1 /* YACC parser for C expressions, for GDB.
2 Copyright (C) 1986-2019 Free Software Foundation, Inc.
4 This file is part of GDB.
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.
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.
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/>. */
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.
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. */
40 #include "expression.h"
42 #include "parser-defs.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 */
51 #include "cp-support.h"
52 #include "macroscope.h"
53 #include "objc-lang.h"
54 #include "typeprint.h"
56 #include "type-stack.h"
58 #define parse_type(ps) builtin_type (ps->gdbarch ())
60 /* Remap normal yacc parser interface names (yyparse, yylex, yyerror,
62 #define GDB_YY_REMAP_PREFIX c_
65 /* The state of the parser, used internally when we are parsing the
68 static struct parser_state *pstate = NULL;
70 /* Data that must be held for the duration of a parse. */
74 /* These are used to hold type lists and type stacks that are
75 allocated during the parse. */
76 std::vector<std::unique_ptr<std::vector<struct type *>>> type_lists;
77 std::vector<std::unique_ptr<struct type_stack>> type_stacks;
79 /* Storage for some strings allocated during the parse. */
80 std::vector<gdb::unique_xmalloc_ptr<char>> strings;
82 /* When we find that lexptr (the global var defined in parse.c) is
83 pointing at a macro invocation, we expand the invocation, and call
84 scan_macro_expansion to save the old lexptr here and point lexptr
85 into the expanded text. When we reach the end of that, we call
86 end_macro_expansion to pop back to the value we saved here. The
87 macro expansion code promises to return only fully-expanded text,
88 so we don't need to "push" more than one level.
90 This is disgusting, of course. It would be cleaner to do all macro
91 expansion beforehand, and then hand that to lexptr. But we don't
92 really know where the expression ends. Remember, in a command like
94 (gdb) break *ADDRESS if CONDITION
96 we evaluate ADDRESS in the scope of the current frame, but we
97 evaluate CONDITION in the scope of the breakpoint's location. So
98 it's simply wrong to try to macro-expand the whole thing at once. */
99 const char *macro_original_text = nullptr;
101 /* We save all intermediate macro expansions on this obstack for the
102 duration of a single parse. The expansion text may sometimes have
103 to live past the end of the expansion, due to yacc lookahead.
104 Rather than try to be clever about saving the data for a single
105 token, we simply keep it all and delete it after parsing has
107 auto_obstack expansion_obstack;
109 /* The type stack. */
110 struct type_stack type_stack;
113 /* This is set and cleared in c_parse. */
115 static struct c_parse_state *cpstate;
119 static int yylex (void);
121 static void yyerror (const char *);
123 static int type_aggregate_p (struct type *);
127 /* Although the yacc "value" of an expression is not used,
128 since the result is stored in the structure being created,
129 other node types do have values. */
144 struct typed_stoken tsval;
146 struct symtoken ssym;
148 const struct block *bval;
149 enum exp_opcode opcode;
151 struct stoken_vector svec;
152 std::vector<struct type *> *tvec;
154 struct type_stack *type_stack;
156 struct objc_class_str theclass;
160 /* YYSTYPE gets defined by %union */
161 static int parse_number (struct parser_state *par_state,
162 const char *, int, int, YYSTYPE *);
163 static struct stoken operator_stoken (const char *);
164 static struct stoken typename_stoken (const char *);
165 static void check_parameter_typelist (std::vector<struct type *> *);
166 static void write_destructor_name (struct parser_state *par_state,
170 static void c_print_token (FILE *file, int type, YYSTYPE value);
171 #define YYPRINT(FILE, TYPE, VALUE) c_print_token (FILE, TYPE, VALUE)
175 %type <voidval> exp exp1 type_exp start variable qualified_name lcurly function_method
177 %type <tval> type typebase
178 %type <tvec> nonempty_typelist func_mod parameter_typelist
179 /* %type <bval> block */
181 /* Fancy type parsing. */
183 %type <lval> array_mod
184 %type <tval> conversion_type_id
186 %type <type_stack> ptr_operator_ts abs_decl direct_abs_decl
188 %token <typed_val_int> INT
189 %token <typed_val_float> FLOAT
191 /* Both NAME and TYPENAME tokens represent symbols in the input,
192 and both convey their data as strings.
193 But a TYPENAME is a string that happens to be defined as a typedef
194 or builtin type name (such as int or char)
195 and a NAME is any other symbol.
196 Contexts where this distinction is not important can use the
197 nonterminal "name", which matches either NAME or TYPENAME. */
199 %token <tsval> STRING
200 %token <sval> NSSTRING /* ObjC Foundation "NSString" literal */
201 %token SELECTOR /* ObjC "@selector" pseudo-operator */
203 %token <ssym> NAME /* BLOCKNAME defined below to give it higher precedence. */
204 %token <ssym> UNKNOWN_CPP_NAME
205 %token <voidval> COMPLETE
206 %token <tsym> TYPENAME
207 %token <theclass> CLASSNAME /* ObjC Class name */
208 %type <sval> name field_name
209 %type <svec> string_exp
210 %type <ssym> name_not_typename
211 %type <tsym> type_name
213 /* This is like a '[' token, but is only generated when parsing
214 Objective C. This lets us reuse the same parser without
215 erroneously parsing ObjC-specific expressions in C. */
218 /* A NAME_OR_INT is a symbol which is not known in the symbol table,
219 but which would parse as a valid number in the current input radix.
220 E.g. "c" when input_radix==16. Depending on the parse, it will be
221 turned into a name or into a number. */
223 %token <ssym> NAME_OR_INT
226 %token STRUCT CLASS UNION ENUM SIZEOF ALIGNOF UNSIGNED COLONCOLON
231 %token REINTERPRET_CAST DYNAMIC_CAST STATIC_CAST CONST_CAST
237 /* Special type cases, put in to allow the parser to distinguish different
239 %token SIGNED_KEYWORD LONG SHORT INT_KEYWORD CONST_KEYWORD VOLATILE_KEYWORD DOUBLE_KEYWORD
241 %token <sval> DOLLAR_VARIABLE
243 %token <opcode> ASSIGN_MODIFY
252 %right '=' ASSIGN_MODIFY
260 %left '<' '>' LEQ GEQ
265 %right UNARY INCREMENT DECREMENT
266 %right ARROW ARROW_STAR '.' DOT_STAR '[' OBJC_LBRAC '('
267 %token <ssym> BLOCKNAME
268 %token <bval> FILENAME
282 { write_exp_elt_opcode(pstate, OP_TYPE);
283 write_exp_elt_type(pstate, $1);
284 write_exp_elt_opcode(pstate, OP_TYPE);}
287 write_exp_elt_opcode (pstate, OP_TYPEOF);
289 | TYPEOF '(' type ')'
291 write_exp_elt_opcode (pstate, OP_TYPE);
292 write_exp_elt_type (pstate, $3);
293 write_exp_elt_opcode (pstate, OP_TYPE);
295 | DECLTYPE '(' exp ')'
297 write_exp_elt_opcode (pstate, OP_DECLTYPE);
301 /* Expressions, including the comma operator. */
304 { write_exp_elt_opcode (pstate, BINOP_COMMA); }
307 /* Expressions, not including the comma operator. */
308 exp : '*' exp %prec UNARY
309 { write_exp_elt_opcode (pstate, UNOP_IND); }
312 exp : '&' exp %prec UNARY
313 { write_exp_elt_opcode (pstate, UNOP_ADDR); }
316 exp : '-' exp %prec UNARY
317 { write_exp_elt_opcode (pstate, UNOP_NEG); }
320 exp : '+' exp %prec UNARY
321 { write_exp_elt_opcode (pstate, UNOP_PLUS); }
324 exp : '!' exp %prec UNARY
325 { write_exp_elt_opcode (pstate, UNOP_LOGICAL_NOT); }
328 exp : '~' exp %prec UNARY
329 { write_exp_elt_opcode (pstate, UNOP_COMPLEMENT); }
332 exp : INCREMENT exp %prec UNARY
333 { write_exp_elt_opcode (pstate, UNOP_PREINCREMENT); }
336 exp : DECREMENT exp %prec UNARY
337 { write_exp_elt_opcode (pstate, UNOP_PREDECREMENT); }
340 exp : exp INCREMENT %prec UNARY
341 { write_exp_elt_opcode (pstate, UNOP_POSTINCREMENT); }
344 exp : exp DECREMENT %prec UNARY
345 { write_exp_elt_opcode (pstate, UNOP_POSTDECREMENT); }
348 exp : TYPEID '(' exp ')' %prec UNARY
349 { write_exp_elt_opcode (pstate, OP_TYPEID); }
352 exp : TYPEID '(' type_exp ')' %prec UNARY
353 { write_exp_elt_opcode (pstate, OP_TYPEID); }
356 exp : SIZEOF exp %prec UNARY
357 { write_exp_elt_opcode (pstate, UNOP_SIZEOF); }
360 exp : ALIGNOF '(' type_exp ')' %prec UNARY
361 { write_exp_elt_opcode (pstate, UNOP_ALIGNOF); }
364 exp : exp ARROW field_name
365 { write_exp_elt_opcode (pstate, STRUCTOP_PTR);
366 write_exp_string (pstate, $3);
367 write_exp_elt_opcode (pstate, STRUCTOP_PTR); }
370 exp : exp ARROW field_name COMPLETE
371 { pstate->mark_struct_expression ();
372 write_exp_elt_opcode (pstate, STRUCTOP_PTR);
373 write_exp_string (pstate, $3);
374 write_exp_elt_opcode (pstate, STRUCTOP_PTR); }
377 exp : exp ARROW COMPLETE
379 pstate->mark_struct_expression ();
380 write_exp_elt_opcode (pstate, STRUCTOP_PTR);
383 write_exp_string (pstate, s);
384 write_exp_elt_opcode (pstate, STRUCTOP_PTR); }
387 exp : exp ARROW '~' name
388 { write_exp_elt_opcode (pstate, STRUCTOP_PTR);
389 write_destructor_name (pstate, $4);
390 write_exp_elt_opcode (pstate, STRUCTOP_PTR); }
393 exp : exp ARROW '~' name COMPLETE
394 { pstate->mark_struct_expression ();
395 write_exp_elt_opcode (pstate, STRUCTOP_PTR);
396 write_destructor_name (pstate, $4);
397 write_exp_elt_opcode (pstate, STRUCTOP_PTR); }
400 exp : exp ARROW qualified_name
401 { /* exp->type::name becomes exp->*(&type::name) */
402 /* Note: this doesn't work if name is a
403 static member! FIXME */
404 write_exp_elt_opcode (pstate, UNOP_ADDR);
405 write_exp_elt_opcode (pstate, STRUCTOP_MPTR); }
408 exp : exp ARROW_STAR exp
409 { write_exp_elt_opcode (pstate, STRUCTOP_MPTR); }
412 exp : exp '.' field_name
413 { write_exp_elt_opcode (pstate, STRUCTOP_STRUCT);
414 write_exp_string (pstate, $3);
415 write_exp_elt_opcode (pstate, STRUCTOP_STRUCT); }
418 exp : exp '.' field_name COMPLETE
419 { pstate->mark_struct_expression ();
420 write_exp_elt_opcode (pstate, STRUCTOP_STRUCT);
421 write_exp_string (pstate, $3);
422 write_exp_elt_opcode (pstate, STRUCTOP_STRUCT); }
425 exp : exp '.' COMPLETE
427 pstate->mark_struct_expression ();
428 write_exp_elt_opcode (pstate, STRUCTOP_STRUCT);
431 write_exp_string (pstate, s);
432 write_exp_elt_opcode (pstate, STRUCTOP_STRUCT); }
435 exp : exp '.' '~' name
436 { write_exp_elt_opcode (pstate, STRUCTOP_STRUCT);
437 write_destructor_name (pstate, $4);
438 write_exp_elt_opcode (pstate, STRUCTOP_STRUCT); }
441 exp : exp '.' '~' name COMPLETE
442 { pstate->mark_struct_expression ();
443 write_exp_elt_opcode (pstate, STRUCTOP_STRUCT);
444 write_destructor_name (pstate, $4);
445 write_exp_elt_opcode (pstate, STRUCTOP_STRUCT); }
448 exp : exp '.' qualified_name
449 { /* exp.type::name becomes exp.*(&type::name) */
450 /* Note: this doesn't work if name is a
451 static member! FIXME */
452 write_exp_elt_opcode (pstate, UNOP_ADDR);
453 write_exp_elt_opcode (pstate, STRUCTOP_MEMBER); }
456 exp : exp DOT_STAR exp
457 { write_exp_elt_opcode (pstate, STRUCTOP_MEMBER); }
460 exp : exp '[' exp1 ']'
461 { write_exp_elt_opcode (pstate, BINOP_SUBSCRIPT); }
464 exp : exp OBJC_LBRAC exp1 ']'
465 { write_exp_elt_opcode (pstate, BINOP_SUBSCRIPT); }
469 * The rules below parse ObjC message calls of the form:
470 * '[' target selector {':' argument}* ']'
473 exp : OBJC_LBRAC TYPENAME
477 std::string copy = copy_name ($2.stoken);
478 theclass = lookup_objc_class (pstate->gdbarch (),
481 error (_("%s is not an ObjC Class"),
483 write_exp_elt_opcode (pstate, OP_LONG);
484 write_exp_elt_type (pstate,
485 parse_type (pstate)->builtin_int);
486 write_exp_elt_longcst (pstate, (LONGEST) theclass);
487 write_exp_elt_opcode (pstate, OP_LONG);
491 { write_exp_elt_opcode (pstate, OP_OBJC_MSGCALL);
492 end_msglist (pstate);
493 write_exp_elt_opcode (pstate, OP_OBJC_MSGCALL);
497 exp : OBJC_LBRAC CLASSNAME
499 write_exp_elt_opcode (pstate, OP_LONG);
500 write_exp_elt_type (pstate,
501 parse_type (pstate)->builtin_int);
502 write_exp_elt_longcst (pstate, (LONGEST) $2.theclass);
503 write_exp_elt_opcode (pstate, OP_LONG);
507 { write_exp_elt_opcode (pstate, OP_OBJC_MSGCALL);
508 end_msglist (pstate);
509 write_exp_elt_opcode (pstate, OP_OBJC_MSGCALL);
516 { write_exp_elt_opcode (pstate, OP_OBJC_MSGCALL);
517 end_msglist (pstate);
518 write_exp_elt_opcode (pstate, OP_OBJC_MSGCALL);
523 { add_msglist(&$1, 0); }
531 msgarg : name ':' exp
532 { add_msglist(&$1, 1); }
533 | ':' exp /* Unnamed arg. */
534 { add_msglist(0, 1); }
535 | ',' exp /* Variable number of args. */
536 { add_msglist(0, 0); }
540 /* This is to save the value of arglist_len
541 being accumulated by an outer function call. */
542 { pstate->start_arglist (); }
543 arglist ')' %prec ARROW
544 { write_exp_elt_opcode (pstate, OP_FUNCALL);
545 write_exp_elt_longcst (pstate,
546 pstate->end_arglist ());
547 write_exp_elt_opcode (pstate, OP_FUNCALL); }
550 /* This is here to disambiguate with the production for
551 "func()::static_var" further below, which uses
552 function_method_void. */
553 exp : exp '(' ')' %prec ARROW
554 { pstate->start_arglist ();
555 write_exp_elt_opcode (pstate, OP_FUNCALL);
556 write_exp_elt_longcst (pstate,
557 pstate->end_arglist ());
558 write_exp_elt_opcode (pstate, OP_FUNCALL); }
562 exp : UNKNOWN_CPP_NAME '('
564 /* This could potentially be a an argument defined
565 lookup function (Koenig). */
566 write_exp_elt_opcode (pstate, OP_ADL_FUNC);
568 (pstate, pstate->expression_context_block);
569 write_exp_elt_sym (pstate,
570 NULL); /* Placeholder. */
571 write_exp_string (pstate, $1.stoken);
572 write_exp_elt_opcode (pstate, OP_ADL_FUNC);
574 /* This is to save the value of arglist_len
575 being accumulated by an outer function call. */
577 pstate->start_arglist ();
579 arglist ')' %prec ARROW
581 write_exp_elt_opcode (pstate, OP_FUNCALL);
582 write_exp_elt_longcst (pstate,
583 pstate->end_arglist ());
584 write_exp_elt_opcode (pstate, OP_FUNCALL);
589 { pstate->start_arglist (); }
596 { pstate->arglist_len = 1; }
599 arglist : arglist ',' exp %prec ABOVE_COMMA
600 { pstate->arglist_len++; }
603 function_method: exp '(' parameter_typelist ')' const_or_volatile
605 std::vector<struct type *> *type_list = $3;
606 LONGEST len = type_list->size ();
608 write_exp_elt_opcode (pstate, TYPE_INSTANCE);
609 /* Save the const/volatile qualifiers as
610 recorded by the const_or_volatile
611 production's actions. */
612 write_exp_elt_longcst
615 .follow_type_instance_flags ()));
616 write_exp_elt_longcst (pstate, len);
617 for (type *type_elt : *type_list)
618 write_exp_elt_type (pstate, type_elt);
619 write_exp_elt_longcst(pstate, len);
620 write_exp_elt_opcode (pstate, TYPE_INSTANCE);
624 function_method_void: exp '(' ')' const_or_volatile
625 { write_exp_elt_opcode (pstate, TYPE_INSTANCE);
627 write_exp_elt_longcst
629 cpstate->type_stack.follow_type_instance_flags ());
630 write_exp_elt_longcst (pstate, 0);
631 write_exp_elt_longcst (pstate, 0);
632 write_exp_elt_opcode (pstate, TYPE_INSTANCE);
636 exp : function_method
639 /* Normally we must interpret "func()" as a function call, instead of
640 a type. The user needs to write func(void) to disambiguate.
641 However, in the "func()::static_var" case, there's no
643 function_method_void_or_typelist: function_method
644 | function_method_void
647 exp : function_method_void_or_typelist COLONCOLON name
649 write_exp_elt_opcode (pstate, OP_FUNC_STATIC_VAR);
650 write_exp_string (pstate, $3);
651 write_exp_elt_opcode (pstate, OP_FUNC_STATIC_VAR);
656 { $$ = pstate->end_arglist () - 1; }
658 exp : lcurly arglist rcurly %prec ARROW
659 { write_exp_elt_opcode (pstate, OP_ARRAY);
660 write_exp_elt_longcst (pstate, (LONGEST) 0);
661 write_exp_elt_longcst (pstate, (LONGEST) $3);
662 write_exp_elt_opcode (pstate, OP_ARRAY); }
665 exp : lcurly type_exp rcurly exp %prec UNARY
666 { write_exp_elt_opcode (pstate, UNOP_MEMVAL_TYPE); }
669 exp : '(' type_exp ')' exp %prec UNARY
670 { write_exp_elt_opcode (pstate, UNOP_CAST_TYPE); }
677 /* Binary operators in order of decreasing precedence. */
680 { write_exp_elt_opcode (pstate, BINOP_REPEAT); }
684 { write_exp_elt_opcode (pstate, BINOP_MUL); }
688 { write_exp_elt_opcode (pstate, BINOP_DIV); }
692 { write_exp_elt_opcode (pstate, BINOP_REM); }
696 { write_exp_elt_opcode (pstate, BINOP_ADD); }
700 { write_exp_elt_opcode (pstate, BINOP_SUB); }
704 { write_exp_elt_opcode (pstate, BINOP_LSH); }
708 { write_exp_elt_opcode (pstate, BINOP_RSH); }
712 { write_exp_elt_opcode (pstate, BINOP_EQUAL); }
715 exp : exp NOTEQUAL exp
716 { write_exp_elt_opcode (pstate, BINOP_NOTEQUAL); }
720 { write_exp_elt_opcode (pstate, BINOP_LEQ); }
724 { write_exp_elt_opcode (pstate, BINOP_GEQ); }
728 { write_exp_elt_opcode (pstate, BINOP_LESS); }
732 { write_exp_elt_opcode (pstate, BINOP_GTR); }
736 { write_exp_elt_opcode (pstate, BINOP_BITWISE_AND); }
740 { write_exp_elt_opcode (pstate, BINOP_BITWISE_XOR); }
744 { write_exp_elt_opcode (pstate, BINOP_BITWISE_IOR); }
748 { write_exp_elt_opcode (pstate, BINOP_LOGICAL_AND); }
752 { write_exp_elt_opcode (pstate, BINOP_LOGICAL_OR); }
755 exp : exp '?' exp ':' exp %prec '?'
756 { write_exp_elt_opcode (pstate, TERNOP_COND); }
760 { write_exp_elt_opcode (pstate, BINOP_ASSIGN); }
763 exp : exp ASSIGN_MODIFY exp
764 { write_exp_elt_opcode (pstate, BINOP_ASSIGN_MODIFY);
765 write_exp_elt_opcode (pstate, $2);
766 write_exp_elt_opcode (pstate,
767 BINOP_ASSIGN_MODIFY); }
771 { write_exp_elt_opcode (pstate, OP_LONG);
772 write_exp_elt_type (pstate, $1.type);
773 write_exp_elt_longcst (pstate, (LONGEST) ($1.val));
774 write_exp_elt_opcode (pstate, OP_LONG); }
779 struct stoken_vector vec;
782 write_exp_string_vector (pstate, $1.type, &vec);
788 parse_number (pstate, $1.stoken.ptr,
789 $1.stoken.length, 0, &val);
790 write_exp_elt_opcode (pstate, OP_LONG);
791 write_exp_elt_type (pstate, val.typed_val_int.type);
792 write_exp_elt_longcst (pstate,
793 (LONGEST) val.typed_val_int.val);
794 write_exp_elt_opcode (pstate, OP_LONG);
800 { write_exp_elt_opcode (pstate, OP_FLOAT);
801 write_exp_elt_type (pstate, $1.type);
802 write_exp_elt_floatcst (pstate, $1.val);
803 write_exp_elt_opcode (pstate, OP_FLOAT); }
809 exp : DOLLAR_VARIABLE
811 write_dollar_variable (pstate, $1);
815 exp : SELECTOR '(' name ')'
817 write_exp_elt_opcode (pstate, OP_OBJC_SELECTOR);
818 write_exp_string (pstate, $3);
819 write_exp_elt_opcode (pstate, OP_OBJC_SELECTOR); }
822 exp : SIZEOF '(' type ')' %prec UNARY
823 { struct type *type = $3;
824 write_exp_elt_opcode (pstate, OP_LONG);
825 write_exp_elt_type (pstate, lookup_signed_typename
826 (pstate->language (),
829 type = check_typedef (type);
831 /* $5.3.3/2 of the C++ Standard (n3290 draft)
832 says of sizeof: "When applied to a reference
833 or a reference type, the result is the size of
834 the referenced type." */
835 if (TYPE_IS_REFERENCE (type))
836 type = check_typedef (TYPE_TARGET_TYPE (type));
837 write_exp_elt_longcst (pstate,
838 (LONGEST) TYPE_LENGTH (type));
839 write_exp_elt_opcode (pstate, OP_LONG); }
842 exp : REINTERPRET_CAST '<' type_exp '>' '(' exp ')' %prec UNARY
843 { write_exp_elt_opcode (pstate,
844 UNOP_REINTERPRET_CAST); }
847 exp : STATIC_CAST '<' type_exp '>' '(' exp ')' %prec UNARY
848 { write_exp_elt_opcode (pstate, UNOP_CAST_TYPE); }
851 exp : DYNAMIC_CAST '<' type_exp '>' '(' exp ')' %prec UNARY
852 { write_exp_elt_opcode (pstate, UNOP_DYNAMIC_CAST); }
855 exp : CONST_CAST '<' type_exp '>' '(' exp ')' %prec UNARY
856 { /* We could do more error checking here, but
857 it doesn't seem worthwhile. */
858 write_exp_elt_opcode (pstate, UNOP_CAST_TYPE); }
864 /* We copy the string here, and not in the
865 lexer, to guarantee that we do not leak a
866 string. Note that we follow the
867 NUL-termination convention of the
869 struct typed_stoken *vec = XNEW (struct typed_stoken);
874 vec->length = $1.length;
875 vec->ptr = (char *) malloc ($1.length + 1);
876 memcpy (vec->ptr, $1.ptr, $1.length + 1);
881 /* Note that we NUL-terminate here, but just
885 $$.tokens = XRESIZEVEC (struct typed_stoken,
888 p = (char *) malloc ($2.length + 1);
889 memcpy (p, $2.ptr, $2.length + 1);
891 $$.tokens[$$.len - 1].type = $2.type;
892 $$.tokens[$$.len - 1].length = $2.length;
893 $$.tokens[$$.len - 1].ptr = p;
900 c_string_type type = C_STRING;
902 for (i = 0; i < $1.len; ++i)
904 switch ($1.tokens[i].type)
912 && type != $1.tokens[i].type)
913 error (_("Undefined string concatenation."));
914 type = (enum c_string_type_values) $1.tokens[i].type;
918 internal_error (__FILE__, __LINE__,
919 "unrecognized type in string concatenation");
923 write_exp_string_vector (pstate, type, &$1);
924 for (i = 0; i < $1.len; ++i)
925 free ($1.tokens[i].ptr);
930 exp : NSSTRING /* ObjC NextStep NSString constant
931 * of the form '@' '"' string '"'.
933 { write_exp_elt_opcode (pstate, OP_OBJC_NSSTRING);
934 write_exp_string (pstate, $1);
935 write_exp_elt_opcode (pstate, OP_OBJC_NSSTRING); }
940 { write_exp_elt_opcode (pstate, OP_LONG);
941 write_exp_elt_type (pstate,
942 parse_type (pstate)->builtin_bool);
943 write_exp_elt_longcst (pstate, (LONGEST) 1);
944 write_exp_elt_opcode (pstate, OP_LONG); }
948 { write_exp_elt_opcode (pstate, OP_LONG);
949 write_exp_elt_type (pstate,
950 parse_type (pstate)->builtin_bool);
951 write_exp_elt_longcst (pstate, (LONGEST) 0);
952 write_exp_elt_opcode (pstate, OP_LONG); }
960 $$ = SYMBOL_BLOCK_VALUE ($1.sym.symbol);
962 error (_("No file or function \"%s\"."),
963 copy_name ($1.stoken).c_str ());
971 block : block COLONCOLON name
973 std::string copy = copy_name ($3);
975 = lookup_symbol (copy.c_str (), $1,
976 VAR_DOMAIN, NULL).symbol;
978 if (!tem || SYMBOL_CLASS (tem) != LOC_BLOCK)
979 error (_("No function \"%s\" in specified context."),
981 $$ = SYMBOL_BLOCK_VALUE (tem); }
984 variable: name_not_typename ENTRY
985 { struct symbol *sym = $1.sym.symbol;
987 if (sym == NULL || !SYMBOL_IS_ARGUMENT (sym)
988 || !symbol_read_needs_frame (sym))
989 error (_("@entry can be used only for function "
990 "parameters, not for \"%s\""),
991 copy_name ($1.stoken).c_str ());
993 write_exp_elt_opcode (pstate, OP_VAR_ENTRY_VALUE);
994 write_exp_elt_sym (pstate, sym);
995 write_exp_elt_opcode (pstate, OP_VAR_ENTRY_VALUE);
999 variable: block COLONCOLON name
1001 std::string copy = copy_name ($3);
1002 struct block_symbol sym
1003 = lookup_symbol (copy.c_str (), $1,
1006 if (sym.symbol == 0)
1007 error (_("No symbol \"%s\" in specified context."),
1009 if (symbol_read_needs_frame (sym.symbol))
1010 pstate->block_tracker->update (sym);
1012 write_exp_elt_opcode (pstate, OP_VAR_VALUE);
1013 write_exp_elt_block (pstate, sym.block);
1014 write_exp_elt_sym (pstate, sym.symbol);
1015 write_exp_elt_opcode (pstate, OP_VAR_VALUE); }
1018 qualified_name: TYPENAME COLONCOLON name
1020 struct type *type = $1.type;
1021 type = check_typedef (type);
1022 if (!type_aggregate_p (type))
1023 error (_("`%s' is not defined as an aggregate type."),
1024 TYPE_SAFE_NAME (type));
1026 write_exp_elt_opcode (pstate, OP_SCOPE);
1027 write_exp_elt_type (pstate, type);
1028 write_exp_string (pstate, $3);
1029 write_exp_elt_opcode (pstate, OP_SCOPE);
1031 | TYPENAME COLONCOLON '~' name
1033 struct type *type = $1.type;
1034 struct stoken tmp_token;
1037 type = check_typedef (type);
1038 if (!type_aggregate_p (type))
1039 error (_("`%s' is not defined as an aggregate type."),
1040 TYPE_SAFE_NAME (type));
1041 buf = (char *) alloca ($4.length + 2);
1042 tmp_token.ptr = buf;
1043 tmp_token.length = $4.length + 1;
1045 memcpy (buf+1, $4.ptr, $4.length);
1046 buf[tmp_token.length] = 0;
1048 /* Check for valid destructor name. */
1049 destructor_name_p (tmp_token.ptr, $1.type);
1050 write_exp_elt_opcode (pstate, OP_SCOPE);
1051 write_exp_elt_type (pstate, type);
1052 write_exp_string (pstate, tmp_token);
1053 write_exp_elt_opcode (pstate, OP_SCOPE);
1055 | TYPENAME COLONCOLON name COLONCOLON name
1057 std::string copy = copy_name ($3);
1058 error (_("No type \"%s\" within class "
1059 "or namespace \"%s\"."),
1060 copy.c_str (), TYPE_SAFE_NAME ($1.type));
1064 variable: qualified_name
1065 | COLONCOLON name_not_typename
1067 std::string name = copy_name ($2.stoken);
1069 struct bound_minimal_symbol msymbol;
1072 = lookup_symbol (name.c_str (),
1073 (const struct block *) NULL,
1074 VAR_DOMAIN, NULL).symbol;
1077 write_exp_elt_opcode (pstate, OP_VAR_VALUE);
1078 write_exp_elt_block (pstate, NULL);
1079 write_exp_elt_sym (pstate, sym);
1080 write_exp_elt_opcode (pstate, OP_VAR_VALUE);
1084 msymbol = lookup_bound_minimal_symbol (name.c_str ());
1085 if (msymbol.minsym != NULL)
1086 write_exp_msymbol (pstate, msymbol);
1087 else if (!have_full_symbols () && !have_partial_symbols ())
1088 error (_("No symbol table is loaded. Use the \"file\" command."));
1090 error (_("No symbol \"%s\" in current context."),
1095 variable: name_not_typename
1096 { struct block_symbol sym = $1.sym;
1100 if (symbol_read_needs_frame (sym.symbol))
1101 pstate->block_tracker->update (sym);
1103 /* If we found a function, see if it's
1104 an ifunc resolver that has the same
1105 address as the ifunc symbol itself.
1106 If so, prefer the ifunc symbol. */
1108 bound_minimal_symbol resolver
1109 = find_gnu_ifunc (sym.symbol);
1110 if (resolver.minsym != NULL)
1111 write_exp_msymbol (pstate, resolver);
1114 write_exp_elt_opcode (pstate, OP_VAR_VALUE);
1115 write_exp_elt_block (pstate, sym.block);
1116 write_exp_elt_sym (pstate, sym.symbol);
1117 write_exp_elt_opcode (pstate, OP_VAR_VALUE);
1120 else if ($1.is_a_field_of_this)
1122 /* C++: it hangs off of `this'. Must
1123 not inadvertently convert from a method call
1125 pstate->block_tracker->update (sym);
1126 write_exp_elt_opcode (pstate, OP_THIS);
1127 write_exp_elt_opcode (pstate, OP_THIS);
1128 write_exp_elt_opcode (pstate, STRUCTOP_PTR);
1129 write_exp_string (pstate, $1.stoken);
1130 write_exp_elt_opcode (pstate, STRUCTOP_PTR);
1134 std::string arg = copy_name ($1.stoken);
1136 bound_minimal_symbol msymbol
1137 = lookup_bound_minimal_symbol (arg.c_str ());
1138 if (msymbol.minsym == NULL)
1140 if (!have_full_symbols () && !have_partial_symbols ())
1141 error (_("No symbol table is loaded. Use the \"file\" command."));
1143 error (_("No symbol \"%s\" in current context."),
1147 /* This minsym might be an alias for
1148 another function. See if we can find
1149 the debug symbol for the target, and
1150 if so, use it instead, since it has
1151 return type / prototype info. This
1152 is important for example for "p
1153 *__errno_location()". */
1154 symbol *alias_target
1155 = ((msymbol.minsym->type != mst_text_gnu_ifunc
1156 && msymbol.minsym->type != mst_data_gnu_ifunc)
1157 ? find_function_alias_target (msymbol)
1159 if (alias_target != NULL)
1161 write_exp_elt_opcode (pstate, OP_VAR_VALUE);
1163 (pstate, SYMBOL_BLOCK_VALUE (alias_target));
1164 write_exp_elt_sym (pstate, alias_target);
1165 write_exp_elt_opcode (pstate, OP_VAR_VALUE);
1168 write_exp_msymbol (pstate, msymbol);
1173 space_identifier : '@' NAME
1175 cpstate->type_stack.insert (pstate,
1176 copy_name ($2.stoken).c_str ());
1180 const_or_volatile: const_or_volatile_noopt
1184 cv_with_space_id : const_or_volatile space_identifier const_or_volatile
1187 const_or_volatile_or_space_identifier_noopt: cv_with_space_id
1188 | const_or_volatile_noopt
1191 const_or_volatile_or_space_identifier:
1192 const_or_volatile_or_space_identifier_noopt
1198 { cpstate->type_stack.insert (tp_pointer); }
1199 const_or_volatile_or_space_identifier
1201 { cpstate->type_stack.insert (tp_pointer); }
1202 const_or_volatile_or_space_identifier
1204 { cpstate->type_stack.insert (tp_reference); }
1206 { cpstate->type_stack.insert (tp_reference); }
1208 { cpstate->type_stack.insert (tp_rvalue_reference); }
1209 | ANDAND ptr_operator
1210 { cpstate->type_stack.insert (tp_rvalue_reference); }
1213 ptr_operator_ts: ptr_operator
1215 $$ = cpstate->type_stack.create ();
1216 cpstate->type_stacks.emplace_back ($$);
1220 abs_decl: ptr_operator_ts direct_abs_decl
1221 { $$ = $2->append ($1); }
1226 direct_abs_decl: '(' abs_decl ')'
1228 | direct_abs_decl array_mod
1230 cpstate->type_stack.push ($1);
1231 cpstate->type_stack.push ($2);
1232 cpstate->type_stack.push (tp_array);
1233 $$ = cpstate->type_stack.create ();
1234 cpstate->type_stacks.emplace_back ($$);
1238 cpstate->type_stack.push ($1);
1239 cpstate->type_stack.push (tp_array);
1240 $$ = cpstate->type_stack.create ();
1241 cpstate->type_stacks.emplace_back ($$);
1244 | direct_abs_decl func_mod
1246 cpstate->type_stack.push ($1);
1247 cpstate->type_stack.push ($2);
1248 $$ = cpstate->type_stack.create ();
1249 cpstate->type_stacks.emplace_back ($$);
1253 cpstate->type_stack.push ($1);
1254 $$ = cpstate->type_stack.create ();
1255 cpstate->type_stacks.emplace_back ($$);
1265 | OBJC_LBRAC INT ']'
1271 $$ = new std::vector<struct type *>;
1272 cpstate->type_lists.emplace_back ($$);
1274 | '(' parameter_typelist ')'
1278 /* We used to try to recognize pointer to member types here, but
1279 that didn't work (shift/reduce conflicts meant that these rules never
1280 got executed). The problem is that
1281 int (foo::bar::baz::bizzle)
1282 is a function type but
1283 int (foo::bar::baz::bizzle::*)
1284 is a pointer to member type. Stroustrup loses again! */
1289 /* Implements (approximately): (type-qualifier)* type-specifier.
1291 When type-specifier is only ever a single word, like 'float' then these
1292 arrive as pre-built TYPENAME tokens thanks to the classify_name
1293 function. However, when a type-specifier can contain multiple words,
1294 for example 'double' can appear as just 'double' or 'long double', and
1295 similarly 'long' can appear as just 'long' or in 'long double', then
1296 these type-specifiers are parsed into their own tokens in the function
1297 lex_one_token and the ident_tokens array. These separate tokens are all
1303 { $$ = lookup_signed_typename (pstate->language (),
1307 { $$ = lookup_signed_typename (pstate->language (),
1311 { $$ = lookup_signed_typename (pstate->language (),
1315 { $$ = lookup_signed_typename (pstate->language (),
1318 | LONG SIGNED_KEYWORD INT_KEYWORD
1319 { $$ = lookup_signed_typename (pstate->language (),
1322 | LONG SIGNED_KEYWORD
1323 { $$ = lookup_signed_typename (pstate->language (),
1326 | SIGNED_KEYWORD LONG INT_KEYWORD
1327 { $$ = lookup_signed_typename (pstate->language (),
1330 | UNSIGNED LONG INT_KEYWORD
1331 { $$ = lookup_unsigned_typename (pstate->language (),
1334 | LONG UNSIGNED INT_KEYWORD
1335 { $$ = lookup_unsigned_typename (pstate->language (),
1339 { $$ = lookup_unsigned_typename (pstate->language (),
1343 { $$ = lookup_signed_typename (pstate->language (),
1346 | LONG LONG INT_KEYWORD
1347 { $$ = lookup_signed_typename (pstate->language (),
1350 | LONG LONG SIGNED_KEYWORD INT_KEYWORD
1351 { $$ = lookup_signed_typename (pstate->language (),
1354 | LONG LONG SIGNED_KEYWORD
1355 { $$ = lookup_signed_typename (pstate->language (),
1358 | SIGNED_KEYWORD LONG LONG
1359 { $$ = lookup_signed_typename (pstate->language (),
1362 | SIGNED_KEYWORD LONG LONG INT_KEYWORD
1363 { $$ = lookup_signed_typename (pstate->language (),
1366 | UNSIGNED LONG LONG
1367 { $$ = lookup_unsigned_typename (pstate->language (),
1370 | UNSIGNED LONG LONG INT_KEYWORD
1371 { $$ = lookup_unsigned_typename (pstate->language (),
1374 | LONG LONG UNSIGNED
1375 { $$ = lookup_unsigned_typename (pstate->language (),
1378 | LONG LONG UNSIGNED INT_KEYWORD
1379 { $$ = lookup_unsigned_typename (pstate->language (),
1383 { $$ = lookup_signed_typename (pstate->language (),
1386 | SHORT SIGNED_KEYWORD INT_KEYWORD
1387 { $$ = lookup_signed_typename (pstate->language (),
1390 | SHORT SIGNED_KEYWORD
1391 { $$ = lookup_signed_typename (pstate->language (),
1394 | UNSIGNED SHORT INT_KEYWORD
1395 { $$ = lookup_unsigned_typename (pstate->language (),
1399 { $$ = lookup_unsigned_typename (pstate->language (),
1402 | SHORT UNSIGNED INT_KEYWORD
1403 { $$ = lookup_unsigned_typename (pstate->language (),
1407 { $$ = lookup_typename (pstate->language (),
1412 | LONG DOUBLE_KEYWORD
1413 { $$ = lookup_typename (pstate->language (),
1420 = lookup_struct (copy_name ($2).c_str (),
1421 pstate->expression_context_block);
1425 pstate->mark_completion_tag (TYPE_CODE_STRUCT,
1429 | STRUCT name COMPLETE
1431 pstate->mark_completion_tag (TYPE_CODE_STRUCT,
1436 { $$ = lookup_struct
1437 (copy_name ($2).c_str (),
1438 pstate->expression_context_block);
1442 pstate->mark_completion_tag (TYPE_CODE_STRUCT,
1446 | CLASS name COMPLETE
1448 pstate->mark_completion_tag (TYPE_CODE_STRUCT,
1454 = lookup_union (copy_name ($2).c_str (),
1455 pstate->expression_context_block);
1459 pstate->mark_completion_tag (TYPE_CODE_UNION,
1463 | UNION name COMPLETE
1465 pstate->mark_completion_tag (TYPE_CODE_UNION,
1470 { $$ = lookup_enum (copy_name ($2).c_str (),
1471 pstate->expression_context_block);
1475 pstate->mark_completion_tag (TYPE_CODE_ENUM, "", 0);
1478 | ENUM name COMPLETE
1480 pstate->mark_completion_tag (TYPE_CODE_ENUM, $2.ptr,
1484 | UNSIGNED type_name
1485 { $$ = lookup_unsigned_typename (pstate->language (),
1487 TYPE_NAME($2.type)); }
1489 { $$ = lookup_unsigned_typename (pstate->language (),
1492 | SIGNED_KEYWORD type_name
1493 { $$ = lookup_signed_typename (pstate->language (),
1495 TYPE_NAME($2.type)); }
1497 { $$ = lookup_signed_typename (pstate->language (),
1500 /* It appears that this rule for templates is never
1501 reduced; template recognition happens by lookahead
1502 in the token processing code in yylex. */
1503 | TEMPLATE name '<' type '>'
1504 { $$ = lookup_template_type
1505 (copy_name($2).c_str (), $4,
1506 pstate->expression_context_block);
1508 | const_or_volatile_or_space_identifier_noopt typebase
1509 { $$ = cpstate->type_stack.follow_types ($2); }
1510 | typebase const_or_volatile_or_space_identifier_noopt
1511 { $$ = cpstate->type_stack.follow_types ($1); }
1517 $$.stoken.ptr = "int";
1518 $$.stoken.length = 3;
1519 $$.type = lookup_signed_typename (pstate->language (),
1525 $$.stoken.ptr = "long";
1526 $$.stoken.length = 4;
1527 $$.type = lookup_signed_typename (pstate->language (),
1533 $$.stoken.ptr = "short";
1534 $$.stoken.length = 5;
1535 $$.type = lookup_signed_typename (pstate->language (),
1543 { check_parameter_typelist ($1); }
1544 | nonempty_typelist ',' DOTDOTDOT
1546 $1->push_back (NULL);
1547 check_parameter_typelist ($1);
1555 std::vector<struct type *> *typelist
1556 = new std::vector<struct type *>;
1557 cpstate->type_lists.emplace_back (typelist);
1559 typelist->push_back ($1);
1562 | nonempty_typelist ',' type
1572 cpstate->type_stack.push ($2);
1573 $$ = cpstate->type_stack.follow_types ($1);
1577 conversion_type_id: typebase conversion_declarator
1578 { $$ = cpstate->type_stack.follow_types ($1); }
1581 conversion_declarator: /* Nothing. */
1582 | ptr_operator conversion_declarator
1585 const_and_volatile: CONST_KEYWORD VOLATILE_KEYWORD
1586 | VOLATILE_KEYWORD CONST_KEYWORD
1589 const_or_volatile_noopt: const_and_volatile
1590 { cpstate->type_stack.insert (tp_const);
1591 cpstate->type_stack.insert (tp_volatile);
1594 { cpstate->type_stack.insert (tp_const); }
1596 { cpstate->type_stack.insert (tp_volatile); }
1600 { $$ = operator_stoken (" new"); }
1602 { $$ = operator_stoken (" delete"); }
1603 | OPERATOR NEW '[' ']'
1604 { $$ = operator_stoken (" new[]"); }
1605 | OPERATOR DELETE '[' ']'
1606 { $$ = operator_stoken (" delete[]"); }
1607 | OPERATOR NEW OBJC_LBRAC ']'
1608 { $$ = operator_stoken (" new[]"); }
1609 | OPERATOR DELETE OBJC_LBRAC ']'
1610 { $$ = operator_stoken (" delete[]"); }
1612 { $$ = operator_stoken ("+"); }
1614 { $$ = operator_stoken ("-"); }
1616 { $$ = operator_stoken ("*"); }
1618 { $$ = operator_stoken ("/"); }
1620 { $$ = operator_stoken ("%"); }
1622 { $$ = operator_stoken ("^"); }
1624 { $$ = operator_stoken ("&"); }
1626 { $$ = operator_stoken ("|"); }
1628 { $$ = operator_stoken ("~"); }
1630 { $$ = operator_stoken ("!"); }
1632 { $$ = operator_stoken ("="); }
1634 { $$ = operator_stoken ("<"); }
1636 { $$ = operator_stoken (">"); }
1637 | OPERATOR ASSIGN_MODIFY
1638 { const char *op = " unknown";
1662 case BINOP_BITWISE_IOR:
1665 case BINOP_BITWISE_AND:
1668 case BINOP_BITWISE_XOR:
1675 $$ = operator_stoken (op);
1678 { $$ = operator_stoken ("<<"); }
1680 { $$ = operator_stoken (">>"); }
1682 { $$ = operator_stoken ("=="); }
1684 { $$ = operator_stoken ("!="); }
1686 { $$ = operator_stoken ("<="); }
1688 { $$ = operator_stoken (">="); }
1690 { $$ = operator_stoken ("&&"); }
1692 { $$ = operator_stoken ("||"); }
1693 | OPERATOR INCREMENT
1694 { $$ = operator_stoken ("++"); }
1695 | OPERATOR DECREMENT
1696 { $$ = operator_stoken ("--"); }
1698 { $$ = operator_stoken (","); }
1699 | OPERATOR ARROW_STAR
1700 { $$ = operator_stoken ("->*"); }
1702 { $$ = operator_stoken ("->"); }
1704 { $$ = operator_stoken ("()"); }
1706 { $$ = operator_stoken ("[]"); }
1707 | OPERATOR OBJC_LBRAC ']'
1708 { $$ = operator_stoken ("[]"); }
1709 | OPERATOR conversion_type_id
1712 c_print_type ($2, NULL, &buf, -1, 0,
1713 &type_print_raw_options);
1715 /* This also needs canonicalization. */
1717 = cp_canonicalize_string (buf.c_str ());
1719 canon = std::move (buf.string ());
1720 $$ = operator_stoken ((" " + canon).c_str ());
1724 /* This rule exists in order to allow some tokens that would not normally
1725 match the 'name' rule to appear as fields within a struct. The example
1726 that initially motivated this was the RISC-V target which models the
1727 floating point registers as a union with fields called 'float' and
1728 'double'. The 'float' string becomes a TYPENAME token and can appear
1729 anywhere a 'name' can, however 'double' is its own token,
1730 DOUBLE_KEYWORD, and doesn't match the 'name' rule.*/
1733 | DOUBLE_KEYWORD { $$ = typename_stoken ("double"); }
1734 | INT_KEYWORD { $$ = typename_stoken ("int"); }
1735 | LONG { $$ = typename_stoken ("long"); }
1736 | SHORT { $$ = typename_stoken ("short"); }
1737 | SIGNED_KEYWORD { $$ = typename_stoken ("signed"); }
1738 | UNSIGNED { $$ = typename_stoken ("unsigned"); }
1741 name : NAME { $$ = $1.stoken; }
1742 | BLOCKNAME { $$ = $1.stoken; }
1743 | TYPENAME { $$ = $1.stoken; }
1744 | NAME_OR_INT { $$ = $1.stoken; }
1745 | UNKNOWN_CPP_NAME { $$ = $1.stoken; }
1749 name_not_typename : NAME
1751 /* These would be useful if name_not_typename was useful, but it is just
1752 a fake for "variable", so these cause reduce/reduce conflicts because
1753 the parser can't tell whether NAME_OR_INT is a name_not_typename (=variable,
1754 =exp) or just an exp. If name_not_typename was ever used in an lvalue
1755 context where only a name could occur, this might be useful.
1760 struct field_of_this_result is_a_field_of_this;
1764 = lookup_symbol ($1.ptr,
1765 pstate->expression_context_block,
1767 &is_a_field_of_this);
1768 $$.is_a_field_of_this
1769 = is_a_field_of_this.type != NULL;
1776 /* Like write_exp_string, but prepends a '~'. */
1779 write_destructor_name (struct parser_state *par_state, struct stoken token)
1781 char *copy = (char *) alloca (token.length + 1);
1784 memcpy (©[1], token.ptr, token.length);
1789 write_exp_string (par_state, token);
1792 /* Returns a stoken of the operator name given by OP (which does not
1793 include the string "operator"). */
1795 static struct stoken
1796 operator_stoken (const char *op)
1798 struct stoken st = { NULL, 0 };
1801 st.length = CP_OPERATOR_LEN + strlen (op);
1802 buf = (char *) malloc (st.length + 1);
1803 strcpy (buf, CP_OPERATOR_STR);
1807 /* The toplevel (c_parse) will free the memory allocated here. */
1808 cpstate->strings.emplace_back (buf);
1812 /* Returns a stoken of the type named TYPE. */
1814 static struct stoken
1815 typename_stoken (const char *type)
1817 struct stoken st = { type, 0 };
1818 st.length = strlen (type);
1822 /* Return true if the type is aggregate-like. */
1825 type_aggregate_p (struct type *type)
1827 return (TYPE_CODE (type) == TYPE_CODE_STRUCT
1828 || TYPE_CODE (type) == TYPE_CODE_UNION
1829 || TYPE_CODE (type) == TYPE_CODE_NAMESPACE
1830 || (TYPE_CODE (type) == TYPE_CODE_ENUM
1831 && TYPE_DECLARED_CLASS (type)));
1834 /* Validate a parameter typelist. */
1837 check_parameter_typelist (std::vector<struct type *> *params)
1842 for (ix = 0; ix < params->size (); ++ix)
1844 type = (*params)[ix];
1845 if (type != NULL && TYPE_CODE (check_typedef (type)) == TYPE_CODE_VOID)
1849 if (params->size () == 1)
1854 error (_("parameter types following 'void'"));
1857 error (_("'void' invalid as parameter type"));
1862 /* Take care of parsing a number (anything that starts with a digit).
1863 Set yylval and return the token type; update lexptr.
1864 LEN is the number of characters in it. */
1866 /*** Needs some error checking for the float case ***/
1869 parse_number (struct parser_state *par_state,
1870 const char *buf, int len, int parsed_float, YYSTYPE *putithere)
1878 int base = input_radix;
1881 /* Number of "L" suffixes encountered. */
1884 /* We have found a "L" or "U" suffix. */
1885 int found_suffix = 0;
1888 struct type *signed_type;
1889 struct type *unsigned_type;
1892 p = (char *) alloca (len);
1893 memcpy (p, buf, len);
1897 /* Handle suffixes for decimal floating-point: "df", "dd" or "dl". */
1898 if (len >= 2 && p[len - 2] == 'd' && p[len - 1] == 'f')
1900 putithere->typed_val_float.type
1901 = parse_type (par_state)->builtin_decfloat;
1904 else if (len >= 2 && p[len - 2] == 'd' && p[len - 1] == 'd')
1906 putithere->typed_val_float.type
1907 = parse_type (par_state)->builtin_decdouble;
1910 else if (len >= 2 && p[len - 2] == 'd' && p[len - 1] == 'l')
1912 putithere->typed_val_float.type
1913 = parse_type (par_state)->builtin_declong;
1916 /* Handle suffixes: 'f' for float, 'l' for long double. */
1917 else if (len >= 1 && TOLOWER (p[len - 1]) == 'f')
1919 putithere->typed_val_float.type
1920 = parse_type (par_state)->builtin_float;
1923 else if (len >= 1 && TOLOWER (p[len - 1]) == 'l')
1925 putithere->typed_val_float.type
1926 = parse_type (par_state)->builtin_long_double;
1929 /* Default type for floating-point literals is double. */
1932 putithere->typed_val_float.type
1933 = parse_type (par_state)->builtin_double;
1936 if (!parse_float (p, len,
1937 putithere->typed_val_float.type,
1938 putithere->typed_val_float.val))
1943 /* Handle base-switching prefixes 0x, 0t, 0d, 0 */
1944 if (p[0] == '0' && len > 1)
1987 if (c >= 'A' && c <= 'Z')
1989 if (c != 'l' && c != 'u')
1991 if (c >= '0' && c <= '9')
1999 if (base > 10 && c >= 'a' && c <= 'f')
2003 n += i = c - 'a' + 10;
2016 return ERROR; /* Char not a digit */
2019 return ERROR; /* Invalid digit in this base */
2021 /* Portably test for overflow (only works for nonzero values, so make
2022 a second check for zero). FIXME: Can't we just make n and prevn
2023 unsigned and avoid this? */
2024 if (c != 'l' && c != 'u' && (prevn >= n) && n != 0)
2025 unsigned_p = 1; /* Try something unsigned */
2027 /* Portably test for unsigned overflow.
2028 FIXME: This check is wrong; for example it doesn't find overflow
2029 on 0x123456789 when LONGEST is 32 bits. */
2030 if (c != 'l' && c != 'u' && n != 0)
2032 if (unsigned_p && prevn >= n)
2033 error (_("Numeric constant too large."));
2038 /* An integer constant is an int, a long, or a long long. An L
2039 suffix forces it to be long; an LL suffix forces it to be long
2040 long. If not forced to a larger size, it gets the first type of
2041 the above that it fits in. To figure out whether it fits, we
2042 shift it right and see whether anything remains. Note that we
2043 can't shift sizeof (LONGEST) * HOST_CHAR_BIT bits or more in one
2044 operation, because many compilers will warn about such a shift
2045 (which always produces a zero result). Sometimes gdbarch_int_bit
2046 or gdbarch_long_bit will be that big, sometimes not. To deal with
2047 the case where it is we just always shift the value more than
2048 once, with fewer bits each time. */
2052 && (un >> (gdbarch_int_bit (par_state->gdbarch ()) - 2)) == 0)
2055 = ((ULONGEST)1) << (gdbarch_int_bit (par_state->gdbarch ()) - 1);
2057 /* A large decimal (not hex or octal) constant (between INT_MAX
2058 and UINT_MAX) is a long or unsigned long, according to ANSI,
2059 never an unsigned int, but this code treats it as unsigned
2060 int. This probably should be fixed. GCC gives a warning on
2063 unsigned_type = parse_type (par_state)->builtin_unsigned_int;
2064 signed_type = parse_type (par_state)->builtin_int;
2066 else if (long_p <= 1
2067 && (un >> (gdbarch_long_bit (par_state->gdbarch ()) - 2)) == 0)
2070 = ((ULONGEST)1) << (gdbarch_long_bit (par_state->gdbarch ()) - 1);
2071 unsigned_type = parse_type (par_state)->builtin_unsigned_long;
2072 signed_type = parse_type (par_state)->builtin_long;
2077 if (sizeof (ULONGEST) * HOST_CHAR_BIT
2078 < gdbarch_long_long_bit (par_state->gdbarch ()))
2079 /* A long long does not fit in a LONGEST. */
2080 shift = (sizeof (ULONGEST) * HOST_CHAR_BIT - 1);
2082 shift = (gdbarch_long_long_bit (par_state->gdbarch ()) - 1);
2083 high_bit = (ULONGEST) 1 << shift;
2084 unsigned_type = parse_type (par_state)->builtin_unsigned_long_long;
2085 signed_type = parse_type (par_state)->builtin_long_long;
2088 putithere->typed_val_int.val = n;
2090 /* If the high bit of the worked out type is set then this number
2091 has to be unsigned. */
2093 if (unsigned_p || (n & high_bit))
2095 putithere->typed_val_int.type = unsigned_type;
2099 putithere->typed_val_int.type = signed_type;
2105 /* Temporary obstack used for holding strings. */
2106 static struct obstack tempbuf;
2107 static int tempbuf_init;
2109 /* Parse a C escape sequence. The initial backslash of the sequence
2110 is at (*PTR)[-1]. *PTR will be updated to point to just after the
2111 last character of the sequence. If OUTPUT is not NULL, the
2112 translated form of the escape sequence will be written there. If
2113 OUTPUT is NULL, no output is written and the call will only affect
2114 *PTR. If an escape sequence is expressed in target bytes, then the
2115 entire sequence will simply be copied to OUTPUT. Return 1 if any
2116 character was emitted, 0 otherwise. */
2119 c_parse_escape (const char **ptr, struct obstack *output)
2121 const char *tokptr = *ptr;
2124 /* Some escape sequences undergo character set conversion. Those we
2128 /* Hex escapes do not undergo character set conversion, so keep
2129 the escape sequence for later. */
2132 obstack_grow_str (output, "\\x");
2134 if (!ISXDIGIT (*tokptr))
2135 error (_("\\x escape without a following hex digit"));
2136 while (ISXDIGIT (*tokptr))
2139 obstack_1grow (output, *tokptr);
2144 /* Octal escapes do not undergo character set conversion, so
2145 keep the escape sequence for later. */
2157 obstack_grow_str (output, "\\");
2159 i < 3 && ISDIGIT (*tokptr) && *tokptr != '8' && *tokptr != '9';
2163 obstack_1grow (output, *tokptr);
2169 /* We handle UCNs later. We could handle them here, but that
2170 would mean a spurious error in the case where the UCN could
2171 be converted to the target charset but not the host
2177 int i, len = c == 'U' ? 8 : 4;
2180 obstack_1grow (output, '\\');
2181 obstack_1grow (output, *tokptr);
2184 if (!ISXDIGIT (*tokptr))
2185 error (_("\\%c escape without a following hex digit"), c);
2186 for (i = 0; i < len && ISXDIGIT (*tokptr); ++i)
2189 obstack_1grow (output, *tokptr);
2195 /* We must pass backslash through so that it does not
2196 cause quoting during the second expansion. */
2199 obstack_grow_str (output, "\\\\");
2203 /* Escapes which undergo conversion. */
2206 obstack_1grow (output, '\a');
2211 obstack_1grow (output, '\b');
2216 obstack_1grow (output, '\f');
2221 obstack_1grow (output, '\n');
2226 obstack_1grow (output, '\r');
2231 obstack_1grow (output, '\t');
2236 obstack_1grow (output, '\v');
2240 /* GCC extension. */
2243 obstack_1grow (output, HOST_ESCAPE_CHAR);
2247 /* Backslash-newline expands to nothing at all. */
2253 /* A few escapes just expand to the character itself. */
2257 /* GCC extensions. */
2262 /* Unrecognized escapes turn into the character itself. */
2265 obstack_1grow (output, *tokptr);
2273 /* Parse a string or character literal from TOKPTR. The string or
2274 character may be wide or unicode. *OUTPTR is set to just after the
2275 end of the literal in the input string. The resulting token is
2276 stored in VALUE. This returns a token value, either STRING or
2277 CHAR, depending on what was parsed. *HOST_CHARS is set to the
2278 number of host characters in the literal. */
2281 parse_string_or_char (const char *tokptr, const char **outptr,
2282 struct typed_stoken *value, int *host_chars)
2288 /* Build the gdb internal form of the input string in tempbuf. Note
2289 that the buffer is null byte terminated *only* for the
2290 convenience of debugging gdb itself and printing the buffer
2291 contents when the buffer contains no embedded nulls. Gdb does
2292 not depend upon the buffer being null byte terminated, it uses
2293 the length string instead. This allows gdb to handle C strings
2294 (as well as strings in other languages) with embedded null
2300 obstack_free (&tempbuf, NULL);
2301 obstack_init (&tempbuf);
2303 /* Record the string type. */
2306 type = C_WIDE_STRING;
2309 else if (*tokptr == 'u')
2314 else if (*tokptr == 'U')
2319 else if (*tokptr == '@')
2321 /* An Objective C string. */
2329 /* Skip the quote. */
2343 *host_chars += c_parse_escape (&tokptr, &tempbuf);
2345 else if (c == quote)
2349 obstack_1grow (&tempbuf, c);
2351 /* FIXME: this does the wrong thing with multi-byte host
2352 characters. We could use mbrlen here, but that would
2353 make "set host-charset" a bit less useful. */
2358 if (*tokptr != quote)
2361 error (_("Unterminated string in expression."));
2363 error (_("Unmatched single quote."));
2368 value->ptr = (char *) obstack_base (&tempbuf);
2369 value->length = obstack_object_size (&tempbuf);
2373 return quote == '"' ? (is_objc ? NSSTRING : STRING) : CHAR;
2376 /* This is used to associate some attributes with a token. */
2380 /* If this bit is set, the token is C++-only. */
2384 /* If this bit is set, the token is conditional: if there is a
2385 symbol of the same name, then the token is a symbol; otherwise,
2386 the token is a keyword. */
2390 DEF_ENUM_FLAGS_TYPE (enum token_flag, token_flags);
2396 enum exp_opcode opcode;
2400 static const struct token tokentab3[] =
2402 {">>=", ASSIGN_MODIFY, BINOP_RSH, 0},
2403 {"<<=", ASSIGN_MODIFY, BINOP_LSH, 0},
2404 {"->*", ARROW_STAR, BINOP_END, FLAG_CXX},
2405 {"...", DOTDOTDOT, BINOP_END, 0}
2408 static const struct token tokentab2[] =
2410 {"+=", ASSIGN_MODIFY, BINOP_ADD, 0},
2411 {"-=", ASSIGN_MODIFY, BINOP_SUB, 0},
2412 {"*=", ASSIGN_MODIFY, BINOP_MUL, 0},
2413 {"/=", ASSIGN_MODIFY, BINOP_DIV, 0},
2414 {"%=", ASSIGN_MODIFY, BINOP_REM, 0},
2415 {"|=", ASSIGN_MODIFY, BINOP_BITWISE_IOR, 0},
2416 {"&=", ASSIGN_MODIFY, BINOP_BITWISE_AND, 0},
2417 {"^=", ASSIGN_MODIFY, BINOP_BITWISE_XOR, 0},
2418 {"++", INCREMENT, BINOP_END, 0},
2419 {"--", DECREMENT, BINOP_END, 0},
2420 {"->", ARROW, BINOP_END, 0},
2421 {"&&", ANDAND, BINOP_END, 0},
2422 {"||", OROR, BINOP_END, 0},
2423 /* "::" is *not* only C++: gdb overrides its meaning in several
2424 different ways, e.g., 'filename'::func, function::variable. */
2425 {"::", COLONCOLON, BINOP_END, 0},
2426 {"<<", LSH, BINOP_END, 0},
2427 {">>", RSH, BINOP_END, 0},
2428 {"==", EQUAL, BINOP_END, 0},
2429 {"!=", NOTEQUAL, BINOP_END, 0},
2430 {"<=", LEQ, BINOP_END, 0},
2431 {">=", GEQ, BINOP_END, 0},
2432 {".*", DOT_STAR, BINOP_END, FLAG_CXX}
2435 /* Identifier-like tokens. Only type-specifiers than can appear in
2436 multi-word type names (for example 'double' can appear in 'long
2437 double') need to be listed here. type-specifiers that are only ever
2438 single word (like 'float') are handled by the classify_name function. */
2439 static const struct token ident_tokens[] =
2441 {"unsigned", UNSIGNED, OP_NULL, 0},
2442 {"template", TEMPLATE, OP_NULL, FLAG_CXX},
2443 {"volatile", VOLATILE_KEYWORD, OP_NULL, 0},
2444 {"struct", STRUCT, OP_NULL, 0},
2445 {"signed", SIGNED_KEYWORD, OP_NULL, 0},
2446 {"sizeof", SIZEOF, OP_NULL, 0},
2447 {"_Alignof", ALIGNOF, OP_NULL, 0},
2448 {"alignof", ALIGNOF, OP_NULL, FLAG_CXX},
2449 {"double", DOUBLE_KEYWORD, OP_NULL, 0},
2450 {"false", FALSEKEYWORD, OP_NULL, FLAG_CXX},
2451 {"class", CLASS, OP_NULL, FLAG_CXX},
2452 {"union", UNION, OP_NULL, 0},
2453 {"short", SHORT, OP_NULL, 0},
2454 {"const", CONST_KEYWORD, OP_NULL, 0},
2455 {"enum", ENUM, OP_NULL, 0},
2456 {"long", LONG, OP_NULL, 0},
2457 {"true", TRUEKEYWORD, OP_NULL, FLAG_CXX},
2458 {"int", INT_KEYWORD, OP_NULL, 0},
2459 {"new", NEW, OP_NULL, FLAG_CXX},
2460 {"delete", DELETE, OP_NULL, FLAG_CXX},
2461 {"operator", OPERATOR, OP_NULL, FLAG_CXX},
2463 {"and", ANDAND, BINOP_END, FLAG_CXX},
2464 {"and_eq", ASSIGN_MODIFY, BINOP_BITWISE_AND, FLAG_CXX},
2465 {"bitand", '&', OP_NULL, FLAG_CXX},
2466 {"bitor", '|', OP_NULL, FLAG_CXX},
2467 {"compl", '~', OP_NULL, FLAG_CXX},
2468 {"not", '!', OP_NULL, FLAG_CXX},
2469 {"not_eq", NOTEQUAL, BINOP_END, FLAG_CXX},
2470 {"or", OROR, BINOP_END, FLAG_CXX},
2471 {"or_eq", ASSIGN_MODIFY, BINOP_BITWISE_IOR, FLAG_CXX},
2472 {"xor", '^', OP_NULL, FLAG_CXX},
2473 {"xor_eq", ASSIGN_MODIFY, BINOP_BITWISE_XOR, FLAG_CXX},
2475 {"const_cast", CONST_CAST, OP_NULL, FLAG_CXX },
2476 {"dynamic_cast", DYNAMIC_CAST, OP_NULL, FLAG_CXX },
2477 {"static_cast", STATIC_CAST, OP_NULL, FLAG_CXX },
2478 {"reinterpret_cast", REINTERPRET_CAST, OP_NULL, FLAG_CXX },
2480 {"__typeof__", TYPEOF, OP_TYPEOF, 0 },
2481 {"__typeof", TYPEOF, OP_TYPEOF, 0 },
2482 {"typeof", TYPEOF, OP_TYPEOF, FLAG_SHADOW },
2483 {"__decltype", DECLTYPE, OP_DECLTYPE, FLAG_CXX },
2484 {"decltype", DECLTYPE, OP_DECLTYPE, FLAG_CXX | FLAG_SHADOW },
2486 {"typeid", TYPEID, OP_TYPEID, FLAG_CXX}
2491 scan_macro_expansion (char *expansion)
2495 /* We'd better not be trying to push the stack twice. */
2496 gdb_assert (! cpstate->macro_original_text);
2498 /* Copy to the obstack, and then free the intermediate
2500 copy = (char *) obstack_copy0 (&cpstate->expansion_obstack, expansion,
2501 strlen (expansion));
2504 /* Save the old lexptr value, so we can return to it when we're done
2505 parsing the expanded text. */
2506 cpstate->macro_original_text = pstate->lexptr;
2507 pstate->lexptr = copy;
2511 scanning_macro_expansion (void)
2513 return cpstate->macro_original_text != 0;
2517 finished_macro_expansion (void)
2519 /* There'd better be something to pop back to. */
2520 gdb_assert (cpstate->macro_original_text);
2522 /* Pop back to the original text. */
2523 pstate->lexptr = cpstate->macro_original_text;
2524 cpstate->macro_original_text = 0;
2527 /* Return true iff the token represents a C++ cast operator. */
2530 is_cast_operator (const char *token, int len)
2532 return (! strncmp (token, "dynamic_cast", len)
2533 || ! strncmp (token, "static_cast", len)
2534 || ! strncmp (token, "reinterpret_cast", len)
2535 || ! strncmp (token, "const_cast", len));
2538 /* The scope used for macro expansion. */
2539 static struct macro_scope *expression_macro_scope;
2541 /* This is set if a NAME token appeared at the very end of the input
2542 string, with no whitespace separating the name from the EOF. This
2543 is used only when parsing to do field name completion. */
2544 static int saw_name_at_eof;
2546 /* This is set if the previously-returned token was a structure
2547 operator -- either '.' or ARROW. */
2548 static bool last_was_structop;
2550 /* Depth of parentheses. */
2551 static int paren_depth;
2553 /* Read one token, getting characters through lexptr. */
2556 lex_one_token (struct parser_state *par_state, bool *is_quoted_name)
2561 const char *tokstart;
2562 bool saw_structop = last_was_structop;
2564 last_was_structop = false;
2565 *is_quoted_name = false;
2569 /* Check if this is a macro invocation that we need to expand. */
2570 if (! scanning_macro_expansion ())
2572 char *expanded = macro_expand_next (&pstate->lexptr,
2573 standard_macro_lookup,
2574 expression_macro_scope);
2577 scan_macro_expansion (expanded);
2580 pstate->prev_lexptr = pstate->lexptr;
2582 tokstart = pstate->lexptr;
2583 /* See if it is a special token of length 3. */
2584 for (i = 0; i < sizeof tokentab3 / sizeof tokentab3[0]; i++)
2585 if (strncmp (tokstart, tokentab3[i].oper, 3) == 0)
2587 if ((tokentab3[i].flags & FLAG_CXX) != 0
2588 && par_state->language ()->la_language != language_cplus)
2591 pstate->lexptr += 3;
2592 yylval.opcode = tokentab3[i].opcode;
2593 return tokentab3[i].token;
2596 /* See if it is a special token of length 2. */
2597 for (i = 0; i < sizeof tokentab2 / sizeof tokentab2[0]; i++)
2598 if (strncmp (tokstart, tokentab2[i].oper, 2) == 0)
2600 if ((tokentab2[i].flags & FLAG_CXX) != 0
2601 && par_state->language ()->la_language != language_cplus)
2604 pstate->lexptr += 2;
2605 yylval.opcode = tokentab2[i].opcode;
2606 if (tokentab2[i].token == ARROW)
2607 last_was_structop = 1;
2608 return tokentab2[i].token;
2611 switch (c = *tokstart)
2614 /* If we were just scanning the result of a macro expansion,
2615 then we need to resume scanning the original text.
2616 If we're parsing for field name completion, and the previous
2617 token allows such completion, return a COMPLETE token.
2618 Otherwise, we were already scanning the original text, and
2619 we're really done. */
2620 if (scanning_macro_expansion ())
2622 finished_macro_expansion ();
2625 else if (saw_name_at_eof)
2627 saw_name_at_eof = 0;
2630 else if (par_state->parse_completion && saw_structop)
2645 if (par_state->language ()->la_language == language_objc
2652 if (paren_depth == 0)
2659 if (pstate->comma_terminates
2661 && ! scanning_macro_expansion ())
2667 /* Might be a floating point number. */
2668 if (pstate->lexptr[1] < '0' || pstate->lexptr[1] > '9')
2670 last_was_structop = true;
2671 goto symbol; /* Nope, must be a symbol. */
2686 /* It's a number. */
2687 int got_dot = 0, got_e = 0, toktype;
2688 const char *p = tokstart;
2689 int hex = input_radix > 10;
2691 if (c == '0' && (p[1] == 'x' || p[1] == 'X'))
2696 else if (c == '0' && (p[1]=='t' || p[1]=='T' || p[1]=='d' || p[1]=='D'))
2704 /* This test includes !hex because 'e' is a valid hex digit
2705 and thus does not indicate a floating point number when
2706 the radix is hex. */
2707 if (!hex && !got_e && (*p == 'e' || *p == 'E'))
2708 got_dot = got_e = 1;
2709 /* This test does not include !hex, because a '.' always indicates
2710 a decimal floating point number regardless of the radix. */
2711 else if (!got_dot && *p == '.')
2713 else if (got_e && (p[-1] == 'e' || p[-1] == 'E')
2714 && (*p == '-' || *p == '+'))
2715 /* This is the sign of the exponent, not the end of the
2718 /* We will take any letters or digits. parse_number will
2719 complain if past the radix, or if L or U are not final. */
2720 else if ((*p < '0' || *p > '9')
2721 && ((*p < 'a' || *p > 'z')
2722 && (*p < 'A' || *p > 'Z')))
2725 toktype = parse_number (par_state, tokstart, p - tokstart,
2726 got_dot|got_e, &yylval);
2727 if (toktype == ERROR)
2729 char *err_copy = (char *) alloca (p - tokstart + 1);
2731 memcpy (err_copy, tokstart, p - tokstart);
2732 err_copy[p - tokstart] = 0;
2733 error (_("Invalid number \"%s\"."), err_copy);
2741 const char *p = &tokstart[1];
2743 if (par_state->language ()->la_language == language_objc)
2745 size_t len = strlen ("selector");
2747 if (strncmp (p, "selector", len) == 0
2748 && (p[len] == '\0' || ISSPACE (p[len])))
2750 pstate->lexptr = p + len;
2757 while (ISSPACE (*p))
2759 size_t len = strlen ("entry");
2760 if (strncmp (p, "entry", len) == 0 && !c_ident_is_alnum (p[len])
2763 pstate->lexptr = &p[len];
2792 if (tokstart[1] != '"' && tokstart[1] != '\'')
2801 int result = parse_string_or_char (tokstart, &pstate->lexptr,
2802 &yylval.tsval, &host_len);
2806 error (_("Empty character constant."));
2807 else if (host_len > 2 && c == '\'')
2810 namelen = pstate->lexptr - tokstart - 1;
2811 *is_quoted_name = true;
2815 else if (host_len > 1)
2816 error (_("Invalid character constant."));
2822 if (!(c == '_' || c == '$' || c_ident_is_alpha (c)))
2823 /* We must have come across a bad character (e.g. ';'). */
2824 error (_("Invalid character '%c' in expression."), c);
2826 /* It's a name. See how long it is. */
2828 for (c = tokstart[namelen];
2829 (c == '_' || c == '$' || c_ident_is_alnum (c) || c == '<');)
2831 /* Template parameter lists are part of the name.
2832 FIXME: This mishandles `print $a<4&&$a>3'. */
2836 if (! is_cast_operator (tokstart, namelen))
2838 /* Scan ahead to get rest of the template specification. Note
2839 that we look ahead only when the '<' adjoins non-whitespace
2840 characters; for comparison expressions, e.g. "a < b > c",
2841 there must be spaces before the '<', etc. */
2842 const char *p = find_template_name_end (tokstart + namelen);
2845 namelen = p - tokstart;
2849 c = tokstart[++namelen];
2852 /* The token "if" terminates the expression and is NOT removed from
2853 the input stream. It doesn't count if it appears in the
2854 expansion of a macro. */
2856 && tokstart[0] == 'i'
2857 && tokstart[1] == 'f'
2858 && ! scanning_macro_expansion ())
2863 /* For the same reason (breakpoint conditions), "thread N"
2864 terminates the expression. "thread" could be an identifier, but
2865 an identifier is never followed by a number without intervening
2866 punctuation. "task" is similar. Handle abbreviations of these,
2867 similarly to breakpoint.c:find_condition_and_thread. */
2869 && (strncmp (tokstart, "thread", namelen) == 0
2870 || strncmp (tokstart, "task", namelen) == 0)
2871 && (tokstart[namelen] == ' ' || tokstart[namelen] == '\t')
2872 && ! scanning_macro_expansion ())
2874 const char *p = tokstart + namelen + 1;
2876 while (*p == ' ' || *p == '\t')
2878 if (*p >= '0' && *p <= '9')
2882 pstate->lexptr += namelen;
2886 yylval.sval.ptr = tokstart;
2887 yylval.sval.length = namelen;
2889 /* Catch specific keywords. */
2890 std::string copy = copy_name (yylval.sval);
2891 for (i = 0; i < sizeof ident_tokens / sizeof ident_tokens[0]; i++)
2892 if (copy == ident_tokens[i].oper)
2894 if ((ident_tokens[i].flags & FLAG_CXX) != 0
2895 && par_state->language ()->la_language != language_cplus)
2898 if ((ident_tokens[i].flags & FLAG_SHADOW) != 0)
2900 struct field_of_this_result is_a_field_of_this;
2902 if (lookup_symbol (copy.c_str (),
2903 pstate->expression_context_block,
2905 (par_state->language ()->la_language
2906 == language_cplus ? &is_a_field_of_this
2910 /* The keyword is shadowed. */
2915 /* It is ok to always set this, even though we don't always
2916 strictly need to. */
2917 yylval.opcode = ident_tokens[i].opcode;
2918 return ident_tokens[i].token;
2921 if (*tokstart == '$')
2922 return DOLLAR_VARIABLE;
2924 if (pstate->parse_completion && *pstate->lexptr == '\0')
2925 saw_name_at_eof = 1;
2927 yylval.ssym.stoken = yylval.sval;
2928 yylval.ssym.sym.symbol = NULL;
2929 yylval.ssym.sym.block = NULL;
2930 yylval.ssym.is_a_field_of_this = 0;
2934 /* An object of this type is pushed on a FIFO by the "outer" lexer. */
2935 struct token_and_value
2941 /* A FIFO of tokens that have been read but not yet returned to the
2943 static std::vector<token_and_value> token_fifo;
2945 /* Non-zero if the lexer should return tokens from the FIFO. */
2948 /* Temporary storage for c_lex; this holds symbol names as they are
2950 auto_obstack name_obstack;
2952 /* Classify a NAME token. The contents of the token are in `yylval'.
2953 Updates yylval and returns the new token type. BLOCK is the block
2954 in which lookups start; this can be NULL to mean the global scope.
2955 IS_QUOTED_NAME is non-zero if the name token was originally quoted
2956 in single quotes. IS_AFTER_STRUCTOP is true if this name follows
2957 a structure operator -- either '.' or ARROW */
2960 classify_name (struct parser_state *par_state, const struct block *block,
2961 bool is_quoted_name, bool is_after_structop)
2963 struct block_symbol bsym;
2964 struct field_of_this_result is_a_field_of_this;
2966 std::string copy = copy_name (yylval.sval);
2968 /* Initialize this in case we *don't* use it in this call; that way
2969 we can refer to it unconditionally below. */
2970 memset (&is_a_field_of_this, 0, sizeof (is_a_field_of_this));
2972 bsym = lookup_symbol (copy.c_str (), block, VAR_DOMAIN,
2973 par_state->language ()->la_name_of_this
2974 ? &is_a_field_of_this : NULL);
2976 if (bsym.symbol && SYMBOL_CLASS (bsym.symbol) == LOC_BLOCK)
2978 yylval.ssym.sym = bsym;
2979 yylval.ssym.is_a_field_of_this = is_a_field_of_this.type != NULL;
2982 else if (!bsym.symbol)
2984 /* If we found a field of 'this', we might have erroneously
2985 found a constructor where we wanted a type name. Handle this
2986 case by noticing that we found a constructor and then look up
2987 the type tag instead. */
2988 if (is_a_field_of_this.type != NULL
2989 && is_a_field_of_this.fn_field != NULL
2990 && TYPE_FN_FIELD_CONSTRUCTOR (is_a_field_of_this.fn_field->fn_fields,
2993 struct field_of_this_result inner_is_a_field_of_this;
2995 bsym = lookup_symbol (copy.c_str (), block, STRUCT_DOMAIN,
2996 &inner_is_a_field_of_this);
2997 if (bsym.symbol != NULL)
2999 yylval.tsym.type = SYMBOL_TYPE (bsym.symbol);
3004 /* If we found a field on the "this" object, or we are looking
3005 up a field on a struct, then we want to prefer it over a
3006 filename. However, if the name was quoted, then it is better
3007 to check for a filename or a block, since this is the only
3008 way the user has of requiring the extension to be used. */
3009 if ((is_a_field_of_this.type == NULL && !is_after_structop)
3012 /* See if it's a file name. */
3013 struct symtab *symtab;
3015 symtab = lookup_symtab (copy.c_str ());
3018 yylval.bval = BLOCKVECTOR_BLOCK (SYMTAB_BLOCKVECTOR (symtab),
3025 if (bsym.symbol && SYMBOL_CLASS (bsym.symbol) == LOC_TYPEDEF)
3027 yylval.tsym.type = SYMBOL_TYPE (bsym.symbol);
3031 /* See if it's an ObjC classname. */
3032 if (par_state->language ()->la_language == language_objc && !bsym.symbol)
3034 CORE_ADDR Class = lookup_objc_class (par_state->gdbarch (),
3040 yylval.theclass.theclass = Class;
3041 sym = lookup_struct_typedef (copy.c_str (),
3042 par_state->expression_context_block, 1);
3044 yylval.theclass.type = SYMBOL_TYPE (sym);
3049 /* Input names that aren't symbols but ARE valid hex numbers, when
3050 the input radix permits them, can be names or numbers depending
3051 on the parse. Note we support radixes > 16 here. */
3053 && ((copy[0] >= 'a' && copy[0] < 'a' + input_radix - 10)
3054 || (copy[0] >= 'A' && copy[0] < 'A' + input_radix - 10)))
3056 YYSTYPE newlval; /* Its value is ignored. */
3057 int hextype = parse_number (par_state, copy.c_str (), yylval.sval.length,
3062 yylval.ssym.sym = bsym;
3063 yylval.ssym.is_a_field_of_this = is_a_field_of_this.type != NULL;
3068 /* Any other kind of symbol */
3069 yylval.ssym.sym = bsym;
3070 yylval.ssym.is_a_field_of_this = is_a_field_of_this.type != NULL;
3072 if (bsym.symbol == NULL
3073 && par_state->language ()->la_language == language_cplus
3074 && is_a_field_of_this.type == NULL
3075 && lookup_minimal_symbol (copy.c_str (), NULL, NULL).minsym == NULL)
3076 return UNKNOWN_CPP_NAME;
3081 /* Like classify_name, but used by the inner loop of the lexer, when a
3082 name might have already been seen. CONTEXT is the context type, or
3083 NULL if this is the first component of a name. */
3086 classify_inner_name (struct parser_state *par_state,
3087 const struct block *block, struct type *context)
3091 if (context == NULL)
3092 return classify_name (par_state, block, false, false);
3094 type = check_typedef (context);
3095 if (!type_aggregate_p (type))
3098 std::string copy = copy_name (yylval.ssym.stoken);
3099 /* N.B. We assume the symbol can only be in VAR_DOMAIN. */
3100 yylval.ssym.sym = cp_lookup_nested_symbol (type, copy.c_str (), block,
3103 /* If no symbol was found, search for a matching base class named
3104 COPY. This will allow users to enter qualified names of class members
3105 relative to the `this' pointer. */
3106 if (yylval.ssym.sym.symbol == NULL)
3108 struct type *base_type = cp_find_type_baseclass_by_name (type,
3111 if (base_type != NULL)
3113 yylval.tsym.type = base_type;
3120 switch (SYMBOL_CLASS (yylval.ssym.sym.symbol))
3124 /* cp_lookup_nested_symbol might have accidentally found a constructor
3125 named COPY when we really wanted a base class of the same name.
3126 Double-check this case by looking for a base class. */
3128 struct type *base_type
3129 = cp_find_type_baseclass_by_name (type, copy.c_str ());
3131 if (base_type != NULL)
3133 yylval.tsym.type = base_type;
3140 yylval.tsym.type = SYMBOL_TYPE (yylval.ssym.sym.symbol);
3146 internal_error (__FILE__, __LINE__, _("not reached"));
3149 /* The outer level of a two-level lexer. This calls the inner lexer
3150 to return tokens. It then either returns these tokens, or
3151 aggregates them into a larger token. This lets us work around a
3152 problem in our parsing approach, where the parser could not
3153 distinguish between qualified names and qualified types at the
3156 This approach is still not ideal, because it mishandles template
3157 types. See the comment in lex_one_token for an example. However,
3158 this is still an improvement over the earlier approach, and will
3159 suffice until we move to better parsing technology. */
3164 token_and_value current;
3165 int first_was_coloncolon, last_was_coloncolon;
3166 struct type *context_type = NULL;
3167 int last_to_examine, next_to_examine, checkpoint;
3168 const struct block *search_block;
3169 bool is_quoted_name, last_lex_was_structop;
3171 if (popping && !token_fifo.empty ())
3175 last_lex_was_structop = last_was_structop;
3177 /* Read the first token and decide what to do. Most of the
3178 subsequent code is C++-only; but also depends on seeing a "::" or
3180 current.token = lex_one_token (pstate, &is_quoted_name);
3181 if (current.token == NAME)
3182 current.token = classify_name (pstate, pstate->expression_context_block,
3183 is_quoted_name, last_lex_was_structop);
3184 if (pstate->language ()->la_language != language_cplus
3185 || (current.token != TYPENAME && current.token != COLONCOLON
3186 && current.token != FILENAME))
3187 return current.token;
3189 /* Read any sequence of alternating "::" and name-like tokens into
3191 current.value = yylval;
3192 token_fifo.push_back (current);
3193 last_was_coloncolon = current.token == COLONCOLON;
3198 /* We ignore quoted names other than the very first one.
3199 Subsequent ones do not have any special meaning. */
3200 current.token = lex_one_token (pstate, &ignore);
3201 current.value = yylval;
3202 token_fifo.push_back (current);
3204 if ((last_was_coloncolon && current.token != NAME)
3205 || (!last_was_coloncolon && current.token != COLONCOLON))
3207 last_was_coloncolon = !last_was_coloncolon;
3211 /* We always read one extra token, so compute the number of tokens
3212 to examine accordingly. */
3213 last_to_examine = token_fifo.size () - 2;
3214 next_to_examine = 0;
3216 current = token_fifo[next_to_examine];
3219 name_obstack.clear ();
3221 if (current.token == FILENAME)
3222 search_block = current.value.bval;
3223 else if (current.token == COLONCOLON)
3224 search_block = NULL;
3227 gdb_assert (current.token == TYPENAME);
3228 search_block = pstate->expression_context_block;
3229 obstack_grow (&name_obstack, current.value.sval.ptr,
3230 current.value.sval.length);
3231 context_type = current.value.tsym.type;
3235 first_was_coloncolon = current.token == COLONCOLON;
3236 last_was_coloncolon = first_was_coloncolon;
3238 while (next_to_examine <= last_to_examine)
3240 token_and_value next;
3242 next = token_fifo[next_to_examine];
3245 if (next.token == NAME && last_was_coloncolon)
3249 yylval = next.value;
3250 classification = classify_inner_name (pstate, search_block,
3252 /* We keep going until we either run out of names, or until
3253 we have a qualified name which is not a type. */
3254 if (classification != TYPENAME && classification != NAME)
3257 /* Accept up to this token. */
3258 checkpoint = next_to_examine;
3260 /* Update the partial name we are constructing. */
3261 if (context_type != NULL)
3263 /* We don't want to put a leading "::" into the name. */
3264 obstack_grow_str (&name_obstack, "::");
3266 obstack_grow (&name_obstack, next.value.sval.ptr,
3267 next.value.sval.length);
3269 yylval.sval.ptr = (const char *) obstack_base (&name_obstack);
3270 yylval.sval.length = obstack_object_size (&name_obstack);
3271 current.value = yylval;
3272 current.token = classification;
3274 last_was_coloncolon = 0;
3276 if (classification == NAME)
3279 context_type = yylval.tsym.type;
3281 else if (next.token == COLONCOLON && !last_was_coloncolon)
3282 last_was_coloncolon = 1;
3285 /* We've reached the end of the name. */
3290 /* If we have a replacement token, install it as the first token in
3291 the FIFO, and delete the other constituent tokens. */
3294 current.value.sval.ptr
3295 = (const char *) obstack_copy0 (&cpstate->expansion_obstack,
3296 current.value.sval.ptr,
3297 current.value.sval.length);
3299 token_fifo[0] = current;
3301 token_fifo.erase (token_fifo.begin () + 1,
3302 token_fifo.begin () + checkpoint);
3306 current = token_fifo[0];
3307 token_fifo.erase (token_fifo.begin ());
3308 yylval = current.value;
3309 return current.token;
3313 c_parse (struct parser_state *par_state)
3315 /* Setting up the parser state. */
3316 scoped_restore pstate_restore = make_scoped_restore (&pstate);
3317 gdb_assert (par_state != NULL);
3320 c_parse_state cstate;
3321 scoped_restore cstate_restore = make_scoped_restore (&cpstate, &cstate);
3323 gdb::unique_xmalloc_ptr<struct macro_scope> macro_scope;
3325 if (par_state->expression_context_block)
3327 = sal_macro_scope (find_pc_line (par_state->expression_context_pc, 0));
3329 macro_scope = default_macro_scope ();
3331 macro_scope = user_macro_scope ();
3333 scoped_restore restore_macro_scope
3334 = make_scoped_restore (&expression_macro_scope, macro_scope.get ());
3336 scoped_restore restore_yydebug = make_scoped_restore (&yydebug,
3339 /* Initialize some state used by the lexer. */
3340 last_was_structop = false;
3341 saw_name_at_eof = 0;
3344 token_fifo.clear ();
3346 name_obstack.clear ();
3353 /* This is called via the YYPRINT macro when parser debugging is
3354 enabled. It prints a token's value. */
3357 c_print_token (FILE *file, int type, YYSTYPE value)
3362 parser_fprintf (file, "typed_val_int<%s, %s>",
3363 TYPE_SAFE_NAME (value.typed_val_int.type),
3364 pulongest (value.typed_val_int.val));
3370 char *copy = (char *) alloca (value.tsval.length + 1);
3372 memcpy (copy, value.tsval.ptr, value.tsval.length);
3373 copy[value.tsval.length] = '\0';
3375 parser_fprintf (file, "tsval<type=%d, %s>", value.tsval.type, copy);
3380 case DOLLAR_VARIABLE:
3381 parser_fprintf (file, "sval<%s>", copy_name (value.sval).c_str ());
3385 parser_fprintf (file, "tsym<type=%s, name=%s>",
3386 TYPE_SAFE_NAME (value.tsym.type),
3387 copy_name (value.tsym.stoken).c_str ());
3391 case UNKNOWN_CPP_NAME:
3394 parser_fprintf (file, "ssym<name=%s, sym=%s, field_of_this=%d>",
3395 copy_name (value.ssym.stoken).c_str (),
3396 (value.ssym.sym.symbol == NULL
3397 ? "(null)" : SYMBOL_PRINT_NAME (value.ssym.sym.symbol)),
3398 value.ssym.is_a_field_of_this);
3402 parser_fprintf (file, "bval<%s>", host_address_to_string (value.bval));
3410 yyerror (const char *msg)
3412 if (pstate->prev_lexptr)
3413 pstate->lexptr = pstate->prev_lexptr;
3415 error (_("A %s in expression, near `%s'."), msg, pstate->lexptr);