This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
Use fnc to force out malformed warnings
[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");
530 else if (PL_oldoldbufptr && isIDFIRST_lazy_if(PL_oldoldbufptr,UTF)) {
f54cb97a 531 const char *t;
8a2bca12 532 for (t = PL_oldoldbufptr; (isWORDCHAR_lazy_if(t,UTF) || *t == ':');
734ab321 533 t += UTF ? UTF8SKIP(t) : 1)
c35e046a 534 NOOP;
56da5a46
RGS
535 if (t < PL_bufptr && isSPACE(*t))
536 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
147e3846 537 "\t(Do you need to predeclare %" UTF8f "?)\n",
b17a0679 538 UTF8fARG(UTF, t - PL_oldoldbufptr, PL_oldoldbufptr));
56da5a46
RGS
539 }
540 else {
541 assert(s >= oldbp);
542 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
147e3846 543 "\t(Missing operator before %" UTF8f "?)\n",
b17a0679 544 UTF8fARG(UTF, s - oldbp, oldbp));
56da5a46 545 }
07c798fb 546 }
3280af22 547 PL_bufptr = oldbp;
8990e307
LW
548}
549
ffb4593c
NT
550/*
551 * S_missingterm
552 * Complain about missing quote/regexp/heredoc terminator.
d4c19fe8 553 * If it's called with NULL then it cauterizes the line buffer.
ffb4593c
NT
554 * If we're in a delimited string and the delimiter is a control
555 * character, it's reformatted into a two-char sequence like ^C.
556 * This is fatal.
557 */
558
76e3520e 559STATIC void
cea2e8a9 560S_missingterm(pTHX_ char *s)
8990e307 561{
e487ff5e 562 char tmpbuf[UTF8_MAXBYTES + 1];
8990e307 563 char q;
cb650135
FC
564 bool uni = FALSE;
565 SV *sv;
8990e307 566 if (s) {
9d4ba2ae 567 char * const nl = strrchr(s,'\n');
d2719217 568 if (nl)
8990e307 569 *nl = '\0';
7f7f7d08 570 uni = UTF;
8990e307 571 }
cb650135 572 else if (PL_multi_close < 32) {
8990e307 573 *tmpbuf = '^';
585ec06d 574 tmpbuf[1] = (char)toCTRL(PL_multi_close);
8990e307
LW
575 tmpbuf[2] = '\0';
576 s = tmpbuf;
577 }
578 else {
cb650135
FC
579 if (LIKELY(PL_multi_close < 256)) {
580 *tmpbuf = (char)PL_multi_close;
581 tmpbuf[1] = '\0';
582 }
583 else {
584 uni = TRUE;
585 *uvchr_to_utf8((U8 *)tmpbuf, PL_multi_close) = 0;
586 }
8990e307
LW
587 s = tmpbuf;
588 }
589 q = strchr(s,'"') ? '\'' : '"';
cb650135
FC
590 sv = sv_2mortal(newSVpv(s,0));
591 if (uni)
592 SvUTF8_on(sv);
147e3846 593 Perl_croak(aTHX_ "Can't find string terminator %c%" SVf
cb650135 594 "%c anywhere before EOF",q,SVfARG(sv),q);
463ee0b2 595}
79072805 596
dd0ac2b9
FC
597#include "feature.h"
598
0d863452 599/*
0d863452
RH
600 * Check whether the named feature is enabled.
601 */
26ea9e12 602bool
3fff3427 603Perl_feature_is_enabled(pTHX_ const char *const name, STRLEN namelen)
0d863452 604{
4a731d7b 605 char he_name[8 + MAX_FEATURE_LEN] = "feature_";
7918f24d
NC
606
607 PERL_ARGS_ASSERT_FEATURE_IS_ENABLED;
ca4d40c4
FC
608
609 assert(CURRENT_FEATURE_BUNDLE == FEATURE_BUNDLE_CUSTOM);
7918f24d 610
26ea9e12
NC
611 if (namelen > MAX_FEATURE_LEN)
612 return FALSE;
3fff3427 613 memcpy(&he_name[8], name, namelen);
7d69d4a6 614
c8ca97b0
NC
615 return cBOOL(cop_hints_fetch_pvn(PL_curcop, he_name, 8 + namelen, 0,
616 REFCOUNTED_HE_EXISTS));
0d863452
RH
617}
618
ffb4593c 619/*
9cbb5ea2
GS
620 * experimental text filters for win32 carriage-returns, utf16-to-utf8 and
621 * utf16-to-utf8-reversed.
ffb4593c
NT
622 */
623
c39cd008
GS
624#ifdef PERL_CR_FILTER
625static void
626strip_return(SV *sv)
627{
eb578fdb
KW
628 const char *s = SvPVX_const(sv);
629 const char * const e = s + SvCUR(sv);
7918f24d
NC
630
631 PERL_ARGS_ASSERT_STRIP_RETURN;
632
c39cd008
GS
633 /* outer loop optimized to do nothing if there are no CR-LFs */
634 while (s < e) {
635 if (*s++ == '\r' && *s == '\n') {
636 /* hit a CR-LF, need to copy the rest */
eb578fdb 637 char *d = s - 1;
c39cd008
GS
638 *d++ = *s++;
639 while (s < e) {
640 if (*s == '\r' && s[1] == '\n')
641 s++;
642 *d++ = *s++;
643 }
644 SvCUR(sv) -= s - d;
645 return;
646 }
647 }
648}
a868473f 649
76e3520e 650STATIC I32
c39cd008 651S_cr_textfilter(pTHX_ int idx, SV *sv, int maxlen)
a868473f 652{
f54cb97a 653 const I32 count = FILTER_READ(idx+1, sv, maxlen);
c39cd008
GS
654 if (count > 0 && !maxlen)
655 strip_return(sv);
656 return count;
a868473f
NIS
657}
658#endif
659
ffb4593c 660/*
8eaa0acf
Z
661=for apidoc Amx|void|lex_start|SV *line|PerlIO *rsfp|U32 flags
662
663Creates and initialises a new lexer/parser state object, supplying
664a context in which to lex and parse from a new source of Perl code.
665A pointer to the new state object is placed in L</PL_parser>. An entry
666is made on the save stack so that upon unwinding the new state object
667will be destroyed and the former value of L</PL_parser> will be restored.
668Nothing else need be done to clean up the parsing context.
669
2d7f6611 670The code to be parsed comes from C<line> and C<rsfp>. C<line>, if
8eaa0acf 671non-null, provides a string (in SV form) containing code to be parsed.
2d7f6611
KW
672A copy of the string is made, so subsequent modification of C<line>
673does not affect parsing. C<rsfp>, if non-null, provides an input stream
8eaa0acf 674from which code will be read to be parsed. If both are non-null, the
2d7f6611
KW
675code in C<line> comes first and must consist of complete lines of input,
676and C<rsfp> supplies the remainder of the source.
8eaa0acf 677
2d7f6611 678The C<flags> parameter is reserved for future use. Currently it is only
e368b3bd 679used by perl internally, so extensions should always pass zero.
8eaa0acf
Z
680
681=cut
682*/
ffb4593c 683
27fcb6ee 684/* LEX_START_SAME_FILTER indicates that this is not a new file, so it
87606032
NC
685 can share filters with the current parser.
686 LEX_START_DONT_CLOSE indicates that the file handle wasn't opened by the
687 caller, hence isn't owned by the parser, so shouldn't be closed on parser
688 destruction. This is used to handle the case of defaulting to reading the
689 script from the standard input because no filename was given on the command
690 line (without getting confused by situation where STDIN has been closed, so
691 the script handle is opened on fd 0) */
27fcb6ee 692
a0d0e21e 693void
8eaa0acf 694Perl_lex_start(pTHX_ SV *line, PerlIO *rsfp, U32 flags)
79072805 695{
6ef55633 696 const char *s = NULL;
5486870f 697 yy_parser *parser, *oparser;
60d63348 698 if (flags && flags & ~LEX_START_FLAGS)
8eaa0acf 699 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_start");
acdf0a21
DM
700
701 /* create and initialise a parser */
702
199e78b7 703 Newxz(parser, 1, yy_parser);
5486870f 704 parser->old_parser = oparser = PL_parser;
acdf0a21
DM
705 PL_parser = parser;
706
28ac2b49 707 parser->stack = NULL;
df13534a 708 parser->stack_max1 = NULL;
28ac2b49 709 parser->ps = NULL;
acdf0a21 710
e3abe207
DM
711 /* on scope exit, free this parser and restore any outer one */
712 SAVEPARSER(parser);
7c4baf47 713 parser->saved_curcop = PL_curcop;
e3abe207 714
acdf0a21 715 /* initialise lexer state */
8990e307 716
fb205e7a 717 parser->nexttoke = 0;
ca4cfd28 718 parser->error_count = oparser ? oparser->error_count : 0;
7f1c3e8c 719 parser->copline = parser->preambling = NOLINE;
5afb0a62 720 parser->lex_state = LEX_NORMAL;
c2598295 721 parser->expect = XSTATE;
2f9285f8 722 parser->rsfp = rsfp;
27fcb6ee
FC
723 parser->rsfp_filters =
724 !(flags & LEX_START_SAME_FILTER) || !oparser
d3cd8e11
FC
725 ? NULL
726 : MUTABLE_AV(SvREFCNT_inc(
727 oparser->rsfp_filters
728 ? oparser->rsfp_filters
729 : (oparser->rsfp_filters = newAV())
730 ));
2f9285f8 731
199e78b7
DM
732 Newx(parser->lex_brackstack, 120, char);
733 Newx(parser->lex_casestack, 12, char);
734 *parser->lex_casestack = '\0';
d794b522 735 Newxz(parser->lex_shared, 1, LEXSHARED);
02b34bbe 736
10efb74f 737 if (line) {
0528fd32 738 STRLEN len;
10efb74f 739 s = SvPV_const(line, len);
0abcdfa4
FC
740 parser->linestr = flags & LEX_START_COPIED
741 ? SvREFCNT_inc_simple_NN(line)
742 : newSVpvn_flags(s, len, SvUTF8(line));
b3dd0aba
FC
743 if (!rsfp)
744 sv_catpvs(parser->linestr, "\n;");
0abcdfa4 745 } else {
bf1b738b 746 parser->linestr = newSVpvn("\n;", rsfp ? 1 : 2);
8990e307 747 }
f06b5848
DM
748 parser->oldoldbufptr =
749 parser->oldbufptr =
750 parser->bufptr =
751 parser->linestart = SvPVX(parser->linestr);
752 parser->bufend = parser->bufptr + SvCUR(parser->linestr);
753 parser->last_lop = parser->last_uni = NULL;
b54f893d 754
6d59e610 755 STATIC_ASSERT_STMT(FITS_IN_8_BITS(LEX_IGNORE_UTF8_HINTS|LEX_EVALBYTES
b54f893d
KW
756 |LEX_DONT_CLOSE_RSFP));
757 parser->lex_flags = (U8) (flags & (LEX_IGNORE_UTF8_HINTS|LEX_EVALBYTES
758 |LEX_DONT_CLOSE_RSFP));
737c24fc 759
60d63348 760 parser->in_pod = parser->filtered = 0;
79072805 761}
a687059c 762
e3abe207
DM
763
764/* delete a parser object */
765
766void
767Perl_parser_free(pTHX_ const yy_parser *parser)
768{
7918f24d
NC
769 PERL_ARGS_ASSERT_PARSER_FREE;
770
7c4baf47 771 PL_curcop = parser->saved_curcop;
bdc0bf6f
DM
772 SvREFCNT_dec(parser->linestr);
773
87606032 774 if (PL_parser->lex_flags & LEX_DONT_CLOSE_RSFP)
2f9285f8 775 PerlIO_clearerr(parser->rsfp);
407f8cf2
KW
776 else if (parser->rsfp && (!parser->old_parser
777 || (parser->old_parser && parser->rsfp != parser->old_parser->rsfp)))
2f9285f8 778 PerlIO_close(parser->rsfp);
5486870f 779 SvREFCNT_dec(parser->rsfp_filters);
10002bc1 780 SvREFCNT_dec(parser->lex_stuff);
7ef70b3d 781 SvREFCNT_dec(parser->lex_sub_repl);
3ac7ff8f
FC
782
783 Safefree(parser->lex_brackstack);
784 Safefree(parser->lex_casestack);
785 Safefree(parser->lex_shared);
786 PL_parser = parser->old_parser;
787 Safefree(parser);
788}
789
790void
791Perl_parser_free_nexttoke_ops(pTHX_ yy_parser *parser, OPSLAB *slab)
792{
3ac7ff8f 793 I32 nexttoke = parser->nexttoke;
3ac7ff8f 794 PERL_ARGS_ASSERT_PARSER_FREE_NEXTTOKE_OPS;
3ce3dcd9 795 while (nexttoke--) {
3ac7ff8f
FC
796 if (S_is_opval_token(parser->nexttype[nexttoke] & 0xffff)
797 && parser->nextval[nexttoke].opval
798 && parser->nextval[nexttoke].opval->op_slabbed
799 && OpSLAB(parser->nextval[nexttoke].opval) == slab) {
3ce3dcd9 800 op_free(parser->nextval[nexttoke].opval);
3ac7ff8f
FC
801 parser->nextval[nexttoke].opval = NULL;
802 }
3ce3dcd9 803 }
e3abe207
DM
804}
805
806
ffb4593c 807/*
f0e67a1d
Z
808=for apidoc AmxU|SV *|PL_parser-E<gt>linestr
809
810Buffer scalar containing the chunk currently under consideration of the
811text currently being lexed. This is always a plain string scalar (for
812which C<SvPOK> is true). It is not intended to be used as a scalar by
813normal scalar means; instead refer to the buffer directly by the pointer
814variables described below.
815
816The lexer maintains various C<char*> pointers to things in the
817C<PL_parser-E<gt>linestr> buffer. If C<PL_parser-E<gt>linestr> is ever
818reallocated, all of these pointers must be updated. Don't attempt to
819do this manually, but rather use L</lex_grow_linestr> if you need to
820reallocate the buffer.
821
822The content of the text chunk in the buffer is commonly exactly one
823complete line of input, up to and including a newline terminator,
824but there are situations where it is otherwise. The octets of the
825buffer may be intended to be interpreted as either UTF-8 or Latin-1.
826The function L</lex_bufutf8> tells you which. Do not use the C<SvUTF8>
827flag on this scalar, which may disagree with it.
828
829For direct examination of the buffer, the variable
830L</PL_parser-E<gt>bufend> points to the end of the buffer. The current
831lexing position is pointed to by L</PL_parser-E<gt>bufptr>. Direct use
832of these pointers is usually preferable to examination of the scalar
833through normal scalar means.
834
835=for apidoc AmxU|char *|PL_parser-E<gt>bufend
836
837Direct pointer to the end of the chunk of text currently being lexed, the
838end of the lexer buffer. This is equal to C<SvPVX(PL_parser-E<gt>linestr)
6602b933 839+ SvCUR(PL_parser-E<gt>linestr)>. A C<NUL> character (zero octet) is
f0e67a1d
Z
840always located at the end of the buffer, and does not count as part of
841the buffer's contents.
842
843=for apidoc AmxU|char *|PL_parser-E<gt>bufptr
844
845Points to the current position of lexing inside the lexer buffer.
846Characters around this point may be freely examined, within
847the range delimited by C<SvPVX(L</PL_parser-E<gt>linestr>)> and
848L</PL_parser-E<gt>bufend>. The octets of the buffer may be intended to be
849interpreted as either UTF-8 or Latin-1, as indicated by L</lex_bufutf8>.
850
851Lexing code (whether in the Perl core or not) moves this pointer past
852the characters that it consumes. It is also expected to perform some
853bookkeeping whenever a newline character is consumed. This movement
854can be more conveniently performed by the function L</lex_read_to>,
855which handles newlines appropriately.
856
857Interpretation of the buffer's octets can be abstracted out by
858using the slightly higher-level functions L</lex_peek_unichar> and
859L</lex_read_unichar>.
860
861=for apidoc AmxU|char *|PL_parser-E<gt>linestart
862
863Points to the start of the current line inside the lexer buffer.
864This is useful for indicating at which column an error occurred, and
865not much else. This must be updated by any lexing code that consumes
866a newline; the function L</lex_read_to> handles this detail.
867
868=cut
869*/
870
871/*
872=for apidoc Amx|bool|lex_bufutf8
873
874Indicates whether the octets in the lexer buffer
875(L</PL_parser-E<gt>linestr>) should be interpreted as the UTF-8 encoding
876of Unicode characters. If not, they should be interpreted as Latin-1
877characters. This is analogous to the C<SvUTF8> flag for scalars.
878
879In UTF-8 mode, it is not guaranteed that the lexer buffer actually
880contains valid UTF-8. Lexing code must be robust in the face of invalid
881encoding.
882
883The actual C<SvUTF8> flag of the L</PL_parser-E<gt>linestr> scalar
884is significant, but not the whole story regarding the input character
885encoding. Normally, when a file is being read, the scalar contains octets
886and its C<SvUTF8> flag is off, but the octets should be interpreted as
887UTF-8 if the C<use utf8> pragma is in effect. During a string eval,
888however, the scalar may have the C<SvUTF8> flag on, and in this case its
889octets should be interpreted as UTF-8 unless the C<use bytes> pragma
890is in effect. This logic may change in the future; use this function
891instead of implementing the logic yourself.
892
893=cut
894*/
895
896bool
897Perl_lex_bufutf8(pTHX)
898{
899 return UTF;
900}
901
902/*
903=for apidoc Amx|char *|lex_grow_linestr|STRLEN len
904
905Reallocates the lexer buffer (L</PL_parser-E<gt>linestr>) to accommodate
2d7f6611 906at least C<len> octets (including terminating C<NUL>). Returns a
f0e67a1d
Z
907pointer to the reallocated buffer. This is necessary before making
908any direct modification of the buffer that would increase its length.
909L</lex_stuff_pvn> provides a more convenient way to insert text into
910the buffer.
911
912Do not use C<SvGROW> or C<sv_grow> directly on C<PL_parser-E<gt>linestr>;
913this function updates all of the lexer's variables that point directly
914into the buffer.
915
916=cut
917*/
918
919char *
920Perl_lex_grow_linestr(pTHX_ STRLEN len)
921{
922 SV *linestr;
923 char *buf;
924 STRLEN bufend_pos, bufptr_pos, oldbufptr_pos, oldoldbufptr_pos;
c7641931 925 STRLEN linestart_pos, last_uni_pos, last_lop_pos, re_eval_start_pos;
98d5e3ef
DM
926 bool current;
927
f0e67a1d
Z
928 linestr = PL_parser->linestr;
929 buf = SvPVX(linestr);
930 if (len <= SvLEN(linestr))
931 return buf;
98d5e3ef
DM
932
933 /* Is the lex_shared linestr SV the same as the current linestr SV?
934 * Only in this case does re_eval_start need adjusting, since it
935 * points within lex_shared->ls_linestr's buffer */
936 current = (linestr == PL_parser->lex_shared->ls_linestr);
937
f0e67a1d
Z
938 bufend_pos = PL_parser->bufend - buf;
939 bufptr_pos = PL_parser->bufptr - buf;
940 oldbufptr_pos = PL_parser->oldbufptr - buf;
941 oldoldbufptr_pos = PL_parser->oldoldbufptr - buf;
942 linestart_pos = PL_parser->linestart - buf;
943 last_uni_pos = PL_parser->last_uni ? PL_parser->last_uni - buf : 0;
944 last_lop_pos = PL_parser->last_lop ? PL_parser->last_lop - buf : 0;
98d5e3ef 945 re_eval_start_pos = (current && PL_parser->lex_shared->re_eval_start) ?
3328ab5a 946 PL_parser->lex_shared->re_eval_start - buf : 0;
c7641931 947
f0e67a1d 948 buf = sv_grow(linestr, len);
c7641931 949
f0e67a1d
Z
950 PL_parser->bufend = buf + bufend_pos;
951 PL_parser->bufptr = buf + bufptr_pos;
952 PL_parser->oldbufptr = buf + oldbufptr_pos;
953 PL_parser->oldoldbufptr = buf + oldoldbufptr_pos;
954 PL_parser->linestart = buf + linestart_pos;
955 if (PL_parser->last_uni)
956 PL_parser->last_uni = buf + last_uni_pos;
957 if (PL_parser->last_lop)
958 PL_parser->last_lop = buf + last_lop_pos;
98d5e3ef 959 if (current && PL_parser->lex_shared->re_eval_start)
3328ab5a 960 PL_parser->lex_shared->re_eval_start = buf + re_eval_start_pos;
f0e67a1d
Z
961 return buf;
962}
963
964/*
83aa740e 965=for apidoc Amx|void|lex_stuff_pvn|const char *pv|STRLEN len|U32 flags
f0e67a1d
Z
966
967Insert characters into the lexer buffer (L</PL_parser-E<gt>linestr>),
968immediately after the current lexing point (L</PL_parser-E<gt>bufptr>),
969reallocating the buffer if necessary. This means that lexing code that
970runs later will see the characters as if they had appeared in the input.
971It is not recommended to do this as part of normal parsing, and most
972uses of this facility run the risk of the inserted characters being
973interpreted in an unintended manner.
974
2d7f6611
KW
975The string to be inserted is represented by C<len> octets starting
976at C<pv>. These octets are interpreted as either UTF-8 or Latin-1,
977according to whether the C<LEX_STUFF_UTF8> flag is set in C<flags>.
f0e67a1d
Z
978The characters are recoded for the lexer buffer, according to how the
979buffer is currently being interpreted (L</lex_bufutf8>). If a string
9dcc53ea 980to be inserted is available as a Perl scalar, the L</lex_stuff_sv>
f0e67a1d
Z
981function is more convenient.
982
983=cut
984*/
985
986void
83aa740e 987Perl_lex_stuff_pvn(pTHX_ const char *pv, STRLEN len, U32 flags)
f0e67a1d 988{
749123ff 989 dVAR;
f0e67a1d
Z
990 char *bufptr;
991 PERL_ARGS_ASSERT_LEX_STUFF_PVN;
992 if (flags & ~(LEX_STUFF_UTF8))
993 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_stuff_pvn");
994 if (UTF) {
995 if (flags & LEX_STUFF_UTF8) {
996 goto plain_copy;
997 } else {
54d004e8 998 STRLEN highhalf = 0; /* Count of variants */
83aa740e 999 const char *p, *e = pv+len;
54d004e8
KW
1000 for (p = pv; p != e; p++) {
1001 if (! UTF8_IS_INVARIANT(*p)) {
1002 highhalf++;
1003 }
1004 }
f0e67a1d
Z
1005 if (!highhalf)
1006 goto plain_copy;
1007 lex_grow_linestr(SvCUR(PL_parser->linestr)+1+len+highhalf);
1008 bufptr = PL_parser->bufptr;
1009 Move(bufptr, bufptr+len+highhalf, PL_parser->bufend+1-bufptr, char);
255fdf19
Z
1010 SvCUR_set(PL_parser->linestr,
1011 SvCUR(PL_parser->linestr) + len+highhalf);
f0e67a1d
Z
1012 PL_parser->bufend += len+highhalf;
1013 for (p = pv; p != e; p++) {
1014 U8 c = (U8)*p;
54d004e8
KW
1015 if (! UTF8_IS_INVARIANT(c)) {
1016 *bufptr++ = UTF8_TWO_BYTE_HI(c);
1017 *bufptr++ = UTF8_TWO_BYTE_LO(c);
f0e67a1d
Z
1018 } else {
1019 *bufptr++ = (char)c;
1020 }
1021 }
1022 }
1023 } else {
1024 if (flags & LEX_STUFF_UTF8) {
1025 STRLEN highhalf = 0;
83aa740e 1026 const char *p, *e = pv+len;
f0e67a1d
Z
1027 for (p = pv; p != e; p++) {
1028 U8 c = (U8)*p;
54d004e8 1029 if (UTF8_IS_ABOVE_LATIN1(c)) {
f0e67a1d
Z
1030 Perl_croak(aTHX_ "Lexing code attempted to stuff "
1031 "non-Latin-1 character into Latin-1 input");
54d004e8 1032 } else if (UTF8_IS_NEXT_CHAR_DOWNGRADEABLE(p, e)) {
f0e67a1d
Z
1033 p++;
1034 highhalf++;
75219bac
KW
1035 } else if (! UTF8_IS_INVARIANT(c)) {
1036 _force_out_malformed_utf8_message((U8 *) p, (U8 *) e,
1037 0,
1038 1 /* 1 means die */ );
1039 NOT_REACHED; /* NOTREACHED */
f0e67a1d
Z
1040 }
1041 }
1042 if (!highhalf)
1043 goto plain_copy;
1044 lex_grow_linestr(SvCUR(PL_parser->linestr)+1+len-highhalf);
1045 bufptr = PL_parser->bufptr;
1046 Move(bufptr, bufptr+len-highhalf, PL_parser->bufend+1-bufptr, char);
255fdf19
Z
1047 SvCUR_set(PL_parser->linestr,
1048 SvCUR(PL_parser->linestr) + len-highhalf);
f0e67a1d 1049 PL_parser->bufend += len-highhalf;
54d004e8
KW
1050 p = pv;
1051 while (p < e) {
1052 if (UTF8_IS_INVARIANT(*p)) {
1053 *bufptr++ = *p;
1054 p++;
f0e67a1d 1055 }
54d004e8
KW
1056 else {
1057 assert(p < e -1 );
a62b247b 1058 *bufptr++ = EIGHT_BIT_UTF8_TO_NATIVE(*p, *(p+1));
54d004e8
KW
1059 p += 2;
1060 }
f0e67a1d
Z
1061 }
1062 } else {
54d004e8 1063 plain_copy:
f0e67a1d
Z
1064 lex_grow_linestr(SvCUR(PL_parser->linestr)+1+len);
1065 bufptr = PL_parser->bufptr;
1066 Move(bufptr, bufptr+len, PL_parser->bufend+1-bufptr, char);
255fdf19 1067 SvCUR_set(PL_parser->linestr, SvCUR(PL_parser->linestr) + len);
f0e67a1d
Z
1068 PL_parser->bufend += len;
1069 Copy(pv, bufptr, len, char);
1070 }
1071 }
1072}
1073
1074/*
9dcc53ea
Z
1075=for apidoc Amx|void|lex_stuff_pv|const char *pv|U32 flags
1076
1077Insert characters into the lexer buffer (L</PL_parser-E<gt>linestr>),
1078immediately after the current lexing point (L</PL_parser-E<gt>bufptr>),
1079reallocating the buffer if necessary. This means that lexing code that
1080runs later will see the characters as if they had appeared in the input.
1081It is not recommended to do this as part of normal parsing, and most
1082uses of this facility run the risk of the inserted characters being
1083interpreted in an unintended manner.
1084
2d7f6611 1085The string to be inserted is represented by octets starting at C<pv>
9dcc53ea
Z
1086and continuing to the first nul. These octets are interpreted as either
1087UTF-8 or Latin-1, according to whether the C<LEX_STUFF_UTF8> flag is set
2d7f6611 1088in C<flags>. The characters are recoded for the lexer buffer, according
9dcc53ea
Z
1089to how the buffer is currently being interpreted (L</lex_bufutf8>).
1090If it is not convenient to nul-terminate a string to be inserted, the
1091L</lex_stuff_pvn> function is more appropriate.
1092
1093=cut
1094*/
1095
1096void
1097Perl_lex_stuff_pv(pTHX_ const char *pv, U32 flags)
1098{
1099 PERL_ARGS_ASSERT_LEX_STUFF_PV;
1100 lex_stuff_pvn(pv, strlen(pv), flags);
1101}
1102
1103/*
f0e67a1d
Z
1104=for apidoc Amx|void|lex_stuff_sv|SV *sv|U32 flags
1105
1106Insert characters into the lexer buffer (L</PL_parser-E<gt>linestr>),
1107immediately after the current lexing point (L</PL_parser-E<gt>bufptr>),
1108reallocating the buffer if necessary. This means that lexing code that
1109runs later will see the characters as if they had appeared in the input.
1110It is not recommended to do this as part of normal parsing, and most
1111uses of this facility run the risk of the inserted characters being
1112interpreted in an unintended manner.
1113
2d7f6611 1114The string to be inserted is the string value of C<sv>. The characters
f0e67a1d 1115are recoded for the lexer buffer, according to how the buffer is currently
9dcc53ea 1116being interpreted (L</lex_bufutf8>). If a string to be inserted is
f0e67a1d
Z
1117not already a Perl scalar, the L</lex_stuff_pvn> function avoids the
1118need to construct a scalar.
1119
1120=cut
1121*/
1122
1123void
1124Perl_lex_stuff_sv(pTHX_ SV *sv, U32 flags)
1125{
1126 char *pv;
1127 STRLEN len;
1128 PERL_ARGS_ASSERT_LEX_STUFF_SV;
1129 if (flags)
1130 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_stuff_sv");
1131 pv = SvPV(sv, len);
1132 lex_stuff_pvn(pv, len, flags | (SvUTF8(sv) ? LEX_STUFF_UTF8 : 0));
1133}
1134
1135/*
1136=for apidoc Amx|void|lex_unstuff|char *ptr
1137
1138Discards text about to be lexed, from L</PL_parser-E<gt>bufptr> up to
2d7f6611 1139C<ptr>. Text following C<ptr> will be moved, and the buffer shortened.
f0e67a1d
Z
1140This hides the discarded text from any lexing code that runs later,
1141as if the text had never appeared.
1142
1143This is not the normal way to consume lexed text. For that, use
1144L</lex_read_to>.
1145
1146=cut
1147*/
1148
1149void
1150Perl_lex_unstuff(pTHX_ char *ptr)
1151{
1152 char *buf, *bufend;
1153 STRLEN unstuff_len;
1154 PERL_ARGS_ASSERT_LEX_UNSTUFF;
1155 buf = PL_parser->bufptr;
1156 if (ptr < buf)
1157 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_unstuff");
1158 if (ptr == buf)
1159 return;
1160 bufend = PL_parser->bufend;
1161 if (ptr > bufend)
1162 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_unstuff");
1163 unstuff_len = ptr - buf;
1164 Move(ptr, buf, bufend+1-ptr, char);
1165 SvCUR_set(PL_parser->linestr, SvCUR(PL_parser->linestr) - unstuff_len);
1166 PL_parser->bufend = bufend - unstuff_len;
1167}
1168
1169/*
1170=for apidoc Amx|void|lex_read_to|char *ptr
1171
1172Consume text in the lexer buffer, from L</PL_parser-E<gt>bufptr> up
2d7f6611 1173to C<ptr>. This advances L</PL_parser-E<gt>bufptr> to match C<ptr>,
f0e67a1d
Z
1174performing the correct bookkeeping whenever a newline character is passed.
1175This is the normal way to consume lexed text.
1176
1177Interpretation of the buffer's octets can be abstracted out by
1178using the slightly higher-level functions L</lex_peek_unichar> and
1179L</lex_read_unichar>.
1180
1181=cut
1182*/
1183
1184void
1185Perl_lex_read_to(pTHX_ char *ptr)
1186{
1187 char *s;
1188 PERL_ARGS_ASSERT_LEX_READ_TO;
1189 s = PL_parser->bufptr;
1190 if (ptr < s || ptr > PL_parser->bufend)
1191 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_read_to");
1192 for (; s != ptr; s++)
1193 if (*s == '\n') {
83944c01 1194 COPLINE_INC_WITH_HERELINES;
f0e67a1d
Z
1195 PL_parser->linestart = s+1;
1196 }
1197 PL_parser->bufptr = ptr;
1198}
1199
1200/*
1201=for apidoc Amx|void|lex_discard_to|char *ptr
1202
1203Discards the first part of the L</PL_parser-E<gt>linestr> buffer,
2d7f6611
KW
1204up to C<ptr>. The remaining content of the buffer will be moved, and
1205all pointers into the buffer updated appropriately. C<ptr> must not
f0e67a1d
Z
1206be later in the buffer than the position of L</PL_parser-E<gt>bufptr>:
1207it is not permitted to discard text that has yet to be lexed.
1208
1209Normally it is not necessarily to do this directly, because it suffices to
1210use the implicit discarding behaviour of L</lex_next_chunk> and things
1211based on it. However, if a token stretches across multiple lines,
1f317c95 1212and the lexing code has kept multiple lines of text in the buffer for
f0e67a1d
Z
1213that purpose, then after completion of the token it would be wise to
1214explicitly discard the now-unneeded earlier lines, to avoid future
1215multi-line tokens growing the buffer without bound.
1216
1217=cut
1218*/
1219
1220void
1221Perl_lex_discard_to(pTHX_ char *ptr)
1222{
1223 char *buf;
1224 STRLEN discard_len;
1225 PERL_ARGS_ASSERT_LEX_DISCARD_TO;
1226 buf = SvPVX(PL_parser->linestr);
1227 if (ptr < buf)
1228 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_discard_to");
1229 if (ptr == buf)
1230 return;
1231 if (ptr > PL_parser->bufptr)
1232 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_discard_to");
1233 discard_len = ptr - buf;
1234 if (PL_parser->oldbufptr < ptr)
1235 PL_parser->oldbufptr = ptr;
1236 if (PL_parser->oldoldbufptr < ptr)
1237 PL_parser->oldoldbufptr = ptr;
1238 if (PL_parser->last_uni && PL_parser->last_uni < ptr)
1239 PL_parser->last_uni = NULL;
1240 if (PL_parser->last_lop && PL_parser->last_lop < ptr)
1241 PL_parser->last_lop = NULL;
1242 Move(ptr, buf, PL_parser->bufend+1-ptr, char);
1243 SvCUR_set(PL_parser->linestr, SvCUR(PL_parser->linestr) - discard_len);
1244 PL_parser->bufend -= discard_len;
1245 PL_parser->bufptr -= discard_len;
1246 PL_parser->oldbufptr -= discard_len;
1247 PL_parser->oldoldbufptr -= discard_len;
1248 if (PL_parser->last_uni)
1249 PL_parser->last_uni -= discard_len;
1250 if (PL_parser->last_lop)
1251 PL_parser->last_lop -= discard_len;
1252}
1253
1254/*
1255=for apidoc Amx|bool|lex_next_chunk|U32 flags
1256
1257Reads in the next chunk of text to be lexed, appending it to
1258L</PL_parser-E<gt>linestr>. This should be called when lexing code has
1259looked to the end of the current chunk and wants to know more. It is
1260usual, but not necessary, for lexing to have consumed the entirety of
1261the current chunk at this time.
1262
1263If L</PL_parser-E<gt>bufptr> is pointing to the very end of the current
1264chunk (i.e., the current chunk has been entirely consumed), normally the
1265current chunk will be discarded at the same time that the new chunk is
c5608a1f 1266read in. If C<flags> has the C<LEX_KEEP_PREVIOUS> bit set, the current chunk
f0e67a1d
Z
1267will not be discarded. If the current chunk has not been entirely
1268consumed, then it will not be discarded regardless of the flag.
1269
1270Returns true if some new text was added to the buffer, or false if the
1271buffer has reached the end of the input text.
1272
1273=cut
1274*/
1275
1276#define LEX_FAKE_EOF 0x80000000
e47d32dc 1277#define LEX_NO_TERM 0x40000000 /* here-doc */
f0e67a1d
Z
1278
1279bool
1280Perl_lex_next_chunk(pTHX_ U32 flags)
1281{
1282 SV *linestr;
1283 char *buf;
1284 STRLEN old_bufend_pos, new_bufend_pos;
1285 STRLEN bufptr_pos, oldbufptr_pos, oldoldbufptr_pos;
1286 STRLEN linestart_pos, last_uni_pos, last_lop_pos;
17cc9359 1287 bool got_some_for_debugger = 0;
f0e67a1d 1288 bool got_some;
112d1284 1289 if (flags & ~(LEX_KEEP_PREVIOUS|LEX_FAKE_EOF|LEX_NO_TERM))
f0e67a1d 1290 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_next_chunk");
d27f4b91 1291 if (!(flags & LEX_NO_TERM) && PL_lex_inwhat)
e47d32dc 1292 return FALSE;
f0e67a1d
Z
1293 linestr = PL_parser->linestr;
1294 buf = SvPVX(linestr);
407f8cf2
KW
1295 if (!(flags & LEX_KEEP_PREVIOUS)
1296 && PL_parser->bufptr == PL_parser->bufend)
1297 {
f0e67a1d
Z
1298 old_bufend_pos = bufptr_pos = oldbufptr_pos = oldoldbufptr_pos = 0;
1299 linestart_pos = 0;
1300 if (PL_parser->last_uni != PL_parser->bufend)
1301 PL_parser->last_uni = NULL;
1302 if (PL_parser->last_lop != PL_parser->bufend)
1303 PL_parser->last_lop = NULL;
1304 last_uni_pos = last_lop_pos = 0;
1305 *buf = 0;
1306 SvCUR(linestr) = 0;
1307 } else {
1308 old_bufend_pos = PL_parser->bufend - buf;
1309 bufptr_pos = PL_parser->bufptr - buf;
1310 oldbufptr_pos = PL_parser->oldbufptr - buf;
1311 oldoldbufptr_pos = PL_parser->oldoldbufptr - buf;
1312 linestart_pos = PL_parser->linestart - buf;
1313 last_uni_pos = PL_parser->last_uni ? PL_parser->last_uni - buf : 0;
1314 last_lop_pos = PL_parser->last_lop ? PL_parser->last_lop - buf : 0;
1315 }
1316 if (flags & LEX_FAKE_EOF) {
1317 goto eof;
60d63348 1318 } else if (!PL_parser->rsfp && !PL_parser->filtered) {
f0e67a1d
Z
1319 got_some = 0;
1320 } else if (filter_gets(linestr, old_bufend_pos)) {
1321 got_some = 1;
17cc9359 1322 got_some_for_debugger = 1;
112d1284
FC
1323 } else if (flags & LEX_NO_TERM) {
1324 got_some = 0;
f0e67a1d 1325 } else {
580561a3 1326 if (!SvPOK(linestr)) /* can get undefined by filter_gets */
847cc851 1327 SvPVCLEAR(linestr);
f0e67a1d
Z
1328 eof:
1329 /* End of real input. Close filehandle (unless it was STDIN),
1330 * then add implicit termination.
1331 */
87606032 1332 if (PL_parser->lex_flags & LEX_DONT_CLOSE_RSFP)
f0e67a1d
Z
1333 PerlIO_clearerr(PL_parser->rsfp);
1334 else if (PL_parser->rsfp)
1335 (void)PerlIO_close(PL_parser->rsfp);
1336 PL_parser->rsfp = NULL;
60d63348 1337 PL_parser->in_pod = PL_parser->filtered = 0;
f0e67a1d
Z
1338 if (!PL_in_eval && PL_minus_p) {
1339 sv_catpvs(linestr,
1340 /*{*/";}continue{print or die qq(-p destination: $!\\n);}");
1341 PL_minus_n = PL_minus_p = 0;
1342 } else if (!PL_in_eval && PL_minus_n) {
1343 sv_catpvs(linestr, /*{*/";}");
1344 PL_minus_n = 0;
1345 } else
1346 sv_catpvs(linestr, ";");
1347 got_some = 1;
1348 }
1349 buf = SvPVX(linestr);
1350 new_bufend_pos = SvCUR(linestr);
1351 PL_parser->bufend = buf + new_bufend_pos;
1352 PL_parser->bufptr = buf + bufptr_pos;
1353 PL_parser->oldbufptr = buf + oldbufptr_pos;
1354 PL_parser->oldoldbufptr = buf + oldoldbufptr_pos;
1355 PL_parser->linestart = buf + linestart_pos;
1356 if (PL_parser->last_uni)
1357 PL_parser->last_uni = buf + last_uni_pos;
1358 if (PL_parser->last_lop)
1359 PL_parser->last_lop = buf + last_lop_pos;
7f1c3e8c
FC
1360 if (PL_parser->preambling != NOLINE) {
1361 CopLINE_set(PL_curcop, PL_parser->preambling + 1);
1362 PL_parser->preambling = NOLINE;
1363 }
407f8cf2
KW
1364 if ( got_some_for_debugger
1365 && PERLDB_LINE_OR_SAVESRC
1366 && PL_curstash != PL_debstash)
1367 {
f0e67a1d
Z
1368 /* debugger active and we're not compiling the debugger code,
1369 * so store the line into the debugger's array of lines
1370 */
1371 update_debugger_info(NULL, buf+old_bufend_pos,
1372 new_bufend_pos-old_bufend_pos);
1373 }
1374 return got_some;
1375}
1376
1377/*
1378=for apidoc Amx|I32|lex_peek_unichar|U32 flags
1379
1380Looks ahead one (Unicode) character in the text currently being lexed.
1381Returns the codepoint (unsigned integer value) of the next character,
1382or -1 if lexing has reached the end of the input text. To consume the
1383peeked character, use L</lex_read_unichar>.
1384
1385If the next character is in (or extends into) the next chunk of input
1386text, the next chunk will be read in. Normally the current chunk will be
c5608a1f
KW
1387discarded at the same time, but if C<flags> has the C<LEX_KEEP_PREVIOUS>
1388bit set, then the current chunk will not be discarded.
f0e67a1d
Z
1389
1390If the input is being interpreted as UTF-8 and a UTF-8 encoding error
1391is encountered, an exception is generated.
1392
1393=cut
1394*/
1395
1396I32
1397Perl_lex_peek_unichar(pTHX_ U32 flags)
1398{
749123ff 1399 dVAR;
f0e67a1d
Z
1400 char *s, *bufend;
1401 if (flags & ~(LEX_KEEP_PREVIOUS))
1402 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_peek_unichar");
1403 s = PL_parser->bufptr;
1404 bufend = PL_parser->bufend;
1405 if (UTF) {
1406 U8 head;
1407 I32 unichar;
1408 STRLEN len, retlen;
1409 if (s == bufend) {
1410 if (!lex_next_chunk(flags))
1411 return -1;
1412 s = PL_parser->bufptr;
1413 bufend = PL_parser->bufend;
1414 }
1415 head = (U8)*s;
54d004e8 1416 if (UTF8_IS_INVARIANT(head))
f0e67a1d 1417 return head;
54d004e8
KW
1418 if (UTF8_IS_START(head)) {
1419 len = UTF8SKIP(&head);
f0e67a1d
Z
1420 while ((STRLEN)(bufend-s) < len) {
1421 if (!lex_next_chunk(flags | LEX_KEEP_PREVIOUS))
1422 break;
1423 s = PL_parser->bufptr;
1424 bufend = PL_parser->bufend;
1425 }
1426 }
c80e42f3 1427 unichar = utf8n_to_uvchr((U8*)s, bufend-s, &retlen, UTF8_CHECK_ONLY);
f0e67a1d 1428 if (retlen == (STRLEN)-1) {
75219bac
KW
1429 _force_out_malformed_utf8_message((U8 *) s,
1430 (U8 *) bufend,
1431 0,
1432 1 /* 1 means die */ );
1433 NOT_REACHED; /* NOTREACHED */
f0e67a1d
Z
1434 }
1435 return unichar;
1436 } else {
1437 if (s == bufend) {
1438 if (!lex_next_chunk(flags))
1439 return -1;
1440 s = PL_parser->bufptr;
1441 }
1442 return (U8)*s;
1443 }
1444}
1445
1446/*
1447=for apidoc Amx|I32|lex_read_unichar|U32 flags
1448
1449Reads the next (Unicode) character in the text currently being lexed.
1450Returns the codepoint (unsigned integer value) of the character read,
1451and moves L</PL_parser-E<gt>bufptr> past the character, or returns -1
1452if lexing has reached the end of the input text. To non-destructively
1453examine the next character, use L</lex_peek_unichar> instead.
1454
1455If the next character is in (or extends into) the next chunk of input
1456text, the next chunk will be read in. Normally the current chunk will be
c5608a1f
KW
1457discarded at the same time, but if C<flags> has the C<LEX_KEEP_PREVIOUS>
1458bit set, then the current chunk will not be discarded.
f0e67a1d
Z
1459
1460If the input is being interpreted as UTF-8 and a UTF-8 encoding error
1461is encountered, an exception is generated.
1462
1463=cut
1464*/
1465
1466I32
1467Perl_lex_read_unichar(pTHX_ U32 flags)
1468{
1469 I32 c;
1470 if (flags & ~(LEX_KEEP_PREVIOUS))
1471 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_read_unichar");
1472 c = lex_peek_unichar(flags);
1473 if (c != -1) {
1474 if (c == '\n')
83944c01 1475 COPLINE_INC_WITH_HERELINES;
d9018cbe
EB
1476 if (UTF)
1477 PL_parser->bufptr += UTF8SKIP(PL_parser->bufptr);
1478 else
1479 ++(PL_parser->bufptr);
f0e67a1d
Z
1480 }
1481 return c;
1482}
1483
1484/*
1485=for apidoc Amx|void|lex_read_space|U32 flags
1486
1487Reads optional spaces, in Perl style, in the text currently being
1488lexed. The spaces may include ordinary whitespace characters and
1489Perl-style comments. C<#line> directives are processed if encountered.
1490L</PL_parser-E<gt>bufptr> is moved past the spaces, so that it points
1491at a non-space character (or the end of the input text).
1492
1493If spaces extend into the next chunk of input text, the next chunk will
1494be read in. Normally the current chunk will be discarded at the same
c5608a1f 1495time, but if C<flags> has the C<LEX_KEEP_PREVIOUS> bit set, then the current
f0e67a1d
Z
1496chunk will not be discarded.
1497
1498=cut
1499*/
1500
21791330 1501#define LEX_NO_INCLINE 0x40000000
f0998909
Z
1502#define LEX_NO_NEXT_CHUNK 0x80000000
1503
f0e67a1d
Z
1504void
1505Perl_lex_read_space(pTHX_ U32 flags)
1506{
1507 char *s, *bufend;
21791330 1508 const bool can_incline = !(flags & LEX_NO_INCLINE);
f0e67a1d 1509 bool need_incline = 0;
21791330 1510 if (flags & ~(LEX_KEEP_PREVIOUS|LEX_NO_NEXT_CHUNK|LEX_NO_INCLINE))
f0e67a1d 1511 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_read_space");
f0e67a1d
Z
1512 s = PL_parser->bufptr;
1513 bufend = PL_parser->bufend;
1514 while (1) {
1515 char c = *s;
1516 if (c == '#') {
1517 do {
1518 c = *++s;
1519 } while (!(c == '\n' || (c == 0 && s == bufend)));
1520 } else if (c == '\n') {
1521 s++;
21791330
FC
1522 if (can_incline) {
1523 PL_parser->linestart = s;
1524 if (s == bufend)
1525 need_incline = 1;
1526 else
1527 incline(s);
1528 }
f0e67a1d
Z
1529 } else if (isSPACE(c)) {
1530 s++;
1531 } else if (c == 0 && s == bufend) {
1532 bool got_more;
65c68e17 1533 line_t l;
f0998909
Z
1534 if (flags & LEX_NO_NEXT_CHUNK)
1535 break;
f0e67a1d 1536 PL_parser->bufptr = s;
65c68e17 1537 l = CopLINE(PL_curcop);
851b527a 1538 CopLINE(PL_curcop) += PL_parser->herelines + 1;
f0e67a1d 1539 got_more = lex_next_chunk(flags);
65c68e17 1540 CopLINE_set(PL_curcop, l);
f0e67a1d
Z
1541 s = PL_parser->bufptr;
1542 bufend = PL_parser->bufend;
1543 if (!got_more)
1544 break;
21791330 1545 if (can_incline && need_incline && PL_parser->rsfp) {
f0e67a1d
Z
1546 incline(s);
1547 need_incline = 0;
1548 }
3c47da3c
FC
1549 } else if (!c) {
1550 s++;
f0e67a1d
Z
1551 } else {
1552 break;
1553 }
1554 }
f0e67a1d
Z
1555 PL_parser->bufptr = s;
1556}
1557
1558/*
fe788d6b
PM
1559
1560=for apidoc EXMp|bool|validate_proto|SV *name|SV *proto|bool warn
1561
1562This function performs syntax checking on a prototype, C<proto>.
1563If C<warn> is true, any illegal characters or mismatched brackets
1564will trigger illegalproto warnings, declaring that they were
1565detected in the prototype for C<name>.
1566
1567The return value is C<true> if this is a valid prototype, and
1568C<false> if it is not, regardless of whether C<warn> was C<true> or
1569C<false>.
1570
1571Note that C<NULL> is a valid C<proto> and will always return C<true>.
1572
1573=cut
1574
1575 */
1576
1577bool
1578Perl_validate_proto(pTHX_ SV *name, SV *proto, bool warn)
1579{
1580 STRLEN len, origlen;
11327fa1 1581 char *p;
fe788d6b
PM
1582 bool bad_proto = FALSE;
1583 bool in_brackets = FALSE;
1584 bool after_slash = FALSE;
1585 char greedy_proto = ' ';
1586 bool proto_after_greedy_proto = FALSE;
1587 bool must_be_last = FALSE;
1588 bool underscore = FALSE;
f791a21a 1589 bool bad_proto_after_underscore = FALSE;
fe788d6b
PM
1590
1591 PERL_ARGS_ASSERT_VALIDATE_PROTO;
1592
1593 if (!proto)
1594 return TRUE;
1595
11327fa1 1596 p = SvPV(proto, len);
fe788d6b
PM
1597 origlen = len;
1598 for (; len--; p++) {
1599 if (!isSPACE(*p)) {
1600 if (must_be_last)
1601 proto_after_greedy_proto = TRUE;
f791a21a
PM
1602 if (underscore) {
1603 if (!strchr(";@%", *p))
1604 bad_proto_after_underscore = TRUE;
1605 underscore = FALSE;
1606 }
fe788d6b
PM
1607 if (!strchr("$@%*;[]&\\_+", *p) || *p == '\0') {
1608 bad_proto = TRUE;
1609 }
1610 else {
fe788d6b
PM
1611 if (*p == '[')
1612 in_brackets = TRUE;
1613 else if (*p == ']')
1614 in_brackets = FALSE;
407f8cf2
KW
1615 else if ((*p == '@' || *p == '%')
1616 && !after_slash
1617 && !in_brackets )
1618 {
fe788d6b
PM
1619 must_be_last = TRUE;
1620 greedy_proto = *p;
1621 }
1622 else if (*p == '_')
f791a21a 1623 underscore = TRUE;
fe788d6b
PM
1624 }
1625 if (*p == '\\')
1626 after_slash = TRUE;
1627 else
1628 after_slash = FALSE;
1629 }
1630 }
1631
1632 if (warn) {
b54d603d 1633 SV *tmpsv = newSVpvs_flags("", SVs_TEMP);
fe788d6b 1634 p -= origlen;
b54d603d
PM
1635 p = SvUTF8(proto)
1636 ? sv_uni_display(tmpsv, newSVpvn_flags(p, origlen, SVs_TEMP | SVf_UTF8),
1637 origlen, UNI_DISPLAY_ISPRINT)
1638 : pv_pretty(tmpsv, p, origlen, 60, NULL, NULL, PERL_PV_ESCAPE_NONASCII);
1639
fe788d6b
PM
1640 if (proto_after_greedy_proto)
1641 Perl_warner(aTHX_ packWARN(WARN_ILLEGALPROTO),
147e3846 1642 "Prototype after '%c' for %" SVf " : %s",
fe788d6b 1643 greedy_proto, SVfARG(name), p);
50278ed0
PM
1644 if (in_brackets)
1645 Perl_warner(aTHX_ packWARN(WARN_ILLEGALPROTO),
147e3846 1646 "Missing ']' in prototype for %" SVf " : %s",
50278ed0 1647 SVfARG(name), p);
b54d603d 1648 if (bad_proto)
fe788d6b 1649 Perl_warner(aTHX_ packWARN(WARN_ILLEGALPROTO),
147e3846 1650 "Illegal character in prototype for %" SVf " : %s",
f791a21a
PM
1651 SVfARG(name), p);
1652 if (bad_proto_after_underscore)
1653 Perl_warner(aTHX_ packWARN(WARN_ILLEGALPROTO),
147e3846 1654 "Illegal character after '_' in prototype for %" SVf " : %s",
f791a21a 1655 SVfARG(name), p);
fe788d6b
PM
1656 }
1657
1658 return (! (proto_after_greedy_proto || bad_proto) );
1659}
1660
1661/*
ffb4593c
NT
1662 * S_incline
1663 * This subroutine has nothing to do with tilting, whether at windmills
1664 * or pinball tables. Its name is short for "increment line". It
57843af0 1665 * increments the current line number in CopLINE(PL_curcop) and checks
ffb4593c 1666 * to see whether the line starts with a comment of the form
9cbb5ea2
GS
1667 * # line 500 "foo.pm"
1668 * If so, it sets the current line number and file to the values in the comment.
ffb4593c
NT
1669 */
1670
76e3520e 1671STATIC void
d9095cec 1672S_incline(pTHX_ const char *s)
463ee0b2 1673{
d9095cec
NC
1674 const char *t;
1675 const char *n;
1676 const char *e;
8818d409 1677 line_t line_num;
22ff3130 1678 UV uv;
463ee0b2 1679
7918f24d
NC
1680 PERL_ARGS_ASSERT_INCLINE;
1681
83944c01 1682 COPLINE_INC_WITH_HERELINES;
451f421f
FC
1683 if (!PL_rsfp && !PL_parser->filtered && PL_lex_state == LEX_NORMAL
1684 && s+1 == PL_bufend && *s == ';') {
1685 /* fake newline in string eval */
1686 CopLINE_dec(PL_curcop);
1687 return;
1688 }
463ee0b2
LW
1689 if (*s++ != '#')
1690 return;
d4c19fe8
AL
1691 while (SPACE_OR_TAB(*s))
1692 s++;
847cc851 1693 if (strEQs(s, "line"))
73659bf1
GS
1694 s += 4;
1695 else
1696 return;
084592ab 1697 if (SPACE_OR_TAB(*s))
73659bf1 1698 s++;
4e553d73 1699 else
73659bf1 1700 return;
d4c19fe8
AL
1701 while (SPACE_OR_TAB(*s))
1702 s++;
463ee0b2
LW
1703 if (!isDIGIT(*s))
1704 return;
d4c19fe8 1705
463ee0b2
LW
1706 n = s;
1707 while (isDIGIT(*s))
1708 s++;
07714eb4 1709 if (!SPACE_OR_TAB(*s) && *s != '\r' && *s != '\n' && *s != '\0')
26b6dc3f 1710 return;
bf4acbe4 1711 while (SPACE_OR_TAB(*s))
463ee0b2 1712 s++;
73659bf1 1713 if (*s == '"' && (t = strchr(s+1, '"'))) {
463ee0b2 1714 s++;
73659bf1
GS
1715 e = t + 1;
1716 }
463ee0b2 1717 else {
c35e046a 1718 t = s;
1bb1a3d6 1719 while (*t && !isSPACE(*t))
c35e046a 1720 t++;
73659bf1 1721 e = t;
463ee0b2 1722 }
bf4acbe4 1723 while (SPACE_OR_TAB(*e) || *e == '\r' || *e == '\f')
73659bf1
GS
1724 e++;
1725 if (*e != '\n' && *e != '\0')
1726 return; /* false alarm */
1727
22ff3130
HS
1728 if (!grok_atoUV(n, &uv, &e))
1729 return;
1730 line_num = ((line_t)uv) - 1;
8818d409 1731
f4dd75d9 1732 if (t - s > 0) {
d9095cec 1733 const STRLEN len = t - s;
3df32bda 1734
d36ee5be 1735 if (!PL_rsfp && !PL_parser->filtered) {
e66cf94c
RGS
1736 /* must copy *{"::_<(eval N)[oldfilename:L]"}
1737 * to *{"::_<newfilename"} */
44867030
NC
1738 /* However, the long form of evals is only turned on by the
1739 debugger - usually they're "(eval %lu)" */
d36ee5be
FC
1740 GV * const cfgv = CopFILEGV(PL_curcop);
1741 if (cfgv) {
38bd7ad8
FC
1742 char smallbuf[128];
1743 STRLEN tmplen2 = len;
44867030 1744 char *tmpbuf2;
449dd039 1745 GV *gv2;
44867030
NC
1746
1747 if (tmplen2 + 2 <= sizeof smallbuf)
1748 tmpbuf2 = smallbuf;
1749 else
1750 Newx(tmpbuf2, tmplen2 + 2, char);
1751
38bd7ad8
FC
1752 tmpbuf2[0] = '_';
1753 tmpbuf2[1] = '<';
44867030
NC
1754
1755 memcpy(tmpbuf2 + 2, s, tmplen2);
1756 tmplen2 += 2;
1757
8a5ee598 1758 gv2 = *(GV**)hv_fetch(PL_defstash, tmpbuf2, tmplen2, TRUE);
e5527e4b 1759 if (!isGV(gv2)) {
8a5ee598 1760 gv_init(gv2, PL_defstash, tmpbuf2, tmplen2, FALSE);
e5527e4b
RGS
1761 /* adjust ${"::_<newfilename"} to store the new file name */
1762 GvSV(gv2) = newSVpvn(tmpbuf2 + 2, tmplen2 - 2);
8818d409
FC
1763 /* The line number may differ. If that is the case,
1764 alias the saved lines that are in the array.
1765 Otherwise alias the whole array. */
1766 if (CopLINE(PL_curcop) == line_num) {
38bd7ad8
FC
1767 GvHV(gv2) = MUTABLE_HV(SvREFCNT_inc(GvHV(cfgv)));
1768 GvAV(gv2) = MUTABLE_AV(SvREFCNT_inc(GvAV(cfgv)));
8818d409 1769 }
38bd7ad8
FC
1770 else if (GvAV(cfgv)) {
1771 AV * const av = GvAV(cfgv);
8818d409
FC
1772 const I32 start = CopLINE(PL_curcop)+1;
1773 I32 items = AvFILLp(av) - start;
1774 if (items > 0) {
1775 AV * const av2 = GvAVn(gv2);
1776 SV **svp = AvARRAY(av) + start;
1777 I32 l = (I32)line_num+1;
1778 while (items--)
1779 av_store(av2, l++, SvREFCNT_inc(*svp++));
1780 }
1781 }
e5527e4b 1782 }
44867030
NC
1783
1784 if (tmpbuf2 != smallbuf) Safefree(tmpbuf2);
d36ee5be 1785 }
e66cf94c 1786 }
05ec9bb3 1787 CopFILE_free(PL_curcop);
449dd039 1788 CopFILE_setn(PL_curcop, s, len);
f4dd75d9 1789 }
8818d409 1790 CopLINE_set(PL_curcop, line_num);
463ee0b2
LW
1791}
1792
80a702cd 1793STATIC void
15f169a1 1794S_update_debugger_info(pTHX_ SV *orig_sv, const char *const buf, STRLEN len)
80a702cd
RGS
1795{
1796 AV *av = CopFILEAVx(PL_curcop);
1797 if (av) {
7f1c3e8c
FC
1798 SV * sv;
1799 if (PL_parser->preambling == NOLINE) sv = newSV_type(SVt_PVMG);
1800 else {
1801 sv = *av_fetch(av, 0, 1);
1802 SvUPGRADE(sv, SVt_PVMG);
1803 }
847cc851 1804 if (!SvPOK(sv)) SvPVCLEAR(sv);
5fa550fb 1805 if (orig_sv)
7f1c3e8c 1806 sv_catsv(sv, orig_sv);
5fa550fb 1807 else
7f1c3e8c
FC
1808 sv_catpvn(sv, buf, len);
1809 if (!SvIOK(sv)) {
1810 (void)SvIOK_on(sv);
1811 SvIV_set(sv, 0);
1812 }
1813 if (PL_parser->preambling == NOLINE)
1814 av_store(av, CopLINE(PL_curcop), sv);
80a702cd
RGS
1815 }
1816}
1817
ffb4593c 1818/*
8c6b0c7d 1819 * skipspace
ffb4593c
NT
1820 * Called to gobble the appropriate amount and type of whitespace.
1821 * Skips comments as well.
71fff7cb 1822 * Returns the next character after the whitespace that is skipped.
8c6b0c7d
FC
1823 *
1824 * peekspace
1825 * Same thing, but look ahead without incrementing line numbers or
1826 * adjusting PL_linestart.
ffb4593c
NT
1827 */
1828
3218e223 1829#define skipspace(s) skipspace_flags(s, 0)
8c6b0c7d 1830#define peekspace(s) skipspace_flags(s, LEX_NO_INCLINE)
3218e223 1831
76e3520e 1832STATIC char *
21791330 1833S_skipspace_flags(pTHX_ char *s, U32 flags)
a687059c 1834{
21791330 1835 PERL_ARGS_ASSERT_SKIPSPACE_FLAGS;
3280af22 1836 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
3c47da3c 1837 while (s < PL_bufend && (SPACE_OR_TAB(*s) || !*s))
463ee0b2 1838 s++;
f0e67a1d
Z
1839 } else {
1840 STRLEN bufptr_pos = PL_bufptr - SvPVX(PL_linestr);
1841 PL_bufptr = s;
21791330 1842 lex_read_space(flags | LEX_KEEP_PREVIOUS |
d27f4b91 1843 (PL_lex_inwhat || PL_lex_state == LEX_FORMLINE ?
f0998909 1844 LEX_NO_NEXT_CHUNK : 0));
3280af22 1845 s = PL_bufptr;
f0e67a1d
Z
1846 PL_bufptr = SvPVX(PL_linestr) + bufptr_pos;
1847 if (PL_linestart > PL_bufptr)
1848 PL_bufptr = PL_linestart;
1849 return s;
463ee0b2 1850 }
5db06880 1851 return s;
a687059c 1852}
378cc40b 1853
ffb4593c
NT
1854/*
1855 * S_check_uni
1856 * Check the unary operators to ensure there's no ambiguity in how they're
1857 * used. An ambiguous piece of code would be:
1858 * rand + 5
1859 * This doesn't mean rand() + 5. Because rand() is a unary operator,
1860 * the +5 is its argument.
1861 */
1862
76e3520e 1863STATIC void
cea2e8a9 1864S_check_uni(pTHX)
ba106d47 1865{
d4c19fe8
AL
1866 const char *s;
1867 const char *t;
2f3197b3 1868
3280af22 1869 if (PL_oldoldbufptr != PL_last_uni)
2f3197b3 1870 return;
3280af22
NIS
1871 while (isSPACE(*PL_last_uni))
1872 PL_last_uni++;
c35e046a 1873 s = PL_last_uni;
8a2bca12 1874 while (isWORDCHAR_lazy_if(s,UTF) || *s == '-')
8ce2ba82 1875 s += UTF ? UTF8SKIP(s) : 1;
3280af22 1876 if ((t = strchr(s, '(')) && t < PL_bufptr)
a0d0e21e 1877 return;
6136c704 1878
9b387841 1879 Perl_ck_warner_d(aTHX_ packWARN(WARN_AMBIGUOUS),
147e3846 1880 "Warning: Use of \"%" UTF8f "\" without parentheses is ambiguous",
b59c097b 1881 UTF8fARG(UTF, (int)(s - PL_last_uni), PL_last_uni));
2f3197b3
LW
1882}
1883
ffb4593c
NT
1884/*
1885 * LOP : macro to build a list operator. Its behaviour has been replaced
1886 * with a subroutine, S_lop() for which LOP is just another name.
1887 */
1888
a0d0e21e
LW
1889#define LOP(f,x) return lop(f,x,s)
1890
ffb4593c
NT
1891/*
1892 * S_lop
1893 * Build a list operator (or something that might be one). The rules:
41e8cbf4
FC
1894 * - if we have a next token, then it's a list operator (no parens) for
1895 * which the next token has already been parsed; e.g.,
1896 * sort foo @args
1897 * sort foo (@args)
ffb4593c
NT
1898 * - if the next thing is an opening paren, then it's a function
1899 * - else it's a list operator
1900 */
1901
76e3520e 1902STATIC I32
11288bb3 1903S_lop(pTHX_ I32 f, U8 x, char *s)
ffed7fef 1904{
7918f24d
NC
1905 PERL_ARGS_ASSERT_LOP;
1906
6154021b 1907 pl_yylval.ival = f;
35c8bce7 1908 CLINE;
3280af22
NIS
1909 PL_bufptr = s;
1910 PL_last_lop = PL_oldbufptr;
eb160463 1911 PL_last_lop_op = (OPCODE)f;
3280af22 1912 if (PL_nexttoke)
78cdf107 1913 goto lstop;
19f1898a 1914 PL_expect = x;
79072805 1915 if (*s == '(')
bbf60fe6 1916 return REPORT(FUNC);
294a536f 1917 s = skipspace(s);
79072805 1918 if (*s == '(')
bbf60fe6 1919 return REPORT(FUNC);
78cdf107
Z
1920 else {
1921 lstop:
1922 if (!PL_lex_allbrackets && PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC)
1923 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC;
bbf60fe6 1924 return REPORT(LSTOP);
78cdf107 1925 }
79072805
LW
1926}
1927
ffb4593c
NT
1928/*
1929 * S_force_next
9cbb5ea2 1930 * When the lexer realizes it knows the next token (for instance,
ffb4593c 1931 * it is reordering tokens for the parser) then it can call S_force_next
9cbb5ea2 1932 * to know what token to return the next time the lexer is called. Caller
b5bbe64a
JH
1933 * will need to set PL_nextval[] and possibly PL_expect to ensure
1934 * the lexer handles the token correctly.
ffb4593c
NT
1935 */
1936
4e553d73 1937STATIC void
cea2e8a9 1938S_force_next(pTHX_ I32 type)
79072805 1939{
704d4215
GG
1940#ifdef DEBUGGING
1941 if (DEBUG_T_TEST) {
1942 PerlIO_printf(Perl_debug_log, "### forced token:\n");
f05d7009 1943 tokereport(type, &NEXTVAL_NEXTTOKE);
704d4215
GG
1944 }
1945#endif
1f7c3e7c 1946 assert(PL_nexttoke < C_ARRAY_LENGTH(PL_nexttype));
3280af22
NIS
1947 PL_nexttype[PL_nexttoke] = type;
1948 PL_nexttoke++;
79072805
LW
1949}
1950
89f35911
FC
1951/*
1952 * S_postderef
1953 *
1954 * This subroutine handles postfix deref syntax after the arrow has already
1955 * been emitted. @* $* etc. are emitted as two separate token right here.
1956 * @[ @{ %[ %{ *{ are emitted also as two tokens, but this function emits
1957 * only the first, leaving yylex to find the next.
89f35911
FC
1958 */
1959
1960static int
ff25e5db 1961S_postderef(pTHX_ int const funny, char const next)
89f35911 1962{
ff25e5db 1963 assert(funny == DOLSHARP || strchr("$@%&*", funny));
89f35911
FC
1964 if (next == '*') {
1965 PL_expect = XOPERATOR;
cc624add 1966 if (PL_lex_state == LEX_INTERPNORMAL && !PL_lex_brackets) {
ff25e5db 1967 assert('@' == funny || '$' == funny || DOLSHARP == funny);
cc624add 1968 PL_lex_state = LEX_INTERPEND;
c3492809
FC
1969 if ('@' == funny)
1970 force_next(POSTJOIN);
cc624add 1971 }
89f35911
FC
1972 force_next(next);
1973 PL_bufptr+=2;
1974 }
1975 else {
760ca746
FC
1976 if ('@' == funny && PL_lex_state == LEX_INTERPNORMAL
1977 && !PL_lex_brackets)
1978 PL_lex_dojoin = 2;
89f35911
FC
1979 PL_expect = XOPERATOR;
1980 PL_bufptr++;
1981 }
1982 return funny;
1983}
1984
28ac2b49
Z
1985void
1986Perl_yyunlex(pTHX)
1987{
a7aaec61
Z
1988 int yyc = PL_parser->yychar;
1989 if (yyc != YYEMPTY) {
1990 if (yyc) {
a7aaec61
Z
1991 NEXTVAL_NEXTTOKE = PL_parser->yylval;
1992 if (yyc == '{'/*}*/ || yyc == HASHBRACK || yyc == '['/*]*/) {
78cdf107 1993 PL_lex_allbrackets--;
a7aaec61 1994 PL_lex_brackets--;
78cdf107
Z
1995 yyc |= (3<<24) | (PL_lex_brackstack[PL_lex_brackets] << 16);
1996 } else if (yyc == '('/*)*/) {
1997 PL_lex_allbrackets--;
1998 yyc |= (2<<24);
a7aaec61
Z
1999 }
2000 force_next(yyc);
2001 }
28ac2b49
Z
2002 PL_parser->yychar = YYEMPTY;
2003 }
2004}
2005
d0a148a6 2006STATIC SV *
15f169a1 2007S_newSV_maybe_utf8(pTHX_ const char *const start, STRLEN len)
d0a148a6 2008{
740cce10 2009 SV * const sv = newSVpvn_utf8(start, len,
c5f058df
KW
2010 !IN_BYTES
2011 && UTF
2012 && !is_utf8_invariant_string((const U8*)start, len)
2013 && is_utf8_string((const U8*)start, len));
d0a148a6
NC
2014 return sv;
2015}
2016
ffb4593c
NT
2017/*
2018 * S_force_word
2019 * When the lexer knows the next thing is a word (for instance, it has
2020 * just seen -> and it knows that the next char is a word char, then
02b34bbe
DM
2021 * it calls S_force_word to stick the next word into the PL_nexttoke/val
2022 * lookahead.
ffb4593c
NT
2023 *
2024 * Arguments:
b1b65b59 2025 * char *start : buffer position (must be within PL_linestr)
185c2e96
DM
2026 * int token : PL_next* will be this type of bare word
2027 * (e.g., METHOD,BAREWORD)
ffb4593c
NT
2028 * int check_keyword : if true, Perl checks to make sure the word isn't
2029 * a keyword (do this if the word is a label, e.g. goto FOO)
2030 * int allow_pack : if true, : characters will also be allowed (require,
2031 * use, etc. do this)
ffb4593c
NT
2032 */
2033
76e3520e 2034STATIC char *
345b3785 2035S_force_word(pTHX_ char *start, int token, int check_keyword, int allow_pack)
79072805 2036{
eb578fdb 2037 char *s;
463ee0b2 2038 STRLEN len;
4e553d73 2039
7918f24d
NC
2040 PERL_ARGS_ASSERT_FORCE_WORD;
2041
294a536f 2042 start = skipspace(start);
463ee0b2 2043 s = start;
407f8cf2 2044 if (isIDFIRST_lazy_if(s,UTF)
e7127e21 2045 || (allow_pack && *s == ':' && s[1] == ':') )
a0d0e21e 2046 {
3280af22 2047 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, allow_pack, &len);
01b5ef50
FC
2048 if (check_keyword) {
2049 char *s2 = PL_tokenbuf;
487e470d 2050 STRLEN len2 = len;
847cc851 2051 if (allow_pack && len > 6 && strEQs(s2, "CORE::"))
487e470d
FC
2052 s2 += 6, len2 -= 6;
2053 if (keyword(s2, len2, 0))
463ee0b2 2054 return start;
01b5ef50 2055 }
463ee0b2 2056 if (token == METHOD) {
294a536f 2057 s = skipspace(s);
463ee0b2 2058 if (*s == '(')
3280af22 2059 PL_expect = XTERM;
463ee0b2 2060 else {
3280af22 2061 PL_expect = XOPERATOR;
463ee0b2 2062 }
79072805 2063 }
9ded7720 2064 NEXTVAL_NEXTTOKE.opval
275103cd 2065 = newSVOP(OP_CONST,0,
d0a148a6 2066 S_newSV_maybe_utf8(aTHX_ PL_tokenbuf, len));
9ded7720 2067 NEXTVAL_NEXTTOKE.opval->op_private |= OPpCONST_BARE;
79072805
LW
2068 force_next(token);
2069 }
2070 return s;
2071}
2072
ffb4593c
NT
2073/*
2074 * S_force_ident
9cbb5ea2 2075 * Called when the lexer wants $foo *foo &foo etc, but the program
ffb4593c
NT
2076 * text only contains the "foo" portion. The first argument is a pointer
2077 * to the "foo", and the second argument is the type symbol to prefix.
185c2e96 2078 * Forces the next token to be a "BAREWORD".
9cbb5ea2 2079 * Creates the symbol if it didn't already exist (via gv_fetchpv()).
ffb4593c
NT
2080 */
2081
76e3520e 2082STATIC void
5aaab254 2083S_force_ident(pTHX_ const char *s, int kind)
79072805 2084{
7918f24d
NC
2085 PERL_ARGS_ASSERT_FORCE_IDENT;
2086
c9b48522
DD
2087 if (s[0]) {
2088 const STRLEN len = s[1] ? strlen(s) : 1; /* s = "\"" see yylex */
275103cd 2089 OP* const o = newSVOP(OP_CONST, 0, newSVpvn_flags(s, len,
728847b1 2090 UTF ? SVf_UTF8 : 0));
9ded7720 2091 NEXTVAL_NEXTTOKE.opval = o;
185c2e96 2092 force_next(BAREWORD);
748a9306 2093 if (kind) {
11343788 2094 o->op_private = OPpCONST_ENTERED;
55497cff
PP
2095 /* XXX see note in pp_entereval() for why we forgo typo
2096 warnings if the symbol must be introduced in an eval.
2097 GSAR 96-10-12 */
90e5519e 2098 gv_fetchpvn_flags(s, len,
4bff32c5 2099 (PL_in_eval ? GV_ADDMULTI
728847b1 2100 : GV_ADD) | ( UTF ? SVf_UTF8 : 0 ),
90e5519e
NC
2101 kind == '$' ? SVt_PV :
2102 kind == '@' ? SVt_PVAV :
2103 kind == '%' ? SVt_PVHV :
a0d0e21e 2104 SVt_PVGV
90e5519e 2105 );
748a9306 2106 }
79072805
LW
2107 }
2108}
2109
3f33d153
FC
2110static void
2111S_force_ident_maybe_lex(pTHX_ char pit)
2112{
3f33d153
FC
2113 NEXTVAL_NEXTTOKE.ival = pit;
2114 force_next('p');
2115}
2116
1571675a
GS
2117NV
2118Perl_str_to_version(pTHX_ SV *sv)
2119{
2120 NV retval = 0.0;
2121 NV nshift = 1.0;
2122 STRLEN len;
cfd0369c 2123 const char *start = SvPV_const(sv,len);
9d4ba2ae 2124 const char * const end = start + len;
504618e9 2125 const bool utf = SvUTF8(sv) ? TRUE : FALSE;
7918f24d
NC
2126
2127 PERL_ARGS_ASSERT_STR_TO_VERSION;
2128
1571675a 2129 while (start < end) {
ba210ebe 2130 STRLEN skip;
1571675a
GS
2131 UV n;
2132 if (utf)
9041c2e3 2133 n = utf8n_to_uvchr((U8*)start, len, &skip, 0);
1571675a
GS
2134 else {
2135 n = *(U8*)start;
2136 skip = 1;
2137 }
2138 retval += ((NV)n)/nshift;
2139 start += skip;
2140 nshift *= 1000;
2141 }
2142 return retval;
2143}
2144
4e553d73 2145/*
ffb4593c
NT
2146 * S_force_version
2147 * Forces the next token to be a version number.
e759cc13
RGS
2148 * If the next token appears to be an invalid version number, (e.g. "v2b"),
2149 * and if "guessing" is TRUE, then no new token is created (and the caller
2150 * must use an alternative parsing method).
ffb4593c
NT
2151 */
2152
76e3520e 2153STATIC char *
e759cc13 2154S_force_version(pTHX_ char *s, int guessing)
89bfa8cd 2155{
5f66b61c 2156 OP *version = NULL;
44dcb63b 2157 char *d;
89bfa8cd 2158
7918f24d
NC
2159 PERL_ARGS_ASSERT_FORCE_VERSION;
2160
294a536f 2161 s = skipspace(s);
89bfa8cd 2162
44dcb63b 2163 d = s;
dd629d5b 2164 if (*d == 'v')
44dcb63b 2165 d++;
44dcb63b 2166 if (isDIGIT(*d)) {
e759cc13
RGS
2167 while (isDIGIT(*d) || *d == '_' || *d == '.')
2168 d++;
4e4da3ac 2169 if (*d == ';' || isSPACE(*d) || *d == '{' || *d == '}' || !*d) {
dd629d5b 2170 SV *ver;
6154021b
RGS
2171 s = scan_num(s, &pl_yylval);
2172 version = pl_yylval.opval;
dd629d5b
GS
2173 ver = cSVOPx(version)->op_sv;
2174 if (SvPOK(ver) && !SvNIOK(ver)) {
862a34c6 2175 SvUPGRADE(ver, SVt_PVNV);
9d6ce603 2176 SvNV_set(ver, str_to_version(ver));
1571675a 2177 SvNOK_on(ver); /* hint that it is a version */
44dcb63b 2178 }
89bfa8cd 2179 }
5db06880 2180 else if (guessing) {
e759cc13 2181 return s;
5db06880 2182 }
89bfa8cd
PP
2183 }
2184
2185 /* NOTE: The parser sees the package name and the VERSION swapped */
9ded7720 2186 NEXTVAL_NEXTTOKE.opval = version;
185c2e96 2187 force_next(BAREWORD);
89bfa8cd 2188
e759cc13 2189 return s;
89bfa8cd
PP
2190}
2191
ffb4593c 2192/*
91152fc1
DG
2193 * S_force_strict_version
2194 * Forces the next token to be a version number using strict syntax rules.
2195 */
2196
2197STATIC char *
2198S_force_strict_version(pTHX_ char *s)
2199{
91152fc1 2200 OP *version = NULL;
91152fc1
DG
2201 const char *errstr = NULL;
2202
2203 PERL_ARGS_ASSERT_FORCE_STRICT_VERSION;
2204
2205 while (isSPACE(*s)) /* leading whitespace */
2206 s++;
2207
2208 if (is_STRICT_VERSION(s,&errstr)) {
2209 SV *ver = newSV(0);
2210 s = (char *)scan_version(s, ver, 0);
2211 version = newSVOP(OP_CONST, 0, ver);
2212 }
407f8cf2
KW
2213 else if ((*s != ';' && *s != '{' && *s != '}' )
2214 && (s = skipspace(s), (*s != ';' && *s != '{' && *s != '}' )))
4e4da3ac 2215 {
91152fc1
DG
2216 PL_bufptr = s;
2217 if (errstr)
2218 yyerror(errstr); /* version required */
2219 return s;
2220 }
2221
91152fc1 2222 /* NOTE: The parser sees the package name and the VERSION swapped */
91152fc1 2223 NEXTVAL_NEXTTOKE.opval = version;
185c2e96 2224 force_next(BAREWORD);
91152fc1
DG
2225
2226 return s;
2227}
2228
2229/*
ffb4593c
NT
2230 * S_tokeq
2231 * Tokenize a quoted string passed in as an SV. It finds the next
2232 * chunk, up to end of string or a backslash. It may make a new
2233 * SV containing that chunk (if HINT_NEW_STRING is on). It also
2234 * turns \\ into \.
2235 */
2236
76e3520e 2237STATIC SV *
cea2e8a9 2238S_tokeq(pTHX_ SV *sv)
79072805 2239{
eb578fdb
KW
2240 char *s;
2241 char *send;
2242 char *d;
b3ac6de7 2243 SV *pv = sv;
79072805 2244
7918f24d
NC
2245 PERL_ARGS_ASSERT_TOKEQ;
2246
279b35ad
FC
2247 assert (SvPOK(sv));
2248 assert (SvLEN(sv));
2249 assert (!SvIsCOW(sv));
307ed071 2250 if (SvTYPE(sv) >= SVt_PVIV && SvIVX(sv) == -1) /* <<'heredoc' */
b3ac6de7 2251 goto finish;
279b35ad
FC
2252 s = SvPVX(sv);
2253 send = SvEND(sv);
dcb21ed6
NC
2254 /* This is relying on the SV being "well formed" with a trailing '\0' */
2255 while (s < send && !(*s == '\\' && s[1] == '\\'))
79072805
LW
2256 s++;
2257 if (s == send)
b3ac6de7 2258 goto finish;
79072805 2259 d = s;
be4731d2 2260 if ( PL_hints & HINT_NEW_STRING ) {
279b35ad
FC
2261 pv = newSVpvn_flags(SvPVX_const(pv), SvCUR(sv),
2262 SVs_TEMP | SvUTF8(sv));
be4731d2 2263 }
79072805
LW
2264 while (s < send) {
2265 if (*s == '\\') {
a0d0e21e 2266 if (s + 1 < send && (s[1] == '\\'))
79072805
LW
2267 s++; /* all that, just for this */
2268 }
2269 *d++ = *s++;
2270 }
2271 *d = '\0';
95a20fc0 2272 SvCUR_set(sv, d - SvPVX_const(sv));
b3ac6de7 2273 finish:
3280af22 2274 if ( PL_hints & HINT_NEW_STRING )
eb0d8d16 2275 return new_constant(NULL, 0, "q", sv, pv, "q", 1);
79072805
LW
2276 return sv;
2277}
2278
ffb4593c
NT
2279/*
2280 * Now come three functions related to double-quote context,
2281 * S_sublex_start, S_sublex_push, and S_sublex_done. They're used when
2282 * converting things like "\u\Lgnat" into ucfirst(lc("gnat")). They
2283 * interact with PL_lex_state, and create fake ( ... ) argument lists
2284 * to handle functions and concatenation.
ecd24171
DM
2285 * For example,
2286 * "foo\lbar"
2287 * is tokenised as
2288 * stringify ( const[foo] concat lcfirst ( const[bar] ) )
ffb4593c
NT
2289 */
2290
2291/*
2292 * S_sublex_start
6154021b 2293 * Assumes that pl_yylval.ival is the op we're creating (e.g. OP_LCFIRST).
ffb4593c
NT
2294 *
2295 * Pattern matching will set PL_lex_op to the pattern-matching op to
6154021b 2296 * make (we return THING if pl_yylval.ival is OP_NULL, PMFUNC otherwise).
ffb4593c 2297 *
16d1d8bd 2298 * OP_CONST is easy--just make the new op and return.
ffb4593c
NT
2299 *
2300 * Everything else becomes a FUNC.
2301 *
16d1d8bd
LM
2302 * Sets PL_lex_state to LEX_INTERPPUSH unless ival was OP_NULL or we
2303 * had an OP_CONST. This just sets us up for a
ffb4593c
NT
2304 * call to S_sublex_push().
2305 */
2306
76e3520e 2307STATIC I32
cea2e8a9 2308S_sublex_start(pTHX)
79072805 2309{
eb578fdb 2310 const I32 op_type = pl_yylval.ival;
79072805
LW
2311
2312 if (op_type == OP_NULL) {
6154021b 2313 pl_yylval.opval = PL_lex_op;
5f66b61c 2314 PL_lex_op = NULL;
79072805
LW
2315 return THING;
2316 }
466112bb 2317 if (op_type == OP_CONST) {
67c71cbb
FC
2318 SV *sv = PL_lex_stuff;
2319 PL_lex_stuff = NULL;
2320 sv = tokeq(sv);
b3ac6de7
IZ
2321
2322 if (SvTYPE(sv) == SVt_PVIV) {
2323 /* Overloaded constants, nothing fancy: Convert to SVt_PV: */
2324 STRLEN len;
96a5add6 2325 const char * const p = SvPV_const(sv, len);
740cce10 2326 SV * const nsv = newSVpvn_flags(p, len, SvUTF8(sv));
b3ac6de7
IZ
2327 SvREFCNT_dec(sv);
2328 sv = nsv;
4e553d73 2329 }
275103cd 2330 pl_yylval.opval = newSVOP(op_type, 0, sv);
79072805
LW
2331 return THING;
2332 }
2333
7ef70b3d
FC
2334 PL_parser->lex_super_state = PL_lex_state;
2335 PL_parser->lex_sub_inwhat = (U16)op_type;
2336 PL_parser->lex_sub_op = PL_lex_op;
3280af22 2337 PL_lex_state = LEX_INTERPPUSH;
55497cff 2338
3280af22
NIS
2339 PL_expect = XTERM;
2340 if (PL_lex_op) {
6154021b 2341 pl_yylval.opval = PL_lex_op;
5f66b61c 2342 PL_lex_op = NULL;
55497cff
PP
2343 return PMFUNC;
2344 }
2345 else
2346 return FUNC;
2347}
2348
ffb4593c
NT
2349/*
2350 * S_sublex_push
2351 * Create a new scope to save the lexing state. The scope will be
2352 * ended in S_sublex_done. Returns a '(', starting the function arguments
2353 * to the uc, lc, etc. found before.
2354 * Sets PL_lex_state to LEX_INTERPCONCAT.
2355 */
2356
76e3520e 2357STATIC I32
cea2e8a9 2358S_sublex_push(pTHX)
55497cff 2359{
78a635de 2360 LEXSHARED *shared;
801d32ac 2361 const bool is_heredoc = PL_multi_close == '<';
f46d017c 2362 ENTER;
55497cff 2363
7ef70b3d 2364 PL_lex_state = PL_parser->lex_super_state;
cc624add 2365 SAVEI8(PL_lex_dojoin);
3280af22 2366 SAVEI32(PL_lex_brackets);
78cdf107 2367 SAVEI32(PL_lex_allbrackets);
b27dce25 2368 SAVEI32(PL_lex_formbrack);
78cdf107 2369 SAVEI8(PL_lex_fakeeof);
3280af22
NIS
2370 SAVEI32(PL_lex_casemods);
2371 SAVEI32(PL_lex_starts);
651b5b28 2372 SAVEI8(PL_lex_state);
7cc34111 2373 SAVESPTR(PL_lex_repl);
7766f137 2374 SAVEVPTR(PL_lex_inpat);
98246f1e 2375 SAVEI16(PL_lex_inwhat);
ffdb8b16 2376 if (is_heredoc)
b42366d4 2377 {
ffdb8b16 2378 SAVECOPLINE(PL_curcop);
b42366d4 2379 SAVEI32(PL_multi_end);
851b527a
FC
2380 SAVEI32(PL_parser->herelines);
2381 PL_parser->herelines = 0;
b42366d4 2382 }
2ca4363d 2383 SAVEIV(PL_multi_close);
3280af22 2384 SAVEPPTR(PL_bufptr);
8452ff4b 2385 SAVEPPTR(PL_bufend);
3280af22
NIS
2386 SAVEPPTR(PL_oldbufptr);
2387 SAVEPPTR(PL_oldoldbufptr);
207e3d1a
JH
2388 SAVEPPTR(PL_last_lop);
2389 SAVEPPTR(PL_last_uni);
3280af22
NIS
2390 SAVEPPTR(PL_linestart);
2391 SAVESPTR(PL_linestr);
8edd5f42
RGS
2392 SAVEGENERICPV(PL_lex_brackstack);
2393 SAVEGENERICPV(PL_lex_casestack);
78a635de 2394 SAVEGENERICPV(PL_parser->lex_shared);
3a54fd60 2395 SAVEBOOL(PL_parser->lex_re_reparsing);
ffdb8b16 2396 SAVEI32(PL_copline);
3280af22 2397
99bd9d90 2398 /* The here-doc parser needs to be able to peek into outer lexing
60f40a38
FC
2399 scopes to find the body of the here-doc. So we put PL_linestr and
2400 PL_bufptr into lex_shared, to ‘share’ those values.
99bd9d90 2401 */
60f40a38
FC
2402 PL_parser->lex_shared->ls_linestr = PL_linestr;
2403 PL_parser->lex_shared->ls_bufptr = PL_bufptr;
99bd9d90 2404
3280af22 2405 PL_linestr = PL_lex_stuff;
7ef70b3d 2406 PL_lex_repl = PL_parser->lex_sub_repl;
a0714e2c 2407 PL_lex_stuff = NULL;
7ef70b3d 2408 PL_parser->lex_sub_repl = NULL;
3280af22 2409
eabab8bc
FC
2410 /* Arrange for PL_lex_stuff to be freed on scope exit, in case it gets
2411 set for an inner quote-like operator and then an error causes scope-
2412 popping. We must not have a PL_lex_stuff value left dangling, as
2413 that breaks assumptions elsewhere. See bug #123617. */
2414 SAVEGENERICSV(PL_lex_stuff);
7ef70b3d 2415 SAVEGENERICSV(PL_parser->lex_sub_repl);
eabab8bc 2416
9cbb5ea2
GS
2417 PL_bufend = PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart
2418 = SvPVX(PL_linestr);
3280af22 2419 PL_bufend += SvCUR(PL_linestr);
bd61b366 2420 PL_last_lop = PL_last_uni = NULL;
3280af22 2421 SAVEFREESV(PL_linestr);
4dc843bc 2422 if (PL_lex_repl) SAVEFREESV(PL_lex_repl);
3280af22
NIS
2423
2424 PL_lex_dojoin = FALSE;
b27dce25 2425 PL_lex_brackets = PL_lex_formbrack = 0;
78cdf107
Z
2426 PL_lex_allbrackets = 0;
2427 PL_lex_fakeeof = LEX_FAKEEOF_NEVER;
a02a5408
JC
2428 Newx(PL_lex_brackstack, 120, char);
2429 Newx(PL_lex_casestack, 12, char);
3280af22
NIS
2430 PL_lex_casemods = 0;
2431 *PL_lex_casestack = '\0';
2432 PL_lex_starts = 0;
2433 PL_lex_state = LEX_INTERPCONCAT;
ffdb8b16 2434 if (is_heredoc)
6ddcf93b 2435 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
ffdb8b16 2436 PL_copline = NOLINE;
78a635de
FC
2437
2438 Newxz(shared, 1, LEXSHARED);
2439 shared->ls_prev = PL_parser->lex_shared;
2440 PL_parser->lex_shared = shared;
3280af22 2441
7ef70b3d 2442 PL_lex_inwhat = PL_parser->lex_sub_inwhat;
bb16bae8 2443 if (PL_lex_inwhat == OP_TRANSR) PL_lex_inwhat = OP_TRANS;
3280af22 2444 if (PL_lex_inwhat == OP_MATCH || PL_lex_inwhat == OP_QR || PL_lex_inwhat == OP_SUBST)
7ef70b3d 2445 PL_lex_inpat = PL_parser->lex_sub_op;
79072805 2446 else
5f66b61c 2447 PL_lex_inpat = NULL;
79072805 2448
3a54fd60
DM
2449 PL_parser->lex_re_reparsing = cBOOL(PL_in_eval & EVAL_RE_REPARSING);
2450 PL_in_eval &= ~EVAL_RE_REPARSING;
2451
55497cff 2452 return '(';
79072805
LW
2453}
2454
ffb4593c
NT
2455/*
2456 * S_sublex_done
2457 * Restores lexer state after a S_sublex_push.
2458 */
2459
76e3520e 2460STATIC I32
cea2e8a9 2461S_sublex_done(pTHX)
79072805 2462{
3280af22 2463 if (!PL_lex_starts++) {
396482e1 2464 SV * const sv = newSVpvs("");
9aa983d2
JH
2465 if (SvUTF8(PL_linestr))
2466 SvUTF8_on(sv);
3280af22 2467 PL_expect = XOPERATOR;
275103cd 2468 pl_yylval.opval = newSVOP(OP_CONST, 0, sv);
79072805
LW
2469 return THING;
2470 }
2471
3280af22
NIS
2472 if (PL_lex_casemods) { /* oops, we've got some unbalanced parens */
2473 PL_lex_state = LEX_INTERPCASEMOD;
cea2e8a9 2474 return yylex();
79072805
LW
2475 }
2476
ffb4593c 2477 /* Is there a right-hand side to take care of? (s//RHS/ or tr//RHS/) */
bb16bae8 2478 assert(PL_lex_inwhat != OP_TRANSR);
5aa91856
FC
2479 if (PL_lex_repl) {
2480 assert (PL_lex_inwhat == OP_SUBST || PL_lex_inwhat == OP_TRANS);
3280af22
NIS
2481 PL_linestr = PL_lex_repl;
2482 PL_lex_inpat = 0;
2483 PL_bufend = PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart = SvPVX(PL_linestr);
2484 PL_bufend += SvCUR(PL_linestr);
bd61b366 2485 PL_last_lop = PL_last_uni = NULL;
3280af22
NIS
2486 PL_lex_dojoin = FALSE;
2487 PL_lex_brackets = 0;
78cdf107
Z
2488 PL_lex_allbrackets = 0;
2489 PL_lex_fakeeof = LEX_FAKEEOF_NEVER;
3280af22
NIS
2490 PL_lex_casemods = 0;
2491 *PL_lex_casestack = '\0';
2492 PL_lex_starts = 0;
25da4f38 2493 if (SvEVALED(PL_lex_repl)) {
3280af22
NIS
2494 PL_lex_state = LEX_INTERPNORMAL;
2495 PL_lex_starts++;
e9fa98b2
HS
2496 /* we don't clear PL_lex_repl here, so that we can check later
2497 whether this is an evalled subst; that means we rely on the
2498 logic to ensure sublex_done() is called again only via the
2499 branch (in yylex()) that clears PL_lex_repl, else we'll loop */
79072805 2500 }
e9fa98b2 2501 else {
3280af22 2502 PL_lex_state = LEX_INTERPCONCAT;
a0714e2c 2503 PL_lex_repl = NULL;
e9fa98b2 2504 }
ffdb8b16
FC
2505 if (SvTYPE(PL_linestr) >= SVt_PVNV) {
2506 CopLINE(PL_curcop) +=
9420b268 2507 ((XPVNV*)SvANY(PL_linestr))->xnv_u.xnv_lines
851b527a
FC
2508 + PL_parser->herelines;
2509 PL_parser->herelines = 0;
ffdb8b16 2510 }
9b6b7be8 2511 return '/';
ffed7fef
LW
2512 }
2513 else {
b42366d4 2514 const line_t l = CopLINE(PL_curcop);
f46d017c 2515 LEAVE;
b42366d4 2516 if (PL_multi_close == '<')
851b527a 2517 PL_parser->herelines += l - PL_multi_end;
3280af22
NIS
2518 PL_bufend = SvPVX(PL_linestr);
2519 PL_bufend += SvCUR(PL_linestr);
2520 PL_expect = XOPERATOR;
79072805 2521 return ')';
ffed7fef
LW
2522 }
2523}
2524
6f613c73
KW
2525PERL_STATIC_INLINE SV*
2526S_get_and_check_backslash_N_name(pTHX_ const char* s, const char* const e)
2527{
140b12ad
KW
2528 /* <s> points to first character of interior of \N{}, <e> to one beyond the
2529 * interior, hence to the "}". Finds what the name resolves to, returning
2530 * an SV* containing it; NULL if no valid one found */
2531
dd2b1b72 2532 SV* res = newSVpvn_flags(s, e - s, UTF ? SVf_UTF8 : 0);
6f613c73 2533
0c415a79
KW
2534 HV * table;
2535 SV **cvp;
2536 SV *cv;
2537 SV *rv;
2538 HV *stash;
2539 const U8* first_bad_char_loc;
2540 const char* backslash_ptr = s - 3; /* Points to the <\> of \N{... */
2541
6f613c73
KW
2542 PERL_ARGS_ASSERT_GET_AND_CHECK_BACKSLASH_N_NAME;
2543
d8d26cac
KW
2544 if (!SvCUR(res)) {
2545 Perl_ck_warner_d(aTHX_ packWARN(WARN_DEPRECATED),
2546 "Unknown charname '' is deprecated");
b7e6151c 2547 return res;
d8d26cac 2548 }
b7e6151c 2549
107160e2
KW
2550 if (UTF && ! is_utf8_string_loc((U8 *) backslash_ptr,
2551 e - backslash_ptr,
2552 &first_bad_char_loc))
2553 {
75219bac
KW
2554 _force_out_malformed_utf8_message(first_bad_char_loc,
2555 (U8 *) PL_parser->bufend,
2556 0,
2557 0 /* 0 means don't die */ );
b59c097b 2558 yyerror_pv(Perl_form(aTHX_
107160e2 2559 "Malformed UTF-8 character immediately after '%.*s'",
b59c097b
AV
2560 (int) (first_bad_char_loc - (U8 *) backslash_ptr), backslash_ptr),
2561 SVf_UTF8);
107160e2
KW
2562 return NULL;
2563 }
2564
2565 res = new_constant( NULL, 0, "charnames", res, NULL, backslash_ptr,
2566 /* include the <}> */
2567 e - backslash_ptr + 1);
6f613c73 2568 if (! SvPOK(res)) {
b6407c49 2569 SvREFCNT_dec_NN(res);
6f613c73
KW
2570 return NULL;
2571 }
2572
0c415a79
KW
2573 /* See if the charnames handler is the Perl core's, and if so, we can skip
2574 * the validation needed for a user-supplied one, as Perl's does its own
2575 * validation. */
2576 table = GvHV(PL_hintgv); /* ^H */
2577 cvp = hv_fetchs(table, "charnames", FALSE);
5882ddb3
FC
2578 if (cvp && (cv = *cvp) && SvROK(cv) && (rv = SvRV(cv),
2579 SvTYPE(rv) == SVt_PVCV) && ((stash = CvSTASH(rv)) != NULL))
0c415a79
KW
2580 {
2581 const char * const name = HvNAME(stash);
6a642c21
FC
2582 if (HvNAMELEN(stash) == sizeof("_charnames")-1
2583 && strEQ(name, "_charnames")) {
0c415a79
KW
2584 return res;
2585 }
2586 }
2587
bde9e88d
KW
2588 /* Here, it isn't Perl's charname handler. We can't rely on a
2589 * user-supplied handler to validate the input name. For non-ut8 input,
2590 * look to see that the first character is legal. Then loop through the
2591 * rest checking that each is a continuation */
6f613c73 2592
36897d64
KW
2593 /* This code makes the reasonable assumption that the only Latin1-range
2594 * characters that begin a character name alias are alphabetic, otherwise
2595 * would have to create a isCHARNAME_BEGIN macro */
b6ba1137 2596
b6ba1137 2597 if (! UTF) {
bde9e88d 2598 if (! isALPHAU(*s)) {
b6ba1137
KW
2599 goto bad_charname;
2600 }
bde9e88d
KW
2601 s++;
2602 while (s < e) {
2603 if (! isCHARNAME_CONT(*s)) {
b6ba1137
KW
2604 goto bad_charname;
2605 }
2d8eb851
KW
2606 if (*s == ' ' && *(s-1) == ' ') {
2607 goto multi_spaces;
bd299e29 2608 }
bde9e88d 2609 s++;
b6ba1137
KW
2610 }
2611 }
2612 else {
bde9e88d
KW
2613 /* Similarly for utf8. For invariants can check directly; for other
2614 * Latin1, can calculate their code point and check; otherwise use a
2615 * swash */
2616 if (UTF8_IS_INVARIANT(*s)) {
2617 if (! isALPHAU(*s)) {
140b12ad
KW
2618 goto bad_charname;
2619 }
bde9e88d
KW
2620 s++;
2621 } else if (UTF8_IS_DOWNGRADEABLE_START(*s)) {
a62b247b 2622 if (! isALPHAU(EIGHT_BIT_UTF8_TO_NATIVE(*s, *(s+1)))) {
b6ba1137 2623 goto bad_charname;
6f613c73 2624 }
bde9e88d 2625 s += 2;
6f613c73 2626 }
bde9e88d
KW
2627 else {
2628 if (! PL_utf8_charname_begin) {
2629 U8 flags = _CORE_SWASH_INIT_ACCEPT_INVLIST;
2630 PL_utf8_charname_begin = _core_swash_init("utf8",
2631 "_Perl_Charname_Begin",
2632 &PL_sv_undef,
2633 1, 0, NULL, &flags);
2634 }
2635 if (! swash_fetch(PL_utf8_charname_begin, (U8 *) s, TRUE)) {
2636 goto bad_charname;
2637 }
2638 s += UTF8SKIP(s);
2639 }
2640
2641 while (s < e) {
2642 if (UTF8_IS_INVARIANT(*s)) {
2643 if (! isCHARNAME_CONT(*s)) {
2644 goto bad_charname;
2645 }
2d8eb851
KW
2646 if (*s == ' ' && *(s-1) == ' ') {
2647 goto multi_spaces;
bd299e29 2648 }
bde9e88d
KW
2649 s++;
2650 }
2651 else if (UTF8_IS_DOWNGRADEABLE_START(*s)) {
a62b247b 2652 if (! isCHARNAME_CONT(EIGHT_BIT_UTF8_TO_NATIVE(*s, *(s+1))))
bde9e88d
KW
2653 {
2654 goto bad_charname;
2655 }
2656 s += 2;
2657 }
2658 else {
2659 if (! PL_utf8_charname_continue) {
2660 U8 flags = _CORE_SWASH_INIT_ACCEPT_INVLIST;
2661 PL_utf8_charname_continue = _core_swash_init("utf8",
2662 "_Perl_Charname_Continue",
2663 &PL_sv_undef,
2664 1, 0, NULL, &flags);
2665 }
2666 if (! swash_fetch(PL_utf8_charname_continue, (U8 *) s, TRUE)) {
2667 goto bad_charname;
2668 }
2669 s += UTF8SKIP(s);
6f613c73
KW
2670 }
2671 }
2d8eb851
KW
2672 }
2673 if (*(s-1) == ' ') {
2674 yyerror_pv(
2675 Perl_form(aTHX_
2676 "charnames alias definitions may not contain trailing "
2677 "white-space; marked by <-- HERE in %.*s<-- HERE %.*s",
2678 (int)(s - backslash_ptr + 1), backslash_ptr,
2679 (int)(e - s + 1), s + 1
2680 ),
2681 UTF ? SVf_UTF8 : 0);
2682 return NULL;
6f613c73
KW
2683 }
2684
94ca1619 2685 if (SvUTF8(res)) { /* Don't accept malformed input */
bde9e88d
KW
2686 const U8* first_bad_char_loc;
2687 STRLEN len;
2688 const char* const str = SvPV_const(res, len);
2689 if (! is_utf8_string_loc((U8 *) str, len, &first_bad_char_loc)) {
75219bac
KW
2690 _force_out_malformed_utf8_message(first_bad_char_loc,
2691 (U8 *) PL_parser->bufend,
2692 0,
2693 0 /* 0 means don't die */ );
bde9e88d
KW
2694 yyerror_pv(
2695 Perl_form(aTHX_
2696 "Malformed UTF-8 returned by %.*s immediately after '%.*s'",
2697 (int) (e - backslash_ptr + 1), backslash_ptr,
2698 (int) ((char *) first_bad_char_loc - str), str
2699 ),
2700 SVf_UTF8);
2701 return NULL;
2702 }
2703 }
140b12ad 2704
bde9e88d 2705 return res;
140b12ad 2706
bde9e88d 2707 bad_charname: {
bde9e88d
KW
2708
2709 /* The final %.*s makes sure that should the trailing NUL be missing
2710 * that this print won't run off the end of the string */
2711 yyerror_pv(
2712 Perl_form(aTHX_
2713 "Invalid character in \\N{...}; marked by <-- HERE in %.*s<-- HERE %.*s",
2d8eb851
KW
2714 (int)(s - backslash_ptr + 1), backslash_ptr,
2715 (int)(e - s + 1), s + 1
bde9e88d
KW
2716 ),
2717 UTF ? SVf_UTF8 : 0);
2718 return NULL;
2719 }
2d8eb851
KW
2720
2721 multi_spaces:
2722 yyerror_pv(
2723 Perl_form(aTHX_
2724 "charnames alias definitions may not contain a sequence of "
2725 "multiple spaces; marked by <-- HERE in %.*s<-- HERE %.*s",
2726 (int)(s - backslash_ptr + 1), backslash_ptr,
2727 (int)(e - s + 1), s + 1
2728 ),
2729 UTF ? SVf_UTF8 : 0);
2730 return NULL;
6f613c73
KW
2731}
2732
02aa26ce
NT
2733/*
2734 scan_const
2735
9da1dd8f
DM
2736 Extracts the next constant part of a pattern, double-quoted string,
2737 or transliteration. This is terrifying code.
2738
2739 For example, in parsing the double-quoted string "ab\x63$d", it would
2740 stop at the '$' and return an OP_CONST containing 'abc'.
02aa26ce 2741
94def140 2742 It looks at PL_lex_inwhat and PL_lex_inpat to find out whether it's
3280af22 2743 processing a pattern (PL_lex_inpat is true), a transliteration
94def140 2744 (PL_lex_inwhat == OP_TRANS is true), or a double-quoted string.
02aa26ce 2745
94def140
ST
2746 Returns a pointer to the character scanned up to. If this is
2747 advanced from the start pointer supplied (i.e. if anything was
9da1dd8f 2748 successfully parsed), will leave an OP_CONST for the substring scanned
6154021b 2749 in pl_yylval. Caller must intuit reason for not parsing further
9b599b2a
GS
2750 by looking at the next characters herself.
2751
02aa26ce 2752 In patterns:
9da1dd8f 2753 expand:
537124e4
KW
2754 \N{FOO} => \N{U+hex_for_character_FOO}
2755 (if FOO expands to multiple characters, expands to \N{U+xx.XX.yy ...})
9da1dd8f
DM
2756
2757 pass through:
2758 all other \-char, including \N and \N{ apart from \N{ABC}
2759
2760 stops on:
2761 @ and $ where it appears to be a var, but not for $ as tail anchor
2762 \l \L \u \U \Q \E
2763 (?{ or (??{
2764
02aa26ce
NT
2765 In transliterations:
2766 characters are VERY literal, except for - not at the start or end
94def140
ST
2767 of the string, which indicates a range. If the range is in bytes,
2768 scan_const expands the range to the full set of intermediate
2769 characters. If the range is in utf8, the hyphen is replaced with
2770 a certain range mark which will be handled by pmtrans() in op.c.
02aa26ce
NT
2771
2772 In double-quoted strings:
2773 backslashes:
2774 double-quoted style: \r and \n
ff3f963a 2775 constants: \x31, etc.
94def140 2776 deprecated backrefs: \1 (in substitution replacements)
02aa26ce
NT
2777 case and quoting: \U \Q \E
2778 stops on @ and $
2779
2780 scan_const does *not* construct ops to handle interpolated strings.
2781 It stops processing as soon as it finds an embedded $ or @ variable
2782 and leaves it to the caller to work out what's going on.
2783
94def140
ST
2784 embedded arrays (whether in pattern or not) could be:
2785 @foo, @::foo, @'foo, @{foo}, @$foo, @+, @-.
2786
2787 $ in double-quoted strings must be the symbol of an embedded scalar.
02aa26ce
NT
2788
2789 $ in pattern could be $foo or could be tail anchor. Assumption:
2790 it's a tail anchor if $ is the last thing in the string, or if it's
94def140 2791 followed by one of "()| \r\n\t"
02aa26ce 2792
9da1dd8f 2793 \1 (backreferences) are turned into $1 in substitutions
02aa26ce
NT
2794
2795 The structure of the code is
2796 while (there's a character to process) {
94def140
ST
2797 handle transliteration ranges
2798 skip regexp comments /(?#comment)/ and codes /(?{code})/
2799 skip #-initiated comments in //x patterns
2800 check for embedded arrays
02aa26ce
NT
2801 check for embedded scalars
2802 if (backslash) {
94def140 2803 deprecate \1 in substitution replacements
02aa26ce
NT
2804 handle string-changing backslashes \l \U \Q \E, etc.
2805 switch (what was escaped) {
94def140 2806 handle \- in a transliteration (becomes a literal -)
ff3f963a 2807 if a pattern and not \N{, go treat as regular character
94def140
ST
2808 handle \132 (octal characters)
2809 handle \x15 and \x{1234} (hex characters)
ff3f963a 2810 handle \N{name} (named characters, also \N{3,5} in a pattern)
94def140
ST
2811 handle \cV (control characters)
2812 handle printf-style backslashes (\f, \r, \n, etc)
02aa26ce 2813 } (end switch)
77a135fe 2814 continue
02aa26ce 2815 } (end if backslash)
77a135fe 2816 handle regular character
02aa26ce 2817 } (end while character to read)
4e553d73 2818
02aa26ce
NT
2819*/
2820
76e3520e 2821STATIC char *
cea2e8a9 2822S_scan_const(pTHX_ char *start)
79072805 2823{
eb578fdb 2824 char *send = PL_bufend; /* end of the constant */
dc023dbb
KW
2825 SV *sv = newSV(send - start); /* sv for the constant. See note below
2826 on sizing. */
eb578fdb
KW
2827 char *s = start; /* start of the constant */
2828 char *d = SvPVX(sv); /* destination for copies */
dc023dbb
KW
2829 bool dorange = FALSE; /* are we in a translit range? */
2830 bool didrange = FALSE; /* did we just finish a range? */
2831 bool in_charclass = FALSE; /* within /[...]/ */
2832 bool has_utf8 = FALSE; /* Output constant is UTF8 */
2833 bool this_utf8 = cBOOL(UTF); /* Is the source string assumed to be
2834 UTF8? But, this can show as true
2835 when the source isn't utf8, as for
2836 example when it is entirely composed
2837 of hex constants */
6f613c73 2838 SV *res; /* result from charnames */
f4240379
KW
2839 STRLEN offset_to_max; /* The offset in the output to where the range
2840 high-end character is temporarily placed */
77a135fe
KW
2841
2842 /* Note on sizing: The scanned constant is placed into sv, which is
2843 * initialized by newSV() assuming one byte of output for every byte of
2844 * input. This routine expects newSV() to allocate an extra byte for a
2845 * trailing NUL, which this routine will append if it gets to the end of
2846 * the input. There may be more bytes of input than output (eg., \N{LATIN
2847 * CAPITAL LETTER A}), or more output than input if the constant ends up
2848 * recoded to utf8, but each time a construct is found that might increase
2849 * the needed size, SvGROW() is called. Its size parameter each time is
2850 * based on the best guess estimate at the time, namely the length used so
2851 * far, plus the length the current construct will occupy, plus room for
2852 * the trailing NUL, plus one byte for every input byte still unscanned */
2853
c3320c2a
KW
2854 UV uv = UV_MAX; /* Initialize to weird value to try to catch any uses
2855 before set */
4c3a8340 2856#ifdef EBCDIC
f4240379
KW
2857 int backslash_N = 0; /* ? was the character from \N{} */
2858 int non_portable_endpoint = 0; /* ? In a range is an endpoint
2859 platform-specific like \x65 */
4c3a8340 2860#endif
012bcf8d 2861
7918f24d
NC
2862 PERL_ARGS_ASSERT_SCAN_CONST;
2863
bb16bae8 2864 assert(PL_lex_inwhat != OP_TRANSR);
7ef70b3d 2865 if (PL_lex_inwhat == OP_TRANS && PL_parser->lex_sub_op) {
2b9d42f0 2866 /* If we are doing a trans and we know we want UTF8 set expectation */
7ef70b3d
FC
2867 has_utf8 = PL_parser->lex_sub_op->op_private & (OPpTRANS_FROM_UTF|OPpTRANS_TO_UTF);
2868 this_utf8 = PL_parser->lex_sub_op->op_private & (PL_lex_repl ? OPpTRANS_FROM_UTF : OPpTRANS_TO_UTF);
2b9d42f0
NIS
2869 }
2870
b899e89d
FC
2871 /* Protect sv from errors and fatal warnings. */
2872 ENTER_with_name("scan_const");
2873 SAVEFREESV(sv);
2b9d42f0 2874
f4240379
KW
2875 while (s < send
2876 || dorange /* Handle tr/// range at right edge of input */
2877 ) {
ff3f963a 2878
02aa26ce 2879 /* get transliterations out of the way (they're most literal) */
3280af22 2880 if (PL_lex_inwhat == OP_TRANS) {
02aa26ce 2881
f4240379
KW
2882 /* But there isn't any special handling necessary unless there is a
2883 * range, so for most cases we just drop down and handle the value
2884 * as any other. There are two exceptions.
2885 *
2886 * 1. A minus sign indicates that we are actually going to have
2887 * a range. In this case, skip the '-', set a flag, then drop
2888 * down to handle what should be the end range value.
2889 * 2. After we've handled that value, the next time through, that
2890 * flag is set and we fix up the range.
2891 *
2892 * Ranges entirely within Latin1 are expanded out entirely, in
2893 * order to avoid the significant overhead of making a swash.
2894 * Ranges that extend above Latin1 have to have a swash, so there
4698efb6
KW
2895 * is no advantage to abbreviating them here, so they are stored
2896 * here as Min, ILLEGAL_UTF8_BYTE, Max. The illegal byte signifies
2897 * a hyphen without any possible ambiguity. On EBCDIC machines, if
f4240379
KW
2898 * the range is expressed as Unicode, the Latin1 portion is
2899 * expanded out even if the entire range extends above Latin1.
2900 * This is because each code point in it has to be processed here
2901 * individually to get its native translation */
2902
2903 if (! dorange) {
2904
2905 /* Here, we don't think we're in a range. If we've processed
2906 * at least one character, then see if this next one is a '-',
2907 * indicating the previous one was the start of a range. But
2908 * don't bother if we're too close to the end for the minus to
2909 * mean that. */
2910 if (*s != '-' || s >= send - 1 || s == start) {
2911
2912 /* A regular character. Process like any other, but first
2913 * clear any flags */
2914 didrange = FALSE;
2915 dorange = FALSE;
e294cc5d 2916#ifdef EBCDIC
f4240379
KW
2917 non_portable_endpoint = 0;
2918 backslash_N = 0;
e294cc5d 2919#endif
f4240379
KW
2920 /* Drops down to generic code to process current byte */
2921 }
2922 else {
2923 if (didrange) { /* Something like y/A-C-Z// */
2924 Perl_croak(aTHX_ "Ambiguous range in transliteration operator");
2925 }
e294cc5d 2926
f4240379 2927 dorange = TRUE;
2b9d42f0 2928
f4240379
KW
2929 s++; /* Skip past the minus */
2930
2931 /* d now points to where the end-range character will be
2932 * placed. Save it so won't have to go finding it later,
2933 * and drop down to get that character. (Actually we
2934 * instead save the offset, to handle the case where a
2935 * realloc in the meantime could change the actual
2936 * pointer). We'll finish processing the range the next
2937 * time through the loop */
2938 offset_to_max = d - SvPVX_const(sv);
2939 }
2940 } /* End of not a range */
2941 else {
2942 /* Here we have parsed a range. Now must handle it. At this
2943 * point:
2944 * 'sv' is a SV* that contains the output string we are
2945 * constructing. The final two characters in that string
2946 * are the range start and range end, in order.
2947 * 'd' points to just beyond the range end in the 'sv' string,
2948 * where we would next place something
2949 * 'offset_to_max' is the offset in 'sv' at which the character
2950 * before 'd' begins.
2951 */
2952 const char * max_ptr = SvPVX_const(sv) + offset_to_max;
2953 const char * min_ptr;
2954 IV range_min;
2955 IV range_max; /* last character in range */
2956 STRLEN save_offset;
2957 STRLEN grow;
11327fa1 2958#ifdef EBCDIC
f4240379
KW
2959 bool convert_unicode;
2960 IV real_range_max = 0;
e294cc5d 2961#endif
f4240379
KW
2962
2963 /* Get the range-ends code point values. */
e294cc5d 2964 if (has_utf8) {
f4240379
KW
2965 /* We know the utf8 is valid, because we just constructed
2966 * it ourselves in previous loop iterations */
2967 min_ptr = (char*) utf8_hop( (U8*) max_ptr, -1);
2968 range_min = valid_utf8_to_uvchr( (U8*) min_ptr, NULL);
2969 range_max = valid_utf8_to_uvchr( (U8*) max_ptr, NULL);
e294cc5d 2970 }
f4240379
KW
2971 else {
2972 min_ptr = max_ptr - 1;
2973 range_min = * (U8*) min_ptr;
2974 range_max = * (U8*) max_ptr;
2975 }
2976
e294cc5d 2977#ifdef EBCDIC
f4240379
KW
2978 /* On EBCDIC platforms, we may have to deal with portable
2979 * ranges. These happen if at least one range endpoint is a
2980 * Unicode value (\N{...}), or if the range is a subset of
2981 * [A-Z] or [a-z], and both ends are literal characters,
2982 * like 'A', and not like \x{C1} */
2983 if ((convert_unicode
2984 = cBOOL(backslash_N) /* \N{} forces Unicode, hence
2985 portable range */
2986 || ( ! non_portable_endpoint
2987 && (( isLOWER_A(range_min) && isLOWER_A(range_max))
2988 || (isUPPER_A(range_min) && isUPPER_A(range_max))))
2989 )) {
2990
2991 /* Special handling is needed for these portable ranges.
2992 * They are defined to all be in Unicode terms, which
2993 * include all Unicode code points between the end points.
2994 * Convert to Unicode to get the Unicode range. Later we
2995 * will convert each code point in the range back to
2996 * native. */
2997 range_min = NATIVE_TO_UNI(range_min);
2998 range_max = NATIVE_TO_UNI(range_max);
2999 }
e294cc5d 3000#endif
8ada0baa 3001
f4240379 3002 if (range_min > range_max) {
11327fa1 3003#ifdef EBCDIC
f4240379
KW
3004 if (convert_unicode) {
3005 /* Need to convert back to native for meaningful
3006 * messages for this platform */
3007 range_min = UNI_TO_NATIVE(range_min);
3008 range_max = UNI_TO_NATIVE(range_max);
3009 }
11327fa1 3010#endif
f4240379
KW
3011
3012 /* Use the characters themselves for the error message if
3013 * ASCII printables; otherwise some visible representation
3014 * of them */
3015 if (isPRINT_A(range_min) && isPRINT_A(range_max)) {
3016 Perl_croak(aTHX_
3017 "Invalid range \"%c-%c\" in transliteration operator",
3018 (char)range_min, (char)range_max);
3019 }
11327fa1 3020#ifdef EBCDIC
f4240379
KW
3021 else if (convert_unicode) {
3022 /* diag_listed_as: Invalid range "%s" in transliteration operator */
3023 Perl_croak(aTHX_
147e3846 3024 "Invalid range \"\\N{U+%04" UVXf "}-\\N{U+%04" UVXf "}\""
f4240379
KW
3025 " in transliteration operator",
3026 range_min, range_max);
3027 }
11327fa1 3028#endif
f4240379
KW
3029 else {
3030 /* diag_listed_as: Invalid range "%s" in transliteration operator */
3031 Perl_croak(aTHX_
147e3846 3032 "Invalid range \"\\x{%04" UVXf "}-\\x{%04" UVXf "}\""
f4240379
KW
3033 " in transliteration operator",
3034 range_min, range_max);
3035 }
c2e66d9e
GS
3036 }
3037
f4240379
KW
3038 if (has_utf8) {
3039
3040 /* We try to avoid creating a swash. If the upper end of
3041 * this range is below 256, this range won't force a swash;
3042 * otherwise it does force a swash, and as long as we have
3043 * to have one, we might as well not expand things out.
3044 * But if it's EBCDIC, we may have to look at each
3045 * character below 256 if we have to convert to/from
3046 * Unicode values */
3047 if (range_max > 255
c7f1f016 3048#ifdef EBCDIC
f4240379 3049 && (range_min > 255 || ! convert_unicode)
8ada0baa 3050#endif
f4240379
KW
3051 ) {
3052 /* Move the high character one byte to the right; then
3053 * insert between it and the range begin, an illegal
3054 * byte which serves to indicate this is a range (using
3055 * a '-' could be ambiguous). */
3056 char *e = d++;
3057 while (e-- > max_ptr) {
3058 *(e + 1) = *e;
e294cc5d 3059 }
f4240379
KW
3060 *(e + 1) = (char) ILLEGAL_UTF8_BYTE;
3061 goto range_done;
3062 }
3063
3064 /* Here, we're going to expand out the range. For EBCDIC
3065 * the range can extend above 255 (not so in ASCII), so
3066 * for EBCDIC, split it into the parts above and below
3067 * 255/256 */
e294cc5d 3068#ifdef EBCDIC
f4240379
KW
3069 if (range_max > 255) {
3070 real_range_max = range_max;
3071 range_max = 255;
3072 }
e294cc5d 3073#endif
f4240379 3074 }
02aa26ce 3075
f4240379
KW
3076 /* Here we need to expand out the string to contain each
3077 * character in the range. Grow the output to handle this */
3078
3079 save_offset = min_ptr - SvPVX_const(sv);
3080
3081 /* The base growth is the number of code points in the range */
3082 grow = range_max - range_min + 1;
3083 if (has_utf8) {
3084
3085 /* But if the output is UTF-8, some of those characters may
3086 * need two bytes (since the maximum range value here is
3087 * 255, the max bytes per character is two). On ASCII
3088 * platforms, it's not much trouble to get an accurate
3089 * count of what's needed. But on EBCDIC, the ones that
3090 * need 2 bytes are scattered around, so just use a worst
3091 * case value instead of calculating for that platform. */
4c3a8340 3092#ifdef EBCDIC
f4240379
KW
3093 grow *= 2;
3094#else
3095 /* Only those above 127 require 2 bytes. This may be
3096 * everything in the range, or not */
3097 if (range_min > 127) {
3098 grow *= 2;
3099 }
3100 else if (range_max > 127) {
3101 grow += range_max - 127;
3102 }
4c3a8340 3103#endif
f4240379
KW
3104 }
3105
3106 /* Subtract 3 for the bytes that were already accounted for
3107 * (min, max, and the hyphen) */
10656159 3108 d = save_offset + SvGROW(sv, SvLEN(sv) + grow - 3);
f4240379 3109
11327fa1
AL
3110#ifdef EBCDIC
3111 /* Here, we expand out the range. */
f4240379
KW
3112 if (convert_unicode) {
3113 IV i;
02aa26ce 3114
f4240379
KW
3115 /* Recall that the min and max are now in Unicode terms, so
3116 * we have to convert each character to its native
3117 * equivalent */
3118 if (has_utf8) {
3119 for (i = range_min; i <= range_max; i++) {
81324705 3120 append_utf8_from_native_byte(LATIN1_TO_NATIVE((U8) i),
f4240379
KW
3121 (U8 **) &d);
3122 }
3123 }
3124 else {
3125 for (i = range_min; i <= range_max; i++) {
81324705 3126 *d++ = (char)LATIN1_TO_NATIVE((U8) i);
f4240379
KW
3127 }
3128 }
01ec43d0 3129 }
11327fa1
AL
3130 else
3131#endif
3132 /* Always gets run for ASCII, and sometimes for EBCDIC. */
3133 {
f4240379
KW
3134 IV i;
3135
3136 /* Here, no conversions are necessary, which means that the
3137 * first character in the range is already in 'd' and
3138 * valid, so we can skip overwriting it */
3139 if (has_utf8) {
3140 d += UTF8SKIP(d);
3141 for (i = range_min + 1; i <= range_max; i++) {
81324705 3142 append_utf8_from_native_byte((U8) i, (U8 **) &d);
f4240379
KW
3143 }
3144 }
3145 else {
3146 d++;
3147 for (i = range_min + 1; i <= range_max; i++) {
3148 *d++ = (char)i;
3149 }
3150 }
a0ed51b3 3151 }
02aa26ce 3152
11327fa1
AL
3153#ifdef EBCDIC
3154 /* If the original range extended above 255, add in that portion. */
f4240379
KW
3155 if (real_range_max) {
3156 *d++ = (char) UTF8_TWO_BYTE_HI(0x100);
3157 *d++ = (char) UTF8_TWO_BYTE_LO(0x100);
3158 if (real_range_max > 0x101)
3159 *d++ = (char) ILLEGAL_UTF8_BYTE;
3160 if (real_range_max > 0x100)
3161 d = (char*)uvchr_to_utf8((U8*)d, real_range_max);
3162 }
11327fa1 3163#endif
02aa26ce 3164
f4240379
KW
3165 range_done:
3166 /* mark the range as done, and continue */
3167 didrange = TRUE;
3168 dorange = FALSE;
3169#ifdef EBCDIC
3170 non_portable_endpoint = 0;
3171 backslash_N = 0;
3172#endif
3173 continue;
3174 } /* End of is a range */
3175 } /* End of transliteration. Joins main code after these else's */
e4a2df84
DM
3176 else if (*s == '[' && PL_lex_inpat && !in_charclass) {
3177 char *s1 = s-1;
3178 int esc = 0;
3179 while (s1 >= start && *s1-- == '\\')
3180 esc = !esc;
3181 if (!esc)
3182 in_charclass = TRUE;
3183 }
2866decb 3184
e4a2df84
DM
3185 else if (*s == ']' && PL_lex_inpat && in_charclass) {
3186 char *s1 = s-1;
3187 int esc = 0;
3188 while (s1 >= start && *s1-- == '\\')
3189 esc = !esc;
3190 if (!esc)
3191 in_charclass = FALSE;
3192 }
2866decb 3193
9da1dd8f
DM
3194 /* skip for regexp comments /(?#comment)/, except for the last
3195 * char, which will be done separately.
3196 * Stop on (?{..}) and friends */
3197
c30fc27b 3198 else if (*s == '(' && PL_lex_inpat && s[1] == '?' && !in_charclass) {
cc6b7395 3199 if (s[2] == '#') {
e994fd66 3200 while (s+1 < send && *s != ')')
5ff03569 3201 *d++ = *s++;
155aba94 3202 }
407f8cf2
KW
3203 else if (!PL_lex_casemods
3204 && ( s[2] == '{' /* This should match regcomp.c */
3205 || (s[2] == '?' && s[3] == '{')))
155aba94 3206 {
9da1dd8f 3207 break;
cc6b7395 3208 }
748a9306 3209 }
02aa26ce
NT
3210
3211 /* likewise skip #-initiated comments in //x patterns */
407f8cf2
KW
3212 else if (*s == '#'
3213 && PL_lex_inpat
3214 && !in_charclass
3215 && ((PMOP*)PL_lex_inpat)->op_pmflags & RXf_PMf_EXTENDED)
3216 {
748a9306 3217 while (s+1 < send && *s != '\n')
5ff03569 3218 *d++ = *s++;
748a9306 3219 }
02aa26ce 3220
9da1dd8f
DM
3221 /* no further processing of single-quoted regex */
3222 else if (PL_lex_inpat && SvIVX(PL_linestr) == '\'')
3223 goto default_action;
3224
5d1d4326 3225 /* check for embedded arrays
da6eedaa 3226 (@foo, @::foo, @'foo, @{foo}, @$foo, @+, @-)
5d1d4326 3227 */
1749ea0d 3228 else if (*s == '@' && s[1]) {
9d58dbc4 3229 if (UTF ? isIDFIRST_utf8((U8*)s+1) : isWORDCHAR_A(s[1]))
1749ea0d
ST
3230 break;
3231 if (strchr(":'{$", s[1]))
3232 break;
3233 if (!PL_lex_inpat && (s[1] == '+' || s[1] == '-'))
3234 break; /* in regexp, neither @+ nor @- are interpolated */
3235 }
02aa26ce
NT
3236
3237 /* check for embedded scalars. only stop if we're sure it's a
3238 variable.
3239 */
79072805 3240 else if (*s == '$') {
3280af22 3241 if (!PL_lex_inpat) /* not a regexp, so $ must be var */
79072805 3242 break;
77772344 3243 if (s + 1 < send && !strchr("()| \r\n\t", s[1])) {
a2a5de95
NC
3244 if (s[1] == '\\') {
3245 Perl_ck_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
3246 "Possible unintended interpolation of $\\ in regex");
77772344 3247 }
79072805 3248 break; /* in regexp, $ might be tail anchor */
77772344 3249 }
79072805 3250 }
02aa26ce 3251
2b9d42f0
NIS
3252 /* End of else if chain - OP_TRANS rejoin rest */
3253
02aa26ce 3254 /* backslashes */
79072805 3255 if (*s == '\\' && s+1 < send) {
ff3f963a
KW
3256 char* e; /* Can be used for ending '}', etc. */
3257
79072805 3258 s++;
02aa26ce 3259
7d0fc23c
KW
3260 /* warn on \1 - \9 in substitution replacements, but note that \11
3261 * is an octal; and \19 is \1 followed by '9' */
407f8cf2
KW
3262 if (PL_lex_inwhat == OP_SUBST
3263 && !PL_lex_inpat
3264 && isDIGIT(*s)
3265 && *s != '0'
3266 && !isDIGIT(s[1]))
79072805 3267 {
c782d7ee 3268 /* diag_listed_as: \%d better written as $%d */
a2a5de95 3269 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX), "\\%c better written as $%c", *s, *s);
79072805
LW
3270 *--s = '$';
3271 break;
3272 }
02aa26ce
NT
3273
3274 /* string-change backslash escapes */
838f2281 3275 if (PL_lex_inwhat != OP_TRANS && *s && strchr("lLuUEQF", *s)) {
79072805
LW
3276 --s;
3277 break;
3278 }
ff3f963a
KW
3279 /* In a pattern, process \N, but skip any other backslash escapes.
3280 * This is because we don't want to translate an escape sequence
3281 * into a meta symbol and have the regex compiler use the meta
3282 * symbol meaning, e.g. \x{2E} would be confused with a dot. But
3283 * in spite of this, we do have to process \N here while the proper
3284 * charnames handler is in scope. See bugs #56444 and #62056.
85fba779 3285 *
ff3f963a
KW
3286 * There is a complication because \N in a pattern may also stand
3287 * for 'match a non-nl', and not mean a charname, in which case its
3288 * processing should be deferred to the regex compiler. To be a
3289 * charname it must be followed immediately by a '{', and not look
3290 * like \N followed by a curly quantifier, i.e., not something like
3291 * \N{3,}. regcurly returns a boolean indicating if it is a legal
3292 * quantifier */
3293 else if (PL_lex_inpat
3294 && (*s != 'N'
3295 || s[1] != '{'
412f55bb 3296 || regcurly(s + 1)))
ff3f963a 3297 {
4d73d076 3298 *d++ = '\\';
cc74c5bd
ST
3299 goto default_action;
3300 }
02aa26ce 3301
79072805 3302 switch (*s) {
79072805 3303 default:
11b8faa4 3304 {
15861f94 3305 if ((isALPHANUMERIC(*s)))
a2a5de95
NC
3306 Perl_ck_warner(aTHX_ packWARN(WARN_MISC),
3307 "Unrecognized escape \\%c passed through",
3308 *s);
11b8faa4 3309 /* default action is to copy the quoted character */
f9a63242 3310 goto default_action;
11b8faa4 3311 }
02aa26ce 3312
632403cc 3313 /* eg. \132 indicates the octal constant 0132 */
79072805
LW
3314 case '0': case '1': case '2': case '3':
3315 case '4': case '5': case '6': case '7':
ba210ebe 3316 {
5e0a247b 3317 I32 flags = PERL_SCAN_SILENT_ILLDIGIT;
53305cf1 3318 STRLEN len = 3;
06972766 3319 uv = grok_oct(s, &len, &flags, NULL);
ba210ebe 3320 s += len;
5e0a247b
KW
3321 if (len < 3 && s < send && isDIGIT(*s)
3322 && ckWARN(WARN_MISC))
3323 {
3324 Perl_warner(aTHX_ packWARN(WARN_MISC),
3325 "%s", form_short_octal_warning(s, len));
3326 }
ba210ebe 3327 }
012bcf8d 3328 goto NUM_ESCAPE_INSERT;
02aa26ce 3329
f0a2b745
KW
3330 /* eg. \o{24} indicates the octal constant \024 */
3331 case 'o':
3332 {
454155d9 3333 const char* error;
f0a2b745 3334
00ce5563 3335 bool valid = grok_bslash_o(&s, &uv, &error,
80f4111b
KW
3336 TRUE, /* Output warning */
3337 FALSE, /* Not strict */
17896a85
KW
3338 TRUE, /* Output warnings for
3339 non-portables */
80f4111b 3340 UTF);
454155d9 3341 if (! valid) {
f0a2b745
KW
3342 yyerror(error);
3343 continue;
3344 }
3345 goto NUM_ESCAPE_INSERT;
3346 }
3347
77a135fe 3348 /* eg. \x24 indicates the hex constant 0x24 */
79072805 3349 case 'x':
a0481293 3350 {
a0481293 3351 const char* error;
355860ce 3352
00ce5563 3353 bool valid = grok_bslash_x(&s, &uv, &error,
80f4111b
KW
3354 TRUE, /* Output warning */
3355 FALSE, /* Not strict */
17896a85
KW
3356 TRUE, /* Output warnings for
3357 non-portables */
80f4111b 3358 UTF);
a0481293
KW
3359 if (! valid) {
3360 yyerror(error);
355860ce 3361 continue;
ba210ebe 3362 }
012bcf8d
GS
3363 }
3364
3365 NUM_ESCAPE_INSERT:
b67fd2c5 3366 /* Insert oct or hex escaped character. */
ba7cea30 3367
06972766 3368 /* Here uv is the ordinal of the next character being added */
f4240379
KW
3369 if (UVCHR_IS_INVARIANT(uv)) {
3370 *d++ = (char) uv;
3371 }
3372 else {
9aa983d2 3373 if (!has_utf8 && uv > 255) {
77a135fe
KW
3374 /* Might need to recode whatever we have accumulated so
3375 * far if it contains any chars variant in utf8 or
3376 * utf-ebcdic. */
3377
3378 SvCUR_set(sv, d - SvPVX_const(sv));
3379 SvPOK_on(sv);
3380 *d = '\0';
77a135fe 3381 /* See Note on sizing above. */
5eef447b 3382 sv_utf8_upgrade_flags_grow(
5f560d8a
KW
3383 sv,
3384 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE
5eef447b
KW
3385 /* Above-latin1 in string
3386 * implies no encoding */
3387 |SV_UTF8_NO_ENCODING,
5f560d8a 3388 UVCHR_SKIP(uv) + (STRLEN)(send - s) + 1);
77a135fe
KW
3389 d = SvPVX(sv) + SvCUR(sv);
3390 has_utf8 = TRUE;
012bcf8d
GS
3391 }
3392
77a135fe 3393 if (has_utf8) {
b67fd2c5
KW
3394 /* Usually, there will already be enough room in 'sv'
3395 * since such escapes are likely longer than any UTF-8
3396 * sequence they can end up as. This isn't the case on
3397 * EBCDIC where \x{40000000} contains 12 bytes, and the
3398 * UTF-8 for it contains 14. And, we have to allow for
3399 * a trailing NUL. It probably can't happen on ASCII
3400 * platforms, but be safe */
3401 const STRLEN needed = d - SvPVX(sv) + UVCHR_SKIP(uv)
3402 + 1;
3403 if (UNLIKELY(needed > SvLEN(sv))) {
3404 SvCUR_set(sv, d - SvPVX_const(sv));
3405 d = sv_grow(sv, needed) + SvCUR(sv);
3406 }
3407
c80e42f3 3408 d = (char*)uvchr_to_utf8((U8*)d, uv);
407f8cf2 3409 if (PL_lex_inwhat == OP_TRANS
7ef70b3d 3410 && PL_parser->lex_sub_op)
407f8cf2 3411 {
7ef70b3d 3412 PL_parser->lex_sub_op->op_private |=
f9a63242
JH
3413 (PL_lex_repl ? OPpTRANS_FROM_UTF
3414 : OPpTRANS_TO_UTF);
f9a63242 3415 }
012bcf8d 3416 }
a0ed51b3 3417 else {
012bcf8d 3418 *d++ = (char)uv;
a0ed51b3 3419 }
012bcf8d 3420 }
f4240379
KW
3421#ifdef EBCDIC
3422 non_portable_endpoint++;
3423#endif
79072805 3424 continue;
02aa26ce 3425
4a2d328f 3426 case 'N':
85fba779
KW
3427 /* In a non-pattern \N must be like \N{U+0041}, or it can be a
3428 * named character, like \N{LATIN SMALL LETTER A}, or a named
3429 * sequence, like \N{LATIN CAPITAL LETTER A WITH MACRON AND
f4240379
KW
3430 * GRAVE} (except y/// can't handle the latter, croaking). For
3431 * convenience all three forms are referred to as "named
3432 * characters" below.
85fba779
KW
3433 *
3434 * For patterns, \N also can mean to match a non-newline. Code
3435 * before this 'switch' statement should already have handled
3436 * this situation, and hence this code only has to deal with
3437 * the named character cases.
3438 *
3439 * For non-patterns, the named characters are converted to
3440 * their string equivalents. In patterns, named characters are
3441 * not converted to their ultimate forms for the same reasons
3442 * that other escapes aren't. Instead, they are converted to
3443 * the \N{U+...} form to get the value from the charnames that
3444 * is in effect right now, while preserving the fact that it
3445 * was a named character, so that the regex compiler knows
3446 * this.
3447 *
3448 * The structure of this section of code (besides checking for
ff3f963a 3449 * errors and upgrading to utf8) is:
f4240379 3450 * If the named character is of the form \N{U+...}, pass it
85fba779
KW
3451 * through if a pattern; otherwise convert the code point
3452 * to utf8
f4240379
KW
3453 * Otherwise must be some \N{NAME}: convert to
3454 * \N{U+c1.c2...} if a pattern; otherwise convert to utf8
3455 *
3456 * Transliteration is an exception. The conversion to utf8 is
3457 * only done if the code point requires it to be representable.
85fba779 3458 *
b6d67071 3459 * Here, 's' points to the 'N'; the test below is guaranteed to
85fba779
KW
3460 * succeed if we are being called on a pattern, as we already
3461 * know from a test above that the next character is a '{'. A
3462 * non-pattern \N must mean 'named character', which requires
3463 * braces */
ff3f963a
KW
3464 s++;
3465 if (*s != '{') {
3466 yyerror("Missing braces on \\N{}");
3467 continue;
3468 }
3469 s++;
3470
0a96133f 3471 /* If there is no matching '}', it is an error. */
ff3f963a
KW
3472 if (! (e = strchr(s, '}'))) {
3473 if (! PL_lex_inpat) {
5777a3f7 3474 yyerror("Missing right brace on \\N{}");
0a96133f 3475 } else {
4407f1b8 3476 yyerror("Missing right brace on \\N{} or unescaped left brace after \\N");
dbc0d4f2 3477 }
0a96133f 3478 continue;
ff3f963a 3479 }
cddc7ef4 3480
ff3f963a 3481 /* Here it looks like a named character */
cddc7ef4 3482
ff3f963a 3483 if (*s == 'U' && s[1] == '+') { /* \N{U+...} */
ff3f963a 3484 s += 2; /* Skip to next char after the 'U+' */
ff3f963a 3485 if (PL_lex_inpat) {
af352bf2
KW
3486
3487 /* In patterns, we can have \N{U+xxxx.yyyy.zzzz...} */
fb2eed93
FC
3488 /* Check the syntax. */
3489 const char *orig_s;
3490 orig_s = s - 5;
3491 if (!isXDIGIT(*s)) {
3492 bad_NU:
3493 yyerror(
3494 "Invalid hexadecimal number in \\N{U+...}"
3495 );
3496 s = e + 1;
3497 continue;
3498 }
3499 while (++s < e) {
3500 if (isXDIGIT(*s))
3501 continue;
3502 else if ((*s == '.' || *s == '_')
3503 && isXDIGIT(s[1]))
3504 continue;
3505 goto bad_NU;
4cbd7e22 3506 }
af352bf2 3507
fb2eed93
FC
3508 /* Pass everything through unchanged.
3509 * +1 is for the '}' */
4cbd7e22
FC
3510 Copy(orig_s, d, e - orig_s + 1, char);
3511 d += e - orig_s + 1;
ff3f963a
KW
3512 }
3513 else { /* Not a pattern: convert the hex to string */
fb2eed93
FC
3514 I32 flags = PERL_SCAN_ALLOW_UNDERSCORES
3515 | PERL_SCAN_SILENT_ILLDIGIT
3516 | PERL_SCAN_DISALLOW_PREFIX;
3517 STRLEN len = e - s;
3518 uv = grok_hex(s, &len, &flags, NULL);
3519 if (len == 0 || (len != (STRLEN)(e - s)))
3520 goto bad_NU;
ff3f963a 3521
f4240379
KW
3522 /* For non-tr///, if the destination is not in utf8,
3523 * unconditionally recode it to be so. This is
3524 * because \N{} implies Unicode semantics, and scalars
3525 * have to be in utf8 to guarantee those semantics.
3526 * tr/// doesn't care about Unicode rules, so no need
3527 * there to upgrade to UTF-8 for small enough code
3528 * points */
3529 if (! has_utf8 && ( uv > 0xFF
3530 || PL_lex_inwhat != OP_TRANS))
3531 {
ff3f963a
KW
3532 SvCUR_set(sv, d - SvPVX_const(sv));
3533 SvPOK_on(sv);
3534 *d = '\0';
3535 /* See Note on sizing above. */
3536 sv_utf8_upgrade_flags_grow(
5f560d8a
KW
3537 sv,
3538 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
b2873c15 3539 OFFUNISKIP(uv) + (STRLEN)(send - e) + 1);
ff3f963a
KW
3540 d = SvPVX(sv) + SvCUR(sv);
3541 has_utf8 = TRUE;
3542 }
3543
a46469e6 3544 /* Add the (Unicode) code point to the output. */
3a34ca0b 3545 if (! has_utf8 || OFFUNI_IS_INVARIANT(uv)) {
a46469e6 3546 *d++ = (char) LATIN1_TO_NATIVE(uv);
ff3f963a 3547 }
a46469e6
KW
3548 else {
3549 d = (char*) uvoffuni_to_utf8_flags((U8*)d, uv, 0);
3550 }
ff3f963a
KW
3551 }
3552 }
6f613c73
KW
3553 else /* Here is \N{NAME} but not \N{U+...}. */
3554 if ((res = get_and_check_backslash_N_name(s, e)))
3555 {
3556 STRLEN len;
3557 const char *str = SvPV_const(res, len);
3558 if (PL_lex_inpat) {
ff3f963a
KW
3559
3560 if (! len) { /* The name resolved to an empty string */
3561 Copy("\\N{}", d, 4, char);
3562 d += 4;
3563 }
3564 else {
3565 /* In order to not lose information for the regex
3566 * compiler, pass the result in the specially made
3567 * syntax: \N{U+c1.c2.c3...}, where c1 etc. are
3568 * the code points in hex of each character
3569 * returned by charnames */
3570
3571 const char *str_end = str + len;
3b721c4f 3572 const STRLEN off = d - SvPVX_const(sv);
94ca1619
KW
3573
3574 if (! SvUTF8(res)) {
3575 /* For the non-UTF-8 case, we can determine the
3576 * exact length needed without having to parse
3577 * through the string. Each character takes up
3578 * 2 hex digits plus either a trailing dot or
3579 * the "}" */
97651d61
KW
3580 const char initial_text[] = "\\N{U+";
3581 const STRLEN initial_len = sizeof(initial_text)
3582 - 1;
94ca1619
KW
3583 d = off + SvGROW(sv, off
3584 + 3 * len
97651d61
KW
3585
3586 /* +1 for trailing NUL */
3587 + initial_len + 1
3588
94ca1619 3589 + (STRLEN)(send - e));
97651d61
KW