This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
Allow space for NUL is UTF-8 array decls
[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" },
7896dde7 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" },
7896dde7 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
7347ee54 459 GCC_DIAG_IGNORE_STMT(-Wformat-nonliteral); /* fmt checked by caller */
b6007c36 460 PerlIO_printf(Perl_debug_log, fmt, pv_display(tmp, s, strlen(s), 0, 60));
7347ee54 461 GCC_DIAG_RESTORE_STMT;
b6007c36
DM
462 SvREFCNT_dec(tmp);
463}
464
8fa7f367
JH
465#endif
466
ffb4593c
NT
467/*
468 * S_ao
469 *
f393a21a
FC
470 * This subroutine looks for an '=' next to the operator that has just been
471 * parsed and turns it into an ASSIGNOP if it finds one.
ffb4593c
NT
472 */
473
76e3520e 474STATIC int
cea2e8a9 475S_ao(pTHX_ int toketype)
a0d0e21e 476{
3280af22
NIS
477 if (*PL_bufptr == '=') {
478 PL_bufptr++;
a0d0e21e 479 if (toketype == ANDAND)
6154021b 480 pl_yylval.ival = OP_ANDASSIGN;
a0d0e21e 481 else if (toketype == OROR)
6154021b 482 pl_yylval.ival = OP_ORASSIGN;
c963b151 483 else if (toketype == DORDOR)
6154021b 484 pl_yylval.ival = OP_DORASSIGN;
a0d0e21e
LW
485 toketype = ASSIGNOP;
486 }
b1764551 487 return REPORT(toketype);
a0d0e21e
LW
488}
489
ffb4593c
NT
490/*
491 * S_no_op
492 * When Perl expects an operator and finds something else, no_op
493 * prints the warning. It always prints "<something> found where
494 * operator expected. It prints "Missing semicolon on previous line?"
495 * if the surprise occurs at the start of the line. "do you need to
496 * predeclare ..." is printed out for code like "sub bar; foo bar $x"
497 * where the compiler doesn't know if foo is a method call or a function.
498 * It prints "Missing operator before end of line" if there's nothing
499 * after the missing operator, or "... before <...>" if there is something
500 * after the missing operator.
488bc579
FC
501 *
502 * PL_bufptr is expected to point to the start of the thing that was found,
503 * and s after the next token or partial token.
ffb4593c
NT
504 */
505
76e3520e 506STATIC void
15f169a1 507S_no_op(pTHX_ const char *const what, char *s)
463ee0b2 508{
9d4ba2ae
AL
509 char * const oldbp = PL_bufptr;
510 const bool is_first = (PL_oldbufptr == PL_linestart);
68dc0745 511
7918f24d
NC
512 PERL_ARGS_ASSERT_NO_OP;
513
1189a94a
GS
514 if (!s)
515 s = oldbp;
07c798fb 516 else
1189a94a 517 PL_bufptr = s;
734ab321 518 yywarn(Perl_form(aTHX_ "%s found where operator expected", what), UTF ? SVf_UTF8 : 0);
56da5a46
RGS
519 if (ckWARN_d(WARN_SYNTAX)) {
520 if (is_first)
521 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
522 "\t(Missing semicolon on previous line?)\n");
fac0f7a3
KW
523 else if (PL_oldoldbufptr && isIDFIRST_lazy_if_safe(PL_oldoldbufptr,
524 PL_bufend,
525 UTF))
526 {
f54cb97a 527 const char *t;
fac0f7a3
KW
528 for (t = PL_oldoldbufptr;
529 (isWORDCHAR_lazy_if_safe(t, PL_bufend, UTF) || *t == ':');
530 t += UTF ? UTF8SKIP(t) : 1)
531 {
c35e046a 532 NOOP;
fac0f7a3 533 }
56da5a46
RGS
534 if (t < PL_bufptr && isSPACE(*t))
535 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
147e3846 536 "\t(Do you need to predeclare %" UTF8f "?)\n",
b17a0679 537 UTF8fARG(UTF, t - PL_oldoldbufptr, PL_oldoldbufptr));
56da5a46
RGS
538 }
539 else {
540 assert(s >= oldbp);
541 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
147e3846 542 "\t(Missing operator before %" UTF8f "?)\n",
b17a0679 543 UTF8fARG(UTF, s - oldbp, oldbp));
56da5a46 544 }
07c798fb 545 }
3280af22 546 PL_bufptr = oldbp;
8990e307
LW
547}
548
ffb4593c
NT
549/*
550 * S_missingterm
551 * Complain about missing quote/regexp/heredoc terminator.
d4c19fe8 552 * If it's called with NULL then it cauterizes the line buffer.
ffb4593c
NT
553 * If we're in a delimited string and the delimiter is a control
554 * character, it's reformatted into a two-char sequence like ^C.
555 * This is fatal.
556 */
557
76e3520e 558STATIC void
1b8d3e0e 559S_missingterm(pTHX_ char *s, STRLEN len)
8990e307 560{
e487ff5e 561 char tmpbuf[UTF8_MAXBYTES + 1];
8990e307 562 char q;
cb650135
FC
563 bool uni = FALSE;
564 SV *sv;
8990e307 565 if (s) {
f88b23d7 566 char * const nl = (char *) my_memrchr(s, '\n', len);
1b8d3e0e
LM
567 if (nl) {
568 *nl = '\0';
569 len = nl - s;
570 }
7f7f7d08 571 uni = UTF;
8990e307 572 }
cb650135 573 else if (PL_multi_close < 32) {
8990e307 574 *tmpbuf = '^';
585ec06d 575 tmpbuf[1] = (char)toCTRL(PL_multi_close);
8990e307
LW
576 tmpbuf[2] = '\0';
577 s = tmpbuf;
1b8d3e0e 578 len = 2;
8990e307
LW
579 }
580 else {
cb650135
FC
581 if (LIKELY(PL_multi_close < 256)) {
582 *tmpbuf = (char)PL_multi_close;
583 tmpbuf[1] = '\0';
1b8d3e0e 584 len = 1;
cb650135
FC
585 }
586 else {
1b8d3e0e
LM
587 char *end = (char *)uvchr_to_utf8((U8 *)tmpbuf, PL_multi_close);
588 *end = '\0';
589 len = end - tmpbuf;
cb650135 590 uni = TRUE;
cb650135 591 }
8990e307
LW
592 s = tmpbuf;
593 }
9f8d27b7 594 q = memchr(s, '"', len) ? '\'' : '"';
1b8d3e0e 595 sv = sv_2mortal(newSVpvn(s, len));
cb650135
FC
596 if (uni)
597 SvUTF8_on(sv);
1b8d3e0e
LM
598 Perl_croak(aTHX_ "Can't find string terminator %c%" SVf "%c"
599 " anywhere before EOF", q, SVfARG(sv), q);
463ee0b2 600}
79072805 601
dd0ac2b9
FC
602#include "feature.h"
603
0d863452 604/*
0d863452
RH
605 * Check whether the named feature is enabled.
606 */
26ea9e12 607bool
3fff3427 608Perl_feature_is_enabled(pTHX_ const char *const name, STRLEN namelen)
0d863452 609{
4a731d7b 610 char he_name[8 + MAX_FEATURE_LEN] = "feature_";
7918f24d
NC
611
612 PERL_ARGS_ASSERT_FEATURE_IS_ENABLED;
ca4d40c4
FC
613
614 assert(CURRENT_FEATURE_BUNDLE == FEATURE_BUNDLE_CUSTOM);
7918f24d 615
26ea9e12
NC
616 if (namelen > MAX_FEATURE_LEN)
617 return FALSE;
3fff3427 618 memcpy(&he_name[8], name, namelen);
7d69d4a6 619
c8ca97b0
NC
620 return cBOOL(cop_hints_fetch_pvn(PL_curcop, he_name, 8 + namelen, 0,
621 REFCOUNTED_HE_EXISTS));
0d863452
RH
622}
623
ffb4593c 624/*
9cbb5ea2
GS
625 * experimental text filters for win32 carriage-returns, utf16-to-utf8 and
626 * utf16-to-utf8-reversed.
ffb4593c
NT
627 */
628
c39cd008
GS
629#ifdef PERL_CR_FILTER
630static void
631strip_return(SV *sv)
632{
eb578fdb
KW
633 const char *s = SvPVX_const(sv);
634 const char * const e = s + SvCUR(sv);
7918f24d
NC
635
636 PERL_ARGS_ASSERT_STRIP_RETURN;
637
c39cd008
GS
638 /* outer loop optimized to do nothing if there are no CR-LFs */
639 while (s < e) {
640 if (*s++ == '\r' && *s == '\n') {
641 /* hit a CR-LF, need to copy the rest */
eb578fdb 642 char *d = s - 1;
c39cd008
GS
643 *d++ = *s++;
644 while (s < e) {
645 if (*s == '\r' && s[1] == '\n')
646 s++;
647 *d++ = *s++;
648 }
649 SvCUR(sv) -= s - d;
650 return;
651 }
652 }
653}
a868473f 654
76e3520e 655STATIC I32
c39cd008 656S_cr_textfilter(pTHX_ int idx, SV *sv, int maxlen)
a868473f 657{
f54cb97a 658 const I32 count = FILTER_READ(idx+1, sv, maxlen);
c39cd008
GS
659 if (count > 0 && !maxlen)
660 strip_return(sv);
661 return count;
a868473f
NIS
662}
663#endif
664
ffb4593c 665/*
8eaa0acf
Z
666=for apidoc Amx|void|lex_start|SV *line|PerlIO *rsfp|U32 flags
667
668Creates and initialises a new lexer/parser state object, supplying
669a context in which to lex and parse from a new source of Perl code.
670A pointer to the new state object is placed in L</PL_parser>. An entry
c83d5090 671is made on the save stack so that upon unwinding, the new state object
8eaa0acf
Z
672will be destroyed and the former value of L</PL_parser> will be restored.
673Nothing else need be done to clean up the parsing context.
674
2d7f6611 675The code to be parsed comes from C<line> and C<rsfp>. C<line>, if
8eaa0acf 676non-null, provides a string (in SV form) containing code to be parsed.
2d7f6611
KW
677A copy of the string is made, so subsequent modification of C<line>
678does not affect parsing. C<rsfp>, if non-null, provides an input stream
8eaa0acf 679from which code will be read to be parsed. If both are non-null, the
2d7f6611
KW
680code in C<line> comes first and must consist of complete lines of input,
681and C<rsfp> supplies the remainder of the source.
8eaa0acf 682
2d7f6611 683The C<flags> parameter is reserved for future use. Currently it is only
e368b3bd 684used by perl internally, so extensions should always pass zero.
8eaa0acf
Z
685
686=cut
687*/
ffb4593c 688
27fcb6ee 689/* LEX_START_SAME_FILTER indicates that this is not a new file, so it
87606032
NC
690 can share filters with the current parser.
691 LEX_START_DONT_CLOSE indicates that the file handle wasn't opened by the
692 caller, hence isn't owned by the parser, so shouldn't be closed on parser
693 destruction. This is used to handle the case of defaulting to reading the
694 script from the standard input because no filename was given on the command
695 line (without getting confused by situation where STDIN has been closed, so
696 the script handle is opened on fd 0) */
27fcb6ee 697
a0d0e21e 698void
8eaa0acf 699Perl_lex_start(pTHX_ SV *line, PerlIO *rsfp, U32 flags)
79072805 700{
6ef55633 701 const char *s = NULL;
5486870f 702 yy_parser *parser, *oparser;
90b58c70 703
60d63348 704 if (flags && flags & ~LEX_START_FLAGS)
8eaa0acf 705 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_start");
acdf0a21
DM
706
707 /* create and initialise a parser */
708
199e78b7 709 Newxz(parser, 1, yy_parser);
5486870f 710 parser->old_parser = oparser = PL_parser;
acdf0a21
DM
711 PL_parser = parser;
712
28ac2b49 713 parser->stack = NULL;
df13534a 714 parser->stack_max1 = NULL;
28ac2b49 715 parser->ps = NULL;
acdf0a21 716
e3abe207
DM
717 /* on scope exit, free this parser and restore any outer one */
718 SAVEPARSER(parser);
7c4baf47 719 parser->saved_curcop = PL_curcop;
e3abe207 720
acdf0a21 721 /* initialise lexer state */
8990e307 722
fb205e7a 723 parser->nexttoke = 0;
ca4cfd28 724 parser->error_count = oparser ? oparser->error_count : 0;
7f1c3e8c 725 parser->copline = parser->preambling = NOLINE;
5afb0a62 726 parser->lex_state = LEX_NORMAL;
c2598295 727 parser->expect = XSTATE;
2f9285f8 728 parser->rsfp = rsfp;
efa571ab 729 parser->recheck_utf8_validity = FALSE;
27fcb6ee
FC
730 parser->rsfp_filters =
731 !(flags & LEX_START_SAME_FILTER) || !oparser
d3cd8e11
FC
732 ? NULL
733 : MUTABLE_AV(SvREFCNT_inc(
734 oparser->rsfp_filters
735 ? oparser->rsfp_filters
736 : (oparser->rsfp_filters = newAV())
737 ));
2f9285f8 738
199e78b7
DM
739 Newx(parser->lex_brackstack, 120, char);
740 Newx(parser->lex_casestack, 12, char);
741 *parser->lex_casestack = '\0';
d794b522 742 Newxz(parser->lex_shared, 1, LEXSHARED);
02b34bbe 743
10efb74f 744 if (line) {
0528fd32 745 STRLEN len;
a6909bd2
KW
746 const U8* first_bad_char_loc;
747
10efb74f 748 s = SvPV_const(line, len);
90b58c70 749
0aab20f2
KW
750 if ( SvUTF8(line)
751 && UNLIKELY(! is_utf8_string_loc((U8 *) s,
752 SvCUR(line),
753 &first_bad_char_loc)))
90b58c70
KW
754 {
755 _force_out_malformed_utf8_message(first_bad_char_loc,
756 (U8 *) s + SvCUR(line),
757 0,
758 1 /* 1 means die */ );
759 NOT_REACHED; /* NOTREACHED */
760 }
761
0abcdfa4
FC
762 parser->linestr = flags & LEX_START_COPIED
763 ? SvREFCNT_inc_simple_NN(line)
764 : newSVpvn_flags(s, len, SvUTF8(line));
b3dd0aba
FC
765 if (!rsfp)
766 sv_catpvs(parser->linestr, "\n;");
0abcdfa4 767 } else {
bf1b738b 768 parser->linestr = newSVpvn("\n;", rsfp ? 1 : 2);
8990e307 769 }
218304f9 770
f06b5848
DM
771 parser->oldoldbufptr =
772 parser->oldbufptr =
773 parser->bufptr =
774 parser->linestart = SvPVX(parser->linestr);
775 parser->bufend = parser->bufptr + SvCUR(parser->linestr);
776 parser->last_lop = parser->last_uni = NULL;
b54f893d 777
6d59e610 778 STATIC_ASSERT_STMT(FITS_IN_8_BITS(LEX_IGNORE_UTF8_HINTS|LEX_EVALBYTES
b54f893d
KW
779 |LEX_DONT_CLOSE_RSFP));
780 parser->lex_flags = (U8) (flags & (LEX_IGNORE_UTF8_HINTS|LEX_EVALBYTES
781 |LEX_DONT_CLOSE_RSFP));
737c24fc 782
60d63348 783 parser->in_pod = parser->filtered = 0;
79072805 784}
a687059c 785
e3abe207
DM
786
787/* delete a parser object */
788
789void
790Perl_parser_free(pTHX_ const yy_parser *parser)
791{
7918f24d
NC
792 PERL_ARGS_ASSERT_PARSER_FREE;
793
7c4baf47 794 PL_curcop = parser->saved_curcop;
bdc0bf6f
DM
795 SvREFCNT_dec(parser->linestr);
796
87606032 797 if (PL_parser->lex_flags & LEX_DONT_CLOSE_RSFP)
2f9285f8 798 PerlIO_clearerr(parser->rsfp);
407f8cf2
KW
799 else if (parser->rsfp && (!parser->old_parser
800 || (parser->old_parser && parser->rsfp != parser->old_parser->rsfp)))
2f9285f8 801 PerlIO_close(parser->rsfp);
5486870f 802 SvREFCNT_dec(parser->rsfp_filters);
10002bc1 803 SvREFCNT_dec(parser->lex_stuff);
7ef70b3d 804 SvREFCNT_dec(parser->lex_sub_repl);
3ac7ff8f
FC
805
806 Safefree(parser->lex_brackstack);
807 Safefree(parser->lex_casestack);
808 Safefree(parser->lex_shared);
809 PL_parser = parser->old_parser;
810 Safefree(parser);
811}
812
813void
814Perl_parser_free_nexttoke_ops(pTHX_ yy_parser *parser, OPSLAB *slab)
815{
3ac7ff8f 816 I32 nexttoke = parser->nexttoke;
3ac7ff8f 817 PERL_ARGS_ASSERT_PARSER_FREE_NEXTTOKE_OPS;
3ce3dcd9 818 while (nexttoke--) {
3ac7ff8f
FC
819 if (S_is_opval_token(parser->nexttype[nexttoke] & 0xffff)
820 && parser->nextval[nexttoke].opval
821 && parser->nextval[nexttoke].opval->op_slabbed
822 && OpSLAB(parser->nextval[nexttoke].opval) == slab) {
3ce3dcd9 823 op_free(parser->nextval[nexttoke].opval);
3ac7ff8f
FC
824 parser->nextval[nexttoke].opval = NULL;
825 }
3ce3dcd9 826 }
e3abe207
DM
827}
828
829
ffb4593c 830/*
f0e67a1d
Z
831=for apidoc AmxU|SV *|PL_parser-E<gt>linestr
832
833Buffer scalar containing the chunk currently under consideration of the
834text currently being lexed. This is always a plain string scalar (for
835which C<SvPOK> is true). It is not intended to be used as a scalar by
836normal scalar means; instead refer to the buffer directly by the pointer
837variables described below.
838
839The lexer maintains various C<char*> pointers to things in the
840C<PL_parser-E<gt>linestr> buffer. If C<PL_parser-E<gt>linestr> is ever
841reallocated, all of these pointers must be updated. Don't attempt to
842do this manually, but rather use L</lex_grow_linestr> if you need to
843reallocate the buffer.
844
845The content of the text chunk in the buffer is commonly exactly one
846complete line of input, up to and including a newline terminator,
847but there are situations where it is otherwise. The octets of the
848buffer may be intended to be interpreted as either UTF-8 or Latin-1.
849The function L</lex_bufutf8> tells you which. Do not use the C<SvUTF8>
850flag on this scalar, which may disagree with it.
851
852For direct examination of the buffer, the variable
853L</PL_parser-E<gt>bufend> points to the end of the buffer. The current
854lexing position is pointed to by L</PL_parser-E<gt>bufptr>. Direct use
855of these pointers is usually preferable to examination of the scalar
856through normal scalar means.
857
858=for apidoc AmxU|char *|PL_parser-E<gt>bufend
859
860Direct pointer to the end of the chunk of text currently being lexed, the
861end of the lexer buffer. This is equal to C<SvPVX(PL_parser-E<gt>linestr)
6602b933 862+ SvCUR(PL_parser-E<gt>linestr)>. A C<NUL> character (zero octet) is
f0e67a1d
Z
863always located at the end of the buffer, and does not count as part of
864the buffer's contents.
865
866=for apidoc AmxU|char *|PL_parser-E<gt>bufptr
867
868Points to the current position of lexing inside the lexer buffer.
869Characters around this point may be freely examined, within
870the range delimited by C<SvPVX(L</PL_parser-E<gt>linestr>)> and
871L</PL_parser-E<gt>bufend>. The octets of the buffer may be intended to be
872interpreted as either UTF-8 or Latin-1, as indicated by L</lex_bufutf8>.
873
874Lexing code (whether in the Perl core or not) moves this pointer past
875the characters that it consumes. It is also expected to perform some
876bookkeeping whenever a newline character is consumed. This movement
877can be more conveniently performed by the function L</lex_read_to>,
878which handles newlines appropriately.
879
880Interpretation of the buffer's octets can be abstracted out by
881using the slightly higher-level functions L</lex_peek_unichar> and
882L</lex_read_unichar>.
883
884=for apidoc AmxU|char *|PL_parser-E<gt>linestart
885
886Points to the start of the current line inside the lexer buffer.
887This is useful for indicating at which column an error occurred, and
888not much else. This must be updated by any lexing code that consumes
889a newline; the function L</lex_read_to> handles this detail.
890
891=cut
892*/
893
894/*
895=for apidoc Amx|bool|lex_bufutf8
896
897Indicates whether the octets in the lexer buffer
898(L</PL_parser-E<gt>linestr>) should be interpreted as the UTF-8 encoding
899of Unicode characters. If not, they should be interpreted as Latin-1
900characters. This is analogous to the C<SvUTF8> flag for scalars.
901
902In UTF-8 mode, it is not guaranteed that the lexer buffer actually
903contains valid UTF-8. Lexing code must be robust in the face of invalid
904encoding.
905
906The actual C<SvUTF8> flag of the L</PL_parser-E<gt>linestr> scalar
907is significant, but not the whole story regarding the input character
908encoding. Normally, when a file is being read, the scalar contains octets
909and its C<SvUTF8> flag is off, but the octets should be interpreted as
910UTF-8 if the C<use utf8> pragma is in effect. During a string eval,
911however, the scalar may have the C<SvUTF8> flag on, and in this case its
912octets should be interpreted as UTF-8 unless the C<use bytes> pragma
913is in effect. This logic may change in the future; use this function
914instead of implementing the logic yourself.
915
916=cut
917*/
918
919bool
920Perl_lex_bufutf8(pTHX)
921{
922 return UTF;
923}
924
925/*
926=for apidoc Amx|char *|lex_grow_linestr|STRLEN len
927
928Reallocates the lexer buffer (L</PL_parser-E<gt>linestr>) to accommodate
2d7f6611 929at least C<len> octets (including terminating C<NUL>). Returns a
f0e67a1d
Z
930pointer to the reallocated buffer. This is necessary before making
931any direct modification of the buffer that would increase its length.
932L</lex_stuff_pvn> provides a more convenient way to insert text into
933the buffer.
934
935Do not use C<SvGROW> or C<sv_grow> directly on C<PL_parser-E<gt>linestr>;
936this function updates all of the lexer's variables that point directly
937into the buffer.
938
939=cut
940*/
941
942char *
943Perl_lex_grow_linestr(pTHX_ STRLEN len)
944{
945 SV *linestr;
946 char *buf;
947 STRLEN bufend_pos, bufptr_pos, oldbufptr_pos, oldoldbufptr_pos;
c7641931 948 STRLEN linestart_pos, last_uni_pos, last_lop_pos, re_eval_start_pos;
98d5e3ef
DM
949 bool current;
950
f0e67a1d
Z
951 linestr = PL_parser->linestr;
952 buf = SvPVX(linestr);
953 if (len <= SvLEN(linestr))
954 return buf;
98d5e3ef
DM
955
956 /* Is the lex_shared linestr SV the same as the current linestr SV?
957 * Only in this case does re_eval_start need adjusting, since it
958 * points within lex_shared->ls_linestr's buffer */
b1b8fb6a
DM
959 current = ( !PL_parser->lex_shared->ls_linestr
960 || linestr == PL_parser->lex_shared->ls_linestr);
98d5e3ef 961
f0e67a1d
Z
962 bufend_pos = PL_parser->bufend - buf;
963 bufptr_pos = PL_parser->bufptr - buf;
964 oldbufptr_pos = PL_parser->oldbufptr - buf;
965 oldoldbufptr_pos = PL_parser->oldoldbufptr - buf;
966 linestart_pos = PL_parser->linestart - buf;
967 last_uni_pos = PL_parser->last_uni ? PL_parser->last_uni - buf : 0;
968 last_lop_pos = PL_parser->last_lop ? PL_parser->last_lop - buf : 0;
98d5e3ef 969 re_eval_start_pos = (current && PL_parser->lex_shared->re_eval_start) ?
3328ab5a 970 PL_parser->lex_shared->re_eval_start - buf : 0;
c7641931 971
f0e67a1d 972 buf = sv_grow(linestr, len);
c7641931 973
f0e67a1d
Z
974 PL_parser->bufend = buf + bufend_pos;
975 PL_parser->bufptr = buf + bufptr_pos;
976 PL_parser->oldbufptr = buf + oldbufptr_pos;
977 PL_parser->oldoldbufptr = buf + oldoldbufptr_pos;
978 PL_parser->linestart = buf + linestart_pos;
979 if (PL_parser->last_uni)
980 PL_parser->last_uni = buf + last_uni_pos;
981 if (PL_parser->last_lop)
982 PL_parser->last_lop = buf + last_lop_pos;
98d5e3ef 983 if (current && PL_parser->lex_shared->re_eval_start)
3328ab5a 984 PL_parser->lex_shared->re_eval_start = buf + re_eval_start_pos;
f0e67a1d
Z
985 return buf;
986}
987
988/*
83aa740e 989=for apidoc Amx|void|lex_stuff_pvn|const char *pv|STRLEN len|U32 flags
f0e67a1d
Z
990
991Insert characters into the lexer buffer (L</PL_parser-E<gt>linestr>),
992immediately after the current lexing point (L</PL_parser-E<gt>bufptr>),
993reallocating the buffer if necessary. This means that lexing code that
994runs later will see the characters as if they had appeared in the input.
995It is not recommended to do this as part of normal parsing, and most
996uses of this facility run the risk of the inserted characters being
997interpreted in an unintended manner.
998
2d7f6611
KW
999The string to be inserted is represented by C<len> octets starting
1000at C<pv>. These octets are interpreted as either UTF-8 or Latin-1,
1001according to whether the C<LEX_STUFF_UTF8> flag is set in C<flags>.
f0e67a1d
Z
1002The characters are recoded for the lexer buffer, according to how the
1003buffer is currently being interpreted (L</lex_bufutf8>). If a string
9dcc53ea 1004to be inserted is available as a Perl scalar, the L</lex_stuff_sv>
f0e67a1d
Z
1005function is more convenient.
1006
1007=cut
1008*/
1009
1010void
83aa740e 1011Perl_lex_stuff_pvn(pTHX_ const char *pv, STRLEN len, U32 flags)
f0e67a1d 1012{
749123ff 1013 dVAR;
f0e67a1d
Z
1014 char *bufptr;
1015 PERL_ARGS_ASSERT_LEX_STUFF_PVN;
1016 if (flags & ~(LEX_STUFF_UTF8))
1017 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_stuff_pvn");
1018 if (UTF) {
1019 if (flags & LEX_STUFF_UTF8) {
1020 goto plain_copy;
1021 } else {
54d004e8 1022 STRLEN highhalf = 0; /* Count of variants */
83aa740e 1023 const char *p, *e = pv+len;
54d004e8
KW
1024 for (p = pv; p != e; p++) {
1025 if (! UTF8_IS_INVARIANT(*p)) {
1026 highhalf++;
1027 }
1028 }
f0e67a1d
Z
1029 if (!highhalf)
1030 goto plain_copy;
1031 lex_grow_linestr(SvCUR(PL_parser->linestr)+1+len+highhalf);
1032 bufptr = PL_parser->bufptr;
1033 Move(bufptr, bufptr+len+highhalf, PL_parser->bufend+1-bufptr, char);
255fdf19
Z
1034 SvCUR_set(PL_parser->linestr,
1035 SvCUR(PL_parser->linestr) + len+highhalf);
f0e67a1d
Z
1036 PL_parser->bufend += len+highhalf;
1037 for (p = pv; p != e; p++) {
27d0a47c 1038 append_utf8_from_native_byte(*p, (U8 **) &bufptr);
f0e67a1d
Z
1039 }
1040 }
1041 } else {
1042 if (flags & LEX_STUFF_UTF8) {
1043 STRLEN highhalf = 0;
83aa740e 1044 const char *p, *e = pv+len;
f0e67a1d
Z
1045 for (p = pv; p != e; p++) {
1046 U8 c = (U8)*p;
54d004e8 1047 if (UTF8_IS_ABOVE_LATIN1(c)) {
f0e67a1d
Z
1048 Perl_croak(aTHX_ "Lexing code attempted to stuff "
1049 "non-Latin-1 character into Latin-1 input");
54d004e8 1050 } else if (UTF8_IS_NEXT_CHAR_DOWNGRADEABLE(p, e)) {
f0e67a1d
Z
1051 p++;
1052 highhalf++;
f88c6466 1053 } else assert(UTF8_IS_INVARIANT(c));
f0e67a1d
Z
1054 }
1055 if (!highhalf)
1056 goto plain_copy;
1057 lex_grow_linestr(SvCUR(PL_parser->linestr)+1+len-highhalf);
1058 bufptr = PL_parser->bufptr;
1059 Move(bufptr, bufptr+len-highhalf, PL_parser->bufend+1-bufptr, char);
255fdf19
Z
1060 SvCUR_set(PL_parser->linestr,
1061 SvCUR(PL_parser->linestr) + len-highhalf);
f0e67a1d 1062 PL_parser->bufend += len-highhalf;
54d004e8
KW
1063 p = pv;
1064 while (p < e) {
1065 if (UTF8_IS_INVARIANT(*p)) {
1066 *bufptr++ = *p;
1067 p++;
f0e67a1d 1068 }
54d004e8
KW
1069 else {
1070 assert(p < e -1 );
a62b247b 1071 *bufptr++ = EIGHT_BIT_UTF8_TO_NATIVE(*p, *(p+1));
54d004e8
KW
1072 p += 2;
1073 }
f0e67a1d
Z
1074 }
1075 } else {
54d004e8 1076 plain_copy:
f0e67a1d
Z
1077 lex_grow_linestr(SvCUR(PL_parser->linestr)+1+len);
1078 bufptr = PL_parser->bufptr;
1079 Move(bufptr, bufptr+len, PL_parser->bufend+1-bufptr, char);
255fdf19 1080 SvCUR_set(PL_parser->linestr, SvCUR(PL_parser->linestr) + len);
f0e67a1d
Z
1081 PL_parser->bufend += len;
1082 Copy(pv, bufptr, len, char);
1083 }
1084 }
1085}
1086
1087/*
9dcc53ea
Z
1088=for apidoc Amx|void|lex_stuff_pv|const char *pv|U32 flags
1089
1090Insert characters into the lexer buffer (L</PL_parser-E<gt>linestr>),
1091immediately after the current lexing point (L</PL_parser-E<gt>bufptr>),
1092reallocating the buffer if necessary. This means that lexing code that
1093runs later will see the characters as if they had appeared in the input.
1094It is not recommended to do this as part of normal parsing, and most
1095uses of this facility run the risk of the inserted characters being
1096interpreted in an unintended manner.
1097
2d7f6611 1098The string to be inserted is represented by octets starting at C<pv>
9dcc53ea
Z
1099and continuing to the first nul. These octets are interpreted as either
1100UTF-8 or Latin-1, according to whether the C<LEX_STUFF_UTF8> flag is set
2d7f6611 1101in C<flags>. The characters are recoded for the lexer buffer, according
9dcc53ea
Z
1102to how the buffer is currently being interpreted (L</lex_bufutf8>).
1103If it is not convenient to nul-terminate a string to be inserted, the
1104L</lex_stuff_pvn> function is more appropriate.
1105
1106=cut
1107*/
1108
1109void
1110Perl_lex_stuff_pv(pTHX_ const char *pv, U32 flags)
1111{
1112 PERL_ARGS_ASSERT_LEX_STUFF_PV;
1113 lex_stuff_pvn(pv, strlen(pv), flags);
1114}
1115
1116/*
f0e67a1d
Z
1117=for apidoc Amx|void|lex_stuff_sv|SV *sv|U32 flags
1118
1119Insert characters into the lexer buffer (L</PL_parser-E<gt>linestr>),
1120immediately after the current lexing point (L</PL_parser-E<gt>bufptr>),
1121reallocating the buffer if necessary. This means that lexing code that
1122runs later will see the characters as if they had appeared in the input.
1123It is not recommended to do this as part of normal parsing, and most
1124uses of this facility run the risk of the inserted characters being
1125interpreted in an unintended manner.
1126
2d7f6611 1127The string to be inserted is the string value of C<sv>. The characters
f0e67a1d 1128are recoded for the lexer buffer, according to how the buffer is currently
9dcc53ea 1129being interpreted (L</lex_bufutf8>). If a string to be inserted is
f0e67a1d
Z
1130not already a Perl scalar, the L</lex_stuff_pvn> function avoids the
1131need to construct a scalar.
1132
1133=cut
1134*/
1135
1136void
1137Perl_lex_stuff_sv(pTHX_ SV *sv, U32 flags)
1138{
1139 char *pv;
1140 STRLEN len;
1141 PERL_ARGS_ASSERT_LEX_STUFF_SV;
1142 if (flags)
1143 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_stuff_sv");
1144 pv = SvPV(sv, len);
1145 lex_stuff_pvn(pv, len, flags | (SvUTF8(sv) ? LEX_STUFF_UTF8 : 0));
1146}
1147
1148/*
1149=for apidoc Amx|void|lex_unstuff|char *ptr
1150
1151Discards text about to be lexed, from L</PL_parser-E<gt>bufptr> up to
2d7f6611 1152C<ptr>. Text following C<ptr> will be moved, and the buffer shortened.
f0e67a1d
Z
1153This hides the discarded text from any lexing code that runs later,
1154as if the text had never appeared.
1155
1156This is not the normal way to consume lexed text. For that, use
1157L</lex_read_to>.
1158
1159=cut
1160*/
1161
1162void
1163Perl_lex_unstuff(pTHX_ char *ptr)
1164{
1165 char *buf, *bufend;
1166 STRLEN unstuff_len;
1167 PERL_ARGS_ASSERT_LEX_UNSTUFF;
1168 buf = PL_parser->bufptr;
1169 if (ptr < buf)
1170 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_unstuff");
1171 if (ptr == buf)
1172 return;
1173 bufend = PL_parser->bufend;
1174 if (ptr > bufend)
1175 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_unstuff");
1176 unstuff_len = ptr - buf;
1177 Move(ptr, buf, bufend+1-ptr, char);
1178 SvCUR_set(PL_parser->linestr, SvCUR(PL_parser->linestr) - unstuff_len);
1179 PL_parser->bufend = bufend - unstuff_len;
1180}
1181
1182/*
1183=for apidoc Amx|void|lex_read_to|char *ptr
1184
1185Consume text in the lexer buffer, from L</PL_parser-E<gt>bufptr> up
2d7f6611 1186to C<ptr>. This advances L</PL_parser-E<gt>bufptr> to match C<ptr>,
f0e67a1d
Z
1187performing the correct bookkeeping whenever a newline character is passed.
1188This is the normal way to consume lexed text.
1189
1190Interpretation of the buffer's octets can be abstracted out by
1191using the slightly higher-level functions L</lex_peek_unichar> and
1192L</lex_read_unichar>.
1193
1194=cut
1195*/
1196
1197void
1198Perl_lex_read_to(pTHX_ char *ptr)
1199{
1200 char *s;
1201 PERL_ARGS_ASSERT_LEX_READ_TO;
1202 s = PL_parser->bufptr;
1203 if (ptr < s || ptr > PL_parser->bufend)
1204 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_read_to");
1205 for (; s != ptr; s++)
1206 if (*s == '\n') {
83944c01 1207 COPLINE_INC_WITH_HERELINES;
f0e67a1d
Z
1208 PL_parser->linestart = s+1;
1209 }
1210 PL_parser->bufptr = ptr;
1211}
1212
1213/*
1214=for apidoc Amx|void|lex_discard_to|char *ptr
1215
1216Discards the first part of the L</PL_parser-E<gt>linestr> buffer,
2d7f6611
KW
1217up to C<ptr>. The remaining content of the buffer will be moved, and
1218all pointers into the buffer updated appropriately. C<ptr> must not
f0e67a1d
Z
1219be later in the buffer than the position of L</PL_parser-E<gt>bufptr>:
1220it is not permitted to discard text that has yet to be lexed.
1221
1222Normally it is not necessarily to do this directly, because it suffices to
1223use the implicit discarding behaviour of L</lex_next_chunk> and things
1224based on it. However, if a token stretches across multiple lines,
1f317c95 1225and the lexing code has kept multiple lines of text in the buffer for
f0e67a1d
Z
1226that purpose, then after completion of the token it would be wise to
1227explicitly discard the now-unneeded earlier lines, to avoid future
1228multi-line tokens growing the buffer without bound.
1229
1230=cut
1231*/
1232
1233void
1234Perl_lex_discard_to(pTHX_ char *ptr)
1235{
1236 char *buf;
1237 STRLEN discard_len;
1238 PERL_ARGS_ASSERT_LEX_DISCARD_TO;
1239 buf = SvPVX(PL_parser->linestr);
1240 if (ptr < buf)
1241 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_discard_to");
1242 if (ptr == buf)
1243 return;
1244 if (ptr > PL_parser->bufptr)
1245 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_discard_to");
1246 discard_len = ptr - buf;
1247 if (PL_parser->oldbufptr < ptr)
1248 PL_parser->oldbufptr = ptr;
1249 if (PL_parser->oldoldbufptr < ptr)
1250 PL_parser->oldoldbufptr = ptr;
1251 if (PL_parser->last_uni && PL_parser->last_uni < ptr)
1252 PL_parser->last_uni = NULL;
1253 if (PL_parser->last_lop && PL_parser->last_lop < ptr)
1254 PL_parser->last_lop = NULL;
1255 Move(ptr, buf, PL_parser->bufend+1-ptr, char);
1256 SvCUR_set(PL_parser->linestr, SvCUR(PL_parser->linestr) - discard_len);
1257 PL_parser->bufend -= discard_len;
1258 PL_parser->bufptr -= discard_len;
1259 PL_parser->oldbufptr -= discard_len;
1260 PL_parser->oldoldbufptr -= discard_len;
1261 if (PL_parser->last_uni)
1262 PL_parser->last_uni -= discard_len;
1263 if (PL_parser->last_lop)
1264 PL_parser->last_lop -= discard_len;
1265}
1266
efa571ab
KW
1267void
1268Perl_notify_parser_that_changed_to_utf8(pTHX)
1269{
1270 /* Called when $^H is changed to indicate that HINT_UTF8 has changed from
1271 * off to on. At compile time, this has the effect of entering a 'use
1272 * utf8' section. This means that any input was not previously checked for
1273 * UTF-8 (because it was off), but now we do need to check it, or our
1274 * assumptions about the input being sane could be wrong, and we could
1275 * segfault. This routine just sets a flag so that the next time we look
1276 * at the input we do the well-formed UTF-8 check. If we aren't in the
1277 * proper phase, there may not be a parser object, but if there is, setting
1278 * the flag is harmless */
1279
1280 if (PL_parser) {
1281 PL_parser->recheck_utf8_validity = TRUE;
1282 }
1283}
1284
f0e67a1d
Z
1285/*
1286=for apidoc Amx|bool|lex_next_chunk|U32 flags
1287
1288Reads in the next chunk of text to be lexed, appending it to
1289L</PL_parser-E<gt>linestr>. This should be called when lexing code has
1290looked to the end of the current chunk and wants to know more. It is
1291usual, but not necessary, for lexing to have consumed the entirety of
1292the current chunk at this time.
1293
1294If L</PL_parser-E<gt>bufptr> is pointing to the very end of the current
1295chunk (i.e., the current chunk has been entirely consumed), normally the
1296current chunk will be discarded at the same time that the new chunk is
c5608a1f 1297read in. If C<flags> has the C<LEX_KEEP_PREVIOUS> bit set, the current chunk
f0e67a1d
Z
1298will not be discarded. If the current chunk has not been entirely
1299consumed, then it will not be discarded regardless of the flag.
1300
1301Returns true if some new text was added to the buffer, or false if the
1302buffer has reached the end of the input text.
1303
1304=cut
1305*/
1306
1307#define LEX_FAKE_EOF 0x80000000
e47d32dc 1308#define LEX_NO_TERM 0x40000000 /* here-doc */
f0e67a1d
Z
1309
1310bool
1311Perl_lex_next_chunk(pTHX_ U32 flags)
1312{
1313 SV *linestr;
1314 char *buf;
1315 STRLEN old_bufend_pos, new_bufend_pos;
1316 STRLEN bufptr_pos, oldbufptr_pos, oldoldbufptr_pos;
1317 STRLEN linestart_pos, last_uni_pos, last_lop_pos;
17cc9359 1318 bool got_some_for_debugger = 0;
f0e67a1d 1319 bool got_some;
6cdc5cd8 1320
112d1284 1321 if (flags & ~(LEX_KEEP_PREVIOUS|LEX_FAKE_EOF|LEX_NO_TERM))
f0e67a1d 1322 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_next_chunk");
d27f4b91 1323 if (!(flags & LEX_NO_TERM) && PL_lex_inwhat)
e47d32dc 1324 return FALSE;
f0e67a1d
Z
1325 linestr = PL_parser->linestr;
1326 buf = SvPVX(linestr);
407f8cf2
KW
1327 if (!(flags & LEX_KEEP_PREVIOUS)
1328 && PL_parser->bufptr == PL_parser->bufend)
1329 {
f0e67a1d
Z
1330 old_bufend_pos = bufptr_pos = oldbufptr_pos = oldoldbufptr_pos = 0;
1331 linestart_pos = 0;
1332 if (PL_parser->last_uni != PL_parser->bufend)
1333 PL_parser->last_uni = NULL;
1334 if (PL_parser->last_lop != PL_parser->bufend)
1335 PL_parser->last_lop = NULL;
1336 last_uni_pos = last_lop_pos = 0;
1337 *buf = 0;
1338 SvCUR(linestr) = 0;
1339 } else {
1340 old_bufend_pos = PL_parser->bufend - buf;
1341 bufptr_pos = PL_parser->bufptr - buf;
1342 oldbufptr_pos = PL_parser->oldbufptr - buf;
1343 oldoldbufptr_pos = PL_parser->oldoldbufptr - buf;
1344 linestart_pos = PL_parser->linestart - buf;
1345 last_uni_pos = PL_parser->last_uni ? PL_parser->last_uni - buf : 0;
1346 last_lop_pos = PL_parser->last_lop ? PL_parser->last_lop - buf : 0;
1347 }
1348 if (flags & LEX_FAKE_EOF) {
1349 goto eof;
60d63348 1350 } else if (!PL_parser->rsfp && !PL_parser->filtered) {
f0e67a1d
Z
1351 got_some = 0;
1352 } else if (filter_gets(linestr, old_bufend_pos)) {
1353 got_some = 1;
17cc9359 1354 got_some_for_debugger = 1;
112d1284
FC
1355 } else if (flags & LEX_NO_TERM) {
1356 got_some = 0;
f0e67a1d 1357 } else {
580561a3 1358 if (!SvPOK(linestr)) /* can get undefined by filter_gets */
847cc851 1359 SvPVCLEAR(linestr);
f0e67a1d
Z
1360 eof:
1361 /* End of real input. Close filehandle (unless it was STDIN),
1362 * then add implicit termination.
1363 */
87606032 1364 if (PL_parser->lex_flags & LEX_DONT_CLOSE_RSFP)
f0e67a1d
Z
1365 PerlIO_clearerr(PL_parser->rsfp);
1366 else if (PL_parser->rsfp)
1367 (void)PerlIO_close(PL_parser->rsfp);
1368 PL_parser->rsfp = NULL;
60d63348 1369 PL_parser->in_pod = PL_parser->filtered = 0;
f0e67a1d
Z
1370 if (!PL_in_eval && PL_minus_p) {
1371 sv_catpvs(linestr,
1372 /*{*/";}continue{print or die qq(-p destination: $!\\n);}");
1373 PL_minus_n = PL_minus_p = 0;
1374 } else if (!PL_in_eval && PL_minus_n) {
1375 sv_catpvs(linestr, /*{*/";}");
1376 PL_minus_n = 0;
1377 } else
1378 sv_catpvs(linestr, ";");
1379 got_some = 1;
1380 }
1381 buf = SvPVX(linestr);
1382 new_bufend_pos = SvCUR(linestr);
1383 PL_parser->bufend = buf + new_bufend_pos;
1384 PL_parser->bufptr = buf + bufptr_pos;
6cdc5cd8 1385
07337b95
KW
1386 if (UTF) {
1387 const U8* first_bad_char_loc;
1388 if (UNLIKELY(! is_utf8_string_loc(
1389 (U8 *) PL_parser->bufptr,
1390 PL_parser->bufend - PL_parser->bufptr,
1391 &first_bad_char_loc)))
1392 {
1393 _force_out_malformed_utf8_message(first_bad_char_loc,
1394 (U8 *) PL_parser->bufend,
1395 0,
1396 1 /* 1 means die */ );
1397 NOT_REACHED; /* NOTREACHED */
1398 }
6cdc5cd8
KW
1399 }
1400
f0e67a1d
Z
1401 PL_parser->oldbufptr = buf + oldbufptr_pos;
1402 PL_parser->oldoldbufptr = buf + oldoldbufptr_pos;
1403 PL_parser->linestart = buf + linestart_pos;
1404 if (PL_parser->last_uni)
1405 PL_parser->last_uni = buf + last_uni_pos;
1406 if (PL_parser->last_lop)
1407 PL_parser->last_lop = buf + last_lop_pos;
7f1c3e8c
FC
1408 if (PL_parser->preambling != NOLINE) {
1409 CopLINE_set(PL_curcop, PL_parser->preambling + 1);
1410 PL_parser->preambling = NOLINE;
1411 }
407f8cf2
KW
1412 if ( got_some_for_debugger
1413 && PERLDB_LINE_OR_SAVESRC
1414 && PL_curstash != PL_debstash)
1415 {
f0e67a1d
Z
1416 /* debugger active and we're not compiling the debugger code,
1417 * so store the line into the debugger's array of lines
1418 */
1419 update_debugger_info(NULL, buf+old_bufend_pos,
1420 new_bufend_pos-old_bufend_pos);
1421 }
1422 return got_some;
1423}
1424
1425/*
1426=for apidoc Amx|I32|lex_peek_unichar|U32 flags
1427
1428Looks ahead one (Unicode) character in the text currently being lexed.
1429Returns the codepoint (unsigned integer value) of the next character,
1430or -1 if lexing has reached the end of the input text. To consume the
1431peeked character, use L</lex_read_unichar>.
1432
1433If the next character is in (or extends into) the next chunk of input
1434text, the next chunk will be read in. Normally the current chunk will be
c5608a1f
KW
1435discarded at the same time, but if C<flags> has the C<LEX_KEEP_PREVIOUS>
1436bit set, then the current chunk will not be discarded.
f0e67a1d
Z
1437
1438If the input is being interpreted as UTF-8 and a UTF-8 encoding error
1439is encountered, an exception is generated.
1440
1441=cut
1442*/
1443
1444I32
1445Perl_lex_peek_unichar(pTHX_ U32 flags)
1446{
749123ff 1447 dVAR;
f0e67a1d
Z
1448 char *s, *bufend;
1449 if (flags & ~(LEX_KEEP_PREVIOUS))
1450 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_peek_unichar");
1451 s = PL_parser->bufptr;
1452 bufend = PL_parser->bufend;
1453 if (UTF) {
1454 U8 head;
1455 I32 unichar;
1456 STRLEN len, retlen;
1457 if (s == bufend) {
1458 if (!lex_next_chunk(flags))
1459 return -1;
1460 s = PL_parser->bufptr;
1461 bufend = PL_parser->bufend;
1462 }
1463 head = (U8)*s;
54d004e8 1464 if (UTF8_IS_INVARIANT(head))
f0e67a1d 1465 return head;
54d004e8
KW
1466 if (UTF8_IS_START(head)) {
1467 len = UTF8SKIP(&head);
f0e67a1d
Z
1468 while ((STRLEN)(bufend-s) < len) {
1469 if (!lex_next_chunk(flags | LEX_KEEP_PREVIOUS))
1470 break;
1471 s = PL_parser->bufptr;
1472 bufend = PL_parser->bufend;
1473 }
1474 }
c80e42f3 1475 unichar = utf8n_to_uvchr((U8*)s, bufend-s, &retlen, UTF8_CHECK_ONLY);
f0e67a1d 1476 if (retlen == (STRLEN)-1) {
75219bac
KW
1477 _force_out_malformed_utf8_message((U8 *) s,
1478 (U8 *) bufend,
1479 0,
1480 1 /* 1 means die */ );
1481 NOT_REACHED; /* NOTREACHED */
f0e67a1d
Z
1482 }
1483 return unichar;
1484 } else {
1485 if (s == bufend) {
1486 if (!lex_next_chunk(flags))
1487 return -1;
1488 s = PL_parser->bufptr;
1489 }
1490 return (U8)*s;
1491 }
1492}
1493
1494/*
1495=for apidoc Amx|I32|lex_read_unichar|U32 flags
1496
1497Reads the next (Unicode) character in the text currently being lexed.
1498Returns the codepoint (unsigned integer value) of the character read,
1499and moves L</PL_parser-E<gt>bufptr> past the character, or returns -1
1500if lexing has reached the end of the input text. To non-destructively
1501examine the next character, use L</lex_peek_unichar> instead.
1502
1503If the next character is in (or extends into) the next chunk of input
1504text, the next chunk will be read in. Normally the current chunk will be
c5608a1f
KW
1505discarded at the same time, but if C<flags> has the C<LEX_KEEP_PREVIOUS>
1506bit set, then the current chunk will not be discarded.
f0e67a1d
Z
1507
1508If the input is being interpreted as UTF-8 and a UTF-8 encoding error
1509is encountered, an exception is generated.
1510
1511=cut
1512*/
1513
1514I32
1515Perl_lex_read_unichar(pTHX_ U32 flags)
1516{
1517 I32 c;
1518 if (flags & ~(LEX_KEEP_PREVIOUS))
1519 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_read_unichar");
1520 c = lex_peek_unichar(flags);
1521 if (c != -1) {
1522 if (c == '\n')
83944c01 1523 COPLINE_INC_WITH_HERELINES;
d9018cbe
EB
1524 if (UTF)
1525 PL_parser->bufptr += UTF8SKIP(PL_parser->bufptr);
1526 else
1527 ++(PL_parser->bufptr);
f0e67a1d
Z
1528 }
1529 return c;
1530}
1531
1532/*
1533=for apidoc Amx|void|lex_read_space|U32 flags
1534
1535Reads optional spaces, in Perl style, in the text currently being
1536lexed. The spaces may include ordinary whitespace characters and
1537Perl-style comments. C<#line> directives are processed if encountered.
1538L</PL_parser-E<gt>bufptr> is moved past the spaces, so that it points
1539at a non-space character (or the end of the input text).
1540
1541If spaces extend into the next chunk of input text, the next chunk will
1542be read in. Normally the current chunk will be discarded at the same
c5608a1f 1543time, but if C<flags> has the C<LEX_KEEP_PREVIOUS> bit set, then the current
f0e67a1d
Z
1544chunk will not be discarded.
1545
1546=cut
1547*/
1548
21791330 1549#define LEX_NO_INCLINE 0x40000000
f0998909
Z
1550#define LEX_NO_NEXT_CHUNK 0x80000000
1551
f0e67a1d
Z
1552void
1553Perl_lex_read_space(pTHX_ U32 flags)
1554{
1555 char *s, *bufend;
21791330 1556 const bool can_incline = !(flags & LEX_NO_INCLINE);
f0e67a1d 1557 bool need_incline = 0;
21791330 1558 if (flags & ~(LEX_KEEP_PREVIOUS|LEX_NO_NEXT_CHUNK|LEX_NO_INCLINE))
f0e67a1d 1559 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_read_space");
f0e67a1d
Z
1560 s = PL_parser->bufptr;
1561 bufend = PL_parser->bufend;
1562 while (1) {
1563 char c = *s;
1564 if (c == '#') {
1565 do {
1566 c = *++s;
1567 } while (!(c == '\n' || (c == 0 && s == bufend)));
1568 } else if (c == '\n') {
1569 s++;
21791330
FC
1570 if (can_incline) {
1571 PL_parser->linestart = s;
1572 if (s == bufend)
1573 need_incline = 1;
1574 else
c6875f94 1575 incline(s, bufend);
21791330 1576 }
f0e67a1d
Z
1577 } else if (isSPACE(c)) {
1578 s++;
1579 } else if (c == 0 && s == bufend) {
1580 bool got_more;
65c68e17 1581 line_t l;
f0998909
Z
1582 if (flags & LEX_NO_NEXT_CHUNK)
1583 break;
f0e67a1d 1584 PL_parser->bufptr = s;
65c68e17 1585 l = CopLINE(PL_curcop);
851b527a 1586 CopLINE(PL_curcop) += PL_parser->herelines + 1;
f0e67a1d 1587 got_more = lex_next_chunk(flags);
65c68e17 1588 CopLINE_set(PL_curcop, l);
f0e67a1d
Z
1589 s = PL_parser->bufptr;
1590 bufend = PL_parser->bufend;
1591 if (!got_more)
1592 break;
21791330 1593 if (can_incline && need_incline && PL_parser->rsfp) {
c6875f94 1594 incline(s, bufend);
f0e67a1d
Z
1595 need_incline = 0;
1596 }
3c47da3c
FC
1597 } else if (!c) {
1598 s++;
f0e67a1d
Z
1599 } else {
1600 break;
1601 }
1602 }
f0e67a1d
Z
1603 PL_parser->bufptr = s;
1604}
1605
1606/*
fe788d6b
PM
1607
1608=for apidoc EXMp|bool|validate_proto|SV *name|SV *proto|bool warn
1609
1610This function performs syntax checking on a prototype, C<proto>.
1611If C<warn> is true, any illegal characters or mismatched brackets
1612will trigger illegalproto warnings, declaring that they were
1613detected in the prototype for C<name>.
1614
1615The return value is C<true> if this is a valid prototype, and
1616C<false> if it is not, regardless of whether C<warn> was C<true> or
1617C<false>.
1618
1619Note that C<NULL> is a valid C<proto> and will always return C<true>.
1620
1621=cut
1622
1623 */
1624
1625bool
5783dc51 1626Perl_validate_proto(pTHX_ SV *name, SV *proto, bool warn, bool curstash)
fe788d6b
PM
1627{
1628 STRLEN len, origlen;
11327fa1 1629 char *p;
fe788d6b
PM
1630 bool bad_proto = FALSE;
1631 bool in_brackets = FALSE;
1632 bool after_slash = FALSE;
1633 char greedy_proto = ' ';
1634 bool proto_after_greedy_proto = FALSE;
1635 bool must_be_last = FALSE;
1636 bool underscore = FALSE;
f791a21a 1637 bool bad_proto_after_underscore = FALSE;
fe788d6b
PM
1638
1639 PERL_ARGS_ASSERT_VALIDATE_PROTO;
1640
1641 if (!proto)
1642 return TRUE;
1643
11327fa1 1644 p = SvPV(proto, len);
fe788d6b
PM
1645 origlen = len;
1646 for (; len--; p++) {
1647 if (!isSPACE(*p)) {
1648 if (must_be_last)
1649 proto_after_greedy_proto = TRUE;
f791a21a
PM
1650 if (underscore) {
1651 if (!strchr(";@%", *p))
1652 bad_proto_after_underscore = TRUE;
1653 underscore = FALSE;
1654 }
fe788d6b
PM
1655 if (!strchr("$@%*;[]&\\_+", *p) || *p == '\0') {
1656 bad_proto = TRUE;
1657 }
1658 else {
fe788d6b
PM
1659 if (*p == '[')
1660 in_brackets = TRUE;
1661 else if (*p == ']')
1662 in_brackets = FALSE;
407f8cf2
KW
1663 else if ((*p == '@' || *p == '%')
1664 && !after_slash
1665 && !in_brackets )
1666 {
fe788d6b
PM
1667 must_be_last = TRUE;
1668 greedy_proto = *p;
1669 }
1670 else if (*p == '_')
f791a21a 1671 underscore = TRUE;
fe788d6b
PM
1672 }
1673 if (*p == '\\')
1674 after_slash = TRUE;
1675 else
1676 after_slash = FALSE;
1677 }
1678 }
1679
1680 if (warn) {
b54d603d 1681 SV *tmpsv = newSVpvs_flags("", SVs_TEMP);
fe788d6b 1682 p -= origlen;
b54d603d
PM
1683 p = SvUTF8(proto)
1684 ? sv_uni_display(tmpsv, newSVpvn_flags(p, origlen, SVs_TEMP | SVf_UTF8),
1685 origlen, UNI_DISPLAY_ISPRINT)
1686 : pv_pretty(tmpsv, p, origlen, 60, NULL, NULL, PERL_PV_ESCAPE_NONASCII);
1687
cbf83791
FC
1688 if (curstash && !memchr(SvPVX(name), ':', SvCUR(name))) {
1689 SV *name2 = sv_2mortal(newSVsv(PL_curstname));
1690 sv_catpvs(name2, "::");
1691 sv_catsv(name2, (SV *)name);
1692 name = name2;
1693 }
1694
fe788d6b
PM
1695 if (proto_after_greedy_proto)
1696 Perl_warner(aTHX_ packWARN(WARN_ILLEGALPROTO),
147e3846 1697 "Prototype after '%c' for %" SVf " : %s",
fe788d6b 1698 greedy_proto, SVfARG(name), p);
50278ed0
PM
1699 if (in_brackets)
1700 Perl_warner(aTHX_ packWARN(WARN_ILLEGALPROTO),
147e3846 1701 "Missing ']' in prototype for %" SVf " : %s",
50278ed0 1702 SVfARG(name), p);
b54d603d 1703 if (bad_proto)
fe788d6b 1704 Perl_warner(aTHX_ packWARN(WARN_ILLEGALPROTO),
147e3846 1705 "Illegal character in prototype for %" SVf " : %s",
f791a21a
PM
1706 SVfARG(name), p);
1707 if (bad_proto_after_underscore)
1708 Perl_warner(aTHX_ packWARN(WARN_ILLEGALPROTO),
147e3846 1709 "Illegal character after '_' in prototype for %" SVf " : %s",
f791a21a 1710 SVfARG(name), p);
fe788d6b
PM
1711 }
1712
1713 return (! (proto_after_greedy_proto || bad_proto) );
1714}
1715
1716/*
ffb4593c
NT
1717 * S_incline
1718 * This subroutine has nothing to do with tilting, whether at windmills
1719 * or pinball tables. Its name is short for "increment line". It
57843af0 1720 * increments the current line number in CopLINE(PL_curcop) and checks
ffb4593c 1721 * to see whether the line starts with a comment of the form
9cbb5ea2
GS
1722 * # line 500 "foo.pm"
1723 * If so, it sets the current line number and file to the values in the comment.
ffb4593c
NT
1724 */
1725
76e3520e 1726STATIC void
c6875f94 1727S_incline(pTHX_ const char *s, const char *end)
463ee0b2 1728{
d9095cec
NC
1729 const char *t;
1730 const char *n;
1731 const char *e;
8818d409 1732 line_t line_num;
22ff3130 1733 UV uv;
463ee0b2 1734
7918f24d
NC
1735 PERL_ARGS_ASSERT_INCLINE;
1736
d77eff5d
KW
1737 assert(end >= s);
1738
83944c01 1739 COPLINE_INC_WITH_HERELINES;
451f421f
FC
1740 if (!PL_rsfp && !PL_parser->filtered && PL_lex_state == LEX_NORMAL
1741 && s+1 == PL_bufend && *s == ';') {
1742 /* fake newline in string eval */
1743 CopLINE_dec(PL_curcop);
1744 return;
1745 }
463ee0b2
LW
1746 if (*s++ != '#')
1747 return;
d4c19fe8
AL
1748 while (SPACE_OR_TAB(*s))
1749 s++;
d77eff5d
KW
1750 if (memBEGINs(s, (STRLEN) (end - s), "line"))
1751 s += sizeof("line") - 1;
73659bf1
GS
1752 else
1753 return;
084592ab 1754 if (SPACE_OR_TAB(*s))
73659bf1 1755 s++;
4e553d73 1756 else
73659bf1 1757 return;
d4c19fe8
AL
1758 while (SPACE_OR_TAB(*s))
1759 s++;
463ee0b2
LW
1760 if (!isDIGIT(*s))
1761 return;
d4c19fe8 1762
463ee0b2
LW
1763 n = s;
1764 while (isDIGIT(*s))
1765 s++;
07714eb4 1766 if (!SPACE_OR_TAB(*s) && *s != '\r' && *s != '\n' && *s != '\0')
26b6dc3f 1767 return;
bf4acbe4 1768 while (SPACE_OR_TAB(*s))
463ee0b2 1769 s++;
85bb8b90 1770 if (*s == '"' && (t = (char *) memchr(s+1, '"', end - s))) {
463ee0b2 1771 s++;
73659bf1
GS
1772 e = t + 1;
1773 }
463ee0b2 1774 else {
c35e046a 1775 t = s;
1bb1a3d6 1776 while (*t && !isSPACE(*t))
c35e046a 1777 t++;
73659bf1 1778 e = t;
463ee0b2 1779 }
bf4acbe4 1780 while (SPACE_OR_TAB(*e) || *e == '\r' || *e == '\f')
73659bf1
GS
1781 e++;
1782 if (*e != '\n' && *e != '\0')
1783 return; /* false alarm */
1784
22ff3130
HS
1785 if (!grok_atoUV(n, &uv, &e))
1786 return;
1787 line_num = ((line_t)uv) - 1;
8818d409 1788
f4dd75d9 1789 if (t - s > 0) {
d9095cec 1790 const STRLEN len = t - s;
3df32bda 1791
d36ee5be 1792 if (!PL_rsfp && !PL_parser->filtered) {
e66cf94c
RGS
1793 /* must copy *{"::_<(eval N)[oldfilename:L]"}
1794 * to *{"::_<newfilename"} */
44867030
NC
1795 /* However, the long form of evals is only turned on by the
1796 debugger - usually they're "(eval %lu)" */
d36ee5be
FC
1797 GV * const cfgv = CopFILEGV(PL_curcop);
1798 if (cfgv) {
38bd7ad8
FC
1799 char smallbuf[128];
1800 STRLEN tmplen2 = len;
44867030 1801 char *tmpbuf2;
449dd039 1802 GV *gv2;
44867030
NC
1803
1804 if (tmplen2 + 2 <= sizeof smallbuf)
1805 tmpbuf2 = smallbuf;
1806 else
1807 Newx(tmpbuf2, tmplen2 + 2, char);
1808
38bd7ad8
FC
1809 tmpbuf2[0] = '_';
1810 tmpbuf2[1] = '<';
44867030
NC
1811
1812 memcpy(tmpbuf2 + 2, s, tmplen2);
1813 tmplen2 += 2;
1814
8a5ee598 1815 gv2 = *(GV**)hv_fetch(PL_defstash, tmpbuf2, tmplen2, TRUE);
e5527e4b 1816 if (!isGV(gv2)) {
8a5ee598 1817 gv_init(gv2, PL_defstash, tmpbuf2, tmplen2, FALSE);
e5527e4b
RGS
1818 /* adjust ${"::_<newfilename"} to store the new file name */
1819 GvSV(gv2) = newSVpvn(tmpbuf2 + 2, tmplen2 - 2);
8818d409
FC
1820 /* The line number may differ. If that is the case,
1821 alias the saved lines that are in the array.
1822 Otherwise alias the whole array. */
1823 if (CopLINE(PL_curcop) == line_num) {
38bd7ad8
FC
1824 GvHV(gv2) = MUTABLE_HV(SvREFCNT_inc(GvHV(cfgv)));
1825 GvAV(gv2) = MUTABLE_AV(SvREFCNT_inc(GvAV(cfgv)));
8818d409 1826 }
38bd7ad8
FC
1827 else if (GvAV(cfgv)) {
1828 AV * const av = GvAV(cfgv);
8818d409
FC
1829 const I32 start = CopLINE(PL_curcop)+1;
1830 I32 items = AvFILLp(av) - start;
1831 if (items > 0) {
1832 AV * const av2 = GvAVn(gv2);
1833 SV **svp = AvARRAY(av) + start;
1834 I32 l = (I32)line_num+1;
1835 while (items--)
1836 av_store(av2, l++, SvREFCNT_inc(*svp++));
1837 }
1838 }
e5527e4b 1839 }
44867030
NC
1840
1841 if (tmpbuf2 != smallbuf) Safefree(tmpbuf2);
d36ee5be 1842 }
e66cf94c 1843 }
05ec9bb3 1844 CopFILE_free(PL_curcop);
449dd039 1845 CopFILE_setn(PL_curcop, s, len);
f4dd75d9 1846 }
8818d409 1847 CopLINE_set(PL_curcop, line_num);
463ee0b2
LW
1848}
1849
80a702cd 1850STATIC void
15f169a1 1851S_update_debugger_info(pTHX_ SV *orig_sv, const char *const buf, STRLEN len)
80a702cd
RGS
1852{
1853 AV *av = CopFILEAVx(PL_curcop);
1854 if (av) {
7f1c3e8c
FC
1855 SV * sv;
1856 if (PL_parser->preambling == NOLINE) sv = newSV_type(SVt_PVMG);
1857 else {
1858 sv = *av_fetch(av, 0, 1);
1859 SvUPGRADE(sv, SVt_PVMG);
1860 }
847cc851 1861 if (!SvPOK(sv)) SvPVCLEAR(sv);
5fa550fb 1862 if (orig_sv)
7f1c3e8c 1863 sv_catsv(sv, orig_sv);
5fa550fb 1864 else
7f1c3e8c
FC
1865 sv_catpvn(sv, buf, len);
1866 if (!SvIOK(sv)) {
1867 (void)SvIOK_on(sv);
1868 SvIV_set(sv, 0);
1869 }
1870 if (PL_parser->preambling == NOLINE)
1871 av_store(av, CopLINE(PL_curcop), sv);
80a702cd
RGS
1872 }
1873}
1874
ffb4593c 1875/*
8c6b0c7d 1876 * skipspace
ffb4593c
NT
1877 * Called to gobble the appropriate amount and type of whitespace.
1878 * Skips comments as well.
71fff7cb 1879 * Returns the next character after the whitespace that is skipped.
8c6b0c7d
FC
1880 *
1881 * peekspace
1882 * Same thing, but look ahead without incrementing line numbers or
1883 * adjusting PL_linestart.
ffb4593c
NT
1884 */
1885
3218e223 1886#define skipspace(s) skipspace_flags(s, 0)
8c6b0c7d 1887#define peekspace(s) skipspace_flags(s, LEX_NO_INCLINE)
3218e223 1888
76e3520e 1889STATIC char *
21791330 1890S_skipspace_flags(pTHX_ char *s, U32 flags)
a687059c 1891{
21791330 1892 PERL_ARGS_ASSERT_SKIPSPACE_FLAGS;
3280af22 1893 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
3c47da3c 1894 while (s < PL_bufend && (SPACE_OR_TAB(*s) || !*s))
463ee0b2 1895 s++;
f0e67a1d
Z
1896 } else {
1897 STRLEN bufptr_pos = PL_bufptr - SvPVX(PL_linestr);
1898 PL_bufptr = s;
21791330 1899 lex_read_space(flags | LEX_KEEP_PREVIOUS |
d27f4b91 1900 (PL_lex_inwhat || PL_lex_state == LEX_FORMLINE ?
f0998909 1901 LEX_NO_NEXT_CHUNK : 0));
3280af22 1902 s = PL_bufptr;
f0e67a1d
Z
1903 PL_bufptr = SvPVX(PL_linestr) + bufptr_pos;
1904 if (PL_linestart > PL_bufptr)
1905 PL_bufptr = PL_linestart;
1906 return s;
463ee0b2 1907 }
5db06880 1908 return s;
a687059c 1909}
378cc40b 1910
ffb4593c
NT
1911/*
1912 * S_check_uni
1913 * Check the unary operators to ensure there's no ambiguity in how they're
1914 * used. An ambiguous piece of code would be:
1915 * rand + 5
1916 * This doesn't mean rand() + 5. Because rand() is a unary operator,
1917 * the +5 is its argument.
1918 */
1919
76e3520e 1920STATIC void
cea2e8a9 1921S_check_uni(pTHX)
ba106d47 1922{
d4c19fe8 1923 const char *s;
2f3197b3 1924
3280af22 1925 if (PL_oldoldbufptr != PL_last_uni)
2f3197b3 1926 return;
3280af22
NIS
1927 while (isSPACE(*PL_last_uni))
1928 PL_last_uni++;
c35e046a 1929 s = PL_last_uni;
fac0f7a3 1930 while (isWORDCHAR_lazy_if_safe(s, PL_bufend, UTF) || *s == '-')
8ce2ba82 1931 s += UTF ? UTF8SKIP(s) : 1;
4efcdc02 1932 if (s < PL_bufptr && memchr(s, '(', PL_bufptr - s))
a0d0e21e 1933 return;
6136c704 1934
9b387841 1935 Perl_ck_warner_d(aTHX_ packWARN(WARN_AMBIGUOUS),
147e3846 1936 "Warning: Use of \"%" UTF8f "\" without parentheses is ambiguous",
b59c097b 1937 UTF8fARG(UTF, (int)(s - PL_last_uni), PL_last_uni));
2f3197b3
LW
1938}
1939
ffb4593c
NT
1940/*
1941 * LOP : macro to build a list operator. Its behaviour has been replaced
1942 * with a subroutine, S_lop() for which LOP is just another name.
1943 */
1944
a0d0e21e
LW
1945#define LOP(f,x) return lop(f,x,s)
1946
ffb4593c
NT
1947/*
1948 * S_lop
1949 * Build a list operator (or something that might be one). The rules:
41e8cbf4
FC
1950 * - if we have a next token, then it's a list operator (no parens) for
1951 * which the next token has already been parsed; e.g.,
1952 * sort foo @args
1953 * sort foo (@args)
ffb4593c
NT
1954 * - if the next thing is an opening paren, then it's a function
1955 * - else it's a list operator
1956 */
1957
76e3520e 1958STATIC I32
11288bb3 1959S_lop(pTHX_ I32 f, U8 x, char *s)
ffed7fef 1960{
7918f24d
NC
1961 PERL_ARGS_ASSERT_LOP;
1962
6154021b 1963 pl_yylval.ival = f;
35c8bce7 1964 CLINE;
3280af22
NIS
1965 PL_bufptr = s;
1966 PL_last_lop = PL_oldbufptr;
eb160463 1967 PL_last_lop_op = (OPCODE)f;
3280af22 1968 if (PL_nexttoke)
78cdf107 1969 goto lstop;
19f1898a 1970 PL_expect = x;
79072805 1971 if (*s == '(')
bbf60fe6 1972 return REPORT(FUNC);
294a536f 1973 s = skipspace(s);
79072805 1974 if (*s == '(')
bbf60fe6 1975 return REPORT(FUNC);
78cdf107
Z
1976 else {
1977 lstop:
1978 if (!PL_lex_allbrackets && PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC)
1979 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC;
bbf60fe6 1980 return REPORT(LSTOP);
78cdf107 1981 }
79072805
LW
1982}
1983
ffb4593c
NT
1984/*
1985 * S_force_next
9cbb5ea2 1986 * When the lexer realizes it knows the next token (for instance,
ffb4593c 1987 * it is reordering tokens for the parser) then it can call S_force_next
9cbb5ea2 1988 * to know what token to return the next time the lexer is called. Caller
b5bbe64a
JH
1989 * will need to set PL_nextval[] and possibly PL_expect to ensure
1990 * the lexer handles the token correctly.
ffb4593c
NT
1991 */
1992
4e553d73 1993STATIC void
cea2e8a9 1994S_force_next(pTHX_ I32 type)
79072805 1995{
704d4215
GG
1996#ifdef DEBUGGING
1997 if (DEBUG_T_TEST) {
1998 PerlIO_printf(Perl_debug_log, "### forced token:\n");
f05d7009 1999 tokereport(type, &NEXTVAL_NEXTTOKE);
704d4215
GG
2000 }
2001#endif
1f7c3e7c 2002 assert(PL_nexttoke < C_ARRAY_LENGTH(PL_nexttype));
3280af22
NIS
2003 PL_nexttype[PL_nexttoke] = type;
2004 PL_nexttoke++;
79072805
LW
2005}
2006
89f35911
FC
2007/*
2008 * S_postderef
2009 *
2010 * This subroutine handles postfix deref syntax after the arrow has already
b3f7b7ad 2011 * been emitted. @* $* etc. are emitted as two separate tokens right here.
89f35911
FC
2012 * @[ @{ %[ %{ *{ are emitted also as two tokens, but this function emits
2013 * only the first, leaving yylex to find the next.
89f35911
FC
2014 */
2015
2016static int
ff25e5db 2017S_postderef(pTHX_ int const funny, char const next)
89f35911 2018{
ff25e5db 2019 assert(funny == DOLSHARP || strchr("$@%&*", funny));
89f35911
FC
2020 if (next == '*') {
2021 PL_expect = XOPERATOR;
cc624add 2022 if (PL_lex_state == LEX_INTERPNORMAL && !PL_lex_brackets) {
ff25e5db 2023 assert('@' == funny || '$' == funny || DOLSHARP == funny);
cc624add 2024 PL_lex_state = LEX_INTERPEND;
c3492809
FC
2025 if ('@' == funny)
2026 force_next(POSTJOIN);
cc624add 2027 }
89f35911
FC
2028 force_next(next);
2029 PL_bufptr+=2;
2030 }
2031 else {
760ca746
FC
2032 if ('@' == funny && PL_lex_state == LEX_INTERPNORMAL
2033 && !PL_lex_brackets)
2034 PL_lex_dojoin = 2;
89f35911
FC
2035 PL_expect = XOPERATOR;
2036 PL_bufptr++;
2037 }
2038 return funny;
2039}
2040
28ac2b49
Z
2041void
2042Perl_yyunlex(pTHX)
2043{
a7aaec61
Z
2044 int yyc = PL_parser->yychar;
2045 if (yyc != YYEMPTY) {
2046 if (yyc) {
a7aaec61
Z
2047 NEXTVAL_NEXTTOKE = PL_parser->yylval;
2048 if (yyc == '{'/*}*/ || yyc == HASHBRACK || yyc == '['/*]*/) {
78cdf107 2049 PL_lex_allbrackets--;
a7aaec61 2050 PL_lex_brackets--;
78cdf107
Z
2051 yyc |= (3<<24) | (PL_lex_brackstack[PL_lex_brackets] << 16);
2052 } else if (yyc == '('/*)*/) {
2053 PL_lex_allbrackets--;
2054 yyc |= (2<<24);
a7aaec61
Z
2055 }
2056 force_next(yyc);
2057 }
28ac2b49
Z
2058 PL_parser->yychar = YYEMPTY;
2059 }
2060}
2061
d0a148a6 2062STATIC SV *
15f169a1 2063S_newSV_maybe_utf8(pTHX_ const char *const start, STRLEN len)
d0a148a6 2064{
740cce10 2065 SV * const sv = newSVpvn_utf8(start, len,
6d492fac
KW
2066 ! IN_BYTES
2067 && UTF
2068 && is_utf8_non_invariant_string((const U8*)start, len));
d0a148a6
NC
2069 return sv;
2070}
2071
ffb4593c
NT
2072/*
2073 * S_force_word
2074 * When the lexer knows the next thing is a word (for instance, it has
2075 * just seen -> and it knows that the next char is a word char, then
02b34bbe
DM
2076 * it calls S_force_word to stick the next word into the PL_nexttoke/val
2077 * lookahead.
ffb4593c
NT
2078 *
2079 * Arguments:
b1b65b59 2080 * char *start : buffer position (must be within PL_linestr)
185c2e96
DM
2081 * int token : PL_next* will be this type of bare word
2082 * (e.g., METHOD,BAREWORD)
ffb4593c
NT
2083 * int check_keyword : if true, Perl checks to make sure the word isn't
2084 * a keyword (do this if the word is a label, e.g. goto FOO)
2085 * int allow_pack : if true, : characters will also be allowed (require,
2086 * use, etc. do this)
ffb4593c
NT
2087 */
2088
76e3520e 2089STATIC char *
345b3785 2090S_force_word(pTHX_ char *start, int token, int check_keyword, int allow_pack)
79072805 2091{
eb578fdb 2092 char *s;
463ee0b2 2093 STRLEN len;
4e553d73 2094
7918f24d
NC
2095 PERL_ARGS_ASSERT_FORCE_WORD;
2096
294a536f 2097 start = skipspace(start);
463ee0b2 2098 s = start;
fac0f7a3 2099 if ( isIDFIRST_lazy_if_safe(s, PL_bufend, UTF)
e7127e21 2100 || (allow_pack && *s == ':' && s[1] == ':') )
a0d0e21e 2101 {
3280af22 2102 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, allow_pack, &len);
01b5ef50
FC
2103 if (check_keyword) {
2104 char *s2 = PL_tokenbuf;
487e470d 2105 STRLEN len2 = len;
de627158
KW
2106 if (allow_pack && memBEGINPs(s2, len, "CORE::")) {
2107 s2 += sizeof("CORE::") - 1;
2108 len2 -= sizeof("CORE::") - 1;
2109 }
487e470d 2110 if (keyword(s2, len2, 0))
463ee0b2 2111 return start;
01b5ef50 2112 }
463ee0b2 2113 if (token == METHOD) {
294a536f 2114 s = skipspace(s);
463ee0b2 2115 if (*s == '(')
3280af22 2116 PL_expect = XTERM;
463ee0b2 2117 else {
3280af22 2118 PL_expect = XOPERATOR;
463ee0b2 2119 }
79072805 2120 }
9ded7720 2121 NEXTVAL_NEXTTOKE.opval
275103cd 2122 = newSVOP(OP_CONST,0,
d0a148a6 2123 S_newSV_maybe_utf8(aTHX_ PL_tokenbuf, len));
9ded7720 2124 NEXTVAL_NEXTTOKE.opval->op_private |= OPpCONST_BARE;
79072805
LW
2125 force_next(token);
2126 }
2127 return s;
2128}
2129
ffb4593c
NT
2130/*
2131 * S_force_ident
9cbb5ea2 2132 * Called when the lexer wants $foo *foo &foo etc, but the program
ffb4593c
NT
2133 * text only contains the "foo" portion. The first argument is a pointer
2134 * to the "foo", and the second argument is the type symbol to prefix.
185c2e96 2135 * Forces the next token to be a "BAREWORD".
9cbb5ea2 2136 * Creates the symbol if it didn't already exist (via gv_fetchpv()).
ffb4593c
NT
2137 */
2138
76e3520e 2139STATIC void
5aaab254 2140S_force_ident(pTHX_ const char *s, int kind)
79072805 2141{
7918f24d
NC
2142 PERL_ARGS_ASSERT_FORCE_IDENT;
2143
c9b48522
DD
2144 if (s[0]) {
2145 const STRLEN len = s[1] ? strlen(s) : 1; /* s = "\"" see yylex */
275103cd 2146 OP* const o = newSVOP(OP_CONST, 0, newSVpvn_flags(s, len,
728847b1 2147 UTF ? SVf_UTF8 : 0));
9ded7720 2148 NEXTVAL_NEXTTOKE.opval = o;
185c2e96 2149 force_next(BAREWORD);
748a9306 2150 if (kind) {
11343788 2151 o->op_private = OPpCONST_ENTERED;
55497cff
PP
2152 /* XXX see note in pp_entereval() for why we forgo typo
2153 warnings if the symbol must be introduced in an eval.
2154 GSAR 96-10-12 */
90e5519e 2155 gv_fetchpvn_flags(s, len,
4bff32c5 2156 (PL_in_eval ? GV_ADDMULTI
728847b1 2157 : GV_ADD) | ( UTF ? SVf_UTF8 : 0 ),
90e5519e
NC
2158 kind == '$' ? SVt_PV :
2159 kind == '@' ? SVt_PVAV :
2160 kind == '%' ? SVt_PVHV :
a0d0e21e 2161 SVt_PVGV
90e5519e 2162 );
748a9306 2163 }
79072805
LW
2164 }
2165}
2166
3f33d153
FC
2167static void
2168S_force_ident_maybe_lex(pTHX_ char pit)
2169{
3f33d153
FC
2170 NEXTVAL_NEXTTOKE.ival = pit;
2171 force_next('p');
2172}
2173
1571675a
GS
2174NV
2175Perl_str_to_version(pTHX_ SV *sv)
2176{
2177 NV retval = 0.0;
2178 NV nshift = 1.0;
2179 STRLEN len;
cfd0369c 2180 const char *start = SvPV_const(sv,len);
9d4ba2ae 2181 const char * const end = start + len;
8298454c 2182 const bool utf = cBOOL(SvUTF8(sv));
7918f24d
NC
2183
2184 PERL_ARGS_ASSERT_STR_TO_VERSION;
2185
1571675a 2186 while (start < end) {
ba210ebe 2187 STRLEN skip;
1571675a
GS
2188 UV n;
2189 if (utf)
9041c2e3 2190 n = utf8n_to_uvchr((U8*)start, len, &skip, 0);
1571675a
GS
2191 else {
2192 n = *(U8*)start;
2193 skip = 1;
2194 }
2195 retval += ((NV)n)/nshift;
2196 start += skip;
2197 nshift *= 1000;
2198 }
2199 return retval;
2200}
2201
4e553d73 2202/*
ffb4593c
NT
2203 * S_force_version
2204 * Forces the next token to be a version number.
e759cc13
RGS
2205 * If the next token appears to be an invalid version number, (e.g. "v2b"),
2206 * and if "guessing" is TRUE, then no new token is created (and the caller
2207 * must use an alternative parsing method).
ffb4593c
NT
2208 */
2209
76e3520e 2210STATIC char *
e759cc13 2211S_force_version(pTHX_ char *s, int guessing)
89bfa8cd 2212{
5f66b61c 2213 OP *version = NULL;
44dcb63b 2214 char *d;
89bfa8cd 2215
7918f24d
NC
2216 PERL_ARGS_ASSERT_FORCE_VERSION;
2217
294a536f 2218 s = skipspace(s);
89bfa8cd 2219
44dcb63b 2220 d = s;
dd629d5b 2221 if (*d == 'v')
44dcb63b 2222 d++;
44dcb63b 2223 if (isDIGIT(*d)) {
e759cc13
RGS
2224 while (isDIGIT(*d) || *d == '_' || *d == '.')
2225 d++;
4e4da3ac 2226 if (*d == ';' || isSPACE(*d) || *d == '{' || *d == '}' || !*d) {
dd629d5b 2227 SV *ver;
6154021b
RGS
2228 s = scan_num(s, &pl_yylval);
2229 version = pl_yylval.opval;
dd629d5b
GS
2230 ver = cSVOPx(version)->op_sv;
2231 if (SvPOK(ver) && !SvNIOK(ver)) {
862a34c6 2232 SvUPGRADE(ver, SVt_PVNV);
9d6ce603 2233 SvNV_set(ver, str_to_version(ver));
1571675a 2234 SvNOK_on(ver); /* hint that it is a version */
44dcb63b 2235 }
89bfa8cd 2236 }
5db06880 2237 else if (guessing) {
e759cc13 2238 return s;
5db06880 2239 }
89bfa8cd
PP
2240 }
2241
2242 /* NOTE: The parser sees the package name and the VERSION swapped */
9ded7720 2243 NEXTVAL_NEXTTOKE.opval = version;
185c2e96 2244 force_next(BAREWORD);
89bfa8cd 2245
e759cc13 2246 return s;
89bfa8cd
PP
2247}
2248
ffb4593c 2249/*
91152fc1
DG
2250 * S_force_strict_version
2251 * Forces the next token to be a version number using strict syntax rules.
2252 */
2253
2254STATIC char *
2255S_force_strict_version(pTHX_ char *s)
2256{
91152fc1 2257 OP *version = NULL;
91152fc1
DG
2258 const char *errstr = NULL;
2259
2260 PERL_ARGS_ASSERT_FORCE_STRICT_VERSION;
2261
2262 while (isSPACE(*s)) /* leading whitespace */
2263 s++;
2264
2265 if (is_STRICT_VERSION(s,&errstr)) {
2266 SV *ver = newSV(0);
2267 s = (char *)scan_version(s, ver, 0);
2268 version = newSVOP(OP_CONST, 0, ver);
2269 }
407f8cf2
KW
2270 else if ((*s != ';' && *s != '{' && *s != '}' )
2271 && (s = skipspace(s), (*s != ';' && *s != '{' && *s != '}' )))
4e4da3ac 2272 {
91152fc1
DG
2273 PL_bufptr = s;
2274 if (errstr)
2275 yyerror(errstr); /* version required */
2276 return s;
2277 }
2278
91152fc1 2279 /* NOTE: The parser sees the package name and the VERSION swapped */
91152fc1 2280 NEXTVAL_NEXTTOKE.opval = version;
185c2e96 2281 force_next(BAREWORD);
91152fc1
DG
2282
2283 return s;
2284}
2285
2286/*
ffb4593c 2287 * S_tokeq
ef3ff34d
KW
2288 * Turns any \\ into \ in a quoted string passed in in 'sv', returning 'sv',
2289 * modified as necessary. However, if HINT_NEW_STRING is on, 'sv' is
2290 * unchanged, and a new SV containing the modified input is returned.
ffb4593c
NT
2291 */
2292
76e3520e 2293STATIC SV *
cea2e8a9 2294S_tokeq(pTHX_ SV *sv)
79072805 2295{
eb578fdb
KW
2296 char *s;
2297 char *send;
2298 char *d;
b3ac6de7 2299 SV *pv = sv;
79072805 2300
7918f24d
NC
2301 PERL_ARGS_ASSERT_TOKEQ;
2302
279b35ad
FC
2303 assert (SvPOK(sv));
2304 assert (SvLEN(sv));
2305 assert (!SvIsCOW(sv));
307ed071 2306 if (SvTYPE(sv) >= SVt_PVIV && SvIVX(sv) == -1) /* <<'heredoc' */
b3ac6de7 2307 goto finish;
279b35ad
FC
2308 s = SvPVX(sv);
2309 send = SvEND(sv);
dcb21ed6
NC
2310 /* This is relying on the SV being "well formed" with a trailing '\0' */
2311 while (s < send && !(*s == '\\' && s[1] == '\\'))
79072805
LW
2312 s++;
2313 if (s == send)
b3ac6de7 2314 goto finish;
79072805 2315 d = s;
be4731d2 2316 if ( PL_hints & HINT_NEW_STRING ) {
279b35ad
FC
2317 pv = newSVpvn_flags(SvPVX_const(pv), SvCUR(sv),
2318 SVs_TEMP | SvUTF8(sv));
be4731d2 2319 }
79072805
LW
2320 while (s < send) {
2321 if (*s == '\\') {
a0d0e21e 2322 if (s + 1 < send && (s[1] == '\\'))
79072805
LW
2323 s++; /* all that, just for this */
2324 }
2325 *d++ = *s++;
2326 }
2327 *d = '\0';
95a20fc0 2328 SvCUR_set(sv, d - SvPVX_const(sv));
b3ac6de7 2329 finish:
3280af22 2330 if ( PL_hints & HINT_NEW_STRING )
eb0d8d16 2331 return new_constant(NULL, 0, "q", sv, pv, "q", 1);
79072805
LW
2332 return sv;
2333}
2334
ffb4593c
NT
2335/*
2336 * Now come three functions related to double-quote context,
2337 * S_sublex_start, S_sublex_push, and S_sublex_done. They're used when
2338 * converting things like "\u\Lgnat" into ucfirst(lc("gnat")). They
2339 * interact with PL_lex_state, and create fake ( ... ) argument lists
2340 * to handle functions and concatenation.
ecd24171
DM
2341 * For example,
2342 * "foo\lbar"
2343 * is tokenised as
2344 * stringify ( const[foo] concat lcfirst ( const[bar] ) )
ffb4593c
NT
2345 */
2346
2347/*
2348 * S_sublex_start
6154021b 2349 * Assumes that pl_yylval.ival is the op we're creating (e.g. OP_LCFIRST).
ffb4593c
NT
2350 *
2351 * Pattern matching will set PL_lex_op to the pattern-matching op to
6154021b 2352 * make (we return THING if pl_yylval.ival is OP_NULL, PMFUNC otherwise).
ffb4593c 2353 *
16d1d8bd 2354 * OP_CONST is easy--just make the new op and return.
ffb4593c
NT
2355 *
2356 * Everything else becomes a FUNC.
2357 *
16d1d8bd
LM
2358 * Sets PL_lex_state to LEX_INTERPPUSH unless ival was OP_NULL or we
2359 * had an OP_CONST. This just sets us up for a
ffb4593c
NT
2360 * call to S_sublex_push().
2361 */
2362
76e3520e 2363STATIC I32
cea2e8a9 2364S_sublex_start(pTHX)
79072805 2365{
eb578fdb 2366 const I32 op_type = pl_yylval.ival;
79072805
LW
2367
2368 if (op_type == OP_NULL) {
6154021b 2369 pl_yylval.opval = PL_lex_op;
5f66b61c 2370 PL_lex_op = NULL;
79072805
LW
2371 return THING;
2372 }
466112bb 2373 if (op_type == OP_CONST) {
67c71cbb
FC
2374 SV *sv = PL_lex_stuff;
2375 PL_lex_stuff = NULL;
2376 sv = tokeq(sv);
b3ac6de7
IZ
2377
2378 if (SvTYPE(sv) == SVt_PVIV) {
2379 /* Overloaded constants, nothing fancy: Convert to SVt_PV: */
2380 STRLEN len;
96a5add6 2381 const char * const p = SvPV_const(sv, len);
740cce10 2382 SV * const nsv = newSVpvn_flags(p, len, SvUTF8(sv));
b3ac6de7
IZ
2383 SvREFCNT_dec(sv);
2384 sv = nsv;
4e553d73 2385 }
275103cd 2386 pl_yylval.opval = newSVOP(op_type, 0, sv);
79072805
LW
2387 return THING;
2388 }
2389
7ef70b3d
FC
2390 PL_parser->lex_super_state = PL_lex_state;
2391 PL_parser->lex_sub_inwhat = (U16)op_type;
2392 PL_parser->lex_sub_op = PL_lex_op;
3280af22 2393 PL_lex_state = LEX_INTERPPUSH;
55497cff 2394
3280af22
NIS
2395 PL_expect = XTERM;
2396 if (PL_lex_op) {
6154021b 2397 pl_yylval.opval = PL_lex_op;
5f66b61c 2398 PL_lex_op = NULL;
55497cff
PP
2399 return PMFUNC;
2400 }
2401 else
2402 return FUNC;
2403}
2404
ffb4593c
NT
2405/*
2406 * S_sublex_push
2407 * Create a new scope to save the lexing state. The scope will be
2408 * ended in S_sublex_done. Returns a '(', starting the function arguments
2409 * to the uc, lc, etc. found before.
2410 * Sets PL_lex_state to LEX_INTERPCONCAT.
2411 */
2412
76e3520e 2413STATIC I32
cea2e8a9 2414S_sublex_push(pTHX)
55497cff 2415{
78a635de 2416 LEXSHARED *shared;
801d32ac 2417 const bool is_heredoc = PL_multi_close == '<';
f46d017c 2418 ENTER;
55497cff 2419
7ef70b3d 2420 PL_lex_state = PL_parser->lex_super_state;
cc624add 2421 SAVEI8(PL_lex_dojoin);
3280af22 2422 SAVEI32(PL_lex_brackets);
78cdf107 2423 SAVEI32(PL_lex_allbrackets);
b27dce25 2424 SAVEI32(PL_lex_formbrack);
78cdf107 2425 SAVEI8(PL_lex_fakeeof);
3280af22
NIS
2426 SAVEI32(PL_lex_casemods);
2427 SAVEI32(PL_lex_starts);
651b5b28 2428 SAVEI8(PL_lex_state);
7cc34111 2429 SAVESPTR(PL_lex_repl);
7766f137 2430 SAVEVPTR(PL_lex_inpat);
98246f1e 2431 SAVEI16(PL_lex_inwhat);
ffdb8b16 2432 if (is_heredoc)
b42366d4 2433 {
ffdb8b16 2434 SAVECOPLINE(PL_curcop);
b42366d4 2435 SAVEI32(PL_multi_end);
851b527a
FC
2436 SAVEI32(PL_parser->herelines);
2437 PL_parser->herelines = 0;
b42366d4 2438 }
2ca4363d 2439 SAVEIV(PL_multi_close);
3280af22 2440 SAVEPPTR(PL_bufptr);
8452ff4b 2441 SAVEPPTR(PL_bufend);
3280af22
NIS
2442 SAVEPPTR(PL_oldbufptr);
2443 SAVEPPTR(PL_oldoldbufptr);
207e3d1a
JH
2444 SAVEPPTR(PL_last_lop);
2445 SAVEPPTR(PL_last_uni);
3280af22
NIS
2446 SAVEPPTR(PL_linestart);
2447 SAVESPTR(PL_linestr);
8edd5f42
RGS
2448 SAVEGENERICPV(PL_lex_brackstack);
2449 SAVEGENERICPV(PL_lex_casestack);
78a635de 2450 SAVEGENERICPV(PL_parser->lex_shared);
3a54fd60 2451 SAVEBOOL(PL_parser->lex_re_reparsing);
ffdb8b16 2452 SAVEI32(PL_copline);
3280af22 2453
99bd9d90 2454 /* The here-doc parser needs to be able to peek into outer lexing
60f40a38
FC
2455 scopes to find the body of the here-doc. So we put PL_linestr and
2456 PL_bufptr into lex_shared, to ‘share’ those values.
99bd9d90 2457 */
60f40a38
FC
2458 PL_parser->lex_shared->ls_linestr = PL_linestr;
2459 PL_parser->lex_shared->ls_bufptr = PL_bufptr;
99bd9d90 2460
3280af22 2461 PL_linestr = PL_lex_stuff;
7ef70b3d 2462 PL_lex_repl = PL_parser->lex_sub_repl;
a0714e2c 2463 PL_lex_stuff = NULL;
7ef70b3d 2464 PL_parser->lex_sub_repl = NULL;
3280af22 2465
eabab8bc
FC
2466 /* Arrange for PL_lex_stuff to be freed on scope exit, in case it gets
2467 set for an inner quote-like operator and then an error causes scope-
2468 popping. We must not have a PL_lex_stuff value left dangling, as
2469 that breaks assumptions elsewhere. See bug #123617. */
2470 SAVEGENERICSV(PL_lex_stuff);
7ef70b3d 2471 SAVEGENERICSV(PL_parser->lex_sub_repl);
eabab8bc 2472
9cbb5ea2
GS
2473 PL_bufend = PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart
2474 = SvPVX(PL_linestr);
3280af22 2475 PL_bufend += SvCUR(PL_linestr);
bd61b366 2476 PL_last_lop = PL_last_uni = NULL;
3280af22 2477 SAVEFREESV(PL_linestr);
4dc843bc 2478 if (PL_lex_repl) SAVEFREESV(PL_lex_repl);
3280af22
NIS
2479
2480 PL_lex_dojoin = FALSE;
b27dce25 2481 PL_lex_brackets = PL_lex_formbrack = 0;
78cdf107
Z
2482 PL_lex_allbrackets = 0;
2483 PL_lex_fakeeof = LEX_FAKEEOF_NEVER;
a02a5408
JC
2484 Newx(PL_lex_brackstack, 120, char);
2485 Newx(PL_lex_casestack, 12, char);
3280af22
NIS
2486 PL_lex_casemods = 0;
2487 *PL_lex_casestack = '\0';
2488 PL_lex_starts = 0;
2489 PL_lex_state = LEX_INTERPCONCAT;
ffdb8b16 2490 if (is_heredoc)
6ddcf93b 2491 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
ffdb8b16 2492 PL_copline = NOLINE;
caae0700 2493
78a635de
FC
2494 Newxz(shared, 1, LEXSHARED);
2495 shared->ls_prev = PL_parser->lex_shared;
2496 PL_parser->lex_shared = shared;
3280af22 2497
7ef70b3d 2498 PL_lex_inwhat = PL_parser->lex_sub_inwhat;
bb16bae8 2499 if (PL_lex_inwhat == OP_TRANSR) PL_lex_inwhat = OP_TRANS;
3280af22 2500 if (PL_lex_inwhat == OP_MATCH || PL_lex_inwhat == OP_QR || PL_lex_inwhat == OP_SUBST)
7ef70b3d 2501 PL_lex_inpat = PL_parser->lex_sub_op;
79072805 2502 else
5f66b61c 2503 PL_lex_inpat = NULL;
79072805 2504
3a54fd60
DM
2505 PL_parser->lex_re_reparsing = cBOOL(PL_in_eval & EVAL_RE_REPARSING);
2506 PL_in_eval &= ~EVAL_RE_REPARSING;
2507
55497cff 2508 return '(';
79072805
LW
2509}
2510
ffb4593c
NT
2511/*
2512 * S_sublex_done
2513 * Restores lexer state after a S_sublex_push.
2514 */
2515
76e3520e 2516STATIC I32
cea2e8a9 2517S_sublex_done(pTHX)
79072805 2518{
3280af22 2519 if (!PL_lex_starts++) {
396482e1 2520 SV * const sv = newSVpvs("");
9aa983d2
JH
2521 if (SvUTF8(PL_linestr))
2522 SvUTF8_on(sv);
3280af22 2523 PL_expect = XOPERATOR;
275103cd 2524 pl_yylval.opval = newSVOP(OP_CONST, 0, sv);
79072805
LW
2525 return THING;
2526 }
2527
3280af22
NIS
2528 if (PL_lex_casemods) { /* oops, we've got some unbalanced parens */
2529 PL_lex_state = LEX_INTERPCASEMOD;
cea2e8a9 2530 return yylex();
79072805
LW
2531 }
2532
ffb4593c 2533 /* Is there a right-hand side to take care of? (s//RHS/ or tr//RHS/) */
bb16bae8 2534 assert(PL_lex_inwhat != OP_TRANSR);
5aa91856
FC
2535 if (PL_lex_repl) {
2536 assert (PL_lex_inwhat == OP_SUBST || PL_lex_inwhat == OP_TRANS);
3280af22
NIS
2537 PL_linestr = PL_lex_repl;
2538 PL_lex_inpat = 0;
2539 PL_bufend = PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart = SvPVX(PL_linestr);
2540 PL_bufend += SvCUR(PL_linestr);
bd61b366 2541 PL_last_lop = PL_last_uni = NULL;
3280af22
NIS
2542 PL_lex_dojoin = FALSE;
2543 PL_lex_brackets = 0;
78cdf107
Z
2544 PL_lex_allbrackets = 0;
2545 PL_lex_fakeeof = LEX_FAKEEOF_NEVER;
3280af22
NIS
2546 PL_lex_casemods = 0;
2547 *PL_lex_casestack = '\0';
2548 PL_lex_starts = 0;
25da4f38 2549 if (SvEVALED(PL_lex_repl)) {
3280af22
NIS
2550 PL_lex_state = LEX_INTERPNORMAL;
2551 PL_lex_starts++;
e9fa98b2
HS
2552 /* we don't clear PL_lex_repl here, so that we can check later
2553 whether this is an evalled subst; that means we rely on the
2554 logic to ensure sublex_done() is called again only via the
2555 branch (in yylex()) that clears PL_lex_repl, else we'll loop */
79072805 2556 }
e9fa98b2 2557 else {
3280af22 2558 PL_lex_state = LEX_INTERPCONCAT;
a0714e2c 2559 PL_lex_repl = NULL;
e9fa98b2 2560 }
ffdb8b16
FC
2561 if (SvTYPE(PL_linestr) >= SVt_PVNV) {
2562 CopLINE(PL_curcop) +=
9420b268 2563 ((XPVNV*)SvANY(PL_linestr))->xnv_u.xnv_lines
851b527a
FC
2564 + PL_parser->herelines;
2565 PL_parser->herelines = 0;
ffdb8b16 2566 }
9b6b7be8 2567 return '/';
ffed7fef
LW
2568 }
2569 else {
b42366d4 2570 const line_t l = CopLINE(PL_curcop);
f46d017c 2571 LEAVE;
b42366d4 2572 if (PL_multi_close == '<')
851b527a 2573 PL_parser->herelines += l - PL_multi_end;
3280af22
NIS
2574 PL_bufend = SvPVX(PL_linestr);
2575 PL_bufend += SvCUR(PL_linestr);
2576 PL_expect = XOPERATOR;
79072805 2577 return ')';
ffed7fef
LW
2578 }
2579}
2580
7a7d14f3 2581STATIC SV*
6f613c73
KW
2582S_get_and_check_backslash_N_name(pTHX_ const char* s, const char* const e)
2583{
140b12ad
KW
2584 /* <s> points to first character of interior of \N{}, <e> to one beyond the
2585 * interior, hence to the "}". Finds what the name resolves to, returning
2586 * an SV* containing it; NULL if no valid one found */
2587
dd2b1b72 2588 SV* res = newSVpvn_flags(s, e - s, UTF ? SVf_UTF8 : 0);
6f613c73 2589
0c415a79
KW
2590 HV * table;
2591 SV **cvp;
2592 SV *cv;
2593 SV *rv;
2594 HV *stash;
0c415a79
KW
2595 const char* backslash_ptr = s - 3; /* Points to the <\> of \N{... */
2596
6f613c73
KW
2597 PERL_ARGS_ASSERT_GET_AND_CHECK_BACKSLASH_N_NAME;
2598
d8d26cac 2599 if (!SvCUR(res)) {
ebcc725e 2600 SvREFCNT_dec_NN(res);
be332ba0
KW
2601 /* diag_listed_as: Unknown charname '%s' */
2602 yyerror("Unknown charname ''");
2603 return NULL;
d8d26cac 2604 }
b7e6151c 2605
107160e2
KW
2606 res = new_constant( NULL, 0, "charnames", res, NULL, backslash_ptr,
2607 /* include the <}> */
2608 e - backslash_ptr + 1);
6f613c73 2609 if (! SvPOK(res)) {
b6407c49 2610 SvREFCNT_dec_NN(res);
6f613c73
KW
2611 return NULL;
2612 }
2613
0c415a79
KW
2614 /* See if the charnames handler is the Perl core's, and if so, we can skip
2615 * the validation needed for a user-supplied one, as Perl's does its own
2616 * validation. */
2617 table = GvHV(PL_hintgv); /* ^H */
2618 cvp = hv_fetchs(table, "charnames", FALSE);
5882ddb3
FC
2619 if (cvp && (cv = *cvp) && SvROK(cv) && (rv = SvRV(cv),
2620 SvTYPE(rv) == SVt_PVCV) && ((stash = CvSTASH(rv)) != NULL))
0c415a79
KW
2621 {
2622 const char * const name = HvNAME(stash);
b59bf0b2 2623 if (memEQs(name, HvNAMELEN(stash), "_charnames")) {
0c415a79
KW
2624 return res;
2625 }
2626 }
2627
bde9e88d
KW
2628 /* Here, it isn't Perl's charname handler. We can't rely on a
2629 * user-supplied handler to validate the input name. For non-ut8 input,
2630 * look to see that the first character is legal. Then loop through the
2631 * rest checking that each is a continuation */
6f613c73 2632
36897d64
KW
2633 /* This code makes the reasonable assumption that the only Latin1-range
2634 * characters that begin a character name alias are alphabetic, otherwise
2635 * would have to create a isCHARNAME_BEGIN macro */
b6ba1137 2636
b6ba1137 2637 if (! UTF) {
bde9e88d 2638 if (! isALPHAU(*s)) {
b6ba1137
KW
2639 goto bad_charname;
2640 }
bde9e88d
KW
2641 s++;
2642 while (s < e) {
2643 if (! isCHARNAME_CONT(*s)) {
b6ba1137
KW
2644 goto bad_charname;
2645 }
2d8eb851
KW
2646 if (*s == ' ' && *(s-1) == ' ') {
2647 goto multi_spaces;
bd299e29 2648 }
bde9e88d 2649 s++;
b6ba1137
KW
2650 }
2651 }
2652 else {
bde9e88d
KW
2653 /* Similarly for utf8. For invariants can check directly; for other
2654 * Latin1, can calculate their code point and check; otherwise use a
2655 * swash */
2656 if (UTF8_IS_INVARIANT(*s)) {
2657 if (! isALPHAU(*s)) {
140b12ad
KW
2658 goto bad_charname;
2659 }
bde9e88d
KW
2660 s++;
2661 } else if (UTF8_IS_DOWNGRADEABLE_START(*s)) {
a62b247b 2662 if (! isALPHAU(EIGHT_BIT_UTF8_TO_NATIVE(*s, *(s+1)))) {
b6ba1137 2663 goto bad_charname;
6f613c73 2664 }
bde9e88d 2665 s += 2;
6f613c73 2666 }
bde9e88d
KW
2667 else {
2668 if (! PL_utf8_charname_begin) {
2669 U8 flags = _CORE_SWASH_INIT_ACCEPT_INVLIST;
2670 PL_utf8_charname_begin = _core_swash_init("utf8",
2671 "_Perl_Charname_Begin",
2672 &PL_sv_undef,
2673 1, 0, NULL, &flags);
2674 }
2675 if (! swash_fetch(PL_utf8_charname_begin, (U8 *) s, TRUE)) {
2676 goto bad_charname;
2677 }
2678 s += UTF8SKIP(s);
2679 }
2680
2681 while (s < e) {
2682 if (UTF8_IS_INVARIANT(*s)) {
2683 if (! isCHARNAME_CONT(*s)) {
2684 goto bad_charname;
2685 }
2d8eb851
KW
2686 if (*s == ' ' && *(s-1) == ' ') {
2687 goto multi_spaces;
bd299e29 2688 }
bde9e88d
KW
2689 s++;
2690 }
2691 else if (UTF8_IS_DOWNGRADEABLE_START(*s)) {
a62b247b 2692 if (! isCHARNAME_CONT(EIGHT_BIT_UTF8_TO_NATIVE(*s, *(s+1))))
bde9e88d
KW
2693 {
2694 goto bad_charname;
2695 }
2696 s += 2;
2697 }
2698 else {
2699 if (! PL_utf8_charname_continue) {
2700 U8 flags = _CORE_SWASH_INIT_ACCEPT_INVLIST;
2701 PL_utf8_charname_continue = _core_swash_init("utf8",
2702 "_Perl_Charname_Continue",
2703 &PL_sv_undef,
2704 1, 0, NULL, &flags);
2705 }
2706 if (! swash_fetch(PL_utf8_charname_continue, (U8 *) s, TRUE)) {
2707 goto bad_charname;
2708 }
2709 s += UTF8SKIP(s);
6f613c73
KW
2710 }
2711 }
2d8eb851
KW
2712 }
2713 if (*(s-1) == ' ') {
8d9d0498
FC
2714 /* diag_listed_as: charnames alias definitions may not contain
2715 trailing white-space; marked by <-- HERE in %s
2716 */
2d8eb851
KW
2717 yyerror_pv(
2718 Perl_form(aTHX_
2719 "charnames alias definitions may not contain trailing "
2720 "white-space; marked by <-- HERE in %.*s<-- HERE %.*s",
2721 (int)(s - backslash_ptr + 1), backslash_ptr,
2722 (int)(e - s + 1), s + 1
2723 ),
2724 UTF ? SVf_UTF8 : 0);
2725 return NULL;
6f613c73
KW
2726 }
2727
94ca1619 2728 if (SvUTF8(res)) { /* Don't accept malformed input */
bde9e88d
KW
2729 const U8* first_bad_char_loc;
2730 STRLEN len;
2731 const char* const str = SvPV_const(res, len);
0aab20f2
KW
2732 if (UNLIKELY(! is_utf8_string_loc((U8 *) str, len,
2733 &first_bad_char_loc)))
2734 {
75219bac
KW
2735 _force_out_malformed_utf8_message(first_bad_char_loc,
2736 (U8 *) PL_parser->bufend,
2737 0,
2738 0 /* 0 means don't die */ );
8d9d0498
FC
2739 /* diag_listed_as: Malformed UTF-8 returned by \N{%s}
2740 immediately after '%s' */
bde9e88d
KW
2741 yyerror_pv(
2742 Perl_form(aTHX_
2743 "Malformed UTF-8 returned by %.*s immediately after '%.*s'",
2744 (int) (e - backslash_ptr + 1), backslash_ptr,
2745 (int) ((char *) first_bad_char_loc - str), str
2746 ),
2747 SVf_UTF8);
2748 return NULL;
2749 }
2750 }
140b12ad 2751
bde9e88d 2752 return res;
140b12ad 2753
bde9e88d 2754 bad_charname: {
bde9e88d
KW
2755
2756 /* The final %.*s makes sure that should the trailing NUL be missing
2757 * that this print won't run off the end of the string */
8d9d0498
FC
2758 /* diag_listed_as: Invalid character in \N{...}; marked by <-- HERE
2759 in \N{%s} */
bde9e88d
KW
2760 yyerror_pv(
2761 Perl_form(aTHX_
2762 "Invalid character in \\N{...}; marked by <-- HERE in %.*s<-- HERE %.*s",
2d8eb851
KW
2763 (int)(s - backslash_ptr + 1), backslash_ptr,
2764 (int)(e - s + 1), s + 1
bde9e88d
KW
2765 ),
2766 UTF ? SVf_UTF8 : 0);
2767 return NULL;
2768 }
2d8eb851
KW
2769
2770 multi_spaces:
8d9d0498
FC
2771 /* diag_listed_as: charnames alias definitions may not contain a
2772 sequence of multiple spaces; marked by <-- HERE
2773 in %s */
2d8eb851
KW
2774 yyerror_pv(
2775 Perl_form(aTHX_
2776 "charnames alias definitions may not contain a sequence of "
2777 "multiple spaces; marked by <-- HERE in %.*s<-- HERE %.*s",
2778 (int)(s - backslash_ptr + 1), backslash_ptr,
2779 (int)(e - s + 1), s + 1
2780 ),
2781 UTF ? SVf_UTF8 : 0);
2782 return NULL;
6f613c73
KW
2783}
2784
02aa26ce
NT
2785/*
2786 scan_const
2787
9da1dd8f
DM
2788 Extracts the next constant part of a pattern, double-quoted string,
2789 or transliteration. This is terrifying code.
2790
2791 For example, in parsing the double-quoted string "ab\x63$d", it would
2792 stop at the '$' and return an OP_CONST containing 'abc'.
02aa26ce 2793
94def140 2794 It looks at PL_lex_inwhat and PL_lex_inpat to find out whether it's
3280af22 2795 processing a pattern (PL_lex_inpat is true), a transliteration
94def140 2796 (PL_lex_inwhat == OP_TRANS is true), or a double-quoted string.
02aa26ce 2797
94def140
ST
2798 Returns a pointer to the character scanned up to. If this is
2799 advanced from the start pointer supplied (i.e. if anything was
9da1dd8f 2800 successfully parsed), will leave an OP_CONST for the substring scanned
6154021b 2801 in pl_yylval. Caller must intuit reason for not parsing further
9b599b2a
GS
2802 by looking at the next characters herself.
2803
02aa26ce 2804 In patterns:
9da1dd8f 2805 expand:
537124e4
KW
2806 \N{FOO} => \N{U+hex_for_character_FOO}
2807 (if FOO expands to multiple characters, expands to \N{U+xx.XX.yy ...})
9da1dd8f
DM
2808
2809 pass through:
2810 all other \-char, including \N and \N{ apart from \N{ABC}
2811
2812 stops on:
2813 @ and $ where it appears to be a var, but not for $ as tail anchor
2814 \l \L \u \U \Q \E
2815 (?{ or (??{
2816
02aa26ce
NT
2817 In transliterations:
2818 characters are VERY literal, except for - not at the start or end
66b09263
KW
2819 of the string, which indicates a range. However some backslash sequences
2820 are recognized: \r, \n, and the like
2821 \007 \o{}, \x{}, \N{}
2822 If all elements in the transliteration are below 256,
94def140
ST
2823 scan_const expands the range to the full set of intermediate
2824 characters. If the range is in utf8, the hyphen is replaced with
2825 a certain range mark which will be handled by pmtrans() in op.c.
02aa26ce
NT
2826
2827 In double-quoted strings:
2828 backslashes:
66b09263 2829 all those recognized in transliterations
94def140 2830 deprecated backrefs: \1 (in substitution replacements)
02aa26ce
NT
2831 case and quoting: \U \Q \E
2832 stops on @ and $
2833
2834 scan_const does *not* construct ops to handle interpolated strings.
2835 It stops processing as soon as it finds an embedded $ or @ variable
2836 and leaves it to the caller to work out what's going on.
2837
94def140
ST
2838 embedded arrays (whether in pattern or not) could be:
2839 @foo, @::foo, @'foo, @{foo}, @$foo, @+, @-.
2840
2841 $ in double-quoted strings must be the symbol of an embedded scalar.
02aa26ce
NT
2842
2843 $ in pattern could be $foo or could be tail anchor. Assumption:
2844 it's a tail anchor if $ is the last thing in the string, or if it's
94def140 2845 followed by one of "()| \r\n\t"
02aa26ce 2846
9da1dd8f 2847 \1 (backreferences) are turned into $1 in substitutions
02aa26ce
NT
2848
2849 The structure of the code is
2850 while (there's a character to process) {
94def140
ST
2851 handle transliteration ranges
2852 skip regexp comments /(?#comment)/ and codes /(?{code})/
2853 skip #-initiated comments in //x patterns
2854 check for embedded arrays
02aa26ce
NT
2855 check for embedded scalars
2856 if (backslash) {
94def140 2857 deprecate \1 in substitution replacements
02aa26ce
NT
2858 handle string-changing backslashes \l \U \Q \E, etc.
2859 switch (what was escaped) {
94def140 2860 handle \- in a transliteration (becomes a literal -)
ff3f963a 2861 if a pattern and not \N{, go treat as regular character
94def140
ST
2862 handle \132 (octal characters)
2863 handle \x15 and \x{1234} (hex characters)
ff3f963a 2864 handle \N{name} (named characters, also \N{3,5} in a pattern)
94def140
ST
2865 handle \cV (control characters)
2866 handle printf-style backslashes (\f, \r, \n, etc)
02aa26ce 2867 } (end switch)
77a135fe 2868 continue
02aa26ce 2869 } (end if backslash)
77a135fe 2870 handle regular character
02aa26ce 2871 } (end while character to read)
02cd137d 2872
02aa26ce
NT
2873*/
2874
76e3520e 2875STATIC char *
cea2e8a9 2876S_scan_const(pTHX_ char *start)
79072805 2877{
eb578fdb 2878 char *send = PL_bufend; /* end of the constant */
dc023dbb
KW
2879 SV *sv = newSV(send - start); /* sv for the constant. See note below
2880 on sizing. */
eb578fdb
KW
2881 char *s = start; /* start of the constant */
2882 char *d = SvPVX(sv); /* destination for copies */
dc023dbb
KW
2883 bool dorange = FALSE; /* are we in a translit range? */
2884 bool didrange = FALSE; /* did we just finish a range? */
2885 bool in_charclass = FALSE; /* within /[...]/ */
2886 bool has_utf8 = FALSE; /* Output constant is UTF8 */
2887 bool this_utf8 = cBOOL(UTF); /* Is the source string assumed to be
2888 UTF8? But, this can show as true
2889 when the source isn't utf8, as for
2890 example when it is entirely composed
2891 of hex constants */
af9be36c
KW
2892 STRLEN utf8_variant_count = 0; /* When not in UTF-8, this counts the
2893 number of characters found so far
2894 that will expand (into 2 bytes)
2895 should we have to convert to
2896 UTF-8) */
6f613c73 2897 SV *res; /* result from charnames */
c89db733
JH
2898 STRLEN offset_to_max = 0; /* The offset in the output to where the range
2899 high-end character is temporarily placed */
77a135fe 2900
fe2ba0a2
KW
2901 /* Does something require special handling in tr/// ? This avoids extra
2902 * work in a less likely case. As such, khw didn't feel it was worth
2903 * adding any branches to the more mainline code to handle this, which
2904 * means that this doesn't get set in some circumstances when things like
2905 * \x{100} get expanded out. As a result there needs to be extra testing
2906 * done in the tr code */
2907 bool has_above_latin1 = FALSE;
2908
77a135fe
KW
2909 /* Note on sizing: The scanned constant is placed into sv, which is
2910 * initialized by newSV() assuming one byte of output for every byte of
2911 * input. This routine expects newSV() to allocate an extra byte for a
2912 * trailing NUL, which this routine will append if it gets to the end of
2913 * the input. There may be more bytes of input than output (eg., \N{LATIN
2914 * CAPITAL LETTER A}), or more output than input if the constant ends up
2915 * recoded to utf8, but each time a construct is found that might increase
2916 * the needed size, SvGROW() is called. Its size parameter each time is
2917 * based on the best guess estimate at the time, namely the length used so
2918 * far, plus the length the current construct will occupy, plus room for
caae0700 2919 * the trailing NUL, plus one byte for every input byte still unscanned */
77a135fe 2920
c3320c2a
KW
2921 UV uv = UV_MAX; /* Initialize to weird value to try to catch any uses
2922 before set */
4c3a8340 2923#ifdef EBCDIC
f4240379
KW
2924 int backslash_N = 0; /* ? was the character from \N{} */
2925 int non_portable_endpoint = 0; /* ? In a range is an endpoint
2926 platform-specific like \x65 */
4c3a8340 2927#endif
012bcf8d 2928
7918f24d
NC
2929 PERL_ARGS_ASSERT_SCAN_CONST;
2930
bb16bae8 2931 assert(PL_lex_inwhat != OP_TRANSR);
7ef70b3d 2932 if (PL_lex_inwhat == OP_TRANS && PL_parser->lex_sub_op) {
2b9d42f0 2933 /* If we are doing a trans and we know we want UTF8 set expectation */
7ef70b3d
FC
2934 has_utf8 = PL_parser->lex_sub_op->op_private & (OPpTRANS_FROM_UTF|OPpTRANS_TO_UTF);
2935 this_utf8 = PL_parser->lex_sub_op->op_private & (PL_lex_repl ? OPpTRANS_FROM_UTF : OPpTRANS_TO_UTF);
2b9d42f0
NIS
2936 }
2937
b899e89d
FC
2938 /* Protect sv from errors and fatal warnings. */
2939 ENTER_with_name("scan_const");
2940 SAVEFREESV(sv);
2b9d42f0 2941
f4240379
KW
2942 while (s < send
2943 || dorange /* Handle tr/// range at right edge of input */
2944 ) {
ff3f963a 2945
02aa26ce 2946 /* get transliterations out of the way (they're most literal) */
3280af22 2947 if (PL_lex_inwhat == OP_TRANS) {
02aa26ce 2948
f4240379
KW
2949 /* But there isn't any special handling necessary unless there is a
2950 * range, so for most cases we just drop down and handle the value
2951 * as any other. There are two exceptions.
2952 *
02cd137d
KW
2953 * 1. A hyphen indicates that we are actually going to have a
2954 * range. In this case, skip the '-', set a flag, then drop
f4240379
KW
2955 * down to handle what should be the end range value.
2956 * 2. After we've handled that value, the next time through, that
2957 * flag is set and we fix up the range.
2958 *
2959 * Ranges entirely within Latin1 are expanded out entirely, in
188d22cf
KW
2960 * order to make the transliteration a simple table look-up.
2961 * Ranges that extend above Latin1 have to be done differently, so
2962 * there is no advantage to expanding them here, so they are
2963 * stored here as Min, ILLEGAL_UTF8_BYTE, Max. The illegal byte
2964 * signifies a hyphen without any possible ambiguity. On EBCDIC
2965 * machines, if the range is expressed as Unicode, the Latin1
2966 * portion is expanded out even if the range extends above
2967 * Latin1. This is because each code point in it has to be
2968 * processed here individually to get its native translation */
f4240379
KW
2969
2970 if (! dorange) {
2971
02cd137d
KW
2972 /* Here, we don't think we're in a range. If the new character
2973 * is not a hyphen; or if it is a hyphen, but it's too close to
e8d55f27
TC
2974 * either edge to indicate a range, or if we haven't output any
2975 * characters yet then it's a regular character. */
2976 if (*s != '-' || s >= send - 1 || s == start || d == SvPVX(sv)) {
f4240379
KW
2977
2978 /* A regular character. Process like any other, but first
2979 * clear any flags */
2980 didrange = FALSE;
2981 dorange = FALSE;
e294cc5d 2982#ifdef EBCDIC
f4240379
KW
2983 non_portable_endpoint = 0;
2984 backslash_N = 0;
e294cc5d 2985#endif
02cd137d
KW
2986 /* The tests here for being above Latin1 and similar ones
2987 * in the following 'else' suffice to find all such
2988 * occurences in the constant, except those added by a
fe2ba0a2 2989 * backslash escape sequence, like \x{100}. Mostly, those
02cd137d 2990 * set 'has_above_latin1' as appropriate */
188d22cf
KW
2991 if (this_utf8 && UTF8_IS_ABOVE_LATIN1(*s)) {
2992 has_above_latin1 = TRUE;
2993 }
2994
f4240379
KW
2995 /* Drops down to generic code to process current byte */
2996 }
02cd137d 2997 else { /* Is a '-' in the context where it means a range */
f4240379 2998 if (didrange) { /* Something like y/A-C-Z// */
02cd137d
KW
2999 Perl_croak(aTHX_ "Ambiguous range in transliteration"
3000 " operator");
f4240379 3001 }
e294cc5d 3002
f4240379 3003 dorange = TRUE;
2b9d42f0 3004
02cd137d 3005 s++; /* Skip past the hyphen */
f4240379
KW
3006
3007 /* d now points to where the end-range character will be
3008 * placed. Save it so won't have to go finding it later,
3009 * and drop down to get that character. (Actually we
3010 * instead save the offset, to handle the case where a
3011 * realloc in the meantime could change the actual
3012 * pointer). We'll finish processing the range the next
3013 * time through the loop */
3014 offset_to_max = d - SvPVX_const(sv);
188d22cf
KW
3015
3016 if (this_utf8 && UTF8_IS_ABOVE_LATIN1(*s)) {
3017 has_above_latin1 = TRUE;
3018 }
02cd137d
KW
3019
3020 /* Drops down to generic code to process current byte */
f4240379
KW
3021 }
3022 } /* End of not a range */
3023 else {
3024 /* Here we have parsed a range. Now must handle it. At this
3025 * point:
3026 * 'sv' is a SV* that contains the output string we are
3027 * constructing. The final two characters in that string
3028 * are the range start and range end, in order.
3029 * 'd' points to just beyond the range end in the 'sv' string,
3030 * where we would next place something
3031 * 'offset_to_max' is the offset in 'sv' at which the character
02cd137d 3032 * (the range's maximum end point) before 'd' begins.
f4240379 3033 */
8efef67c 3034 char * max_ptr = SvPVX(sv) + offset_to_max;
5ca37d54 3035 char * min_ptr;
f4240379
KW
3036 IV range_min;
3037 IV range_max; /* last character in range */
f4240379 3038 STRLEN grow;
5ca37d54
KW
3039 Size_t offset_to_min = 0;
3040 Size_t extras = 0;
11327fa1 3041#ifdef EBCDIC
f4240379
KW
3042 bool convert_unicode;
3043 IV real_range_max = 0;
e294cc5d 3044#endif
02cd137d 3045 /* Get the code point values of the range ends. */
e294cc5d 3046 if (has_utf8) {
f4240379
KW
3047 /* We know the utf8 is valid, because we just constructed
3048 * it ourselves in previous loop iterations */
3049 min_ptr = (char*) utf8_hop( (U8*) max_ptr, -1);
3050 range_min = valid_utf8_to_uvchr( (U8*) min_ptr, NULL);
3051 range_max = valid_utf8_to_uvchr( (U8*) max_ptr, NULL);
fe2ba0a2
KW
3052
3053 /* This compensates for not all code setting
3054 * 'has_above_latin1', so that we don't skip stuff that
3055 * should be executed */
3056 if (range_max > 255) {
3057 has_above_latin1 = TRUE;
3058 }
e294cc5d 3059 }
f4240379
KW
3060 else {
3061 min_ptr = max_ptr - 1;
3062 range_min = * (U8*) min_ptr;
3063 range_max = * (U8*) max_ptr;
3064 }
3065
8efef67c
KW
3066 /* If the range is just a single code point, like tr/a-a/.../,
3067 * that code point is already in the output, twice. We can
3068 * just back up over the second instance and avoid all the rest
3069 * of the work. But if it is a variant character, it's been
218304f9
KW
3070 * counted twice, so decrement. (This unlikely scenario is
3071 * special cased, like the one for a range of 2 code points
3072 * below, only because the main-line code below needs a range
3073 * of 3 or more to work without special casing. Might as well
3074 * get it out of the way now.) */
8efef67c
KW
3075 if (UNLIKELY(range_max == range_min)) {
3076 d = max_ptr;
3077 if (! has_utf8 && ! UVCHR_IS_INVARIANT(range_max)) {
3078 utf8_variant_count--;
3079 }
3080 goto range_done;
3081 }
3082
e294cc5d 3083#ifdef EBCDIC
f4240379
KW
3084 /* On EBCDIC platforms, we may have to deal with portable
3085 * ranges. These happen if at least one range endpoint is a
3086 * Unicode value (\N{...}), or if the range is a subset of
3087 * [A-Z] or [a-z], and both ends are literal characters,
3088 * like 'A', and not like \x{C1} */
87dd6ea7 3089 convert_unicode =
02cd137d
KW
3090 cBOOL(backslash_N) /* \N{} forces Unicode,
3091 hence portable range */
3092 || ( ! non_portable_endpoint
3093 && (( isLOWER_A(range_min) && isLOWER_A(range_max))
3094 || (isUPPER_A(range_min) && isUPPER_A(range_max))));
87dd6ea7 3095 if (convert_unicode) {
f4240379
KW
3096
3097 /* Special handling is needed for these portable ranges.
02cd137d
KW
3098 * They are defined to be in Unicode terms, which includes
3099 * all the Unicode code points between the end points.
f4240379
KW
3100 * Convert to Unicode to get the Unicode range. Later we
3101 * will convert each code point in the range back to
3102 * native. */
3103 range_min = NATIVE_TO_UNI(range_min);
3104 range_max = NATIVE_TO_UNI(range_max);
3105 }
e294cc5d 3106#endif
8ada0baa 3107
f4240379 3108 if (range_min > range_max) {
11327fa1 3109#ifdef EBCDIC
f4240379
KW
3110 if (convert_unicode) {
3111 /* Need to convert back to native for meaningful
3112 * messages for this platform */
3113 range_min = UNI_TO_NATIVE(range_min);
3114 range_max = UNI_TO_NATIVE(range_max);
3115 }
11327fa1 3116#endif
f4240379
KW
3117 /* Use the characters themselves for the error message if
3118 * ASCII printables; otherwise some visible representation
3119 * of them */
3120 if (isPRINT_A(range_min) && isPRINT_A(range_max)) {
3121 Perl_croak(aTHX_
3122 "Invalid range \"%c-%c\" in transliteration operator",
3123 (char)range_min, (char)range_max);
3124 }
11327fa1 3125#ifdef EBCDIC
f4240379 3126 else if (convert_unicode) {
02cd137d 3127 /* diag_listed_as: Invalid range "%s" in transliteration operator */
f4240379 3128 Perl_croak(aTHX_
02cd137d
KW
3129 "Invalid range \"\\N{U+%04" UVXf "}-\\N{U+%04"
3130 UVXf "}\" in transliteration operator",
3131 range_min, range_max);
f4240379 3132 }
11327fa1 3133#endif
f4240379 3134 else {
02cd137d 3135 /* diag_listed_as: Invalid range "%s" in transliteration operator */
f4240379 3136 Perl_croak(aTHX_
02cd137d
KW
3137 "Invalid range \"\\x{%04" UVXf "}-\\x{%04" UVXf "}\""
3138 " in transliteration operator",
3139 range_min, range_max);
f4240379 3140 }
c2e66d9e
GS
3141 }
3142
aca41667
KW
3143 /* If the range is exactly two code points long, they are
3144 * already both in the output */
3145 if (UNLIKELY(range_min + 1 == range_max)) {
3146 goto range_done;
3147 }
3148
3149 /* Here the range contains at least 3 code points */
3150
f4240379
KW
3151 if (has_utf8) {
3152
188d22cf
KW
3153 /* If everything in the transliteration is below 256, we
3154 * can avoid special handling later. A translation table
02cd137d
KW
3155 * for each of those bytes is created by op.c. So we
3156 * expand out all ranges to their constituent code points.
3157 * But if we've encountered something above 255, the
3158 * expanding won't help, so skip doing that. But if it's
3159 * EBCDIC, we may have to look at each character below 256
3160 * if we have to convert to/from Unicode values */
188d22cf 3161 if ( has_above_latin1
c7f1f016 3162#ifdef EBCDIC
f4240379 3163 && (range_min > 255 || ! convert_unicode)
8ada0baa 3164#endif
f4240379
KW
3165 ) {
3166 /* Move the high character one byte to the right; then
3167 * insert between it and the range begin, an illegal
3168 * byte which serves to indicate this is a range (using
02cd137d 3169 * a '-' would be ambiguous). */
f4240379
KW
3170 char *e = d++;
3171 while (e-- > max_ptr) {
3172 *(e + 1) = *e;
e294cc5d 3173 }
f4240379
KW
3174 *(e + 1) = (char) ILLEGAL_UTF8_BYTE;
3175 goto range_done;
3176 }
3177
3178 /* Here, we're going to expand out the range. For EBCDIC
3179 * the range can extend above 255 (not so in ASCII), so
3180 * for EBCDIC, split it into the parts above and below
3181 * 255/256 */
e294cc5d 3182#ifdef EBCDIC
f4240379
KW
3183 if (range_max > 255) {
3184 real_range_max = range_max;
3185 range_max = 255;
3186 }
e294cc5d 3187#endif
f4240379 3188 }
02aa26ce 3189
f4240379 3190 /* Here we need to expand out the string to contain each
5ca37d54
KW
3191 * character in the range. Grow the output to handle this.
3192 * For non-UTF8, we need a byte for each code point in the
3193 * range, minus the three that we've already allocated for: the
3194 * hyphen, the min, and the max. For UTF-8, we need this
3195 * plus an extra byte for each code point that occupies two
3196 * bytes (is variant) when in UTF-8 (except we've already
3197 * allocated for the end points, including if they are
3198 * variants). For ASCII platforms and Unicode ranges on EBCDIC
3199 * platforms, it's easy to calculate a precise number. To
3200 * start, we count the variants in the range, which we need
3201 * elsewhere in this function anyway. (For the case where it
3202 * isn't easy to calculate, 'extras' has been initialized to 0,
3203 * and the calculation is done in a loop further down.) */
3204#ifdef EBCDIC
3205 if (convert_unicode)
3206#endif
3207 {
3208 /* This is executed unconditionally on ASCII, and for
3209 * Unicode ranges on EBCDIC. Under these conditions, all
3210 * code points above a certain value are variant; and none
3211 * under that value are. We just need to find out how much
3212 * of the range is above that value. We don't count the
3213 * end points here, as they will already have been counted
3214 * as they were parsed. */
3215 if (range_min >= UTF_CONTINUATION_MARK) {
3216
3217 /* The whole range is made up of variants */
3218 extras = (range_max - 1) - (range_min + 1) + 1;
3219 }
3220 else if (range_max >= UTF_CONTINUATION_MARK) {
f4240379 3221
5ca37d54
KW
3222 /* Only the higher portion of the range is variants */
3223 extras = (range_max - 1) - UTF_CONTINUATION_MARK + 1;
3224 }
f4240379 3225
5ca37d54
KW
3226 utf8_variant_count += extras;
3227 }
3228
3229 /* The base growth is the number of code points in the range,
3230 * not including the endpoints, which have already been sized
3231 * for (and output). We don't subtract for the hyphen, as it
3232 * has been parsed but not output, and the SvGROW below is
3233 * based only on what's been output plus what's left to parse.
3234 * */
3235 grow = (range_max - 1) - (range_min + 1) + 1;
f4240379 3236
5ca37d54 3237 if (has_utf8) {
4c3a8340 3238#ifdef EBCDIC
5ca37d54
KW
3239 /* In some cases in EBCDIC, we haven't yet calculated a
3240 * precise amount needed for the UTF-8 variants. Just
3241 * assume the worst case, that everything will expand by a
3242 * byte */
3243 if (! convert_unicode) {
f4240379
KW
3244 grow *= 2;
3245 }
5ca37d54 3246 else
4c3a8340 3247#endif
5ca37d54
KW
3248 {
3249 /* Otherwise we know exactly how many variants there
3250 * are in the range. */
3251 grow += extras;
3252 }
f4240379
KW
3253 }
3254
5ca37d54
KW
3255 /* Grow, but position the output to overwrite the range min end
3256 * point, because in some cases we overwrite that */
3257 SvCUR_set(sv, d - SvPVX_const(sv));
3258 offset_to_min = min_ptr - SvPVX_const(sv);
3259
3260 /* See Note on sizing above. */
3261 d = offset_to_min + SvGROW(sv, SvCUR(sv)
3262 + (send - s)
3263 + grow
3264 + 1 /* Trailing NUL */ );
f4240379 3265
5ca37d54 3266 /* Now, we can expand out the range. */
11327fa1 3267#ifdef EBCDIC
f4240379 3268 if (convert_unicode) {
5ca37d54 3269 SSize_t i;
02aa26ce 3270
f4240379
KW
3271 /* Recall that the min and max are now in Unicode terms, so
3272 * we have to convert each character to its native
3273 * equivalent */
3274 if (has_utf8) {
3275 for (i = range_min; i <= range_max; i++) {
02cd137d
KW
3276 append_utf8_from_native_byte(
3277 LATIN1_TO_NATIVE((U8) i),
3278 (U8 **) &d);
f4240379
KW
3279 }
3280 }
3281 else {
3282 for (i = range_min; i <= range_max; i++) {
81324705 3283 *d++ = (char)LATIN1_TO_NATIVE((U8) i);
f4240379
KW
3284 }
3285 }
01ec43d0 3286 }
11327fa1
AL
3287 else
3288#endif
3289 /* Always gets run for ASCII, and sometimes for EBCDIC. */
3290 {
f4240379
KW
3291 /* Here, no conversions are necessary, which means that the
3292 * first character in the range is already in 'd' and
3293 * valid, so we can skip overwriting it */
3294 if (has_utf8) {
19742f39 3295 SSize_t i;
f4240379
KW
3296 d += UTF8SKIP(d);
3297 for (i = range_min + 1; i <= range_max; i++) {
81324705 3298 append_utf8_from_native_byte((U8) i, (U8 **) &d);
f4240379
KW
3299 }
3300 }
3301 else {
19742f39 3302 SSize_t i;
f4240379 3303 d++;
5ca37d54
KW
3304 assert(range_min + 1 <= range_max);
3305 for (i = range_min + 1; i < range_max; i++) {
3306#ifdef EBCDIC
3307 /* In this case on EBCDIC, we haven't calculated
3308 * the variants. Do it here, as we go along */
3309 if (! UVCHR_IS_INVARIANT(i)) {
3310 utf8_variant_count++;
3311 }
3312#endif
f4240379
KW
3313 *d++ = (char)i;
3314 }
5ca37d54
KW
3315
3316 /* The range_max is done outside the loop so as to
3317 * avoid having to special case not incrementing
3318 * 'utf8_variant_count' on EBCDIC (it's already been
3319 * counted when originally parsed) */
3320 *d++ = (char) range_max;
f4240379 3321 }
a0ed51b3 3322 }
02aa26ce 3323
11327fa1 3324#ifdef EBCDIC
02cd137d
KW
3325 /* If the original range extended above 255, add in that
3326 * portion. */
f4240379
KW
3327 if (real_range_max) {
3328 *d++ = (char) UTF8_TWO_BYTE_HI(0x100);
3329 *d++ = (char) UTF8_TWO_BYTE_LO(0x100);
043af6f6
KW
3330 if (real_range_max > 0x100) {
3331 if (real_range_max > 0x101) {
3332 *d++ = (char) ILLEGAL_UTF8_BYTE;
3333 }
f4240379 3334 d = (char*)uvchr_to_utf8((U8*)d, real_range_max);
043af6f6 3335 }
f4240379 3336 }
11327fa1 3337#endif
02aa26ce 3338
f4240379
KW
3339 range_done:
3340 /* mark the range as done, and continue */
3341 didrange = TRUE;
3342 dorange = FALSE;
3343#ifdef EBCDIC
3344 non_portable_endpoint = 0;
3345 backslash_N = 0;
3346#endif
3347 continue;
3348 } /* End of is a range */
3349 } /* End of transliteration. Joins main code after these else's */
e4a2df84
DM
3350 else if (*s == '[' && PL_lex_inpat && !in_charclass) {
3351 char *s1 = s-1;
3352 int esc = 0;
3353 while (s1 >= start && *s1-- == '\\')
3354 esc = !esc;
3355 if (!esc)
3356 in_charclass = TRUE;
3357 }
1e02a175 3358 else if (*s == ']' && PL_lex_inpat && in_charclass) {
e4a2df84
DM
3359 char *s1 = s-1;
3360 int esc = 0;
3361 while (s1 >= start && *s1-- == '\\')
3362 esc = !esc;
3363 if (!esc)
3364 in_charclass = FALSE;
3365 }
02cd137d
KW
3366 /* skip for regexp comments /(?#comment)/, except for the last
3367 * char, which will be done separately. Stop on (?{..}) and
3368 * friends */
c30fc27b 3369 else if (*s == '(' && PL_lex_inpat && s[1] == '?' && !in_charclass) {
cc6b7395 3370 if (s[2] == '#') {
e994fd66 3371 while (s+1 < send && *s != ')')
5ff03569 3372 *d++ = *s++;
155aba94 3373 }
407f8cf2
KW
3374 else if (!PL_lex_casemods
3375 && ( s[2] == '{' /* This should match regcomp.c */
3376 || (s[2] == '?' && s[3] == '{')))
155aba94 3377 {
9da1dd8f 3378 break;
cc6b7395 3379 }
748a9306 3380 }
02cd137d 3381 /* likewise skip #-initiated comments in //x patterns */
407f8cf2
KW
3382 else if (*s == '#'
3383 && PL_lex_inpat
3384 && !in_charclass
3385 && ((PMOP*)PL_lex_inpat)->op_pmflags & RXf_PMf_EXTENDED)
3386 {
8faf4f30 3387 while (s < send && *s != '\n')
5ff03569 3388 *d++ = *s++;
748a9306 3389 }
02cd137d 3390 /* no further processing of single-quoted regex */
9da1dd8f
DM
3391 else if (PL_lex_inpat && SvIVX(PL_linestr) == '\'')
3392 goto default_action;
3393
02cd137d
KW
3394 /* check for embedded arrays
3395 * (@foo, @::foo, @'foo, @{foo}, @$foo, @+, @-)
3396 */
1749ea0d 3397 else if (*s == '@' && s[1]) {
fac0f7a3
KW
3398 if (UTF
3399 ? isIDFIRST_utf8_safe(s+1, send)
3400 : isWORDCHAR_A(s[1]))
3401 {
1749ea0d 3402 break;
fac0f7a3 3403 }
1749ea0d
ST
3404 if (strchr(":'{$", s[1]))
3405 break;
3406 if (!PL_lex_inpat && (s[1] == '+' || s[1] == '-'))
3407 break; /* in regexp, neither @+ nor @- are interpolated */
3408 }
02cd137d
KW
3409 /* check for embedded scalars. only stop if we're sure it's a
3410 * variable. */
79072805 3411 else if (*s == '$') {
3280af22 3412 if (!PL_lex_inpat) /* not a regexp, so $ must be var */
79072805 3413 break;
77772344 3414 if (s + 1 < send && !strchr("()| \r\n\t", s[1])) {
a2a5de95
NC
3415 if (s[1] == '\\') {
3416 Perl_ck_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
3417 "Possible unintended interpolation of $\\ in regex");
77772344 3418 }
79072805 3419 break; /* in regexp, $ might be tail anchor */
77772344 3420 }
79072805 3421 }
02aa26ce 3422
2b9d42f0
NIS
3423 /* End of else if chain - OP_TRANS rejoin rest */
3424
8faf4f30
HS
3425 if (UNLIKELY(s >= send)) {
3426 assert(s == send);
3427 break;
3428 }
3429
02aa26ce 3430 /* backslashes */
79072805 3431 if (*s == '\\' && s+1 < send) {
ff3f963a
KW
3432 char* e; /* Can be used for ending '}', etc. */
3433
79072805 3434 s++;
02aa26ce 3435
7d0fc23c
KW
3436 /* warn on \1 - \9 in substitution replacements, but note that \11
3437 * is an octal; and \19 is \1 followed by '9' */
407f8cf2
KW
3438 if (PL_lex_inwhat == OP_SUBST
3439 && !PL_lex_inpat
3440 && isDIGIT(*s)
3441 && *s != '0'
3442 && !isDIGIT(s[1]))
79072805 3443 {
c782d7ee 3444 /* diag_listed_as: \%d better written as $%d */
a2a5de95 3445 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX), "\\%c better written as $%c", *s, *s);
79072805
LW
3446 *--s = '$';
3447 break;
3448 }
02aa26ce
NT
3449
3450 /* string-change backslash escapes */
838f2281 3451 if (PL_lex_inwhat != OP_TRANS && *s && strchr("lLuUEQF", *s)) {
79072805
LW
3452 --s;
3453 break;
3454 }
ff3f963a
KW
3455 /* In a pattern, process \N, but skip any other backslash escapes.
3456 * This is because we don't want to translate an escape sequence
3457 * into a meta symbol and have the regex compiler use the meta
3458 * symbol meaning, e.g. \x{2E} would be confused with a dot. But
3459 * in spite of this, we do have to process \N here while the proper
3460 * charnames handler is in scope. See bugs #56444 and #62056.
85fba779 3461 *
ff3f963a
KW
3462 * There is a complication because \N in a pattern may also stand
3463 * for 'match a non-nl', and not mean a charname, in which case its
3464 * processing should be deferred to the regex compiler. To be a
3465 * charname it must be followed immediately by a '{', and not look
3466 * like \N followed by a curly quantifier, i.e., not something like
3467 * \N{3,}. regcurly returns a boolean indicating if it is a legal
3468 * quantifier */
3469 else if (PL_lex_inpat
3470 && (*s != 'N'
3471 || s[1] != '{'
412f55bb 3472 || regcurly(s + 1)))
ff3f963a 3473 {
4d73d076 3474 *d++ = '\\';
cc74c5bd
ST
3475 goto default_action;
3476 }
02aa26ce 3477
79072805 3478 switch (*s) {
79072805 3479 default:
11b8faa4 3480 {
15861f94 3481 if ((isALPHANUMERIC(*s)))
a2a5de95
NC
3482 Perl_ck_warner(aTHX_ packWARN(WARN_MISC),
3483 "Unrecognized escape \\%c passed through",
3484 *s);
11b8faa4 3485 /* default action is to copy the quoted character */
f9a63242 3486 goto default_action;
11b8faa4 3487 }
02aa26ce 3488
632403cc 3489 /* eg. \132 indicates the octal constant 0132 */
79072805
LW
3490 case '0': case '1': case '2': case '3':
3491 case '4': case '5': case '6': case '7':
ba210ebe 3492 {
5e0a247b 3493 I32 flags = PERL_SCAN_SILENT_ILLDIGIT;
53305cf1 3494 STRLEN len = 3;
06972766 3495 uv = grok_oct(s, &len, &flags, NULL);
ba210ebe 3496 s += len;
5e0a247b
KW
3497 if (len < 3 && s < send && isDIGIT(*s)
3498 && ckWARN(WARN_MISC))
3499 {
3500 Perl_warner(aTHX_ packWARN(WARN_MISC),
3501 "%s", form_short_octal_warning(s, len));
3502 }
ba210ebe 3503 }
012bcf8d 3504 goto NUM_ESCAPE_INSERT;
02aa26ce 3505
f0a2b745
KW
3506 /* eg. \o{24} indicates the octal constant \024 */
3507 case 'o':
3508 {
454155d9 3509 const char* error;
f0a2b745 3510
e8278639
KW
3511 bool valid = grok_bslash_o(&s, PL_bufend,
3512 &uv, &error,
80f4111b
KW
3513 TRUE, /* Output warning */
3514 FALSE, /* Not strict */
17896a85
KW
3515 TRUE, /* Output warnings for
3516 non-portables */
80f4111b 3517 UTF);
454155d9 3518 if (! valid) {
f0a2b745 3519 yyerror(error);
3dd4eaeb 3520 uv = 0; /* drop through to ensure range ends are set */
f0a2b745
KW
3521 }
3522 goto NUM_ESCAPE_INSERT;
3523 }
3524
77a135fe 3525 /* eg. \x24 indicates the hex constant 0x24 */
79072805 3526 case 'x':
a0481293 3527 {
a0481293 3528 const char* error;
355860ce 3529
e8278639
KW
3530 bool valid = grok_bslash_x(&s, PL_bufend,
3531 &uv, &error,
80f4111b
KW
3532 TRUE, /* Output warning */
3533 FALSE, /* Not strict */
17896a85
KW
3534 TRUE, /* Output warnings for
3535 non-portables */
80f4111b 3536 UTF);
a0481293
KW
3537 if (! valid) {
3538 yyerror(error);
3dd4eaeb 3539 uv = 0; /* drop through to ensure range ends are set */
ba210ebe 3540 }
012bcf8d
GS
3541 }
3542
3543 NUM_ESCAPE_INSERT:
b67fd2c5 3544 /* Insert oct or hex escaped character. */
02cd137d 3545
06972766 3546 /* Here uv is the ordinal of the next character being added */
f4240379
KW
3547 if (UVCHR_IS_INVARIANT(uv)) {
3548 *d++ = (char) uv;
3549 }
3550 else {
9aa983d2 3551 if (!has_utf8 && uv > 255) {
caae0700