projects / binutils-gdb.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
history | raw | HEAD
Rewrite the Rust expression parser
[binutils-gdb.git] / gdb / c-exp.y
1 /* YACC parser for C expressions, for GDB.
2 Copyright (C) 1986-2021 Free Software Foundation, Inc.
3
4 This file is part of GDB.
5
6 This program is free software; you can redistribute it and/or modify
7 it under the terms of the GNU General Public License as published by
8 the Free Software Foundation; either version 3 of the License, or
9 (at your option) any later version.
10
11 This program is distributed in the hope that it will be useful,
12 but WITHOUT ANY WARRANTY; without even the implied warranty of
13 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
14 GNU General Public License for more details.
15
16 You should have received a copy of the GNU General Public License
17 along with this program. If not, see <http://www.gnu.org/licenses/>. */
18
19 /* Parse a C expression from text in a string,
20 and return the result as a struct expression pointer.
21 That structure contains arithmetic operations in reverse polish,
22 with constants represented by operations that are followed by special data.
23 See expression.h for the details of the format.
24 What is important here is that it can be built up sequentially
25 during the process of parsing; the lower levels of the tree always
26 come first in the result.
27
28 Note that malloc's and realloc's in this file are transformed to
29 xmalloc and xrealloc respectively by the same sed command in the
30 makefile that remaps any other malloc/realloc inserted by the parser
31 generator. Doing this with #defines and trying to control the interaction
32 with include files (<malloc.h> and <stdlib.h> for example) just became
33 too messy, particularly when such includes can be inserted at random
34 times by the parser generator. */
35
36 %{
37
38 #include "defs.h"
39 #include <ctype.h>
40 #include "expression.h"
41 #include "value.h"
42 #include "parser-defs.h"
43 #include "language.h"
44 #include "c-lang.h"
45 #include "c-support.h"
46 #include "bfd.h" /* Required by objfiles.h. */
47 #include "symfile.h" /* Required by objfiles.h. */
48 #include "objfiles.h" /* For have_full_symbols and have_partial_symbols */
49 #include "charset.h"
50 #include "block.h"
51 #include "cp-support.h"
52 #include "macroscope.h"
53 #include "objc-lang.h"
54 #include "typeprint.h"
55 #include "cp-abi.h"
56 #include "type-stack.h"
57 #include "target-float.h"
58 #include "c-exp.h"
59
60 #define parse_type(ps) builtin_type (ps->gdbarch ())
61
62 /* Remap normal yacc parser interface names (yyparse, yylex, yyerror,
63 etc). */
64 #define GDB_YY_REMAP_PREFIX c_
65 #include "yy-remap.h"
66
67 /* The state of the parser, used internally when we are parsing the
68 expression. */
69
70 static struct parser_state *pstate = NULL;
71
72 /* Data that must be held for the duration of a parse. */
73
74 struct c_parse_state
75 {
76 /* These are used to hold type lists and type stacks that are
77 allocated during the parse. */
78 std::vector<std::unique_ptr<std::vector<struct type *>>> type_lists;
79 std::vector<std::unique_ptr<struct type_stack>> type_stacks;
80
81 /* Storage for some strings allocated during the parse. */
82 std::vector<gdb::unique_xmalloc_ptr<char>> strings;
83
84 /* When we find that lexptr (the global var defined in parse.c) is
85 pointing at a macro invocation, we expand the invocation, and call
86 scan_macro_expansion to save the old lexptr here and point lexptr
87 into the expanded text. When we reach the end of that, we call
88 end_macro_expansion to pop back to the value we saved here. The
89 macro expansion code promises to return only fully-expanded text,
90 so we don't need to "push" more than one level.
91
92 This is disgusting, of course. It would be cleaner to do all macro
93 expansion beforehand, and then hand that to lexptr. But we don't
94 really know where the expression ends. Remember, in a command like
95
96 (gdb) break *ADDRESS if CONDITION
97
98 we evaluate ADDRESS in the scope of the current frame, but we
99 evaluate CONDITION in the scope of the breakpoint's location. So
100 it's simply wrong to try to macro-expand the whole thing at once. */
101 const char *macro_original_text = nullptr;
102
103 /* We save all intermediate macro expansions on this obstack for the
104 duration of a single parse. The expansion text may sometimes have
105 to live past the end of the expansion, due to yacc lookahead.
106 Rather than try to be clever about saving the data for a single
107 token, we simply keep it all and delete it after parsing has
108 completed. */
109 auto_obstack expansion_obstack;
110
111 /* The type stack. */
112 struct type_stack type_stack;
113 };
114
115 /* This is set and cleared in c_parse. */
116
117 static struct c_parse_state *cpstate;
118
119 int yyparse (void);
120
121 static int yylex (void);
122
123 static void yyerror (const char *);
124
125 static int type_aggregate_p (struct type *);
126
127 using namespace expr;
128 %}
129
130 /* Although the yacc "value" of an expression is not used,
131 since the result is stored in the structure being created,
132 other node types do have values. */
133
134 %union
135 {
136 LONGEST lval;
137 struct {
138 LONGEST val;
139 struct type *type;
140 } typed_val_int;
141 struct {
142 gdb_byte val[16];
143 struct type *type;
144 } typed_val_float;
145 struct type *tval;
146 struct stoken sval;
147 struct typed_stoken tsval;
148 struct ttype tsym;
149 struct symtoken ssym;
150 int voidval;
151 const struct block *bval;
152 enum exp_opcode opcode;
153
154 struct stoken_vector svec;
155 std::vector<struct type *> *tvec;
156
157 struct type_stack *type_stack;
158
159 struct objc_class_str theclass;
160 }
161
162 %{
163 /* YYSTYPE gets defined by %union */
164 static int parse_number (struct parser_state *par_state,
165 const char *, int, int, YYSTYPE *);
166 static struct stoken operator_stoken (const char *);
167 static struct stoken typename_stoken (const char *);
168 static void check_parameter_typelist (std::vector<struct type *> *);
169
170 #ifdef YYBISON
171 static void c_print_token (FILE *file, int type, YYSTYPE value);
172 #define YYPRINT(FILE, TYPE, VALUE) c_print_token (FILE, TYPE, VALUE)
173 #endif
174 %}
175
176 %type <voidval> exp exp1 type_exp start variable qualified_name lcurly function_method
177 %type <lval> rcurly
178 %type <tval> type typebase scalar_type
179 %type <tvec> nonempty_typelist func_mod parameter_typelist
180 /* %type <bval> block */
181
182 /* Fancy type parsing. */
183 %type <tval> ptype
184 %type <lval> array_mod
185 %type <tval> conversion_type_id
186
187 %type <type_stack> ptr_operator_ts abs_decl direct_abs_decl
188
189 %token <typed_val_int> INT COMPLEX_INT
190 %token <typed_val_float> FLOAT COMPLEX_FLOAT
191
192 /* Both NAME and TYPENAME tokens represent symbols in the input,
193 and both convey their data as strings.
194 But a TYPENAME is a string that happens to be defined as a typedef
195 or builtin type name (such as int or char)
196 and a NAME is any other symbol.
197 Contexts where this distinction is not important can use the
198 nonterminal "name", which matches either NAME or TYPENAME. */
199
200 %token <tsval> STRING
201 %token <sval> NSSTRING /* ObjC Foundation "NSString" literal */
202 %token SELECTOR /* ObjC "@selector" pseudo-operator */
203 %token <tsval> CHAR
204 %token <ssym> NAME /* BLOCKNAME defined below to give it higher precedence. */
205 %token <ssym> UNKNOWN_CPP_NAME
206 %token <voidval> COMPLETE
207 %token <tsym> TYPENAME
208 %token <theclass> CLASSNAME /* ObjC Class name */
209 %type <sval> name field_name
210 %type <svec> string_exp
211 %type <ssym> name_not_typename
212 %type <tsym> type_name
213
214 /* This is like a '[' token, but is only generated when parsing
215 Objective C. This lets us reuse the same parser without
216 erroneously parsing ObjC-specific expressions in C. */
217 %token OBJC_LBRAC
218
219 /* A NAME_OR_INT is a symbol which is not known in the symbol table,
220 but which would parse as a valid number in the current input radix.
221 E.g. "c" when input_radix==16. Depending on the parse, it will be
222 turned into a name or into a number. */
223
224 %token <ssym> NAME_OR_INT
225
226 %token OPERATOR
227 %token STRUCT CLASS UNION ENUM SIZEOF ALIGNOF UNSIGNED COLONCOLON
228 %token TEMPLATE
229 %token ERROR
230 %token NEW DELETE
231 %type <sval> oper
232 %token REINTERPRET_CAST DYNAMIC_CAST STATIC_CAST CONST_CAST
233 %token ENTRY
234 %token TYPEOF
235 %token DECLTYPE
236 %token TYPEID
237
238 /* Special type cases, put in to allow the parser to distinguish different
239 legal basetypes. */
240 %token SIGNED_KEYWORD LONG SHORT INT_KEYWORD CONST_KEYWORD VOLATILE_KEYWORD DOUBLE_KEYWORD
241 %token RESTRICT ATOMIC
242 %token FLOAT_KEYWORD COMPLEX
243
244 %token <sval> DOLLAR_VARIABLE
245
246 %token <opcode> ASSIGN_MODIFY
247
248 /* C++ */
249 %token TRUEKEYWORD
250 %token FALSEKEYWORD
251
252
253 %left ','
254 %left ABOVE_COMMA
255 %right '=' ASSIGN_MODIFY
256 %right '?'
257 %left OROR
258 %left ANDAND
259 %left '|'
260 %left '^'
261 %left '&'
262 %left EQUAL NOTEQUAL
263 %left '<' '>' LEQ GEQ
264 %left LSH RSH
265 %left '@'
266 %left '+' '-'
267 %left '*' '/' '%'
268 %right UNARY INCREMENT DECREMENT
269 %right ARROW ARROW_STAR '.' DOT_STAR '[' OBJC_LBRAC '('
270 %token <ssym> BLOCKNAME
271 %token <bval> FILENAME
272 %type <bval> block
273 %left COLONCOLON
274
275 %token DOTDOTDOT
276
277 \f
278 %%
279
280 start : exp1
281 | type_exp
282 ;
283
284 type_exp: type
285 {
286 pstate->push_new<type_operation> ($1);
287 }
288 | TYPEOF '(' exp ')'
289 {
290 pstate->wrap<typeof_operation> ();
291 }
292 | TYPEOF '(' type ')'
293 {
294 pstate->push_new<type_operation> ($3);
295 }
296 | DECLTYPE '(' exp ')'
297 {
298 pstate->wrap<decltype_operation> ();
299 }
300 ;
301
302 /* Expressions, including the comma operator. */
303 exp1 : exp
304 | exp1 ',' exp
305 { pstate->wrap2<comma_operation> (); }
306 ;
307
308 /* Expressions, not including the comma operator. */
309 exp : '*' exp %prec UNARY
310 { pstate->wrap<unop_ind_operation> (); }
311 ;
312
313 exp : '&' exp %prec UNARY
314 { pstate->wrap<unop_addr_operation> (); }
315 ;
316
317 exp : '-' exp %prec UNARY
318 { pstate->wrap<unary_neg_operation> (); }
319 ;
320
321 exp : '+' exp %prec UNARY
322 { pstate->wrap<unary_plus_operation> (); }
323 ;
324
325 exp : '!' exp %prec UNARY
326 {
327 if (pstate->language ()->la_language
328 == language_opencl)
329 pstate->wrap<opencl_not_operation> ();
330 else
331 pstate->wrap<unary_logical_not_operation> ();
332 }
333 ;
334
335 exp : '~' exp %prec UNARY
336 { pstate->wrap<unary_complement_operation> (); }
337 ;
338
339 exp : INCREMENT exp %prec UNARY
340 { pstate->wrap<preinc_operation> (); }
341 ;
342
343 exp : DECREMENT exp %prec UNARY
344 { pstate->wrap<predec_operation> (); }
345 ;
346
347 exp : exp INCREMENT %prec UNARY
348 { pstate->wrap<postinc_operation> (); }
349 ;
350
351 exp : exp DECREMENT %prec UNARY
352 { pstate->wrap<postdec_operation> (); }
353 ;
354
355 exp : TYPEID '(' exp ')' %prec UNARY
356 { pstate->wrap<typeid_operation> (); }
357 ;
358
359 exp : TYPEID '(' type_exp ')' %prec UNARY
360 { pstate->wrap<typeid_operation> (); }
361 ;
362
363 exp : SIZEOF exp %prec UNARY
364 { pstate->wrap<unop_sizeof_operation> (); }
365 ;
366
367 exp : ALIGNOF '(' type_exp ')' %prec UNARY
368 { pstate->wrap<unop_alignof_operation> (); }
369 ;
370
371 exp : exp ARROW field_name
372 {
373 pstate->push_new<structop_ptr_operation>
374 (pstate->pop (), copy_name ($3));
375 }
376 ;
377
378 exp : exp ARROW field_name COMPLETE
379 {
380 structop_base_operation *op
381 = new structop_ptr_operation (pstate->pop (),
382 copy_name ($3));
383 pstate->mark_struct_expression (op);
384 pstate->push (operation_up (op));
385 }
386 ;
387
388 exp : exp ARROW COMPLETE
389 {
390 structop_base_operation *op
391 = new structop_ptr_operation (pstate->pop (), "");
392 pstate->mark_struct_expression (op);
393 pstate->push (operation_up (op));
394 }
395 ;
396
397 exp : exp ARROW '~' name
398 {
399 pstate->push_new<structop_ptr_operation>
400 (pstate->pop (), "~" + copy_name ($4));
401 }
402 ;
403
404 exp : exp ARROW '~' name COMPLETE
405 {
406 structop_base_operation *op
407 = new structop_ptr_operation (pstate->pop (),
408 "~" + copy_name ($4));
409 pstate->mark_struct_expression (op);
410 pstate->push (operation_up (op));
411 }
412 ;
413
414 exp : exp ARROW qualified_name
415 { /* exp->type::name becomes exp->*(&type::name) */
416 /* Note: this doesn't work if name is a
417 static member! FIXME */
418 pstate->wrap<unop_addr_operation> ();
419 pstate->wrap2<structop_mptr_operation> (); }
420 ;
421
422 exp : exp ARROW_STAR exp
423 { pstate->wrap2<structop_mptr_operation> (); }
424 ;
425
426 exp : exp '.' field_name
427 {
428 if (pstate->language ()->la_language
429 == language_opencl)
430 pstate->push_new<opencl_structop_operation>
431 (pstate->pop (), copy_name ($3));
432 else
433 pstate->push_new<structop_operation>
434 (pstate->pop (), copy_name ($3));
435 }
436 ;
437
438 exp : exp '.' field_name COMPLETE
439 {
440 structop_base_operation *op
441 = new structop_operation (pstate->pop (),
442 copy_name ($3));
443 pstate->mark_struct_expression (op);
444 pstate->push (operation_up (op));
445 }
446 ;
447
448 exp : exp '.' COMPLETE
449 {
450 structop_base_operation *op
451 = new structop_operation (pstate->pop (), "");
452 pstate->mark_struct_expression (op);
453 pstate->push (operation_up (op));
454 }
455 ;
456
457 exp : exp '.' '~' name
458 {
459 pstate->push_new<structop_operation>
460 (pstate->pop (), "~" + copy_name ($4));
461 }
462 ;
463
464 exp : exp '.' '~' name COMPLETE
465 {
466 structop_base_operation *op
467 = new structop_operation (pstate->pop (),
468 "~" + copy_name ($4));
469 pstate->mark_struct_expression (op);
470 pstate->push (operation_up (op));
471 }
472 ;
473
474 exp : exp '.' qualified_name
475 { /* exp.type::name becomes exp.*(&type::name) */
476 /* Note: this doesn't work if name is a
477 static member! FIXME */
478 pstate->wrap<unop_addr_operation> ();
479 pstate->wrap2<structop_member_operation> (); }
480 ;
481
482 exp : exp DOT_STAR exp
483 { pstate->wrap2<structop_member_operation> (); }
484 ;
485
486 exp : exp '[' exp1 ']'
487 { pstate->wrap2<subscript_operation> (); }
488 ;
489
490 exp : exp OBJC_LBRAC exp1 ']'
491 { pstate->wrap2<subscript_operation> (); }
492 ;
493
494 /*
495 * The rules below parse ObjC message calls of the form:
496 * '[' target selector {':' argument}* ']'
497 */
498
499 exp : OBJC_LBRAC TYPENAME
500 {
501 CORE_ADDR theclass;
502
503 std::string copy = copy_name ($2.stoken);
504 theclass = lookup_objc_class (pstate->gdbarch (),
505 copy.c_str ());
506 if (theclass == 0)
507 error (_("%s is not an ObjC Class"),
508 copy.c_str ());
509 pstate->push_new<long_const_operation>
510 (parse_type (pstate)->builtin_int,
511 (LONGEST) theclass);
512 start_msglist();
513 }
514 msglist ']'
515 { end_msglist (pstate); }
516 ;
517
518 exp : OBJC_LBRAC CLASSNAME
519 {
520 pstate->push_new<long_const_operation>
521 (parse_type (pstate)->builtin_int,
522 (LONGEST) $2.theclass);
523 start_msglist();
524 }
525 msglist ']'
526 { end_msglist (pstate); }
527 ;
528
529 exp : OBJC_LBRAC exp
530 { start_msglist(); }
531 msglist ']'
532 { end_msglist (pstate); }
533 ;
534
535 msglist : name
536 { add_msglist(&$1, 0); }
537 | msgarglist
538 ;
539
540 msgarglist : msgarg
541 | msgarglist msgarg
542 ;
543
544 msgarg : name ':' exp
545 { add_msglist(&$1, 1); }
546 | ':' exp /* Unnamed arg. */
547 { add_msglist(0, 1); }
548 | ',' exp /* Variable number of args. */
549 { add_msglist(0, 0); }
550 ;
551
552 exp : exp '('
553 /* This is to save the value of arglist_len
554 being accumulated by an outer function call. */
555 { pstate->start_arglist (); }
556 arglist ')' %prec ARROW
557 {
558 std::vector<operation_up> args
559 = pstate->pop_vector (pstate->end_arglist ());
560 pstate->push_new<funcall_operation>
561 (pstate->pop (), std::move (args));
562 }
563 ;
564
565 /* This is here to disambiguate with the production for
566 "func()::static_var" further below, which uses
567 function_method_void. */
568 exp : exp '(' ')' %prec ARROW
569 {
570 pstate->push_new<funcall_operation>
571 (pstate->pop (), std::vector<operation_up> ());
572 }
573 ;
574
575
576 exp : UNKNOWN_CPP_NAME '('
577 {
578 /* This could potentially be a an argument defined
579 lookup function (Koenig). */
580 /* This is to save the value of arglist_len
581 being accumulated by an outer function call. */
582 pstate->start_arglist ();
583 }
584 arglist ')' %prec ARROW
585 {
586 std::vector<operation_up> args
587 = pstate->pop_vector (pstate->end_arglist ());
588 pstate->push_new<adl_func_operation>
589 (copy_name ($1.stoken),
590 pstate->expression_context_block,
591 std::move (args));
592 }
593 ;
594
595 lcurly : '{'
596 { pstate->start_arglist (); }
597 ;
598
599 arglist :
600 ;
601
602 arglist : exp
603 { pstate->arglist_len = 1; }
604 ;
605
606 arglist : arglist ',' exp %prec ABOVE_COMMA
607 { pstate->arglist_len++; }
608 ;
609
610 function_method: exp '(' parameter_typelist ')' const_or_volatile
611 {
612 std::vector<struct type *> *type_list = $3;
613 /* Save the const/volatile qualifiers as
614 recorded by the const_or_volatile
615 production's actions. */
616 type_instance_flags flags
617 = (cpstate->type_stack
618 .follow_type_instance_flags ());
619 pstate->push_new<type_instance_operation>
620 (flags, std::move (*type_list),
621 pstate->pop ());
622 }
623 ;
624
625 function_method_void: exp '(' ')' const_or_volatile
626 {
627 type_instance_flags flags
628 = (cpstate->type_stack
629 .follow_type_instance_flags ());
630 pstate->push_new<type_instance_operation>
631 (flags, std::vector<type *> (), pstate->pop ());
632 }
633 ;
634
635 exp : function_method
636 ;
637
638 /* Normally we must interpret "func()" as a function call, instead of
639 a type. The user needs to write func(void) to disambiguate.
640 However, in the "func()::static_var" case, there's no
641 ambiguity. */
642 function_method_void_or_typelist: function_method
643 | function_method_void
644 ;
645
646 exp : function_method_void_or_typelist COLONCOLON name
647 {
648 pstate->push_new<func_static_var_operation>
649 (pstate->pop (), copy_name ($3));
650 }
651 ;
652
653 rcurly : '}'
654 { $$ = pstate->end_arglist () - 1; }
655 ;
656 exp : lcurly arglist rcurly %prec ARROW
657 {
658 std::vector<operation_up> args
659 = pstate->pop_vector ($3 + 1);
660 pstate->push_new<array_operation> (0, $3,
661 std::move (args));
662 }
663 ;
664
665 exp : lcurly type_exp rcurly exp %prec UNARY
666 { pstate->wrap2<unop_memval_type_operation> (); }
667 ;
668
669 exp : '(' type_exp ')' exp %prec UNARY
670 {
671 if (pstate->language ()->la_language
672 == language_opencl)
673 pstate->wrap2<opencl_cast_type_operation> ();
674 else
675 pstate->wrap2<unop_cast_type_operation> ();
676 }
677 ;
678
679 exp : '(' exp1 ')'
680 { }
681 ;
682
683 /* Binary operators in order of decreasing precedence. */
684
685 exp : exp '@' exp
686 { pstate->wrap2<repeat_operation> (); }
687 ;
688
689 exp : exp '*' exp
690 { pstate->wrap2<mul_operation> (); }
691 ;
692
693 exp : exp '/' exp
694 { pstate->wrap2<div_operation> (); }
695 ;
696
697 exp : exp '%' exp
698 { pstate->wrap2<rem_operation> (); }
699 ;
700
701 exp : exp '+' exp
702 { pstate->wrap2<add_operation> (); }
703 ;
704
705 exp : exp '-' exp
706 { pstate->wrap2<sub_operation> (); }
707 ;
708
709 exp : exp LSH exp
710 { pstate->wrap2<lsh_operation> (); }
711 ;
712
713 exp : exp RSH exp
714 { pstate->wrap2<rsh_operation> (); }
715 ;
716
717 exp : exp EQUAL exp
718 {
719 if (pstate->language ()->la_language
720 == language_opencl)
721 pstate->wrap2<opencl_equal_operation> ();
722 else
723 pstate->wrap2<equal_operation> ();
724 }
725 ;
726
727 exp : exp NOTEQUAL exp
728 {
729 if (pstate->language ()->la_language
730 == language_opencl)
731 pstate->wrap2<opencl_notequal_operation> ();
732 else
733 pstate->wrap2<notequal_operation> ();
734 }
735 ;
736
737 exp : exp LEQ exp
738 {
739 if (pstate->language ()->la_language
740 == language_opencl)
741 pstate->wrap2<opencl_leq_operation> ();
742 else
743 pstate->wrap2<leq_operation> ();
744 }
745 ;
746
747 exp : exp GEQ exp
748 {
749 if (pstate->language ()->la_language
750 == language_opencl)
751 pstate->wrap2<opencl_geq_operation> ();
752 else
753 pstate->wrap2<geq_operation> ();
754 }
755 ;
756
757 exp : exp '<' exp
758 {
759 if (pstate->language ()->la_language
760 == language_opencl)
761 pstate->wrap2<opencl_less_operation> ();
762 else
763 pstate->wrap2<less_operation> ();
764 }
765 ;
766
767 exp : exp '>' exp
768 {
769 if (pstate->language ()->la_language
770 == language_opencl)
771 pstate->wrap2<opencl_gtr_operation> ();
772 else
773 pstate->wrap2<gtr_operation> ();
774 }
775 ;
776
777 exp : exp '&' exp
778 { pstate->wrap2<bitwise_and_operation> (); }
779 ;
780
781 exp : exp '^' exp
782 { pstate->wrap2<bitwise_xor_operation> (); }
783 ;
784
785 exp : exp '|' exp
786 { pstate->wrap2<bitwise_ior_operation> (); }
787 ;
788
789 exp : exp ANDAND exp
790 {
791 if (pstate->language ()->la_language
792 == language_opencl)
793 {
794 operation_up rhs = pstate->pop ();
795 operation_up lhs = pstate->pop ();
796 pstate->push_new<opencl_logical_binop_operation>
797 (BINOP_LOGICAL_AND, std::move (lhs),
798 std::move (rhs));
799 }
800 else
801 pstate->wrap2<logical_and_operation> ();
802 }
803 ;
804
805 exp : exp OROR exp
806 {
807 if (pstate->language ()->la_language
808 == language_opencl)
809 {
810 operation_up rhs = pstate->pop ();
811 operation_up lhs = pstate->pop ();
812 pstate->push_new<opencl_logical_binop_operation>
813 (BINOP_LOGICAL_OR, std::move (lhs),
814 std::move (rhs));
815 }
816 else
817 pstate->wrap2<logical_or_operation> ();
818 }
819 ;
820
821 exp : exp '?' exp ':' exp %prec '?'
822 {
823 operation_up last = pstate->pop ();
824 operation_up mid = pstate->pop ();
825 operation_up first = pstate->pop ();
826 if (pstate->language ()->la_language
827 == language_opencl)
828 pstate->push_new<opencl_ternop_cond_operation>
829 (std::move (first), std::move (mid),
830 std::move (last));
831 else
832 pstate->push_new<ternop_cond_operation>
833 (std::move (first), std::move (mid),
834 std::move (last));
835 }
836 ;
837
838 exp : exp '=' exp
839 {
840 if (pstate->language ()->la_language
841 == language_opencl)
842 pstate->wrap2<opencl_assign_operation> ();
843 else
844 pstate->wrap2<assign_operation> ();
845 }
846 ;
847
848 exp : exp ASSIGN_MODIFY exp
849 {
850 operation_up rhs = pstate->pop ();
851 operation_up lhs = pstate->pop ();
852 pstate->push_new<assign_modify_operation>
853 ($2, std::move (lhs), std::move (rhs));
854 }
855 ;
856
857 exp : INT
858 {
859 pstate->push_new<long_const_operation>
860 ($1.type, $1.val);
861 }
862 ;
863
864 exp : COMPLEX_INT
865 {
866 operation_up real
867 = (make_operation<long_const_operation>
868 (TYPE_TARGET_TYPE ($1.type), 0));
869 operation_up imag
870 = (make_operation<long_const_operation>
871 (TYPE_TARGET_TYPE ($1.type), $1.val));
872 pstate->push_new<complex_operation>
873 (std::move (real), std::move (imag), $1.type);
874 }
875 ;
876
877 exp : CHAR
878 {
879 struct stoken_vector vec;
880 vec.len = 1;
881 vec.tokens = &$1;
882 pstate->push_c_string ($1.type, &vec);
883 }
884 ;
885
886 exp : NAME_OR_INT
887 { YYSTYPE val;
888 parse_number (pstate, $1.stoken.ptr,
889 $1.stoken.length, 0, &val);
890 pstate->push_new<long_const_operation>
891 (val.typed_val_int.type,
892 val.typed_val_int.val);
893 }
894 ;
895
896
897 exp : FLOAT
898 {
899 float_data data;
900 std::copy (std::begin ($1.val), std::end ($1.val),
901 std::begin (data));
902 pstate->push_new<float_const_operation> ($1.type, data);
903 }
904 ;
905
906 exp : COMPLEX_FLOAT
907 {
908 struct type *underlying
909 = TYPE_TARGET_TYPE ($1.type);
910
911 float_data val;
912 target_float_from_host_double (val.data (),
913 underlying, 0);
914 operation_up real
915 = (make_operation<float_const_operation>
916 (underlying, val));
917
918 std::copy (std::begin ($1.val), std::end ($1.val),
919 std::begin (val));
920 operation_up imag
921 = (make_operation<float_const_operation>
922 (underlying, val));
923
924 pstate->push_new<complex_operation>
925 (std::move (real), std::move (imag),
926 $1.type);
927 }
928 ;
929
930 exp : variable
931 ;
932
933 exp : DOLLAR_VARIABLE
934 {
935 pstate->push_dollar ($1);
936 }
937 ;
938
939 exp : SELECTOR '(' name ')'
940 {
941 pstate->push_new<objc_selector_operation>
942 (copy_name ($3));
943 }
944 ;
945
946 exp : SIZEOF '(' type ')' %prec UNARY
947 { struct type *type = $3;
948 struct type *int_type
949 = lookup_signed_typename (pstate->language (),
950 "int");
951 type = check_typedef (type);
952
953 /* $5.3.3/2 of the C++ Standard (n3290 draft)
954 says of sizeof: "When applied to a reference
955 or a reference type, the result is the size of
956 the referenced type." */
957 if (TYPE_IS_REFERENCE (type))
958 type = check_typedef (TYPE_TARGET_TYPE (type));
959 pstate->push_new<long_const_operation>
960 (int_type, TYPE_LENGTH (type));
961 }
962 ;
963
964 exp : REINTERPRET_CAST '<' type_exp '>' '(' exp ')' %prec UNARY
965 { pstate->wrap2<reinterpret_cast_operation> (); }
966 ;
967
968 exp : STATIC_CAST '<' type_exp '>' '(' exp ')' %prec UNARY
969 { pstate->wrap2<unop_cast_type_operation> (); }
970 ;
971
972 exp : DYNAMIC_CAST '<' type_exp '>' '(' exp ')' %prec UNARY
973 { pstate->wrap2<dynamic_cast_operation> (); }
974 ;
975
976 exp : CONST_CAST '<' type_exp '>' '(' exp ')' %prec UNARY
977 { /* We could do more error checking here, but
978 it doesn't seem worthwhile. */
979 pstate->wrap2<unop_cast_type_operation> (); }
980 ;
981
982 string_exp:
983 STRING
984 {
985 /* We copy the string here, and not in the
986 lexer, to guarantee that we do not leak a
987 string. Note that we follow the
988 NUL-termination convention of the
989 lexer. */
990 struct typed_stoken *vec = XNEW (struct typed_stoken);
991 $$.len = 1;
992 $$.tokens = vec;
993
994 vec->type = $1.type;
995 vec->length = $1.length;
996 vec->ptr = (char *) malloc ($1.length + 1);
997 memcpy (vec->ptr, $1.ptr, $1.length + 1);
998 }
999
1000 | string_exp STRING
1001 {
1002 /* Note that we NUL-terminate here, but just
1003 for convenience. */
1004 char *p;
1005 ++$$.len;
1006 $$.tokens = XRESIZEVEC (struct typed_stoken,
1007 $$.tokens, $$.len);
1008
1009 p = (char *) malloc ($2.length + 1);
1010 memcpy (p, $2.ptr, $2.length + 1);
1011
1012 $$.tokens[$$.len - 1].type = $2.type;
1013 $$.tokens[$$.len - 1].length = $2.length;
1014 $$.tokens[$$.len - 1].ptr = p;
1015 }
1016 ;
1017
1018 exp : string_exp
1019 {
1020 int i;
1021 c_string_type type = C_STRING;
1022
1023 for (i = 0; i < $1.len; ++i)
1024 {
1025 switch ($1.tokens[i].type)
1026 {
1027 case C_STRING:
1028 break;
1029 case C_WIDE_STRING:
1030 case C_STRING_16:
1031 case C_STRING_32:
1032 if (type != C_STRING
1033 && type != $1.tokens[i].type)
1034 error (_("Undefined string concatenation."));
1035 type = (enum c_string_type_values) $1.tokens[i].type;
1036 break;
1037 default:
1038 /* internal error */
1039 internal_error (__FILE__, __LINE__,
1040 "unrecognized type in string concatenation");
1041 }
1042 }
1043
1044 pstate->push_c_string (type, &$1);
1045 for (i = 0; i < $1.len; ++i)
1046 free ($1.tokens[i].ptr);
1047 free ($1.tokens);
1048 }
1049 ;
1050
1051 exp : NSSTRING /* ObjC NextStep NSString constant
1052 * of the form '@' '"' string '"'.
1053 */
1054 {
1055 pstate->push_new<objc_nsstring_operation>
1056 (copy_name ($1));
1057 }
1058 ;
1059
1060 /* C++. */
1061 exp : TRUEKEYWORD
1062 { pstate->push_new<long_const_operation>
1063 (parse_type (pstate)->builtin_bool, 1);
1064 }
1065 ;
1066
1067 exp : FALSEKEYWORD
1068 { pstate->push_new<long_const_operation>
1069 (parse_type (pstate)->builtin_bool, 0);
1070 }
1071 ;
1072
1073 /* end of C++. */
1074
1075 block : BLOCKNAME
1076 {
1077 if ($1.sym.symbol)
1078 $$ = SYMBOL_BLOCK_VALUE ($1.sym.symbol);
1079 else
1080 error (_("No file or function \"%s\"."),
1081 copy_name ($1.stoken).c_str ());
1082 }
1083 | FILENAME
1084 {
1085 $$ = $1;
1086 }
1087 ;
1088
1089 block : block COLONCOLON name
1090 {
1091 std::string copy = copy_name ($3);
1092 struct symbol *tem
1093 = lookup_symbol (copy.c_str (), $1,
1094 VAR_DOMAIN, NULL).symbol;
1095
1096 if (!tem || SYMBOL_CLASS (tem) != LOC_BLOCK)
1097 error (_("No function \"%s\" in specified context."),
1098 copy.c_str ());
1099 $$ = SYMBOL_BLOCK_VALUE (tem); }
1100 ;
1101
1102 variable: name_not_typename ENTRY
1103 { struct symbol *sym = $1.sym.symbol;
1104
1105 if (sym == NULL || !SYMBOL_IS_ARGUMENT (sym)
1106 || !symbol_read_needs_frame (sym))
1107 error (_("@entry can be used only for function "
1108 "parameters, not for \"%s\""),
1109 copy_name ($1.stoken).c_str ());
1110
1111 pstate->push_new<var_entry_value_operation> (sym);
1112 }
1113 ;
1114
1115 variable: block COLONCOLON name
1116 {
1117 std::string copy = copy_name ($3);
1118 struct block_symbol sym
1119 = lookup_symbol (copy.c_str (), $1,
1120 VAR_DOMAIN, NULL);
1121
1122 if (sym.symbol == 0)
1123 error (_("No symbol \"%s\" in specified context."),
1124 copy.c_str ());
1125 if (symbol_read_needs_frame (sym.symbol))
1126 pstate->block_tracker->update (sym);
1127
1128 pstate->push_new<var_value_operation> (sym);
1129 }
1130 ;
1131
1132 qualified_name: TYPENAME COLONCOLON name
1133 {
1134 struct type *type = $1.type;
1135 type = check_typedef (type);
1136 if (!type_aggregate_p (type))
1137 error (_("`%s' is not defined as an aggregate type."),
1138 TYPE_SAFE_NAME (type));
1139
1140 pstate->push_new<scope_operation> (type,
1141 copy_name ($3));
1142 }
1143 | TYPENAME COLONCOLON '~' name
1144 {
1145 struct type *type = $1.type;
1146
1147 type = check_typedef (type);
1148 if (!type_aggregate_p (type))
1149 error (_("`%s' is not defined as an aggregate type."),
1150 TYPE_SAFE_NAME (type));
1151 std::string name = "~" + std::string ($4.ptr,
1152 $4.length);
1153
1154 /* Check for valid destructor name. */
1155 destructor_name_p (name.c_str (), $1.type);
1156 pstate->push_new<scope_operation> (type,
1157 std::move (name));
1158 }
1159 | TYPENAME COLONCOLON name COLONCOLON name
1160 {
1161 std::string copy = copy_name ($3);
1162 error (_("No type \"%s\" within class "
1163 "or namespace \"%s\"."),
1164 copy.c_str (), TYPE_SAFE_NAME ($1.type));
1165 }
1166 ;
1167
1168 variable: qualified_name
1169 | COLONCOLON name_not_typename
1170 {
1171 std::string name = copy_name ($2.stoken);
1172 struct block_symbol sym
1173 = lookup_symbol (name.c_str (),
1174 (const struct block *) NULL,
1175 VAR_DOMAIN, NULL);
1176 pstate->push_symbol (name.c_str (), sym);
1177 }
1178 ;
1179
1180 variable: name_not_typename
1181 { struct block_symbol sym = $1.sym;
1182
1183 if (sym.symbol)
1184 {
1185 if (symbol_read_needs_frame (sym.symbol))
1186 pstate->block_tracker->update (sym);
1187
1188 /* If we found a function, see if it's
1189 an ifunc resolver that has the same
1190 address as the ifunc symbol itself.
1191 If so, prefer the ifunc symbol. */
1192
1193 bound_minimal_symbol resolver
1194 = find_gnu_ifunc (sym.symbol);
1195 if (resolver.minsym != NULL)
1196 pstate->push_new<var_msym_value_operation>
1197 (resolver);
1198 else
1199 pstate->push_new<var_value_operation> (sym);
1200 }
1201 else if ($1.is_a_field_of_this)
1202 {
1203 /* C++: it hangs off of `this'. Must
1204 not inadvertently convert from a method call
1205 to data ref. */
1206 pstate->block_tracker->update (sym);
1207 operation_up thisop
1208 = make_operation<op_this_operation> ();
1209 pstate->push_new<structop_ptr_operation>
1210 (std::move (thisop), copy_name ($1.stoken));
1211 }
1212 else
1213 {
1214 std::string arg = copy_name ($1.stoken);
1215
1216 bound_minimal_symbol msymbol
1217 = lookup_bound_minimal_symbol (arg.c_str ());
1218 if (msymbol.minsym == NULL)
1219 {
1220 if (!have_full_symbols () && !have_partial_symbols ())
1221 error (_("No symbol table is loaded. Use the \"file\" command."));
1222 else
1223 error (_("No symbol \"%s\" in current context."),
1224 arg.c_str ());
1225 }
1226
1227 /* This minsym might be an alias for
1228 another function. See if we can find
1229 the debug symbol for the target, and
1230 if so, use it instead, since it has
1231 return type / prototype info. This
1232 is important for example for "p
1233 *__errno_location()". */
1234 symbol *alias_target
1235 = ((msymbol.minsym->type != mst_text_gnu_ifunc
1236 && msymbol.minsym->type != mst_data_gnu_ifunc)
1237 ? find_function_alias_target (msymbol)
1238 : NULL);
1239 if (alias_target != NULL)
1240 {
1241 block_symbol bsym { alias_target,
1242 SYMBOL_BLOCK_VALUE (alias_target) };
1243 pstate->push_new<var_value_operation> (bsym);
1244 }
1245 else
1246 pstate->push_new<var_msym_value_operation>
1247 (msymbol);
1248 }
1249 }
1250 ;
1251
1252 const_or_volatile: const_or_volatile_noopt
1253 |
1254 ;
1255
1256 single_qualifier:
1257 CONST_KEYWORD
1258 { cpstate->type_stack.insert (tp_const); }
1259 | VOLATILE_KEYWORD
1260 { cpstate->type_stack.insert (tp_volatile); }
1261 | ATOMIC
1262 { cpstate->type_stack.insert (tp_atomic); }
1263 | RESTRICT
1264 { cpstate->type_stack.insert (tp_restrict); }
1265 | '@' NAME
1266 {
1267 cpstate->type_stack.insert (pstate,
1268 copy_name ($2.stoken).c_str ());
1269 }
1270 ;
1271
1272 qualifier_seq_noopt:
1273 single_qualifier
1274 | qualifier_seq single_qualifier
1275 ;
1276
1277 qualifier_seq:
1278 qualifier_seq_noopt
1279 |
1280 ;
1281
1282 ptr_operator:
1283 ptr_operator '*'
1284 { cpstate->type_stack.insert (tp_pointer); }
1285 qualifier_seq
1286 | '*'
1287 { cpstate->type_stack.insert (tp_pointer); }
1288 qualifier_seq
1289 | '&'
1290 { cpstate->type_stack.insert (tp_reference); }
1291 | '&' ptr_operator
1292 { cpstate->type_stack.insert (tp_reference); }
1293 | ANDAND
1294 { cpstate->type_stack.insert (tp_rvalue_reference); }
1295 | ANDAND ptr_operator
1296 { cpstate->type_stack.insert (tp_rvalue_reference); }
1297 ;
1298
1299 ptr_operator_ts: ptr_operator
1300 {
1301 $$ = cpstate->type_stack.create ();
1302 cpstate->type_stacks.emplace_back ($$);
1303 }
1304 ;
1305
1306 abs_decl: ptr_operator_ts direct_abs_decl
1307 { $$ = $2->append ($1); }
1308 | ptr_operator_ts
1309 | direct_abs_decl
1310 ;
1311
1312 direct_abs_decl: '(' abs_decl ')'
1313 { $$ = $2; }
1314 | direct_abs_decl array_mod
1315 {
1316 cpstate->type_stack.push ($1);
1317 cpstate->type_stack.push ($2);
1318 cpstate->type_stack.push (tp_array);
1319 $$ = cpstate->type_stack.create ();
1320 cpstate->type_stacks.emplace_back ($$);
1321 }
1322 | array_mod
1323 {
1324 cpstate->type_stack.push ($1);
1325 cpstate->type_stack.push (tp_array);
1326 $$ = cpstate->type_stack.create ();
1327 cpstate->type_stacks.emplace_back ($$);
1328 }
1329
1330 | direct_abs_decl func_mod
1331 {
1332 cpstate->type_stack.push ($1);
1333 cpstate->type_stack.push ($2);
1334 $$ = cpstate->type_stack.create ();
1335 cpstate->type_stacks.emplace_back ($$);
1336 }
1337 | func_mod
1338 {
1339 cpstate->type_stack.push ($1);
1340 $$ = cpstate->type_stack.create ();
1341 cpstate->type_stacks.emplace_back ($$);
1342 }
1343 ;
1344
1345 array_mod: '[' ']'
1346 { $$ = -1; }
1347 | OBJC_LBRAC ']'
1348 { $$ = -1; }
1349 | '[' INT ']'
1350 { $$ = $2.val; }
1351 | OBJC_LBRAC INT ']'
1352 { $$ = $2.val; }
1353 ;
1354
1355 func_mod: '(' ')'
1356 {
1357 $$ = new std::vector<struct type *>;
1358 cpstate->type_lists.emplace_back ($$);
1359 }
1360 | '(' parameter_typelist ')'
1361 { $$ = $2; }
1362 ;
1363
1364 /* We used to try to recognize pointer to member types here, but
1365 that didn't work (shift/reduce conflicts meant that these rules never
1366 got executed). The problem is that
1367 int (foo::bar::baz::bizzle)
1368 is a function type but
1369 int (foo::bar::baz::bizzle::*)
1370 is a pointer to member type. Stroustrup loses again! */
1371
1372 type : ptype
1373 ;
1374
1375 /* A helper production that recognizes scalar types that can validly
1376 be used with _Complex. */
1377
1378 scalar_type:
1379 INT_KEYWORD
1380 { $$ = lookup_signed_typename (pstate->language (),
1381 "int"); }
1382 | LONG
1383 { $$ = lookup_signed_typename (pstate->language (),
1384 "long"); }
1385 | SHORT
1386 { $$ = lookup_signed_typename (pstate->language (),
1387 "short"); }
1388 | LONG INT_KEYWORD
1389 { $$ = lookup_signed_typename (pstate->language (),
1390 "long"); }
1391 | LONG SIGNED_KEYWORD INT_KEYWORD
1392 { $$ = lookup_signed_typename (pstate->language (),
1393 "long"); }
1394 | LONG SIGNED_KEYWORD
1395 { $$ = lookup_signed_typename (pstate->language (),
1396 "long"); }
1397 | SIGNED_KEYWORD LONG INT_KEYWORD
1398 { $$ = lookup_signed_typename (pstate->language (),
1399 "long"); }
1400 | UNSIGNED LONG INT_KEYWORD
1401 { $$ = lookup_unsigned_typename (pstate->language (),
1402 "long"); }
1403 | LONG UNSIGNED INT_KEYWORD
1404 { $$ = lookup_unsigned_typename (pstate->language (),
1405 "long"); }
1406 | LONG UNSIGNED
1407 { $$ = lookup_unsigned_typename (pstate->language (),
1408 "long"); }
1409 | LONG LONG
1410 { $$ = lookup_signed_typename (pstate->language (),
1411 "long long"); }
1412 | LONG LONG INT_KEYWORD
1413 { $$ = lookup_signed_typename (pstate->language (),
1414 "long long"); }
1415 | LONG LONG SIGNED_KEYWORD INT_KEYWORD
1416 { $$ = lookup_signed_typename (pstate->language (),
1417 "long long"); }
1418 | LONG LONG SIGNED_KEYWORD
1419 { $$ = lookup_signed_typename (pstate->language (),
1420 "long long"); }
1421 | SIGNED_KEYWORD LONG LONG
1422 { $$ = lookup_signed_typename (pstate->language (),
1423 "long long"); }
1424 | SIGNED_KEYWORD LONG LONG INT_KEYWORD
1425 { $$ = lookup_signed_typename (pstate->language (),
1426 "long long"); }
1427 | UNSIGNED LONG LONG
1428 { $$ = lookup_unsigned_typename (pstate->language (),
1429 "long long"); }
1430 | UNSIGNED LONG LONG INT_KEYWORD
1431 { $$ = lookup_unsigned_typename (pstate->language (),
1432 "long long"); }
1433 | LONG LONG UNSIGNED
1434 { $$ = lookup_unsigned_typename (pstate->language (),
1435 "long long"); }
1436 | LONG LONG UNSIGNED INT_KEYWORD
1437 { $$ = lookup_unsigned_typename (pstate->language (),
1438 "long long"); }
1439 | SHORT INT_KEYWORD
1440 { $$ = lookup_signed_typename (pstate->language (),
1441 "short"); }
1442 | SHORT SIGNED_KEYWORD INT_KEYWORD
1443 { $$ = lookup_signed_typename (pstate->language (),
1444 "short"); }
1445 | SHORT SIGNED_KEYWORD
1446 { $$ = lookup_signed_typename (pstate->language (),
1447 "short"); }
1448 | UNSIGNED SHORT INT_KEYWORD
1449 { $$ = lookup_unsigned_typename (pstate->language (),
1450 "short"); }
1451 | SHORT UNSIGNED
1452 { $$ = lookup_unsigned_typename (pstate->language (),
1453 "short"); }
1454 | SHORT UNSIGNED INT_KEYWORD
1455 { $$ = lookup_unsigned_typename (pstate->language (),
1456 "short"); }
1457 | DOUBLE_KEYWORD
1458 { $$ = lookup_typename (pstate->language (),
1459 "double",
1460 NULL,
1461 0); }
1462 | FLOAT_KEYWORD
1463 { $$ = lookup_typename (pstate->language (),
1464 "float",
1465 NULL,
1466 0); }
1467 | LONG DOUBLE_KEYWORD
1468 { $$ = lookup_typename (pstate->language (),
1469 "long double",
1470 NULL,
1471 0); }
1472 | UNSIGNED type_name
1473 { $$ = lookup_unsigned_typename (pstate->language (),
1474 $2.type->name ()); }
1475 | UNSIGNED
1476 { $$ = lookup_unsigned_typename (pstate->language (),
1477 "int"); }
1478 | SIGNED_KEYWORD type_name
1479 { $$ = lookup_signed_typename (pstate->language (),
1480 $2.type->name ()); }
1481 | SIGNED_KEYWORD
1482 { $$ = lookup_signed_typename (pstate->language (),
1483 "int"); }
1484 ;
1485
1486 /* Implements (approximately): (type-qualifier)* type-specifier.
1487
1488 When type-specifier is only ever a single word, like 'float' then these
1489 arrive as pre-built TYPENAME tokens thanks to the classify_name
1490 function. However, when a type-specifier can contain multiple words,
1491 for example 'double' can appear as just 'double' or 'long double', and
1492 similarly 'long' can appear as just 'long' or in 'long double', then
1493 these type-specifiers are parsed into their own tokens in the function
1494 lex_one_token and the ident_tokens array. These separate tokens are all
1495 recognised here. */
1496 typebase
1497 : TYPENAME
1498 { $$ = $1.type; }
1499 | scalar_type
1500 { $$ = $1; }
1501 | COMPLEX scalar_type
1502 {
1503 $$ = init_complex_type (nullptr, $2);
1504 }
1505 | STRUCT name
1506 { $$
1507 = lookup_struct (copy_name ($2).c_str (),
1508 pstate->expression_context_block);
1509 }
1510 | STRUCT COMPLETE
1511 {
1512 pstate->mark_completion_tag (TYPE_CODE_STRUCT,
1513 "", 0);
1514 $$ = NULL;
1515 }
1516 | STRUCT name COMPLETE
1517 {
1518 pstate->mark_completion_tag (TYPE_CODE_STRUCT,
1519 $2.ptr, $2.length);
1520 $$ = NULL;
1521 }
1522 | CLASS name
1523 { $$ = lookup_struct
1524 (copy_name ($2).c_str (),
1525 pstate->expression_context_block);
1526 }
1527 | CLASS COMPLETE
1528 {
1529 pstate->mark_completion_tag (TYPE_CODE_STRUCT,
1530 "", 0);
1531 $$ = NULL;
1532 }
1533 | CLASS name COMPLETE
1534 {
1535 pstate->mark_completion_tag (TYPE_CODE_STRUCT,
1536 $2.ptr, $2.length);
1537 $$ = NULL;
1538 }
1539 | UNION name
1540 { $$
1541 = lookup_union (copy_name ($2).c_str (),
1542 pstate->expression_context_block);
1543 }
1544 | UNION COMPLETE
1545 {
1546 pstate->mark_completion_tag (TYPE_CODE_UNION,
1547 "", 0);
1548 $$ = NULL;
1549 }
1550 | UNION name COMPLETE
1551 {
1552 pstate->mark_completion_tag (TYPE_CODE_UNION,
1553 $2.ptr, $2.length);
1554 $$ = NULL;
1555 }
1556 | ENUM name
1557 { $$ = lookup_enum (copy_name ($2).c_str (),
1558 pstate->expression_context_block);
1559 }
1560 | ENUM COMPLETE
1561 {
1562 pstate->mark_completion_tag (TYPE_CODE_ENUM, "", 0);
1563 $$ = NULL;
1564 }
1565 | ENUM name COMPLETE
1566 {
1567 pstate->mark_completion_tag (TYPE_CODE_ENUM, $2.ptr,
1568 $2.length);
1569 $$ = NULL;
1570 }
1571 /* It appears that this rule for templates is never
1572 reduced; template recognition happens by lookahead
1573 in the token processing code in yylex. */
1574 | TEMPLATE name '<' type '>'
1575 { $$ = lookup_template_type
1576 (copy_name($2).c_str (), $4,
1577 pstate->expression_context_block);
1578 }
1579 | qualifier_seq_noopt typebase
1580 { $$ = cpstate->type_stack.follow_types ($2); }
1581 | typebase qualifier_seq_noopt
1582 { $$ = cpstate->type_stack.follow_types ($1); }
1583 ;
1584
1585 type_name: TYPENAME
1586 | INT_KEYWORD
1587 {
1588 $$.stoken.ptr = "int";
1589 $$.stoken.length = 3;
1590 $$.type = lookup_signed_typename (pstate->language (),
1591 "int");
1592 }
1593 | LONG
1594 {
1595 $$.stoken.ptr = "long";
1596 $$.stoken.length = 4;
1597 $$.type = lookup_signed_typename (pstate->language (),
1598 "long");
1599 }
1600 | SHORT
1601 {
1602 $$.stoken.ptr = "short";
1603 $$.stoken.length = 5;
1604 $$.type = lookup_signed_typename (pstate->language (),
1605 "short");
1606 }
1607 ;
1608
1609 parameter_typelist:
1610 nonempty_typelist
1611 { check_parameter_typelist ($1); }
1612 | nonempty_typelist ',' DOTDOTDOT
1613 {
1614 $1->push_back (NULL);
1615 check_parameter_typelist ($1);
1616 $$ = $1;
1617 }
1618 ;
1619
1620 nonempty_typelist
1621 : type
1622 {
1623 std::vector<struct type *> *typelist
1624 = new std::vector<struct type *>;
1625 cpstate->type_lists.emplace_back (typelist);
1626
1627 typelist->push_back ($1);
1628 $$ = typelist;
1629 }
1630 | nonempty_typelist ',' type
1631 {
1632 $1->push_back ($3);
1633 $$ = $1;
1634 }
1635 ;
1636
1637 ptype : typebase
1638 | ptype abs_decl
1639 {
1640 cpstate->type_stack.push ($2);
1641 $$ = cpstate->type_stack.follow_types ($1);
1642 }
1643 ;
1644
1645 conversion_type_id: typebase conversion_declarator
1646 { $$ = cpstate->type_stack.follow_types ($1); }
1647 ;
1648
1649 conversion_declarator: /* Nothing. */
1650 | ptr_operator conversion_declarator
1651 ;
1652
1653 const_and_volatile: CONST_KEYWORD VOLATILE_KEYWORD
1654 | VOLATILE_KEYWORD CONST_KEYWORD
1655 ;
1656
1657 const_or_volatile_noopt: const_and_volatile
1658 { cpstate->type_stack.insert (tp_const);
1659 cpstate->type_stack.insert (tp_volatile);
1660 }
1661 | CONST_KEYWORD
1662 { cpstate->type_stack.insert (tp_const); }
1663 | VOLATILE_KEYWORD
1664 { cpstate->type_stack.insert (tp_volatile); }
1665 ;
1666
1667 oper: OPERATOR NEW
1668 { $$ = operator_stoken (" new"); }
1669 | OPERATOR DELETE
1670 { $$ = operator_stoken (" delete"); }
1671 | OPERATOR NEW '[' ']'
1672 { $$ = operator_stoken (" new[]"); }
1673 | OPERATOR DELETE '[' ']'
1674 { $$ = operator_stoken (" delete[]"); }
1675 | OPERATOR NEW OBJC_LBRAC ']'
1676 { $$ = operator_stoken (" new[]"); }
1677 | OPERATOR DELETE OBJC_LBRAC ']'
1678 { $$ = operator_stoken (" delete[]"); }
1679 | OPERATOR '+'
1680 { $$ = operator_stoken ("+"); }
1681 | OPERATOR '-'
1682 { $$ = operator_stoken ("-"); }
1683 | OPERATOR '*'
1684 { $$ = operator_stoken ("*"); }
1685 | OPERATOR '/'
1686 { $$ = operator_stoken ("/"); }
1687 | OPERATOR '%'
1688 { $$ = operator_stoken ("%"); }
1689 | OPERATOR '^'
1690 { $$ = operator_stoken ("^"); }
1691 | OPERATOR '&'
1692 { $$ = operator_stoken ("&"); }
1693 | OPERATOR '|'
1694 { $$ = operator_stoken ("|"); }
1695 | OPERATOR '~'
1696 { $$ = operator_stoken ("~"); }
1697 | OPERATOR '!'
1698 { $$ = operator_stoken ("!"); }
1699 | OPERATOR '='
1700 { $$ = operator_stoken ("="); }
1701 | OPERATOR '<'
1702 { $$ = operator_stoken ("<"); }
1703 | OPERATOR '>'
1704 { $$ = operator_stoken (">"); }
1705 | OPERATOR ASSIGN_MODIFY
1706 { const char *op = " unknown";
1707 switch ($2)
1708 {
1709 case BINOP_RSH:
1710 op = ">>=";
1711 break;
1712 case BINOP_LSH:
1713 op = "<<=";
1714 break;
1715 case BINOP_ADD:
1716 op = "+=";
1717 break;
1718 case BINOP_SUB:
1719 op = "-=";
1720 break;
1721 case BINOP_MUL:
1722 op = "*=";
1723 break;
1724 case BINOP_DIV:
1725 op = "/=";
1726 break;
1727 case BINOP_REM:
1728 op = "%=";
1729 break;
1730 case BINOP_BITWISE_IOR:
1731 op = "|=";
1732 break;
1733 case BINOP_BITWISE_AND:
1734 op = "&=";
1735 break;
1736 case BINOP_BITWISE_XOR:
1737 op = "^=";
1738 break;
1739 default:
1740 break;
1741 }
1742
1743 $$ = operator_stoken (op);
1744 }
1745 | OPERATOR LSH
1746 { $$ = operator_stoken ("<<"); }
1747 | OPERATOR RSH
1748 { $$ = operator_stoken (">>"); }
1749 | OPERATOR EQUAL
1750 { $$ = operator_stoken ("=="); }
1751 | OPERATOR NOTEQUAL
1752 { $$ = operator_stoken ("!="); }
1753 | OPERATOR LEQ
1754 { $$ = operator_stoken ("<="); }
1755 | OPERATOR GEQ
1756 { $$ = operator_stoken (">="); }
1757 | OPERATOR ANDAND
1758 { $$ = operator_stoken ("&&"); }
1759 | OPERATOR OROR
1760 { $$ = operator_stoken ("||"); }
1761 | OPERATOR INCREMENT
1762 { $$ = operator_stoken ("++"); }
1763 | OPERATOR DECREMENT
1764 { $$ = operator_stoken ("--"); }
1765 | OPERATOR ','
1766 { $$ = operator_stoken (","); }
1767 | OPERATOR ARROW_STAR
1768 { $$ = operator_stoken ("->*"); }
1769 | OPERATOR ARROW
1770 { $$ = operator_stoken ("->"); }
1771 | OPERATOR '(' ')'
1772 { $$ = operator_stoken ("()"); }
1773 | OPERATOR '[' ']'
1774 { $$ = operator_stoken ("[]"); }
1775 | OPERATOR OBJC_LBRAC ']'
1776 { $$ = operator_stoken ("[]"); }
1777 | OPERATOR conversion_type_id
1778 { string_file buf;
1779
1780 c_print_type ($2, NULL, &buf, -1, 0,
1781 &type_print_raw_options);
1782 std::string name = std::move (buf.string ());
1783
1784 /* This also needs canonicalization. */
1785 gdb::unique_xmalloc_ptr<char> canon
1786 = cp_canonicalize_string (name.c_str ());
1787 if (canon != nullptr)
1788 name = canon.get ();
1789 $$ = operator_stoken ((" " + name).c_str ());
1790 }
1791 ;
1792
1793 /* This rule exists in order to allow some tokens that would not normally
1794 match the 'name' rule to appear as fields within a struct. The example
1795 that initially motivated this was the RISC-V target which models the
1796 floating point registers as a union with fields called 'float' and
1797 'double'. */
1798 field_name
1799 : name
1800 | DOUBLE_KEYWORD { $$ = typename_stoken ("double"); }
1801 | FLOAT_KEYWORD { $$ = typename_stoken ("float"); }
1802 | INT_KEYWORD { $$ = typename_stoken ("int"); }
1803 | LONG { $$ = typename_stoken ("long"); }
1804 | SHORT { $$ = typename_stoken ("short"); }
1805 | SIGNED_KEYWORD { $$ = typename_stoken ("signed"); }
1806 | UNSIGNED { $$ = typename_stoken ("unsigned"); }
1807 ;
1808
1809 name : NAME { $$ = $1.stoken; }
1810 | BLOCKNAME { $$ = $1.stoken; }
1811 | TYPENAME { $$ = $1.stoken; }
1812 | NAME_OR_INT { $$ = $1.stoken; }
1813 | UNKNOWN_CPP_NAME { $$ = $1.stoken; }
1814 | oper { $$ = $1; }
1815 ;
1816
1817 name_not_typename : NAME
1818 | BLOCKNAME
1819 /* These would be useful if name_not_typename was useful, but it is just
1820 a fake for "variable", so these cause reduce/reduce conflicts because
1821 the parser can't tell whether NAME_OR_INT is a name_not_typename (=variable,
1822 =exp) or just an exp. If name_not_typename was ever used in an lvalue
1823 context where only a name could occur, this might be useful.
1824 | NAME_OR_INT
1825 */
1826 | oper
1827 {
1828 struct field_of_this_result is_a_field_of_this;
1829
1830 $$.stoken = $1;
1831 $$.sym
1832 = lookup_symbol ($1.ptr,
1833 pstate->expression_context_block,
1834 VAR_DOMAIN,
1835 &is_a_field_of_this);
1836 $$.is_a_field_of_this
1837 = is_a_field_of_this.type != NULL;
1838 }
1839 | UNKNOWN_CPP_NAME
1840 ;
1841
1842 %%
1843
1844 /* Returns a stoken of the operator name given by OP (which does not
1845 include the string "operator"). */
1846
1847 static struct stoken
1848 operator_stoken (const char *op)
1849 {
1850 struct stoken st = { NULL, 0 };
1851 char *buf;
1852
1853 st.length = CP_OPERATOR_LEN + strlen (op);
1854 buf = (char *) malloc (st.length + 1);
1855 strcpy (buf, CP_OPERATOR_STR);
1856 strcat (buf, op);
1857 st.ptr = buf;
1858
1859 /* The toplevel (c_parse) will free the memory allocated here. */
1860 cpstate->strings.emplace_back (buf);
1861 return st;
1862 };
1863
1864 /* Returns a stoken of the type named TYPE. */
1865
1866 static struct stoken
1867 typename_stoken (const char *type)
1868 {
1869 struct stoken st = { type, 0 };
1870 st.length = strlen (type);
1871 return st;
1872 };
1873
1874 /* Return true if the type is aggregate-like. */
1875
1876 static int
1877 type_aggregate_p (struct type *type)
1878 {
1879 return (type->code () == TYPE_CODE_STRUCT
1880 || type->code () == TYPE_CODE_UNION
1881 || type->code () == TYPE_CODE_NAMESPACE
1882 || (type->code () == TYPE_CODE_ENUM
1883 && type->is_declared_class ()));
1884 }
1885
1886 /* Validate a parameter typelist. */
1887
1888 static void
1889 check_parameter_typelist (std::vector<struct type *> *params)
1890 {
1891 struct type *type;
1892 int ix;
1893
1894 for (ix = 0; ix < params->size (); ++ix)
1895 {
1896 type = (*params)[ix];
1897 if (type != NULL && check_typedef (type)->code () == TYPE_CODE_VOID)
1898 {
1899 if (ix == 0)
1900 {
1901 if (params->size () == 1)
1902 {
1903 /* Ok. */
1904 break;
1905 }
1906 error (_("parameter types following 'void'"));
1907 }
1908 else
1909 error (_("'void' invalid as parameter type"));
1910 }
1911 }
1912 }
1913
1914 /* Take care of parsing a number (anything that starts with a digit).
1915 Set yylval and return the token type; update lexptr.
1916 LEN is the number of characters in it. */
1917
1918 /*** Needs some error checking for the float case ***/
1919
1920 static int
1921 parse_number (struct parser_state *par_state,
1922 const char *buf, int len, int parsed_float, YYSTYPE *putithere)
1923 {
1924 ULONGEST n = 0;
1925 ULONGEST prevn = 0;
1926 ULONGEST un;
1927
1928 int i = 0;
1929 int c;
1930 int base = input_radix;
1931 int unsigned_p = 0;
1932
1933 /* Number of "L" suffixes encountered. */
1934 int long_p = 0;
1935
1936 /* Imaginary number. */
1937 bool imaginary_p = false;
1938
1939 /* We have found a "L" or "U" (or "i") suffix. */
1940 int found_suffix = 0;
1941
1942 ULONGEST high_bit;
1943 struct type *signed_type;
1944 struct type *unsigned_type;
1945 char *p;
1946
1947 p = (char *) alloca (len);
1948 memcpy (p, buf, len);
1949
1950 if (parsed_float)
1951 {
1952 if (len >= 1 && p[len - 1] == 'i')
1953 {
1954 imaginary_p = true;
1955 --len;
1956 }
1957
1958 /* Handle suffixes for decimal floating-point: "df", "dd" or "dl". */
1959 if (len >= 2 && p[len - 2] == 'd' && p[len - 1] == 'f')
1960 {
1961 putithere->typed_val_float.type
1962 = parse_type (par_state)->builtin_decfloat;
1963 len -= 2;
1964 }
1965 else if (len >= 2 && p[len - 2] == 'd' && p[len - 1] == 'd')
1966 {
1967 putithere->typed_val_float.type
1968 = parse_type (par_state)->builtin_decdouble;
1969 len -= 2;
1970 }
1971 else if (len >= 2 && p[len - 2] == 'd' && p[len - 1] == 'l')
1972 {
1973 putithere->typed_val_float.type
1974 = parse_type (par_state)->builtin_declong;
1975 len -= 2;
1976 }
1977 /* Handle suffixes: 'f' for float, 'l' for long double. */
1978 else if (len >= 1 && TOLOWER (p[len - 1]) == 'f')
1979 {
1980 putithere->typed_val_float.type
1981 = parse_type (par_state)->builtin_float;
1982 len -= 1;
1983 }
1984 else if (len >= 1 && TOLOWER (p[len - 1]) == 'l')
1985 {
1986 putithere->typed_val_float.type
1987 = parse_type (par_state)->builtin_long_double;
1988 len -= 1;
1989 }
1990 /* Default type for floating-point literals is double. */
1991 else
1992 {
1993 putithere->typed_val_float.type
1994 = parse_type (par_state)->builtin_double;
1995 }
1996
1997 if (!parse_float (p, len,
1998 putithere->typed_val_float.type,
1999 putithere->typed_val_float.val))
2000 return ERROR;
2001
2002 if (imaginary_p)
2003 putithere->typed_val_float.type
2004 = init_complex_type (nullptr, putithere->typed_val_float.type);
2005
2006 return imaginary_p ? COMPLEX_FLOAT : FLOAT;
2007 }
2008
2009 /* Handle base-switching prefixes 0x, 0t, 0d, 0 */
2010 if (p[0] == '0' && len > 1)
2011 switch (p[1])
2012 {
2013 case 'x':
2014 case 'X':
2015 if (len >= 3)
2016 {
2017 p += 2;
2018 base = 16;
2019 len -= 2;
2020 }
2021 break;
2022
2023 case 'b':
2024 case 'B':
2025 if (len >= 3)
2026 {
2027 p += 2;
2028 base = 2;
2029 len -= 2;
2030 }
2031 break;
2032
2033 case 't':
2034 case 'T':
2035 case 'd':
2036 case 'D':
2037 if (len >= 3)
2038 {
2039 p += 2;
2040 base = 10;
2041 len -= 2;
2042 }
2043 break;
2044
2045 default:
2046 base = 8;
2047 break;
2048 }
2049
2050 while (len-- > 0)
2051 {
2052 c = *p++;
2053 if (c >= 'A' && c <= 'Z')
2054 c += 'a' - 'A';
2055 if (c != 'l' && c != 'u' && c != 'i')
2056 n *= base;
2057 if (c >= '0' && c <= '9')
2058 {
2059 if (found_suffix)
2060 return ERROR;
2061 n += i = c - '0';
2062 }
2063 else
2064 {
2065 if (base > 10 && c >= 'a' && c <= 'f')
2066 {
2067 if (found_suffix)
2068 return ERROR;
2069 n += i = c - 'a' + 10;
2070 }
2071 else if (c == 'l')
2072 {
2073 ++long_p;
2074 found_suffix = 1;
2075 }
2076 else if (c == 'u')
2077 {
2078 unsigned_p = 1;
2079 found_suffix = 1;
2080 }
2081 else if (c == 'i')
2082 {
2083 imaginary_p = true;
2084 found_suffix = 1;
2085 }
2086 else
2087 return ERROR; /* Char not a digit */
2088 }
2089 if (i >= base)
2090 return ERROR; /* Invalid digit in this base */
2091
2092 /* Portably test for overflow (only works for nonzero values, so make
2093 a second check for zero). FIXME: Can't we just make n and prevn
2094 unsigned and avoid this? */
2095 if (c != 'l' && c != 'u' && c != 'i' && (prevn >= n) && n != 0)
2096 unsigned_p = 1; /* Try something unsigned */
2097
2098 /* Portably test for unsigned overflow.
2099 FIXME: This check is wrong; for example it doesn't find overflow
2100 on 0x123456789 when LONGEST is 32 bits. */
2101 if (c != 'l' && c != 'u' && c != 'i' && n != 0)
2102 {
2103 if (unsigned_p && prevn >= n)
2104 error (_("Numeric constant too large."));
2105 }
2106 prevn = n;
2107 }
2108
2109 /* An integer constant is an int, a long, or a long long. An L
2110 suffix forces it to be long; an LL suffix forces it to be long
2111 long. If not forced to a larger size, it gets the first type of
2112 the above that it fits in. To figure out whether it fits, we
2113 shift it right and see whether anything remains. Note that we
2114 can't shift sizeof (LONGEST) * HOST_CHAR_BIT bits or more in one
2115 operation, because many compilers will warn about such a shift
2116 (which always produces a zero result). Sometimes gdbarch_int_bit
2117 or gdbarch_long_bit will be that big, sometimes not. To deal with
2118 the case where it is we just always shift the value more than
2119 once, with fewer bits each time. */
2120
2121 un = n >> 2;
2122 if (long_p == 0
2123 && (un >> (gdbarch_int_bit (par_state->gdbarch ()) - 2)) == 0)
2124 {
2125 high_bit
2126 = ((ULONGEST)1) << (gdbarch_int_bit (par_state->gdbarch ()) - 1);
2127
2128 /* A large decimal (not hex or octal) constant (between INT_MAX
2129 and UINT_MAX) is a long or unsigned long, according to ANSI,
2130 never an unsigned int, but this code treats it as unsigned
2131 int. This probably should be fixed. GCC gives a warning on
2132 such constants. */
2133
2134 unsigned_type = parse_type (par_state)->builtin_unsigned_int;
2135 signed_type = parse_type (par_state)->builtin_int;
2136 }
2137 else if (long_p <= 1
2138 && (un >> (gdbarch_long_bit (par_state->gdbarch ()) - 2)) == 0)
2139 {
2140 high_bit
2141 = ((ULONGEST)1) << (gdbarch_long_bit (par_state->gdbarch ()) - 1);
2142 unsigned_type = parse_type (par_state)->builtin_unsigned_long;
2143 signed_type = parse_type (par_state)->builtin_long;
2144 }
2145 else
2146 {
2147 int shift;
2148 if (sizeof (ULONGEST) * HOST_CHAR_BIT
2149 < gdbarch_long_long_bit (par_state->gdbarch ()))
2150 /* A long long does not fit in a LONGEST. */
2151 shift = (sizeof (ULONGEST) * HOST_CHAR_BIT - 1);
2152 else
2153 shift = (gdbarch_long_long_bit (par_state->gdbarch ()) - 1);
2154 high_bit = (ULONGEST) 1 << shift;
2155 unsigned_type = parse_type (par_state)->builtin_unsigned_long_long;
2156 signed_type = parse_type (par_state)->builtin_long_long;
2157 }
2158
2159 putithere->typed_val_int.val = n;
2160
2161 /* If the high bit of the worked out type is set then this number
2162 has to be unsigned. */
2163
2164 if (unsigned_p || (n & high_bit))
2165 {
2166 putithere->typed_val_int.type = unsigned_type;
2167 }
2168 else
2169 {
2170 putithere->typed_val_int.type = signed_type;
2171 }
2172
2173 if (imaginary_p)
2174 putithere->typed_val_int.type
2175 = init_complex_type (nullptr, putithere->typed_val_int.type);
2176
2177 return imaginary_p ? COMPLEX_INT : INT;
2178 }
2179
2180 /* Temporary obstack used for holding strings. */
2181 static struct obstack tempbuf;
2182 static int tempbuf_init;
2183
2184 /* Parse a C escape sequence. The initial backslash of the sequence
2185 is at (*PTR)[-1]. *PTR will be updated to point to just after the
2186 last character of the sequence. If OUTPUT is not NULL, the
2187 translated form of the escape sequence will be written there. If
2188 OUTPUT is NULL, no output is written and the call will only affect
2189 *PTR. If an escape sequence is expressed in target bytes, then the
2190 entire sequence will simply be copied to OUTPUT. Return 1 if any
2191 character was emitted, 0 otherwise. */
2192
2193 int
2194 c_parse_escape (const char **ptr, struct obstack *output)
2195 {
2196 const char *tokptr = *ptr;
2197 int result = 1;
2198
2199 /* Some escape sequences undergo character set conversion. Those we
2200 translate here. */
2201 switch (*tokptr)
2202 {
2203 /* Hex escapes do not undergo character set conversion, so keep
2204 the escape sequence for later. */
2205 case 'x':
2206 if (output)
2207 obstack_grow_str (output, "\\x");
2208 ++tokptr;
2209 if (!ISXDIGIT (*tokptr))
2210 error (_("\\x escape without a following hex digit"));
2211 while (ISXDIGIT (*tokptr))
2212 {
2213 if (output)
2214 obstack_1grow (output, *tokptr);
2215 ++tokptr;
2216 }
2217 break;
2218
2219 /* Octal escapes do not undergo character set conversion, so
2220 keep the escape sequence for later. */
2221 case '0':
2222 case '1':
2223 case '2':
2224 case '3':
2225 case '4':
2226 case '5':
2227 case '6':
2228 case '7':
2229 {
2230 int i;
2231 if (output)
2232 obstack_grow_str (output, "\\");
2233 for (i = 0;
2234 i < 3 && ISDIGIT (*tokptr) && *tokptr != '8' && *tokptr != '9';
2235 ++i)
2236 {
2237 if (output)
2238 obstack_1grow (output, *tokptr);
2239 ++tokptr;
2240 }
2241 }
2242 break;
2243
2244 /* We handle UCNs later. We could handle them here, but that
2245 would mean a spurious error in the case where the UCN could
2246 be converted to the target charset but not the host
2247 charset. */
2248 case 'u':
2249 case 'U':
2250 {
2251 char c = *tokptr;
2252 int i, len = c == 'U' ? 8 : 4;
2253 if (output)
2254 {
2255 obstack_1grow (output, '\\');
2256 obstack_1grow (output, *tokptr);
2257 }
2258 ++tokptr;
2259 if (!ISXDIGIT (*tokptr))
2260 error (_("\\%c escape without a following hex digit"), c);
2261 for (i = 0; i < len && ISXDIGIT (*tokptr); ++i)
2262 {
2263 if (output)
2264 obstack_1grow (output, *tokptr);
2265 ++tokptr;
2266 }
2267 }
2268 break;
2269
2270 /* We must pass backslash through so that it does not
2271 cause quoting during the second expansion. */
2272 case '\\':
2273 if (output)
2274 obstack_grow_str (output, "\\\\");
2275 ++tokptr;
2276 break;
2277
2278 /* Escapes which undergo conversion. */
2279 case 'a':
2280 if (output)
2281 obstack_1grow (output, '\a');
2282 ++tokptr;
2283 break;
2284 case 'b':
2285 if (output)
2286 obstack_1grow (output, '\b');
2287 ++tokptr;
2288 break;
2289 case 'f':
2290 if (output)
2291 obstack_1grow (output, '\f');
2292 ++tokptr;
2293 break;
2294 case 'n':
2295 if (output)
2296 obstack_1grow (output, '\n');
2297 ++tokptr;
2298 break;
2299 case 'r':
2300 if (output)
2301 obstack_1grow (output, '\r');
2302 ++tokptr;
2303 break;
2304 case 't':
2305 if (output)
2306 obstack_1grow (output, '\t');
2307 ++tokptr;
2308 break;
2309 case 'v':
2310 if (output)
2311 obstack_1grow (output, '\v');
2312 ++tokptr;
2313 break;
2314
2315 /* GCC extension. */
2316 case 'e':
2317 if (output)
2318 obstack_1grow (output, HOST_ESCAPE_CHAR);
2319 ++tokptr;
2320 break;
2321
2322 /* Backslash-newline expands to nothing at all. */
2323 case '\n':
2324 ++tokptr;
2325 result = 0;
2326 break;
2327
2328 /* A few escapes just expand to the character itself. */
2329 case '\'':
2330 case '\"':
2331 case '?':
2332 /* GCC extensions. */
2333 case '(':
2334 case '{':
2335 case '[':
2336 case '%':
2337 /* Unrecognized escapes turn into the character itself. */
2338 default:
2339 if (output)
2340 obstack_1grow (output, *tokptr);
2341 ++tokptr;
2342 break;
2343 }
2344 *ptr = tokptr;
2345 return result;
2346 }
2347
2348 /* Parse a string or character literal from TOKPTR. The string or
2349 character may be wide or unicode. *OUTPTR is set to just after the
2350 end of the literal in the input string. The resulting token is
2351 stored in VALUE. This returns a token value, either STRING or
2352 CHAR, depending on what was parsed. *HOST_CHARS is set to the
2353 number of host characters in the literal. */
2354
2355 static int
2356 parse_string_or_char (const char *tokptr, const char **outptr,
2357 struct typed_stoken *value, int *host_chars)
2358 {
2359 int quote;
2360 c_string_type type;
2361 int is_objc = 0;
2362
2363 /* Build the gdb internal form of the input string in tempbuf. Note
2364 that the buffer is null byte terminated *only* for the
2365 convenience of debugging gdb itself and printing the buffer
2366 contents when the buffer contains no embedded nulls. Gdb does
2367 not depend upon the buffer being null byte terminated, it uses
2368 the length string instead. This allows gdb to handle C strings
2369 (as well as strings in other languages) with embedded null
2370 bytes */
2371
2372 if (!tempbuf_init)
2373 tempbuf_init = 1;
2374 else
2375 obstack_free (&tempbuf, NULL);
2376 obstack_init (&tempbuf);
2377
2378 /* Record the string type. */
2379 if (*tokptr == 'L')
2380 {
2381 type = C_WIDE_STRING;
2382 ++tokptr;
2383 }
2384 else if (*tokptr == 'u')
2385 {
2386 type = C_STRING_16;
2387 ++tokptr;
2388 }
2389 else if (*tokptr == 'U')
2390 {
2391 type = C_STRING_32;
2392 ++tokptr;
2393 }
2394 else if (*tokptr == '@')
2395 {
2396 /* An Objective C string. */
2397 is_objc = 1;
2398 type = C_STRING;
2399 ++tokptr;
2400 }
2401 else
2402 type = C_STRING;
2403
2404 /* Skip the quote. */
2405 quote = *tokptr;
2406 if (quote == '\'')
2407 type |= C_CHAR;
2408 ++tokptr;
2409
2410 *host_chars = 0;
2411
2412 while (*tokptr)
2413 {
2414 char c = *tokptr;
2415 if (c == '\\')
2416 {
2417 ++tokptr;
2418 *host_chars += c_parse_escape (&tokptr, &tempbuf);
2419 }
2420 else if (c == quote)
2421 break;
2422 else
2423 {
2424 obstack_1grow (&tempbuf, c);
2425 ++tokptr;
2426 /* FIXME: this does the wrong thing with multi-byte host
2427 characters. We could use mbrlen here, but that would
2428 make "set host-charset" a bit less useful. */
2429 ++*host_chars;
2430 }
2431 }
2432
2433 if (*tokptr != quote)
2434 {
2435 if (quote == '"')
2436 error (_("Unterminated string in expression."));
2437 else
2438 error (_("Unmatched single quote."));
2439 }
2440 ++tokptr;
2441
2442 value->type = type;
2443 value->ptr = (char *) obstack_base (&tempbuf);
2444 value->length = obstack_object_size (&tempbuf);
2445
2446 *outptr = tokptr;
2447
2448 return quote == '"' ? (is_objc ? NSSTRING : STRING) : CHAR;
2449 }
2450
2451 /* This is used to associate some attributes with a token. */
2452
2453 enum token_flag
2454 {
2455 /* If this bit is set, the token is C++-only. */
2456
2457 FLAG_CXX = 1,
2458
2459 /* If this bit is set, the token is C-only. */
2460
2461 FLAG_C = 2,
2462
2463 /* If this bit is set, the token is conditional: if there is a
2464 symbol of the same name, then the token is a symbol; otherwise,
2465 the token is a keyword. */
2466
2467 FLAG_SHADOW = 4
2468 };
2469 DEF_ENUM_FLAGS_TYPE (enum token_flag, token_flags);
2470
2471 struct token
2472 {
2473 const char *oper;
2474 int token;
2475 enum exp_opcode opcode;
2476 token_flags flags;
2477 };
2478
2479 static const struct token tokentab3[] =
2480 {
2481 {">>=", ASSIGN_MODIFY, BINOP_RSH, 0},
2482 {"<<=", ASSIGN_MODIFY, BINOP_LSH, 0},
2483 {"->*", ARROW_STAR, OP_NULL, FLAG_CXX},
2484 {"...", DOTDOTDOT, OP_NULL, 0}
2485 };
2486
2487 static const struct token tokentab2[] =
2488 {
2489 {"+=", ASSIGN_MODIFY, BINOP_ADD, 0},
2490 {"-=", ASSIGN_MODIFY, BINOP_SUB, 0},
2491 {"*=", ASSIGN_MODIFY, BINOP_MUL, 0},
2492 {"/=", ASSIGN_MODIFY, BINOP_DIV, 0},
2493 {"%=", ASSIGN_MODIFY, BINOP_REM, 0},
2494 {"|=", ASSIGN_MODIFY, BINOP_BITWISE_IOR, 0},
2495 {"&=", ASSIGN_MODIFY, BINOP_BITWISE_AND, 0},
2496 {"^=", ASSIGN_MODIFY, BINOP_BITWISE_XOR, 0},
2497 {"++", INCREMENT, OP_NULL, 0},
2498 {"--", DECREMENT, OP_NULL, 0},
2499 {"->", ARROW, OP_NULL, 0},
2500 {"&&", ANDAND, OP_NULL, 0},
2501 {"||", OROR, OP_NULL, 0},
2502 /* "::" is *not* only C++: gdb overrides its meaning in several
2503 different ways, e.g., 'filename'::func, function::variable. */
2504 {"::", COLONCOLON, OP_NULL, 0},
2505 {"<<", LSH, OP_NULL, 0},
2506 {">>", RSH, OP_NULL, 0},
2507 {"==", EQUAL, OP_NULL, 0},
2508 {"!=", NOTEQUAL, OP_NULL, 0},
2509 {"<=", LEQ, OP_NULL, 0},
2510 {">=", GEQ, OP_NULL, 0},
2511 {".*", DOT_STAR, OP_NULL, FLAG_CXX}
2512 };
2513
2514 /* Identifier-like tokens. Only type-specifiers than can appear in
2515 multi-word type names (for example 'double' can appear in 'long
2516 double') need to be listed here. type-specifiers that are only ever
2517 single word (like 'char') are handled by the classify_name function. */
2518 static const struct token ident_tokens[] =
2519 {
2520 {"unsigned", UNSIGNED, OP_NULL, 0},
2521 {"template", TEMPLATE, OP_NULL, FLAG_CXX},
2522 {"volatile", VOLATILE_KEYWORD, OP_NULL, 0},
2523 {"struct", STRUCT, OP_NULL, 0},
2524 {"signed", SIGNED_KEYWORD, OP_NULL, 0},
2525 {"sizeof", SIZEOF, OP_NULL, 0},
2526 {"_Alignof", ALIGNOF, OP_NULL, 0},
2527 {"alignof", ALIGNOF, OP_NULL, FLAG_CXX},
2528 {"double", DOUBLE_KEYWORD, OP_NULL, 0},
2529 {"float", FLOAT_KEYWORD, OP_NULL, 0},
2530 {"false", FALSEKEYWORD, OP_NULL, FLAG_CXX},
2531 {"class", CLASS, OP_NULL, FLAG_CXX},
2532 {"union", UNION, OP_NULL, 0},
2533 {"short", SHORT, OP_NULL, 0},
2534 {"const", CONST_KEYWORD, OP_NULL, 0},
2535 {"restrict", RESTRICT, OP_NULL, FLAG_C | FLAG_SHADOW},
2536 {"__restrict__", RESTRICT, OP_NULL, 0},
2537 {"__restrict", RESTRICT, OP_NULL, 0},
2538 {"_Atomic", ATOMIC, OP_NULL, 0},
2539 {"enum", ENUM, OP_NULL, 0},
2540 {"long", LONG, OP_NULL, 0},
2541 {"_Complex", COMPLEX, OP_NULL, 0},
2542 {"__complex__", COMPLEX, OP_NULL, 0},
2543
2544 {"true", TRUEKEYWORD, OP_NULL, FLAG_CXX},
2545 {"int", INT_KEYWORD, OP_NULL, 0},
2546 {"new", NEW, OP_NULL, FLAG_CXX},
2547 {"delete", DELETE, OP_NULL, FLAG_CXX},
2548 {"operator", OPERATOR, OP_NULL, FLAG_CXX},
2549
2550 {"and", ANDAND, OP_NULL, FLAG_CXX},
2551 {"and_eq", ASSIGN_MODIFY, BINOP_BITWISE_AND, FLAG_CXX},
2552 {"bitand", '&', OP_NULL, FLAG_CXX},
2553 {"bitor", '|', OP_NULL, FLAG_CXX},
2554 {"compl", '~', OP_NULL, FLAG_CXX},
2555 {"not", '!', OP_NULL, FLAG_CXX},
2556 {"not_eq", NOTEQUAL, OP_NULL, FLAG_CXX},
2557 {"or", OROR, OP_NULL, FLAG_CXX},
2558 {"or_eq", ASSIGN_MODIFY, BINOP_BITWISE_IOR, FLAG_CXX},
2559 {"xor", '^', OP_NULL, FLAG_CXX},
2560 {"xor_eq", ASSIGN_MODIFY, BINOP_BITWISE_XOR, FLAG_CXX},
2561
2562 {"const_cast", CONST_CAST, OP_NULL, FLAG_CXX },
2563 {"dynamic_cast", DYNAMIC_CAST, OP_NULL, FLAG_CXX },
2564 {"static_cast", STATIC_CAST, OP_NULL, FLAG_CXX },
2565 {"reinterpret_cast", REINTERPRET_CAST, OP_NULL, FLAG_CXX },
2566
2567 {"__typeof__", TYPEOF, OP_TYPEOF, 0 },
2568 {"__typeof", TYPEOF, OP_TYPEOF, 0 },
2569 {"typeof", TYPEOF, OP_TYPEOF, FLAG_SHADOW },
2570 {"__decltype", DECLTYPE, OP_DECLTYPE, FLAG_CXX },
2571 {"decltype", DECLTYPE, OP_DECLTYPE, FLAG_CXX | FLAG_SHADOW },
2572
2573 {"typeid", TYPEID, OP_TYPEID, FLAG_CXX}
2574 };
2575
2576
2577 static void
2578 scan_macro_expansion (const char *expansion)
2579 {
2580 /* We'd better not be trying to push the stack twice. */
2581 gdb_assert (! cpstate->macro_original_text);
2582
2583 /* Copy to the obstack. */
2584 const char *copy = obstack_strdup (&cpstate->expansion_obstack, expansion);
2585
2586 /* Save the old lexptr value, so we can return to it when we're done
2587 parsing the expanded text. */
2588 cpstate->macro_original_text = pstate->lexptr;
2589 pstate->lexptr = copy;
2590 }
2591
2592 static int
2593 scanning_macro_expansion (void)
2594 {
2595 return cpstate->macro_original_text != 0;
2596 }
2597
2598 static void
2599 finished_macro_expansion (void)
2600 {
2601 /* There'd better be something to pop back to. */
2602 gdb_assert (cpstate->macro_original_text);
2603
2604 /* Pop back to the original text. */
2605 pstate->lexptr = cpstate->macro_original_text;
2606 cpstate->macro_original_text = 0;
2607 }
2608
2609 /* Return true iff the token represents a C++ cast operator. */
2610
2611 static int
2612 is_cast_operator (const char *token, int len)
2613 {
2614 return (! strncmp (token, "dynamic_cast", len)
2615 || ! strncmp (token, "static_cast", len)
2616 || ! strncmp (token, "reinterpret_cast", len)
2617 || ! strncmp (token, "const_cast", len));
2618 }
2619
2620 /* The scope used for macro expansion. */
2621 static struct macro_scope *expression_macro_scope;
2622
2623 /* This is set if a NAME token appeared at the very end of the input
2624 string, with no whitespace separating the name from the EOF. This
2625 is used only when parsing to do field name completion. */
2626 static int saw_name_at_eof;
2627
2628 /* This is set if the previously-returned token was a structure
2629 operator -- either '.' or ARROW. */
2630 static bool last_was_structop;
2631
2632 /* Depth of parentheses. */
2633 static int paren_depth;
2634
2635 /* Read one token, getting characters through lexptr. */
2636
2637 static int
2638 lex_one_token (struct parser_state *par_state, bool *is_quoted_name)
2639 {
2640 int c;
2641 int namelen;
2642 unsigned int i;
2643 const char *tokstart;
2644 bool saw_structop = last_was_structop;
2645
2646 last_was_structop = false;
2647 *is_quoted_name = false;
2648
2649 retry:
2650
2651 /* Check if this is a macro invocation that we need to expand. */
2652 if (! scanning_macro_expansion ())
2653 {
2654 gdb::unique_xmalloc_ptr<char> expanded
2655 = macro_expand_next (&pstate->lexptr, *expression_macro_scope);
2656
2657 if (expanded != nullptr)
2658 scan_macro_expansion (expanded.get ());
2659 }
2660
2661 pstate->prev_lexptr = pstate->lexptr;
2662
2663 tokstart = pstate->lexptr;
2664 /* See if it is a special token of length 3. */
2665 for (i = 0; i < sizeof tokentab3 / sizeof tokentab3[0]; i++)
2666 if (strncmp (tokstart, tokentab3[i].oper, 3) == 0)
2667 {
2668 if ((tokentab3[i].flags & FLAG_CXX) != 0
2669 && par_state->language ()->la_language != language_cplus)
2670 break;
2671 gdb_assert ((tokentab3[i].flags & FLAG_C) == 0);
2672
2673 pstate->lexptr += 3;
2674 yylval.opcode = tokentab3[i].opcode;
2675 return tokentab3[i].token;
2676 }
2677
2678 /* See if it is a special token of length 2. */
2679 for (i = 0; i < sizeof tokentab2 / sizeof tokentab2[0]; i++)
2680 if (strncmp (tokstart, tokentab2[i].oper, 2) == 0)
2681 {
2682 if ((tokentab2[i].flags & FLAG_CXX) != 0
2683 && par_state->language ()->la_language != language_cplus)
2684 break;
2685 gdb_assert ((tokentab2[i].flags & FLAG_C) == 0);
2686
2687 pstate->lexptr += 2;
2688 yylval.opcode = tokentab2[i].opcode;
2689 if (tokentab2[i].token == ARROW)
2690 last_was_structop = 1;
2691 return tokentab2[i].token;
2692 }
2693
2694 switch (c = *tokstart)
2695 {
2696 case 0:
2697 /* If we were just scanning the result of a macro expansion,
2698 then we need to resume scanning the original text.
2699 If we're parsing for field name completion, and the previous
2700 token allows such completion, return a COMPLETE token.
2701 Otherwise, we were already scanning the original text, and
2702 we're really done. */
2703 if (scanning_macro_expansion ())
2704 {
2705 finished_macro_expansion ();
2706 goto retry;
2707 }
2708 else if (saw_name_at_eof)
2709 {
2710 saw_name_at_eof = 0;
2711 return COMPLETE;
2712 }
2713 else if (par_state->parse_completion && saw_structop)
2714 return COMPLETE;
2715 else
2716 return 0;
2717
2718 case ' ':
2719 case '\t':
2720 case '\n':
2721 pstate->lexptr++;
2722 goto retry;
2723
2724 case '[':
2725 case '(':
2726 paren_depth++;
2727 pstate->lexptr++;
2728 if (par_state->language ()->la_language == language_objc
2729 && c == '[')
2730 return OBJC_LBRAC;
2731 return c;
2732
2733 case ']':
2734 case ')':
2735 if (paren_depth == 0)
2736 return 0;
2737 paren_depth--;
2738 pstate->lexptr++;
2739 return c;
2740
2741 case ',':
2742 if (pstate->comma_terminates
2743 && paren_depth == 0
2744 && ! scanning_macro_expansion ())
2745 return 0;
2746 pstate->lexptr++;
2747 return c;
2748
2749 case '.':
2750 /* Might be a floating point number. */
2751 if (pstate->lexptr[1] < '0' || pstate->lexptr[1] > '9')
2752 {
2753 last_was_structop = true;
2754 goto symbol; /* Nope, must be a symbol. */
2755 }
2756 /* FALL THRU. */
2757
2758 case '0':
2759 case '1':
2760 case '2':
2761 case '3':
2762 case '4':
2763 case '5':
2764 case '6':
2765 case '7':
2766 case '8':
2767 case '9':
2768 {
2769 /* It's a number. */
2770 int got_dot = 0, got_e = 0, got_p = 0, toktype;
2771 const char *p = tokstart;
2772 int hex = input_radix > 10;
2773
2774 if (c == '0' && (p[1] == 'x' || p[1] == 'X'))
2775 {
2776 p += 2;
2777 hex = 1;
2778 }
2779 else if (c == '0' && (p[1]=='t' || p[1]=='T' || p[1]=='d' || p[1]=='D'))
2780 {
2781 p += 2;
2782 hex = 0;
2783 }
2784
2785 for (;; ++p)
2786 {
2787 /* This test includes !hex because 'e' is a valid hex digit
2788 and thus does not indicate a floating point number when
2789 the radix is hex. */
2790 if (!hex && !got_e && !got_p && (*p == 'e' || *p == 'E'))
2791 got_dot = got_e = 1;
2792 else if (!got_e && !got_p && (*p == 'p' || *p == 'P'))
2793 got_dot = got_p = 1;
2794 /* This test does not include !hex, because a '.' always indicates
2795 a decimal floating point number regardless of the radix. */
2796 else if (!got_dot && *p == '.')
2797 got_dot = 1;
2798 else if (((got_e && (p[-1] == 'e' || p[-1] == 'E'))
2799 || (got_p && (p[-1] == 'p' || p[-1] == 'P')))
2800 && (*p == '-' || *p == '+'))
2801 /* This is the sign of the exponent, not the end of the
2802 number. */
2803 continue;
2804 /* We will take any letters or digits. parse_number will
2805 complain if past the radix, or if L or U are not final. */
2806 else if ((*p < '0' || *p > '9')
2807 && ((*p < 'a' || *p > 'z')
2808 && (*p < 'A' || *p > 'Z')))
2809 break;
2810 }
2811 toktype = parse_number (par_state, tokstart, p - tokstart,
2812 got_dot | got_e | got_p, &yylval);
2813 if (toktype == ERROR)
2814 {
2815 char *err_copy = (char *) alloca (p - tokstart + 1);
2816
2817 memcpy (err_copy, tokstart, p - tokstart);
2818 err_copy[p - tokstart] = 0;
2819 error (_("Invalid number \"%s\"."), err_copy);
2820 }
2821 pstate->lexptr = p;
2822 return toktype;
2823 }
2824
2825 case '@':
2826 {
2827 const char *p = &tokstart[1];
2828
2829 if (par_state->language ()->la_language == language_objc)
2830 {
2831 size_t len = strlen ("selector");
2832
2833 if (strncmp (p, "selector", len) == 0
2834 && (p[len] == '\0' || ISSPACE (p[len])))
2835 {
2836 pstate->lexptr = p + len;
2837 return SELECTOR;
2838 }
2839 else if (*p == '"')
2840 goto parse_string;
2841 }
2842
2843 while (ISSPACE (*p))
2844 p++;
2845 size_t len = strlen ("entry");
2846 if (strncmp (p, "entry", len) == 0 && !c_ident_is_alnum (p[len])
2847 && p[len] != '_')
2848 {
2849 pstate->lexptr = &p[len];
2850 return ENTRY;
2851 }
2852 }
2853 /* FALLTHRU */
2854 case '+':
2855 case '-':
2856 case '*':
2857 case '/':
2858 case '%':
2859 case '|':
2860 case '&':
2861 case '^':
2862 case '~':
2863 case '!':
2864 case '<':
2865 case '>':
2866 case '?':
2867 case ':':
2868 case '=':
2869 case '{':
2870 case '}':
2871 symbol:
2872 pstate->lexptr++;
2873 return c;
2874
2875 case 'L':
2876 case 'u':
2877 case 'U':
2878 if (tokstart[1] != '"' && tokstart[1] != '\'')
2879 break;
2880 /* Fall through. */
2881 case '\'':
2882 case '"':
2883
2884 parse_string:
2885 {
2886 int host_len;
2887 int result = parse_string_or_char (tokstart, &pstate->lexptr,
2888 &yylval.tsval, &host_len);
2889 if (result == CHAR)
2890 {
2891 if (host_len == 0)
2892 error (_("Empty character constant."));
2893 else if (host_len > 2 && c == '\'')
2894 {
2895 ++tokstart;
2896 namelen = pstate->lexptr - tokstart - 1;
2897 *is_quoted_name = true;
2898
2899 goto tryname;
2900 }
2901 else if (host_len > 1)
2902 error (_("Invalid character constant."));
2903 }
2904 return result;
2905 }
2906 }
2907
2908 if (!(c == '_' || c == '$' || c_ident_is_alpha (c)))
2909 /* We must have come across a bad character (e.g. ';'). */
2910 error (_("Invalid character '%c' in expression."), c);
2911
2912 /* It's a name. See how long it is. */
2913 namelen = 0;
2914 for (c = tokstart[namelen];
2915 (c == '_' || c == '$' || c_ident_is_alnum (c) || c == '<');)
2916 {
2917 /* Template parameter lists are part of the name.
2918 FIXME: This mishandles `print $a<4&&$a>3'. */
2919
2920 if (c == '<')
2921 {
2922 if (! is_cast_operator (tokstart, namelen))
2923 {
2924 /* Scan ahead to get rest of the template specification. Note
2925 that we look ahead only when the '<' adjoins non-whitespace
2926 characters; for comparison expressions, e.g. "a < b > c",
2927 there must be spaces before the '<', etc. */
2928 const char *p = find_template_name_end (tokstart + namelen);
2929
2930 if (p)
2931 namelen = p - tokstart;
2932 }
2933 break;
2934 }
2935 c = tokstart[++namelen];
2936 }
2937
2938 /* The token "if" terminates the expression and is NOT removed from
2939 the input stream. It doesn't count if it appears in the
2940 expansion of a macro. */
2941 if (namelen == 2
2942 && tokstart[0] == 'i'
2943 && tokstart[1] == 'f'
2944 && ! scanning_macro_expansion ())
2945 {
2946 return 0;
2947 }
2948
2949 /* For the same reason (breakpoint conditions), "thread N"
2950 terminates the expression. "thread" could be an identifier, but
2951 an identifier is never followed by a number without intervening
2952 punctuation. "task" is similar. Handle abbreviations of these,
2953 similarly to breakpoint.c:find_condition_and_thread. */
2954 if (namelen >= 1
2955 && (strncmp (tokstart, "thread", namelen) == 0
2956 || strncmp (tokstart, "task", namelen) == 0)
2957 && (tokstart[namelen] == ' ' || tokstart[namelen] == '\t')
2958 && ! scanning_macro_expansion ())
2959 {
2960 const char *p = tokstart + namelen + 1;
2961
2962 while (*p == ' ' || *p == '\t')
2963 p++;
2964 if (*p >= '0' && *p <= '9')
2965 return 0;
2966 }
2967
2968 pstate->lexptr += namelen;
2969
2970 tryname:
2971
2972 yylval.sval.ptr = tokstart;
2973 yylval.sval.length = namelen;
2974
2975 /* Catch specific keywords. */
2976 std::string copy = copy_name (yylval.sval);
2977 for (i = 0; i < sizeof ident_tokens / sizeof ident_tokens[0]; i++)
2978 if (copy == ident_tokens[i].oper)
2979 {
2980 if ((ident_tokens[i].flags & FLAG_CXX) != 0
2981 && par_state->language ()->la_language != language_cplus)
2982 break;
2983 if ((ident_tokens[i].flags & FLAG_C) != 0
2984 && par_state->language ()->la_language != language_c
2985 && par_state->language ()->la_language != language_objc)
2986 break;
2987
2988 if ((ident_tokens[i].flags & FLAG_SHADOW) != 0)
2989 {
2990 struct field_of_this_result is_a_field_of_this;
2991
2992 if (lookup_symbol (copy.c_str (),
2993 pstate->expression_context_block,
2994 VAR_DOMAIN,
2995 (par_state->language ()->la_language
2996 == language_cplus ? &is_a_field_of_this
2997 : NULL)).symbol
2998 != NULL)
2999 {
3000 /* The keyword is shadowed. */
3001 break;
3002 }
3003 }
3004
3005 /* It is ok to always set this, even though we don't always
3006 strictly need to. */
3007 yylval.opcode = ident_tokens[i].opcode;
3008 return ident_tokens[i].token;
3009 }
3010
3011 if (*tokstart == '$')
3012 return DOLLAR_VARIABLE;
3013
3014 if (pstate->parse_completion && *pstate->lexptr == '\0')
3015 saw_name_at_eof = 1;
3016
3017 yylval.ssym.stoken = yylval.sval;
3018 yylval.ssym.sym.symbol = NULL;
3019 yylval.ssym.sym.block = NULL;
3020 yylval.ssym.is_a_field_of_this = 0;
3021 return NAME;
3022 }
3023
3024 /* An object of this type is pushed on a FIFO by the "outer" lexer. */
3025 struct token_and_value
3026 {
3027 int token;
3028 YYSTYPE value;
3029 };
3030
3031 /* A FIFO of tokens that have been read but not yet returned to the
3032 parser. */
3033 static std::vector<token_and_value> token_fifo;
3034
3035 /* Non-zero if the lexer should return tokens from the FIFO. */
3036 static int popping;
3037
3038 /* Temporary storage for c_lex; this holds symbol names as they are
3039 built up. */
3040 static auto_obstack name_obstack;
3041
3042 /* Classify a NAME token. The contents of the token are in `yylval'.
3043 Updates yylval and returns the new token type. BLOCK is the block
3044 in which lookups start; this can be NULL to mean the global scope.
3045 IS_QUOTED_NAME is non-zero if the name token was originally quoted
3046 in single quotes. IS_AFTER_STRUCTOP is true if this name follows
3047 a structure operator -- either '.' or ARROW */
3048
3049 static int
3050 classify_name (struct parser_state *par_state, const struct block *block,
3051 bool is_quoted_name, bool is_after_structop)
3052 {
3053 struct block_symbol bsym;
3054 struct field_of_this_result is_a_field_of_this;
3055
3056 std::string copy = copy_name (yylval.sval);
3057
3058 /* Initialize this in case we *don't* use it in this call; that way
3059 we can refer to it unconditionally below. */
3060 memset (&is_a_field_of_this, 0, sizeof (is_a_field_of_this));
3061
3062 bsym = lookup_symbol (copy.c_str (), block, VAR_DOMAIN,
3063 par_state->language ()->name_of_this ()
3064 ? &is_a_field_of_this : NULL);
3065
3066 if (bsym.symbol && SYMBOL_CLASS (bsym.symbol) == LOC_BLOCK)
3067 {
3068 yylval.ssym.sym = bsym;
3069 yylval.ssym.is_a_field_of_this = is_a_field_of_this.type != NULL;
3070 return BLOCKNAME;
3071 }
3072 else if (!bsym.symbol)
3073 {
3074 /* If we found a field of 'this', we might have erroneously
3075 found a constructor where we wanted a type name. Handle this
3076 case by noticing that we found a constructor and then look up
3077 the type tag instead. */
3078 if (is_a_field_of_this.type != NULL
3079 && is_a_field_of_this.fn_field != NULL
3080 && TYPE_FN_FIELD_CONSTRUCTOR (is_a_field_of_this.fn_field->fn_fields,
3081 0))
3082 {
3083 struct field_of_this_result inner_is_a_field_of_this;
3084
3085 bsym = lookup_symbol (copy.c_str (), block, STRUCT_DOMAIN,
3086 &inner_is_a_field_of_this);
3087 if (bsym.symbol != NULL)
3088 {
3089 yylval.tsym.type = SYMBOL_TYPE (bsym.symbol);
3090 return TYPENAME;
3091 }
3092 }
3093
3094 /* If we found a field on the "this" object, or we are looking
3095 up a field on a struct, then we want to prefer it over a
3096 filename. However, if the name was quoted, then it is better
3097 to check for a filename or a block, since this is the only
3098 way the user has of requiring the extension to be used. */
3099 if ((is_a_field_of_this.type == NULL && !is_after_structop)
3100 || is_quoted_name)
3101 {
3102 /* See if it's a file name. */
3103 struct symtab *symtab;
3104
3105 symtab = lookup_symtab (copy.c_str ());
3106 if (symtab)
3107 {
3108 yylval.bval = BLOCKVECTOR_BLOCK (SYMTAB_BLOCKVECTOR (symtab),
3109 STATIC_BLOCK);
3110 return FILENAME;
3111 }
3112 }
3113 }
3114
3115 if (bsym.symbol && SYMBOL_CLASS (bsym.symbol) == LOC_TYPEDEF)
3116 {
3117 yylval.tsym.type = SYMBOL_TYPE (bsym.symbol);
3118 return TYPENAME;
3119 }
3120
3121 /* See if it's an ObjC classname. */
3122 if (par_state->language ()->la_language == language_objc && !bsym.symbol)
3123 {
3124 CORE_ADDR Class = lookup_objc_class (par_state->gdbarch (),
3125 copy.c_str ());
3126 if (Class)
3127 {
3128 struct symbol *sym;
3129
3130 yylval.theclass.theclass = Class;
3131 sym = lookup_struct_typedef (copy.c_str (),
3132 par_state->expression_context_block, 1);
3133 if (sym)
3134 yylval.theclass.type = SYMBOL_TYPE (sym);
3135 return CLASSNAME;
3136 }
3137 }
3138
3139 /* Input names that aren't symbols but ARE valid hex numbers, when
3140 the input radix permits them, can be names or numbers depending
3141 on the parse. Note we support radixes > 16 here. */
3142 if (!bsym.symbol
3143 && ((copy[0] >= 'a' && copy[0] < 'a' + input_radix - 10)
3144 || (copy[0] >= 'A' && copy[0] < 'A' + input_radix - 10)))
3145 {
3146 YYSTYPE newlval; /* Its value is ignored. */
3147 int hextype = parse_number (par_state, copy.c_str (), yylval.sval.length,
3148 0, &newlval);
3149
3150 if (hextype == INT)
3151 {
3152 yylval.ssym.sym = bsym;
3153 yylval.ssym.is_a_field_of_this = is_a_field_of_this.type != NULL;
3154 return NAME_OR_INT;
3155 }
3156 }
3157
3158 /* Any other kind of symbol */
3159 yylval.ssym.sym = bsym;
3160 yylval.ssym.is_a_field_of_this = is_a_field_of_this.type != NULL;
3161
3162 if (bsym.symbol == NULL
3163 && par_state->language ()->la_language == language_cplus
3164 && is_a_field_of_this.type == NULL
3165 && lookup_minimal_symbol (copy.c_str (), NULL, NULL).minsym == NULL)
3166 return UNKNOWN_CPP_NAME;
3167
3168 return NAME;
3169 }
3170
3171 /* Like classify_name, but used by the inner loop of the lexer, when a
3172 name might have already been seen. CONTEXT is the context type, or
3173 NULL if this is the first component of a name. */
3174
3175 static int
3176 classify_inner_name (struct parser_state *par_state,
3177 const struct block *block, struct type *context)
3178 {
3179 struct type *type;
3180
3181 if (context == NULL)
3182 return classify_name (par_state, block, false, false);
3183
3184 type = check_typedef (context);
3185 if (!type_aggregate_p (type))
3186 return ERROR;
3187
3188 std::string copy = copy_name (yylval.ssym.stoken);
3189 /* N.B. We assume the symbol can only be in VAR_DOMAIN. */
3190 yylval.ssym.sym = cp_lookup_nested_symbol (type, copy.c_str (), block,
3191 VAR_DOMAIN);
3192
3193 /* If no symbol was found, search for a matching base class named
3194 COPY. This will allow users to enter qualified names of class members
3195 relative to the `this' pointer. */
3196 if (yylval.ssym.sym.symbol == NULL)
3197 {
3198 struct type *base_type = cp_find_type_baseclass_by_name (type,
3199 copy.c_str ());
3200
3201 if (base_type != NULL)
3202 {
3203 yylval.tsym.type = base_type;
3204 return TYPENAME;
3205 }
3206
3207 return ERROR;
3208 }
3209
3210 switch (SYMBOL_CLASS (yylval.ssym.sym.symbol))
3211 {
3212 case LOC_BLOCK:
3213 case LOC_LABEL:
3214 /* cp_lookup_nested_symbol might have accidentally found a constructor
3215 named COPY when we really wanted a base class of the same name.
3216 Double-check this case by looking for a base class. */
3217 {
3218 struct type *base_type
3219 = cp_find_type_baseclass_by_name (type, copy.c_str ());
3220
3221 if (base_type != NULL)
3222 {
3223 yylval.tsym.type = base_type;
3224 return TYPENAME;
3225 }
3226 }
3227 return ERROR;
3228
3229 case LOC_TYPEDEF:
3230 yylval.tsym.type = SYMBOL_TYPE (yylval.ssym.sym.symbol);
3231 return TYPENAME;
3232
3233 default:
3234 return NAME;
3235 }
3236 internal_error (__FILE__, __LINE__, _("not reached"));
3237 }
3238
3239 /* The outer level of a two-level lexer. This calls the inner lexer
3240 to return tokens. It then either returns these tokens, or
3241 aggregates them into a larger token. This lets us work around a
3242 problem in our parsing approach, where the parser could not
3243 distinguish between qualified names and qualified types at the
3244 right point.
3245
3246 This approach is still not ideal, because it mishandles template
3247 types. See the comment in lex_one_token for an example. However,
3248 this is still an improvement over the earlier approach, and will
3249 suffice until we move to better parsing technology. */
3250
3251 static int
3252 yylex (void)
3253 {
3254 token_and_value current;
3255 int first_was_coloncolon, last_was_coloncolon;
3256 struct type *context_type = NULL;
3257 int last_to_examine, next_to_examine, checkpoint;
3258 const struct block *search_block;
3259 bool is_quoted_name, last_lex_was_structop;
3260
3261 if (popping && !token_fifo.empty ())
3262 goto do_pop;
3263 popping = 0;
3264
3265 last_lex_was_structop = last_was_structop;
3266
3267 /* Read the first token and decide what to do. Most of the
3268 subsequent code is C++-only; but also depends on seeing a "::" or
3269 name-like token. */
3270 current.token = lex_one_token (pstate, &is_quoted_name);
3271 if (current.token == NAME)
3272 current.token = classify_name (pstate, pstate->expression_context_block,
3273 is_quoted_name, last_lex_was_structop);
3274 if (pstate->language ()->la_language != language_cplus
3275 || (current.token != TYPENAME && current.token != COLONCOLON
3276 && current.token != FILENAME))
3277 return current.token;
3278
3279 /* Read any sequence of alternating "::" and name-like tokens into
3280 the token FIFO. */
3281 current.value = yylval;
3282 token_fifo.push_back (current);
3283 last_was_coloncolon = current.token == COLONCOLON;
3284 while (1)
3285 {
3286 bool ignore;
3287
3288 /* We ignore quoted names other than the very first one.
3289 Subsequent ones do not have any special meaning. */
3290 current.token = lex_one_token (pstate, &ignore);
3291 current.value = yylval;
3292 token_fifo.push_back (current);
3293
3294 if ((last_was_coloncolon && current.token != NAME)
3295 || (!last_was_coloncolon && current.token != COLONCOLON))
3296 break;
3297 last_was_coloncolon = !last_was_coloncolon;
3298 }
3299 popping = 1;
3300
3301 /* We always read one extra token, so compute the number of tokens
3302 to examine accordingly. */
3303 last_to_examine = token_fifo.size () - 2;
3304 next_to_examine = 0;
3305
3306 current = token_fifo[next_to_examine];
3307 ++next_to_examine;
3308
3309 name_obstack.clear ();
3310 checkpoint = 0;
3311 if (current.token == FILENAME)
3312 search_block = current.value.bval;
3313 else if (current.token == COLONCOLON)
3314 search_block = NULL;
3315 else
3316 {
3317 gdb_assert (current.token == TYPENAME);
3318 search_block = pstate->expression_context_block;
3319 obstack_grow (&name_obstack, current.value.sval.ptr,
3320 current.value.sval.length);
3321 context_type = current.value.tsym.type;
3322 checkpoint = 1;
3323 }
3324
3325 first_was_coloncolon = current.token == COLONCOLON;
3326 last_was_coloncolon = first_was_coloncolon;
3327
3328 while (next_to_examine <= last_to_examine)
3329 {
3330 token_and_value next;
3331
3332 next = token_fifo[next_to_examine];
3333 ++next_to_examine;
3334
3335 if (next.token == NAME && last_was_coloncolon)
3336 {
3337 int classification;
3338
3339 yylval = next.value;
3340 classification = classify_inner_name (pstate, search_block,
3341 context_type);
3342 /* We keep going until we either run out of names, or until
3343 we have a qualified name which is not a type. */
3344 if (classification != TYPENAME && classification != NAME)
3345 break;
3346
3347 /* Accept up to this token. */
3348 checkpoint = next_to_examine;
3349
3350 /* Update the partial name we are constructing. */
3351 if (context_type != NULL)
3352 {
3353 /* We don't want to put a leading "::" into the name. */
3354 obstack_grow_str (&name_obstack, "::");
3355 }
3356 obstack_grow (&name_obstack, next.value.sval.ptr,
3357 next.value.sval.length);
3358
3359 yylval.sval.ptr = (const char *) obstack_base (&name_obstack);
3360 yylval.sval.length = obstack_object_size (&name_obstack);
3361 current.value = yylval;
3362 current.token = classification;
3363
3364 last_was_coloncolon = 0;
3365
3366 if (classification == NAME)
3367 break;
3368
3369 context_type = yylval.tsym.type;
3370 }
3371 else if (next.token == COLONCOLON && !last_was_coloncolon)
3372 last_was_coloncolon = 1;
3373 else
3374 {
3375 /* We've reached the end of the name. */
3376 break;
3377 }
3378 }
3379
3380 /* If we have a replacement token, install it as the first token in
3381 the FIFO, and delete the other constituent tokens. */
3382 if (checkpoint > 0)
3383 {
3384 current.value.sval.ptr
3385 = obstack_strndup (&cpstate->expansion_obstack,
3386 current.value.sval.ptr,
3387 current.value.sval.length);
3388
3389 token_fifo[0] = current;
3390 if (checkpoint > 1)
3391 token_fifo.erase (token_fifo.begin () + 1,
3392 token_fifo.begin () + checkpoint);
3393 }
3394
3395 do_pop:
3396 current = token_fifo[0];
3397 token_fifo.erase (token_fifo.begin ());
3398 yylval = current.value;
3399 return current.token;
3400 }
3401
3402 int
3403 c_parse (struct parser_state *par_state)
3404 {
3405 /* Setting up the parser state. */
3406 scoped_restore pstate_restore = make_scoped_restore (&pstate);
3407 gdb_assert (par_state != NULL);
3408 pstate = par_state;
3409
3410 c_parse_state cstate;
3411 scoped_restore cstate_restore = make_scoped_restore (&cpstate, &cstate);
3412
3413 gdb::unique_xmalloc_ptr<struct macro_scope> macro_scope;
3414
3415 if (par_state->expression_context_block)
3416 macro_scope
3417 = sal_macro_scope (find_pc_line (par_state->expression_context_pc, 0));
3418 else
3419 macro_scope = default_macro_scope ();
3420 if (! macro_scope)
3421 macro_scope = user_macro_scope ();
3422
3423 scoped_restore restore_macro_scope
3424 = make_scoped_restore (&expression_macro_scope, macro_scope.get ());
3425
3426 scoped_restore restore_yydebug = make_scoped_restore (&yydebug,
3427 parser_debug);
3428
3429 /* Initialize some state used by the lexer. */
3430 last_was_structop = false;
3431 saw_name_at_eof = 0;
3432 paren_depth = 0;
3433
3434 token_fifo.clear ();
3435 popping = 0;
3436 name_obstack.clear ();
3437
3438 int result = yyparse ();
3439 if (!result)
3440 pstate->set_operation (pstate->pop ());
3441 return result;
3442 }
3443
3444 #ifdef YYBISON
3445
3446 /* This is called via the YYPRINT macro when parser debugging is
3447 enabled. It prints a token's value. */
3448
3449 static void
3450 c_print_token (FILE *file, int type, YYSTYPE value)
3451 {
3452 switch (type)
3453 {
3454 case INT:
3455 parser_fprintf (file, "typed_val_int<%s, %s>",
3456 TYPE_SAFE_NAME (value.typed_val_int.type),
3457 pulongest (value.typed_val_int.val));
3458 break;
3459
3460 case CHAR:
3461 case STRING:
3462 {
3463 char *copy = (char *) alloca (value.tsval.length + 1);
3464
3465 memcpy (copy, value.tsval.ptr, value.tsval.length);
3466 copy[value.tsval.length] = '\0';
3467
3468 parser_fprintf (file, "tsval<type=%d, %s>", value.tsval.type, copy);
3469 }
3470 break;
3471
3472 case NSSTRING:
3473 case DOLLAR_VARIABLE:
3474 parser_fprintf (file, "sval<%s>", copy_name (value.sval).c_str ());
3475 break;
3476
3477 case TYPENAME:
3478 parser_fprintf (file, "tsym<type=%s, name=%s>",
3479 TYPE_SAFE_NAME (value.tsym.type),
3480 copy_name (value.tsym.stoken).c_str ());
3481 break;
3482
3483 case NAME:
3484 case UNKNOWN_CPP_NAME:
3485 case NAME_OR_INT:
3486 case BLOCKNAME:
3487 parser_fprintf (file, "ssym<name=%s, sym=%s, field_of_this=%d>",
3488 copy_name (value.ssym.stoken).c_str (),
3489 (value.ssym.sym.symbol == NULL
3490 ? "(null)" : value.ssym.sym.symbol->print_name ()),
3491 value.ssym.is_a_field_of_this);
3492 break;
3493
3494 case FILENAME:
3495 parser_fprintf (file, "bval<%s>", host_address_to_string (value.bval));
3496 break;
3497 }
3498 }
3499
3500 #endif
3501
3502 static void
3503 yyerror (const char *msg)
3504 {
3505 if (pstate->prev_lexptr)
3506 pstate->lexptr = pstate->prev_lexptr;
3507
3508 error (_("A %s in expression, near `%s'."), msg, pstate->lexptr);
3509 }