2012-05-18 Sergio Durigan Junior <sergiodj@redhat.com>
[deliverable/binutils-gdb.git] / gdb / c-exp.y
1 /* YACC parser for C expressions, for GDB.
2 Copyright (C) 1986, 1989-2000, 2003-2004, 2006-2012 Free Software
3 Foundation, Inc.
4
5 This file is part of GDB.
6
7 This program is free software; you can redistribute it and/or modify
8 it under the terms of the GNU General Public License as published by
9 the Free Software Foundation; either version 3 of the License, or
10 (at your option) any later version.
11
12 This program is distributed in the hope that it will be useful,
13 but WITHOUT ANY WARRANTY; without even the implied warranty of
14 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
15 GNU General Public License for more details.
16
17 You should have received a copy of the GNU General Public License
18 along with this program. If not, see <http://www.gnu.org/licenses/>. */
19
20 /* Parse a C expression from text in a string,
21 and return the result as a struct expression pointer.
22 That structure contains arithmetic operations in reverse polish,
23 with constants represented by operations that are followed by special data.
24 See expression.h for the details of the format.
25 What is important here is that it can be built up sequentially
26 during the process of parsing; the lower levels of the tree always
27 come first in the result.
28
29 Note that malloc's and realloc's in this file are transformed to
30 xmalloc and xrealloc respectively by the same sed command in the
31 makefile that remaps any other malloc/realloc inserted by the parser
32 generator. Doing this with #defines and trying to control the interaction
33 with include files (<malloc.h> and <stdlib.h> for example) just became
34 too messy, particularly when such includes can be inserted at random
35 times by the parser generator. */
36
37 %{
38
39 #include "defs.h"
40 #include "gdb_string.h"
41 #include <ctype.h>
42 #include "expression.h"
43 #include "value.h"
44 #include "parser-defs.h"
45 #include "language.h"
46 #include "c-lang.h"
47 #include "bfd.h" /* Required by objfiles.h. */
48 #include "symfile.h" /* Required by objfiles.h. */
49 #include "objfiles.h" /* For have_full_symbols and have_partial_symbols */
50 #include "charset.h"
51 #include "block.h"
52 #include "cp-support.h"
53 #include "dfp.h"
54 #include "gdb_assert.h"
55 #include "macroscope.h"
56
57 #define parse_type builtin_type (parse_gdbarch)
58
59 /* Remap normal yacc parser interface names (yyparse, yylex, yyerror, etc),
60 as well as gratuitiously global symbol names, so we can have multiple
61 yacc generated parsers in gdb. Note that these are only the variables
62 produced by yacc. If other parser generators (bison, byacc, etc) produce
63 additional global names that conflict at link time, then those parser
64 generators need to be fixed instead of adding those names to this list. */
65
66 #define yymaxdepth c_maxdepth
67 #define yyparse c_parse_internal
68 #define yylex c_lex
69 #define yyerror c_error
70 #define yylval c_lval
71 #define yychar c_char
72 #define yydebug c_debug
73 #define yypact c_pact
74 #define yyr1 c_r1
75 #define yyr2 c_r2
76 #define yydef c_def
77 #define yychk c_chk
78 #define yypgo c_pgo
79 #define yyact c_act
80 #define yyexca c_exca
81 #define yyerrflag c_errflag
82 #define yynerrs c_nerrs
83 #define yyps c_ps
84 #define yypv c_pv
85 #define yys c_s
86 #define yy_yys c_yys
87 #define yystate c_state
88 #define yytmp c_tmp
89 #define yyv c_v
90 #define yy_yyv c_yyv
91 #define yyval c_val
92 #define yylloc c_lloc
93 #define yyreds c_reds /* With YYDEBUG defined */
94 #define yytoks c_toks /* With YYDEBUG defined */
95 #define yyname c_name /* With YYDEBUG defined */
96 #define yyrule c_rule /* With YYDEBUG defined */
97 #define yylhs c_yylhs
98 #define yylen c_yylen
99 #define yydefred c_yydefred
100 #define yydgoto c_yydgoto
101 #define yysindex c_yysindex
102 #define yyrindex c_yyrindex
103 #define yygindex c_yygindex
104 #define yytable c_yytable
105 #define yycheck c_yycheck
106 #define yyss c_yyss
107 #define yysslim c_yysslim
108 #define yyssp c_yyssp
109 #define yystacksize c_yystacksize
110 #define yyvs c_yyvs
111 #define yyvsp c_yyvsp
112
113 #ifndef YYDEBUG
114 #define YYDEBUG 1 /* Default to yydebug support */
115 #endif
116
117 #define YYFPRINTF parser_fprintf
118
119 int yyparse (void);
120
121 static int yylex (void);
122
123 void yyerror (char *);
124
125 %}
126
127 /* Although the yacc "value" of an expression is not used,
128 since the result is stored in the structure being created,
129 other node types do have values. */
130
131 %union
132 {
133 LONGEST lval;
134 struct {
135 LONGEST val;
136 struct type *type;
137 } typed_val_int;
138 struct {
139 DOUBLEST dval;
140 struct type *type;
141 } typed_val_float;
142 struct {
143 gdb_byte val[16];
144 struct type *type;
145 } typed_val_decfloat;
146 struct symbol *sym;
147 struct type *tval;
148 struct stoken sval;
149 struct typed_stoken tsval;
150 struct ttype tsym;
151 struct symtoken ssym;
152 int voidval;
153 struct block *bval;
154 enum exp_opcode opcode;
155 struct internalvar *ivar;
156
157 struct stoken_vector svec;
158 struct type **tvec;
159 int *ivec;
160 }
161
162 %{
163 /* YYSTYPE gets defined by %union */
164 static int parse_number (char *, int, int, YYSTYPE *);
165 static struct stoken operator_stoken (const char *);
166 %}
167
168 %type <voidval> exp exp1 type_exp start variable qualified_name lcurly
169 %type <lval> rcurly
170 %type <tval> type typebase
171 %type <tvec> nonempty_typelist
172 /* %type <bval> block */
173
174 /* Fancy type parsing. */
175 %type <voidval> func_mod direct_abs_decl abs_decl
176 %type <tval> ptype
177 %type <lval> array_mod
178
179 %token <typed_val_int> INT
180 %token <typed_val_float> FLOAT
181 %token <typed_val_decfloat> DECFLOAT
182
183 /* Both NAME and TYPENAME tokens represent symbols in the input,
184 and both convey their data as strings.
185 But a TYPENAME is a string that happens to be defined as a typedef
186 or builtin type name (such as int or char)
187 and a NAME is any other symbol.
188 Contexts where this distinction is not important can use the
189 nonterminal "name", which matches either NAME or TYPENAME. */
190
191 %token <tsval> STRING
192 %token <tsval> CHAR
193 %token <ssym> NAME /* BLOCKNAME defined below to give it higher precedence. */
194 %token <ssym> UNKNOWN_CPP_NAME
195 %token <voidval> COMPLETE
196 %token <tsym> TYPENAME
197 %type <sval> name
198 %type <svec> string_exp
199 %type <ssym> name_not_typename
200 %type <tsym> typename
201
202 /* A NAME_OR_INT is a symbol which is not known in the symbol table,
203 but which would parse as a valid number in the current input radix.
204 E.g. "c" when input_radix==16. Depending on the parse, it will be
205 turned into a name or into a number. */
206
207 %token <ssym> NAME_OR_INT
208
209 %token OPERATOR
210 %token STRUCT CLASS UNION ENUM SIZEOF UNSIGNED COLONCOLON
211 %token TEMPLATE
212 %token ERROR
213 %token NEW DELETE
214 %type <sval> operator
215 %token REINTERPRET_CAST DYNAMIC_CAST STATIC_CAST CONST_CAST
216 %token ENTRY
217
218 /* Special type cases, put in to allow the parser to distinguish different
219 legal basetypes. */
220 %token SIGNED_KEYWORD LONG SHORT INT_KEYWORD CONST_KEYWORD VOLATILE_KEYWORD DOUBLE_KEYWORD
221
222 %token <sval> VARIABLE
223
224 %token <opcode> ASSIGN_MODIFY
225
226 /* C++ */
227 %token TRUEKEYWORD
228 %token FALSEKEYWORD
229
230
231 %left ','
232 %left ABOVE_COMMA
233 %right '=' ASSIGN_MODIFY
234 %right '?'
235 %left OROR
236 %left ANDAND
237 %left '|'
238 %left '^'
239 %left '&'
240 %left EQUAL NOTEQUAL
241 %left '<' '>' LEQ GEQ
242 %left LSH RSH
243 %left '@'
244 %left '+' '-'
245 %left '*' '/' '%'
246 %right UNARY INCREMENT DECREMENT
247 %right ARROW ARROW_STAR '.' DOT_STAR '[' '('
248 %token <ssym> BLOCKNAME
249 %token <bval> FILENAME
250 %type <bval> block
251 %left COLONCOLON
252
253 \f
254 %%
255
256 start : exp1
257 | type_exp
258 ;
259
260 type_exp: type
261 { write_exp_elt_opcode(OP_TYPE);
262 write_exp_elt_type($1);
263 write_exp_elt_opcode(OP_TYPE);}
264 ;
265
266 /* Expressions, including the comma operator. */
267 exp1 : exp
268 | exp1 ',' exp
269 { write_exp_elt_opcode (BINOP_COMMA); }
270 ;
271
272 /* Expressions, not including the comma operator. */
273 exp : '*' exp %prec UNARY
274 { write_exp_elt_opcode (UNOP_IND); }
275 ;
276
277 exp : '&' exp %prec UNARY
278 { write_exp_elt_opcode (UNOP_ADDR); }
279 ;
280
281 exp : '-' exp %prec UNARY
282 { write_exp_elt_opcode (UNOP_NEG); }
283 ;
284
285 exp : '+' exp %prec UNARY
286 { write_exp_elt_opcode (UNOP_PLUS); }
287 ;
288
289 exp : '!' exp %prec UNARY
290 { write_exp_elt_opcode (UNOP_LOGICAL_NOT); }
291 ;
292
293 exp : '~' exp %prec UNARY
294 { write_exp_elt_opcode (UNOP_COMPLEMENT); }
295 ;
296
297 exp : INCREMENT exp %prec UNARY
298 { write_exp_elt_opcode (UNOP_PREINCREMENT); }
299 ;
300
301 exp : DECREMENT exp %prec UNARY
302 { write_exp_elt_opcode (UNOP_PREDECREMENT); }
303 ;
304
305 exp : exp INCREMENT %prec UNARY
306 { write_exp_elt_opcode (UNOP_POSTINCREMENT); }
307 ;
308
309 exp : exp DECREMENT %prec UNARY
310 { write_exp_elt_opcode (UNOP_POSTDECREMENT); }
311 ;
312
313 exp : SIZEOF exp %prec UNARY
314 { write_exp_elt_opcode (UNOP_SIZEOF); }
315 ;
316
317 exp : exp ARROW name
318 { write_exp_elt_opcode (STRUCTOP_PTR);
319 write_exp_string ($3);
320 write_exp_elt_opcode (STRUCTOP_PTR); }
321 ;
322
323 exp : exp ARROW name COMPLETE
324 { mark_struct_expression ();
325 write_exp_elt_opcode (STRUCTOP_PTR);
326 write_exp_string ($3);
327 write_exp_elt_opcode (STRUCTOP_PTR); }
328 ;
329
330 exp : exp ARROW COMPLETE
331 { struct stoken s;
332 mark_struct_expression ();
333 write_exp_elt_opcode (STRUCTOP_PTR);
334 s.ptr = "";
335 s.length = 0;
336 write_exp_string (s);
337 write_exp_elt_opcode (STRUCTOP_PTR); }
338 ;
339
340 exp : exp ARROW qualified_name
341 { /* exp->type::name becomes exp->*(&type::name) */
342 /* Note: this doesn't work if name is a
343 static member! FIXME */
344 write_exp_elt_opcode (UNOP_ADDR);
345 write_exp_elt_opcode (STRUCTOP_MPTR); }
346 ;
347
348 exp : exp ARROW_STAR exp
349 { write_exp_elt_opcode (STRUCTOP_MPTR); }
350 ;
351
352 exp : exp '.' name
353 { write_exp_elt_opcode (STRUCTOP_STRUCT);
354 write_exp_string ($3);
355 write_exp_elt_opcode (STRUCTOP_STRUCT); }
356 ;
357
358 exp : exp '.' name COMPLETE
359 { mark_struct_expression ();
360 write_exp_elt_opcode (STRUCTOP_STRUCT);
361 write_exp_string ($3);
362 write_exp_elt_opcode (STRUCTOP_STRUCT); }
363 ;
364
365 exp : exp '.' COMPLETE
366 { struct stoken s;
367 mark_struct_expression ();
368 write_exp_elt_opcode (STRUCTOP_STRUCT);
369 s.ptr = "";
370 s.length = 0;
371 write_exp_string (s);
372 write_exp_elt_opcode (STRUCTOP_STRUCT); }
373 ;
374
375 exp : exp '.' qualified_name
376 { /* exp.type::name becomes exp.*(&type::name) */
377 /* Note: this doesn't work if name is a
378 static member! FIXME */
379 write_exp_elt_opcode (UNOP_ADDR);
380 write_exp_elt_opcode (STRUCTOP_MEMBER); }
381 ;
382
383 exp : exp DOT_STAR exp
384 { write_exp_elt_opcode (STRUCTOP_MEMBER); }
385 ;
386
387 exp : exp '[' exp1 ']'
388 { write_exp_elt_opcode (BINOP_SUBSCRIPT); }
389 ;
390
391 exp : exp '('
392 /* This is to save the value of arglist_len
393 being accumulated by an outer function call. */
394 { start_arglist (); }
395 arglist ')' %prec ARROW
396 { write_exp_elt_opcode (OP_FUNCALL);
397 write_exp_elt_longcst ((LONGEST) end_arglist ());
398 write_exp_elt_opcode (OP_FUNCALL); }
399 ;
400
401 exp : UNKNOWN_CPP_NAME '('
402 {
403 /* This could potentially be a an argument defined
404 lookup function (Koenig). */
405 write_exp_elt_opcode (OP_ADL_FUNC);
406 write_exp_elt_block (expression_context_block);
407 write_exp_elt_sym (NULL); /* Placeholder. */
408 write_exp_string ($1.stoken);
409 write_exp_elt_opcode (OP_ADL_FUNC);
410
411 /* This is to save the value of arglist_len
412 being accumulated by an outer function call. */
413
414 start_arglist ();
415 }
416 arglist ')' %prec ARROW
417 {
418 write_exp_elt_opcode (OP_FUNCALL);
419 write_exp_elt_longcst ((LONGEST) end_arglist ());
420 write_exp_elt_opcode (OP_FUNCALL);
421 }
422 ;
423
424 lcurly : '{'
425 { start_arglist (); }
426 ;
427
428 arglist :
429 ;
430
431 arglist : exp
432 { arglist_len = 1; }
433 ;
434
435 arglist : arglist ',' exp %prec ABOVE_COMMA
436 { arglist_len++; }
437 ;
438
439 exp : exp '(' nonempty_typelist ')' const_or_volatile
440 { int i;
441 write_exp_elt_opcode (TYPE_INSTANCE);
442 write_exp_elt_longcst ((LONGEST) $<ivec>3[0]);
443 for (i = 0; i < $<ivec>3[0]; ++i)
444 write_exp_elt_type ($<tvec>3[i + 1]);
445 write_exp_elt_longcst((LONGEST) $<ivec>3[0]);
446 write_exp_elt_opcode (TYPE_INSTANCE);
447 free ($3);
448 }
449 ;
450
451 rcurly : '}'
452 { $$ = end_arglist () - 1; }
453 ;
454 exp : lcurly arglist rcurly %prec ARROW
455 { write_exp_elt_opcode (OP_ARRAY);
456 write_exp_elt_longcst ((LONGEST) 0);
457 write_exp_elt_longcst ((LONGEST) $3);
458 write_exp_elt_opcode (OP_ARRAY); }
459 ;
460
461 exp : lcurly type rcurly exp %prec UNARY
462 { write_exp_elt_opcode (UNOP_MEMVAL);
463 write_exp_elt_type ($2);
464 write_exp_elt_opcode (UNOP_MEMVAL); }
465 ;
466
467 exp : '(' type ')' exp %prec UNARY
468 { write_exp_elt_opcode (UNOP_CAST);
469 write_exp_elt_type ($2);
470 write_exp_elt_opcode (UNOP_CAST); }
471 ;
472
473 exp : '(' exp1 ')'
474 { }
475 ;
476
477 /* Binary operators in order of decreasing precedence. */
478
479 exp : exp '@' exp
480 { write_exp_elt_opcode (BINOP_REPEAT); }
481 ;
482
483 exp : exp '*' exp
484 { write_exp_elt_opcode (BINOP_MUL); }
485 ;
486
487 exp : exp '/' exp
488 { write_exp_elt_opcode (BINOP_DIV); }
489 ;
490
491 exp : exp '%' exp
492 { write_exp_elt_opcode (BINOP_REM); }
493 ;
494
495 exp : exp '+' exp
496 { write_exp_elt_opcode (BINOP_ADD); }
497 ;
498
499 exp : exp '-' exp
500 { write_exp_elt_opcode (BINOP_SUB); }
501 ;
502
503 exp : exp LSH exp
504 { write_exp_elt_opcode (BINOP_LSH); }
505 ;
506
507 exp : exp RSH exp
508 { write_exp_elt_opcode (BINOP_RSH); }
509 ;
510
511 exp : exp EQUAL exp
512 { write_exp_elt_opcode (BINOP_EQUAL); }
513 ;
514
515 exp : exp NOTEQUAL exp
516 { write_exp_elt_opcode (BINOP_NOTEQUAL); }
517 ;
518
519 exp : exp LEQ exp
520 { write_exp_elt_opcode (BINOP_LEQ); }
521 ;
522
523 exp : exp GEQ exp
524 { write_exp_elt_opcode (BINOP_GEQ); }
525 ;
526
527 exp : exp '<' exp
528 { write_exp_elt_opcode (BINOP_LESS); }
529 ;
530
531 exp : exp '>' exp
532 { write_exp_elt_opcode (BINOP_GTR); }
533 ;
534
535 exp : exp '&' exp
536 { write_exp_elt_opcode (BINOP_BITWISE_AND); }
537 ;
538
539 exp : exp '^' exp
540 { write_exp_elt_opcode (BINOP_BITWISE_XOR); }
541 ;
542
543 exp : exp '|' exp
544 { write_exp_elt_opcode (BINOP_BITWISE_IOR); }
545 ;
546
547 exp : exp ANDAND exp
548 { write_exp_elt_opcode (BINOP_LOGICAL_AND); }
549 ;
550
551 exp : exp OROR exp
552 { write_exp_elt_opcode (BINOP_LOGICAL_OR); }
553 ;
554
555 exp : exp '?' exp ':' exp %prec '?'
556 { write_exp_elt_opcode (TERNOP_COND); }
557 ;
558
559 exp : exp '=' exp
560 { write_exp_elt_opcode (BINOP_ASSIGN); }
561 ;
562
563 exp : exp ASSIGN_MODIFY exp
564 { write_exp_elt_opcode (BINOP_ASSIGN_MODIFY);
565 write_exp_elt_opcode ($2);
566 write_exp_elt_opcode (BINOP_ASSIGN_MODIFY); }
567 ;
568
569 exp : INT
570 { write_exp_elt_opcode (OP_LONG);
571 write_exp_elt_type ($1.type);
572 write_exp_elt_longcst ((LONGEST)($1.val));
573 write_exp_elt_opcode (OP_LONG); }
574 ;
575
576 exp : CHAR
577 {
578 struct stoken_vector vec;
579 vec.len = 1;
580 vec.tokens = &$1;
581 write_exp_string_vector ($1.type, &vec);
582 }
583 ;
584
585 exp : NAME_OR_INT
586 { YYSTYPE val;
587 parse_number ($1.stoken.ptr, $1.stoken.length, 0, &val);
588 write_exp_elt_opcode (OP_LONG);
589 write_exp_elt_type (val.typed_val_int.type);
590 write_exp_elt_longcst ((LONGEST)val.typed_val_int.val);
591 write_exp_elt_opcode (OP_LONG);
592 }
593 ;
594
595
596 exp : FLOAT
597 { write_exp_elt_opcode (OP_DOUBLE);
598 write_exp_elt_type ($1.type);
599 write_exp_elt_dblcst ($1.dval);
600 write_exp_elt_opcode (OP_DOUBLE); }
601 ;
602
603 exp : DECFLOAT
604 { write_exp_elt_opcode (OP_DECFLOAT);
605 write_exp_elt_type ($1.type);
606 write_exp_elt_decfloatcst ($1.val);
607 write_exp_elt_opcode (OP_DECFLOAT); }
608 ;
609
610 exp : variable
611 ;
612
613 exp : VARIABLE
614 {
615 write_dollar_variable ($1);
616 }
617 ;
618
619 exp : SIZEOF '(' type ')' %prec UNARY
620 { write_exp_elt_opcode (OP_LONG);
621 write_exp_elt_type (lookup_signed_typename
622 (parse_language, parse_gdbarch,
623 "int"));
624 CHECK_TYPEDEF ($3);
625 write_exp_elt_longcst ((LONGEST) TYPE_LENGTH ($3));
626 write_exp_elt_opcode (OP_LONG); }
627 ;
628
629 exp : REINTERPRET_CAST '<' type '>' '(' exp ')' %prec UNARY
630 { write_exp_elt_opcode (UNOP_REINTERPRET_CAST);
631 write_exp_elt_type ($3);
632 write_exp_elt_opcode (UNOP_REINTERPRET_CAST); }
633 ;
634
635 exp : STATIC_CAST '<' type '>' '(' exp ')' %prec UNARY
636 { write_exp_elt_opcode (UNOP_CAST);
637 write_exp_elt_type ($3);
638 write_exp_elt_opcode (UNOP_CAST); }
639 ;
640
641 exp : DYNAMIC_CAST '<' type '>' '(' exp ')' %prec UNARY
642 { write_exp_elt_opcode (UNOP_DYNAMIC_CAST);
643 write_exp_elt_type ($3);
644 write_exp_elt_opcode (UNOP_DYNAMIC_CAST); }
645 ;
646
647 exp : CONST_CAST '<' type '>' '(' exp ')' %prec UNARY
648 { /* We could do more error checking here, but
649 it doesn't seem worthwhile. */
650 write_exp_elt_opcode (UNOP_CAST);
651 write_exp_elt_type ($3);
652 write_exp_elt_opcode (UNOP_CAST); }
653 ;
654
655 string_exp:
656 STRING
657 {
658 /* We copy the string here, and not in the
659 lexer, to guarantee that we do not leak a
660 string. Note that we follow the
661 NUL-termination convention of the
662 lexer. */
663 struct typed_stoken *vec = XNEW (struct typed_stoken);
664 $$.len = 1;
665 $$.tokens = vec;
666
667 vec->type = $1.type;
668 vec->length = $1.length;
669 vec->ptr = malloc ($1.length + 1);
670 memcpy (vec->ptr, $1.ptr, $1.length + 1);
671 }
672
673 | string_exp STRING
674 {
675 /* Note that we NUL-terminate here, but just
676 for convenience. */
677 char *p;
678 ++$$.len;
679 $$.tokens = realloc ($$.tokens,
680 $$.len * sizeof (struct typed_stoken));
681
682 p = malloc ($2.length + 1);
683 memcpy (p, $2.ptr, $2.length + 1);
684
685 $$.tokens[$$.len - 1].type = $2.type;
686 $$.tokens[$$.len - 1].length = $2.length;
687 $$.tokens[$$.len - 1].ptr = p;
688 }
689 ;
690
691 exp : string_exp
692 {
693 int i;
694 enum c_string_type type = C_STRING;
695
696 for (i = 0; i < $1.len; ++i)
697 {
698 switch ($1.tokens[i].type)
699 {
700 case C_STRING:
701 break;
702 case C_WIDE_STRING:
703 case C_STRING_16:
704 case C_STRING_32:
705 if (type != C_STRING
706 && type != $1.tokens[i].type)
707 error (_("Undefined string concatenation."));
708 type = $1.tokens[i].type;
709 break;
710 default:
711 /* internal error */
712 internal_error (__FILE__, __LINE__,
713 "unrecognized type in string concatenation");
714 }
715 }
716
717 write_exp_string_vector (type, &$1);
718 for (i = 0; i < $1.len; ++i)
719 free ($1.tokens[i].ptr);
720 free ($1.tokens);
721 }
722 ;
723
724 /* C++. */
725 exp : TRUEKEYWORD
726 { write_exp_elt_opcode (OP_LONG);
727 write_exp_elt_type (parse_type->builtin_bool);
728 write_exp_elt_longcst ((LONGEST) 1);
729 write_exp_elt_opcode (OP_LONG); }
730 ;
731
732 exp : FALSEKEYWORD
733 { write_exp_elt_opcode (OP_LONG);
734 write_exp_elt_type (parse_type->builtin_bool);
735 write_exp_elt_longcst ((LONGEST) 0);
736 write_exp_elt_opcode (OP_LONG); }
737 ;
738
739 /* end of C++. */
740
741 block : BLOCKNAME
742 {
743 if ($1.sym)
744 $$ = SYMBOL_BLOCK_VALUE ($1.sym);
745 else
746 error (_("No file or function \"%s\"."),
747 copy_name ($1.stoken));
748 }
749 | FILENAME
750 {
751 $$ = $1;
752 }
753 ;
754
755 block : block COLONCOLON name
756 { struct symbol *tem
757 = lookup_symbol (copy_name ($3), $1,
758 VAR_DOMAIN, (int *) NULL);
759 if (!tem || SYMBOL_CLASS (tem) != LOC_BLOCK)
760 error (_("No function \"%s\" in specified context."),
761 copy_name ($3));
762 $$ = SYMBOL_BLOCK_VALUE (tem); }
763 ;
764
765 variable: name_not_typename ENTRY
766 { struct symbol *sym = $1.sym;
767
768 if (sym == NULL || !SYMBOL_IS_ARGUMENT (sym)
769 || !symbol_read_needs_frame (sym))
770 error (_("@entry can be used only for function "
771 "parameters, not for \"%s\""),
772 copy_name ($1.stoken));
773
774 write_exp_elt_opcode (OP_VAR_ENTRY_VALUE);
775 write_exp_elt_sym (sym);
776 write_exp_elt_opcode (OP_VAR_ENTRY_VALUE);
777 }
778 ;
779
780 variable: block COLONCOLON name
781 { struct symbol *sym;
782 sym = lookup_symbol (copy_name ($3), $1,
783 VAR_DOMAIN, (int *) NULL);
784 if (sym == 0)
785 error (_("No symbol \"%s\" in specified context."),
786 copy_name ($3));
787 if (symbol_read_needs_frame (sym))
788 {
789 if (innermost_block == 0
790 || contained_in (block_found,
791 innermost_block))
792 innermost_block = block_found;
793 }
794
795 write_exp_elt_opcode (OP_VAR_VALUE);
796 /* block_found is set by lookup_symbol. */
797 write_exp_elt_block (block_found);
798 write_exp_elt_sym (sym);
799 write_exp_elt_opcode (OP_VAR_VALUE); }
800 ;
801
802 qualified_name: TYPENAME COLONCOLON name
803 {
804 struct type *type = $1.type;
805 CHECK_TYPEDEF (type);
806 if (TYPE_CODE (type) != TYPE_CODE_STRUCT
807 && TYPE_CODE (type) != TYPE_CODE_UNION
808 && TYPE_CODE (type) != TYPE_CODE_NAMESPACE)
809 error (_("`%s' is not defined as an aggregate type."),
810 TYPE_NAME (type));
811
812 write_exp_elt_opcode (OP_SCOPE);
813 write_exp_elt_type (type);
814 write_exp_string ($3);
815 write_exp_elt_opcode (OP_SCOPE);
816 }
817 | TYPENAME COLONCOLON '~' name
818 {
819 struct type *type = $1.type;
820 struct stoken tmp_token;
821 CHECK_TYPEDEF (type);
822 if (TYPE_CODE (type) != TYPE_CODE_STRUCT
823 && TYPE_CODE (type) != TYPE_CODE_UNION
824 && TYPE_CODE (type) != TYPE_CODE_NAMESPACE)
825 error (_("`%s' is not defined as an aggregate type."),
826 TYPE_NAME (type));
827
828 tmp_token.ptr = (char*) alloca ($4.length + 2);
829 tmp_token.length = $4.length + 1;
830 tmp_token.ptr[0] = '~';
831 memcpy (tmp_token.ptr+1, $4.ptr, $4.length);
832 tmp_token.ptr[tmp_token.length] = 0;
833
834 /* Check for valid destructor name. */
835 destructor_name_p (tmp_token.ptr, $1.type);
836 write_exp_elt_opcode (OP_SCOPE);
837 write_exp_elt_type (type);
838 write_exp_string (tmp_token);
839 write_exp_elt_opcode (OP_SCOPE);
840 }
841 | TYPENAME COLONCOLON name COLONCOLON name
842 {
843 char *copy = copy_name ($3);
844 error (_("No type \"%s\" within class "
845 "or namespace \"%s\"."),
846 copy, TYPE_NAME ($1.type));
847 }
848 ;
849
850 variable: qualified_name
851 | COLONCOLON name_not_typename
852 {
853 char *name = copy_name ($2.stoken);
854 struct symbol *sym;
855 struct minimal_symbol *msymbol;
856
857 sym =
858 lookup_symbol (name, (const struct block *) NULL,
859 VAR_DOMAIN, (int *) NULL);
860 if (sym)
861 {
862 write_exp_elt_opcode (OP_VAR_VALUE);
863 write_exp_elt_block (NULL);
864 write_exp_elt_sym (sym);
865 write_exp_elt_opcode (OP_VAR_VALUE);
866 break;
867 }
868
869 msymbol = lookup_minimal_symbol (name, NULL, NULL);
870 if (msymbol != NULL)
871 write_exp_msymbol (msymbol);
872 else if (!have_full_symbols () && !have_partial_symbols ())
873 error (_("No symbol table is loaded. Use the \"file\" command."));
874 else
875 error (_("No symbol \"%s\" in current context."), name);
876 }
877 ;
878
879 variable: name_not_typename
880 { struct symbol *sym = $1.sym;
881
882 if (sym)
883 {
884 if (symbol_read_needs_frame (sym))
885 {
886 if (innermost_block == 0
887 || contained_in (block_found,
888 innermost_block))
889 innermost_block = block_found;
890 }
891
892 write_exp_elt_opcode (OP_VAR_VALUE);
893 /* We want to use the selected frame, not
894 another more inner frame which happens to
895 be in the same block. */
896 write_exp_elt_block (NULL);
897 write_exp_elt_sym (sym);
898 write_exp_elt_opcode (OP_VAR_VALUE);
899 }
900 else if ($1.is_a_field_of_this)
901 {
902 /* C++: it hangs off of `this'. Must
903 not inadvertently convert from a method call
904 to data ref. */
905 if (innermost_block == 0
906 || contained_in (block_found,
907 innermost_block))
908 innermost_block = block_found;
909 write_exp_elt_opcode (OP_THIS);
910 write_exp_elt_opcode (OP_THIS);
911 write_exp_elt_opcode (STRUCTOP_PTR);
912 write_exp_string ($1.stoken);
913 write_exp_elt_opcode (STRUCTOP_PTR);
914 }
915 else
916 {
917 struct minimal_symbol *msymbol;
918 char *arg = copy_name ($1.stoken);
919
920 msymbol =
921 lookup_minimal_symbol (arg, NULL, NULL);
922 if (msymbol != NULL)
923 write_exp_msymbol (msymbol);
924 else if (!have_full_symbols () && !have_partial_symbols ())
925 error (_("No symbol table is loaded. Use the \"file\" command."));
926 else
927 error (_("No symbol \"%s\" in current context."),
928 copy_name ($1.stoken));
929 }
930 }
931 ;
932
933 space_identifier : '@' NAME
934 { push_type_address_space (copy_name ($2.stoken));
935 push_type (tp_space_identifier);
936 }
937 ;
938
939 const_or_volatile: const_or_volatile_noopt
940 |
941 ;
942
943 cv_with_space_id : const_or_volatile space_identifier const_or_volatile
944 ;
945
946 const_or_volatile_or_space_identifier_noopt: cv_with_space_id
947 | const_or_volatile_noopt
948 ;
949
950 const_or_volatile_or_space_identifier:
951 const_or_volatile_or_space_identifier_noopt
952 |
953 ;
954
955 abs_decl: '*'
956 { push_type (tp_pointer); $$ = 0; }
957 | '*' abs_decl
958 { push_type (tp_pointer); $$ = $2; }
959 | '&'
960 { push_type (tp_reference); $$ = 0; }
961 | '&' abs_decl
962 { push_type (tp_reference); $$ = $2; }
963 | direct_abs_decl
964 ;
965
966 direct_abs_decl: '(' abs_decl ')'
967 { $$ = $2; }
968 | direct_abs_decl array_mod
969 {
970 push_type_int ($2);
971 push_type (tp_array);
972 }
973 | array_mod
974 {
975 push_type_int ($1);
976 push_type (tp_array);
977 $$ = 0;
978 }
979
980 | direct_abs_decl func_mod
981 { push_type (tp_function); }
982 | func_mod
983 { push_type (tp_function); }
984 ;
985
986 array_mod: '[' ']'
987 { $$ = -1; }
988 | '[' INT ']'
989 { $$ = $2.val; }
990 ;
991
992 func_mod: '(' ')'
993 { $$ = 0; }
994 | '(' nonempty_typelist ')'
995 { free ($2); $$ = 0; }
996 ;
997
998 /* We used to try to recognize pointer to member types here, but
999 that didn't work (shift/reduce conflicts meant that these rules never
1000 got executed). The problem is that
1001 int (foo::bar::baz::bizzle)
1002 is a function type but
1003 int (foo::bar::baz::bizzle::*)
1004 is a pointer to member type. Stroustrup loses again! */
1005
1006 type : ptype
1007 ;
1008
1009 typebase /* Implements (approximately): (type-qualifier)* type-specifier */
1010 : TYPENAME
1011 { $$ = $1.type; }
1012 | INT_KEYWORD
1013 { $$ = lookup_signed_typename (parse_language,
1014 parse_gdbarch,
1015 "int"); }
1016 | LONG
1017 { $$ = lookup_signed_typename (parse_language,
1018 parse_gdbarch,
1019 "long"); }
1020 | SHORT
1021 { $$ = lookup_signed_typename (parse_language,
1022 parse_gdbarch,
1023 "short"); }
1024 | LONG INT_KEYWORD
1025 { $$ = lookup_signed_typename (parse_language,
1026 parse_gdbarch,
1027 "long"); }
1028 | LONG SIGNED_KEYWORD INT_KEYWORD
1029 { $$ = lookup_signed_typename (parse_language,
1030 parse_gdbarch,
1031 "long"); }
1032 | LONG SIGNED_KEYWORD
1033 { $$ = lookup_signed_typename (parse_language,
1034 parse_gdbarch,
1035 "long"); }
1036 | SIGNED_KEYWORD LONG INT_KEYWORD
1037 { $$ = lookup_signed_typename (parse_language,
1038 parse_gdbarch,
1039 "long"); }
1040 | UNSIGNED LONG INT_KEYWORD
1041 { $$ = lookup_unsigned_typename (parse_language,
1042 parse_gdbarch,
1043 "long"); }
1044 | LONG UNSIGNED INT_KEYWORD
1045 { $$ = lookup_unsigned_typename (parse_language,
1046 parse_gdbarch,
1047 "long"); }
1048 | LONG UNSIGNED
1049 { $$ = lookup_unsigned_typename (parse_language,
1050 parse_gdbarch,
1051 "long"); }
1052 | LONG LONG
1053 { $$ = lookup_signed_typename (parse_language,
1054 parse_gdbarch,
1055 "long long"); }
1056 | LONG LONG INT_KEYWORD
1057 { $$ = lookup_signed_typename (parse_language,
1058 parse_gdbarch,
1059 "long long"); }
1060 | LONG LONG SIGNED_KEYWORD INT_KEYWORD
1061 { $$ = lookup_signed_typename (parse_language,
1062 parse_gdbarch,
1063 "long long"); }
1064 | LONG LONG SIGNED_KEYWORD
1065 { $$ = lookup_signed_typename (parse_language,
1066 parse_gdbarch,
1067 "long long"); }
1068 | SIGNED_KEYWORD LONG LONG
1069 { $$ = lookup_signed_typename (parse_language,
1070 parse_gdbarch,
1071 "long long"); }
1072 | SIGNED_KEYWORD LONG LONG INT_KEYWORD
1073 { $$ = lookup_signed_typename (parse_language,
1074 parse_gdbarch,
1075 "long long"); }
1076 | UNSIGNED LONG LONG
1077 { $$ = lookup_unsigned_typename (parse_language,
1078 parse_gdbarch,
1079 "long long"); }
1080 | UNSIGNED LONG LONG INT_KEYWORD
1081 { $$ = lookup_unsigned_typename (parse_language,
1082 parse_gdbarch,
1083 "long long"); }
1084 | LONG LONG UNSIGNED
1085 { $$ = lookup_unsigned_typename (parse_language,
1086 parse_gdbarch,
1087 "long long"); }
1088 | LONG LONG UNSIGNED INT_KEYWORD
1089 { $$ = lookup_unsigned_typename (parse_language,
1090 parse_gdbarch,
1091 "long long"); }
1092 | SHORT INT_KEYWORD
1093 { $$ = lookup_signed_typename (parse_language,
1094 parse_gdbarch,
1095 "short"); }
1096 | SHORT SIGNED_KEYWORD INT_KEYWORD
1097 { $$ = lookup_signed_typename (parse_language,
1098 parse_gdbarch,
1099 "short"); }
1100 | SHORT SIGNED_KEYWORD
1101 { $$ = lookup_signed_typename (parse_language,
1102 parse_gdbarch,
1103 "short"); }
1104 | UNSIGNED SHORT INT_KEYWORD
1105 { $$ = lookup_unsigned_typename (parse_language,
1106 parse_gdbarch,
1107 "short"); }
1108 | SHORT UNSIGNED
1109 { $$ = lookup_unsigned_typename (parse_language,
1110 parse_gdbarch,
1111 "short"); }
1112 | SHORT UNSIGNED INT_KEYWORD
1113 { $$ = lookup_unsigned_typename (parse_language,
1114 parse_gdbarch,
1115 "short"); }
1116 | DOUBLE_KEYWORD
1117 { $$ = lookup_typename (parse_language, parse_gdbarch,
1118 "double", (struct block *) NULL,
1119 0); }
1120 | LONG DOUBLE_KEYWORD
1121 { $$ = lookup_typename (parse_language, parse_gdbarch,
1122 "long double",
1123 (struct block *) NULL, 0); }
1124 | STRUCT name
1125 { $$ = lookup_struct (copy_name ($2),
1126 expression_context_block); }
1127 | CLASS name
1128 { $$ = lookup_struct (copy_name ($2),
1129 expression_context_block); }
1130 | UNION name
1131 { $$ = lookup_union (copy_name ($2),
1132 expression_context_block); }
1133 | ENUM name
1134 { $$ = lookup_enum (copy_name ($2),
1135 expression_context_block); }
1136 | UNSIGNED typename
1137 { $$ = lookup_unsigned_typename (parse_language,
1138 parse_gdbarch,
1139 TYPE_NAME($2.type)); }
1140 | UNSIGNED
1141 { $$ = lookup_unsigned_typename (parse_language,
1142 parse_gdbarch,
1143 "int"); }
1144 | SIGNED_KEYWORD typename
1145 { $$ = lookup_signed_typename (parse_language,
1146 parse_gdbarch,
1147 TYPE_NAME($2.type)); }
1148 | SIGNED_KEYWORD
1149 { $$ = lookup_signed_typename (parse_language,
1150 parse_gdbarch,
1151 "int"); }
1152 /* It appears that this rule for templates is never
1153 reduced; template recognition happens by lookahead
1154 in the token processing code in yylex. */
1155 | TEMPLATE name '<' type '>'
1156 { $$ = lookup_template_type(copy_name($2), $4,
1157 expression_context_block);
1158 }
1159 | const_or_volatile_or_space_identifier_noopt typebase
1160 { $$ = follow_types ($2); }
1161 | typebase const_or_volatile_or_space_identifier_noopt
1162 { $$ = follow_types ($1); }
1163 ;
1164
1165 typename: TYPENAME
1166 | INT_KEYWORD
1167 {
1168 $$.stoken.ptr = "int";
1169 $$.stoken.length = 3;
1170 $$.type = lookup_signed_typename (parse_language,
1171 parse_gdbarch,
1172 "int");
1173 }
1174 | LONG
1175 {
1176 $$.stoken.ptr = "long";
1177 $$.stoken.length = 4;
1178 $$.type = lookup_signed_typename (parse_language,
1179 parse_gdbarch,
1180 "long");
1181 }
1182 | SHORT
1183 {
1184 $$.stoken.ptr = "short";
1185 $$.stoken.length = 5;
1186 $$.type = lookup_signed_typename (parse_language,
1187 parse_gdbarch,
1188 "short");
1189 }
1190 ;
1191
1192 nonempty_typelist
1193 : type
1194 { $$ = (struct type **) malloc (sizeof (struct type *) * 2);
1195 $<ivec>$[0] = 1; /* Number of types in vector */
1196 $$[1] = $1;
1197 }
1198 | nonempty_typelist ',' type
1199 { int len = sizeof (struct type *) * (++($<ivec>1[0]) + 1);
1200 $$ = (struct type **) realloc ((char *) $1, len);
1201 $$[$<ivec>$[0]] = $3;
1202 }
1203 ;
1204
1205 ptype : typebase
1206 | ptype const_or_volatile_or_space_identifier abs_decl const_or_volatile_or_space_identifier
1207 { $$ = follow_types ($1); }
1208 ;
1209
1210 const_and_volatile: CONST_KEYWORD VOLATILE_KEYWORD
1211 | VOLATILE_KEYWORD CONST_KEYWORD
1212 ;
1213
1214 const_or_volatile_noopt: const_and_volatile
1215 { push_type (tp_const);
1216 push_type (tp_volatile);
1217 }
1218 | CONST_KEYWORD
1219 { push_type (tp_const); }
1220 | VOLATILE_KEYWORD
1221 { push_type (tp_volatile); }
1222 ;
1223
1224 operator: OPERATOR NEW
1225 { $$ = operator_stoken (" new"); }
1226 | OPERATOR DELETE
1227 { $$ = operator_stoken (" delete "); }
1228 | OPERATOR NEW '[' ']'
1229 { $$ = operator_stoken (" new[]"); }
1230 | OPERATOR DELETE '[' ']'
1231 { $$ = operator_stoken (" delete[] "); }
1232 | OPERATOR '+'
1233 { $$ = operator_stoken ("+"); }
1234 | OPERATOR '-'
1235 { $$ = operator_stoken ("-"); }
1236 | OPERATOR '*'
1237 { $$ = operator_stoken ("*"); }
1238 | OPERATOR '/'
1239 { $$ = operator_stoken ("/"); }
1240 | OPERATOR '%'
1241 { $$ = operator_stoken ("%"); }
1242 | OPERATOR '^'
1243 { $$ = operator_stoken ("^"); }
1244 | OPERATOR '&'
1245 { $$ = operator_stoken ("&"); }
1246 | OPERATOR '|'
1247 { $$ = operator_stoken ("|"); }
1248 | OPERATOR '~'
1249 { $$ = operator_stoken ("~"); }
1250 | OPERATOR '!'
1251 { $$ = operator_stoken ("!"); }
1252 | OPERATOR '='
1253 { $$ = operator_stoken ("="); }
1254 | OPERATOR '<'
1255 { $$ = operator_stoken ("<"); }
1256 | OPERATOR '>'
1257 { $$ = operator_stoken (">"); }
1258 | OPERATOR ASSIGN_MODIFY
1259 { const char *op = "unknown";
1260 switch ($2)
1261 {
1262 case BINOP_RSH:
1263 op = ">>=";
1264 break;
1265 case BINOP_LSH:
1266 op = "<<=";
1267 break;
1268 case BINOP_ADD:
1269 op = "+=";
1270 break;
1271 case BINOP_SUB:
1272 op = "-=";
1273 break;
1274 case BINOP_MUL:
1275 op = "*=";
1276 break;
1277 case BINOP_DIV:
1278 op = "/=";
1279 break;
1280 case BINOP_REM:
1281 op = "%=";
1282 break;
1283 case BINOP_BITWISE_IOR:
1284 op = "|=";
1285 break;
1286 case BINOP_BITWISE_AND:
1287 op = "&=";
1288 break;
1289 case BINOP_BITWISE_XOR:
1290 op = "^=";
1291 break;
1292 default:
1293 break;
1294 }
1295
1296 $$ = operator_stoken (op);
1297 }
1298 | OPERATOR LSH
1299 { $$ = operator_stoken ("<<"); }
1300 | OPERATOR RSH
1301 { $$ = operator_stoken (">>"); }
1302 | OPERATOR EQUAL
1303 { $$ = operator_stoken ("=="); }
1304 | OPERATOR NOTEQUAL
1305 { $$ = operator_stoken ("!="); }
1306 | OPERATOR LEQ
1307 { $$ = operator_stoken ("<="); }
1308 | OPERATOR GEQ
1309 { $$ = operator_stoken (">="); }
1310 | OPERATOR ANDAND
1311 { $$ = operator_stoken ("&&"); }
1312 | OPERATOR OROR
1313 { $$ = operator_stoken ("||"); }
1314 | OPERATOR INCREMENT
1315 { $$ = operator_stoken ("++"); }
1316 | OPERATOR DECREMENT
1317 { $$ = operator_stoken ("--"); }
1318 | OPERATOR ','
1319 { $$ = operator_stoken (","); }
1320 | OPERATOR ARROW_STAR
1321 { $$ = operator_stoken ("->*"); }
1322 | OPERATOR ARROW
1323 { $$ = operator_stoken ("->"); }
1324 | OPERATOR '(' ')'
1325 { $$ = operator_stoken ("()"); }
1326 | OPERATOR '[' ']'
1327 { $$ = operator_stoken ("[]"); }
1328 | OPERATOR ptype
1329 { char *name;
1330 long length;
1331 struct ui_file *buf = mem_fileopen ();
1332
1333 c_print_type ($2, NULL, buf, -1, 0);
1334 name = ui_file_xstrdup (buf, &length);
1335 ui_file_delete (buf);
1336 $$ = operator_stoken (name);
1337 free (name);
1338 }
1339 ;
1340
1341
1342
1343 name : NAME { $$ = $1.stoken; }
1344 | BLOCKNAME { $$ = $1.stoken; }
1345 | TYPENAME { $$ = $1.stoken; }
1346 | NAME_OR_INT { $$ = $1.stoken; }
1347 | UNKNOWN_CPP_NAME { $$ = $1.stoken; }
1348 | operator { $$ = $1; }
1349 ;
1350
1351 name_not_typename : NAME
1352 | BLOCKNAME
1353 /* These would be useful if name_not_typename was useful, but it is just
1354 a fake for "variable", so these cause reduce/reduce conflicts because
1355 the parser can't tell whether NAME_OR_INT is a name_not_typename (=variable,
1356 =exp) or just an exp. If name_not_typename was ever used in an lvalue
1357 context where only a name could occur, this might be useful.
1358 | NAME_OR_INT
1359 */
1360 | operator
1361 {
1362 $$.stoken = $1;
1363 $$.sym = lookup_symbol ($1.ptr,
1364 expression_context_block,
1365 VAR_DOMAIN,
1366 &$$.is_a_field_of_this);
1367 }
1368 | UNKNOWN_CPP_NAME
1369 ;
1370
1371 %%
1372
1373 /* Returns a stoken of the operator name given by OP (which does not
1374 include the string "operator"). */
1375 static struct stoken
1376 operator_stoken (const char *op)
1377 {
1378 static const char *operator_string = "operator";
1379 struct stoken st = { NULL, 0 };
1380 st.length = strlen (operator_string) + strlen (op);
1381 st.ptr = malloc (st.length + 1);
1382 strcpy (st.ptr, operator_string);
1383 strcat (st.ptr, op);
1384
1385 /* The toplevel (c_parse) will free the memory allocated here. */
1386 make_cleanup (free, st.ptr);
1387 return st;
1388 };
1389
1390 /* Take care of parsing a number (anything that starts with a digit).
1391 Set yylval and return the token type; update lexptr.
1392 LEN is the number of characters in it. */
1393
1394 /*** Needs some error checking for the float case ***/
1395
1396 static int
1397 parse_number (char *p, int len, int parsed_float, YYSTYPE *putithere)
1398 {
1399 /* FIXME: Shouldn't these be unsigned? We don't deal with negative values
1400 here, and we do kind of silly things like cast to unsigned. */
1401 LONGEST n = 0;
1402 LONGEST prevn = 0;
1403 ULONGEST un;
1404
1405 int i = 0;
1406 int c;
1407 int base = input_radix;
1408 int unsigned_p = 0;
1409
1410 /* Number of "L" suffixes encountered. */
1411 int long_p = 0;
1412
1413 /* We have found a "L" or "U" suffix. */
1414 int found_suffix = 0;
1415
1416 ULONGEST high_bit;
1417 struct type *signed_type;
1418 struct type *unsigned_type;
1419
1420 if (parsed_float)
1421 {
1422 /* If it ends at "df", "dd" or "dl", take it as type of decimal floating
1423 point. Return DECFLOAT. */
1424
1425 if (len >= 2 && p[len - 2] == 'd' && p[len - 1] == 'f')
1426 {
1427 p[len - 2] = '\0';
1428 putithere->typed_val_decfloat.type
1429 = parse_type->builtin_decfloat;
1430 decimal_from_string (putithere->typed_val_decfloat.val, 4,
1431 gdbarch_byte_order (parse_gdbarch), p);
1432 p[len - 2] = 'd';
1433 return DECFLOAT;
1434 }
1435
1436 if (len >= 2 && p[len - 2] == 'd' && p[len - 1] == 'd')
1437 {
1438 p[len - 2] = '\0';
1439 putithere->typed_val_decfloat.type
1440 = parse_type->builtin_decdouble;
1441 decimal_from_string (putithere->typed_val_decfloat.val, 8,
1442 gdbarch_byte_order (parse_gdbarch), p);
1443 p[len - 2] = 'd';
1444 return DECFLOAT;
1445 }
1446
1447 if (len >= 2 && p[len - 2] == 'd' && p[len - 1] == 'l')
1448 {
1449 p[len - 2] = '\0';
1450 putithere->typed_val_decfloat.type
1451 = parse_type->builtin_declong;
1452 decimal_from_string (putithere->typed_val_decfloat.val, 16,
1453 gdbarch_byte_order (parse_gdbarch), p);
1454 p[len - 2] = 'd';
1455 return DECFLOAT;
1456 }
1457
1458 if (! parse_c_float (parse_gdbarch, p, len,
1459 &putithere->typed_val_float.dval,
1460 &putithere->typed_val_float.type))
1461 return ERROR;
1462 return FLOAT;
1463 }
1464
1465 /* Handle base-switching prefixes 0x, 0t, 0d, 0 */
1466 if (p[0] == '0')
1467 switch (p[1])
1468 {
1469 case 'x':
1470 case 'X':
1471 if (len >= 3)
1472 {
1473 p += 2;
1474 base = 16;
1475 len -= 2;
1476 }
1477 break;
1478
1479 case 'b':
1480 case 'B':
1481 if (len >= 3)
1482 {
1483 p += 2;
1484 base = 2;
1485 len -= 2;
1486 }
1487 break;
1488
1489 case 't':
1490 case 'T':
1491 case 'd':
1492 case 'D':
1493 if (len >= 3)
1494 {
1495 p += 2;
1496 base = 10;
1497 len -= 2;
1498 }
1499 break;
1500
1501 default:
1502 base = 8;
1503 break;
1504 }
1505
1506 while (len-- > 0)
1507 {
1508 c = *p++;
1509 if (c >= 'A' && c <= 'Z')
1510 c += 'a' - 'A';
1511 if (c != 'l' && c != 'u')
1512 n *= base;
1513 if (c >= '0' && c <= '9')
1514 {
1515 if (found_suffix)
1516 return ERROR;
1517 n += i = c - '0';
1518 }
1519 else
1520 {
1521 if (base > 10 && c >= 'a' && c <= 'f')
1522 {
1523 if (found_suffix)
1524 return ERROR;
1525 n += i = c - 'a' + 10;
1526 }
1527 else if (c == 'l')
1528 {
1529 ++long_p;
1530 found_suffix = 1;
1531 }
1532 else if (c == 'u')
1533 {
1534 unsigned_p = 1;
1535 found_suffix = 1;
1536 }
1537 else
1538 return ERROR; /* Char not a digit */
1539 }
1540 if (i >= base)
1541 return ERROR; /* Invalid digit in this base */
1542
1543 /* Portably test for overflow (only works for nonzero values, so make
1544 a second check for zero). FIXME: Can't we just make n and prevn
1545 unsigned and avoid this? */
1546 if (c != 'l' && c != 'u' && (prevn >= n) && n != 0)
1547 unsigned_p = 1; /* Try something unsigned */
1548
1549 /* Portably test for unsigned overflow.
1550 FIXME: This check is wrong; for example it doesn't find overflow
1551 on 0x123456789 when LONGEST is 32 bits. */
1552 if (c != 'l' && c != 'u' && n != 0)
1553 {
1554 if ((unsigned_p && (ULONGEST) prevn >= (ULONGEST) n))
1555 error (_("Numeric constant too large."));
1556 }
1557 prevn = n;
1558 }
1559
1560 /* An integer constant is an int, a long, or a long long. An L
1561 suffix forces it to be long; an LL suffix forces it to be long
1562 long. If not forced to a larger size, it gets the first type of
1563 the above that it fits in. To figure out whether it fits, we
1564 shift it right and see whether anything remains. Note that we
1565 can't shift sizeof (LONGEST) * HOST_CHAR_BIT bits or more in one
1566 operation, because many compilers will warn about such a shift
1567 (which always produces a zero result). Sometimes gdbarch_int_bit
1568 or gdbarch_long_bit will be that big, sometimes not. To deal with
1569 the case where it is we just always shift the value more than
1570 once, with fewer bits each time. */
1571
1572 un = (ULONGEST)n >> 2;
1573 if (long_p == 0
1574 && (un >> (gdbarch_int_bit (parse_gdbarch) - 2)) == 0)
1575 {
1576 high_bit = ((ULONGEST)1) << (gdbarch_int_bit (parse_gdbarch) - 1);
1577
1578 /* A large decimal (not hex or octal) constant (between INT_MAX
1579 and UINT_MAX) is a long or unsigned long, according to ANSI,
1580 never an unsigned int, but this code treats it as unsigned
1581 int. This probably should be fixed. GCC gives a warning on
1582 such constants. */
1583
1584 unsigned_type = parse_type->builtin_unsigned_int;
1585 signed_type = parse_type->builtin_int;
1586 }
1587 else if (long_p <= 1
1588 && (un >> (gdbarch_long_bit (parse_gdbarch) - 2)) == 0)
1589 {
1590 high_bit = ((ULONGEST)1) << (gdbarch_long_bit (parse_gdbarch) - 1);
1591 unsigned_type = parse_type->builtin_unsigned_long;
1592 signed_type = parse_type->builtin_long;
1593 }
1594 else
1595 {
1596 int shift;
1597 if (sizeof (ULONGEST) * HOST_CHAR_BIT
1598 < gdbarch_long_long_bit (parse_gdbarch))
1599 /* A long long does not fit in a LONGEST. */
1600 shift = (sizeof (ULONGEST) * HOST_CHAR_BIT - 1);
1601 else
1602 shift = (gdbarch_long_long_bit (parse_gdbarch) - 1);
1603 high_bit = (ULONGEST) 1 << shift;
1604 unsigned_type = parse_type->builtin_unsigned_long_long;
1605 signed_type = parse_type->builtin_long_long;
1606 }
1607
1608 putithere->typed_val_int.val = n;
1609
1610 /* If the high bit of the worked out type is set then this number
1611 has to be unsigned. */
1612
1613 if (unsigned_p || (n & high_bit))
1614 {
1615 putithere->typed_val_int.type = unsigned_type;
1616 }
1617 else
1618 {
1619 putithere->typed_val_int.type = signed_type;
1620 }
1621
1622 return INT;
1623 }
1624
1625 /* Temporary obstack used for holding strings. */
1626 static struct obstack tempbuf;
1627 static int tempbuf_init;
1628
1629 /* Parse a C escape sequence. The initial backslash of the sequence
1630 is at (*PTR)[-1]. *PTR will be updated to point to just after the
1631 last character of the sequence. If OUTPUT is not NULL, the
1632 translated form of the escape sequence will be written there. If
1633 OUTPUT is NULL, no output is written and the call will only affect
1634 *PTR. If an escape sequence is expressed in target bytes, then the
1635 entire sequence will simply be copied to OUTPUT. Return 1 if any
1636 character was emitted, 0 otherwise. */
1637
1638 int
1639 c_parse_escape (char **ptr, struct obstack *output)
1640 {
1641 char *tokptr = *ptr;
1642 int result = 1;
1643
1644 /* Some escape sequences undergo character set conversion. Those we
1645 translate here. */
1646 switch (*tokptr)
1647 {
1648 /* Hex escapes do not undergo character set conversion, so keep
1649 the escape sequence for later. */
1650 case 'x':
1651 if (output)
1652 obstack_grow_str (output, "\\x");
1653 ++tokptr;
1654 if (!isxdigit (*tokptr))
1655 error (_("\\x escape without a following hex digit"));
1656 while (isxdigit (*tokptr))
1657 {
1658 if (output)
1659 obstack_1grow (output, *tokptr);
1660 ++tokptr;
1661 }
1662 break;
1663
1664 /* Octal escapes do not undergo character set conversion, so
1665 keep the escape sequence for later. */
1666 case '0':
1667 case '1':
1668 case '2':
1669 case '3':
1670 case '4':
1671 case '5':
1672 case '6':
1673 case '7':
1674 {
1675 int i;
1676 if (output)
1677 obstack_grow_str (output, "\\");
1678 for (i = 0;
1679 i < 3 && isdigit (*tokptr) && *tokptr != '8' && *tokptr != '9';
1680 ++i)
1681 {
1682 if (output)
1683 obstack_1grow (output, *tokptr);
1684 ++tokptr;
1685 }
1686 }
1687 break;
1688
1689 /* We handle UCNs later. We could handle them here, but that
1690 would mean a spurious error in the case where the UCN could
1691 be converted to the target charset but not the host
1692 charset. */
1693 case 'u':
1694 case 'U':
1695 {
1696 char c = *tokptr;
1697 int i, len = c == 'U' ? 8 : 4;
1698 if (output)
1699 {
1700 obstack_1grow (output, '\\');
1701 obstack_1grow (output, *tokptr);
1702 }
1703 ++tokptr;
1704 if (!isxdigit (*tokptr))
1705 error (_("\\%c escape without a following hex digit"), c);
1706 for (i = 0; i < len && isxdigit (*tokptr); ++i)
1707 {
1708 if (output)
1709 obstack_1grow (output, *tokptr);
1710 ++tokptr;
1711 }
1712 }
1713 break;
1714
1715 /* We must pass backslash through so that it does not
1716 cause quoting during the second expansion. */
1717 case '\\':
1718 if (output)
1719 obstack_grow_str (output, "\\\\");
1720 ++tokptr;
1721 break;
1722
1723 /* Escapes which undergo conversion. */
1724 case 'a':
1725 if (output)
1726 obstack_1grow (output, '\a');
1727 ++tokptr;
1728 break;
1729 case 'b':
1730 if (output)
1731 obstack_1grow (output, '\b');
1732 ++tokptr;
1733 break;
1734 case 'f':
1735 if (output)
1736 obstack_1grow (output, '\f');
1737 ++tokptr;
1738 break;
1739 case 'n':
1740 if (output)
1741 obstack_1grow (output, '\n');
1742 ++tokptr;
1743 break;
1744 case 'r':
1745 if (output)
1746 obstack_1grow (output, '\r');
1747 ++tokptr;
1748 break;
1749 case 't':
1750 if (output)
1751 obstack_1grow (output, '\t');
1752 ++tokptr;
1753 break;
1754 case 'v':
1755 if (output)
1756 obstack_1grow (output, '\v');
1757 ++tokptr;
1758 break;
1759
1760 /* GCC extension. */
1761 case 'e':
1762 if (output)
1763 obstack_1grow (output, HOST_ESCAPE_CHAR);
1764 ++tokptr;
1765 break;
1766
1767 /* Backslash-newline expands to nothing at all. */
1768 case '\n':
1769 ++tokptr;
1770 result = 0;
1771 break;
1772
1773 /* A few escapes just expand to the character itself. */
1774 case '\'':
1775 case '\"':
1776 case '?':
1777 /* GCC extensions. */
1778 case '(':
1779 case '{':
1780 case '[':
1781 case '%':
1782 /* Unrecognized escapes turn into the character itself. */
1783 default:
1784 if (output)
1785 obstack_1grow (output, *tokptr);
1786 ++tokptr;
1787 break;
1788 }
1789 *ptr = tokptr;
1790 return result;
1791 }
1792
1793 /* Parse a string or character literal from TOKPTR. The string or
1794 character may be wide or unicode. *OUTPTR is set to just after the
1795 end of the literal in the input string. The resulting token is
1796 stored in VALUE. This returns a token value, either STRING or
1797 CHAR, depending on what was parsed. *HOST_CHARS is set to the
1798 number of host characters in the literal. */
1799 static int
1800 parse_string_or_char (char *tokptr, char **outptr, struct typed_stoken *value,
1801 int *host_chars)
1802 {
1803 int quote;
1804 enum c_string_type type;
1805
1806 /* Build the gdb internal form of the input string in tempbuf. Note
1807 that the buffer is null byte terminated *only* for the
1808 convenience of debugging gdb itself and printing the buffer
1809 contents when the buffer contains no embedded nulls. Gdb does
1810 not depend upon the buffer being null byte terminated, it uses
1811 the length string instead. This allows gdb to handle C strings
1812 (as well as strings in other languages) with embedded null
1813 bytes */
1814
1815 if (!tempbuf_init)
1816 tempbuf_init = 1;
1817 else
1818 obstack_free (&tempbuf, NULL);
1819 obstack_init (&tempbuf);
1820
1821 /* Record the string type. */
1822 if (*tokptr == 'L')
1823 {
1824 type = C_WIDE_STRING;
1825 ++tokptr;
1826 }
1827 else if (*tokptr == 'u')
1828 {
1829 type = C_STRING_16;
1830 ++tokptr;
1831 }
1832 else if (*tokptr == 'U')
1833 {
1834 type = C_STRING_32;
1835 ++tokptr;
1836 }
1837 else
1838 type = C_STRING;
1839
1840 /* Skip the quote. */
1841 quote = *tokptr;
1842 if (quote == '\'')
1843 type |= C_CHAR;
1844 ++tokptr;
1845
1846 *host_chars = 0;
1847
1848 while (*tokptr)
1849 {
1850 char c = *tokptr;
1851 if (c == '\\')
1852 {
1853 ++tokptr;
1854 *host_chars += c_parse_escape (&tokptr, &tempbuf);
1855 }
1856 else if (c == quote)
1857 break;
1858 else
1859 {
1860 obstack_1grow (&tempbuf, c);
1861 ++tokptr;
1862 /* FIXME: this does the wrong thing with multi-byte host
1863 characters. We could use mbrlen here, but that would
1864 make "set host-charset" a bit less useful. */
1865 ++*host_chars;
1866 }
1867 }
1868
1869 if (*tokptr != quote)
1870 {
1871 if (quote == '"')
1872 error (_("Unterminated string in expression."));
1873 else
1874 error (_("Unmatched single quote."));
1875 }
1876 ++tokptr;
1877
1878 value->type = type;
1879 value->ptr = obstack_base (&tempbuf);
1880 value->length = obstack_object_size (&tempbuf);
1881
1882 *outptr = tokptr;
1883
1884 return quote == '"' ? STRING : CHAR;
1885 }
1886
1887 struct token
1888 {
1889 char *operator;
1890 int token;
1891 enum exp_opcode opcode;
1892 int cxx_only;
1893 };
1894
1895 static const struct token tokentab3[] =
1896 {
1897 {">>=", ASSIGN_MODIFY, BINOP_RSH, 0},
1898 {"<<=", ASSIGN_MODIFY, BINOP_LSH, 0},
1899 {"->*", ARROW_STAR, BINOP_END, 1}
1900 };
1901
1902 static const struct token tokentab2[] =
1903 {
1904 {"+=", ASSIGN_MODIFY, BINOP_ADD, 0},
1905 {"-=", ASSIGN_MODIFY, BINOP_SUB, 0},
1906 {"*=", ASSIGN_MODIFY, BINOP_MUL, 0},
1907 {"/=", ASSIGN_MODIFY, BINOP_DIV, 0},
1908 {"%=", ASSIGN_MODIFY, BINOP_REM, 0},
1909 {"|=", ASSIGN_MODIFY, BINOP_BITWISE_IOR, 0},
1910 {"&=", ASSIGN_MODIFY, BINOP_BITWISE_AND, 0},
1911 {"^=", ASSIGN_MODIFY, BINOP_BITWISE_XOR, 0},
1912 {"++", INCREMENT, BINOP_END, 0},
1913 {"--", DECREMENT, BINOP_END, 0},
1914 {"->", ARROW, BINOP_END, 0},
1915 {"&&", ANDAND, BINOP_END, 0},
1916 {"||", OROR, BINOP_END, 0},
1917 /* "::" is *not* only C++: gdb overrides its meaning in several
1918 different ways, e.g., 'filename'::func, function::variable. */
1919 {"::", COLONCOLON, BINOP_END, 0},
1920 {"<<", LSH, BINOP_END, 0},
1921 {">>", RSH, BINOP_END, 0},
1922 {"==", EQUAL, BINOP_END, 0},
1923 {"!=", NOTEQUAL, BINOP_END, 0},
1924 {"<=", LEQ, BINOP_END, 0},
1925 {">=", GEQ, BINOP_END, 0},
1926 {".*", DOT_STAR, BINOP_END, 1}
1927 };
1928
1929 /* Identifier-like tokens. */
1930 static const struct token ident_tokens[] =
1931 {
1932 {"unsigned", UNSIGNED, OP_NULL, 0},
1933 {"template", TEMPLATE, OP_NULL, 1},
1934 {"volatile", VOLATILE_KEYWORD, OP_NULL, 0},
1935 {"struct", STRUCT, OP_NULL, 0},
1936 {"signed", SIGNED_KEYWORD, OP_NULL, 0},
1937 {"sizeof", SIZEOF, OP_NULL, 0},
1938 {"double", DOUBLE_KEYWORD, OP_NULL, 0},
1939 {"false", FALSEKEYWORD, OP_NULL, 1},
1940 {"class", CLASS, OP_NULL, 1},
1941 {"union", UNION, OP_NULL, 0},
1942 {"short", SHORT, OP_NULL, 0},
1943 {"const", CONST_KEYWORD, OP_NULL, 0},
1944 {"enum", ENUM, OP_NULL, 0},
1945 {"long", LONG, OP_NULL, 0},
1946 {"true", TRUEKEYWORD, OP_NULL, 1},
1947 {"int", INT_KEYWORD, OP_NULL, 0},
1948 {"new", NEW, OP_NULL, 1},
1949 {"delete", DELETE, OP_NULL, 1},
1950 {"operator", OPERATOR, OP_NULL, 1},
1951
1952 {"and", ANDAND, BINOP_END, 1},
1953 {"and_eq", ASSIGN_MODIFY, BINOP_BITWISE_AND, 1},
1954 {"bitand", '&', OP_NULL, 1},
1955 {"bitor", '|', OP_NULL, 1},
1956 {"compl", '~', OP_NULL, 1},
1957 {"not", '!', OP_NULL, 1},
1958 {"not_eq", NOTEQUAL, BINOP_END, 1},
1959 {"or", OROR, BINOP_END, 1},
1960 {"or_eq", ASSIGN_MODIFY, BINOP_BITWISE_IOR, 1},
1961 {"xor", '^', OP_NULL, 1},
1962 {"xor_eq", ASSIGN_MODIFY, BINOP_BITWISE_XOR, 1},
1963
1964 {"const_cast", CONST_CAST, OP_NULL, 1 },
1965 {"dynamic_cast", DYNAMIC_CAST, OP_NULL, 1 },
1966 {"static_cast", STATIC_CAST, OP_NULL, 1 },
1967 {"reinterpret_cast", REINTERPRET_CAST, OP_NULL, 1 }
1968 };
1969
1970 /* When we find that lexptr (the global var defined in parse.c) is
1971 pointing at a macro invocation, we expand the invocation, and call
1972 scan_macro_expansion to save the old lexptr here and point lexptr
1973 into the expanded text. When we reach the end of that, we call
1974 end_macro_expansion to pop back to the value we saved here. The
1975 macro expansion code promises to return only fully-expanded text,
1976 so we don't need to "push" more than one level.
1977
1978 This is disgusting, of course. It would be cleaner to do all macro
1979 expansion beforehand, and then hand that to lexptr. But we don't
1980 really know where the expression ends. Remember, in a command like
1981
1982 (gdb) break *ADDRESS if CONDITION
1983
1984 we evaluate ADDRESS in the scope of the current frame, but we
1985 evaluate CONDITION in the scope of the breakpoint's location. So
1986 it's simply wrong to try to macro-expand the whole thing at once. */
1987 static char *macro_original_text;
1988
1989 /* We save all intermediate macro expansions on this obstack for the
1990 duration of a single parse. The expansion text may sometimes have
1991 to live past the end of the expansion, due to yacc lookahead.
1992 Rather than try to be clever about saving the data for a single
1993 token, we simply keep it all and delete it after parsing has
1994 completed. */
1995 static struct obstack expansion_obstack;
1996
1997 static void
1998 scan_macro_expansion (char *expansion)
1999 {
2000 char *copy;
2001
2002 /* We'd better not be trying to push the stack twice. */
2003 gdb_assert (! macro_original_text);
2004
2005 /* Copy to the obstack, and then free the intermediate
2006 expansion. */
2007 copy = obstack_copy0 (&expansion_obstack, expansion, strlen (expansion));
2008 xfree (expansion);
2009
2010 /* Save the old lexptr value, so we can return to it when we're done
2011 parsing the expanded text. */
2012 macro_original_text = lexptr;
2013 lexptr = copy;
2014 }
2015
2016
2017 static int
2018 scanning_macro_expansion (void)
2019 {
2020 return macro_original_text != 0;
2021 }
2022
2023
2024 static void
2025 finished_macro_expansion (void)
2026 {
2027 /* There'd better be something to pop back to. */
2028 gdb_assert (macro_original_text);
2029
2030 /* Pop back to the original text. */
2031 lexptr = macro_original_text;
2032 macro_original_text = 0;
2033 }
2034
2035
2036 static void
2037 scan_macro_cleanup (void *dummy)
2038 {
2039 if (macro_original_text)
2040 finished_macro_expansion ();
2041
2042 obstack_free (&expansion_obstack, NULL);
2043 }
2044
2045 /* Return true iff the token represents a C++ cast operator. */
2046
2047 static int
2048 is_cast_operator (const char *token, int len)
2049 {
2050 return (! strncmp (token, "dynamic_cast", len)
2051 || ! strncmp (token, "static_cast", len)
2052 || ! strncmp (token, "reinterpret_cast", len)
2053 || ! strncmp (token, "const_cast", len));
2054 }
2055
2056 /* The scope used for macro expansion. */
2057 static struct macro_scope *expression_macro_scope;
2058
2059 /* This is set if a NAME token appeared at the very end of the input
2060 string, with no whitespace separating the name from the EOF. This
2061 is used only when parsing to do field name completion. */
2062 static int saw_name_at_eof;
2063
2064 /* This is set if the previously-returned token was a structure
2065 operator -- either '.' or ARROW. This is used only when parsing to
2066 do field name completion. */
2067 static int last_was_structop;
2068
2069 /* Read one token, getting characters through lexptr. */
2070
2071 static int
2072 lex_one_token (void)
2073 {
2074 int c;
2075 int namelen;
2076 unsigned int i;
2077 char *tokstart;
2078 int saw_structop = last_was_structop;
2079 char *copy;
2080
2081 last_was_structop = 0;
2082
2083 retry:
2084
2085 /* Check if this is a macro invocation that we need to expand. */
2086 if (! scanning_macro_expansion ())
2087 {
2088 char *expanded = macro_expand_next (&lexptr,
2089 standard_macro_lookup,
2090 expression_macro_scope);
2091
2092 if (expanded)
2093 scan_macro_expansion (expanded);
2094 }
2095
2096 prev_lexptr = lexptr;
2097
2098 tokstart = lexptr;
2099 /* See if it is a special token of length 3. */
2100 for (i = 0; i < sizeof tokentab3 / sizeof tokentab3[0]; i++)
2101 if (strncmp (tokstart, tokentab3[i].operator, 3) == 0)
2102 {
2103 if (tokentab3[i].cxx_only
2104 && parse_language->la_language != language_cplus)
2105 break;
2106
2107 lexptr += 3;
2108 yylval.opcode = tokentab3[i].opcode;
2109 return tokentab3[i].token;
2110 }
2111
2112 /* See if it is a special token of length 2. */
2113 for (i = 0; i < sizeof tokentab2 / sizeof tokentab2[0]; i++)
2114 if (strncmp (tokstart, tokentab2[i].operator, 2) == 0)
2115 {
2116 if (tokentab2[i].cxx_only
2117 && parse_language->la_language != language_cplus)
2118 break;
2119
2120 lexptr += 2;
2121 yylval.opcode = tokentab2[i].opcode;
2122 if (in_parse_field && tokentab2[i].token == ARROW)
2123 last_was_structop = 1;
2124 return tokentab2[i].token;
2125 }
2126
2127 switch (c = *tokstart)
2128 {
2129 case 0:
2130 /* If we were just scanning the result of a macro expansion,
2131 then we need to resume scanning the original text.
2132 If we're parsing for field name completion, and the previous
2133 token allows such completion, return a COMPLETE token.
2134 Otherwise, we were already scanning the original text, and
2135 we're really done. */
2136 if (scanning_macro_expansion ())
2137 {
2138 finished_macro_expansion ();
2139 goto retry;
2140 }
2141 else if (saw_name_at_eof)
2142 {
2143 saw_name_at_eof = 0;
2144 return COMPLETE;
2145 }
2146 else if (saw_structop)
2147 return COMPLETE;
2148 else
2149 return 0;
2150
2151 case ' ':
2152 case '\t':
2153 case '\n':
2154 lexptr++;
2155 goto retry;
2156
2157 case '[':
2158 case '(':
2159 paren_depth++;
2160 lexptr++;
2161 return c;
2162
2163 case ']':
2164 case ')':
2165 if (paren_depth == 0)
2166 return 0;
2167 paren_depth--;
2168 lexptr++;
2169 return c;
2170
2171 case ',':
2172 if (comma_terminates
2173 && paren_depth == 0
2174 && ! scanning_macro_expansion ())
2175 return 0;
2176 lexptr++;
2177 return c;
2178
2179 case '.':
2180 /* Might be a floating point number. */
2181 if (lexptr[1] < '0' || lexptr[1] > '9')
2182 {
2183 if (in_parse_field)
2184 last_was_structop = 1;
2185 goto symbol; /* Nope, must be a symbol. */
2186 }
2187 /* FALL THRU into number case. */
2188
2189 case '0':
2190 case '1':
2191 case '2':
2192 case '3':
2193 case '4':
2194 case '5':
2195 case '6':
2196 case '7':
2197 case '8':
2198 case '9':
2199 {
2200 /* It's a number. */
2201 int got_dot = 0, got_e = 0, toktype;
2202 char *p = tokstart;
2203 int hex = input_radix > 10;
2204
2205 if (c == '0' && (p[1] == 'x' || p[1] == 'X'))
2206 {
2207 p += 2;
2208 hex = 1;
2209 }
2210 else if (c == '0' && (p[1]=='t' || p[1]=='T' || p[1]=='d' || p[1]=='D'))
2211 {
2212 p += 2;
2213 hex = 0;
2214 }
2215
2216 for (;; ++p)
2217 {
2218 /* This test includes !hex because 'e' is a valid hex digit
2219 and thus does not indicate a floating point number when
2220 the radix is hex. */
2221 if (!hex && !got_e && (*p == 'e' || *p == 'E'))
2222 got_dot = got_e = 1;
2223 /* This test does not include !hex, because a '.' always indicates
2224 a decimal floating point number regardless of the radix. */
2225 else if (!got_dot && *p == '.')
2226 got_dot = 1;
2227 else if (got_e && (p[-1] == 'e' || p[-1] == 'E')
2228 && (*p == '-' || *p == '+'))
2229 /* This is the sign of the exponent, not the end of the
2230 number. */
2231 continue;
2232 /* We will take any letters or digits. parse_number will
2233 complain if past the radix, or if L or U are not final. */
2234 else if ((*p < '0' || *p > '9')
2235 && ((*p < 'a' || *p > 'z')
2236 && (*p < 'A' || *p > 'Z')))
2237 break;
2238 }
2239 toktype = parse_number (tokstart, p - tokstart, got_dot|got_e, &yylval);
2240 if (toktype == ERROR)
2241 {
2242 char *err_copy = (char *) alloca (p - tokstart + 1);
2243
2244 memcpy (err_copy, tokstart, p - tokstart);
2245 err_copy[p - tokstart] = 0;
2246 error (_("Invalid number \"%s\"."), err_copy);
2247 }
2248 lexptr = p;
2249 return toktype;
2250 }
2251
2252 case '@':
2253 {
2254 char *p = &tokstart[1];
2255 size_t len = strlen ("entry");
2256
2257 while (isspace (*p))
2258 p++;
2259 if (strncmp (p, "entry", len) == 0 && !isalnum (p[len])
2260 && p[len] != '_')
2261 {
2262 lexptr = &p[len];
2263 return ENTRY;
2264 }
2265 }
2266 /* FALLTHRU */
2267 case '+':
2268 case '-':
2269 case '*':
2270 case '/':
2271 case '%':
2272 case '|':
2273 case '&':
2274 case '^':
2275 case '~':
2276 case '!':
2277 case '<':
2278 case '>':
2279 case '?':
2280 case ':':
2281 case '=':
2282 case '{':
2283 case '}':
2284 symbol:
2285 lexptr++;
2286 return c;
2287
2288 case 'L':
2289 case 'u':
2290 case 'U':
2291 if (tokstart[1] != '"' && tokstart[1] != '\'')
2292 break;
2293 /* Fall through. */
2294 case '\'':
2295 case '"':
2296 {
2297 int host_len;
2298 int result = parse_string_or_char (tokstart, &lexptr, &yylval.tsval,
2299 &host_len);
2300 if (result == CHAR)
2301 {
2302 if (host_len == 0)
2303 error (_("Empty character constant."));
2304 else if (host_len > 2 && c == '\'')
2305 {
2306 ++tokstart;
2307 namelen = lexptr - tokstart - 1;
2308 goto tryname;
2309 }
2310 else if (host_len > 1)
2311 error (_("Invalid character constant."));
2312 }
2313 return result;
2314 }
2315 }
2316
2317 if (!(c == '_' || c == '$'
2318 || (c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z')))
2319 /* We must have come across a bad character (e.g. ';'). */
2320 error (_("Invalid character '%c' in expression."), c);
2321
2322 /* It's a name. See how long it is. */
2323 namelen = 0;
2324 for (c = tokstart[namelen];
2325 (c == '_' || c == '$' || (c >= '0' && c <= '9')
2326 || (c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z') || c == '<');)
2327 {
2328 /* Template parameter lists are part of the name.
2329 FIXME: This mishandles `print $a<4&&$a>3'. */
2330
2331 if (c == '<')
2332 {
2333 if (! is_cast_operator (tokstart, namelen))
2334 {
2335 /* Scan ahead to get rest of the template specification. Note
2336 that we look ahead only when the '<' adjoins non-whitespace
2337 characters; for comparison expressions, e.g. "a < b > c",
2338 there must be spaces before the '<', etc. */
2339
2340 char * p = find_template_name_end (tokstart + namelen);
2341 if (p)
2342 namelen = p - tokstart;
2343 }
2344 break;
2345 }
2346 c = tokstart[++namelen];
2347 }
2348
2349 /* The token "if" terminates the expression and is NOT removed from
2350 the input stream. It doesn't count if it appears in the
2351 expansion of a macro. */
2352 if (namelen == 2
2353 && tokstart[0] == 'i'
2354 && tokstart[1] == 'f'
2355 && ! scanning_macro_expansion ())
2356 {
2357 return 0;
2358 }
2359
2360 /* For the same reason (breakpoint conditions), "thread N"
2361 terminates the expression. "thread" could be an identifier, but
2362 an identifier is never followed by a number without intervening
2363 punctuation. "task" is similar. Handle abbreviations of these,
2364 similarly to breakpoint.c:find_condition_and_thread. */
2365 if (namelen >= 1
2366 && (strncmp (tokstart, "thread", namelen) == 0
2367 || strncmp (tokstart, "task", namelen) == 0)
2368 && (tokstart[namelen] == ' ' || tokstart[namelen] == '\t')
2369 && ! scanning_macro_expansion ())
2370 {
2371 char *p = tokstart + namelen + 1;
2372 while (*p == ' ' || *p == '\t')
2373 p++;
2374 if (*p >= '0' && *p <= '9')
2375 return 0;
2376 }
2377
2378 lexptr += namelen;
2379
2380 tryname:
2381
2382 yylval.sval.ptr = tokstart;
2383 yylval.sval.length = namelen;
2384
2385 /* Catch specific keywords. */
2386 copy = copy_name (yylval.sval);
2387 for (i = 0; i < sizeof ident_tokens / sizeof ident_tokens[0]; i++)
2388 if (strcmp (copy, ident_tokens[i].operator) == 0)
2389 {
2390 if (ident_tokens[i].cxx_only
2391 && parse_language->la_language != language_cplus)
2392 break;
2393
2394 /* It is ok to always set this, even though we don't always
2395 strictly need to. */
2396 yylval.opcode = ident_tokens[i].opcode;
2397 return ident_tokens[i].token;
2398 }
2399
2400 if (*tokstart == '$')
2401 return VARIABLE;
2402
2403 if (in_parse_field && *lexptr == '\0')
2404 saw_name_at_eof = 1;
2405 return NAME;
2406 }
2407
2408 /* An object of this type is pushed on a FIFO by the "outer" lexer. */
2409 typedef struct
2410 {
2411 int token;
2412 YYSTYPE value;
2413 } token_and_value;
2414
2415 DEF_VEC_O (token_and_value);
2416
2417 /* A FIFO of tokens that have been read but not yet returned to the
2418 parser. */
2419 static VEC (token_and_value) *token_fifo;
2420
2421 /* Non-zero if the lexer should return tokens from the FIFO. */
2422 static int popping;
2423
2424 /* Temporary storage for c_lex; this holds symbol names as they are
2425 built up. */
2426 static struct obstack name_obstack;
2427
2428 /* Classify a NAME token. The contents of the token are in `yylval'.
2429 Updates yylval and returns the new token type. BLOCK is the block
2430 in which lookups start; this can be NULL to mean the global
2431 scope. */
2432 static int
2433 classify_name (struct block *block)
2434 {
2435 struct symbol *sym;
2436 char *copy;
2437 int is_a_field_of_this = 0;
2438
2439 copy = copy_name (yylval.sval);
2440
2441 sym = lookup_symbol (copy, block, VAR_DOMAIN,
2442 parse_language->la_language == language_cplus
2443 ? &is_a_field_of_this : (int *) NULL);
2444
2445 if (sym && SYMBOL_CLASS (sym) == LOC_BLOCK)
2446 {
2447 yylval.ssym.sym = sym;
2448 yylval.ssym.is_a_field_of_this = is_a_field_of_this;
2449 return BLOCKNAME;
2450 }
2451 else if (!sym)
2452 {
2453 /* See if it's a file name. */
2454 struct symtab *symtab;
2455
2456 symtab = lookup_symtab (copy);
2457 if (symtab)
2458 {
2459 yylval.bval = BLOCKVECTOR_BLOCK (BLOCKVECTOR (symtab), STATIC_BLOCK);
2460 return FILENAME;
2461 }
2462 }
2463
2464 if (sym && SYMBOL_CLASS (sym) == LOC_TYPEDEF)
2465 {
2466 yylval.tsym.type = SYMBOL_TYPE (sym);
2467 return TYPENAME;
2468 }
2469
2470 yylval.tsym.type
2471 = language_lookup_primitive_type_by_name (parse_language,
2472 parse_gdbarch, copy);
2473 if (yylval.tsym.type != NULL)
2474 return TYPENAME;
2475
2476 /* Input names that aren't symbols but ARE valid hex numbers, when
2477 the input radix permits them, can be names or numbers depending
2478 on the parse. Note we support radixes > 16 here. */
2479 if (!sym
2480 && ((copy[0] >= 'a' && copy[0] < 'a' + input_radix - 10)
2481 || (copy[0] >= 'A' && copy[0] < 'A' + input_radix - 10)))
2482 {
2483 YYSTYPE newlval; /* Its value is ignored. */
2484 int hextype = parse_number (copy, yylval.sval.length, 0, &newlval);
2485 if (hextype == INT)
2486 {
2487 yylval.ssym.sym = sym;
2488 yylval.ssym.is_a_field_of_this = is_a_field_of_this;
2489 return NAME_OR_INT;
2490 }
2491 }
2492
2493 /* Any other kind of symbol */
2494 yylval.ssym.sym = sym;
2495 yylval.ssym.is_a_field_of_this = is_a_field_of_this;
2496
2497 if (sym == NULL
2498 && parse_language->la_language == language_cplus
2499 && !is_a_field_of_this
2500 && !lookup_minimal_symbol (copy, NULL, NULL))
2501 return UNKNOWN_CPP_NAME;
2502
2503 return NAME;
2504 }
2505
2506 /* Like classify_name, but used by the inner loop of the lexer, when a
2507 name might have already been seen. FIRST_NAME is true if the token
2508 in `yylval' is the first component of a name, false otherwise. If
2509 this function returns NAME, it might not have updated `yylval'.
2510 This is ok because the caller only cares about TYPENAME. */
2511 static int
2512 classify_inner_name (struct block *block, int first_name)
2513 {
2514 struct type *type, *new_type;
2515 char *copy;
2516
2517 if (first_name)
2518 return classify_name (block);
2519
2520 type = check_typedef (yylval.tsym.type);
2521 if (TYPE_CODE (type) != TYPE_CODE_STRUCT
2522 && TYPE_CODE (type) != TYPE_CODE_UNION
2523 && TYPE_CODE (type) != TYPE_CODE_NAMESPACE)
2524 /* We know the caller won't expect us to update yylval. */
2525 return NAME;
2526
2527 copy = copy_name (yylval.tsym.stoken);
2528 new_type = cp_lookup_nested_type (yylval.tsym.type, copy, block);
2529
2530 if (new_type == NULL)
2531 /* We know the caller won't expect us to update yylval. */
2532 return NAME;
2533
2534 yylval.tsym.type = new_type;
2535 return TYPENAME;
2536 }
2537
2538 /* The outer level of a two-level lexer. This calls the inner lexer
2539 to return tokens. It then either returns these tokens, or
2540 aggregates them into a larger token. This lets us work around a
2541 problem in our parsing approach, where the parser could not
2542 distinguish between qualified names and qualified types at the
2543 right point.
2544
2545 This approach is still not ideal, because it mishandles template
2546 types. See the comment in lex_one_token for an example. However,
2547 this is still an improvement over the earlier approach, and will
2548 suffice until we move to better parsing technology. */
2549 static int
2550 yylex (void)
2551 {
2552 token_and_value current;
2553 int first_was_coloncolon, last_was_coloncolon, first_iter;
2554
2555 if (popping && !VEC_empty (token_and_value, token_fifo))
2556 {
2557 token_and_value tv = *VEC_index (token_and_value, token_fifo, 0);
2558 VEC_ordered_remove (token_and_value, token_fifo, 0);
2559 yylval = tv.value;
2560 return tv.token;
2561 }
2562 popping = 0;
2563
2564 current.token = lex_one_token ();
2565 if (current.token == NAME)
2566 current.token = classify_name (expression_context_block);
2567 if (parse_language->la_language != language_cplus
2568 || (current.token != TYPENAME && current.token != COLONCOLON))
2569 return current.token;
2570
2571 first_was_coloncolon = current.token == COLONCOLON;
2572 last_was_coloncolon = first_was_coloncolon;
2573 obstack_free (&name_obstack, obstack_base (&name_obstack));
2574 if (!last_was_coloncolon)
2575 obstack_grow (&name_obstack, yylval.sval.ptr, yylval.sval.length);
2576 current.value = yylval;
2577 first_iter = 1;
2578 while (1)
2579 {
2580 token_and_value next;
2581
2582 next.token = lex_one_token ();
2583 next.value = yylval;
2584
2585 if (next.token == NAME && last_was_coloncolon)
2586 {
2587 int classification;
2588
2589 classification = classify_inner_name (first_was_coloncolon
2590 ? NULL
2591 : expression_context_block,
2592 first_iter);
2593 /* We keep going until we either run out of names, or until
2594 we have a qualified name which is not a type. */
2595 if (classification != TYPENAME)
2596 {
2597 /* Push the final component and leave the loop. */
2598 VEC_safe_push (token_and_value, token_fifo, &next);
2599 break;
2600 }
2601
2602 /* Update the partial name we are constructing. */
2603 if (!first_iter)
2604 {
2605 /* We don't want to put a leading "::" into the name. */
2606 obstack_grow_str (&name_obstack, "::");
2607 }
2608 obstack_grow (&name_obstack, next.value.sval.ptr,
2609 next.value.sval.length);
2610
2611 yylval.sval.ptr = obstack_base (&name_obstack);
2612 yylval.sval.length = obstack_object_size (&name_obstack);
2613 current.value = yylval;
2614 current.token = classification;
2615
2616 last_was_coloncolon = 0;
2617 }
2618 else if (next.token == COLONCOLON && !last_was_coloncolon)
2619 last_was_coloncolon = 1;
2620 else
2621 {
2622 /* We've reached the end of the name. */
2623 VEC_safe_push (token_and_value, token_fifo, &next);
2624 break;
2625 }
2626
2627 first_iter = 0;
2628 }
2629
2630 popping = 1;
2631
2632 /* If we ended with a "::", insert it too. */
2633 if (last_was_coloncolon)
2634 {
2635 token_and_value cc;
2636 memset (&cc, 0, sizeof (token_and_value));
2637 if (first_was_coloncolon && first_iter)
2638 {
2639 yylval = cc.value;
2640 return COLONCOLON;
2641 }
2642 cc.token = COLONCOLON;
2643 VEC_safe_insert (token_and_value, token_fifo, 0, &cc);
2644 }
2645
2646 yylval = current.value;
2647 yylval.sval.ptr = obstack_copy0 (&expansion_obstack,
2648 yylval.sval.ptr,
2649 yylval.sval.length);
2650 return current.token;
2651 }
2652
2653 int
2654 c_parse (void)
2655 {
2656 int result;
2657 struct cleanup *back_to = make_cleanup (free_current_contents,
2658 &expression_macro_scope);
2659
2660 /* Set up the scope for macro expansion. */
2661 expression_macro_scope = NULL;
2662
2663 if (expression_context_block)
2664 expression_macro_scope
2665 = sal_macro_scope (find_pc_line (expression_context_pc, 0));
2666 else
2667 expression_macro_scope = default_macro_scope ();
2668 if (! expression_macro_scope)
2669 expression_macro_scope = user_macro_scope ();
2670
2671 /* Initialize macro expansion code. */
2672 obstack_init (&expansion_obstack);
2673 gdb_assert (! macro_original_text);
2674 make_cleanup (scan_macro_cleanup, 0);
2675
2676 make_cleanup_restore_integer (&yydebug);
2677 yydebug = parser_debug;
2678
2679 /* Initialize some state used by the lexer. */
2680 last_was_structop = 0;
2681 saw_name_at_eof = 0;
2682
2683 VEC_free (token_and_value, token_fifo);
2684 popping = 0;
2685 obstack_init (&name_obstack);
2686 make_cleanup_obstack_free (&name_obstack);
2687
2688 result = yyparse ();
2689 do_cleanups (back_to);
2690 return result;
2691 }
2692
2693
2694 void
2695 yyerror (char *msg)
2696 {
2697 if (prev_lexptr)
2698 lexptr = prev_lexptr;
2699
2700 error (_("A %s in expression, near `%s'."), (msg ? msg : "error"), lexptr);
2701 }
This page took 0.108891 seconds and 5 git commands to generate.