*** empty log message ***
[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 const char *suffix;
1423 int suffix_len;
1424
1425 /* If it ends at "df", "dd" or "dl", take it as type of decimal floating
1426 point. Return DECFLOAT. */
1427
1428 if (len >= 2 && p[len - 2] == 'd' && p[len - 1] == 'f')
1429 {
1430 p[len - 2] = '\0';
1431 putithere->typed_val_decfloat.type
1432 = parse_type->builtin_decfloat;
1433 decimal_from_string (putithere->typed_val_decfloat.val, 4,
1434 gdbarch_byte_order (parse_gdbarch), p);
1435 p[len - 2] = 'd';
1436 return DECFLOAT;
1437 }
1438
1439 if (len >= 2 && p[len - 2] == 'd' && p[len - 1] == 'd')
1440 {
1441 p[len - 2] = '\0';
1442 putithere->typed_val_decfloat.type
1443 = parse_type->builtin_decdouble;
1444 decimal_from_string (putithere->typed_val_decfloat.val, 8,
1445 gdbarch_byte_order (parse_gdbarch), p);
1446 p[len - 2] = 'd';
1447 return DECFLOAT;
1448 }
1449
1450 if (len >= 2 && p[len - 2] == 'd' && p[len - 1] == 'l')
1451 {
1452 p[len - 2] = '\0';
1453 putithere->typed_val_decfloat.type
1454 = parse_type->builtin_declong;
1455 decimal_from_string (putithere->typed_val_decfloat.val, 16,
1456 gdbarch_byte_order (parse_gdbarch), p);
1457 p[len - 2] = 'd';
1458 return DECFLOAT;
1459 }
1460
1461 if (! parse_c_float (parse_gdbarch, p, len,
1462 &putithere->typed_val_float.dval,
1463 &putithere->typed_val_float.type))
1464 return ERROR;
1465 return FLOAT;
1466 }
1467
1468 /* Handle base-switching prefixes 0x, 0t, 0d, 0 */
1469 if (p[0] == '0')
1470 switch (p[1])
1471 {
1472 case 'x':
1473 case 'X':
1474 if (len >= 3)
1475 {
1476 p += 2;
1477 base = 16;
1478 len -= 2;
1479 }
1480 break;
1481
1482 case 'b':
1483 case 'B':
1484 if (len >= 3)
1485 {
1486 p += 2;
1487 base = 2;
1488 len -= 2;
1489 }
1490 break;
1491
1492 case 't':
1493 case 'T':
1494 case 'd':
1495 case 'D':
1496 if (len >= 3)
1497 {
1498 p += 2;
1499 base = 10;
1500 len -= 2;
1501 }
1502 break;
1503
1504 default:
1505 base = 8;
1506 break;
1507 }
1508
1509 while (len-- > 0)
1510 {
1511 c = *p++;
1512 if (c >= 'A' && c <= 'Z')
1513 c += 'a' - 'A';
1514 if (c != 'l' && c != 'u')
1515 n *= base;
1516 if (c >= '0' && c <= '9')
1517 {
1518 if (found_suffix)
1519 return ERROR;
1520 n += i = c - '0';
1521 }
1522 else
1523 {
1524 if (base > 10 && c >= 'a' && c <= 'f')
1525 {
1526 if (found_suffix)
1527 return ERROR;
1528 n += i = c - 'a' + 10;
1529 }
1530 else if (c == 'l')
1531 {
1532 ++long_p;
1533 found_suffix = 1;
1534 }
1535 else if (c == 'u')
1536 {
1537 unsigned_p = 1;
1538 found_suffix = 1;
1539 }
1540 else
1541 return ERROR; /* Char not a digit */
1542 }
1543 if (i >= base)
1544 return ERROR; /* Invalid digit in this base */
1545
1546 /* Portably test for overflow (only works for nonzero values, so make
1547 a second check for zero). FIXME: Can't we just make n and prevn
1548 unsigned and avoid this? */
1549 if (c != 'l' && c != 'u' && (prevn >= n) && n != 0)
1550 unsigned_p = 1; /* Try something unsigned */
1551
1552 /* Portably test for unsigned overflow.
1553 FIXME: This check is wrong; for example it doesn't find overflow
1554 on 0x123456789 when LONGEST is 32 bits. */
1555 if (c != 'l' && c != 'u' && n != 0)
1556 {
1557 if ((unsigned_p && (ULONGEST) prevn >= (ULONGEST) n))
1558 error (_("Numeric constant too large."));
1559 }
1560 prevn = n;
1561 }
1562
1563 /* An integer constant is an int, a long, or a long long. An L
1564 suffix forces it to be long; an LL suffix forces it to be long
1565 long. If not forced to a larger size, it gets the first type of
1566 the above that it fits in. To figure out whether it fits, we
1567 shift it right and see whether anything remains. Note that we
1568 can't shift sizeof (LONGEST) * HOST_CHAR_BIT bits or more in one
1569 operation, because many compilers will warn about such a shift
1570 (which always produces a zero result). Sometimes gdbarch_int_bit
1571 or gdbarch_long_bit will be that big, sometimes not. To deal with
1572 the case where it is we just always shift the value more than
1573 once, with fewer bits each time. */
1574
1575 un = (ULONGEST)n >> 2;
1576 if (long_p == 0
1577 && (un >> (gdbarch_int_bit (parse_gdbarch) - 2)) == 0)
1578 {
1579 high_bit = ((ULONGEST)1) << (gdbarch_int_bit (parse_gdbarch) - 1);
1580
1581 /* A large decimal (not hex or octal) constant (between INT_MAX
1582 and UINT_MAX) is a long or unsigned long, according to ANSI,
1583 never an unsigned int, but this code treats it as unsigned
1584 int. This probably should be fixed. GCC gives a warning on
1585 such constants. */
1586
1587 unsigned_type = parse_type->builtin_unsigned_int;
1588 signed_type = parse_type->builtin_int;
1589 }
1590 else if (long_p <= 1
1591 && (un >> (gdbarch_long_bit (parse_gdbarch) - 2)) == 0)
1592 {
1593 high_bit = ((ULONGEST)1) << (gdbarch_long_bit (parse_gdbarch) - 1);
1594 unsigned_type = parse_type->builtin_unsigned_long;
1595 signed_type = parse_type->builtin_long;
1596 }
1597 else
1598 {
1599 int shift;
1600 if (sizeof (ULONGEST) * HOST_CHAR_BIT
1601 < gdbarch_long_long_bit (parse_gdbarch))
1602 /* A long long does not fit in a LONGEST. */
1603 shift = (sizeof (ULONGEST) * HOST_CHAR_BIT - 1);
1604 else
1605 shift = (gdbarch_long_long_bit (parse_gdbarch) - 1);
1606 high_bit = (ULONGEST) 1 << shift;
1607 unsigned_type = parse_type->builtin_unsigned_long_long;
1608 signed_type = parse_type->builtin_long_long;
1609 }
1610
1611 putithere->typed_val_int.val = n;
1612
1613 /* If the high bit of the worked out type is set then this number
1614 has to be unsigned. */
1615
1616 if (unsigned_p || (n & high_bit))
1617 {
1618 putithere->typed_val_int.type = unsigned_type;
1619 }
1620 else
1621 {
1622 putithere->typed_val_int.type = signed_type;
1623 }
1624
1625 return INT;
1626 }
1627
1628 /* Temporary obstack used for holding strings. */
1629 static struct obstack tempbuf;
1630 static int tempbuf_init;
1631
1632 /* Parse a C escape sequence. The initial backslash of the sequence
1633 is at (*PTR)[-1]. *PTR will be updated to point to just after the
1634 last character of the sequence. If OUTPUT is not NULL, the
1635 translated form of the escape sequence will be written there. If
1636 OUTPUT is NULL, no output is written and the call will only affect
1637 *PTR. If an escape sequence is expressed in target bytes, then the
1638 entire sequence will simply be copied to OUTPUT. Return 1 if any
1639 character was emitted, 0 otherwise. */
1640
1641 int
1642 c_parse_escape (char **ptr, struct obstack *output)
1643 {
1644 char *tokptr = *ptr;
1645 int result = 1;
1646
1647 /* Some escape sequences undergo character set conversion. Those we
1648 translate here. */
1649 switch (*tokptr)
1650 {
1651 /* Hex escapes do not undergo character set conversion, so keep
1652 the escape sequence for later. */
1653 case 'x':
1654 if (output)
1655 obstack_grow_str (output, "\\x");
1656 ++tokptr;
1657 if (!isxdigit (*tokptr))
1658 error (_("\\x escape without a following hex digit"));
1659 while (isxdigit (*tokptr))
1660 {
1661 if (output)
1662 obstack_1grow (output, *tokptr);
1663 ++tokptr;
1664 }
1665 break;
1666
1667 /* Octal escapes do not undergo character set conversion, so
1668 keep the escape sequence for later. */
1669 case '0':
1670 case '1':
1671 case '2':
1672 case '3':
1673 case '4':
1674 case '5':
1675 case '6':
1676 case '7':
1677 {
1678 int i;
1679 if (output)
1680 obstack_grow_str (output, "\\");
1681 for (i = 0;
1682 i < 3 && isdigit (*tokptr) && *tokptr != '8' && *tokptr != '9';
1683 ++i)
1684 {
1685 if (output)
1686 obstack_1grow (output, *tokptr);
1687 ++tokptr;
1688 }
1689 }
1690 break;
1691
1692 /* We handle UCNs later. We could handle them here, but that
1693 would mean a spurious error in the case where the UCN could
1694 be converted to the target charset but not the host
1695 charset. */
1696 case 'u':
1697 case 'U':
1698 {
1699 char c = *tokptr;
1700 int i, len = c == 'U' ? 8 : 4;
1701 if (output)
1702 {
1703 obstack_1grow (output, '\\');
1704 obstack_1grow (output, *tokptr);
1705 }
1706 ++tokptr;
1707 if (!isxdigit (*tokptr))
1708 error (_("\\%c escape without a following hex digit"), c);
1709 for (i = 0; i < len && isxdigit (*tokptr); ++i)
1710 {
1711 if (output)
1712 obstack_1grow (output, *tokptr);
1713 ++tokptr;
1714 }
1715 }
1716 break;
1717
1718 /* We must pass backslash through so that it does not
1719 cause quoting during the second expansion. */
1720 case '\\':
1721 if (output)
1722 obstack_grow_str (output, "\\\\");
1723 ++tokptr;
1724 break;
1725
1726 /* Escapes which undergo conversion. */
1727 case 'a':
1728 if (output)
1729 obstack_1grow (output, '\a');
1730 ++tokptr;
1731 break;
1732 case 'b':
1733 if (output)
1734 obstack_1grow (output, '\b');
1735 ++tokptr;
1736 break;
1737 case 'f':
1738 if (output)
1739 obstack_1grow (output, '\f');
1740 ++tokptr;
1741 break;
1742 case 'n':
1743 if (output)
1744 obstack_1grow (output, '\n');
1745 ++tokptr;
1746 break;
1747 case 'r':
1748 if (output)
1749 obstack_1grow (output, '\r');
1750 ++tokptr;
1751 break;
1752 case 't':
1753 if (output)
1754 obstack_1grow (output, '\t');
1755 ++tokptr;
1756 break;
1757 case 'v':
1758 if (output)
1759 obstack_1grow (output, '\v');
1760 ++tokptr;
1761 break;
1762
1763 /* GCC extension. */
1764 case 'e':
1765 if (output)
1766 obstack_1grow (output, HOST_ESCAPE_CHAR);
1767 ++tokptr;
1768 break;
1769
1770 /* Backslash-newline expands to nothing at all. */
1771 case '\n':
1772 ++tokptr;
1773 result = 0;
1774 break;
1775
1776 /* A few escapes just expand to the character itself. */
1777 case '\'':
1778 case '\"':
1779 case '?':
1780 /* GCC extensions. */
1781 case '(':
1782 case '{':
1783 case '[':
1784 case '%':
1785 /* Unrecognized escapes turn into the character itself. */
1786 default:
1787 if (output)
1788 obstack_1grow (output, *tokptr);
1789 ++tokptr;
1790 break;
1791 }
1792 *ptr = tokptr;
1793 return result;
1794 }
1795
1796 /* Parse a string or character literal from TOKPTR. The string or
1797 character may be wide or unicode. *OUTPTR is set to just after the
1798 end of the literal in the input string. The resulting token is
1799 stored in VALUE. This returns a token value, either STRING or
1800 CHAR, depending on what was parsed. *HOST_CHARS is set to the
1801 number of host characters in the literal. */
1802 static int
1803 parse_string_or_char (char *tokptr, char **outptr, struct typed_stoken *value,
1804 int *host_chars)
1805 {
1806 int quote;
1807 enum c_string_type type;
1808
1809 /* Build the gdb internal form of the input string in tempbuf. Note
1810 that the buffer is null byte terminated *only* for the
1811 convenience of debugging gdb itself and printing the buffer
1812 contents when the buffer contains no embedded nulls. Gdb does
1813 not depend upon the buffer being null byte terminated, it uses
1814 the length string instead. This allows gdb to handle C strings
1815 (as well as strings in other languages) with embedded null
1816 bytes */
1817
1818 if (!tempbuf_init)
1819 tempbuf_init = 1;
1820 else
1821 obstack_free (&tempbuf, NULL);
1822 obstack_init (&tempbuf);
1823
1824 /* Record the string type. */
1825 if (*tokptr == 'L')
1826 {
1827 type = C_WIDE_STRING;
1828 ++tokptr;
1829 }
1830 else if (*tokptr == 'u')
1831 {
1832 type = C_STRING_16;
1833 ++tokptr;
1834 }
1835 else if (*tokptr == 'U')
1836 {
1837 type = C_STRING_32;
1838 ++tokptr;
1839 }
1840 else
1841 type = C_STRING;
1842
1843 /* Skip the quote. */
1844 quote = *tokptr;
1845 if (quote == '\'')
1846 type |= C_CHAR;
1847 ++tokptr;
1848
1849 *host_chars = 0;
1850
1851 while (*tokptr)
1852 {
1853 char c = *tokptr;
1854 if (c == '\\')
1855 {
1856 ++tokptr;
1857 *host_chars += c_parse_escape (&tokptr, &tempbuf);
1858 }
1859 else if (c == quote)
1860 break;
1861 else
1862 {
1863 obstack_1grow (&tempbuf, c);
1864 ++tokptr;
1865 /* FIXME: this does the wrong thing with multi-byte host
1866 characters. We could use mbrlen here, but that would
1867 make "set host-charset" a bit less useful. */
1868 ++*host_chars;
1869 }
1870 }
1871
1872 if (*tokptr != quote)
1873 {
1874 if (quote == '"')
1875 error (_("Unterminated string in expression."));
1876 else
1877 error (_("Unmatched single quote."));
1878 }
1879 ++tokptr;
1880
1881 value->type = type;
1882 value->ptr = obstack_base (&tempbuf);
1883 value->length = obstack_object_size (&tempbuf);
1884
1885 *outptr = tokptr;
1886
1887 return quote == '"' ? STRING : CHAR;
1888 }
1889
1890 struct token
1891 {
1892 char *operator;
1893 int token;
1894 enum exp_opcode opcode;
1895 int cxx_only;
1896 };
1897
1898 static const struct token tokentab3[] =
1899 {
1900 {">>=", ASSIGN_MODIFY, BINOP_RSH, 0},
1901 {"<<=", ASSIGN_MODIFY, BINOP_LSH, 0},
1902 {"->*", ARROW_STAR, BINOP_END, 1}
1903 };
1904
1905 static const struct token tokentab2[] =
1906 {
1907 {"+=", ASSIGN_MODIFY, BINOP_ADD, 0},
1908 {"-=", ASSIGN_MODIFY, BINOP_SUB, 0},
1909 {"*=", ASSIGN_MODIFY, BINOP_MUL, 0},
1910 {"/=", ASSIGN_MODIFY, BINOP_DIV, 0},
1911 {"%=", ASSIGN_MODIFY, BINOP_REM, 0},
1912 {"|=", ASSIGN_MODIFY, BINOP_BITWISE_IOR, 0},
1913 {"&=", ASSIGN_MODIFY, BINOP_BITWISE_AND, 0},
1914 {"^=", ASSIGN_MODIFY, BINOP_BITWISE_XOR, 0},
1915 {"++", INCREMENT, BINOP_END, 0},
1916 {"--", DECREMENT, BINOP_END, 0},
1917 {"->", ARROW, BINOP_END, 0},
1918 {"&&", ANDAND, BINOP_END, 0},
1919 {"||", OROR, BINOP_END, 0},
1920 /* "::" is *not* only C++: gdb overrides its meaning in several
1921 different ways, e.g., 'filename'::func, function::variable. */
1922 {"::", COLONCOLON, BINOP_END, 0},
1923 {"<<", LSH, BINOP_END, 0},
1924 {">>", RSH, BINOP_END, 0},
1925 {"==", EQUAL, BINOP_END, 0},
1926 {"!=", NOTEQUAL, BINOP_END, 0},
1927 {"<=", LEQ, BINOP_END, 0},
1928 {">=", GEQ, BINOP_END, 0},
1929 {".*", DOT_STAR, BINOP_END, 1}
1930 };
1931
1932 /* Identifier-like tokens. */
1933 static const struct token ident_tokens[] =
1934 {
1935 {"unsigned", UNSIGNED, OP_NULL, 0},
1936 {"template", TEMPLATE, OP_NULL, 1},
1937 {"volatile", VOLATILE_KEYWORD, OP_NULL, 0},
1938 {"struct", STRUCT, OP_NULL, 0},
1939 {"signed", SIGNED_KEYWORD, OP_NULL, 0},
1940 {"sizeof", SIZEOF, OP_NULL, 0},
1941 {"double", DOUBLE_KEYWORD, OP_NULL, 0},
1942 {"false", FALSEKEYWORD, OP_NULL, 1},
1943 {"class", CLASS, OP_NULL, 1},
1944 {"union", UNION, OP_NULL, 0},
1945 {"short", SHORT, OP_NULL, 0},
1946 {"const", CONST_KEYWORD, OP_NULL, 0},
1947 {"enum", ENUM, OP_NULL, 0},
1948 {"long", LONG, OP_NULL, 0},
1949 {"true", TRUEKEYWORD, OP_NULL, 1},
1950 {"int", INT_KEYWORD, OP_NULL, 0},
1951 {"new", NEW, OP_NULL, 1},
1952 {"delete", DELETE, OP_NULL, 1},
1953 {"operator", OPERATOR, OP_NULL, 1},
1954
1955 {"and", ANDAND, BINOP_END, 1},
1956 {"and_eq", ASSIGN_MODIFY, BINOP_BITWISE_AND, 1},
1957 {"bitand", '&', OP_NULL, 1},
1958 {"bitor", '|', OP_NULL, 1},
1959 {"compl", '~', OP_NULL, 1},
1960 {"not", '!', OP_NULL, 1},
1961 {"not_eq", NOTEQUAL, BINOP_END, 1},
1962 {"or", OROR, BINOP_END, 1},
1963 {"or_eq", ASSIGN_MODIFY, BINOP_BITWISE_IOR, 1},
1964 {"xor", '^', OP_NULL, 1},
1965 {"xor_eq", ASSIGN_MODIFY, BINOP_BITWISE_XOR, 1},
1966
1967 {"const_cast", CONST_CAST, OP_NULL, 1 },
1968 {"dynamic_cast", DYNAMIC_CAST, OP_NULL, 1 },
1969 {"static_cast", STATIC_CAST, OP_NULL, 1 },
1970 {"reinterpret_cast", REINTERPRET_CAST, OP_NULL, 1 }
1971 };
1972
1973 /* When we find that lexptr (the global var defined in parse.c) is
1974 pointing at a macro invocation, we expand the invocation, and call
1975 scan_macro_expansion to save the old lexptr here and point lexptr
1976 into the expanded text. When we reach the end of that, we call
1977 end_macro_expansion to pop back to the value we saved here. The
1978 macro expansion code promises to return only fully-expanded text,
1979 so we don't need to "push" more than one level.
1980
1981 This is disgusting, of course. It would be cleaner to do all macro
1982 expansion beforehand, and then hand that to lexptr. But we don't
1983 really know where the expression ends. Remember, in a command like
1984
1985 (gdb) break *ADDRESS if CONDITION
1986
1987 we evaluate ADDRESS in the scope of the current frame, but we
1988 evaluate CONDITION in the scope of the breakpoint's location. So
1989 it's simply wrong to try to macro-expand the whole thing at once. */
1990 static char *macro_original_text;
1991
1992 /* We save all intermediate macro expansions on this obstack for the
1993 duration of a single parse. The expansion text may sometimes have
1994 to live past the end of the expansion, due to yacc lookahead.
1995 Rather than try to be clever about saving the data for a single
1996 token, we simply keep it all and delete it after parsing has
1997 completed. */
1998 static struct obstack expansion_obstack;
1999
2000 static void
2001 scan_macro_expansion (char *expansion)
2002 {
2003 char *copy;
2004
2005 /* We'd better not be trying to push the stack twice. */
2006 gdb_assert (! macro_original_text);
2007
2008 /* Copy to the obstack, and then free the intermediate
2009 expansion. */
2010 copy = obstack_copy0 (&expansion_obstack, expansion, strlen (expansion));
2011 xfree (expansion);
2012
2013 /* Save the old lexptr value, so we can return to it when we're done
2014 parsing the expanded text. */
2015 macro_original_text = lexptr;
2016 lexptr = copy;
2017 }
2018
2019
2020 static int
2021 scanning_macro_expansion (void)
2022 {
2023 return macro_original_text != 0;
2024 }
2025
2026
2027 static void
2028 finished_macro_expansion (void)
2029 {
2030 /* There'd better be something to pop back to. */
2031 gdb_assert (macro_original_text);
2032
2033 /* Pop back to the original text. */
2034 lexptr = macro_original_text;
2035 macro_original_text = 0;
2036 }
2037
2038
2039 static void
2040 scan_macro_cleanup (void *dummy)
2041 {
2042 if (macro_original_text)
2043 finished_macro_expansion ();
2044
2045 obstack_free (&expansion_obstack, NULL);
2046 }
2047
2048 /* Return true iff the token represents a C++ cast operator. */
2049
2050 static int
2051 is_cast_operator (const char *token, int len)
2052 {
2053 return (! strncmp (token, "dynamic_cast", len)
2054 || ! strncmp (token, "static_cast", len)
2055 || ! strncmp (token, "reinterpret_cast", len)
2056 || ! strncmp (token, "const_cast", len));
2057 }
2058
2059 /* The scope used for macro expansion. */
2060 static struct macro_scope *expression_macro_scope;
2061
2062 /* This is set if a NAME token appeared at the very end of the input
2063 string, with no whitespace separating the name from the EOF. This
2064 is used only when parsing to do field name completion. */
2065 static int saw_name_at_eof;
2066
2067 /* This is set if the previously-returned token was a structure
2068 operator -- either '.' or ARROW. This is used only when parsing to
2069 do field name completion. */
2070 static int last_was_structop;
2071
2072 /* Read one token, getting characters through lexptr. */
2073
2074 static int
2075 lex_one_token (void)
2076 {
2077 int c;
2078 int namelen;
2079 unsigned int i;
2080 char *tokstart;
2081 int saw_structop = last_was_structop;
2082 char *copy;
2083
2084 last_was_structop = 0;
2085
2086 retry:
2087
2088 /* Check if this is a macro invocation that we need to expand. */
2089 if (! scanning_macro_expansion ())
2090 {
2091 char *expanded = macro_expand_next (&lexptr,
2092 standard_macro_lookup,
2093 expression_macro_scope);
2094
2095 if (expanded)
2096 scan_macro_expansion (expanded);
2097 }
2098
2099 prev_lexptr = lexptr;
2100
2101 tokstart = lexptr;
2102 /* See if it is a special token of length 3. */
2103 for (i = 0; i < sizeof tokentab3 / sizeof tokentab3[0]; i++)
2104 if (strncmp (tokstart, tokentab3[i].operator, 3) == 0)
2105 {
2106 if (tokentab3[i].cxx_only
2107 && parse_language->la_language != language_cplus)
2108 break;
2109
2110 lexptr += 3;
2111 yylval.opcode = tokentab3[i].opcode;
2112 return tokentab3[i].token;
2113 }
2114
2115 /* See if it is a special token of length 2. */
2116 for (i = 0; i < sizeof tokentab2 / sizeof tokentab2[0]; i++)
2117 if (strncmp (tokstart, tokentab2[i].operator, 2) == 0)
2118 {
2119 if (tokentab2[i].cxx_only
2120 && parse_language->la_language != language_cplus)
2121 break;
2122
2123 lexptr += 2;
2124 yylval.opcode = tokentab2[i].opcode;
2125 if (in_parse_field && tokentab2[i].token == ARROW)
2126 last_was_structop = 1;
2127 return tokentab2[i].token;
2128 }
2129
2130 switch (c = *tokstart)
2131 {
2132 case 0:
2133 /* If we were just scanning the result of a macro expansion,
2134 then we need to resume scanning the original text.
2135 If we're parsing for field name completion, and the previous
2136 token allows such completion, return a COMPLETE token.
2137 Otherwise, we were already scanning the original text, and
2138 we're really done. */
2139 if (scanning_macro_expansion ())
2140 {
2141 finished_macro_expansion ();
2142 goto retry;
2143 }
2144 else if (saw_name_at_eof)
2145 {
2146 saw_name_at_eof = 0;
2147 return COMPLETE;
2148 }
2149 else if (saw_structop)
2150 return COMPLETE;
2151 else
2152 return 0;
2153
2154 case ' ':
2155 case '\t':
2156 case '\n':
2157 lexptr++;
2158 goto retry;
2159
2160 case '[':
2161 case '(':
2162 paren_depth++;
2163 lexptr++;
2164 return c;
2165
2166 case ']':
2167 case ')':
2168 if (paren_depth == 0)
2169 return 0;
2170 paren_depth--;
2171 lexptr++;
2172 return c;
2173
2174 case ',':
2175 if (comma_terminates
2176 && paren_depth == 0
2177 && ! scanning_macro_expansion ())
2178 return 0;
2179 lexptr++;
2180 return c;
2181
2182 case '.':
2183 /* Might be a floating point number. */
2184 if (lexptr[1] < '0' || lexptr[1] > '9')
2185 {
2186 if (in_parse_field)
2187 last_was_structop = 1;
2188 goto symbol; /* Nope, must be a symbol. */
2189 }
2190 /* FALL THRU into number case. */
2191
2192 case '0':
2193 case '1':
2194 case '2':
2195 case '3':
2196 case '4':
2197 case '5':
2198 case '6':
2199 case '7':
2200 case '8':
2201 case '9':
2202 {
2203 /* It's a number. */
2204 int got_dot = 0, got_e = 0, toktype;
2205 char *p = tokstart;
2206 int hex = input_radix > 10;
2207
2208 if (c == '0' && (p[1] == 'x' || p[1] == 'X'))
2209 {
2210 p += 2;
2211 hex = 1;
2212 }
2213 else if (c == '0' && (p[1]=='t' || p[1]=='T' || p[1]=='d' || p[1]=='D'))
2214 {
2215 p += 2;
2216 hex = 0;
2217 }
2218
2219 for (;; ++p)
2220 {
2221 /* This test includes !hex because 'e' is a valid hex digit
2222 and thus does not indicate a floating point number when
2223 the radix is hex. */
2224 if (!hex && !got_e && (*p == 'e' || *p == 'E'))
2225 got_dot = got_e = 1;
2226 /* This test does not include !hex, because a '.' always indicates
2227 a decimal floating point number regardless of the radix. */
2228 else if (!got_dot && *p == '.')
2229 got_dot = 1;
2230 else if (got_e && (p[-1] == 'e' || p[-1] == 'E')
2231 && (*p == '-' || *p == '+'))
2232 /* This is the sign of the exponent, not the end of the
2233 number. */
2234 continue;
2235 /* We will take any letters or digits. parse_number will
2236 complain if past the radix, or if L or U are not final. */
2237 else if ((*p < '0' || *p > '9')
2238 && ((*p < 'a' || *p > 'z')
2239 && (*p < 'A' || *p > 'Z')))
2240 break;
2241 }
2242 toktype = parse_number (tokstart, p - tokstart, got_dot|got_e, &yylval);
2243 if (toktype == ERROR)
2244 {
2245 char *err_copy = (char *) alloca (p - tokstart + 1);
2246
2247 memcpy (err_copy, tokstart, p - tokstart);
2248 err_copy[p - tokstart] = 0;
2249 error (_("Invalid number \"%s\"."), err_copy);
2250 }
2251 lexptr = p;
2252 return toktype;
2253 }
2254
2255 case '@':
2256 {
2257 char *p = &tokstart[1];
2258 size_t len = strlen ("entry");
2259
2260 while (isspace (*p))
2261 p++;
2262 if (strncmp (p, "entry", len) == 0 && !isalnum (p[len])
2263 && p[len] != '_')
2264 {
2265 lexptr = &p[len];
2266 return ENTRY;
2267 }
2268 }
2269 /* FALLTHRU */
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 case '=':
2285 case '{':
2286 case '}':
2287 symbol:
2288 lexptr++;
2289 return c;
2290
2291 case 'L':
2292 case 'u':
2293 case 'U':
2294 if (tokstart[1] != '"' && tokstart[1] != '\'')
2295 break;
2296 /* Fall through. */
2297 case '\'':
2298 case '"':
2299 {
2300 int host_len;
2301 int result = parse_string_or_char (tokstart, &lexptr, &yylval.tsval,
2302 &host_len);
2303 if (result == CHAR)
2304 {
2305 if (host_len == 0)
2306 error (_("Empty character constant."));
2307 else if (host_len > 2 && c == '\'')
2308 {
2309 ++tokstart;
2310 namelen = lexptr - tokstart - 1;
2311 goto tryname;
2312 }
2313 else if (host_len > 1)
2314 error (_("Invalid character constant."));
2315 }
2316 return result;
2317 }
2318 }
2319
2320 if (!(c == '_' || c == '$'
2321 || (c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z')))
2322 /* We must have come across a bad character (e.g. ';'). */
2323 error (_("Invalid character '%c' in expression."), c);
2324
2325 /* It's a name. See how long it is. */
2326 namelen = 0;
2327 for (c = tokstart[namelen];
2328 (c == '_' || c == '$' || (c >= '0' && c <= '9')
2329 || (c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z') || c == '<');)
2330 {
2331 /* Template parameter lists are part of the name.
2332 FIXME: This mishandles `print $a<4&&$a>3'. */
2333
2334 if (c == '<')
2335 {
2336 if (! is_cast_operator (tokstart, namelen))
2337 {
2338 /* Scan ahead to get rest of the template specification. Note
2339 that we look ahead only when the '<' adjoins non-whitespace
2340 characters; for comparison expressions, e.g. "a < b > c",
2341 there must be spaces before the '<', etc. */
2342
2343 char * p = find_template_name_end (tokstart + namelen);
2344 if (p)
2345 namelen = p - tokstart;
2346 }
2347 break;
2348 }
2349 c = tokstart[++namelen];
2350 }
2351
2352 /* The token "if" terminates the expression and is NOT removed from
2353 the input stream. It doesn't count if it appears in the
2354 expansion of a macro. */
2355 if (namelen == 2
2356 && tokstart[0] == 'i'
2357 && tokstart[1] == 'f'
2358 && ! scanning_macro_expansion ())
2359 {
2360 return 0;
2361 }
2362
2363 /* For the same reason (breakpoint conditions), "thread N"
2364 terminates the expression. "thread" could be an identifier, but
2365 an identifier is never followed by a number without intervening
2366 punctuation. "task" is similar. Handle abbreviations of these,
2367 similarly to breakpoint.c:find_condition_and_thread. */
2368 if (namelen >= 1
2369 && (strncmp (tokstart, "thread", namelen) == 0
2370 || strncmp (tokstart, "task", namelen) == 0)
2371 && (tokstart[namelen] == ' ' || tokstart[namelen] == '\t')
2372 && ! scanning_macro_expansion ())
2373 {
2374 char *p = tokstart + namelen + 1;
2375 while (*p == ' ' || *p == '\t')
2376 p++;
2377 if (*p >= '0' && *p <= '9')
2378 return 0;
2379 }
2380
2381 lexptr += namelen;
2382
2383 tryname:
2384
2385 yylval.sval.ptr = tokstart;
2386 yylval.sval.length = namelen;
2387
2388 /* Catch specific keywords. */
2389 copy = copy_name (yylval.sval);
2390 for (i = 0; i < sizeof ident_tokens / sizeof ident_tokens[0]; i++)
2391 if (strcmp (copy, ident_tokens[i].operator) == 0)
2392 {
2393 if (ident_tokens[i].cxx_only
2394 && parse_language->la_language != language_cplus)
2395 break;
2396
2397 /* It is ok to always set this, even though we don't always
2398 strictly need to. */
2399 yylval.opcode = ident_tokens[i].opcode;
2400 return ident_tokens[i].token;
2401 }
2402
2403 if (*tokstart == '$')
2404 return VARIABLE;
2405
2406 if (in_parse_field && *lexptr == '\0')
2407 saw_name_at_eof = 1;
2408 return NAME;
2409 }
2410
2411 /* An object of this type is pushed on a FIFO by the "outer" lexer. */
2412 typedef struct
2413 {
2414 int token;
2415 YYSTYPE value;
2416 } token_and_value;
2417
2418 DEF_VEC_O (token_and_value);
2419
2420 /* A FIFO of tokens that have been read but not yet returned to the
2421 parser. */
2422 static VEC (token_and_value) *token_fifo;
2423
2424 /* Non-zero if the lexer should return tokens from the FIFO. */
2425 static int popping;
2426
2427 /* Temporary storage for c_lex; this holds symbol names as they are
2428 built up. */
2429 static struct obstack name_obstack;
2430
2431 /* Classify a NAME token. The contents of the token are in `yylval'.
2432 Updates yylval and returns the new token type. BLOCK is the block
2433 in which lookups start; this can be NULL to mean the global
2434 scope. */
2435 static int
2436 classify_name (struct block *block)
2437 {
2438 struct symbol *sym;
2439 char *copy;
2440 int is_a_field_of_this = 0;
2441
2442 copy = copy_name (yylval.sval);
2443
2444 sym = lookup_symbol (copy, block, VAR_DOMAIN,
2445 parse_language->la_language == language_cplus
2446 ? &is_a_field_of_this : (int *) NULL);
2447
2448 if (sym && SYMBOL_CLASS (sym) == LOC_BLOCK)
2449 {
2450 yylval.ssym.sym = sym;
2451 yylval.ssym.is_a_field_of_this = is_a_field_of_this;
2452 return BLOCKNAME;
2453 }
2454 else if (!sym)
2455 {
2456 /* See if it's a file name. */
2457 struct symtab *symtab;
2458
2459 symtab = lookup_symtab (copy);
2460 if (symtab)
2461 {
2462 yylval.bval = BLOCKVECTOR_BLOCK (BLOCKVECTOR (symtab), STATIC_BLOCK);
2463 return FILENAME;
2464 }
2465 }
2466
2467 if (sym && SYMBOL_CLASS (sym) == LOC_TYPEDEF)
2468 {
2469 yylval.tsym.type = SYMBOL_TYPE (sym);
2470 return TYPENAME;
2471 }
2472
2473 yylval.tsym.type
2474 = language_lookup_primitive_type_by_name (parse_language,
2475 parse_gdbarch, copy);
2476 if (yylval.tsym.type != NULL)
2477 return TYPENAME;
2478
2479 /* Input names that aren't symbols but ARE valid hex numbers, when
2480 the input radix permits them, can be names or numbers depending
2481 on the parse. Note we support radixes > 16 here. */
2482 if (!sym
2483 && ((copy[0] >= 'a' && copy[0] < 'a' + input_radix - 10)
2484 || (copy[0] >= 'A' && copy[0] < 'A' + input_radix - 10)))
2485 {
2486 YYSTYPE newlval; /* Its value is ignored. */
2487 int hextype = parse_number (copy, yylval.sval.length, 0, &newlval);
2488 if (hextype == INT)
2489 {
2490 yylval.ssym.sym = sym;
2491 yylval.ssym.is_a_field_of_this = is_a_field_of_this;
2492 return NAME_OR_INT;
2493 }
2494 }
2495
2496 /* Any other kind of symbol */
2497 yylval.ssym.sym = sym;
2498 yylval.ssym.is_a_field_of_this = is_a_field_of_this;
2499
2500 if (sym == NULL
2501 && parse_language->la_language == language_cplus
2502 && !is_a_field_of_this
2503 && !lookup_minimal_symbol (copy, NULL, NULL))
2504 return UNKNOWN_CPP_NAME;
2505
2506 return NAME;
2507 }
2508
2509 /* Like classify_name, but used by the inner loop of the lexer, when a
2510 name might have already been seen. FIRST_NAME is true if the token
2511 in `yylval' is the first component of a name, false otherwise. If
2512 this function returns NAME, it might not have updated `yylval'.
2513 This is ok because the caller only cares about TYPENAME. */
2514 static int
2515 classify_inner_name (struct block *block, int first_name)
2516 {
2517 struct type *type, *new_type;
2518 char *copy;
2519
2520 if (first_name)
2521 return classify_name (block);
2522
2523 type = check_typedef (yylval.tsym.type);
2524 if (TYPE_CODE (type) != TYPE_CODE_STRUCT
2525 && TYPE_CODE (type) != TYPE_CODE_UNION
2526 && TYPE_CODE (type) != TYPE_CODE_NAMESPACE)
2527 /* We know the caller won't expect us to update yylval. */
2528 return NAME;
2529
2530 copy = copy_name (yylval.tsym.stoken);
2531 new_type = cp_lookup_nested_type (yylval.tsym.type, copy, block);
2532
2533 if (new_type == NULL)
2534 /* We know the caller won't expect us to update yylval. */
2535 return NAME;
2536
2537 yylval.tsym.type = new_type;
2538 return TYPENAME;
2539 }
2540
2541 /* The outer level of a two-level lexer. This calls the inner lexer
2542 to return tokens. It then either returns these tokens, or
2543 aggregates them into a larger token. This lets us work around a
2544 problem in our parsing approach, where the parser could not
2545 distinguish between qualified names and qualified types at the
2546 right point.
2547
2548 This approach is still not ideal, because it mishandles template
2549 types. See the comment in lex_one_token for an example. However,
2550 this is still an improvement over the earlier approach, and will
2551 suffice until we move to better parsing technology. */
2552 static int
2553 yylex (void)
2554 {
2555 token_and_value current;
2556 int first_was_coloncolon, last_was_coloncolon, first_iter;
2557
2558 if (popping && !VEC_empty (token_and_value, token_fifo))
2559 {
2560 token_and_value tv = *VEC_index (token_and_value, token_fifo, 0);
2561 VEC_ordered_remove (token_and_value, token_fifo, 0);
2562 yylval = tv.value;
2563 return tv.token;
2564 }
2565 popping = 0;
2566
2567 current.token = lex_one_token ();
2568 if (current.token == NAME)
2569 current.token = classify_name (expression_context_block);
2570 if (parse_language->la_language != language_cplus
2571 || (current.token != TYPENAME && current.token != COLONCOLON))
2572 return current.token;
2573
2574 first_was_coloncolon = current.token == COLONCOLON;
2575 last_was_coloncolon = first_was_coloncolon;
2576 obstack_free (&name_obstack, obstack_base (&name_obstack));
2577 if (!last_was_coloncolon)
2578 obstack_grow (&name_obstack, yylval.sval.ptr, yylval.sval.length);
2579 current.value = yylval;
2580 first_iter = 1;
2581 while (1)
2582 {
2583 token_and_value next;
2584
2585 next.token = lex_one_token ();
2586 next.value = yylval;
2587
2588 if (next.token == NAME && last_was_coloncolon)
2589 {
2590 int classification;
2591
2592 classification = classify_inner_name (first_was_coloncolon
2593 ? NULL
2594 : expression_context_block,
2595 first_iter);
2596 /* We keep going until we either run out of names, or until
2597 we have a qualified name which is not a type. */
2598 if (classification != TYPENAME)
2599 {
2600 /* Push the final component and leave the loop. */
2601 VEC_safe_push (token_and_value, token_fifo, &next);
2602 break;
2603 }
2604
2605 /* Update the partial name we are constructing. */
2606 if (!first_iter)
2607 {
2608 /* We don't want to put a leading "::" into the name. */
2609 obstack_grow_str (&name_obstack, "::");
2610 }
2611 obstack_grow (&name_obstack, next.value.sval.ptr,
2612 next.value.sval.length);
2613
2614 yylval.sval.ptr = obstack_base (&name_obstack);
2615 yylval.sval.length = obstack_object_size (&name_obstack);
2616 current.value = yylval;
2617 current.token = classification;
2618
2619 last_was_coloncolon = 0;
2620 }
2621 else if (next.token == COLONCOLON && !last_was_coloncolon)
2622 last_was_coloncolon = 1;
2623 else
2624 {
2625 /* We've reached the end of the name. */
2626 VEC_safe_push (token_and_value, token_fifo, &next);
2627 break;
2628 }
2629
2630 first_iter = 0;
2631 }
2632
2633 popping = 1;
2634
2635 /* If we ended with a "::", insert it too. */
2636 if (last_was_coloncolon)
2637 {
2638 token_and_value cc;
2639 memset (&cc, 0, sizeof (token_and_value));
2640 if (first_was_coloncolon && first_iter)
2641 {
2642 yylval = cc.value;
2643 return COLONCOLON;
2644 }
2645 cc.token = COLONCOLON;
2646 VEC_safe_insert (token_and_value, token_fifo, 0, &cc);
2647 }
2648
2649 yylval = current.value;
2650 yylval.sval.ptr = obstack_copy0 (&expansion_obstack,
2651 yylval.sval.ptr,
2652 yylval.sval.length);
2653 return current.token;
2654 }
2655
2656 int
2657 c_parse (void)
2658 {
2659 int result;
2660 struct cleanup *back_to = make_cleanup (free_current_contents,
2661 &expression_macro_scope);
2662
2663 /* Set up the scope for macro expansion. */
2664 expression_macro_scope = NULL;
2665
2666 if (expression_context_block)
2667 expression_macro_scope
2668 = sal_macro_scope (find_pc_line (expression_context_pc, 0));
2669 else
2670 expression_macro_scope = default_macro_scope ();
2671 if (! expression_macro_scope)
2672 expression_macro_scope = user_macro_scope ();
2673
2674 /* Initialize macro expansion code. */
2675 obstack_init (&expansion_obstack);
2676 gdb_assert (! macro_original_text);
2677 make_cleanup (scan_macro_cleanup, 0);
2678
2679 make_cleanup_restore_integer (&yydebug);
2680 yydebug = parser_debug;
2681
2682 /* Initialize some state used by the lexer. */
2683 last_was_structop = 0;
2684 saw_name_at_eof = 0;
2685
2686 VEC_free (token_and_value, token_fifo);
2687 popping = 0;
2688 obstack_init (&name_obstack);
2689 make_cleanup_obstack_free (&name_obstack);
2690
2691 result = yyparse ();
2692 do_cleanups (back_to);
2693 return result;
2694 }
2695
2696
2697 void
2698 yyerror (char *msg)
2699 {
2700 if (prev_lexptr)
2701 lexptr = prev_lexptr;
2702
2703 error (_("A %s in expression, near `%s'."), (msg ? msg : "error"), lexptr);
2704 }
This page took 0.087074 seconds and 4 git commands to generate.