
e759cc13
RGS
2188 while (isDIGIT(*d) || *d == '_' || *d == '.')
2189 d++;
4e4da3ac 2190 if (*d == ';' || isSPACE(*d) || *d == '{' || *d == '}' || !*d) {
dd629d5b 2191 SV *ver;
6154021b
RGS
2192 s = scan_num(s, &pl_yylval);
2193 version = pl_yylval.opval;
dd629d5b
GS
2194 ver = cSVOPx(version)->op_sv;
2195 if (SvPOK(ver) && !SvNIOK(ver)) {
862a34c6 2196 SvUPGRADE(ver, SVt_PVNV);
9d6ce603 2197 SvNV_set(ver, str_to_version(ver));
1571675a 2198 SvNOK_on(ver); /* hint that it is a version */
44dcb63b 2199 }
89bfa8cd 2200 }
5db06880 2201 else if (guessing) {
e759cc13 2202 return s;
5db06880 2203 }
89bfa8cd
PP
2204 }
2205
2206 /* NOTE: The parser sees the package name and the VERSION swapped */
9ded7720 2207 NEXTVAL_NEXTTOKE.opval = version;
185c2e96 2208 force_next(BAREWORD);
89bfa8cd 2209
e759cc13 2210 return s;
89bfa8cd
PP
2211}
2212
ffb4593c 2213/*
91152fc1
DG
2214 * S_force_strict_version
2215 * Forces the next token to be a version number using strict syntax rules.
2216 */
2217
2218STATIC char *
2219S_force_strict_version(pTHX_ char *s)
2220{
91152fc1 2221 OP *version = NULL;
91152fc1
DG
2222 const char *errstr = NULL;
2223
2224 PERL_ARGS_ASSERT_FORCE_STRICT_VERSION;
2225
2226 while (isSPACE(*s)) /* leading whitespace */
2227 s++;
2228
2229 if (is_STRICT_VERSION(s,&errstr)) {
2230 SV *ver = newSV(0);
2231 s = (char *)scan_version(s, ver, 0);
2232 version = newSVOP(OP_CONST, 0, ver);
2233 }
407f8cf2
KW
2234 else if ((*s != ';' && *s != '{' && *s != '}' )
2235 && (s = skipspace(s), (*s != ';' && *s != '{' && *s != '}' )))
4e4da3ac 2236 {
91152fc1
DG
2237 PL_bufptr = s;
2238 if (errstr)
2239 yyerror(errstr); /* version required */
2240 return s;
2241 }
2242
91152fc1 2243 /* NOTE: The parser sees the package name and the VERSION swapped */
91152fc1 2244 NEXTVAL_NEXTTOKE.opval = version;
185c2e96 2245 force_next(BAREWORD);
91152fc1
DG
2246
2247 return s;
2248}
2249
2250/*
ffb4593c
NT
2251 * S_tokeq
2252 * Tokenize a quoted string passed in as an SV. It finds the next
2253 * chunk, up to end of string or a backslash. It may make a new
2254 * SV containing that chunk (if HINT_NEW_STRING is on). It also
2255 * turns \\ into \.
2256 */
2257
76e3520e 2258STATIC SV *
cea2e8a9 2259S_tokeq(pTHX_ SV *sv)
79072805 2260{
eb578fdb
KW
2261 char *s;
2262 char *send;
2263 char *d;
b3ac6de7 2264 SV *pv = sv;
79072805 2265
7918f24d
NC
2266 PERL_ARGS_ASSERT_TOKEQ;
2267
279b35ad
FC
2268 assert (SvPOK(sv));
2269 assert (SvLEN(sv));
2270 assert (!SvIsCOW(sv));
307ed071 2271 if (SvTYPE(sv) >= SVt_PVIV && SvIVX(sv) == -1) /* <<'heredoc' */
b3ac6de7 2272 goto finish;
279b35ad
FC
2273 s = SvPVX(sv);
2274 send = SvEND(sv);
dcb21ed6
NC
2275 /* This is relying on the SV being "well formed" with a trailing '\0' */
2276 while (s < send && !(*s == '\\' && s[1] == '\\'))
79072805
LW
2277 s++;
2278 if (s == send)
b3ac6de7 2279 goto finish;
79072805 2280 d = s;
be4731d2 2281 if ( PL_hints & HINT_NEW_STRING ) {
279b35ad
FC
2282 pv = newSVpvn_flags(SvPVX_const(pv), SvCUR(sv),
2283 SVs_TEMP | SvUTF8(sv));
be4731d2 2284 }
79072805
LW
2285 while (s < send) {
2286 if (*s == '\\') {
a0d0e21e 2287 if (s + 1 < send && (s[1] == '\\'))
79072805
LW
2288 s++; /* all that, just for this */
2289 }
2290 *d++ = *s++;
2291 }
2292 *d = '\0';
95a20fc0 2293 SvCUR_set(sv, d - SvPVX_const(sv));
b3ac6de7 2294 finish:
3280af22 2295 if ( PL_hints & HINT_NEW_STRING )
eb0d8d16 2296 return new_constant(NULL, 0, "q", sv, pv, "q", 1);
79072805
LW
2297 return sv;
2298}
2299
ffb4593c
NT
2300/*
2301 * Now come three functions related to double-quote context,
2302 * S_sublex_start, S_sublex_push, and S_sublex_done. They're used when
2303 * converting things like "\u\Lgnat" into ucfirst(lc("gnat")). They
2304 * interact with PL_lex_state, and create fake ( ... ) argument lists
2305 * to handle functions and concatenation.
ecd24171
DM
2306 * For example,
2307 * "foo\lbar"
2308 * is tokenised as
2309 * stringify ( const[foo] concat lcfirst ( const[bar] ) )
ffb4593c
NT
2310 */
2311
2312/*
2313 * S_sublex_start
6154021b 2314 * Assumes that pl_yylval.ival is the op we're creating (e.g. OP_LCFIRST).
ffb4593c
NT
2315 *
2316 * Pattern matching will set PL_lex_op to the pattern-matching op to
6154021b 2317 * make (we return THING if pl_yylval.ival is OP_NULL, PMFUNC otherwise).
ffb4593c 2318 *
16d1d8bd 2319 * OP_CONST is easy--just make the new op and return.
ffb4593c
NT
2320 *
2321 * Everything else becomes a FUNC.
2322 *
16d1d8bd
LM
2323 * Sets PL_lex_state to LEX_INTERPPUSH unless ival was OP_NULL or we
2324 * had an OP_CONST. This just sets us up for a
ffb4593c
NT
2325 * call to S_sublex_push().
2326 */
2327
76e3520e 2328STATIC I32
cea2e8a9 2329S_sublex_start(pTHX)
79072805 2330{
eb578fdb 2331 const I32 op_type = pl_yylval.ival;
79072805
LW
2332
2333 if (op_type == OP_NULL) {
6154021b 2334 pl_yylval.opval = PL_lex_op;
5f66b61c 2335 PL_lex_op = NULL;
79072805
LW
2336 return THING;
2337 }
466112bb 2338 if (op_type == OP_CONST) {
67c71cbb
FC
2339 SV *sv = PL_lex_stuff;
2340 PL_lex_stuff = NULL;
2341 sv = tokeq(sv);
b3ac6de7
IZ
2342
2343 if (SvTYPE(sv) == SVt_PVIV) {
2344 /* Overloaded constants, nothing fancy: Convert to SVt_PV: */
2345 STRLEN len;
96a5add6 2346 const char * const p = SvPV_const(sv, len);
740cce10 2347 SV * const nsv = newSVpvn_flags(p, len, SvUTF8(sv));
b3ac6de7
IZ
2348 SvREFCNT_dec(sv);
2349 sv = nsv;
4e553d73 2350 }
275103cd 2351 pl_yylval.opval = newSVOP(op_type, 0, sv);
79072805
LW
2352 return THING;
2353 }
2354
7ef70b3d
FC
2355 PL_parser->lex_super_state = PL_lex_state;
2356 PL_parser->lex_sub_inwhat = (U16)op_type;
2357 PL_parser->lex_sub_op = PL_lex_op;
3280af22 2358 PL_lex_state = LEX_INTERPPUSH;
55497cff 2359
3280af22
NIS
2360 PL_expect = XTERM;
2361 if (PL_lex_op) {
6154021b 2362 pl_yylval.opval = PL_lex_op;
5f66b61c 2363 PL_lex_op = NULL;
55497cff
PP
2364 return PMFUNC;
2365 }
2366 else
2367 return FUNC;
2368}
2369
ffb4593c
NT
2370/*
2371 * S_sublex_push
2372 * Create a new scope to save the lexing state. The scope will be
2373 * ended in S_sublex_done. Returns a '(', starting the function arguments
2374 * to the uc, lc, etc. found before.
2375 * Sets PL_lex_state to LEX_INTERPCONCAT.
2376 */
2377
76e3520e 2378STATIC I32
cea2e8a9 2379S_sublex_push(pTHX)
55497cff 2380{
78a635de 2381 LEXSHARED *shared;
801d32ac 2382 const bool is_heredoc = PL_multi_close == '<';
f46d017c 2383 ENTER;
55497cff 2384
7ef70b3d 2385 PL_lex_state = PL_parser->lex_super_state;
cc624add 2386 SAVEI8(PL_lex_dojoin);
3280af22 2387 SAVEI32(PL_lex_brackets);
78cdf107 2388 SAVEI32(PL_lex_allbrackets);
b27dce25 2389 SAVEI32(PL_lex_formbrack);
78cdf107 2390 SAVEI8(PL_lex_fakeeof);
3280af22
NIS
2391 SAVEI32(PL_lex_casemods);
2392 SAVEI32(PL_lex_starts);
651b5b28 2393 SAVEI8(PL_lex_state);
7cc34111 2394 SAVESPTR(PL_lex_repl);
7766f137 2395 SAVEVPTR(PL_lex_inpat);
98246f1e 2396 SAVEI16(PL_lex_inwhat);
ffdb8b16 2397 if (is_heredoc)
b42366d4 2398 {
ffdb8b16 2399 SAVECOPLINE(PL_curcop);
b42366d4 2400 SAVEI32(PL_multi_end);
851b527a
FC
2401 SAVEI32(PL_parser->herelines);
2402 PL_parser->herelines = 0;
b42366d4 2403 }
2ca4363d 2404 SAVEIV(PL_multi_close);
3280af22 2405 SAVEPPTR(PL_bufptr);
8452ff4b 2406 SAVEPPTR(PL_bufend);
3280af22
NIS
2407 SAVEPPTR(PL_oldbufptr);
2408 SAVEPPTR(PL_oldoldbufptr);
207e3d1a
JH
2409 SAVEPPTR(PL_last_lop);
2410 SAVEPPTR(PL_last_uni);
3280af22
NIS
2411 SAVEPPTR(PL_linestart);
2412 SAVESPTR(PL_linestr);
8edd5f42
RGS
2413 SAVEGENERICPV(PL_lex_brackstack);
2414 SAVEGENERICPV(PL_lex_casestack);
78a635de 2415 SAVEGENERICPV(PL_parser->lex_shared);
3a54fd60 2416 SAVEBOOL(PL_parser->lex_re_reparsing);
ffdb8b16 2417 SAVEI32(PL_copline);
3280af22 2418
99bd9d90 2419 /* The here-doc parser needs to be able to peek into outer lexing
60f40a38
FC
2420 scopes to find the body of the here-doc. So we put PL_linestr and
2421 PL_bufptr into lex_shared, to ‘share’ those values.
99bd9d90 2422 */
60f40a38
FC
2423 PL_parser->lex_shared->ls_linestr = PL_linestr;
2424 PL_parser->lex_shared->ls_bufptr = PL_bufptr;
99bd9d90 2425
3280af22 2426 PL_linestr = PL_lex_stuff;
7ef70b3d 2427 PL_lex_repl = PL_parser->lex_sub_repl;
a0714e2c 2428 PL_lex_stuff = NULL;
7ef70b3d 2429 PL_parser->lex_sub_repl = NULL;
3280af22 2430
eabab8bc
FC
2431 /* Arrange for PL_lex_stuff to be freed on scope exit, in case it gets
2432 set for an inner quote-like operator and then an error causes scope-
2433 popping. We must not have a PL_lex_stuff value left dangling, as
2434 that breaks assumptions elsewhere. See bug #123617. */
2435 SAVEGENERICSV(PL_lex_stuff);
7ef70b3d 2436 SAVEGENERICSV(PL_parser->lex_sub_repl);
eabab8bc 2437
9cbb5ea2
GS
2438 PL_bufend = PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart
2439 = SvPVX(PL_linestr);
3280af22 2440 PL_bufend += SvCUR(PL_linestr);
bd61b366 2441 PL_last_lop = PL_last_uni = NULL;
3280af22 2442 SAVEFREESV(PL_linestr);
4dc843bc 2443 if (PL_lex_repl) SAVEFREESV(PL_lex_repl);
3280af22
NIS
2444
2445 PL_lex_dojoin = FALSE;
b27dce25 2446 PL_lex_brackets = PL_lex_formbrack = 0;
78cdf107
Z
2447 PL_lex_allbrackets = 0;
2448 PL_lex_fakeeof = LEX_FAKEEOF_NEVER;
a02a5408
JC
2449 Newx(PL_lex_brackstack, 120, char);
2450 Newx(PL_lex_casestack, 12, char);
3280af22
NIS
2451 PL_lex_casemods = 0;
2452 *PL_lex_casestack = '\0';
2453 PL_lex_starts = 0;
2454 PL_lex_state = LEX_INTERPCONCAT;
ffdb8b16 2455 if (is_heredoc)
6ddcf93b 2456 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
ffdb8b16 2457 PL_copline = NOLINE;
caae0700 2458
78a635de
FC
2459 Newxz(shared, 1, LEXSHARED);
2460 shared->ls_prev = PL_parser->lex_shared;
2461 PL_parser->lex_shared = shared;
3280af22 2462
7ef70b3d 2463 PL_lex_inwhat = PL_parser->lex_sub_inwhat;
bb16bae8 2464 if (PL_lex_inwhat == OP_TRANSR) PL_lex_inwhat = OP_TRANS;
3280af22 2465 if (PL_lex_inwhat == OP_MATCH || PL_lex_inwhat == OP_QR || PL_lex_inwhat == OP_SUBST)
7ef70b3d 2466 PL_lex_inpat = PL_parser->lex_sub_op;
79072805 2467 else
5f66b61c 2468 PL_lex_inpat = NULL;
79072805 2469
3a54fd60
DM
2470 PL_parser->lex_re_reparsing = cBOOL(PL_in_eval & EVAL_RE_REPARSING);
2471 PL_in_eval &= ~EVAL_RE_REPARSING;
2472
55497cff 2473 return '(';
79072805
LW
2474}
2475
ffb4593c
NT
2476/*
2477 * S_sublex_done
2478 * Restores lexer state after a S_sublex_push.
2479 */
2480
76e3520e 2481STATIC I32
cea2e8a9 2482S_sublex_done(pTHX)
79072805 2483{
3280af22 2484 if (!PL_lex_starts++) {
396482e1 2485 SV * const sv = newSVpvs("");
9aa983d2
JH
2486 if (SvUTF8(PL_linestr))
2487 SvUTF8_on(sv);
3280af22 2488 PL_expect = XOPERATOR;
275103cd 2489 pl_yylval.opval = newSVOP(OP_CONST, 0, sv);
79072805
LW
2490 return THING;
2491 }
2492
3280af22
NIS
2493 if (PL_lex_casemods) { /* oops, we've got some unbalanced parens */
2494 PL_lex_state = LEX_INTERPCASEMOD;
cea2e8a9 2495 return yylex();
79072805
LW
2496 }
2497
ffb4593c 2498 /* Is there a right-hand side to take care of? (s//RHS/ or tr//RHS/) */
bb16bae8 2499 assert(PL_lex_inwhat != OP_TRANSR);
5aa91856
FC
2500 if (PL_lex_repl) {
2501 assert (PL_lex_inwhat == OP_SUBST || PL_lex_inwhat == OP_TRANS);
3280af22
NIS
2502 PL_linestr = PL_lex_repl;
2503 PL_lex_inpat = 0;
2504 PL_bufend = PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart = SvPVX(PL_linestr);
2505 PL_bufend += SvCUR(PL_linestr);
bd61b366 2506 PL_last_lop = PL_last_uni = NULL;
3280af22
NIS
2507 PL_lex_dojoin = FALSE;
2508 PL_lex_brackets = 0;
78cdf107
Z
2509 PL_lex_allbrackets = 0;
2510 PL_lex_fakeeof = LEX_FAKEEOF_NEVER;
3280af22
NIS
2511 PL_lex_casemods = 0;
2512 *PL_lex_casestack = '\0';
2513 PL_lex_starts = 0;
25da4f38 2514 if (SvEVALED(PL_lex_repl)) {
3280af22
NIS
2515 PL_lex_state = LEX_INTERPNORMAL;
2516 PL_lex_starts++;
e9fa98b2
HS
2517 /* we don't clear PL_lex_repl here, so that we can check later
2518 whether this is an evalled subst; that means we rely on the
2519 logic to ensure sublex_done() is called again only via the
2520 branch (in yylex()) that clears PL_lex_repl, else we'll loop */
79072805 2521 }
e9fa98b2 2522 else {
3280af22 2523 PL_lex_state = LEX_INTERPCONCAT;
a0714e2c 2524 PL_lex_repl = NULL;
e9fa98b2 2525 }
ffdb8b16
FC
2526 if (SvTYPE(PL_linestr) >= SVt_PVNV) {
2527 CopLINE(PL_curcop) +=
9420b268 2528 ((XPVNV*)SvANY(PL_linestr))->xnv_u.xnv_lines
851b527a
FC
2529 + PL_parser->herelines;
2530 PL_parser->herelines = 0;
ffdb8b16 2531 }
9b6b7be8 2532 return '/';
ffed7fef
LW
2533 }
2534 else {
b42366d4 2535 const line_t l = CopLINE(PL_curcop);
f46d017c 2536 LEAVE;
b42366d4 2537 if (PL_multi_close == '<')
851b527a 2538 PL_parser->herelines += l - PL_multi_end;
3280af22
NIS
2539 PL_bufend = SvPVX(PL_linestr);
2540 PL_bufend += SvCUR(PL_linestr);
2541 PL_expect = XOPERATOR;
79072805 2542 return ')';
ffed7fef
LW
2543 }
2544}
2545
6f613c73
KW
2546PERL_STATIC_INLINE SV*
2547S_get_and_check_backslash_N_name(pTHX_ const char* s, const char* const e)
2548{
140b12ad
KW
2549 /* <s> points to first character of interior of \N{}, <e> to one beyond the
2550 * interior, hence to the "}". Finds what the name resolves to, returning
2551 * an SV* containing it; NULL if no valid one found */
2552
dd2b1b72 2553 SV* res = newSVpvn_flags(s, e - s, UTF ? SVf_UTF8 : 0);
6f613c73 2554
0c415a79
KW
2555 HV * table;
2556 SV **cvp;
2557 SV *cv;
2558 SV *rv;
2559 HV *stash;
2560 const U8* first_bad_char_loc;
2561 const char* backslash_ptr = s - 3; /* Points to the <\> of \N{... */
2562
6f613c73
KW
2563 PERL_ARGS_ASSERT_GET_AND_CHECK_BACKSLASH_N_NAME;
2564
d8d26cac
KW
2565 if (!SvCUR(res)) {
2566 Perl_ck_warner_d(aTHX_ packWARN(WARN_DEPRECATED),
2567 "Unknown charname '' is deprecated");
b7e6151c 2568 return res;
d8d26cac 2569 }
b7e6151c 2570
107160e2
KW
2571 if (UTF && ! is_utf8_string_loc((U8 *) backslash_ptr,
2572 e - backslash_ptr,
2573 &first_bad_char_loc))
2574 {
75219bac
KW
2575 _force_out_malformed_utf8_message(first_bad_char_loc,
2576 (U8 *) PL_parser->bufend,
2577 0,
2578 0 /* 0 means don't die */ );
b59c097b 2579 yyerror_pv(Perl_form(aTHX_
107160e2 2580 "Malformed UTF-8 character immediately after '%.*s'",
b59c097b
AV
2581 (int) (first_bad_char_loc - (U8 *) backslash_ptr), backslash_ptr),
2582 SVf_UTF8);
107160e2
KW
2583 return NULL;
2584 }
2585
2586 res = new_constant( NULL, 0, "charnames", res, NULL, backslash_ptr,
2587 /* include the <}> */
2588 e - backslash_ptr + 1);
6f613c73 2589 if (! SvPOK(res)) {
b6407c49 2590 SvREFCNT_dec_NN(res);
6f613c73
KW
2591 return NULL;
2592 }
2593
0c415a79
KW
2594 /* See if the charnames handler is the Perl core's, and if so, we can skip
2595 * the validation needed for a user-supplied one, as Perl's does its own
2596 * validation. */
2597 table = GvHV(PL_hintgv); /* ^H */
2598 cvp = hv_fetchs(table, "charnames", FALSE);
5882ddb3
FC
2599 if (cvp && (cv = *cvp) && SvROK(cv) && (rv = SvRV(cv),
2600 SvTYPE(rv) == SVt_PVCV) && ((stash = CvSTASH(rv)) != NULL))
0c415a79
KW
2601 {
2602 const char * const name = HvNAME(stash);
6a642c21
FC
2603 if (HvNAMELEN(stash) == sizeof("_charnames")-1
2604 && strEQ(name, "_charnames")) {
0c415a79
KW
2605 return res;
2606 }
2607 }
2608
bde9e88d
KW
2609 /* Here, it isn't Perl's charname handler. We can't rely on a
2610 * user-supplied handler to validate the input name. For non-ut8 input,
2611 * look to see that the first character is legal. Then loop through the
2612 * rest checking that each is a continuation */
6f613c73 2613
36897d64
KW
2614 /* This code makes the reasonable assumption that the only Latin1-range
2615 * characters that begin a character name alias are alphabetic, otherwise
2616 * would have to create a isCHARNAME_BEGIN macro */
b6ba1137 2617
b6ba1137 2618 if (! UTF) {
bde9e88d 2619 if (! isALPHAU(*s)) {
b6ba1137
KW
2620 goto bad_charname;
2621 }
bde9e88d
KW
2622 s++;
2623 while (s < e) {
2624 if (! isCHARNAME_CONT(*s)) {
b6ba1137
KW
2625 goto bad_charname;
2626 }
2d8eb851
KW
2627 if (*s == ' ' && *(s-1) == ' ') {
2628 goto multi_spaces;
bd299e29 2629 }
bde9e88d 2630 s++;
b6ba1137
KW
2631 }
2632 }
2633 else {
bde9e88d
KW
2634 /* Similarly for utf8. For invariants can check directly; for other
2635 * Latin1, can calculate their code point and check; otherwise use a
2636 * swash */
2637 if (UTF8_IS_INVARIANT(*s)) {
2638 if (! isALPHAU(*s)) {
140b12ad
KW
2639 goto bad_charname;
2640 }
bde9e88d
KW
2641 s++;
2642 } else if (UTF8_IS_DOWNGRADEABLE_START(*s)) {
a62b247b 2643 if (! isALPHAU(EIGHT_BIT_UTF8_TO_NATIVE(*s, *(s+1)))) {
b6ba1137 2644 goto bad_charname;
6f613c73 2645 }
bde9e88d 2646 s += 2;
6f613c73 2647 }
bde9e88d
KW
2648 else {
2649 if (! PL_utf8_charname_begin) {
2650 U8 flags = _CORE_SWASH_INIT_ACCEPT_INVLIST;
2651 PL_utf8_charname_begin = _core_swash_init("utf8",
2652 "_Perl_Charname_Begin",
2653 &PL_sv_undef,
2654 1, 0, NULL, &flags);
2655 }
2656 if (! swash_fetch(PL_utf8_charname_begin, (U8 *) s, TRUE)) {
2657 goto bad_charname;
2658 }
2659 s += UTF8SKIP(s);
2660 }
2661
2662 while (s < e) {
2663 if (UTF8_IS_INVARIANT(*s)) {
2664 if (! isCHARNAME_CONT(*s)) {
2665 goto bad_charname;
2666 }
2d8eb851
KW
2667 if (*s == ' ' && *(s-1) == ' ') {
2668 goto multi_spaces;
bd299e29 2669 }
bde9e88d
KW
2670 s++;
2671 }
2672 else if (UTF8_IS_DOWNGRADEABLE_START(*s)) {
a62b247b 2673 if (! isCHARNAME_CONT(EIGHT_BIT_UTF8_TO_NATIVE(*s, *(s+1))))
bde9e88d
KW
2674 {
2675 goto bad_charname;
2676 }
2677 s += 2;
2678 }
2679 else {
2680 if (! PL_utf8_charname_continue) {
2681 U8 flags = _CORE_SWASH_INIT_ACCEPT_INVLIST;
2682 PL_utf8_charname_continue = _core_swash_init("utf8",
2683 "_Perl_Charname_Continue",
2684 &PL_sv_undef,
2685 1, 0, NULL, &flags);
2686 }
2687 if (! swash_fetch(PL_utf8_charname_continue, (U8 *) s, TRUE)) {
2688 goto bad_charname;
2689 }
2690 s += UTF8SKIP(s);
6f613c73
KW
2691 }
2692 }
2d8eb851
KW
2693 }
2694 if (*(s-1) == ' ') {
2695 yyerror_pv(
2696 Perl_form(aTHX_
2697 "charnames alias definitions may not contain trailing "
2698 "white-space; marked by <-- HERE in %.*s<-- HERE %.*s",
2699 (int)(s - backslash_ptr + 1), backslash_ptr,
2700 (int)(e - s + 1), s + 1
2701 ),
2702 UTF ? SVf_UTF8 : 0);
2703 return NULL;
6f613c73
KW
2704 }
2705
94ca1619 2706 if (SvUTF8(res)) { /* Don't accept malformed input */
bde9e88d
KW
2707 const U8* first_bad_char_loc;
2708 STRLEN len;
2709 const char* const str = SvPV_const(res, len);
2710 if (! is_utf8_string_loc((U8 *) str, len, &first_bad_char_loc)) {
75219bac
KW
2711 _force_out_malformed_utf8_message(first_bad_char_loc,
2712 (U8 *) PL_parser->bufend,
2713 0,
2714 0 /* 0 means don't die */ );
bde9e88d
KW
2715 yyerror_pv(
2716 Perl_form(aTHX_
2717 "Malformed UTF-8 returned by %.*s immediately after '%.*s'",
2718 (int) (e - backslash_ptr + 1), backslash_ptr,
2719 (int) ((char *) first_bad_char_loc - str), str
2720 ),
2721 SVf_UTF8);
2722 return NULL;
2723 }
2724 }
140b12ad 2725
bde9e88d 2726 return res;
140b12ad 2727
bde9e88d 2728 bad_charname: {
bde9e88d
KW
2729
2730 /* The final %.*s makes sure that should the trailing NUL be missing
2731 * that this print won't run off the end of the string */
2732 yyerror_pv(
2733 Perl_form(aTHX_
2734 "Invalid character in \\N{...}; marked by <-- HERE in %.*s<-- HERE %.*s",
2d8eb851
KW
2735 (int)(s - backslash_ptr + 1), backslash_ptr,
2736 (int)(e - s + 1), s + 1
bde9e88d
KW
2737 ),
2738 UTF ? SVf_UTF8 : 0);
2739 return NULL;
2740 }
2d8eb851
KW
2741
2742 multi_spaces:
2743 yyerror_pv(
2744 Perl_form(aTHX_
2745 "charnames alias definitions may not contain a sequence of "
2746 "multiple spaces; marked by <-- HERE in %.*s<-- HERE %.*s",
2747 (int)(s - backslash_ptr + 1), backslash_ptr,
2748 (int)(e - s + 1), s + 1
2749 ),
2750 UTF ? SVf_UTF8 : 0);
2751 return NULL;
6f613c73
KW
2752}
2753
02aa26ce
NT
2754/*
2755 scan_const
2756
9da1dd8f
DM
2757 Extracts the next constant part of a pattern, double-quoted string,
2758 or transliteration. This is terrifying code.
2759
2760 For example, in parsing the double-quoted string "ab\x63$d", it would
2761 stop at the '$' and return an OP_CONST containing 'abc'.
02aa26ce 2762
94def140 2763 It looks at PL_lex_inwhat and PL_lex_inpat to find out whether it's
3280af22 2764 processing a pattern (PL_lex_inpat is true), a transliteration
94def140 2765 (PL_lex_inwhat == OP_TRANS is true), or a double-quoted string.
02aa26ce 2766
94def140
ST
2767 Returns a pointer to the character scanned up to. If this is
2768 advanced from the start pointer supplied (i.e. if anything was
9da1dd8f 2769 successfully parsed), will leave an OP_CONST for the substring scanned
6154021b 2770 in pl_yylval. Caller must intuit reason for not parsing further
9b599b2a
GS
2771 by looking at the next characters herself.
2772
02aa26ce 2773 In patterns:
9da1dd8f 2774 expand:
537124e4
KW
2775 \N{FOO} => \N{U+hex_for_character_FOO}
2776 (if FOO expands to multiple characters, expands to \N{U+xx.XX.yy ...})
9da1dd8f
DM
2777
2778 pass through:
2779 all other \-char, including \N and \N{ apart from \N{ABC}
2780
2781 stops on:
2782 @ and $ where it appears to be a var, but not for $ as tail anchor
2783 \l \L \u \U \Q \E
2784 (?{ or (??{
2785
02aa26ce
NT
2786 In transliterations:
2787 characters are VERY literal, except for - not at the start or end
94def140
ST
2788 of the string, which indicates a range. If the range is in bytes,
2789 scan_const expands the range to the full set of intermediate
2790 characters. If the range is in utf8, the hyphen is replaced with
2791 a certain range mark which will be handled by pmtrans() in op.c.
02aa26ce
NT
2792
2793 In double-quoted strings:
2794 backslashes:
2795 double-quoted style: \r and \n
ff3f963a 2796 constants: \x31, etc.
94def140 2797 deprecated backrefs: \1 (in substitution replacements)
02aa26ce
NT
2798 case and quoting: \U \Q \E
2799 stops on @ and $
2800
2801 scan_const does *not* construct ops to handle interpolated strings.
2802 It stops processing as soon as it finds an embedded $ or @ variable
2803 and leaves it to the caller to work out what's going on.
2804
94def140
ST
2805 embedded arrays (whether in pattern or not) could be:
2806 @foo, @::foo, @'foo, @{foo}, @$foo, @+, @-.
2807
2808 $ in double-quoted strings must be the symbol of an embedded scalar.
02aa26ce
NT
2809
2810 $ in pattern could be $foo or could be tail anchor. Assumption:
2811 it's a tail anchor if $ is the last thing in the string, or if it's
94def140 2812 followed by one of "()| \r\n\t"
02aa26ce 2813
9da1dd8f 2814 \1 (backreferences) are turned into $1 in substitutions
02aa26ce
NT
2815
2816 The structure of the code is
2817 while (there's a character to process) {
94def140
ST
2818 handle transliteration ranges
2819 skip regexp comments /(?#comment)/ and codes /(?{code})/
2820 skip #-initiated comments in //x patterns
2821 check for embedded arrays
02aa26ce
NT
2822 check for embedded scalars
2823 if (backslash) {
94def140 2824 deprecate \1 in substitution replacements
02aa26ce
NT
2825 handle string-changing backslashes \l \U \Q \E, etc.
2826 switch (what was escaped) {
94def140 2827 handle \- in a transliteration (becomes a literal -)
ff3f963a 2828 if a pattern and not \N{, go treat as regular character
94def140
ST
2829 handle \132 (octal characters)
2830 handle \x15 and \x{1234} (hex characters)
ff3f963a 2831 handle \N{name} (named characters, also \N{3,5} in a pattern)
94def140
ST
2832 handle \cV (control characters)
2833 handle printf-style backslashes (\f, \r, \n, etc)
02aa26ce 2834 } (end switch)
77a135fe 2835 continue
02aa26ce 2836 } (end if backslash)
77a135fe 2837 handle regular character
02aa26ce 2838 } (end while character to read)
4e553d73 2839
02aa26ce
NT
2840*/
2841
76e3520e 2842STATIC char *
cea2e8a9 2843S_scan_const(pTHX_ char *start)
79072805 2844{
eb578fdb 2845 char *send = PL_bufend; /* end of the constant */
dc023dbb
KW
2846 SV *sv = newSV(send - start); /* sv for the constant. See note below
2847 on sizing. */
eb578fdb
KW
2848 char *s = start; /* start of the constant */
2849 char *d = SvPVX(sv); /* destination for copies */
dc023dbb
KW
2850 bool dorange = FALSE; /* are we in a translit range? */
2851 bool didrange = FALSE; /* did we just finish a range? */
2852 bool in_charclass = FALSE; /* within /[...]/ */
2853 bool has_utf8 = FALSE; /* Output constant is UTF8 */
2854 bool this_utf8 = cBOOL(UTF); /* Is the source string assumed to be
2855 UTF8? But, this can show as true
2856 when the source isn't utf8, as for
2857 example when it is entirely composed
2858 of hex constants */
af9be36c
KW
2859 STRLEN utf8_variant_count = 0; /* When not in UTF-8, this counts the
2860 number of characters found so far
2861 that will expand (into 2 bytes)
2862 should we have to convert to
2863 UTF-8) */
6f613c73 2864 SV *res; /* result from charnames */
f4240379
KW
2865 STRLEN offset_to_max; /* The offset in the output to where the range
2866 high-end character is temporarily placed */
77a135fe
KW
2867
2868 /* Note on sizing: The scanned constant is placed into sv, which is
2869 * initialized by newSV() assuming one byte of output for every byte of
2870 * input. This routine expects newSV() to allocate an extra byte for a
2871 * trailing NUL, which this routine will append if it gets to the end of
2872 * the input. There may be more bytes of input than output (eg., \N{LATIN
2873 * CAPITAL LETTER A}), or more output than input if the constant ends up
2874 * recoded to utf8, but each time a construct is found that might increase
2875 * the needed size, SvGROW() is called. Its size parameter each time is
2876 * based on the best guess estimate at the time, namely the length used so
2877 * far, plus the length the current construct will occupy, plus room for
caae0700 2878 * the trailing NUL, plus one byte for every input byte still unscanned */
77a135fe 2879
c3320c2a
KW
2880 UV uv = UV_MAX; /* Initialize to weird value to try to catch any uses
2881 before set */
4c3a8340 2882#ifdef EBCDIC
f4240379
KW
2883 int backslash_N = 0; /* ? was the character from \N{} */
2884 int non_portable_endpoint = 0; /* ? In a range is an endpoint
2885 platform-specific like \x65 */
4c3a8340 2886#endif
012bcf8d 2887
7918f24d
NC
2888 PERL_ARGS_ASSERT_SCAN_CONST;
2889
bb16bae8 2890 assert(PL_lex_inwhat != OP_TRANSR);
7ef70b3d 2891 if (PL_lex_inwhat == OP_TRANS && PL_parser->lex_sub_op) {
2b9d42f0 2892 /* If we are doing a trans and we know we want UTF8 set expectation */
7ef70b3d
FC
2893 has_utf8 = PL_parser->lex_sub_op->op_private & (OPpTRANS_FROM_UTF|OPpTRANS_TO_UTF);
2894 this_utf8 = PL_parser->lex_sub_op->op_private & (PL_lex_repl ? OPpTRANS_FROM_UTF : OPpTRANS_TO_UTF);
2b9d42f0
NIS
2895 }
2896
b899e89d
FC
2897 /* Protect sv from errors and fatal warnings. */
2898 ENTER_with_name("scan_const");
2899 SAVEFREESV(sv);
2b9d42f0 2900
f4240379
KW
2901 while (s < send
2902 || dorange /* Handle tr/// range at right edge of input */
2903 ) {
ff3f963a 2904
02aa26ce 2905 /* get transliterations out of the way (they're most literal) */
3280af22 2906 if (PL_lex_inwhat == OP_TRANS) {
02aa26ce 2907
f4240379
KW
2908 /* But there isn't any special handling necessary unless there is a
2909 * range, so for most cases we just drop down and handle the value
2910 * as any other. There are two exceptions.
2911 *
2912 * 1. A minus sign indicates that we are actually going to have
2913 * a range. In this case, skip the '-', set a flag, then drop
2914 * down to handle what should be the end range value.
2915 * 2. After we've handled that value, the next time through, that
2916 * flag is set and we fix up the range.
2917 *
2918 * Ranges entirely within Latin1 are expanded out entirely, in
2919 * order to avoid the significant overhead of making a swash.
2920 * Ranges that extend above Latin1 have to have a swash, so there
4698efb6
KW
2921 * is no advantage to abbreviating them here, so they are stored
2922 * here as Min, ILLEGAL_UTF8_BYTE, Max. The illegal byte signifies
2923 * a hyphen without any possible ambiguity. On EBCDIC machines, if
f4240379
KW
2924 * the range is expressed as Unicode, the Latin1 portion is
2925 * expanded out even if the entire range extends above Latin1.
2926 * This is because each code point in it has to be processed here
2927 * individually to get its native translation */
2928
2929 if (! dorange) {
2930
2931 /* Here, we don't think we're in a range. If we've processed
2932 * at least one character, then see if this next one is a '-',
2933 * indicating the previous one was the start of a range. But
2934 * don't bother if we're too close to the end for the minus to
2935 * mean that. */
2936 if (*s != '-' || s >= send - 1 || s == start) {
2937
2938 /* A regular character. Process like any other, but first
2939 * clear any flags */
2940 didrange = FALSE;
2941 dorange = FALSE;
e294cc5d 2942#ifdef EBCDIC
f4240379
KW
2943 non_portable_endpoint = 0;
2944 backslash_N = 0;
e294cc5d 2945#endif
f4240379
KW
2946 /* Drops down to generic code to process current byte */
2947 }
2948 else {
2949 if (didrange) { /* Something like y/A-C-Z// */
2950 Perl_croak(aTHX_ "Ambiguous range in transliteration operator");
2951 }
e294cc5d 2952
f4240379 2953 dorange = TRUE;
2b9d42f0 2954
f4240379
KW
2955 s++; /* Skip past the minus */
2956
2957 /* d now points to where the end-range character will be
2958 * placed. Save it so won't have to go finding it later,
2959 * and drop down to get that character. (Actually we
2960 * instead save the offset, to handle the case where a
2961 * realloc in the meantime could change the actual
2962 * pointer). We'll finish processing the range the next
2963 * time through the loop */
2964 offset_to_max = d - SvPVX_const(sv);
2965 }
2966 } /* End of not a range */
2967 else {
2968 /* Here we have parsed a range. Now must handle it. At this
2969 * point:
2970 * 'sv' is a SV* that contains the output string we are
2971 * constructing. The final two characters in that string
2972 * are the range start and range end, in order.
2973 * 'd' points to just beyond the range end in the 'sv' string,
2974 * where we would next place something
2975 * 'offset_to_max' is the offset in 'sv' at which the character
2976 * before 'd' begins.
2977 */
2978 const char * max_ptr = SvPVX_const(sv) + offset_to_max;
2979 const char * min_ptr;
2980 IV range_min;
2981 IV range_max; /* last character in range */
2982 STRLEN save_offset;
2983 STRLEN grow;
11327fa1 2984#ifdef EBCDIC
f4240379
KW
2985 bool convert_unicode;
2986 IV real_range_max = 0;
e294cc5d 2987#endif
f4240379
KW
2988
2989 /* Get the range-ends code point values. */
e294cc5d 2990 if (has_utf8) {
f4240379
KW
2991 /* We know the utf8 is valid, because we just constructed
2992 * it ourselves in previous loop iterations */
2993 min_ptr = (char*) utf8_hop( (U8*) max_ptr, -1);
2994 range_min = valid_utf8_to_uvchr( (U8*) min_ptr, NULL);
2995 range_max = valid_utf8_to_uvchr( (U8*) max_ptr, NULL);
e294cc5d 2996 }
f4240379
KW
2997 else {
2998 min_ptr = max_ptr - 1;
2999 range_min = * (U8*) min_ptr;
3000 range_max = * (U8*) max_ptr;
3001 }
3002
e294cc5d 3003#ifdef EBCDIC
f4240379
KW
3004 /* On EBCDIC platforms, we may have to deal with portable
3005 * ranges. These happen if at least one range endpoint is a
3006 * Unicode value (\N{...}), or if the range is a subset of
3007 * [A-Z] or [a-z], and both ends are literal characters,
3008 * like 'A', and not like \x{C1} */
3009 if ((convert_unicode
3010 = cBOOL(backslash_N) /* \N{} forces Unicode, hence
3011 portable range */
3012 || ( ! non_portable_endpoint
3013 && (( isLOWER_A(range_min) && isLOWER_A(range_max))
3014 || (isUPPER_A(range_min) && isUPPER_A(range_max))))
3015 )) {
3016
3017 /* Special handling is needed for these portable ranges.
3018 * They are defined to all be in Unicode terms, which
3019 * include all Unicode code points between the end points.
3020 * Convert to Unicode to get the Unicode range. Later we
3021 * will convert each code point in the range back to
3022 * native. */
3023 range_min = NATIVE_TO_UNI(range_min);
3024 range_max = NATIVE_TO_UNI(range_max);
3025 }
e294cc5d 3026#endif
8ada0baa 3027
f4240379 3028 if (range_min > range_max) {
11327fa1 3029#ifdef EBCDIC
f4240379
KW
3030 if (convert_unicode) {
3031 /* Need to convert back to native for meaningful
3032 * messages for this platform */
3033 range_min = UNI_TO_NATIVE(range_min);
3034 range_max = UNI_TO_NATIVE(range_max);
3035 }
11327fa1 3036#endif
f4240379
KW
3037
3038 /* Use the characters themselves for the error message if
3039 * ASCII printables; otherwise some visible representation
3040 * of them */
3041 if (isPRINT_A(range_min) && isPRINT_A(range_max)) {
3042 Perl_croak(aTHX_
3043 "Invalid range \"%c-%c\" in transliteration operator",
3044 (char)range_min, (char)range_max);
3045 }
11327fa1 3046#ifdef EBCDIC
f4240379
KW
3047 else if (convert_unicode) {
3048 /* diag_listed_as: Invalid range "%s" in transliteration operator */
3049 Perl_croak(aTHX_
147e3846 3050 "Invalid range \"\\N{U+%04" UVXf "}-\\N{U+%04" UVXf "}\""
f4240379
KW
3051 " in transliteration operator",
3052 range_min, range_max);
3053 }
11327fa1 3054#endif
f4240379
KW
3055 else {
3056 /* diag_listed_as: Invalid range "%s" in transliteration operator */
3057 Perl_croak(aTHX_
147e3846 3058 "Invalid range \"\\x{%04" UVXf "}-\\x{%04" UVXf "}\""
f4240379
KW
3059 " in transliteration operator",
3060 range_min, range_max);
3061 }
c2e66d9e
GS
3062 }
3063
f4240379
KW
3064 if (has_utf8) {
3065
3066 /* We try to avoid creating a swash. If the upper end of
3067 * this range is below 256, this range won't force a swash;
3068 * otherwise it does force a swash, and as long as we have
3069 * to have one, we might as well not expand things out.
3070 * But if it's EBCDIC, we may have to look at each
3071 * character below 256 if we have to convert to/from
3072 * Unicode values */
3073 if (range_max > 255
c7f1f016 3074#ifdef EBCDIC
f4240379 3075 && (range_min > 255 || ! convert_unicode)
8ada0baa 3076#endif
f4240379
KW
3077 ) {
3078 /* Move the high character one byte to the right; then
3079 * insert between it and the range begin, an illegal
3080 * byte which serves to indicate this is a range (using
3081 * a '-' could be ambiguous). */
3082 char *e = d++;
3083 while (e-- > max_ptr) {
3084 *(e + 1) = *e;
e294cc5d 3085 }
f4240379
KW
3086 *(e + 1) = (char) ILLEGAL_UTF8_BYTE;
3087 goto range_done;
3088 }
3089
3090 /* Here, we're going to expand out the range. For EBCDIC
3091 * the range can extend above 255 (not so in ASCII), so
3092 * for EBCDIC, split it into the parts above and below
3093 * 255/256 */
e294cc5d 3094#ifdef EBCDIC
f4240379
KW
3095 if (range_max > 255) {
3096 real_range_max = range_max;
3097 range_max = 255;
3098 }
e294cc5d 3099#endif
f4240379 3100 }
02aa26ce 3101
f4240379
KW
3102 /* Here we need to expand out the string to contain each
3103 * character in the range. Grow the output to handle this */
3104
3105 save_offset = min_ptr - SvPVX_const(sv);
3106
3107 /* The base growth is the number of code points in the range */
3108 grow = range_max - range_min + 1;
3109 if (has_utf8) {
3110
3111 /* But if the output is UTF-8, some of those characters may
3112 * need two bytes (since the maximum range value here is
3113 * 255, the max bytes per character is two). On ASCII
3114 * platforms, it's not much trouble to get an accurate
3115 * count of what's needed. But on EBCDIC, the ones that
3116 * need 2 bytes are scattered around, so just use a worst
3117 * case value instead of calculating for that platform. */
4c3a8340 3118#ifdef EBCDIC
f4240379
KW
3119 grow *= 2;
3120#else
3121 /* Only those above 127 require 2 bytes. This may be
3122 * everything in the range, or not */
3123 if (range_min > 127) {
3124 grow *= 2;
3125 }
3126 else if (range_max > 127) {
3127 grow += range_max - 127;
3128 }
4c3a8340 3129#endif
f4240379
KW
3130 }
3131
3132 /* Subtract 3 for the bytes that were already accounted for
3133 * (min, max, and the hyphen) */
10656159 3134 d = save_offset + SvGROW(sv, SvLEN(sv) + grow - 3);
f4240379 3135
11327fa1
AL
3136#ifdef EBCDIC
3137 /* Here, we expand out the range. */
f4240379
KW
3138 if (convert_unicode) {
3139 IV i;
02aa26ce 3140
f4240379
KW
3141 /* Recall that the min and max are now in Unicode terms, so
3142 * we have to convert each character to its native
3143 * equivalent */
3144 if (has_utf8) {
3145 for (i = range_min; i <= range_max; i++) {
81324705 3146 append_utf8_from_native_byte(LATIN1_TO_NATIVE((U8) i),
f4240379
KW
3147 (U8 **) &d);
3148 }
3149 }
3150 else {
3151 for (i = range_min; i <= range_max; i++) {
81324705 3152 *d++ = (char)LATIN1_TO_NATIVE((U8) i);
f4240379
KW
3153 }
3154 }
01ec43d0 3155 }
11327fa1
AL
3156 else
3157#endif
3158 /* Always gets run for ASCII, and sometimes for EBCDIC. */
3159 {
f4240379
KW
3160 IV i;
3161
3162 /* Here, no conversions are necessary, which means that the
3163 * first character in the range is already in 'd' and
3164 * valid, so we can skip overwriting it */
3165 if (has_utf8) {
3166 d += UTF8SKIP(d);
3167 for (i = range_min + 1; i <= range_max; i++) {
81324705 3168 append_utf8_from_native_byte((U8) i, (U8 **) &d);
f4240379
KW
3169 }
3170 }
3171 else {
3172 d++;
3173 for (i = range_min + 1; i <= range_max; i++) {
3174 *d++ = (char)i;
3175 }
3176 }
a0ed51b3 3177 }
02aa26ce 3178
11327fa1
AL
3179#ifdef EBCDIC
3180 /* If the original range extended above 255, add in that portion. */
f4240379
KW
3181 if (real_range_max) {
3182 *d++ = (char) UTF8_TWO_BYTE_HI(0x100);
3183 *d++ = (char) UTF8_TWO_BYTE_LO(0x100);
3184 if (real_range_max > 0x101)
3185 *d++ = (char) ILLEGAL_UTF8_BYTE;
3186 if (real_range_max > 0x100)
3187 d = (char*)uvchr_to_utf8((U8*)d, real_range_max);
3188 }
11327fa1 3189#endif
02aa26ce 3190
f4240379
KW
3191 range_done:
3192 /* mark the range as done, and continue */
3193 didrange = TRUE;
3194 dorange = FALSE;
3195#ifdef EBCDIC
3196 non_portable_endpoint = 0;
3197 backslash_N = 0;
3198#endif
3199 continue;
3200 } /* End of is a range */
3201 } /* End of transliteration. Joins main code after these else's */
e4a2df84
DM
3202 else if (*s == '[' && PL_lex_inpat && !in_charclass) {
3203 char *s1 = s-1;
3204 int esc = 0;
3205 while (s1 >= start && *s1-- == '\\')
3206 esc = !esc;
3207 if (!esc)
3208 in_charclass = TRUE;
3209 }
1e02a175 3210 else if (*s == ']' && PL_lex_inpat && in_charclass) {
e4a2df84
DM
3211 char *s1 = s-1;
3212 int esc = 0;
3213 while (s1 >= start && *s1-- == '\\')
3214 esc = !esc;
3215 if (!esc)
3216 in_charclass = FALSE;
3217 }
2866decb 3218
9da1dd8f
DM
3219 /* skip for regexp comments /(?#comment)/, except for the last
3220 * char, which will be done separately.
3221 * Stop on (?{..}) and friends */
3222
c30fc27b 3223 else if (*s == '(' && PL_lex_inpat && s[1] == '?' && !in_charclass) {
cc6b7395 3224 if (s[2] == '#') {
e994fd66 3225 while (s+1 < send && *s != ')')
5ff03569 3226 *d++ = *s++;
155aba94 3227 }
407f8cf2
KW
3228 else if (!PL_lex_casemods
3229 && ( s[2] == '{' /* This should match regcomp.c */
3230 || (s[2] == '?' && s[3] == '{')))
155aba94 3231 {
9da1dd8f 3232 break;
cc6b7395 3233 }
748a9306 3234 }
02aa26ce
NT
3235
3236 /* likewise skip #-initiated comments in //x patterns */
407f8cf2
KW
3237 else if (*s == '#'
3238 && PL_lex_inpat
3239 && !in_charclass
3240 && ((PMOP*)PL_lex_inpat)->op_pmflags & RXf_PMf_EXTENDED)
3241 {
748a9306 3242 while (s+1 < send && *s != '\n')
5ff03569 3243 *d++ = *s++;
748a9306 3244 }
02aa26ce 3245
9da1dd8f
DM
3246 /* no further processing of single-quoted regex */
3247 else if (PL_lex_inpat && SvIVX(PL_linestr) == '\'')
3248 goto default_action;
3249
5d1d4326 3250 /* check for embedded arrays
da6eedaa 3251 (@foo, @::foo, @'foo, @{foo}, @$foo, @+, @-)
5d1d4326 3252 */
1749ea0d 3253 else if (*s == '@' && s[1]) {
fac0f7a3
KW
3254 if (UTF
3255 ? isIDFIRST_utf8_safe(s+1, send)
3256 : isWORDCHAR_A(s[1]))
3257 {
1749ea0d 3258 break;
fac0f7a3 3259 }
1749ea0d
ST
3260 if (strchr(":'{$", s[1]))
3261 break;
3262 if (!PL_lex_inpat && (s[1] == '+' || s[1] == '-'))
3263 break; /* in regexp, neither @+ nor @- are interpolated */
3264 }
02aa26ce
NT
3265
3266 /* check for embedded scalars. only stop if we're sure it's a
3267 variable.
3268 */
79072805 3269 else if (*s == '$') {
3280af22 3270 if (!PL_lex_inpat) /* not a regexp, so $ must be var */
79072805 3271 break;
77772344 3272 if (s + 1 < send && !strchr("()| \r\n\t", s[1])) {
a2a5de95
NC
3273 if (s[1] == '\\') {
3274 Perl_ck_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
3275 "Possible unintended interpolation of $\\ in regex");
77772344 3276 }
79072805 3277 break; /* in regexp, $ might be tail anchor */
77772344 3278 }
79072805 3279 }
02aa26ce 3280
2b9d42f0
NIS
3281 /* End of else if chain - OP_TRANS rejoin rest */
3282
02aa26ce 3283 /* backslashes */
79072805 3284 if (*s == '\\' && s+1 < send) {
ff3f963a
KW
3285 char* e; /* Can be used for ending '}', etc. */
3286
79072805 3287 s++;
02aa26ce 3288
7d0fc23c
KW
3289 /* warn on \1 - \9 in substitution replacements, but note that \11
3290 * is an octal; and \19 is \1 followed by '9' */
407f8cf2
KW
3291 if (PL_lex_inwhat == OP_SUBST
3292 && !PL_lex_inpat
3293 && isDIGIT(*s)
3294 && *s != '0'
3295 && !isDIGIT(s[1]))
79072805 3296 {
c782d7ee 3297 /* diag_listed_as: \%d better written as $%d */
a2a5de95 3298 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX), "\\%c better written as $%c", *s, *s);
79072805
LW
3299 *--s = '$';
3300 break;
3301 }
02aa26ce
NT
3302
3303 /* string-change backslash escapes */
838f2281 3304 if (PL_lex_inwhat != OP_TRANS && *s && strchr("lLuUEQF", *s)) {
79072805
LW
3305 --s;
3306 break;
3307 }
ff3f963a
KW
3308 /* In a pattern, process \N, but skip any other backslash escapes.
3309 * This is because we don't want to translate an escape sequence
3310 * into a meta symbol and have the regex compiler use the meta
3311 * symbol meaning, e.g. \x{2E} would be confused with a dot. But
3312 * in spite of this, we do have to process \N here while the proper
3313 * charnames handler is in scope. See bugs #56444 and #62056.
85fba779 3314 *
ff3f963a
KW
3315 * There is a complication because \N in a pattern may also stand
3316 * for 'match a non-nl', and not mean a charname, in which case its
3317 * processing should be deferred to the regex compiler. To be a
3318 * charname it must be followed immediately by a '{', and not look
3319 * like \N followed by a curly quantifier, i.e., not something like
3320 * \N{3,}. regcurly returns a boolean indicating if it is a legal
3321 * quantifier */
3322 else if (PL_lex_inpat
3323 && (*s != 'N'
3324 || s[1] != '{'
412f55bb 3325 || regcurly(s + 1)))
ff3f963a 3326 {
4d73d076 3327 *d++ = '\\';
cc74c5bd
ST
3328 goto default_action;
3329 }
02aa26ce 3330
79072805 3331 switch (*s) {
79072805 3332 default:
11b8faa4 3333 {
15861f94 3334 if ((isALPHANUMERIC(*s)))
a2a5de95
NC
3335 Perl_ck_warner(aTHX_ packWARN(WARN_MISC),
3336 "Unrecognized escape \\%c passed through",
3337 *s);
11b8faa4 3338 /* default action is to copy the quoted character */
f9a63242 3339 goto default_action;
11b8faa4 3340 }
02aa26ce 3341
632403cc 3342 /* eg. \132 indicates the octal constant 0132 */
79072805
LW
3343 case '0': case '1': case '2': case '3':
3344 case '4': case '5': case '6': case '7':
ba210ebe 3345 {
5e0a247b 3346 I32 flags = PERL_SCAN_SILENT_ILLDIGIT;
53305cf1 3347 STRLEN len = 3;
06972766 3348 uv = grok_oct(s, &len, &flags, NULL);
ba210ebe 3349 s += len;
5e0a247b
KW
3350 if (len < 3 && s < send && isDIGIT(*s)
3351 && ckWARN(WARN_MISC))
3352 {
3353 Perl_warner(aTHX_ packWARN(WARN_MISC),
3354 "%s", form_short_octal_warning(s, len));
3355 }
ba210ebe 3356 }
012bcf8d 3357 goto NUM_ESCAPE_INSERT;
02aa26ce 3358
f0a2b745
KW
3359 /* eg. \o{24} indicates the octal constant \024 */
3360 case 'o':
3361 {
454155d9 3362 const char* error;
f0a2b745 3363
00ce5563 3364 bool valid = grok_bslash_o(&s, &uv, &error,
80f4111b
KW
3365 TRUE, /* Output warning */
3366 FALSE, /* Not strict */
17896a85
KW
3367 TRUE, /* Output warnings for
3368 non-portables */
80f4111b 3369 UTF);
454155d9 3370 if (! valid) {
f0a2b745
KW
3371 yyerror(error);
3372 continue;
3373 }
3374 goto NUM_ESCAPE_INSERT;
3375 }
3376
77a135fe 3377 /* eg. \x24 indicates the hex constant 0x24 */
79072805 3378 case 'x':
a0481293 3379 {
a0481293 3380 const char* error;
355860ce 3381
00ce5563 3382 bool valid = grok_bslash_x(&s, &uv, &error,
80f4111b
KW
3383 TRUE, /* Output warning */
3384 FALSE, /* Not strict */
17896a85
KW
3385 TRUE, /* Output warnings for
3386 non-portables */
80f4111b 3387 UTF);
a0481293
KW
3388 if (! valid) {
3389 yyerror(error);
355860ce 3390 continue;
ba210ebe 3391 }
012bcf8d
GS
3392 }
3393
3394 NUM_ESCAPE_INSERT:
b67fd2c5 3395 /* Insert oct or hex escaped character. */
ba7cea30 3396
06972766 3397 /* Here uv is the ordinal of the next character being added */
f4240379
KW
3398 if (UVCHR_IS_INVARIANT(uv)) {
3399 *d++ = (char) uv;
3400 }
3401 else {
9aa983d2 3402 if (!has_utf8 && uv > 255) {
caae0700 3403
af9be36c
KW
3404 /* Here, 'uv' won't fit unless we convert to UTF-8.
3405 * If we've only seen invariants so far, all we have to
3406 * do is turn on the flag */
3407 if (utf8_variant_count == 0) {
3408 SvUTF8_on(sv);
3409 }
3410 else {
1e02a175
KW
3411 SvCUR_set(sv, d - SvPVX_const(sv));
3412 SvPOK_on(sv);
3413 *d = '\0';
3414
3415 sv_utf8_upgrade_flags_grow(
3416 sv,
3417 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE
3418 /* Above-latin1 in string
3419 * implies no encoding */
3420 |SV_UTF8_NO_ENCODING,
3421
2916094f
KW
3422 /* Since we're having to grow here,
3423 * make sure we have enough room for
3424 * this escape and a NUL, so the
3425 * code immediately below won't have
3426 * to actually grow again */
1e02a175
KW
3427 UVCHR_SKIP(uv)
3428 + (STRLEN)(send - s) + 1);
3429 d = SvPVX(sv) + SvCUR(sv);
af9be36c
KW
3430 }
3431
1e02a175 3432 has_utf8 = TRUE;
012bcf8d
GS
3433 }
3434
8d0042a8
KW
3435 if (! has_utf8) {
3436 *d++ = (char)uv;
af9be36c 3437 utf8_variant_count++;
8d0042a8
KW
3438 }
3439 else {
b67fd2c5
KW
3440 /* Usually, there will already be enough room in 'sv'
3441 * since such escapes are likely longer than any UTF-8
3442 * sequence they can end up as. This isn't the case on
3443 * EBCDIC where \x{40000000} contains 12 bytes, and the
3444 * UTF-8 for it contains 14. And, we have to allow for
3445 * a trailing NUL. It probably can't happen on ASCII
2916094f
KW
3446 * platforms, but be safe. See Note on sizing above. */
3447 const STRLEN needed = d - SvPVX(sv)
3448 + UVCHR_SKIP(uv)
3449 + (send - s)
b67fd2c5
KW
3450 + 1;
3451 if (UNLIKELY(needed > SvLEN(sv))) {
3452 SvCUR_set(sv, d - SvPVX_const(sv));
94beb101 3453 d = SvCUR(sv) + SvGROW(sv, needed);
b67fd2c5
KW
3454 }
3455
c80e42f3 3456 d = (char*)uvchr_to_utf8((U8*)d, uv);
407f8cf2 3457 if (PL_lex_inwhat == OP_TRANS
7ef70b3d 3458 && PL_parser->lex_sub_op)
407f8cf2 3459 {
7ef70b3d 3460 PL_parser->lex_sub_op->op_private |=
f9a63242
JH
3461 (PL_lex_repl ? OPpTRANS_FROM_UTF
3462 : OPpTRANS_TO_UTF);
f9a63242 3463 }
a0ed51b3 3464 }
012bcf8d 3465 }
f4240379
KW
3466#ifdef EBCDIC
3467 non_portable_endpoint++;
3468#endif
79072805 3469 continue;
02aa26ce 3470
4a2d328f 3471 case 'N':
85fba779
KW
3472 /* In a non-pattern \N must be like \N{U+0041}, or it can be a
3473 * named character, like \N{LATIN SMALL LETTER A}, or a named
3474 * sequence, like \N{LATIN CAPITAL LETTER A WITH MACRON AND
f4240379
KW
3475 * GRAVE} (except y/// can't handle the latter, croaking). For
3476 * convenience all three forms are referred to as "named
3477 * characters" below.
85fba779
KW
3478 *
3479 * For patterns, \N also can mean to match a non-newline. Code
3480 * before this 'switch' statement should already have handled
3481 * this situation, and hence this code only has to deal with
3482 * the named character cases.
3483 *
3484 * For non-patterns, the named characters are converted to
3485 * their string equivalents. In patterns, named characters are
3486 * not converted to their ultimate forms for the same reasons
3487 * that other escapes aren't. Instead, they are converted to
3488 * the \N{U+...} form to get the value from the charnames that
3489 * is in effect right now, while preserving the fact that it
3490 * was a named character, so that the regex compiler knows
3491 * this.
3492 *
3493 * The structure of this section of code (besides checking for
ff3f963a 3494 * errors and upgrading to utf8) is:
f4240379 3495 * If the named character is of the form \N{U+...}, pass it
85fba779
KW
3496 * through if a pattern; otherwise convert the code point
3497 * to utf8
f4240379
KW
3498 * Otherwise must be some \N{NAME}: convert to
3499 * \N{U+c1.c2...} if a pattern; otherwise convert to utf8
3500 *
3501 * Transliteration is an exception. The conversion to utf8 is
3502 * only done if the code point requires it to be representable.
85fba779 3503 *
b6d67071 3504 * Here, 's' points to the 'N'; the test below is guaranteed to
85fba779
KW
3505 * succeed if we are being called on a pattern, as we already
3506 * know from a test above that the next character is a '{'. A
3507 * non-pattern \N must mean 'named character', which requires
3508 * braces */
ff3f963a
KW
3509 s++;
3510 if (*s != '{') {
caae0700 3511 yyerror("Missing braces on \\N{}");
ff3f963a
KW
3512 continue;
3513 }
3514 s++;
3515
0a96133f 3516 /* If there is no matching '}', it is an error. */
ff3f963a
KW
3517 if (! (e = strchr(s, '}'))) {
3518 if (! PL_lex_inpat) {
5777a3f7 3519 yyerror("Missing right brace on \\N{}");
0a96133f 3520 } else {
4407f1b8 3521 yyerror("Missing right brace on \\N{} or unescaped left brace after \\N");
dbc0d4f2 3522 }
0a96133f 3523 continue;
ff3f963a 3524 }
cddc7ef4 3525
ff3f963a 3526 /* Here it looks like a named character */
cddc7ef4 3527
ff3f963a 3528 if (*s == 'U' && s[1] == '+') { /* \N{U+...} */
ff3f963a 3529 s += 2; /* Skip to next char after the 'U+' */
ff3f963a 3530 if (PL_lex_inpat) {
af352bf2
KW
3531
3532 /* In patterns, we can have \N{U+xxxx.yyyy.zzzz...} */
fb2eed93
FC
3533 /* Check the syntax. */
3534 const char *orig_s;
3535 orig_s = s - 5;
3536 if (!isXDIGIT(*s)) {
3537 bad_NU:
3538 yyerror(
3539 "Invalid hexadecimal number in \\N{U+...}"
3540 );
3541 s = e + 1;
3542 continue;
3543 }
3544 while (++s < e) {
3545 if (isXDIGIT(*s))
3546 continue;
3547 else if ((*s == '.' || *s == '_')
3548 && isXDIGIT(s[1]))
3549 continue;
3550 goto bad_NU;
4cbd7e22 3551 }
af352bf2 3552
fb2eed93
FC
3553 /* Pass everything through unchanged.
3554 * +1 is for the '}' */
4cbd7e22
FC
3555 Copy(orig_s, d, e - orig_s + 1, char);
3556 d += e - orig_s + 1;
ff3f963a
KW
3557 }
3558 else { /* Not a pattern: convert the hex to string */
fb2eed93
FC
3559 I32 flags = PERL_SCAN_ALLOW_UNDERSCORES
3560 | PERL_SCAN_SILENT_ILLDIGIT
3561 | PERL_SCAN_DISALLOW_PREFIX;
3562 STRLEN len = e - s;
3563 uv = grok_hex(s, &len, &flags, NULL);
3564 if (len == 0 || (len != (STRLEN)(e - s)))
3565 goto bad_NU;
ff3f963a 3566
f4240379
KW
3567 /* For non-tr///, if the destination is not in utf8,
3568 * unconditionally recode it to be so. This is
3569 * because \N{} implies Unicode semantics, and scalars
3570 * have to be in utf8 to guarantee those semantics.
3571 * tr/// doesn't care about Unicode rules, so no need
3572 * there to upgrade to UTF-8 for small enough code
3573 * points */
3574 if (! has_utf8 && ( uv > 0xFF
3575 || PL_lex_inwhat != OP_TRANS))
3576 {
af9be36c
KW
3577 /* See Note on sizing above. */
3578 const STRLEN extra = OFFUNISKIP(uv) + (send - e) + 1;
3579
ff3f963a
KW
3580 SvCUR_set(sv, d - SvPVX_const(sv));
3581 SvPOK_on(sv);
3582 *d = '\0';
af9be36c
KW
3583
3584 if (utf8_variant_count == 0) {
3585 SvUT