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 theclass = lookup_objc_class (pstate->gdbarch (),
478 copy_name ($2.stoken));
480 error (_("%s is not an ObjC Class"),
481 copy_name ($2.stoken));
482 write_exp_elt_opcode (pstate, OP_LONG);
483 write_exp_elt_type (pstate,
484 parse_type (pstate)->builtin_int);
485 write_exp_elt_longcst (pstate, (LONGEST) theclass);
486 write_exp_elt_opcode (pstate, OP_LONG);
490 { write_exp_elt_opcode (pstate, OP_OBJC_MSGCALL);
491 end_msglist (pstate);
492 write_exp_elt_opcode (pstate, OP_OBJC_MSGCALL);
496 exp : OBJC_LBRAC CLASSNAME
498 write_exp_elt_opcode (pstate, OP_LONG);
499 write_exp_elt_type (pstate,
500 parse_type (pstate)->builtin_int);
501 write_exp_elt_longcst (pstate, (LONGEST) $2.theclass);
502 write_exp_elt_opcode (pstate, OP_LONG);
506 { write_exp_elt_opcode (pstate, OP_OBJC_MSGCALL);
507 end_msglist (pstate);
508 write_exp_elt_opcode (pstate, OP_OBJC_MSGCALL);
515 { write_exp_elt_opcode (pstate, OP_OBJC_MSGCALL);
516 end_msglist (pstate);
517 write_exp_elt_opcode (pstate, OP_OBJC_MSGCALL);
522 { add_msglist(&$1, 0); }
530 msgarg : name ':' exp
531 { add_msglist(&$1, 1); }
532 | ':' exp /* Unnamed arg. */
533 { add_msglist(0, 1); }
534 | ',' exp /* Variable number of args. */
535 { add_msglist(0, 0); }
539 /* This is to save the value of arglist_len
540 being accumulated by an outer function call. */
541 { pstate->start_arglist (); }
542 arglist ')' %prec ARROW
543 { write_exp_elt_opcode (pstate, OP_FUNCALL);
544 write_exp_elt_longcst (pstate,
545 pstate->end_arglist ());
546 write_exp_elt_opcode (pstate, OP_FUNCALL); }
549 /* This is here to disambiguate with the production for
550 "func()::static_var" further below, which uses
551 function_method_void. */
552 exp : exp '(' ')' %prec ARROW
553 { pstate->start_arglist ();
554 write_exp_elt_opcode (pstate, OP_FUNCALL);
555 write_exp_elt_longcst (pstate,
556 pstate->end_arglist ());
557 write_exp_elt_opcode (pstate, OP_FUNCALL); }
561 exp : UNKNOWN_CPP_NAME '('
563 /* This could potentially be a an argument defined
564 lookup function (Koenig). */
565 write_exp_elt_opcode (pstate, OP_ADL_FUNC);
567 (pstate, pstate->expression_context_block);
568 write_exp_elt_sym (pstate,
569 NULL); /* Placeholder. */
570 write_exp_string (pstate, $1.stoken);
571 write_exp_elt_opcode (pstate, OP_ADL_FUNC);
573 /* This is to save the value of arglist_len
574 being accumulated by an outer function call. */
576 pstate->start_arglist ();
578 arglist ')' %prec ARROW
580 write_exp_elt_opcode (pstate, OP_FUNCALL);
581 write_exp_elt_longcst (pstate,
582 pstate->end_arglist ());
583 write_exp_elt_opcode (pstate, OP_FUNCALL);
588 { pstate->start_arglist (); }
595 { pstate->arglist_len = 1; }
598 arglist : arglist ',' exp %prec ABOVE_COMMA
599 { pstate->arglist_len++; }
602 function_method: exp '(' parameter_typelist ')' const_or_volatile
604 std::vector<struct type *> *type_list = $3;
605 LONGEST len = type_list->size ();
607 write_exp_elt_opcode (pstate, TYPE_INSTANCE);
608 /* Save the const/volatile qualifiers as
609 recorded by the const_or_volatile
610 production's actions. */
611 write_exp_elt_longcst
614 .follow_type_instance_flags ()));
615 write_exp_elt_longcst (pstate, len);
616 for (type *type_elt : *type_list)
617 write_exp_elt_type (pstate, type_elt);
618 write_exp_elt_longcst(pstate, len);
619 write_exp_elt_opcode (pstate, TYPE_INSTANCE);
623 function_method_void: exp '(' ')' const_or_volatile
624 { write_exp_elt_opcode (pstate, TYPE_INSTANCE);
626 write_exp_elt_longcst
628 cpstate->type_stack.follow_type_instance_flags ());
629 write_exp_elt_longcst (pstate, 0);
630 write_exp_elt_longcst (pstate, 0);
631 write_exp_elt_opcode (pstate, TYPE_INSTANCE);
635 exp : function_method
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
642 function_method_void_or_typelist: function_method
643 | function_method_void
646 exp : function_method_void_or_typelist COLONCOLON name
648 write_exp_elt_opcode (pstate, OP_FUNC_STATIC_VAR);
649 write_exp_string (pstate, $3);
650 write_exp_elt_opcode (pstate, OP_FUNC_STATIC_VAR);
655 { $$ = pstate->end_arglist () - 1; }
657 exp : lcurly arglist rcurly %prec ARROW
658 { write_exp_elt_opcode (pstate, OP_ARRAY);
659 write_exp_elt_longcst (pstate, (LONGEST) 0);
660 write_exp_elt_longcst (pstate, (LONGEST) $3);
661 write_exp_elt_opcode (pstate, OP_ARRAY); }
664 exp : lcurly type_exp rcurly exp %prec UNARY
665 { write_exp_elt_opcode (pstate, UNOP_MEMVAL_TYPE); }
668 exp : '(' type_exp ')' exp %prec UNARY
669 { write_exp_elt_opcode (pstate, UNOP_CAST_TYPE); }
676 /* Binary operators in order of decreasing precedence. */
679 { write_exp_elt_opcode (pstate, BINOP_REPEAT); }
683 { write_exp_elt_opcode (pstate, BINOP_MUL); }
687 { write_exp_elt_opcode (pstate, BINOP_DIV); }
691 { write_exp_elt_opcode (pstate, BINOP_REM); }
695 { write_exp_elt_opcode (pstate, BINOP_ADD); }
699 { write_exp_elt_opcode (pstate, BINOP_SUB); }
703 { write_exp_elt_opcode (pstate, BINOP_LSH); }
707 { write_exp_elt_opcode (pstate, BINOP_RSH); }
711 { write_exp_elt_opcode (pstate, BINOP_EQUAL); }
714 exp : exp NOTEQUAL exp
715 { write_exp_elt_opcode (pstate, BINOP_NOTEQUAL); }
719 { write_exp_elt_opcode (pstate, BINOP_LEQ); }
723 { write_exp_elt_opcode (pstate, BINOP_GEQ); }
727 { write_exp_elt_opcode (pstate, BINOP_LESS); }
731 { write_exp_elt_opcode (pstate, BINOP_GTR); }
735 { write_exp_elt_opcode (pstate, BINOP_BITWISE_AND); }
739 { write_exp_elt_opcode (pstate, BINOP_BITWISE_XOR); }
743 { write_exp_elt_opcode (pstate, BINOP_BITWISE_IOR); }
747 { write_exp_elt_opcode (pstate, BINOP_LOGICAL_AND); }
751 { write_exp_elt_opcode (pstate, BINOP_LOGICAL_OR); }
754 exp : exp '?' exp ':' exp %prec '?'
755 { write_exp_elt_opcode (pstate, TERNOP_COND); }
759 { write_exp_elt_opcode (pstate, BINOP_ASSIGN); }
762 exp : exp ASSIGN_MODIFY exp
763 { write_exp_elt_opcode (pstate, BINOP_ASSIGN_MODIFY);
764 write_exp_elt_opcode (pstate, $2);
765 write_exp_elt_opcode (pstate,
766 BINOP_ASSIGN_MODIFY); }
770 { write_exp_elt_opcode (pstate, OP_LONG);
771 write_exp_elt_type (pstate, $1.type);
772 write_exp_elt_longcst (pstate, (LONGEST) ($1.val));
773 write_exp_elt_opcode (pstate, OP_LONG); }
778 struct stoken_vector vec;
781 write_exp_string_vector (pstate, $1.type, &vec);
787 parse_number (pstate, $1.stoken.ptr,
788 $1.stoken.length, 0, &val);
789 write_exp_elt_opcode (pstate, OP_LONG);
790 write_exp_elt_type (pstate, val.typed_val_int.type);
791 write_exp_elt_longcst (pstate,
792 (LONGEST) val.typed_val_int.val);
793 write_exp_elt_opcode (pstate, OP_LONG);
799 { write_exp_elt_opcode (pstate, OP_FLOAT);
800 write_exp_elt_type (pstate, $1.type);
801 write_exp_elt_floatcst (pstate, $1.val);
802 write_exp_elt_opcode (pstate, OP_FLOAT); }
808 exp : DOLLAR_VARIABLE
810 write_dollar_variable (pstate, $1);
814 exp : SELECTOR '(' name ')'
816 write_exp_elt_opcode (pstate, OP_OBJC_SELECTOR);
817 write_exp_string (pstate, $3);
818 write_exp_elt_opcode (pstate, OP_OBJC_SELECTOR); }
821 exp : SIZEOF '(' type ')' %prec UNARY
822 { struct type *type = $3;
823 write_exp_elt_opcode (pstate, OP_LONG);
824 write_exp_elt_type (pstate, lookup_signed_typename
825 (pstate->language (),
828 type = check_typedef (type);
830 /* $5.3.3/2 of the C++ Standard (n3290 draft)
831 says of sizeof: "When applied to a reference
832 or a reference type, the result is the size of
833 the referenced type." */
834 if (TYPE_IS_REFERENCE (type))
835 type = check_typedef (TYPE_TARGET_TYPE (type));
836 write_exp_elt_longcst (pstate,
837 (LONGEST) TYPE_LENGTH (type));
838 write_exp_elt_opcode (pstate, OP_LONG); }
841 exp : REINTERPRET_CAST '<' type_exp '>' '(' exp ')' %prec UNARY
842 { write_exp_elt_opcode (pstate,
843 UNOP_REINTERPRET_CAST); }
846 exp : STATIC_CAST '<' type_exp '>' '(' exp ')' %prec UNARY
847 { write_exp_elt_opcode (pstate, UNOP_CAST_TYPE); }
850 exp : DYNAMIC_CAST '<' type_exp '>' '(' exp ')' %prec UNARY
851 { write_exp_elt_opcode (pstate, UNOP_DYNAMIC_CAST); }
854 exp : CONST_CAST '<' type_exp '>' '(' exp ')' %prec UNARY
855 { /* We could do more error checking here, but
856 it doesn't seem worthwhile. */
857 write_exp_elt_opcode (pstate, UNOP_CAST_TYPE); }
863 /* We copy the string here, and not in the
864 lexer, to guarantee that we do not leak a
865 string. Note that we follow the
866 NUL-termination convention of the
868 struct typed_stoken *vec = XNEW (struct typed_stoken);
873 vec->length = $1.length;
874 vec->ptr = (char *) malloc ($1.length + 1);
875 memcpy (vec->ptr, $1.ptr, $1.length + 1);
880 /* Note that we NUL-terminate here, but just
884 $$.tokens = XRESIZEVEC (struct typed_stoken,
887 p = (char *) malloc ($2.length + 1);
888 memcpy (p, $2.ptr, $2.length + 1);
890 $$.tokens[$$.len - 1].type = $2.type;
891 $$.tokens[$$.len - 1].length = $2.length;
892 $$.tokens[$$.len - 1].ptr = p;
899 c_string_type type = C_STRING;
901 for (i = 0; i < $1.len; ++i)
903 switch ($1.tokens[i].type)
911 && type != $1.tokens[i].type)
912 error (_("Undefined string concatenation."));
913 type = (enum c_string_type_values) $1.tokens[i].type;
917 internal_error (__FILE__, __LINE__,
918 "unrecognized type in string concatenation");
922 write_exp_string_vector (pstate, type, &$1);
923 for (i = 0; i < $1.len; ++i)
924 free ($1.tokens[i].ptr);
929 exp : NSSTRING /* ObjC NextStep NSString constant
930 * of the form '@' '"' string '"'.
932 { write_exp_elt_opcode (pstate, OP_OBJC_NSSTRING);
933 write_exp_string (pstate, $1);
934 write_exp_elt_opcode (pstate, OP_OBJC_NSSTRING); }
939 { write_exp_elt_opcode (pstate, OP_LONG);
940 write_exp_elt_type (pstate,
941 parse_type (pstate)->builtin_bool);
942 write_exp_elt_longcst (pstate, (LONGEST) 1);
943 write_exp_elt_opcode (pstate, OP_LONG); }
947 { write_exp_elt_opcode (pstate, OP_LONG);
948 write_exp_elt_type (pstate,
949 parse_type (pstate)->builtin_bool);
950 write_exp_elt_longcst (pstate, (LONGEST) 0);
951 write_exp_elt_opcode (pstate, OP_LONG); }
959 $$ = SYMBOL_BLOCK_VALUE ($1.sym.symbol);
961 error (_("No file or function \"%s\"."),
962 copy_name ($1.stoken));
970 block : block COLONCOLON name
972 = lookup_symbol (copy_name ($3), $1,
973 VAR_DOMAIN, NULL).symbol;
975 if (!tem || SYMBOL_CLASS (tem) != LOC_BLOCK)
976 error (_("No function \"%s\" in specified context."),
978 $$ = SYMBOL_BLOCK_VALUE (tem); }
981 variable: name_not_typename ENTRY
982 { struct symbol *sym = $1.sym.symbol;
984 if (sym == NULL || !SYMBOL_IS_ARGUMENT (sym)
985 || !symbol_read_needs_frame (sym))
986 error (_("@entry can be used only for function "
987 "parameters, not for \"%s\""),
988 copy_name ($1.stoken));
990 write_exp_elt_opcode (pstate, OP_VAR_ENTRY_VALUE);
991 write_exp_elt_sym (pstate, sym);
992 write_exp_elt_opcode (pstate, OP_VAR_ENTRY_VALUE);
996 variable: block COLONCOLON name
997 { struct block_symbol sym
998 = lookup_symbol (copy_name ($3), $1,
1001 if (sym.symbol == 0)
1002 error (_("No symbol \"%s\" in specified context."),
1004 if (symbol_read_needs_frame (sym.symbol))
1005 pstate->block_tracker->update (sym);
1007 write_exp_elt_opcode (pstate, OP_VAR_VALUE);
1008 write_exp_elt_block (pstate, sym.block);
1009 write_exp_elt_sym (pstate, sym.symbol);
1010 write_exp_elt_opcode (pstate, OP_VAR_VALUE); }
1013 qualified_name: TYPENAME COLONCOLON name
1015 struct type *type = $1.type;
1016 type = check_typedef (type);
1017 if (!type_aggregate_p (type))
1018 error (_("`%s' is not defined as an aggregate type."),
1019 TYPE_SAFE_NAME (type));
1021 write_exp_elt_opcode (pstate, OP_SCOPE);
1022 write_exp_elt_type (pstate, type);
1023 write_exp_string (pstate, $3);
1024 write_exp_elt_opcode (pstate, OP_SCOPE);
1026 | TYPENAME COLONCOLON '~' name
1028 struct type *type = $1.type;
1029 struct stoken tmp_token;
1032 type = check_typedef (type);
1033 if (!type_aggregate_p (type))
1034 error (_("`%s' is not defined as an aggregate type."),
1035 TYPE_SAFE_NAME (type));
1036 buf = (char *) alloca ($4.length + 2);
1037 tmp_token.ptr = buf;
1038 tmp_token.length = $4.length + 1;
1040 memcpy (buf+1, $4.ptr, $4.length);
1041 buf[tmp_token.length] = 0;
1043 /* Check for valid destructor name. */
1044 destructor_name_p (tmp_token.ptr, $1.type);
1045 write_exp_elt_opcode (pstate, OP_SCOPE);
1046 write_exp_elt_type (pstate, type);
1047 write_exp_string (pstate, tmp_token);
1048 write_exp_elt_opcode (pstate, OP_SCOPE);
1050 | TYPENAME COLONCOLON name COLONCOLON name
1052 char *copy = copy_name ($3);
1053 error (_("No type \"%s\" within class "
1054 "or namespace \"%s\"."),
1055 copy, TYPE_SAFE_NAME ($1.type));
1059 variable: qualified_name
1060 | COLONCOLON name_not_typename
1062 char *name = copy_name ($2.stoken);
1064 struct bound_minimal_symbol msymbol;
1067 = lookup_symbol (name, (const struct block *) NULL,
1068 VAR_DOMAIN, NULL).symbol;
1071 write_exp_elt_opcode (pstate, OP_VAR_VALUE);
1072 write_exp_elt_block (pstate, NULL);
1073 write_exp_elt_sym (pstate, sym);
1074 write_exp_elt_opcode (pstate, OP_VAR_VALUE);
1078 msymbol = lookup_bound_minimal_symbol (name);
1079 if (msymbol.minsym != NULL)
1080 write_exp_msymbol (pstate, msymbol);
1081 else if (!have_full_symbols () && !have_partial_symbols ())
1082 error (_("No symbol table is loaded. Use the \"file\" command."));
1084 error (_("No symbol \"%s\" in current context."), name);
1088 variable: name_not_typename
1089 { struct block_symbol sym = $1.sym;
1093 if (symbol_read_needs_frame (sym.symbol))
1094 pstate->block_tracker->update (sym);
1096 /* If we found a function, see if it's
1097 an ifunc resolver that has the same
1098 address as the ifunc symbol itself.
1099 If so, prefer the ifunc symbol. */
1101 bound_minimal_symbol resolver
1102 = find_gnu_ifunc (sym.symbol);
1103 if (resolver.minsym != NULL)
1104 write_exp_msymbol (pstate, resolver);
1107 write_exp_elt_opcode (pstate, OP_VAR_VALUE);
1108 write_exp_elt_block (pstate, sym.block);
1109 write_exp_elt_sym (pstate, sym.symbol);
1110 write_exp_elt_opcode (pstate, OP_VAR_VALUE);
1113 else if ($1.is_a_field_of_this)
1115 /* C++: it hangs off of `this'. Must
1116 not inadvertently convert from a method call
1118 pstate->block_tracker->update (sym);
1119 write_exp_elt_opcode (pstate, OP_THIS);
1120 write_exp_elt_opcode (pstate, OP_THIS);
1121 write_exp_elt_opcode (pstate, STRUCTOP_PTR);
1122 write_exp_string (pstate, $1.stoken);
1123 write_exp_elt_opcode (pstate, STRUCTOP_PTR);
1127 char *arg = copy_name ($1.stoken);
1129 bound_minimal_symbol msymbol
1130 = lookup_bound_minimal_symbol (arg);
1131 if (msymbol.minsym == NULL)
1133 if (!have_full_symbols () && !have_partial_symbols ())
1134 error (_("No symbol table is loaded. Use the \"file\" command."));
1136 error (_("No symbol \"%s\" in current context."),
1137 copy_name ($1.stoken));
1140 /* This minsym might be an alias for
1141 another function. See if we can find
1142 the debug symbol for the target, and
1143 if so, use it instead, since it has
1144 return type / prototype info. This
1145 is important for example for "p
1146 *__errno_location()". */
1147 symbol *alias_target
1148 = ((msymbol.minsym->type != mst_text_gnu_ifunc
1149 && msymbol.minsym->type != mst_data_gnu_ifunc)
1150 ? find_function_alias_target (msymbol)
1152 if (alias_target != NULL)
1154 write_exp_elt_opcode (pstate, OP_VAR_VALUE);
1156 (pstate, SYMBOL_BLOCK_VALUE (alias_target));
1157 write_exp_elt_sym (pstate, alias_target);
1158 write_exp_elt_opcode (pstate, OP_VAR_VALUE);
1161 write_exp_msymbol (pstate, msymbol);
1166 space_identifier : '@' NAME
1168 cpstate->type_stack.insert (pstate, copy_name ($2.stoken));
1172 const_or_volatile: const_or_volatile_noopt
1176 cv_with_space_id : const_or_volatile space_identifier const_or_volatile
1179 const_or_volatile_or_space_identifier_noopt: cv_with_space_id
1180 | const_or_volatile_noopt
1183 const_or_volatile_or_space_identifier:
1184 const_or_volatile_or_space_identifier_noopt
1190 { cpstate->type_stack.insert (tp_pointer); }
1191 const_or_volatile_or_space_identifier
1193 { cpstate->type_stack.insert (tp_pointer); }
1194 const_or_volatile_or_space_identifier
1196 { cpstate->type_stack.insert (tp_reference); }
1198 { cpstate->type_stack.insert (tp_reference); }
1200 { cpstate->type_stack.insert (tp_rvalue_reference); }
1201 | ANDAND ptr_operator
1202 { cpstate->type_stack.insert (tp_rvalue_reference); }
1205 ptr_operator_ts: ptr_operator
1207 $$ = cpstate->type_stack.create ();
1208 cpstate->type_stacks.emplace_back ($$);
1212 abs_decl: ptr_operator_ts direct_abs_decl
1213 { $$ = $2->append ($1); }
1218 direct_abs_decl: '(' abs_decl ')'
1220 | direct_abs_decl array_mod
1222 cpstate->type_stack.push ($1);
1223 cpstate->type_stack.push ($2);
1224 cpstate->type_stack.push (tp_array);
1225 $$ = cpstate->type_stack.create ();
1226 cpstate->type_stacks.emplace_back ($$);
1230 cpstate->type_stack.push ($1);
1231 cpstate->type_stack.push (tp_array);
1232 $$ = cpstate->type_stack.create ();
1233 cpstate->type_stacks.emplace_back ($$);
1236 | direct_abs_decl func_mod
1238 cpstate->type_stack.push ($1);
1239 cpstate->type_stack.push ($2);
1240 $$ = cpstate->type_stack.create ();
1241 cpstate->type_stacks.emplace_back ($$);
1245 cpstate->type_stack.push ($1);
1246 $$ = cpstate->type_stack.create ();
1247 cpstate->type_stacks.emplace_back ($$);
1257 | OBJC_LBRAC INT ']'
1263 $$ = new std::vector<struct type *>;
1264 cpstate->type_lists.emplace_back ($$);
1266 | '(' parameter_typelist ')'
1270 /* We used to try to recognize pointer to member types here, but
1271 that didn't work (shift/reduce conflicts meant that these rules never
1272 got executed). The problem is that
1273 int (foo::bar::baz::bizzle)
1274 is a function type but
1275 int (foo::bar::baz::bizzle::*)
1276 is a pointer to member type. Stroustrup loses again! */
1281 /* Implements (approximately): (type-qualifier)* type-specifier.
1283 When type-specifier is only ever a single word, like 'float' then these
1284 arrive as pre-built TYPENAME tokens thanks to the classify_name
1285 function. However, when a type-specifier can contain multiple words,
1286 for example 'double' can appear as just 'double' or 'long double', and
1287 similarly 'long' can appear as just 'long' or in 'long double', then
1288 these type-specifiers are parsed into their own tokens in the function
1289 lex_one_token and the ident_tokens array. These separate tokens are all
1295 { $$ = lookup_signed_typename (pstate->language (),
1299 { $$ = lookup_signed_typename (pstate->language (),
1303 { $$ = lookup_signed_typename (pstate->language (),
1307 { $$ = lookup_signed_typename (pstate->language (),
1310 | LONG SIGNED_KEYWORD INT_KEYWORD
1311 { $$ = lookup_signed_typename (pstate->language (),
1314 | LONG SIGNED_KEYWORD
1315 { $$ = lookup_signed_typename (pstate->language (),
1318 | SIGNED_KEYWORD LONG INT_KEYWORD
1319 { $$ = lookup_signed_typename (pstate->language (),
1322 | UNSIGNED LONG INT_KEYWORD
1323 { $$ = lookup_unsigned_typename (pstate->language (),
1326 | LONG UNSIGNED INT_KEYWORD
1327 { $$ = lookup_unsigned_typename (pstate->language (),
1331 { $$ = lookup_unsigned_typename (pstate->language (),
1335 { $$ = lookup_signed_typename (pstate->language (),
1338 | LONG LONG INT_KEYWORD
1339 { $$ = lookup_signed_typename (pstate->language (),
1342 | LONG LONG SIGNED_KEYWORD INT_KEYWORD
1343 { $$ = lookup_signed_typename (pstate->language (),
1346 | LONG LONG SIGNED_KEYWORD
1347 { $$ = lookup_signed_typename (pstate->language (),
1350 | SIGNED_KEYWORD LONG LONG
1351 { $$ = lookup_signed_typename (pstate->language (),
1354 | SIGNED_KEYWORD LONG LONG INT_KEYWORD
1355 { $$ = lookup_signed_typename (pstate->language (),
1358 | UNSIGNED LONG LONG
1359 { $$ = lookup_unsigned_typename (pstate->language (),
1362 | UNSIGNED LONG LONG INT_KEYWORD
1363 { $$ = lookup_unsigned_typename (pstate->language (),
1366 | LONG LONG UNSIGNED
1367 { $$ = lookup_unsigned_typename (pstate->language (),
1370 | LONG LONG UNSIGNED INT_KEYWORD
1371 { $$ = lookup_unsigned_typename (pstate->language (),
1375 { $$ = lookup_signed_typename (pstate->language (),
1378 | SHORT SIGNED_KEYWORD INT_KEYWORD
1379 { $$ = lookup_signed_typename (pstate->language (),
1382 | SHORT SIGNED_KEYWORD
1383 { $$ = lookup_signed_typename (pstate->language (),
1386 | UNSIGNED SHORT INT_KEYWORD
1387 { $$ = lookup_unsigned_typename (pstate->language (),
1391 { $$ = lookup_unsigned_typename (pstate->language (),
1394 | SHORT UNSIGNED INT_KEYWORD
1395 { $$ = lookup_unsigned_typename (pstate->language (),
1399 { $$ = lookup_typename (pstate->language (),
1404 | LONG DOUBLE_KEYWORD
1405 { $$ = lookup_typename (pstate->language (),
1412 = lookup_struct (copy_name ($2),
1413 pstate->expression_context_block);
1417 pstate->mark_completion_tag (TYPE_CODE_STRUCT,
1421 | STRUCT name COMPLETE
1423 pstate->mark_completion_tag (TYPE_CODE_STRUCT,
1428 { $$ = lookup_struct
1429 (copy_name ($2), pstate->expression_context_block);
1433 pstate->mark_completion_tag (TYPE_CODE_STRUCT,
1437 | CLASS name COMPLETE
1439 pstate->mark_completion_tag (TYPE_CODE_STRUCT,
1445 = lookup_union (copy_name ($2),
1446 pstate->expression_context_block);
1450 pstate->mark_completion_tag (TYPE_CODE_UNION,
1454 | UNION name COMPLETE
1456 pstate->mark_completion_tag (TYPE_CODE_UNION,
1461 { $$ = lookup_enum (copy_name ($2),
1462 pstate->expression_context_block);
1466 pstate->mark_completion_tag (TYPE_CODE_ENUM, "", 0);
1469 | ENUM name COMPLETE
1471 pstate->mark_completion_tag (TYPE_CODE_ENUM, $2.ptr,
1475 | UNSIGNED type_name
1476 { $$ = lookup_unsigned_typename (pstate->language (),
1478 TYPE_NAME($2.type)); }
1480 { $$ = lookup_unsigned_typename (pstate->language (),
1483 | SIGNED_KEYWORD type_name
1484 { $$ = lookup_signed_typename (pstate->language (),
1486 TYPE_NAME($2.type)); }
1488 { $$ = lookup_signed_typename (pstate->language (),
1491 /* It appears that this rule for templates is never
1492 reduced; template recognition happens by lookahead
1493 in the token processing code in yylex. */
1494 | TEMPLATE name '<' type '>'
1495 { $$ = lookup_template_type
1497 pstate->expression_context_block);
1499 | const_or_volatile_or_space_identifier_noopt typebase
1500 { $$ = cpstate->type_stack.follow_types ($2); }
1501 | typebase const_or_volatile_or_space_identifier_noopt
1502 { $$ = cpstate->type_stack.follow_types ($1); }
1508 $$.stoken.ptr = "int";
1509 $$.stoken.length = 3;
1510 $$.type = lookup_signed_typename (pstate->language (),
1516 $$.stoken.ptr = "long";
1517 $$.stoken.length = 4;
1518 $$.type = lookup_signed_typename (pstate->language (),
1524 $$.stoken.ptr = "short";
1525 $$.stoken.length = 5;
1526 $$.type = lookup_signed_typename (pstate->language (),
1534 { check_parameter_typelist ($1); }
1535 | nonempty_typelist ',' DOTDOTDOT
1537 $1->push_back (NULL);
1538 check_parameter_typelist ($1);
1546 std::vector<struct type *> *typelist
1547 = new std::vector<struct type *>;
1548 cpstate->type_lists.emplace_back (typelist);
1550 typelist->push_back ($1);
1553 | nonempty_typelist ',' type
1563 cpstate->type_stack.push ($2);
1564 $$ = cpstate->type_stack.follow_types ($1);
1568 conversion_type_id: typebase conversion_declarator
1569 { $$ = cpstate->type_stack.follow_types ($1); }
1572 conversion_declarator: /* Nothing. */
1573 | ptr_operator conversion_declarator
1576 const_and_volatile: CONST_KEYWORD VOLATILE_KEYWORD
1577 | VOLATILE_KEYWORD CONST_KEYWORD
1580 const_or_volatile_noopt: const_and_volatile
1581 { cpstate->type_stack.insert (tp_const);
1582 cpstate->type_stack.insert (tp_volatile);
1585 { cpstate->type_stack.insert (tp_const); }
1587 { cpstate->type_stack.insert (tp_volatile); }
1591 { $$ = operator_stoken (" new"); }
1593 { $$ = operator_stoken (" delete"); }
1594 | OPERATOR NEW '[' ']'
1595 { $$ = operator_stoken (" new[]"); }
1596 | OPERATOR DELETE '[' ']'
1597 { $$ = operator_stoken (" delete[]"); }
1598 | OPERATOR NEW OBJC_LBRAC ']'
1599 { $$ = operator_stoken (" new[]"); }
1600 | OPERATOR DELETE OBJC_LBRAC ']'
1601 { $$ = operator_stoken (" delete[]"); }
1603 { $$ = operator_stoken ("+"); }
1605 { $$ = operator_stoken ("-"); }
1607 { $$ = operator_stoken ("*"); }
1609 { $$ = operator_stoken ("/"); }
1611 { $$ = operator_stoken ("%"); }
1613 { $$ = operator_stoken ("^"); }
1615 { $$ = operator_stoken ("&"); }
1617 { $$ = operator_stoken ("|"); }
1619 { $$ = operator_stoken ("~"); }
1621 { $$ = operator_stoken ("!"); }
1623 { $$ = operator_stoken ("="); }
1625 { $$ = operator_stoken ("<"); }
1627 { $$ = operator_stoken (">"); }
1628 | OPERATOR ASSIGN_MODIFY
1629 { const char *op = " unknown";
1653 case BINOP_BITWISE_IOR:
1656 case BINOP_BITWISE_AND:
1659 case BINOP_BITWISE_XOR:
1666 $$ = operator_stoken (op);
1669 { $$ = operator_stoken ("<<"); }
1671 { $$ = operator_stoken (">>"); }
1673 { $$ = operator_stoken ("=="); }
1675 { $$ = operator_stoken ("!="); }
1677 { $$ = operator_stoken ("<="); }
1679 { $$ = operator_stoken (">="); }
1681 { $$ = operator_stoken ("&&"); }
1683 { $$ = operator_stoken ("||"); }
1684 | OPERATOR INCREMENT
1685 { $$ = operator_stoken ("++"); }
1686 | OPERATOR DECREMENT
1687 { $$ = operator_stoken ("--"); }
1689 { $$ = operator_stoken (","); }
1690 | OPERATOR ARROW_STAR
1691 { $$ = operator_stoken ("->*"); }
1693 { $$ = operator_stoken ("->"); }
1695 { $$ = operator_stoken ("()"); }
1697 { $$ = operator_stoken ("[]"); }
1698 | OPERATOR OBJC_LBRAC ']'
1699 { $$ = operator_stoken ("[]"); }
1700 | OPERATOR conversion_type_id
1703 c_print_type ($2, NULL, &buf, -1, 0,
1704 &type_print_raw_options);
1706 /* This also needs canonicalization. */
1708 = cp_canonicalize_string (buf.c_str ());
1710 canon = std::move (buf.string ());
1711 $$ = operator_stoken ((" " + canon).c_str ());
1715 /* This rule exists in order to allow some tokens that would not normally
1716 match the 'name' rule to appear as fields within a struct. The example
1717 that initially motivated this was the RISC-V target which models the
1718 floating point registers as a union with fields called 'float' and
1719 'double'. The 'float' string becomes a TYPENAME token and can appear
1720 anywhere a 'name' can, however 'double' is its own token,
1721 DOUBLE_KEYWORD, and doesn't match the 'name' rule.*/
1724 | DOUBLE_KEYWORD { $$ = typename_stoken ("double"); }
1725 | INT_KEYWORD { $$ = typename_stoken ("int"); }
1726 | LONG { $$ = typename_stoken ("long"); }
1727 | SHORT { $$ = typename_stoken ("short"); }
1728 | SIGNED_KEYWORD { $$ = typename_stoken ("signed"); }
1729 | UNSIGNED { $$ = typename_stoken ("unsigned"); }
1732 name : NAME { $$ = $1.stoken; }
1733 | BLOCKNAME { $$ = $1.stoken; }
1734 | TYPENAME { $$ = $1.stoken; }
1735 | NAME_OR_INT { $$ = $1.stoken; }
1736 | UNKNOWN_CPP_NAME { $$ = $1.stoken; }
1740 name_not_typename : NAME
1742 /* These would be useful if name_not_typename was useful, but it is just
1743 a fake for "variable", so these cause reduce/reduce conflicts because
1744 the parser can't tell whether NAME_OR_INT is a name_not_typename (=variable,
1745 =exp) or just an exp. If name_not_typename was ever used in an lvalue
1746 context where only a name could occur, this might be useful.
1751 struct field_of_this_result is_a_field_of_this;
1755 = lookup_symbol ($1.ptr,
1756 pstate->expression_context_block,
1758 &is_a_field_of_this);
1759 $$.is_a_field_of_this
1760 = is_a_field_of_this.type != NULL;
1767 /* Like write_exp_string, but prepends a '~'. */
1770 write_destructor_name (struct parser_state *par_state, struct stoken token)
1772 char *copy = (char *) alloca (token.length + 1);
1775 memcpy (©[1], token.ptr, token.length);
1780 write_exp_string (par_state, token);
1783 /* Returns a stoken of the operator name given by OP (which does not
1784 include the string "operator"). */
1786 static struct stoken
1787 operator_stoken (const char *op)
1789 struct stoken st = { NULL, 0 };
1792 st.length = CP_OPERATOR_LEN + strlen (op);
1793 buf = (char *) malloc (st.length + 1);
1794 strcpy (buf, CP_OPERATOR_STR);
1798 /* The toplevel (c_parse) will free the memory allocated here. */
1799 cpstate->strings.emplace_back (buf);
1803 /* Returns a stoken of the type named TYPE. */
1805 static struct stoken
1806 typename_stoken (const char *type)
1808 struct stoken st = { type, 0 };
1809 st.length = strlen (type);
1813 /* Return true if the type is aggregate-like. */
1816 type_aggregate_p (struct type *type)
1818 return (TYPE_CODE (type) == TYPE_CODE_STRUCT
1819 || TYPE_CODE (type) == TYPE_CODE_UNION
1820 || TYPE_CODE (type) == TYPE_CODE_NAMESPACE
1821 || (TYPE_CODE (type) == TYPE_CODE_ENUM
1822 && TYPE_DECLARED_CLASS (type)));
1825 /* Validate a parameter typelist. */
1828 check_parameter_typelist (std::vector<struct type *> *params)
1833 for (ix = 0; ix < params->size (); ++ix)
1835 type = (*params)[ix];
1836 if (type != NULL && TYPE_CODE (check_typedef (type)) == TYPE_CODE_VOID)
1840 if (params->size () == 1)
1845 error (_("parameter types following 'void'"));
1848 error (_("'void' invalid as parameter type"));
1853 /* Take care of parsing a number (anything that starts with a digit).
1854 Set yylval and return the token type; update lexptr.
1855 LEN is the number of characters in it. */
1857 /*** Needs some error checking for the float case ***/
1860 parse_number (struct parser_state *par_state,
1861 const char *buf, int len, int parsed_float, YYSTYPE *putithere)
1869 int base = input_radix;
1872 /* Number of "L" suffixes encountered. */
1875 /* We have found a "L" or "U" suffix. */
1876 int found_suffix = 0;
1879 struct type *signed_type;
1880 struct type *unsigned_type;
1883 p = (char *) alloca (len);
1884 memcpy (p, buf, len);
1888 /* Handle suffixes for decimal floating-point: "df", "dd" or "dl". */
1889 if (len >= 2 && p[len - 2] == 'd' && p[len - 1] == 'f')
1891 putithere->typed_val_float.type
1892 = parse_type (par_state)->builtin_decfloat;
1895 else if (len >= 2 && p[len - 2] == 'd' && p[len - 1] == 'd')
1897 putithere->typed_val_float.type
1898 = parse_type (par_state)->builtin_decdouble;
1901 else if (len >= 2 && p[len - 2] == 'd' && p[len - 1] == 'l')
1903 putithere->typed_val_float.type
1904 = parse_type (par_state)->builtin_declong;
1907 /* Handle suffixes: 'f' for float, 'l' for long double. */
1908 else if (len >= 1 && TOLOWER (p[len - 1]) == 'f')
1910 putithere->typed_val_float.type
1911 = parse_type (par_state)->builtin_float;
1914 else if (len >= 1 && TOLOWER (p[len - 1]) == 'l')
1916 putithere->typed_val_float.type
1917 = parse_type (par_state)->builtin_long_double;
1920 /* Default type for floating-point literals is double. */
1923 putithere->typed_val_float.type
1924 = parse_type (par_state)->builtin_double;
1927 if (!parse_float (p, len,
1928 putithere->typed_val_float.type,
1929 putithere->typed_val_float.val))
1934 /* Handle base-switching prefixes 0x, 0t, 0d, 0 */
1935 if (p[0] == '0' && len > 1)
1978 if (c >= 'A' && c <= 'Z')
1980 if (c != 'l' && c != 'u')
1982 if (c >= '0' && c <= '9')
1990 if (base > 10 && c >= 'a' && c <= 'f')
1994 n += i = c - 'a' + 10;
2007 return ERROR; /* Char not a digit */
2010 return ERROR; /* Invalid digit in this base */
2012 /* Portably test for overflow (only works for nonzero values, so make
2013 a second check for zero). FIXME: Can't we just make n and prevn
2014 unsigned and avoid this? */
2015 if (c != 'l' && c != 'u' && (prevn >= n) && n != 0)
2016 unsigned_p = 1; /* Try something unsigned */
2018 /* Portably test for unsigned overflow.
2019 FIXME: This check is wrong; for example it doesn't find overflow
2020 on 0x123456789 when LONGEST is 32 bits. */
2021 if (c != 'l' && c != 'u' && n != 0)
2023 if (unsigned_p && prevn >= n)
2024 error (_("Numeric constant too large."));
2029 /* An integer constant is an int, a long, or a long long. An L
2030 suffix forces it to be long; an LL suffix forces it to be long
2031 long. If not forced to a larger size, it gets the first type of
2032 the above that it fits in. To figure out whether it fits, we
2033 shift it right and see whether anything remains. Note that we
2034 can't shift sizeof (LONGEST) * HOST_CHAR_BIT bits or more in one
2035 operation, because many compilers will warn about such a shift
2036 (which always produces a zero result). Sometimes gdbarch_int_bit
2037 or gdbarch_long_bit will be that big, sometimes not. To deal with
2038 the case where it is we just always shift the value more than
2039 once, with fewer bits each time. */
2043 && (un >> (gdbarch_int_bit (par_state->gdbarch ()) - 2)) == 0)
2046 = ((ULONGEST)1) << (gdbarch_int_bit (par_state->gdbarch ()) - 1);
2048 /* A large decimal (not hex or octal) constant (between INT_MAX
2049 and UINT_MAX) is a long or unsigned long, according to ANSI,
2050 never an unsigned int, but this code treats it as unsigned
2051 int. This probably should be fixed. GCC gives a warning on
2054 unsigned_type = parse_type (par_state)->builtin_unsigned_int;
2055 signed_type = parse_type (par_state)->builtin_int;
2057 else if (long_p <= 1
2058 && (un >> (gdbarch_long_bit (par_state->gdbarch ()) - 2)) == 0)
2061 = ((ULONGEST)1) << (gdbarch_long_bit (par_state->gdbarch ()) - 1);
2062 unsigned_type = parse_type (par_state)->builtin_unsigned_long;
2063 signed_type = parse_type (par_state)->builtin_long;
2068 if (sizeof (ULONGEST) * HOST_CHAR_BIT
2069 < gdbarch_long_long_bit (par_state->gdbarch ()))
2070 /* A long long does not fit in a LONGEST. */
2071 shift = (sizeof (ULONGEST) * HOST_CHAR_BIT - 1);
2073 shift = (gdbarch_long_long_bit (par_state->gdbarch ()) - 1);
2074 high_bit = (ULONGEST) 1 << shift;
2075 unsigned_type = parse_type (par_state)->builtin_unsigned_long_long;
2076 signed_type = parse_type (par_state)->builtin_long_long;
2079 putithere->typed_val_int.val = n;
2081 /* If the high bit of the worked out type is set then this number
2082 has to be unsigned. */
2084 if (unsigned_p || (n & high_bit))
2086 putithere->typed_val_int.type = unsigned_type;
2090 putithere->typed_val_int.type = signed_type;
2096 /* Temporary obstack used for holding strings. */
2097 static struct obstack tempbuf;
2098 static int tempbuf_init;
2100 /* Parse a C escape sequence. The initial backslash of the sequence
2101 is at (*PTR)[-1]. *PTR will be updated to point to just after the
2102 last character of the sequence. If OUTPUT is not NULL, the
2103 translated form of the escape sequence will be written there. If
2104 OUTPUT is NULL, no output is written and the call will only affect
2105 *PTR. If an escape sequence is expressed in target bytes, then the
2106 entire sequence will simply be copied to OUTPUT. Return 1 if any
2107 character was emitted, 0 otherwise. */
2110 c_parse_escape (const char **ptr, struct obstack *output)
2112 const char *tokptr = *ptr;
2115 /* Some escape sequences undergo character set conversion. Those we
2119 /* Hex escapes do not undergo character set conversion, so keep
2120 the escape sequence for later. */
2123 obstack_grow_str (output, "\\x");
2125 if (!ISXDIGIT (*tokptr))
2126 error (_("\\x escape without a following hex digit"));
2127 while (ISXDIGIT (*tokptr))
2130 obstack_1grow (output, *tokptr);
2135 /* Octal escapes do not undergo character set conversion, so
2136 keep the escape sequence for later. */
2148 obstack_grow_str (output, "\\");
2150 i < 3 && ISDIGIT (*tokptr) && *tokptr != '8' && *tokptr != '9';
2154 obstack_1grow (output, *tokptr);
2160 /* We handle UCNs later. We could handle them here, but that
2161 would mean a spurious error in the case where the UCN could
2162 be converted to the target charset but not the host
2168 int i, len = c == 'U' ? 8 : 4;
2171 obstack_1grow (output, '\\');
2172 obstack_1grow (output, *tokptr);
2175 if (!ISXDIGIT (*tokptr))
2176 error (_("\\%c escape without a following hex digit"), c);
2177 for (i = 0; i < len && ISXDIGIT (*tokptr); ++i)
2180 obstack_1grow (output, *tokptr);
2186 /* We must pass backslash through so that it does not
2187 cause quoting during the second expansion. */
2190 obstack_grow_str (output, "\\\\");
2194 /* Escapes which undergo conversion. */
2197 obstack_1grow (output, '\a');
2202 obstack_1grow (output, '\b');
2207 obstack_1grow (output, '\f');
2212 obstack_1grow (output, '\n');
2217 obstack_1grow (output, '\r');
2222 obstack_1grow (output, '\t');
2227 obstack_1grow (output, '\v');
2231 /* GCC extension. */
2234 obstack_1grow (output, HOST_ESCAPE_CHAR);
2238 /* Backslash-newline expands to nothing at all. */
2244 /* A few escapes just expand to the character itself. */
2248 /* GCC extensions. */
2253 /* Unrecognized escapes turn into the character itself. */
2256 obstack_1grow (output, *tokptr);
2264 /* Parse a string or character literal from TOKPTR. The string or
2265 character may be wide or unicode. *OUTPTR is set to just after the
2266 end of the literal in the input string. The resulting token is
2267 stored in VALUE. This returns a token value, either STRING or
2268 CHAR, depending on what was parsed. *HOST_CHARS is set to the
2269 number of host characters in the literal. */
2272 parse_string_or_char (const char *tokptr, const char **outptr,
2273 struct typed_stoken *value, int *host_chars)
2279 /* Build the gdb internal form of the input string in tempbuf. Note
2280 that the buffer is null byte terminated *only* for the
2281 convenience of debugging gdb itself and printing the buffer
2282 contents when the buffer contains no embedded nulls. Gdb does
2283 not depend upon the buffer being null byte terminated, it uses
2284 the length string instead. This allows gdb to handle C strings
2285 (as well as strings in other languages) with embedded null
2291 obstack_free (&tempbuf, NULL);
2292 obstack_init (&tempbuf);
2294 /* Record the string type. */
2297 type = C_WIDE_STRING;
2300 else if (*tokptr == 'u')
2305 else if (*tokptr == 'U')
2310 else if (*tokptr == '@')
2312 /* An Objective C string. */
2320 /* Skip the quote. */
2334 *host_chars += c_parse_escape (&tokptr, &tempbuf);
2336 else if (c == quote)
2340 obstack_1grow (&tempbuf, c);
2342 /* FIXME: this does the wrong thing with multi-byte host
2343 characters. We could use mbrlen here, but that would
2344 make "set host-charset" a bit less useful. */
2349 if (*tokptr != quote)
2352 error (_("Unterminated string in expression."));
2354 error (_("Unmatched single quote."));
2359 value->ptr = (char *) obstack_base (&tempbuf);
2360 value->length = obstack_object_size (&tempbuf);
2364 return quote == '"' ? (is_objc ? NSSTRING : STRING) : CHAR;
2367 /* This is used to associate some attributes with a token. */
2371 /* If this bit is set, the token is C++-only. */
2375 /* If this bit is set, the token is conditional: if there is a
2376 symbol of the same name, then the token is a symbol; otherwise,
2377 the token is a keyword. */
2381 DEF_ENUM_FLAGS_TYPE (enum token_flag, token_flags);
2387 enum exp_opcode opcode;
2391 static const struct token tokentab3[] =
2393 {">>=", ASSIGN_MODIFY, BINOP_RSH, 0},
2394 {"<<=", ASSIGN_MODIFY, BINOP_LSH, 0},
2395 {"->*", ARROW_STAR, BINOP_END, FLAG_CXX},
2396 {"...", DOTDOTDOT, BINOP_END, 0}
2399 static const struct token tokentab2[] =
2401 {"+=", ASSIGN_MODIFY, BINOP_ADD, 0},
2402 {"-=", ASSIGN_MODIFY, BINOP_SUB, 0},
2403 {"*=", ASSIGN_MODIFY, BINOP_MUL, 0},
2404 {"/=", ASSIGN_MODIFY, BINOP_DIV, 0},
2405 {"%=", ASSIGN_MODIFY, BINOP_REM, 0},
2406 {"|=", ASSIGN_MODIFY, BINOP_BITWISE_IOR, 0},
2407 {"&=", ASSIGN_MODIFY, BINOP_BITWISE_AND, 0},
2408 {"^=", ASSIGN_MODIFY, BINOP_BITWISE_XOR, 0},
2409 {"++", INCREMENT, BINOP_END, 0},
2410 {"--", DECREMENT, BINOP_END, 0},
2411 {"->", ARROW, BINOP_END, 0},
2412 {"&&", ANDAND, BINOP_END, 0},
2413 {"||", OROR, BINOP_END, 0},
2414 /* "::" is *not* only C++: gdb overrides its meaning in several
2415 different ways, e.g., 'filename'::func, function::variable. */
2416 {"::", COLONCOLON, BINOP_END, 0},
2417 {"<<", LSH, BINOP_END, 0},
2418 {">>", RSH, BINOP_END, 0},
2419 {"==", EQUAL, BINOP_END, 0},
2420 {"!=", NOTEQUAL, BINOP_END, 0},
2421 {"<=", LEQ, BINOP_END, 0},
2422 {">=", GEQ, BINOP_END, 0},
2423 {".*", DOT_STAR, BINOP_END, FLAG_CXX}
2426 /* Identifier-like tokens. Only type-specifiers than can appear in
2427 multi-word type names (for example 'double' can appear in 'long
2428 double') need to be listed here. type-specifiers that are only ever
2429 single word (like 'float') are handled by the classify_name function. */
2430 static const struct token ident_tokens[] =
2432 {"unsigned", UNSIGNED, OP_NULL, 0},
2433 {"template", TEMPLATE, OP_NULL, FLAG_CXX},
2434 {"volatile", VOLATILE_KEYWORD, OP_NULL, 0},
2435 {"struct", STRUCT, OP_NULL, 0},
2436 {"signed", SIGNED_KEYWORD, OP_NULL, 0},
2437 {"sizeof", SIZEOF, OP_NULL, 0},
2438 {"_Alignof", ALIGNOF, OP_NULL, 0},
2439 {"alignof", ALIGNOF, OP_NULL, FLAG_CXX},
2440 {"double", DOUBLE_KEYWORD, OP_NULL, 0},
2441 {"false", FALSEKEYWORD, OP_NULL, FLAG_CXX},
2442 {"class", CLASS, OP_NULL, FLAG_CXX},
2443 {"union", UNION, OP_NULL, 0},
2444 {"short", SHORT, OP_NULL, 0},
2445 {"const", CONST_KEYWORD, OP_NULL, 0},
2446 {"enum", ENUM, OP_NULL, 0},
2447 {"long", LONG, OP_NULL, 0},
2448 {"true", TRUEKEYWORD, OP_NULL, FLAG_CXX},
2449 {"int", INT_KEYWORD, OP_NULL, 0},
2450 {"new", NEW, OP_NULL, FLAG_CXX},
2451 {"delete", DELETE, OP_NULL, FLAG_CXX},
2452 {"operator", OPERATOR, OP_NULL, FLAG_CXX},
2454 {"and", ANDAND, BINOP_END, FLAG_CXX},
2455 {"and_eq", ASSIGN_MODIFY, BINOP_BITWISE_AND, FLAG_CXX},
2456 {"bitand", '&', OP_NULL, FLAG_CXX},
2457 {"bitor", '|', OP_NULL, FLAG_CXX},
2458 {"compl", '~', OP_NULL, FLAG_CXX},
2459 {"not", '!', OP_NULL, FLAG_CXX},
2460 {"not_eq", NOTEQUAL, BINOP_END, FLAG_CXX},
2461 {"or", OROR, BINOP_END, FLAG_CXX},
2462 {"or_eq", ASSIGN_MODIFY, BINOP_BITWISE_IOR, FLAG_CXX},
2463 {"xor", '^', OP_NULL, FLAG_CXX},
2464 {"xor_eq", ASSIGN_MODIFY, BINOP_BITWISE_XOR, FLAG_CXX},
2466 {"const_cast", CONST_CAST, OP_NULL, FLAG_CXX },
2467 {"dynamic_cast", DYNAMIC_CAST, OP_NULL, FLAG_CXX },
2468 {"static_cast", STATIC_CAST, OP_NULL, FLAG_CXX },
2469 {"reinterpret_cast", REINTERPRET_CAST, OP_NULL, FLAG_CXX },
2471 {"__typeof__", TYPEOF, OP_TYPEOF, 0 },
2472 {"__typeof", TYPEOF, OP_TYPEOF, 0 },
2473 {"typeof", TYPEOF, OP_TYPEOF, FLAG_SHADOW },
2474 {"__decltype", DECLTYPE, OP_DECLTYPE, FLAG_CXX },
2475 {"decltype", DECLTYPE, OP_DECLTYPE, FLAG_CXX | FLAG_SHADOW },
2477 {"typeid", TYPEID, OP_TYPEID, FLAG_CXX}
2482 scan_macro_expansion (char *expansion)
2486 /* We'd better not be trying to push the stack twice. */
2487 gdb_assert (! cpstate->macro_original_text);
2489 /* Copy to the obstack, and then free the intermediate
2491 copy = (char *) obstack_copy0 (&cpstate->expansion_obstack, expansion,
2492 strlen (expansion));
2495 /* Save the old lexptr value, so we can return to it when we're done
2496 parsing the expanded text. */
2497 cpstate->macro_original_text = pstate->lexptr;
2498 pstate->lexptr = copy;
2502 scanning_macro_expansion (void)
2504 return cpstate->macro_original_text != 0;
2508 finished_macro_expansion (void)
2510 /* There'd better be something to pop back to. */
2511 gdb_assert (cpstate->macro_original_text);
2513 /* Pop back to the original text. */
2514 pstate->lexptr = cpstate->macro_original_text;
2515 cpstate->macro_original_text = 0;
2518 /* Return true iff the token represents a C++ cast operator. */
2521 is_cast_operator (const char *token, int len)
2523 return (! strncmp (token, "dynamic_cast", len)
2524 || ! strncmp (token, "static_cast", len)
2525 || ! strncmp (token, "reinterpret_cast", len)
2526 || ! strncmp (token, "const_cast", len));
2529 /* The scope used for macro expansion. */
2530 static struct macro_scope *expression_macro_scope;
2532 /* This is set if a NAME token appeared at the very end of the input
2533 string, with no whitespace separating the name from the EOF. This
2534 is used only when parsing to do field name completion. */
2535 static int saw_name_at_eof;
2537 /* This is set if the previously-returned token was a structure
2538 operator -- either '.' or ARROW. */
2539 static bool last_was_structop;
2541 /* Depth of parentheses. */
2542 static int paren_depth;
2544 /* Read one token, getting characters through lexptr. */
2547 lex_one_token (struct parser_state *par_state, bool *is_quoted_name)
2552 const char *tokstart;
2553 bool saw_structop = last_was_structop;
2556 last_was_structop = false;
2557 *is_quoted_name = false;
2561 /* Check if this is a macro invocation that we need to expand. */
2562 if (! scanning_macro_expansion ())
2564 char *expanded = macro_expand_next (&pstate->lexptr,
2565 standard_macro_lookup,
2566 expression_macro_scope);
2569 scan_macro_expansion (expanded);
2572 pstate->prev_lexptr = pstate->lexptr;
2574 tokstart = pstate->lexptr;
2575 /* See if it is a special token of length 3. */
2576 for (i = 0; i < sizeof tokentab3 / sizeof tokentab3[0]; i++)
2577 if (strncmp (tokstart, tokentab3[i].oper, 3) == 0)
2579 if ((tokentab3[i].flags & FLAG_CXX) != 0
2580 && par_state->language ()->la_language != language_cplus)
2583 pstate->lexptr += 3;
2584 yylval.opcode = tokentab3[i].opcode;
2585 return tokentab3[i].token;
2588 /* See if it is a special token of length 2. */
2589 for (i = 0; i < sizeof tokentab2 / sizeof tokentab2[0]; i++)
2590 if (strncmp (tokstart, tokentab2[i].oper, 2) == 0)
2592 if ((tokentab2[i].flags & FLAG_CXX) != 0
2593 && par_state->language ()->la_language != language_cplus)
2596 pstate->lexptr += 2;
2597 yylval.opcode = tokentab2[i].opcode;
2598 if (tokentab2[i].token == ARROW)
2599 last_was_structop = 1;
2600 return tokentab2[i].token;
2603 switch (c = *tokstart)
2606 /* If we were just scanning the result of a macro expansion,
2607 then we need to resume scanning the original text.
2608 If we're parsing for field name completion, and the previous
2609 token allows such completion, return a COMPLETE token.
2610 Otherwise, we were already scanning the original text, and
2611 we're really done. */
2612 if (scanning_macro_expansion ())
2614 finished_macro_expansion ();
2617 else if (saw_name_at_eof)
2619 saw_name_at_eof = 0;
2622 else if (par_state->parse_completion && saw_structop)
2637 if (par_state->language ()->la_language == language_objc
2644 if (paren_depth == 0)
2651 if (pstate->comma_terminates
2653 && ! scanning_macro_expansion ())
2659 /* Might be a floating point number. */
2660 if (pstate->lexptr[1] < '0' || pstate->lexptr[1] > '9')
2662 last_was_structop = true;
2663 goto symbol; /* Nope, must be a symbol. */
2678 /* It's a number. */
2679 int got_dot = 0, got_e = 0, toktype;
2680 const char *p = tokstart;
2681 int hex = input_radix > 10;
2683 if (c == '0' && (p[1] == 'x' || p[1] == 'X'))
2688 else if (c == '0' && (p[1]=='t' || p[1]=='T' || p[1]=='d' || p[1]=='D'))
2696 /* This test includes !hex because 'e' is a valid hex digit
2697 and thus does not indicate a floating point number when
2698 the radix is hex. */
2699 if (!hex && !got_e && (*p == 'e' || *p == 'E'))
2700 got_dot = got_e = 1;
2701 /* This test does not include !hex, because a '.' always indicates
2702 a decimal floating point number regardless of the radix. */
2703 else if (!got_dot && *p == '.')
2705 else if (got_e && (p[-1] == 'e' || p[-1] == 'E')
2706 && (*p == '-' || *p == '+'))
2707 /* This is the sign of the exponent, not the end of the
2710 /* We will take any letters or digits. parse_number will
2711 complain if past the radix, or if L or U are not final. */
2712 else if ((*p < '0' || *p > '9')
2713 && ((*p < 'a' || *p > 'z')
2714 && (*p < 'A' || *p > 'Z')))
2717 toktype = parse_number (par_state, tokstart, p - tokstart,
2718 got_dot|got_e, &yylval);
2719 if (toktype == ERROR)
2721 char *err_copy = (char *) alloca (p - tokstart + 1);
2723 memcpy (err_copy, tokstart, p - tokstart);
2724 err_copy[p - tokstart] = 0;
2725 error (_("Invalid number \"%s\"."), err_copy);
2733 const char *p = &tokstart[1];
2735 if (par_state->language ()->la_language == language_objc)
2737 size_t len = strlen ("selector");
2739 if (strncmp (p, "selector", len) == 0
2740 && (p[len] == '\0' || ISSPACE (p[len])))
2742 pstate->lexptr = p + len;
2749 while (ISSPACE (*p))
2751 size_t len = strlen ("entry");
2752 if (strncmp (p, "entry", len) == 0 && !c_ident_is_alnum (p[len])
2755 pstate->lexptr = &p[len];
2784 if (tokstart[1] != '"' && tokstart[1] != '\'')
2793 int result = parse_string_or_char (tokstart, &pstate->lexptr,
2794 &yylval.tsval, &host_len);
2798 error (_("Empty character constant."));
2799 else if (host_len > 2 && c == '\'')
2802 namelen = pstate->lexptr - tokstart - 1;
2803 *is_quoted_name = true;
2807 else if (host_len > 1)
2808 error (_("Invalid character constant."));
2814 if (!(c == '_' || c == '$' || c_ident_is_alpha (c)))
2815 /* We must have come across a bad character (e.g. ';'). */
2816 error (_("Invalid character '%c' in expression."), c);
2818 /* It's a name. See how long it is. */
2820 for (c = tokstart[namelen];
2821 (c == '_' || c == '$' || c_ident_is_alnum (c) || c == '<');)
2823 /* Template parameter lists are part of the name.
2824 FIXME: This mishandles `print $a<4&&$a>3'. */
2828 if (! is_cast_operator (tokstart, namelen))
2830 /* Scan ahead to get rest of the template specification. Note
2831 that we look ahead only when the '<' adjoins non-whitespace
2832 characters; for comparison expressions, e.g. "a < b > c",
2833 there must be spaces before the '<', etc. */
2834 const char *p = find_template_name_end (tokstart + namelen);
2837 namelen = p - tokstart;
2841 c = tokstart[++namelen];
2844 /* The token "if" terminates the expression and is NOT removed from
2845 the input stream. It doesn't count if it appears in the
2846 expansion of a macro. */
2848 && tokstart[0] == 'i'
2849 && tokstart[1] == 'f'
2850 && ! scanning_macro_expansion ())
2855 /* For the same reason (breakpoint conditions), "thread N"
2856 terminates the expression. "thread" could be an identifier, but
2857 an identifier is never followed by a number without intervening
2858 punctuation. "task" is similar. Handle abbreviations of these,
2859 similarly to breakpoint.c:find_condition_and_thread. */
2861 && (strncmp (tokstart, "thread", namelen) == 0
2862 || strncmp (tokstart, "task", namelen) == 0)
2863 && (tokstart[namelen] == ' ' || tokstart[namelen] == '\t')
2864 && ! scanning_macro_expansion ())
2866 const char *p = tokstart + namelen + 1;
2868 while (*p == ' ' || *p == '\t')
2870 if (*p >= '0' && *p <= '9')
2874 pstate->lexptr += namelen;
2878 yylval.sval.ptr = tokstart;
2879 yylval.sval.length = namelen;
2881 /* Catch specific keywords. */
2882 copy = copy_name (yylval.sval);
2883 for (i = 0; i < sizeof ident_tokens / sizeof ident_tokens[0]; i++)
2884 if (strcmp (copy, ident_tokens[i].oper) == 0)
2886 if ((ident_tokens[i].flags & FLAG_CXX) != 0
2887 && par_state->language ()->la_language != language_cplus)
2890 if ((ident_tokens[i].flags & FLAG_SHADOW) != 0)
2892 struct field_of_this_result is_a_field_of_this;
2894 if (lookup_symbol (copy,
2895 pstate->expression_context_block,
2897 (par_state->language ()->la_language
2898 == language_cplus ? &is_a_field_of_this
2902 /* The keyword is shadowed. */
2907 /* It is ok to always set this, even though we don't always
2908 strictly need to. */
2909 yylval.opcode = ident_tokens[i].opcode;
2910 return ident_tokens[i].token;
2913 if (*tokstart == '$')
2914 return DOLLAR_VARIABLE;
2916 if (pstate->parse_completion && *pstate->lexptr == '\0')
2917 saw_name_at_eof = 1;
2919 yylval.ssym.stoken = yylval.sval;
2920 yylval.ssym.sym.symbol = NULL;
2921 yylval.ssym.sym.block = NULL;
2922 yylval.ssym.is_a_field_of_this = 0;
2926 /* An object of this type is pushed on a FIFO by the "outer" lexer. */
2927 struct token_and_value
2933 /* A FIFO of tokens that have been read but not yet returned to the
2935 static std::vector<token_and_value> token_fifo;
2937 /* Non-zero if the lexer should return tokens from the FIFO. */
2940 /* Temporary storage for c_lex; this holds symbol names as they are
2942 auto_obstack name_obstack;
2944 /* Classify a NAME token. The contents of the token are in `yylval'.
2945 Updates yylval and returns the new token type. BLOCK is the block
2946 in which lookups start; this can be NULL to mean the global scope.
2947 IS_QUOTED_NAME is non-zero if the name token was originally quoted
2948 in single quotes. IS_AFTER_STRUCTOP is true if this name follows
2949 a structure operator -- either '.' or ARROW */
2952 classify_name (struct parser_state *par_state, const struct block *block,
2953 bool is_quoted_name, bool is_after_structop)
2955 struct block_symbol bsym;
2957 struct field_of_this_result is_a_field_of_this;
2959 copy = copy_name (yylval.sval);
2961 /* Initialize this in case we *don't* use it in this call; that way
2962 we can refer to it unconditionally below. */
2963 memset (&is_a_field_of_this, 0, sizeof (is_a_field_of_this));
2965 bsym = lookup_symbol (copy, block, VAR_DOMAIN,
2966 par_state->language ()->la_name_of_this
2967 ? &is_a_field_of_this : NULL);
2969 if (bsym.symbol && SYMBOL_CLASS (bsym.symbol) == LOC_BLOCK)
2971 yylval.ssym.sym = bsym;
2972 yylval.ssym.is_a_field_of_this = is_a_field_of_this.type != NULL;
2975 else if (!bsym.symbol)
2977 /* If we found a field of 'this', we might have erroneously
2978 found a constructor where we wanted a type name. Handle this
2979 case by noticing that we found a constructor and then look up
2980 the type tag instead. */
2981 if (is_a_field_of_this.type != NULL
2982 && is_a_field_of_this.fn_field != NULL
2983 && TYPE_FN_FIELD_CONSTRUCTOR (is_a_field_of_this.fn_field->fn_fields,
2986 struct field_of_this_result inner_is_a_field_of_this;
2988 bsym = lookup_symbol (copy, block, STRUCT_DOMAIN,
2989 &inner_is_a_field_of_this);
2990 if (bsym.symbol != NULL)
2992 yylval.tsym.type = SYMBOL_TYPE (bsym.symbol);
2997 /* If we found a field on the "this" object, or we are looking
2998 up a field on a struct, then we want to prefer it over a
2999 filename. However, if the name was quoted, then it is better
3000 to check for a filename or a block, since this is the only
3001 way the user has of requiring the extension to be used. */
3002 if ((is_a_field_of_this.type == NULL && !is_after_structop)
3005 /* See if it's a file name. */
3006 struct symtab *symtab;
3008 symtab = lookup_symtab (copy);
3011 yylval.bval = BLOCKVECTOR_BLOCK (SYMTAB_BLOCKVECTOR (symtab),
3018 if (bsym.symbol && SYMBOL_CLASS (bsym.symbol) == LOC_TYPEDEF)
3020 yylval.tsym.type = SYMBOL_TYPE (bsym.symbol);
3024 /* See if it's an ObjC classname. */
3025 if (par_state->language ()->la_language == language_objc && !bsym.symbol)
3027 CORE_ADDR Class = lookup_objc_class (par_state->gdbarch (), copy);
3032 yylval.theclass.theclass = Class;
3033 sym = lookup_struct_typedef (copy,
3034 par_state->expression_context_block, 1);
3036 yylval.theclass.type = SYMBOL_TYPE (sym);
3041 /* Input names that aren't symbols but ARE valid hex numbers, when
3042 the input radix permits them, can be names or numbers depending
3043 on the parse. Note we support radixes > 16 here. */
3045 && ((copy[0] >= 'a' && copy[0] < 'a' + input_radix - 10)
3046 || (copy[0] >= 'A' && copy[0] < 'A' + input_radix - 10)))
3048 YYSTYPE newlval; /* Its value is ignored. */
3049 int hextype = parse_number (par_state, copy, yylval.sval.length,
3054 yylval.ssym.sym = bsym;
3055 yylval.ssym.is_a_field_of_this = is_a_field_of_this.type != NULL;
3060 /* Any other kind of symbol */
3061 yylval.ssym.sym = bsym;
3062 yylval.ssym.is_a_field_of_this = is_a_field_of_this.type != NULL;
3064 if (bsym.symbol == NULL
3065 && par_state->language ()->la_language == language_cplus
3066 && is_a_field_of_this.type == NULL
3067 && lookup_minimal_symbol (copy, NULL, NULL).minsym == NULL)
3068 return UNKNOWN_CPP_NAME;
3073 /* Like classify_name, but used by the inner loop of the lexer, when a
3074 name might have already been seen. CONTEXT is the context type, or
3075 NULL if this is the first component of a name. */
3078 classify_inner_name (struct parser_state *par_state,
3079 const struct block *block, struct type *context)
3084 if (context == NULL)
3085 return classify_name (par_state, block, false, false);
3087 type = check_typedef (context);
3088 if (!type_aggregate_p (type))
3091 copy = copy_name (yylval.ssym.stoken);
3092 /* N.B. We assume the symbol can only be in VAR_DOMAIN. */
3093 yylval.ssym.sym = cp_lookup_nested_symbol (type, copy, block, VAR_DOMAIN);
3095 /* If no symbol was found, search for a matching base class named
3096 COPY. This will allow users to enter qualified names of class members
3097 relative to the `this' pointer. */
3098 if (yylval.ssym.sym.symbol == NULL)
3100 struct type *base_type = cp_find_type_baseclass_by_name (type, copy);
3102 if (base_type != NULL)
3104 yylval.tsym.type = base_type;
3111 switch (SYMBOL_CLASS (yylval.ssym.sym.symbol))
3115 /* cp_lookup_nested_symbol might have accidentally found a constructor
3116 named COPY when we really wanted a base class of the same name.
3117 Double-check this case by looking for a base class. */
3119 struct type *base_type = cp_find_type_baseclass_by_name (type, copy);
3121 if (base_type != NULL)
3123 yylval.tsym.type = base_type;
3130 yylval.tsym.type = SYMBOL_TYPE (yylval.ssym.sym.symbol);
3136 internal_error (__FILE__, __LINE__, _("not reached"));
3139 /* The outer level of a two-level lexer. This calls the inner lexer
3140 to return tokens. It then either returns these tokens, or
3141 aggregates them into a larger token. This lets us work around a
3142 problem in our parsing approach, where the parser could not
3143 distinguish between qualified names and qualified types at the
3146 This approach is still not ideal, because it mishandles template
3147 types. See the comment in lex_one_token for an example. However,
3148 this is still an improvement over the earlier approach, and will
3149 suffice until we move to better parsing technology. */
3154 token_and_value current;
3155 int first_was_coloncolon, last_was_coloncolon;
3156 struct type *context_type = NULL;
3157 int last_to_examine, next_to_examine, checkpoint;
3158 const struct block *search_block;
3159 bool is_quoted_name, last_lex_was_structop;
3161 if (popping && !token_fifo.empty ())
3165 last_lex_was_structop = last_was_structop;
3167 /* Read the first token and decide what to do. Most of the
3168 subsequent code is C++-only; but also depends on seeing a "::" or
3170 current.token = lex_one_token (pstate, &is_quoted_name);
3171 if (current.token == NAME)
3172 current.token = classify_name (pstate, pstate->expression_context_block,
3173 is_quoted_name, last_lex_was_structop);
3174 if (pstate->language ()->la_language != language_cplus
3175 || (current.token != TYPENAME && current.token != COLONCOLON
3176 && current.token != FILENAME))
3177 return current.token;
3179 /* Read any sequence of alternating "::" and name-like tokens into
3181 current.value = yylval;
3182 token_fifo.push_back (current);
3183 last_was_coloncolon = current.token == COLONCOLON;
3188 /* We ignore quoted names other than the very first one.
3189 Subsequent ones do not have any special meaning. */
3190 current.token = lex_one_token (pstate, &ignore);
3191 current.value = yylval;
3192 token_fifo.push_back (current);
3194 if ((last_was_coloncolon && current.token != NAME)
3195 || (!last_was_coloncolon && current.token != COLONCOLON))
3197 last_was_coloncolon = !last_was_coloncolon;
3201 /* We always read one extra token, so compute the number of tokens
3202 to examine accordingly. */
3203 last_to_examine = token_fifo.size () - 2;
3204 next_to_examine = 0;
3206 current = token_fifo[next_to_examine];
3209 name_obstack.clear ();
3211 if (current.token == FILENAME)
3212 search_block = current.value.bval;
3213 else if (current.token == COLONCOLON)
3214 search_block = NULL;
3217 gdb_assert (current.token == TYPENAME);
3218 search_block = pstate->expression_context_block;
3219 obstack_grow (&name_obstack, current.value.sval.ptr,
3220 current.value.sval.length);
3221 context_type = current.value.tsym.type;
3225 first_was_coloncolon = current.token == COLONCOLON;
3226 last_was_coloncolon = first_was_coloncolon;
3228 while (next_to_examine <= last_to_examine)
3230 token_and_value next;
3232 next = token_fifo[next_to_examine];
3235 if (next.token == NAME && last_was_coloncolon)
3239 yylval = next.value;
3240 classification = classify_inner_name (pstate, search_block,
3242 /* We keep going until we either run out of names, or until
3243 we have a qualified name which is not a type. */
3244 if (classification != TYPENAME && classification != NAME)
3247 /* Accept up to this token. */
3248 checkpoint = next_to_examine;
3250 /* Update the partial name we are constructing. */
3251 if (context_type != NULL)
3253 /* We don't want to put a leading "::" into the name. */
3254 obstack_grow_str (&name_obstack, "::");
3256 obstack_grow (&name_obstack, next.value.sval.ptr,
3257 next.value.sval.length);
3259 yylval.sval.ptr = (const char *) obstack_base (&name_obstack);
3260 yylval.sval.length = obstack_object_size (&name_obstack);
3261 current.value = yylval;
3262 current.token = classification;
3264 last_was_coloncolon = 0;
3266 if (classification == NAME)
3269 context_type = yylval.tsym.type;
3271 else if (next.token == COLONCOLON && !last_was_coloncolon)
3272 last_was_coloncolon = 1;
3275 /* We've reached the end of the name. */
3280 /* If we have a replacement token, install it as the first token in
3281 the FIFO, and delete the other constituent tokens. */
3284 current.value.sval.ptr
3285 = (const char *) obstack_copy0 (&cpstate->expansion_obstack,
3286 current.value.sval.ptr,
3287 current.value.sval.length);
3289 token_fifo[0] = current;
3291 token_fifo.erase (token_fifo.begin () + 1,
3292 token_fifo.begin () + checkpoint);
3296 current = token_fifo[0];
3297 token_fifo.erase (token_fifo.begin ());
3298 yylval = current.value;
3299 return current.token;
3303 c_parse (struct parser_state *par_state)
3305 /* Setting up the parser state. */
3306 scoped_restore pstate_restore = make_scoped_restore (&pstate);
3307 gdb_assert (par_state != NULL);
3310 c_parse_state cstate;
3311 scoped_restore cstate_restore = make_scoped_restore (&cpstate, &cstate);
3313 gdb::unique_xmalloc_ptr<struct macro_scope> macro_scope;
3315 if (par_state->expression_context_block)
3317 = sal_macro_scope (find_pc_line (par_state->expression_context_pc, 0));
3319 macro_scope = default_macro_scope ();
3321 macro_scope = user_macro_scope ();
3323 scoped_restore restore_macro_scope
3324 = make_scoped_restore (&expression_macro_scope, macro_scope.get ());
3326 scoped_restore restore_yydebug = make_scoped_restore (&yydebug,
3329 /* Initialize some state used by the lexer. */
3330 last_was_structop = false;
3331 saw_name_at_eof = 0;
3334 token_fifo.clear ();
3336 name_obstack.clear ();
3343 /* This is called via the YYPRINT macro when parser debugging is
3344 enabled. It prints a token's value. */
3347 c_print_token (FILE *file, int type, YYSTYPE value)
3352 parser_fprintf (file, "typed_val_int<%s, %s>",
3353 TYPE_SAFE_NAME (value.typed_val_int.type),
3354 pulongest (value.typed_val_int.val));
3360 char *copy = (char *) alloca (value.tsval.length + 1);
3362 memcpy (copy, value.tsval.ptr, value.tsval.length);
3363 copy[value.tsval.length] = '\0';
3365 parser_fprintf (file, "tsval<type=%d, %s>", value.tsval.type, copy);
3370 case DOLLAR_VARIABLE:
3371 parser_fprintf (file, "sval<%s>", copy_name (value.sval));
3375 parser_fprintf (file, "tsym<type=%s, name=%s>",
3376 TYPE_SAFE_NAME (value.tsym.type),
3377 copy_name (value.tsym.stoken));
3381 case UNKNOWN_CPP_NAME:
3384 parser_fprintf (file, "ssym<name=%s, sym=%s, field_of_this=%d>",
3385 copy_name (value.ssym.stoken),
3386 (value.ssym.sym.symbol == NULL
3387 ? "(null)" : SYMBOL_PRINT_NAME (value.ssym.sym.symbol)),
3388 value.ssym.is_a_field_of_this);
3392 parser_fprintf (file, "bval<%s>", host_address_to_string (value.bval));
3400 yyerror (const char *msg)
3402 if (pstate->prev_lexptr)
3403 pstate->lexptr = pstate->prev_lexptr;
3405 error (_("A %s in expression, near `%s'."), msg, pstate->lexptr);