44fa6905446380c8676c514f33266f7fbb4c9535
[deliverable/binutils-gdb.git] / gdb / c-exp.y
1 /* YACC parser for C expressions, for GDB.
2 Copyright (C) 1986, 1989, 1990, 1991, 1992, 1993, 1994, 1995, 1996, 1997,
3 1998, 1999, 2000, 2003, 2004, 2006, 2007, 2008, 2009, 2010
4 Free Software Foundation, Inc.
5
6 This file is part of GDB.
7
8 This program is free software; you can redistribute it and/or modify
9 it under the terms of the GNU General Public License as published by
10 the Free Software Foundation; either version 3 of the License, or
11 (at your option) any later version.
12
13 This program is distributed in the hope that it will be useful,
14 but WITHOUT ANY WARRANTY; without even the implied warranty of
15 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
16 GNU General Public License for more details.
17
18 You should have received a copy of the GNU General Public License
19 along with this program. If not, see <http://www.gnu.org/licenses/>. */
20
21 /* Parse a C expression from text in a string,
22 and return the result as a struct expression pointer.
23 That structure contains arithmetic operations in reverse polish,
24 with constants represented by operations that are followed by special data.
25 See expression.h for the details of the format.
26 What is important here is that it can be built up sequentially
27 during the process of parsing; the lower levels of the tree always
28 come first in the result.
29
30 Note that malloc's and realloc's in this file are transformed to
31 xmalloc and xrealloc respectively by the same sed command in the
32 makefile that remaps any other malloc/realloc inserted by the parser
33 generator. Doing this with #defines and trying to control the interaction
34 with include files (<malloc.h> and <stdlib.h> for example) just became
35 too messy, particularly when such includes can be inserted at random
36 times by the parser generator. */
37
38 %{
39
40 #include "defs.h"
41 #include "gdb_string.h"
42 #include <ctype.h>
43 #include "expression.h"
44 #include "value.h"
45 #include "parser-defs.h"
46 #include "language.h"
47 #include "c-lang.h"
48 #include "bfd.h" /* Required by objfiles.h. */
49 #include "symfile.h" /* Required by objfiles.h. */
50 #include "objfiles.h" /* For have_full_symbols and have_partial_symbols */
51 #include "charset.h"
52 #include "block.h"
53 #include "cp-support.h"
54 #include "dfp.h"
55 #include "gdb_assert.h"
56 #include "macroscope.h"
57
58 #define parse_type builtin_type (parse_gdbarch)
59
60 /* Remap normal yacc parser interface names (yyparse, yylex, yyerror, etc),
61 as well as gratuitiously global symbol names, so we can have multiple
62 yacc generated parsers in gdb. Note that these are only the variables
63 produced by yacc. If other parser generators (bison, byacc, etc) produce
64 additional global names that conflict at link time, then those parser
65 generators need to be fixed instead of adding those names to this list. */
66
67 #define yymaxdepth c_maxdepth
68 #define yyparse c_parse_internal
69 #define yylex c_lex
70 #define yyerror c_error
71 #define yylval c_lval
72 #define yychar c_char
73 #define yydebug c_debug
74 #define yypact c_pact
75 #define yyr1 c_r1
76 #define yyr2 c_r2
77 #define yydef c_def
78 #define yychk c_chk
79 #define yypgo c_pgo
80 #define yyact c_act
81 #define yyexca c_exca
82 #define yyerrflag c_errflag
83 #define yynerrs c_nerrs
84 #define yyps c_ps
85 #define yypv c_pv
86 #define yys c_s
87 #define yy_yys c_yys
88 #define yystate c_state
89 #define yytmp c_tmp
90 #define yyv c_v
91 #define yy_yyv c_yyv
92 #define yyval c_val
93 #define yylloc c_lloc
94 #define yyreds c_reds /* With YYDEBUG defined */
95 #define yytoks c_toks /* With YYDEBUG defined */
96 #define yyname c_name /* With YYDEBUG defined */
97 #define yyrule c_rule /* With YYDEBUG defined */
98 #define yylhs c_yylhs
99 #define yylen c_yylen
100 #define yydefred c_yydefred
101 #define yydgoto c_yydgoto
102 #define yysindex c_yysindex
103 #define yyrindex c_yyrindex
104 #define yygindex c_yygindex
105 #define yytable c_yytable
106 #define yycheck c_yycheck
107
108 #ifndef YYDEBUG
109 #define YYDEBUG 1 /* Default to yydebug support */
110 #endif
111
112 #define YYFPRINTF parser_fprintf
113
114 int yyparse (void);
115
116 static int yylex (void);
117
118 void yyerror (char *);
119
120 %}
121
122 /* Although the yacc "value" of an expression is not used,
123 since the result is stored in the structure being created,
124 other node types do have values. */
125
126 %union
127 {
128 LONGEST lval;
129 struct {
130 LONGEST val;
131 struct type *type;
132 } typed_val_int;
133 struct {
134 DOUBLEST dval;
135 struct type *type;
136 } typed_val_float;
137 struct {
138 gdb_byte val[16];
139 struct type *type;
140 } typed_val_decfloat;
141 struct symbol *sym;
142 struct type *tval;
143 struct stoken sval;
144 struct typed_stoken tsval;
145 struct ttype tsym;
146 struct symtoken ssym;
147 int voidval;
148 struct block *bval;
149 enum exp_opcode opcode;
150 struct internalvar *ivar;
151
152 struct stoken_vector svec;
153 struct type **tvec;
154 int *ivec;
155 }
156
157 %{
158 /* YYSTYPE gets defined by %union */
159 static int parse_number (char *, int, int, YYSTYPE *);
160 static struct stoken operator_stoken (const char *);
161 %}
162
163 %type <voidval> exp exp1 type_exp start variable qualified_name lcurly
164 %type <lval> rcurly
165 %type <tval> type typebase
166 %type <tvec> nonempty_typelist
167 /* %type <bval> block */
168
169 /* Fancy type parsing. */
170 %type <voidval> func_mod direct_abs_decl abs_decl
171 %type <tval> ptype
172 %type <lval> array_mod
173
174 %token <typed_val_int> INT
175 %token <typed_val_float> FLOAT
176 %token <typed_val_decfloat> DECFLOAT
177
178 /* Both NAME and TYPENAME tokens represent symbols in the input,
179 and both convey their data as strings.
180 But a TYPENAME is a string that happens to be defined as a typedef
181 or builtin type name (such as int or char)
182 and a NAME is any other symbol.
183 Contexts where this distinction is not important can use the
184 nonterminal "name", which matches either NAME or TYPENAME. */
185
186 %token <tsval> STRING
187 %token <tsval> CHAR
188 %token <ssym> NAME /* BLOCKNAME defined below to give it higher precedence. */
189 %token <voidval> COMPLETE
190 %token <tsym> TYPENAME
191 %type <sval> name
192 %type <svec> string_exp
193 %type <ssym> name_not_typename
194 %type <tsym> typename
195
196 /* A NAME_OR_INT is a symbol which is not known in the symbol table,
197 but which would parse as a valid number in the current input radix.
198 E.g. "c" when input_radix==16. Depending on the parse, it will be
199 turned into a name or into a number. */
200
201 %token <ssym> NAME_OR_INT
202
203 %token OPERATOR
204 %token STRUCT CLASS UNION ENUM SIZEOF UNSIGNED COLONCOLON
205 %token TEMPLATE
206 %token ERROR
207 %token NEW DELETE
208 %type <sval> operator
209 %token REINTERPRET_CAST DYNAMIC_CAST STATIC_CAST CONST_CAST
210
211 /* Special type cases, put in to allow the parser to distinguish different
212 legal basetypes. */
213 %token SIGNED_KEYWORD LONG SHORT INT_KEYWORD CONST_KEYWORD VOLATILE_KEYWORD DOUBLE_KEYWORD
214
215 %token <sval> VARIABLE
216
217 %token <opcode> ASSIGN_MODIFY
218
219 /* C++ */
220 %token TRUEKEYWORD
221 %token FALSEKEYWORD
222
223
224 %left ','
225 %left ABOVE_COMMA
226 %right '=' ASSIGN_MODIFY
227 %right '?'
228 %left OROR
229 %left ANDAND
230 %left '|'
231 %left '^'
232 %left '&'
233 %left EQUAL NOTEQUAL
234 %left '<' '>' LEQ GEQ
235 %left LSH RSH
236 %left '@'
237 %left '+' '-'
238 %left '*' '/' '%'
239 %right UNARY INCREMENT DECREMENT
240 %right ARROW ARROW_STAR '.' DOT_STAR '[' '('
241 %token <ssym> BLOCKNAME
242 %token <bval> FILENAME
243 %type <bval> block
244 %left COLONCOLON
245
246 \f
247 %%
248
249 start : exp1
250 | type_exp
251 ;
252
253 type_exp: type
254 { write_exp_elt_opcode(OP_TYPE);
255 write_exp_elt_type($1);
256 write_exp_elt_opcode(OP_TYPE);}
257 ;
258
259 /* Expressions, including the comma operator. */
260 exp1 : exp
261 | exp1 ',' exp
262 { write_exp_elt_opcode (BINOP_COMMA); }
263 ;
264
265 /* Expressions, not including the comma operator. */
266 exp : '*' exp %prec UNARY
267 { write_exp_elt_opcode (UNOP_IND); }
268 ;
269
270 exp : '&' exp %prec UNARY
271 { write_exp_elt_opcode (UNOP_ADDR); }
272 ;
273
274 exp : '-' exp %prec UNARY
275 { write_exp_elt_opcode (UNOP_NEG); }
276 ;
277
278 exp : '+' exp %prec UNARY
279 { write_exp_elt_opcode (UNOP_PLUS); }
280 ;
281
282 exp : '!' exp %prec UNARY
283 { write_exp_elt_opcode (UNOP_LOGICAL_NOT); }
284 ;
285
286 exp : '~' exp %prec UNARY
287 { write_exp_elt_opcode (UNOP_COMPLEMENT); }
288 ;
289
290 exp : INCREMENT exp %prec UNARY
291 { write_exp_elt_opcode (UNOP_PREINCREMENT); }
292 ;
293
294 exp : DECREMENT exp %prec UNARY
295 { write_exp_elt_opcode (UNOP_PREDECREMENT); }
296 ;
297
298 exp : exp INCREMENT %prec UNARY
299 { write_exp_elt_opcode (UNOP_POSTINCREMENT); }
300 ;
301
302 exp : exp DECREMENT %prec UNARY
303 { write_exp_elt_opcode (UNOP_POSTDECREMENT); }
304 ;
305
306 exp : SIZEOF exp %prec UNARY
307 { write_exp_elt_opcode (UNOP_SIZEOF); }
308 ;
309
310 exp : exp ARROW name
311 { write_exp_elt_opcode (STRUCTOP_PTR);
312 write_exp_string ($3);
313 write_exp_elt_opcode (STRUCTOP_PTR); }
314 ;
315
316 exp : exp ARROW name COMPLETE
317 { mark_struct_expression ();
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 COMPLETE
324 { struct stoken s;
325 mark_struct_expression ();
326 write_exp_elt_opcode (STRUCTOP_PTR);
327 s.ptr = "";
328 s.length = 0;
329 write_exp_string (s);
330 write_exp_elt_opcode (STRUCTOP_PTR); }
331 ;
332
333 exp : exp ARROW qualified_name
334 { /* exp->type::name becomes exp->*(&type::name) */
335 /* Note: this doesn't work if name is a
336 static member! FIXME */
337 write_exp_elt_opcode (UNOP_ADDR);
338 write_exp_elt_opcode (STRUCTOP_MPTR); }
339 ;
340
341 exp : exp ARROW_STAR exp
342 { write_exp_elt_opcode (STRUCTOP_MPTR); }
343 ;
344
345 exp : exp '.' name
346 { write_exp_elt_opcode (STRUCTOP_STRUCT);
347 write_exp_string ($3);
348 write_exp_elt_opcode (STRUCTOP_STRUCT); }
349 ;
350
351 exp : exp '.' name COMPLETE
352 { mark_struct_expression ();
353 write_exp_elt_opcode (STRUCTOP_STRUCT);
354 write_exp_string ($3);
355 write_exp_elt_opcode (STRUCTOP_STRUCT); }
356 ;
357
358 exp : exp '.' COMPLETE
359 { struct stoken s;
360 mark_struct_expression ();
361 write_exp_elt_opcode (STRUCTOP_STRUCT);
362 s.ptr = "";
363 s.length = 0;
364 write_exp_string (s);
365 write_exp_elt_opcode (STRUCTOP_STRUCT); }
366 ;
367
368 exp : exp '.' qualified_name
369 { /* exp.type::name becomes exp.*(&type::name) */
370 /* Note: this doesn't work if name is a
371 static member! FIXME */
372 write_exp_elt_opcode (UNOP_ADDR);
373 write_exp_elt_opcode (STRUCTOP_MEMBER); }
374 ;
375
376 exp : exp DOT_STAR exp
377 { write_exp_elt_opcode (STRUCTOP_MEMBER); }
378 ;
379
380 exp : exp '[' exp1 ']'
381 { write_exp_elt_opcode (BINOP_SUBSCRIPT); }
382 ;
383
384 exp : exp '('
385 /* This is to save the value of arglist_len
386 being accumulated by an outer function call. */
387 { start_arglist (); }
388 arglist ')' %prec ARROW
389 { write_exp_elt_opcode (OP_FUNCALL);
390 write_exp_elt_longcst ((LONGEST) end_arglist ());
391 write_exp_elt_opcode (OP_FUNCALL); }
392 ;
393
394 lcurly : '{'
395 { start_arglist (); }
396 ;
397
398 arglist :
399 ;
400
401 arglist : exp
402 { arglist_len = 1; }
403 ;
404
405 arglist : arglist ',' exp %prec ABOVE_COMMA
406 { arglist_len++; }
407 ;
408
409 exp : exp '(' nonempty_typelist ')' const_or_volatile
410 { int i;
411 write_exp_elt_opcode (TYPE_INSTANCE);
412 write_exp_elt_longcst ((LONGEST) $<ivec>3[0]);
413 for (i = 0; i < $<ivec>3[0]; ++i)
414 write_exp_elt_type ($<tvec>3[i + 1]);
415 write_exp_elt_longcst((LONGEST) $<ivec>3[0]);
416 write_exp_elt_opcode (TYPE_INSTANCE);
417 free ($3);
418 }
419 ;
420
421 rcurly : '}'
422 { $$ = end_arglist () - 1; }
423 ;
424 exp : lcurly arglist rcurly %prec ARROW
425 { write_exp_elt_opcode (OP_ARRAY);
426 write_exp_elt_longcst ((LONGEST) 0);
427 write_exp_elt_longcst ((LONGEST) $3);
428 write_exp_elt_opcode (OP_ARRAY); }
429 ;
430
431 exp : lcurly type rcurly exp %prec UNARY
432 { write_exp_elt_opcode (UNOP_MEMVAL);
433 write_exp_elt_type ($2);
434 write_exp_elt_opcode (UNOP_MEMVAL); }
435 ;
436
437 exp : '(' type ')' exp %prec UNARY
438 { write_exp_elt_opcode (UNOP_CAST);
439 write_exp_elt_type ($2);
440 write_exp_elt_opcode (UNOP_CAST); }
441 ;
442
443 exp : '(' exp1 ')'
444 { }
445 ;
446
447 /* Binary operators in order of decreasing precedence. */
448
449 exp : exp '@' exp
450 { write_exp_elt_opcode (BINOP_REPEAT); }
451 ;
452
453 exp : exp '*' exp
454 { write_exp_elt_opcode (BINOP_MUL); }
455 ;
456
457 exp : exp '/' exp
458 { write_exp_elt_opcode (BINOP_DIV); }
459 ;
460
461 exp : exp '%' exp
462 { write_exp_elt_opcode (BINOP_REM); }
463 ;
464
465 exp : exp '+' exp
466 { write_exp_elt_opcode (BINOP_ADD); }
467 ;
468
469 exp : exp '-' exp
470 { write_exp_elt_opcode (BINOP_SUB); }
471 ;
472
473 exp : exp LSH exp
474 { write_exp_elt_opcode (BINOP_LSH); }
475 ;
476
477 exp : exp RSH exp
478 { write_exp_elt_opcode (BINOP_RSH); }
479 ;
480
481 exp : exp EQUAL exp
482 { write_exp_elt_opcode (BINOP_EQUAL); }
483 ;
484
485 exp : exp NOTEQUAL exp
486 { write_exp_elt_opcode (BINOP_NOTEQUAL); }
487 ;
488
489 exp : exp LEQ exp
490 { write_exp_elt_opcode (BINOP_LEQ); }
491 ;
492
493 exp : exp GEQ exp
494 { write_exp_elt_opcode (BINOP_GEQ); }
495 ;
496
497 exp : exp '<' exp
498 { write_exp_elt_opcode (BINOP_LESS); }
499 ;
500
501 exp : exp '>' exp
502 { write_exp_elt_opcode (BINOP_GTR); }
503 ;
504
505 exp : exp '&' exp
506 { write_exp_elt_opcode (BINOP_BITWISE_AND); }
507 ;
508
509 exp : exp '^' exp
510 { write_exp_elt_opcode (BINOP_BITWISE_XOR); }
511 ;
512
513 exp : exp '|' exp
514 { write_exp_elt_opcode (BINOP_BITWISE_IOR); }
515 ;
516
517 exp : exp ANDAND exp
518 { write_exp_elt_opcode (BINOP_LOGICAL_AND); }
519 ;
520
521 exp : exp OROR exp
522 { write_exp_elt_opcode (BINOP_LOGICAL_OR); }
523 ;
524
525 exp : exp '?' exp ':' exp %prec '?'
526 { write_exp_elt_opcode (TERNOP_COND); }
527 ;
528
529 exp : exp '=' exp
530 { write_exp_elt_opcode (BINOP_ASSIGN); }
531 ;
532
533 exp : exp ASSIGN_MODIFY exp
534 { write_exp_elt_opcode (BINOP_ASSIGN_MODIFY);
535 write_exp_elt_opcode ($2);
536 write_exp_elt_opcode (BINOP_ASSIGN_MODIFY); }
537 ;
538
539 exp : INT
540 { write_exp_elt_opcode (OP_LONG);
541 write_exp_elt_type ($1.type);
542 write_exp_elt_longcst ((LONGEST)($1.val));
543 write_exp_elt_opcode (OP_LONG); }
544 ;
545
546 exp : CHAR
547 {
548 struct stoken_vector vec;
549 vec.len = 1;
550 vec.tokens = &$1;
551 write_exp_string_vector ($1.type, &vec);
552 }
553 ;
554
555 exp : NAME_OR_INT
556 { YYSTYPE val;
557 parse_number ($1.stoken.ptr, $1.stoken.length, 0, &val);
558 write_exp_elt_opcode (OP_LONG);
559 write_exp_elt_type (val.typed_val_int.type);
560 write_exp_elt_longcst ((LONGEST)val.typed_val_int.val);
561 write_exp_elt_opcode (OP_LONG);
562 }
563 ;
564
565
566 exp : FLOAT
567 { write_exp_elt_opcode (OP_DOUBLE);
568 write_exp_elt_type ($1.type);
569 write_exp_elt_dblcst ($1.dval);
570 write_exp_elt_opcode (OP_DOUBLE); }
571 ;
572
573 exp : DECFLOAT
574 { write_exp_elt_opcode (OP_DECFLOAT);
575 write_exp_elt_type ($1.type);
576 write_exp_elt_decfloatcst ($1.val);
577 write_exp_elt_opcode (OP_DECFLOAT); }
578 ;
579
580 exp : variable
581 ;
582
583 exp : VARIABLE
584 {
585 write_dollar_variable ($1);
586 }
587 ;
588
589 exp : SIZEOF '(' type ')' %prec UNARY
590 { write_exp_elt_opcode (OP_LONG);
591 write_exp_elt_type (parse_type->builtin_int);
592 CHECK_TYPEDEF ($3);
593 write_exp_elt_longcst ((LONGEST) TYPE_LENGTH ($3));
594 write_exp_elt_opcode (OP_LONG); }
595 ;
596
597 exp : REINTERPRET_CAST '<' type '>' '(' exp ')' %prec UNARY
598 { write_exp_elt_opcode (UNOP_REINTERPRET_CAST);
599 write_exp_elt_type ($3);
600 write_exp_elt_opcode (UNOP_REINTERPRET_CAST); }
601 ;
602
603 exp : STATIC_CAST '<' type '>' '(' exp ')' %prec UNARY
604 { write_exp_elt_opcode (UNOP_CAST);
605 write_exp_elt_type ($3);
606 write_exp_elt_opcode (UNOP_CAST); }
607 ;
608
609 exp : DYNAMIC_CAST '<' type '>' '(' exp ')' %prec UNARY
610 { write_exp_elt_opcode (UNOP_DYNAMIC_CAST);
611 write_exp_elt_type ($3);
612 write_exp_elt_opcode (UNOP_DYNAMIC_CAST); }
613 ;
614
615 exp : CONST_CAST '<' type '>' '(' exp ')' %prec UNARY
616 { /* We could do more error checking here, but
617 it doesn't seem worthwhile. */
618 write_exp_elt_opcode (UNOP_CAST);
619 write_exp_elt_type ($3);
620 write_exp_elt_opcode (UNOP_CAST); }
621 ;
622
623 string_exp:
624 STRING
625 {
626 /* We copy the string here, and not in the
627 lexer, to guarantee that we do not leak a
628 string. Note that we follow the
629 NUL-termination convention of the
630 lexer. */
631 struct typed_stoken *vec = XNEW (struct typed_stoken);
632 $$.len = 1;
633 $$.tokens = vec;
634
635 vec->type = $1.type;
636 vec->length = $1.length;
637 vec->ptr = malloc ($1.length + 1);
638 memcpy (vec->ptr, $1.ptr, $1.length + 1);
639 }
640
641 | string_exp STRING
642 {
643 /* Note that we NUL-terminate here, but just
644 for convenience. */
645 char *p;
646 ++$$.len;
647 $$.tokens = realloc ($$.tokens,
648 $$.len * sizeof (struct typed_stoken));
649
650 p = malloc ($2.length + 1);
651 memcpy (p, $2.ptr, $2.length + 1);
652
653 $$.tokens[$$.len - 1].type = $2.type;
654 $$.tokens[$$.len - 1].length = $2.length;
655 $$.tokens[$$.len - 1].ptr = p;
656 }
657 ;
658
659 exp : string_exp
660 {
661 int i;
662 enum c_string_type type = C_STRING;
663
664 for (i = 0; i < $1.len; ++i)
665 {
666 switch ($1.tokens[i].type)
667 {
668 case C_STRING:
669 break;
670 case C_WIDE_STRING:
671 case C_STRING_16:
672 case C_STRING_32:
673 if (type != C_STRING
674 && type != $1.tokens[i].type)
675 error ("Undefined string concatenation.");
676 type = $1.tokens[i].type;
677 break;
678 default:
679 /* internal error */
680 internal_error (__FILE__, __LINE__,
681 "unrecognized type in string concatenation");
682 }
683 }
684
685 write_exp_string_vector (type, &$1);
686 for (i = 0; i < $1.len; ++i)
687 free ($1.tokens[i].ptr);
688 free ($1.tokens);
689 }
690 ;
691
692 /* C++. */
693 exp : TRUEKEYWORD
694 { write_exp_elt_opcode (OP_LONG);
695 write_exp_elt_type (parse_type->builtin_bool);
696 write_exp_elt_longcst ((LONGEST) 1);
697 write_exp_elt_opcode (OP_LONG); }
698 ;
699
700 exp : FALSEKEYWORD
701 { write_exp_elt_opcode (OP_LONG);
702 write_exp_elt_type (parse_type->builtin_bool);
703 write_exp_elt_longcst ((LONGEST) 0);
704 write_exp_elt_opcode (OP_LONG); }
705 ;
706
707 /* end of C++. */
708
709 block : BLOCKNAME
710 {
711 if ($1.sym)
712 $$ = SYMBOL_BLOCK_VALUE ($1.sym);
713 else
714 error ("No file or function \"%s\".",
715 copy_name ($1.stoken));
716 }
717 | FILENAME
718 {
719 $$ = $1;
720 }
721 ;
722
723 block : block COLONCOLON name
724 { struct symbol *tem
725 = lookup_symbol (copy_name ($3), $1,
726 VAR_DOMAIN, (int *) NULL);
727 if (!tem || SYMBOL_CLASS (tem) != LOC_BLOCK)
728 error ("No function \"%s\" in specified context.",
729 copy_name ($3));
730 $$ = SYMBOL_BLOCK_VALUE (tem); }
731 ;
732
733 variable: block COLONCOLON name
734 { struct symbol *sym;
735 sym = lookup_symbol (copy_name ($3), $1,
736 VAR_DOMAIN, (int *) NULL);
737 if (sym == 0)
738 error ("No symbol \"%s\" in specified context.",
739 copy_name ($3));
740
741 write_exp_elt_opcode (OP_VAR_VALUE);
742 /* block_found is set by lookup_symbol. */
743 write_exp_elt_block (block_found);
744 write_exp_elt_sym (sym);
745 write_exp_elt_opcode (OP_VAR_VALUE); }
746 ;
747
748 qualified_name: TYPENAME COLONCOLON name
749 {
750 struct type *type = $1.type;
751 CHECK_TYPEDEF (type);
752 if (TYPE_CODE (type) != TYPE_CODE_STRUCT
753 && TYPE_CODE (type) != TYPE_CODE_UNION
754 && TYPE_CODE (type) != TYPE_CODE_NAMESPACE)
755 error ("`%s' is not defined as an aggregate type.",
756 TYPE_NAME (type));
757
758 write_exp_elt_opcode (OP_SCOPE);
759 write_exp_elt_type (type);
760 write_exp_string ($3);
761 write_exp_elt_opcode (OP_SCOPE);
762 }
763 | TYPENAME COLONCOLON '~' name
764 {
765 struct type *type = $1.type;
766 struct stoken tmp_token;
767 CHECK_TYPEDEF (type);
768 if (TYPE_CODE (type) != TYPE_CODE_STRUCT
769 && TYPE_CODE (type) != TYPE_CODE_UNION
770 && TYPE_CODE (type) != TYPE_CODE_NAMESPACE)
771 error ("`%s' is not defined as an aggregate type.",
772 TYPE_NAME (type));
773
774 tmp_token.ptr = (char*) alloca ($4.length + 2);
775 tmp_token.length = $4.length + 1;
776 tmp_token.ptr[0] = '~';
777 memcpy (tmp_token.ptr+1, $4.ptr, $4.length);
778 tmp_token.ptr[tmp_token.length] = 0;
779
780 /* Check for valid destructor name. */
781 destructor_name_p (tmp_token.ptr, type);
782 write_exp_elt_opcode (OP_SCOPE);
783 write_exp_elt_type (type);
784 write_exp_string (tmp_token);
785 write_exp_elt_opcode (OP_SCOPE);
786 }
787 | TYPENAME COLONCOLON name COLONCOLON name
788 {
789 char *copy = copy_name ($3);
790 error (_("No type \"%s\" within class "
791 "or namespace \"%s\"."),
792 copy, TYPE_NAME ($1.type));
793 }
794 ;
795
796 variable: qualified_name
797 | COLONCOLON name_not_typename
798 {
799 char *name = copy_name ($2.stoken);
800 struct symbol *sym;
801 struct minimal_symbol *msymbol;
802
803 sym =
804 lookup_symbol (name, (const struct block *) NULL,
805 VAR_DOMAIN, (int *) NULL);
806 if (sym)
807 {
808 write_exp_elt_opcode (OP_VAR_VALUE);
809 write_exp_elt_block (NULL);
810 write_exp_elt_sym (sym);
811 write_exp_elt_opcode (OP_VAR_VALUE);
812 break;
813 }
814
815 msymbol = lookup_minimal_symbol (name, NULL, NULL);
816 if (msymbol != NULL)
817 write_exp_msymbol (msymbol);
818 else if (!have_full_symbols () && !have_partial_symbols ())
819 error ("No symbol table is loaded. Use the \"file\" command.");
820 else
821 error ("No symbol \"%s\" in current context.", name);
822 }
823 ;
824
825 variable: name_not_typename
826 { struct symbol *sym = $1.sym;
827
828 if (sym)
829 {
830 if (symbol_read_needs_frame (sym))
831 {
832 if (innermost_block == 0
833 || contained_in (block_found,
834 innermost_block))
835 innermost_block = block_found;
836 }
837
838 write_exp_elt_opcode (OP_VAR_VALUE);
839 /* We want to use the selected frame, not
840 another more inner frame which happens to
841 be in the same block. */
842 write_exp_elt_block (NULL);
843 write_exp_elt_sym (sym);
844 write_exp_elt_opcode (OP_VAR_VALUE);
845 }
846 else if ($1.is_a_field_of_this)
847 {
848 /* C++: it hangs off of `this'. Must
849 not inadvertently convert from a method call
850 to data ref. */
851 if (innermost_block == 0
852 || contained_in (block_found,
853 innermost_block))
854 innermost_block = block_found;
855 write_exp_elt_opcode (OP_THIS);
856 write_exp_elt_opcode (OP_THIS);
857 write_exp_elt_opcode (STRUCTOP_PTR);
858 write_exp_string ($1.stoken);
859 write_exp_elt_opcode (STRUCTOP_PTR);
860 }
861 else
862 {
863 struct minimal_symbol *msymbol;
864 char *arg = copy_name ($1.stoken);
865
866 msymbol =
867 lookup_minimal_symbol (arg, NULL, NULL);
868 if (msymbol != NULL)
869 write_exp_msymbol (msymbol);
870 else if (!have_full_symbols () && !have_partial_symbols ())
871 error ("No symbol table is loaded. Use the \"file\" command.");
872 else
873 error ("No symbol \"%s\" in current context.",
874 copy_name ($1.stoken));
875 }
876 }
877 ;
878
879 space_identifier : '@' NAME
880 { push_type_address_space (copy_name ($2.stoken));
881 push_type (tp_space_identifier);
882 }
883 ;
884
885 const_or_volatile: const_or_volatile_noopt
886 |
887 ;
888
889 cv_with_space_id : const_or_volatile space_identifier const_or_volatile
890 ;
891
892 const_or_volatile_or_space_identifier_noopt: cv_with_space_id
893 | const_or_volatile_noopt
894 ;
895
896 const_or_volatile_or_space_identifier:
897 const_or_volatile_or_space_identifier_noopt
898 |
899 ;
900
901 abs_decl: '*'
902 { push_type (tp_pointer); $$ = 0; }
903 | '*' abs_decl
904 { push_type (tp_pointer); $$ = $2; }
905 | '&'
906 { push_type (tp_reference); $$ = 0; }
907 | '&' abs_decl
908 { push_type (tp_reference); $$ = $2; }
909 | direct_abs_decl
910 ;
911
912 direct_abs_decl: '(' abs_decl ')'
913 { $$ = $2; }
914 | direct_abs_decl array_mod
915 {
916 push_type_int ($2);
917 push_type (tp_array);
918 }
919 | array_mod
920 {
921 push_type_int ($1);
922 push_type (tp_array);
923 $$ = 0;
924 }
925
926 | direct_abs_decl func_mod
927 { push_type (tp_function); }
928 | func_mod
929 { push_type (tp_function); }
930 ;
931
932 array_mod: '[' ']'
933 { $$ = -1; }
934 | '[' INT ']'
935 { $$ = $2.val; }
936 ;
937
938 func_mod: '(' ')'
939 { $$ = 0; }
940 | '(' nonempty_typelist ')'
941 { free ($2); $$ = 0; }
942 ;
943
944 /* We used to try to recognize pointer to member types here, but
945 that didn't work (shift/reduce conflicts meant that these rules never
946 got executed). The problem is that
947 int (foo::bar::baz::bizzle)
948 is a function type but
949 int (foo::bar::baz::bizzle::*)
950 is a pointer to member type. Stroustrup loses again! */
951
952 type : ptype
953 ;
954
955 typebase /* Implements (approximately): (type-qualifier)* type-specifier */
956 : TYPENAME
957 { $$ = $1.type; }
958 | INT_KEYWORD
959 { $$ = parse_type->builtin_int; }
960 | LONG
961 { $$ = parse_type->builtin_long; }
962 | SHORT
963 { $$ = parse_type->builtin_short; }
964 | LONG INT_KEYWORD
965 { $$ = parse_type->builtin_long; }
966 | LONG SIGNED_KEYWORD INT_KEYWORD
967 { $$ = parse_type->builtin_long; }
968 | LONG SIGNED_KEYWORD
969 { $$ = parse_type->builtin_long; }
970 | SIGNED_KEYWORD LONG INT_KEYWORD
971 { $$ = parse_type->builtin_long; }
972 | UNSIGNED LONG INT_KEYWORD
973 { $$ = parse_type->builtin_unsigned_long; }
974 | LONG UNSIGNED INT_KEYWORD
975 { $$ = parse_type->builtin_unsigned_long; }
976 | LONG UNSIGNED
977 { $$ = parse_type->builtin_unsigned_long; }
978 | LONG LONG
979 { $$ = parse_type->builtin_long_long; }
980 | LONG LONG INT_KEYWORD
981 { $$ = parse_type->builtin_long_long; }
982 | LONG LONG SIGNED_KEYWORD INT_KEYWORD
983 { $$ = parse_type->builtin_long_long; }
984 | LONG LONG SIGNED_KEYWORD
985 { $$ = parse_type->builtin_long_long; }
986 | SIGNED_KEYWORD LONG LONG
987 { $$ = parse_type->builtin_long_long; }
988 | SIGNED_KEYWORD LONG LONG INT_KEYWORD
989 { $$ = parse_type->builtin_long_long; }
990 | UNSIGNED LONG LONG
991 { $$ = parse_type->builtin_unsigned_long_long; }
992 | UNSIGNED LONG LONG INT_KEYWORD
993 { $$ = parse_type->builtin_unsigned_long_long; }
994 | LONG LONG UNSIGNED
995 { $$ = parse_type->builtin_unsigned_long_long; }
996 | LONG LONG UNSIGNED INT_KEYWORD
997 { $$ = parse_type->builtin_unsigned_long_long; }
998 | SHORT INT_KEYWORD
999 { $$ = parse_type->builtin_short; }
1000 | SHORT SIGNED_KEYWORD INT_KEYWORD
1001 { $$ = parse_type->builtin_short; }
1002 | SHORT SIGNED_KEYWORD
1003 { $$ = parse_type->builtin_short; }
1004 | UNSIGNED SHORT INT_KEYWORD
1005 { $$ = parse_type->builtin_unsigned_short; }
1006 | SHORT UNSIGNED
1007 { $$ = parse_type->builtin_unsigned_short; }
1008 | SHORT UNSIGNED INT_KEYWORD
1009 { $$ = parse_type->builtin_unsigned_short; }
1010 | DOUBLE_KEYWORD
1011 { $$ = parse_type->builtin_double; }
1012 | LONG DOUBLE_KEYWORD
1013 { $$ = parse_type->builtin_long_double; }
1014 | STRUCT name
1015 { $$ = lookup_struct (copy_name ($2),
1016 expression_context_block); }
1017 | CLASS name
1018 { $$ = lookup_struct (copy_name ($2),
1019 expression_context_block); }
1020 | UNION name
1021 { $$ = lookup_union (copy_name ($2),
1022 expression_context_block); }
1023 | ENUM name
1024 { $$ = lookup_enum (copy_name ($2),
1025 expression_context_block); }
1026 | UNSIGNED typename
1027 { $$ = lookup_unsigned_typename (parse_language,
1028 parse_gdbarch,
1029 TYPE_NAME($2.type)); }
1030 | UNSIGNED
1031 { $$ = parse_type->builtin_unsigned_int; }
1032 | SIGNED_KEYWORD typename
1033 { $$ = lookup_signed_typename (parse_language,
1034 parse_gdbarch,
1035 TYPE_NAME($2.type)); }
1036 | SIGNED_KEYWORD
1037 { $$ = parse_type->builtin_int; }
1038 /* It appears that this rule for templates is never
1039 reduced; template recognition happens by lookahead
1040 in the token processing code in yylex. */
1041 | TEMPLATE name '<' type '>'
1042 { $$ = lookup_template_type(copy_name($2), $4,
1043 expression_context_block);
1044 }
1045 | const_or_volatile_or_space_identifier_noopt typebase
1046 { $$ = follow_types ($2); }
1047 | typebase const_or_volatile_or_space_identifier_noopt
1048 { $$ = follow_types ($1); }
1049 ;
1050
1051 typename: TYPENAME
1052 | INT_KEYWORD
1053 {
1054 $$.stoken.ptr = "int";
1055 $$.stoken.length = 3;
1056 $$.type = parse_type->builtin_int;
1057 }
1058 | LONG
1059 {
1060 $$.stoken.ptr = "long";
1061 $$.stoken.length = 4;
1062 $$.type = parse_type->builtin_long;
1063 }
1064 | SHORT
1065 {
1066 $$.stoken.ptr = "short";
1067 $$.stoken.length = 5;
1068 $$.type = parse_type->builtin_short;
1069 }
1070 ;
1071
1072 nonempty_typelist
1073 : type
1074 { $$ = (struct type **) malloc (sizeof (struct type *) * 2);
1075 $<ivec>$[0] = 1; /* Number of types in vector */
1076 $$[1] = $1;
1077 }
1078 | nonempty_typelist ',' type
1079 { int len = sizeof (struct type *) * (++($<ivec>1[0]) + 1);
1080 $$ = (struct type **) realloc ((char *) $1, len);
1081 $$[$<ivec>$[0]] = $3;
1082 }
1083 ;
1084
1085 ptype : typebase
1086 | ptype const_or_volatile_or_space_identifier abs_decl const_or_volatile_or_space_identifier
1087 { $$ = follow_types ($1); }
1088 ;
1089
1090 const_and_volatile: CONST_KEYWORD VOLATILE_KEYWORD
1091 | VOLATILE_KEYWORD CONST_KEYWORD
1092 ;
1093
1094 const_or_volatile_noopt: const_and_volatile
1095 { push_type (tp_const);
1096 push_type (tp_volatile);
1097 }
1098 | CONST_KEYWORD
1099 { push_type (tp_const); }
1100 | VOLATILE_KEYWORD
1101 { push_type (tp_volatile); }
1102 ;
1103
1104 operator: OPERATOR NEW
1105 { $$ = operator_stoken (" new"); }
1106 | OPERATOR DELETE
1107 { $$ = operator_stoken (" delete"); }
1108 | OPERATOR NEW '[' ']'
1109 { $$ = operator_stoken (" new[]"); }
1110 | OPERATOR DELETE '[' ']'
1111 { $$ = operator_stoken (" delete[]"); }
1112 | OPERATOR '+'
1113 { $$ = operator_stoken ("+"); }
1114 | OPERATOR '-'
1115 { $$ = operator_stoken ("-"); }
1116 | OPERATOR '*'
1117 { $$ = operator_stoken ("*"); }
1118 | OPERATOR '/'
1119 { $$ = operator_stoken ("/"); }
1120 | OPERATOR '%'
1121 { $$ = operator_stoken ("%"); }
1122 | OPERATOR '^'
1123 { $$ = operator_stoken ("^"); }
1124 | OPERATOR '&'
1125 { $$ = operator_stoken ("&"); }
1126 | OPERATOR '|'
1127 { $$ = operator_stoken ("|"); }
1128 | OPERATOR '~'
1129 { $$ = operator_stoken ("~"); }
1130 | OPERATOR '!'
1131 { $$ = operator_stoken ("!"); }
1132 | OPERATOR '='
1133 { $$ = operator_stoken ("="); }
1134 | OPERATOR '<'
1135 { $$ = operator_stoken ("<"); }
1136 | OPERATOR '>'
1137 { $$ = operator_stoken (">"); }
1138 | OPERATOR ASSIGN_MODIFY
1139 { const char *op = "unknown";
1140 switch ($2)
1141 {
1142 case BINOP_RSH:
1143 op = ">>=";
1144 break;
1145 case BINOP_LSH:
1146 op = "<<=";
1147 break;
1148 case BINOP_ADD:
1149 op = "+=";
1150 break;
1151 case BINOP_SUB:
1152 op = "-=";
1153 break;
1154 case BINOP_MUL:
1155 op = "*=";
1156 break;
1157 case BINOP_DIV:
1158 op = "/=";
1159 break;
1160 case BINOP_REM:
1161 op = "%=";
1162 break;
1163 case BINOP_BITWISE_IOR:
1164 op = "|=";
1165 break;
1166 case BINOP_BITWISE_AND:
1167 op = "&=";
1168 break;
1169 case BINOP_BITWISE_XOR:
1170 op = "^=";
1171 break;
1172 default:
1173 break;
1174 }
1175
1176 $$ = operator_stoken (op);
1177 }
1178 | OPERATOR LSH
1179 { $$ = operator_stoken ("<<"); }
1180 | OPERATOR RSH
1181 { $$ = operator_stoken (">>"); }
1182 | OPERATOR EQUAL
1183 { $$ = operator_stoken ("=="); }
1184 | OPERATOR NOTEQUAL
1185 { $$ = operator_stoken ("!="); }
1186 | OPERATOR LEQ
1187 { $$ = operator_stoken ("<="); }
1188 | OPERATOR GEQ
1189 { $$ = operator_stoken (">="); }
1190 | OPERATOR ANDAND
1191 { $$ = operator_stoken ("&&"); }
1192 | OPERATOR OROR
1193 { $$ = operator_stoken ("||"); }
1194 | OPERATOR INCREMENT
1195 { $$ = operator_stoken ("++"); }
1196 | OPERATOR DECREMENT
1197 { $$ = operator_stoken ("--"); }
1198 | OPERATOR ','
1199 { $$ = operator_stoken (","); }
1200 | OPERATOR ARROW_STAR
1201 { $$ = operator_stoken ("->*"); }
1202 | OPERATOR ARROW
1203 { $$ = operator_stoken ("->"); }
1204 | OPERATOR '(' ')'
1205 { $$ = operator_stoken ("()"); }
1206 | OPERATOR '[' ']'
1207 { $$ = operator_stoken ("[]"); }
1208 | OPERATOR ptype
1209 { char *name;
1210 long length;
1211 struct ui_file *buf = mem_fileopen ();
1212
1213 c_print_type ($2, NULL, buf, -1, 0);
1214 name = ui_file_xstrdup (buf, &length);
1215 ui_file_delete (buf);
1216 $$ = operator_stoken (name);
1217 free (name);
1218 }
1219 ;
1220
1221
1222
1223 name : NAME { $$ = $1.stoken; }
1224 | BLOCKNAME { $$ = $1.stoken; }
1225 | TYPENAME { $$ = $1.stoken; }
1226 | NAME_OR_INT { $$ = $1.stoken; }
1227 | operator { $$ = $1; }
1228 ;
1229
1230 name_not_typename : NAME
1231 | BLOCKNAME
1232 /* These would be useful if name_not_typename was useful, but it is just
1233 a fake for "variable", so these cause reduce/reduce conflicts because
1234 the parser can't tell whether NAME_OR_INT is a name_not_typename (=variable,
1235 =exp) or just an exp. If name_not_typename was ever used in an lvalue
1236 context where only a name could occur, this might be useful.
1237 | NAME_OR_INT
1238 */
1239 | operator
1240 {
1241 $$.stoken = $1;
1242 $$.sym = lookup_symbol ($1.ptr,
1243 expression_context_block,
1244 VAR_DOMAIN,
1245 &$$.is_a_field_of_this);
1246 }
1247 ;
1248
1249 %%
1250
1251 /* Returns a stoken of the operator name given by OP (which does not
1252 include the string "operator"). */
1253 static struct stoken
1254 operator_stoken (const char *op)
1255 {
1256 static const char *operator_string = "operator";
1257 struct stoken st = { NULL, 0 };
1258 st.length = strlen (operator_string) + strlen (op);
1259 st.ptr = malloc (st.length + 1);
1260 strcpy (st.ptr, operator_string);
1261 strcat (st.ptr, op);
1262
1263 /* The toplevel (c_parse) will free the memory allocated here. */
1264 make_cleanup (free, st.ptr);
1265 return st;
1266 };
1267
1268 /* Take care of parsing a number (anything that starts with a digit).
1269 Set yylval and return the token type; update lexptr.
1270 LEN is the number of characters in it. */
1271
1272 /*** Needs some error checking for the float case ***/
1273
1274 static int
1275 parse_number (char *p, int len, int parsed_float, YYSTYPE *putithere)
1276 {
1277 /* FIXME: Shouldn't these be unsigned? We don't deal with negative values
1278 here, and we do kind of silly things like cast to unsigned. */
1279 LONGEST n = 0;
1280 LONGEST prevn = 0;
1281 ULONGEST un;
1282
1283 int i = 0;
1284 int c;
1285 int base = input_radix;
1286 int unsigned_p = 0;
1287
1288 /* Number of "L" suffixes encountered. */
1289 int long_p = 0;
1290
1291 /* We have found a "L" or "U" suffix. */
1292 int found_suffix = 0;
1293
1294 ULONGEST high_bit;
1295 struct type *signed_type;
1296 struct type *unsigned_type;
1297
1298 if (parsed_float)
1299 {
1300 /* It's a float since it contains a point or an exponent. */
1301 char *s;
1302 int num; /* number of tokens scanned by scanf */
1303 char saved_char;
1304
1305 /* If it ends at "df", "dd" or "dl", take it as type of decimal floating
1306 point. Return DECFLOAT. */
1307
1308 if (len >= 2 && p[len - 2] == 'd' && p[len - 1] == 'f')
1309 {
1310 p[len - 2] = '\0';
1311 putithere->typed_val_decfloat.type
1312 = parse_type->builtin_decfloat;
1313 decimal_from_string (putithere->typed_val_decfloat.val, 4,
1314 gdbarch_byte_order (parse_gdbarch), p);
1315 p[len - 2] = 'd';
1316 return DECFLOAT;
1317 }
1318
1319 if (len >= 2 && p[len - 2] == 'd' && p[len - 1] == 'd')
1320 {
1321 p[len - 2] = '\0';
1322 putithere->typed_val_decfloat.type
1323 = parse_type->builtin_decdouble;
1324 decimal_from_string (putithere->typed_val_decfloat.val, 8,
1325 gdbarch_byte_order (parse_gdbarch), p);
1326 p[len - 2] = 'd';
1327 return DECFLOAT;
1328 }
1329
1330 if (len >= 2 && p[len - 2] == 'd' && p[len - 1] == 'l')
1331 {
1332 p[len - 2] = '\0';
1333 putithere->typed_val_decfloat.type
1334 = parse_type->builtin_declong;
1335 decimal_from_string (putithere->typed_val_decfloat.val, 16,
1336 gdbarch_byte_order (parse_gdbarch), p);
1337 p[len - 2] = 'd';
1338 return DECFLOAT;
1339 }
1340
1341 s = malloc (len);
1342 saved_char = p[len];
1343 p[len] = 0; /* null-terminate the token */
1344 num = sscanf (p, "%" DOUBLEST_SCAN_FORMAT "%s",
1345 &putithere->typed_val_float.dval, s);
1346 p[len] = saved_char; /* restore the input stream */
1347
1348 if (num == 1)
1349 putithere->typed_val_float.type =
1350 parse_type->builtin_double;
1351
1352 if (num == 2 )
1353 {
1354 /* See if it has any float suffix: 'f' for float, 'l' for long
1355 double. */
1356 if (!strcasecmp (s, "f"))
1357 putithere->typed_val_float.type =
1358 parse_type->builtin_float;
1359 else if (!strcasecmp (s, "l"))
1360 putithere->typed_val_float.type =
1361 parse_type->builtin_long_double;
1362 else
1363 {
1364 free (s);
1365 return ERROR;
1366 }
1367 }
1368
1369 free (s);
1370 return FLOAT;
1371 }
1372
1373 /* Handle base-switching prefixes 0x, 0t, 0d, 0 */
1374 if (p[0] == '0')
1375 switch (p[1])
1376 {
1377 case 'x':
1378 case 'X':
1379 if (len >= 3)
1380 {
1381 p += 2;
1382 base = 16;
1383 len -= 2;
1384 }
1385 break;
1386
1387 case 'b':
1388 case 'B':
1389 if (len >= 3)
1390 {
1391 p += 2;
1392 base = 2;
1393 len -= 2;
1394 }
1395 break;
1396
1397 case 't':
1398 case 'T':
1399 case 'd':
1400 case 'D':
1401 if (len >= 3)
1402 {
1403 p += 2;
1404 base = 10;
1405 len -= 2;
1406 }
1407 break;
1408
1409 default:
1410 base = 8;
1411 break;
1412 }
1413
1414 while (len-- > 0)
1415 {
1416 c = *p++;
1417 if (c >= 'A' && c <= 'Z')
1418 c += 'a' - 'A';
1419 if (c != 'l' && c != 'u')
1420 n *= base;
1421 if (c >= '0' && c <= '9')
1422 {
1423 if (found_suffix)
1424 return ERROR;
1425 n += i = c - '0';
1426 }
1427 else
1428 {
1429 if (base > 10 && c >= 'a' && c <= 'f')
1430 {
1431 if (found_suffix)
1432 return ERROR;
1433 n += i = c - 'a' + 10;
1434 }
1435 else if (c == 'l')
1436 {
1437 ++long_p;
1438 found_suffix = 1;
1439 }
1440 else if (c == 'u')
1441 {
1442 unsigned_p = 1;
1443 found_suffix = 1;
1444 }
1445 else
1446 return ERROR; /* Char not a digit */
1447 }
1448 if (i >= base)
1449 return ERROR; /* Invalid digit in this base */
1450
1451 /* Portably test for overflow (only works for nonzero values, so make
1452 a second check for zero). FIXME: Can't we just make n and prevn
1453 unsigned and avoid this? */
1454 if (c != 'l' && c != 'u' && (prevn >= n) && n != 0)
1455 unsigned_p = 1; /* Try something unsigned */
1456
1457 /* Portably test for unsigned overflow.
1458 FIXME: This check is wrong; for example it doesn't find overflow
1459 on 0x123456789 when LONGEST is 32 bits. */
1460 if (c != 'l' && c != 'u' && n != 0)
1461 {
1462 if ((unsigned_p && (ULONGEST) prevn >= (ULONGEST) n))
1463 error ("Numeric constant too large.");
1464 }
1465 prevn = n;
1466 }
1467
1468 /* An integer constant is an int, a long, or a long long. An L
1469 suffix forces it to be long; an LL suffix forces it to be long
1470 long. If not forced to a larger size, it gets the first type of
1471 the above that it fits in. To figure out whether it fits, we
1472 shift it right and see whether anything remains. Note that we
1473 can't shift sizeof (LONGEST) * HOST_CHAR_BIT bits or more in one
1474 operation, because many compilers will warn about such a shift
1475 (which always produces a zero result). Sometimes gdbarch_int_bit
1476 or gdbarch_long_bit will be that big, sometimes not. To deal with
1477 the case where it is we just always shift the value more than
1478 once, with fewer bits each time. */
1479
1480 un = (ULONGEST)n >> 2;
1481 if (long_p == 0
1482 && (un >> (gdbarch_int_bit (parse_gdbarch) - 2)) == 0)
1483 {
1484 high_bit = ((ULONGEST)1) << (gdbarch_int_bit (parse_gdbarch) - 1);
1485
1486 /* A large decimal (not hex or octal) constant (between INT_MAX
1487 and UINT_MAX) is a long or unsigned long, according to ANSI,
1488 never an unsigned int, but this code treats it as unsigned
1489 int. This probably should be fixed. GCC gives a warning on
1490 such constants. */
1491
1492 unsigned_type = parse_type->builtin_unsigned_int;
1493 signed_type = parse_type->builtin_int;
1494 }
1495 else if (long_p <= 1
1496 && (un >> (gdbarch_long_bit (parse_gdbarch) - 2)) == 0)
1497 {
1498 high_bit = ((ULONGEST)1) << (gdbarch_long_bit (parse_gdbarch) - 1);
1499 unsigned_type = parse_type->builtin_unsigned_long;
1500 signed_type = parse_type->builtin_long;
1501 }
1502 else
1503 {
1504 int shift;
1505 if (sizeof (ULONGEST) * HOST_CHAR_BIT
1506 < gdbarch_long_long_bit (parse_gdbarch))
1507 /* A long long does not fit in a LONGEST. */
1508 shift = (sizeof (ULONGEST) * HOST_CHAR_BIT - 1);
1509 else
1510 shift = (gdbarch_long_long_bit (parse_gdbarch) - 1);
1511 high_bit = (ULONGEST) 1 << shift;
1512 unsigned_type = parse_type->builtin_unsigned_long_long;
1513 signed_type = parse_type->builtin_long_long;
1514 }
1515
1516 putithere->typed_val_int.val = n;
1517
1518 /* If the high bit of the worked out type is set then this number
1519 has to be unsigned. */
1520
1521 if (unsigned_p || (n & high_bit))
1522 {
1523 putithere->typed_val_int.type = unsigned_type;
1524 }
1525 else
1526 {
1527 putithere->typed_val_int.type = signed_type;
1528 }
1529
1530 return INT;
1531 }
1532
1533 /* Temporary obstack used for holding strings. */
1534 static struct obstack tempbuf;
1535 static int tempbuf_init;
1536
1537 /* Parse a C escape sequence. The initial backslash of the sequence
1538 is at (*PTR)[-1]. *PTR will be updated to point to just after the
1539 last character of the sequence. If OUTPUT is not NULL, the
1540 translated form of the escape sequence will be written there. If
1541 OUTPUT is NULL, no output is written and the call will only affect
1542 *PTR. If an escape sequence is expressed in target bytes, then the
1543 entire sequence will simply be copied to OUTPUT. Return 1 if any
1544 character was emitted, 0 otherwise. */
1545
1546 int
1547 c_parse_escape (char **ptr, struct obstack *output)
1548 {
1549 char *tokptr = *ptr;
1550 int result = 1;
1551
1552 /* Some escape sequences undergo character set conversion. Those we
1553 translate here. */
1554 switch (*tokptr)
1555 {
1556 /* Hex escapes do not undergo character set conversion, so keep
1557 the escape sequence for later. */
1558 case 'x':
1559 if (output)
1560 obstack_grow_str (output, "\\x");
1561 ++tokptr;
1562 if (!isxdigit (*tokptr))
1563 error (_("\\x escape without a following hex digit"));
1564 while (isxdigit (*tokptr))
1565 {
1566 if (output)
1567 obstack_1grow (output, *tokptr);
1568 ++tokptr;
1569 }
1570 break;
1571
1572 /* Octal escapes do not undergo character set conversion, so
1573 keep the escape sequence for later. */
1574 case '0':
1575 case '1':
1576 case '2':
1577 case '3':
1578 case '4':
1579 case '5':
1580 case '6':
1581 case '7':
1582 {
1583 int i;
1584 if (output)
1585 obstack_grow_str (output, "\\");
1586 for (i = 0;
1587 i < 3 && isdigit (*tokptr) && *tokptr != '8' && *tokptr != '9';
1588 ++i)
1589 {
1590 if (output)
1591 obstack_1grow (output, *tokptr);
1592 ++tokptr;
1593 }
1594 }
1595 break;
1596
1597 /* We handle UCNs later. We could handle them here, but that
1598 would mean a spurious error in the case where the UCN could
1599 be converted to the target charset but not the host
1600 charset. */
1601 case 'u':
1602 case 'U':
1603 {
1604 char c = *tokptr;
1605 int i, len = c == 'U' ? 8 : 4;
1606 if (output)
1607 {
1608 obstack_1grow (output, '\\');
1609 obstack_1grow (output, *tokptr);
1610 }
1611 ++tokptr;
1612 if (!isxdigit (*tokptr))
1613 error (_("\\%c escape without a following hex digit"), c);
1614 for (i = 0; i < len && isxdigit (*tokptr); ++i)
1615 {
1616 if (output)
1617 obstack_1grow (output, *tokptr);
1618 ++tokptr;
1619 }
1620 }
1621 break;
1622
1623 /* We must pass backslash through so that it does not
1624 cause quoting during the second expansion. */
1625 case '\\':
1626 if (output)
1627 obstack_grow_str (output, "\\\\");
1628 ++tokptr;
1629 break;
1630
1631 /* Escapes which undergo conversion. */
1632 case 'a':
1633 if (output)
1634 obstack_1grow (output, '\a');
1635 ++tokptr;
1636 break;
1637 case 'b':
1638 if (output)
1639 obstack_1grow (output, '\b');
1640 ++tokptr;
1641 break;
1642 case 'f':
1643 if (output)
1644 obstack_1grow (output, '\f');
1645 ++tokptr;
1646 break;
1647 case 'n':
1648 if (output)
1649 obstack_1grow (output, '\n');
1650 ++tokptr;
1651 break;
1652 case 'r':
1653 if (output)
1654 obstack_1grow (output, '\r');
1655 ++tokptr;
1656 break;
1657 case 't':
1658 if (output)
1659 obstack_1grow (output, '\t');
1660 ++tokptr;
1661 break;
1662 case 'v':
1663 if (output)
1664 obstack_1grow (output, '\v');
1665 ++tokptr;
1666 break;
1667
1668 /* GCC extension. */
1669 case 'e':
1670 if (output)
1671 obstack_1grow (output, HOST_ESCAPE_CHAR);
1672 ++tokptr;
1673 break;
1674
1675 /* Backslash-newline expands to nothing at all. */
1676 case '\n':
1677 ++tokptr;
1678 result = 0;
1679 break;
1680
1681 /* A few escapes just expand to the character itself. */
1682 case '\'':
1683 case '\"':
1684 case '?':
1685 /* GCC extensions. */
1686 case '(':
1687 case '{':
1688 case '[':
1689 case '%':
1690 /* Unrecognized escapes turn into the character itself. */
1691 default:
1692 if (output)
1693 obstack_1grow (output, *tokptr);
1694 ++tokptr;
1695 break;
1696 }
1697 *ptr = tokptr;
1698 return result;
1699 }
1700
1701 /* Parse a string or character literal from TOKPTR. The string or
1702 character may be wide or unicode. *OUTPTR is set to just after the
1703 end of the literal in the input string. The resulting token is
1704 stored in VALUE. This returns a token value, either STRING or
1705 CHAR, depending on what was parsed. *HOST_CHARS is set to the
1706 number of host characters in the literal. */
1707 static int
1708 parse_string_or_char (char *tokptr, char **outptr, struct typed_stoken *value,
1709 int *host_chars)
1710 {
1711 int quote;
1712 enum c_string_type type;
1713
1714 /* Build the gdb internal form of the input string in tempbuf. Note
1715 that the buffer is null byte terminated *only* for the
1716 convenience of debugging gdb itself and printing the buffer
1717 contents when the buffer contains no embedded nulls. Gdb does
1718 not depend upon the buffer being null byte terminated, it uses
1719 the length string instead. This allows gdb to handle C strings
1720 (as well as strings in other languages) with embedded null
1721 bytes */
1722
1723 if (!tempbuf_init)
1724 tempbuf_init = 1;
1725 else
1726 obstack_free (&tempbuf, NULL);
1727 obstack_init (&tempbuf);
1728
1729 /* Record the string type. */
1730 if (*tokptr == 'L')
1731 {
1732 type = C_WIDE_STRING;
1733 ++tokptr;
1734 }
1735 else if (*tokptr == 'u')
1736 {
1737 type = C_STRING_16;
1738 ++tokptr;
1739 }
1740 else if (*tokptr == 'U')
1741 {
1742 type = C_STRING_32;
1743 ++tokptr;
1744 }
1745 else
1746 type = C_STRING;
1747
1748 /* Skip the quote. */
1749 quote = *tokptr;
1750 if (quote == '\'')
1751 type |= C_CHAR;
1752 ++tokptr;
1753
1754 *host_chars = 0;
1755
1756 while (*tokptr)
1757 {
1758 char c = *tokptr;
1759 if (c == '\\')
1760 {
1761 ++tokptr;
1762 *host_chars += c_parse_escape (&tokptr, &tempbuf);
1763 }
1764 else if (c == quote)
1765 break;
1766 else
1767 {
1768 obstack_1grow (&tempbuf, c);
1769 ++tokptr;
1770 /* FIXME: this does the wrong thing with multi-byte host
1771 characters. We could use mbrlen here, but that would
1772 make "set host-charset" a bit less useful. */
1773 ++*host_chars;
1774 }
1775 }
1776
1777 if (*tokptr != quote)
1778 {
1779 if (quote == '"')
1780 error ("Unterminated string in expression.");
1781 else
1782 error ("Unmatched single quote.");
1783 }
1784 ++tokptr;
1785
1786 value->type = type;
1787 value->ptr = obstack_base (&tempbuf);
1788 value->length = obstack_object_size (&tempbuf);
1789
1790 *outptr = tokptr;
1791
1792 return quote == '"' ? STRING : CHAR;
1793 }
1794
1795 struct token
1796 {
1797 char *operator;
1798 int token;
1799 enum exp_opcode opcode;
1800 int cxx_only;
1801 };
1802
1803 static const struct token tokentab3[] =
1804 {
1805 {">>=", ASSIGN_MODIFY, BINOP_RSH, 0},
1806 {"<<=", ASSIGN_MODIFY, BINOP_LSH, 0},
1807 {"->*", ARROW_STAR, BINOP_END, 1}
1808 };
1809
1810 static const struct token tokentab2[] =
1811 {
1812 {"+=", ASSIGN_MODIFY, BINOP_ADD, 0},
1813 {"-=", ASSIGN_MODIFY, BINOP_SUB, 0},
1814 {"*=", ASSIGN_MODIFY, BINOP_MUL, 0},
1815 {"/=", ASSIGN_MODIFY, BINOP_DIV, 0},
1816 {"%=", ASSIGN_MODIFY, BINOP_REM, 0},
1817 {"|=", ASSIGN_MODIFY, BINOP_BITWISE_IOR, 0},
1818 {"&=", ASSIGN_MODIFY, BINOP_BITWISE_AND, 0},
1819 {"^=", ASSIGN_MODIFY, BINOP_BITWISE_XOR, 0},
1820 {"++", INCREMENT, BINOP_END, 0},
1821 {"--", DECREMENT, BINOP_END, 0},
1822 {"->", ARROW, BINOP_END, 0},
1823 {"&&", ANDAND, BINOP_END, 0},
1824 {"||", OROR, BINOP_END, 0},
1825 /* "::" is *not* only C++: gdb overrides its meaning in several
1826 different ways, e.g., 'filename'::func, function::variable. */
1827 {"::", COLONCOLON, BINOP_END, 0},
1828 {"<<", LSH, BINOP_END, 0},
1829 {">>", RSH, BINOP_END, 0},
1830 {"==", EQUAL, BINOP_END, 0},
1831 {"!=", NOTEQUAL, BINOP_END, 0},
1832 {"<=", LEQ, BINOP_END, 0},
1833 {">=", GEQ, BINOP_END, 0},
1834 {".*", DOT_STAR, BINOP_END, 1}
1835 };
1836
1837 /* Identifier-like tokens. */
1838 static const struct token ident_tokens[] =
1839 {
1840 {"unsigned", UNSIGNED, OP_NULL, 0},
1841 {"template", TEMPLATE, OP_NULL, 1},
1842 {"volatile", VOLATILE_KEYWORD, OP_NULL, 0},
1843 {"struct", STRUCT, OP_NULL, 0},
1844 {"signed", SIGNED_KEYWORD, OP_NULL, 0},
1845 {"sizeof", SIZEOF, OP_NULL, 0},
1846 {"double", DOUBLE_KEYWORD, OP_NULL, 0},
1847 {"false", FALSEKEYWORD, OP_NULL, 1},
1848 {"class", CLASS, OP_NULL, 1},
1849 {"union", UNION, OP_NULL, 0},
1850 {"short", SHORT, OP_NULL, 0},
1851 {"const", CONST_KEYWORD, OP_NULL, 0},
1852 {"enum", ENUM, OP_NULL, 0},
1853 {"long", LONG, OP_NULL, 0},
1854 {"true", TRUEKEYWORD, OP_NULL, 1},
1855 {"int", INT_KEYWORD, OP_NULL, 0},
1856 {"new", NEW, OP_NULL, 1},
1857 {"delete", DELETE, OP_NULL, 1},
1858 {"operator", OPERATOR, OP_NULL, 1},
1859
1860 {"and", ANDAND, BINOP_END, 1},
1861 {"and_eq", ASSIGN_MODIFY, BINOP_BITWISE_AND, 1},
1862 {"bitand", '&', OP_NULL, 1},
1863 {"bitor", '|', OP_NULL, 1},
1864 {"compl", '~', OP_NULL, 1},
1865 {"not", '!', OP_NULL, 1},
1866 {"not_eq", NOTEQUAL, BINOP_END, 1},
1867 {"or", OROR, BINOP_END, 1},
1868 {"or_eq", ASSIGN_MODIFY, BINOP_BITWISE_IOR, 1},
1869 {"xor", '^', OP_NULL, 1},
1870 {"xor_eq", ASSIGN_MODIFY, BINOP_BITWISE_XOR, 1},
1871
1872 {"const_cast", CONST_CAST, OP_NULL, 1 },
1873 {"dynamic_cast", DYNAMIC_CAST, OP_NULL, 1 },
1874 {"static_cast", STATIC_CAST, OP_NULL, 1 },
1875 {"reinterpret_cast", REINTERPRET_CAST, OP_NULL, 1 }
1876 };
1877
1878 /* When we find that lexptr (the global var defined in parse.c) is
1879 pointing at a macro invocation, we expand the invocation, and call
1880 scan_macro_expansion to save the old lexptr here and point lexptr
1881 into the expanded text. When we reach the end of that, we call
1882 end_macro_expansion to pop back to the value we saved here. The
1883 macro expansion code promises to return only fully-expanded text,
1884 so we don't need to "push" more than one level.
1885
1886 This is disgusting, of course. It would be cleaner to do all macro
1887 expansion beforehand, and then hand that to lexptr. But we don't
1888 really know where the expression ends. Remember, in a command like
1889
1890 (gdb) break *ADDRESS if CONDITION
1891
1892 we evaluate ADDRESS in the scope of the current frame, but we
1893 evaluate CONDITION in the scope of the breakpoint's location. So
1894 it's simply wrong to try to macro-expand the whole thing at once. */
1895 static char *macro_original_text;
1896
1897 /* We save all intermediate macro expansions on this obstack for the
1898 duration of a single parse. The expansion text may sometimes have
1899 to live past the end of the expansion, due to yacc lookahead.
1900 Rather than try to be clever about saving the data for a single
1901 token, we simply keep it all and delete it after parsing has
1902 completed. */
1903 static struct obstack expansion_obstack;
1904
1905 static void
1906 scan_macro_expansion (char *expansion)
1907 {
1908 char *copy;
1909
1910 /* We'd better not be trying to push the stack twice. */
1911 gdb_assert (! macro_original_text);
1912
1913 /* Copy to the obstack, and then free the intermediate
1914 expansion. */
1915 copy = obstack_copy0 (&expansion_obstack, expansion, strlen (expansion));
1916 xfree (expansion);
1917
1918 /* Save the old lexptr value, so we can return to it when we're done
1919 parsing the expanded text. */
1920 macro_original_text = lexptr;
1921 lexptr = copy;
1922 }
1923
1924
1925 static int
1926 scanning_macro_expansion (void)
1927 {
1928 return macro_original_text != 0;
1929 }
1930
1931
1932 static void
1933 finished_macro_expansion (void)
1934 {
1935 /* There'd better be something to pop back to. */
1936 gdb_assert (macro_original_text);
1937
1938 /* Pop back to the original text. */
1939 lexptr = macro_original_text;
1940 macro_original_text = 0;
1941 }
1942
1943
1944 static void
1945 scan_macro_cleanup (void *dummy)
1946 {
1947 if (macro_original_text)
1948 finished_macro_expansion ();
1949
1950 obstack_free (&expansion_obstack, NULL);
1951 }
1952
1953 /* Return true iff the token represents a C++ cast operator. */
1954
1955 static int
1956 is_cast_operator (const char *token, int len)
1957 {
1958 return (! strncmp (token, "dynamic_cast", len)
1959 || ! strncmp (token, "static_cast", len)
1960 || ! strncmp (token, "reinterpret_cast", len)
1961 || ! strncmp (token, "const_cast", len));
1962 }
1963
1964 /* The scope used for macro expansion. */
1965 static struct macro_scope *expression_macro_scope;
1966
1967 /* This is set if a NAME token appeared at the very end of the input
1968 string, with no whitespace separating the name from the EOF. This
1969 is used only when parsing to do field name completion. */
1970 static int saw_name_at_eof;
1971
1972 /* This is set if the previously-returned token was a structure
1973 operator -- either '.' or ARROW. This is used only when parsing to
1974 do field name completion. */
1975 static int last_was_structop;
1976
1977 /* Read one token, getting characters through lexptr. */
1978
1979 static int
1980 lex_one_token (void)
1981 {
1982 int c;
1983 int namelen;
1984 unsigned int i;
1985 char *tokstart;
1986 int saw_structop = last_was_structop;
1987 char *copy;
1988
1989 last_was_structop = 0;
1990
1991 retry:
1992
1993 /* Check if this is a macro invocation that we need to expand. */
1994 if (! scanning_macro_expansion ())
1995 {
1996 char *expanded = macro_expand_next (&lexptr,
1997 standard_macro_lookup,
1998 expression_macro_scope);
1999
2000 if (expanded)
2001 scan_macro_expansion (expanded);
2002 }
2003
2004 prev_lexptr = lexptr;
2005
2006 tokstart = lexptr;
2007 /* See if it is a special token of length 3. */
2008 for (i = 0; i < sizeof tokentab3 / sizeof tokentab3[0]; i++)
2009 if (strncmp (tokstart, tokentab3[i].operator, 3) == 0)
2010 {
2011 if (tokentab3[i].cxx_only
2012 && parse_language->la_language != language_cplus)
2013 break;
2014
2015 lexptr += 3;
2016 yylval.opcode = tokentab3[i].opcode;
2017 return tokentab3[i].token;
2018 }
2019
2020 /* See if it is a special token of length 2. */
2021 for (i = 0; i < sizeof tokentab2 / sizeof tokentab2[0]; i++)
2022 if (strncmp (tokstart, tokentab2[i].operator, 2) == 0)
2023 {
2024 if (tokentab2[i].cxx_only
2025 && parse_language->la_language != language_cplus)
2026 break;
2027
2028 lexptr += 2;
2029 yylval.opcode = tokentab2[i].opcode;
2030 if (in_parse_field && tokentab2[i].token == ARROW)
2031 last_was_structop = 1;
2032 return tokentab2[i].token;
2033 }
2034
2035 switch (c = *tokstart)
2036 {
2037 case 0:
2038 /* If we were just scanning the result of a macro expansion,
2039 then we need to resume scanning the original text.
2040 If we're parsing for field name completion, and the previous
2041 token allows such completion, return a COMPLETE token.
2042 Otherwise, we were already scanning the original text, and
2043 we're really done. */
2044 if (scanning_macro_expansion ())
2045 {
2046 finished_macro_expansion ();
2047 goto retry;
2048 }
2049 else if (saw_name_at_eof)
2050 {
2051 saw_name_at_eof = 0;
2052 return COMPLETE;
2053 }
2054 else if (saw_structop)
2055 return COMPLETE;
2056 else
2057 return 0;
2058
2059 case ' ':
2060 case '\t':
2061 case '\n':
2062 lexptr++;
2063 goto retry;
2064
2065 case '[':
2066 case '(':
2067 paren_depth++;
2068 lexptr++;
2069 return c;
2070
2071 case ']':
2072 case ')':
2073 if (paren_depth == 0)
2074 return 0;
2075 paren_depth--;
2076 lexptr++;
2077 return c;
2078
2079 case ',':
2080 if (comma_terminates
2081 && paren_depth == 0
2082 && ! scanning_macro_expansion ())
2083 return 0;
2084 lexptr++;
2085 return c;
2086
2087 case '.':
2088 /* Might be a floating point number. */
2089 if (lexptr[1] < '0' || lexptr[1] > '9')
2090 {
2091 if (in_parse_field)
2092 last_was_structop = 1;
2093 goto symbol; /* Nope, must be a symbol. */
2094 }
2095 /* FALL THRU into number case. */
2096
2097 case '0':
2098 case '1':
2099 case '2':
2100 case '3':
2101 case '4':
2102 case '5':
2103 case '6':
2104 case '7':
2105 case '8':
2106 case '9':
2107 {
2108 /* It's a number. */
2109 int got_dot = 0, got_e = 0, toktype;
2110 char *p = tokstart;
2111 int hex = input_radix > 10;
2112
2113 if (c == '0' && (p[1] == 'x' || p[1] == 'X'))
2114 {
2115 p += 2;
2116 hex = 1;
2117 }
2118 else if (c == '0' && (p[1]=='t' || p[1]=='T' || p[1]=='d' || p[1]=='D'))
2119 {
2120 p += 2;
2121 hex = 0;
2122 }
2123
2124 for (;; ++p)
2125 {
2126 /* This test includes !hex because 'e' is a valid hex digit
2127 and thus does not indicate a floating point number when
2128 the radix is hex. */
2129 if (!hex && !got_e && (*p == 'e' || *p == 'E'))
2130 got_dot = got_e = 1;
2131 /* This test does not include !hex, because a '.' always indicates
2132 a decimal floating point number regardless of the radix. */
2133 else if (!got_dot && *p == '.')
2134 got_dot = 1;
2135 else if (got_e && (p[-1] == 'e' || p[-1] == 'E')
2136 && (*p == '-' || *p == '+'))
2137 /* This is the sign of the exponent, not the end of the
2138 number. */
2139 continue;
2140 /* We will take any letters or digits. parse_number will
2141 complain if past the radix, or if L or U are not final. */
2142 else if ((*p < '0' || *p > '9')
2143 && ((*p < 'a' || *p > 'z')
2144 && (*p < 'A' || *p > 'Z')))
2145 break;
2146 }
2147 toktype = parse_number (tokstart, p - tokstart, got_dot|got_e, &yylval);
2148 if (toktype == ERROR)
2149 {
2150 char *err_copy = (char *) alloca (p - tokstart + 1);
2151
2152 memcpy (err_copy, tokstart, p - tokstart);
2153 err_copy[p - tokstart] = 0;
2154 error ("Invalid number \"%s\".", err_copy);
2155 }
2156 lexptr = p;
2157 return toktype;
2158 }
2159
2160 case '+':
2161 case '-':
2162 case '*':
2163 case '/':
2164 case '%':
2165 case '|':
2166 case '&':
2167 case '^':
2168 case '~':
2169 case '!':
2170 case '@':
2171 case '<':
2172 case '>':
2173 case '?':
2174 case ':':
2175 case '=':
2176 case '{':
2177 case '}':
2178 symbol:
2179 lexptr++;
2180 return c;
2181
2182 case 'L':
2183 case 'u':
2184 case 'U':
2185 if (tokstart[1] != '"' && tokstart[1] != '\'')
2186 break;
2187 /* Fall through. */
2188 case '\'':
2189 case '"':
2190 {
2191 int host_len;
2192 int result = parse_string_or_char (tokstart, &lexptr, &yylval.tsval,
2193 &host_len);
2194 if (result == CHAR)
2195 {
2196 if (host_len == 0)
2197 error ("Empty character constant.");
2198 else if (host_len > 2 && c == '\'')
2199 {
2200 ++tokstart;
2201 namelen = lexptr - tokstart - 1;
2202 goto tryname;
2203 }
2204 else if (host_len > 1)
2205 error ("Invalid character constant.");
2206 }
2207 return result;
2208 }
2209 }
2210
2211 if (!(c == '_' || c == '$'
2212 || (c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z')))
2213 /* We must have come across a bad character (e.g. ';'). */
2214 error ("Invalid character '%c' in expression.", c);
2215
2216 /* It's a name. See how long it is. */
2217 namelen = 0;
2218 for (c = tokstart[namelen];
2219 (c == '_' || c == '$' || (c >= '0' && c <= '9')
2220 || (c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z') || c == '<');)
2221 {
2222 /* Template parameter lists are part of the name.
2223 FIXME: This mishandles `print $a<4&&$a>3'. */
2224
2225 if (c == '<')
2226 {
2227 if (! is_cast_operator (tokstart, namelen))
2228 {
2229 /* Scan ahead to get rest of the template specification. Note
2230 that we look ahead only when the '<' adjoins non-whitespace
2231 characters; for comparison expressions, e.g. "a < b > c",
2232 there must be spaces before the '<', etc. */
2233
2234 char * p = find_template_name_end (tokstart + namelen);
2235 if (p)
2236 namelen = p - tokstart;
2237 }
2238 break;
2239 }
2240 c = tokstart[++namelen];
2241 }
2242
2243 /* The token "if" terminates the expression and is NOT removed from
2244 the input stream. It doesn't count if it appears in the
2245 expansion of a macro. */
2246 if (namelen == 2
2247 && tokstart[0] == 'i'
2248 && tokstart[1] == 'f'
2249 && ! scanning_macro_expansion ())
2250 {
2251 return 0;
2252 }
2253
2254 /* For the same reason (breakpoint conditions), "thread N"
2255 terminates the expression. "thread" could be an identifier, but
2256 an identifier is never followed by a number without intervening
2257 punctuation. "task" is similar. Handle abbreviations of these,
2258 similarly to breakpoint.c:find_condition_and_thread. */
2259 if (namelen >= 1
2260 && (strncmp (tokstart, "thread", namelen) == 0
2261 || strncmp (tokstart, "task", namelen) == 0)
2262 && (tokstart[namelen] == ' ' || tokstart[namelen] == '\t')
2263 && ! scanning_macro_expansion ())
2264 {
2265 char *p = tokstart + namelen + 1;
2266 while (*p == ' ' || *p == '\t')
2267 p++;
2268 if (*p >= '0' && *p <= '9')
2269 return 0;
2270 }
2271
2272 lexptr += namelen;
2273
2274 tryname:
2275
2276 yylval.sval.ptr = tokstart;
2277 yylval.sval.length = namelen;
2278
2279 /* Catch specific keywords. */
2280 copy = copy_name (yylval.sval);
2281 for (i = 0; i < sizeof ident_tokens / sizeof ident_tokens[0]; i++)
2282 if (strcmp (copy, ident_tokens[i].operator) == 0)
2283 {
2284 if (ident_tokens[i].cxx_only
2285 && parse_language->la_language != language_cplus)
2286 break;
2287
2288 /* It is ok to always set this, even though we don't always
2289 strictly need to. */
2290 yylval.opcode = ident_tokens[i].opcode;
2291 return ident_tokens[i].token;
2292 }
2293
2294 if (*tokstart == '$')
2295 return VARIABLE;
2296
2297 if (in_parse_field && *lexptr == '\0')
2298 saw_name_at_eof = 1;
2299 return NAME;
2300 }
2301
2302 /* An object of this type is pushed on a FIFO by the "outer" lexer. */
2303 typedef struct
2304 {
2305 int token;
2306 YYSTYPE value;
2307 } token_and_value;
2308
2309 DEF_VEC_O (token_and_value);
2310
2311 /* A FIFO of tokens that have been read but not yet returned to the
2312 parser. */
2313 static VEC (token_and_value) *token_fifo;
2314
2315 /* Non-zero if the lexer should return tokens from the FIFO. */
2316 static int popping;
2317
2318 /* Temporary storage for c_lex; this holds symbol names as they are
2319 built up. */
2320 static struct obstack name_obstack;
2321
2322 /* Classify a NAME token. The contents of the token are in `yylval'.
2323 Updates yylval and returns the new token type. BLOCK is the block
2324 in which lookups start; this can be NULL to mean the global
2325 scope. */
2326 static int
2327 classify_name (struct block *block)
2328 {
2329 struct symbol *sym;
2330 char *copy;
2331 int is_a_field_of_this = 0;
2332
2333 copy = copy_name (yylval.sval);
2334
2335 sym = lookup_symbol (copy, block, VAR_DOMAIN,
2336 parse_language->la_language == language_cplus
2337 ? &is_a_field_of_this : (int *) NULL);
2338
2339 if (sym && SYMBOL_CLASS (sym) == LOC_BLOCK)
2340 {
2341 yylval.ssym.sym = sym;
2342 yylval.ssym.is_a_field_of_this = is_a_field_of_this;
2343 return BLOCKNAME;
2344 }
2345 else if (!sym)
2346 {
2347 /* See if it's a file name. */
2348 struct symtab *symtab;
2349
2350 symtab = lookup_symtab (copy);
2351 if (symtab)
2352 {
2353 yylval.bval = BLOCKVECTOR_BLOCK (BLOCKVECTOR (symtab), STATIC_BLOCK);
2354 return FILENAME;
2355 }
2356 }
2357
2358 if (sym && SYMBOL_CLASS (sym) == LOC_TYPEDEF)
2359 {
2360 yylval.tsym.type = SYMBOL_TYPE (sym);
2361 return TYPENAME;
2362 }
2363
2364 yylval.tsym.type
2365 = language_lookup_primitive_type_by_name (parse_language,
2366 parse_gdbarch, copy);
2367 if (yylval.tsym.type != NULL)
2368 return TYPENAME;
2369
2370 /* Input names that aren't symbols but ARE valid hex numbers, when
2371 the input radix permits them, can be names or numbers depending
2372 on the parse. Note we support radixes > 16 here. */
2373 if (!sym
2374 && ((copy[0] >= 'a' && copy[0] < 'a' + input_radix - 10)
2375 || (copy[0] >= 'A' && copy[0] < 'A' + input_radix - 10)))
2376 {
2377 YYSTYPE newlval; /* Its value is ignored. */
2378 int hextype = parse_number (copy, yylval.sval.length, 0, &newlval);
2379 if (hextype == INT)
2380 {
2381 yylval.ssym.sym = sym;
2382 yylval.ssym.is_a_field_of_this = is_a_field_of_this;
2383 return NAME_OR_INT;
2384 }
2385 }
2386
2387 /* Any other kind of symbol */
2388 yylval.ssym.sym = sym;
2389 yylval.ssym.is_a_field_of_this = is_a_field_of_this;
2390 return NAME;
2391 }
2392
2393 /* Like classify_name, but used by the inner loop of the lexer, when a
2394 name might have already been seen. FIRST_NAME is true if the token
2395 in `yylval' is the first component of a name, false otherwise. If
2396 this function returns NAME, it might not have updated `yylval'.
2397 This is ok because the caller only cares about TYPENAME. */
2398 static int
2399 classify_inner_name (struct block *block, int first_name)
2400 {
2401 struct type *type, *new_type;
2402 char *copy;
2403
2404 if (first_name)
2405 return classify_name (block);
2406
2407 type = check_typedef (yylval.tsym.type);
2408 if (TYPE_CODE (type) != TYPE_CODE_STRUCT
2409 && TYPE_CODE (type) != TYPE_CODE_UNION
2410 && TYPE_CODE (type) != TYPE_CODE_NAMESPACE)
2411 /* We know the caller won't expect us to update yylval. */
2412 return NAME;
2413
2414 copy = copy_name (yylval.tsym.stoken);
2415 new_type = cp_lookup_nested_type (type, copy, block);
2416
2417 if (new_type == NULL)
2418 /* We know the caller won't expect us to update yylval. */
2419 return NAME;
2420
2421 yylval.tsym.type = new_type;
2422 return TYPENAME;
2423 }
2424
2425 /* The outer level of a two-level lexer. This calls the inner lexer
2426 to return tokens. It then either returns these tokens, or
2427 aggregates them into a larger token. This lets us work around a
2428 problem in our parsing approach, where the parser could not
2429 distinguish between qualified names and qualified types at the
2430 right point.
2431
2432 This approach is still not ideal, because it mishandles template
2433 types. See the comment in lex_one_token for an example. However,
2434 this is still an improvement over the earlier approach, and will
2435 suffice until we move to better parsing technology. */
2436 static int
2437 yylex (void)
2438 {
2439 token_and_value current;
2440 int first_was_coloncolon, last_was_coloncolon, first_iter;
2441
2442 if (popping && !VEC_empty (token_and_value, token_fifo))
2443 {
2444 token_and_value tv = *VEC_index (token_and_value, token_fifo, 0);
2445 VEC_ordered_remove (token_and_value, token_fifo, 0);
2446 yylval = tv.value;
2447 return tv.token;
2448 }
2449 popping = 0;
2450
2451 current.token = lex_one_token ();
2452 if (current.token == NAME)
2453 current.token = classify_name (expression_context_block);
2454 if (parse_language->la_language != language_cplus
2455 || (current.token != TYPENAME && current.token != COLONCOLON))
2456 return current.token;
2457
2458 first_was_coloncolon = current.token == COLONCOLON;
2459 last_was_coloncolon = first_was_coloncolon;
2460 obstack_free (&name_obstack, obstack_base (&name_obstack));
2461 if (!last_was_coloncolon)
2462 obstack_grow (&name_obstack, yylval.sval.ptr, yylval.sval.length);
2463 current.value = yylval;
2464 first_iter = 1;
2465 while (1)
2466 {
2467 token_and_value next;
2468
2469 next.token = lex_one_token ();
2470 next.value = yylval;
2471
2472 if (next.token == NAME && last_was_coloncolon)
2473 {
2474 int classification;
2475
2476 classification = classify_inner_name (first_was_coloncolon
2477 ? NULL
2478 : expression_context_block,
2479 first_iter);
2480 /* We keep going until we either run out of names, or until
2481 we have a qualified name which is not a type. */
2482 if (classification != TYPENAME)
2483 {
2484 /* Push the final component and leave the loop. */
2485 VEC_safe_push (token_and_value, token_fifo, &next);
2486 break;
2487 }
2488
2489 /* Update the partial name we are constructing. */
2490 if (!first_iter)
2491 {
2492 /* We don't want to put a leading "::" into the name. */
2493 obstack_grow_str (&name_obstack, "::");
2494 }
2495 obstack_grow (&name_obstack, next.value.sval.ptr,
2496 next.value.sval.length);
2497
2498 yylval.sval.ptr = obstack_base (&name_obstack);
2499 yylval.sval.length = obstack_object_size (&name_obstack);
2500 current.value = yylval;
2501 current.token = classification;
2502
2503 last_was_coloncolon = 0;
2504 }
2505 else if (next.token == COLONCOLON && !last_was_coloncolon)
2506 last_was_coloncolon = 1;
2507 else
2508 {
2509 /* We've reached the end of the name. */
2510 VEC_safe_push (token_and_value, token_fifo, &next);
2511 break;
2512 }
2513
2514 first_iter = 0;
2515 }
2516
2517 popping = 1;
2518
2519 /* If we ended with a "::", insert it too. */
2520 if (last_was_coloncolon)
2521 {
2522 token_and_value cc;
2523 memset (&cc, 0, sizeof (token_and_value));
2524 if (first_was_coloncolon)
2525 {
2526 yylval = cc.value;
2527 return COLONCOLON;
2528 }
2529 cc.token = COLONCOLON;
2530 VEC_safe_insert (token_and_value, token_fifo, 0, &cc);
2531 }
2532
2533 yylval = current.value;
2534 yylval.sval.ptr = obstack_copy0 (&expansion_obstack,
2535 yylval.sval.ptr,
2536 yylval.sval.length);
2537 return current.token;
2538 }
2539
2540 int
2541 c_parse (void)
2542 {
2543 int result;
2544 struct cleanup *back_to = make_cleanup (free_current_contents,
2545 &expression_macro_scope);
2546
2547 /* Set up the scope for macro expansion. */
2548 expression_macro_scope = NULL;
2549
2550 if (expression_context_block)
2551 expression_macro_scope
2552 = sal_macro_scope (find_pc_line (expression_context_pc, 0));
2553 else
2554 expression_macro_scope = default_macro_scope ();
2555 if (! expression_macro_scope)
2556 expression_macro_scope = user_macro_scope ();
2557
2558 /* Initialize macro expansion code. */
2559 obstack_init (&expansion_obstack);
2560 gdb_assert (! macro_original_text);
2561 make_cleanup (scan_macro_cleanup, 0);
2562
2563 make_cleanup_restore_integer (&yydebug);
2564 yydebug = parser_debug;
2565
2566 /* Initialize some state used by the lexer. */
2567 last_was_structop = 0;
2568 saw_name_at_eof = 0;
2569
2570 VEC_free (token_and_value, token_fifo);
2571 popping = 0;
2572 obstack_init (&name_obstack);
2573 make_cleanup_obstack_free (&name_obstack);
2574
2575 result = yyparse ();
2576 do_cleanups (back_to);
2577 return result;
2578 }
2579
2580
2581 void
2582 yyerror (char *msg)
2583 {
2584 if (prev_lexptr)
2585 lexptr = prev_lexptr;
2586
2587 error ("A %s in expression, near `%s'.", (msg ? msg : "error"), lexptr);
2588 }
This page took 0.126017 seconds and 3 git commands to generate.