1 /* YACC parser for C expressions, for GDB.
2 Copyright (C) 1986-2021 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"
57 #include "target-float.h"
60 #define parse_type(ps) builtin_type (ps->gdbarch ())
62 /* Remap normal yacc parser interface names (yyparse, yylex, yyerror,
64 #define GDB_YY_REMAP_PREFIX c_
67 /* The state of the parser, used internally when we are parsing the
70 static struct parser_state *pstate = NULL;
72 /* Data that must be held for the duration of a parse. */
76 /* These are used to hold type lists and type stacks that are
77 allocated during the parse. */
78 std::vector<std::unique_ptr<std::vector<struct type *>>> type_lists;
79 std::vector<std::unique_ptr<struct type_stack>> type_stacks;
81 /* Storage for some strings allocated during the parse. */
82 std::vector<gdb::unique_xmalloc_ptr<char>> strings;
84 /* When we find that lexptr (the global var defined in parse.c) is
85 pointing at a macro invocation, we expand the invocation, and call
86 scan_macro_expansion to save the old lexptr here and point lexptr
87 into the expanded text. When we reach the end of that, we call
88 end_macro_expansion to pop back to the value we saved here. The
89 macro expansion code promises to return only fully-expanded text,
90 so we don't need to "push" more than one level.
92 This is disgusting, of course. It would be cleaner to do all macro
93 expansion beforehand, and then hand that to lexptr. But we don't
94 really know where the expression ends. Remember, in a command like
96 (gdb) break *ADDRESS if CONDITION
98 we evaluate ADDRESS in the scope of the current frame, but we
99 evaluate CONDITION in the scope of the breakpoint's location. So
100 it's simply wrong to try to macro-expand the whole thing at once. */
101 const char *macro_original_text = nullptr;
103 /* We save all intermediate macro expansions on this obstack for the
104 duration of a single parse. The expansion text may sometimes have
105 to live past the end of the expansion, due to yacc lookahead.
106 Rather than try to be clever about saving the data for a single
107 token, we simply keep it all and delete it after parsing has
109 auto_obstack expansion_obstack;
111 /* The type stack. */
112 struct type_stack type_stack;
115 /* This is set and cleared in c_parse. */
117 static struct c_parse_state *cpstate;
121 static int yylex (void);
123 static void yyerror (const char *);
125 static int type_aggregate_p (struct type *);
127 using namespace expr;
130 /* Although the yacc "value" of an expression is not used,
131 since the result is stored in the structure being created,
132 other node types do have values. */
147 struct typed_stoken tsval;
149 struct symtoken ssym;
151 const struct block *bval;
152 enum exp_opcode opcode;
154 struct stoken_vector svec;
155 std::vector<struct type *> *tvec;
157 struct type_stack *type_stack;
159 struct objc_class_str theclass;
163 /* YYSTYPE gets defined by %union */
164 static int parse_number (struct parser_state *par_state,
165 const char *, int, int, YYSTYPE *);
166 static struct stoken operator_stoken (const char *);
167 static struct stoken typename_stoken (const char *);
168 static void check_parameter_typelist (std::vector<struct type *> *);
171 static void c_print_token (FILE *file, int type, YYSTYPE value);
172 #define YYPRINT(FILE, TYPE, VALUE) c_print_token (FILE, TYPE, VALUE)
176 %type <voidval> exp exp1 type_exp start variable qualified_name lcurly function_method
178 %type <tval> type typebase scalar_type
179 %type <tvec> nonempty_typelist func_mod parameter_typelist
180 /* %type <bval> block */
182 /* Fancy type parsing. */
184 %type <lval> array_mod
185 %type <tval> conversion_type_id
187 %type <type_stack> ptr_operator_ts abs_decl direct_abs_decl
189 %token <typed_val_int> INT COMPLEX_INT
190 %token <typed_val_float> FLOAT COMPLEX_FLOAT
192 /* Both NAME and TYPENAME tokens represent symbols in the input,
193 and both convey their data as strings.
194 But a TYPENAME is a string that happens to be defined as a typedef
195 or builtin type name (such as int or char)
196 and a NAME is any other symbol.
197 Contexts where this distinction is not important can use the
198 nonterminal "name", which matches either NAME or TYPENAME. */
200 %token <tsval> STRING
201 %token <sval> NSSTRING /* ObjC Foundation "NSString" literal */
202 %token SELECTOR /* ObjC "@selector" pseudo-operator */
204 %token <ssym> NAME /* BLOCKNAME defined below to give it higher precedence. */
205 %token <ssym> UNKNOWN_CPP_NAME
206 %token <voidval> COMPLETE
207 %token <tsym> TYPENAME
208 %token <theclass> CLASSNAME /* ObjC Class name */
209 %type <sval> name field_name
210 %type <svec> string_exp
211 %type <ssym> name_not_typename
212 %type <tsym> type_name
214 /* This is like a '[' token, but is only generated when parsing
215 Objective C. This lets us reuse the same parser without
216 erroneously parsing ObjC-specific expressions in C. */
219 /* A NAME_OR_INT is a symbol which is not known in the symbol table,
220 but which would parse as a valid number in the current input radix.
221 E.g. "c" when input_radix==16. Depending on the parse, it will be
222 turned into a name or into a number. */
224 %token <ssym> NAME_OR_INT
227 %token STRUCT CLASS UNION ENUM SIZEOF ALIGNOF UNSIGNED COLONCOLON
232 %token REINTERPRET_CAST DYNAMIC_CAST STATIC_CAST CONST_CAST
238 /* Special type cases, put in to allow the parser to distinguish different
240 %token SIGNED_KEYWORD LONG SHORT INT_KEYWORD CONST_KEYWORD VOLATILE_KEYWORD DOUBLE_KEYWORD
241 %token RESTRICT ATOMIC
242 %token FLOAT_KEYWORD COMPLEX
244 %token <sval> DOLLAR_VARIABLE
246 %token <opcode> ASSIGN_MODIFY
255 %right '=' ASSIGN_MODIFY
263 %left '<' '>' LEQ GEQ
268 %right UNARY INCREMENT DECREMENT
269 %right ARROW ARROW_STAR '.' DOT_STAR '[' OBJC_LBRAC '('
270 %token <ssym> BLOCKNAME
271 %token <bval> FILENAME
286 pstate->push_new<type_operation> ($1);
290 pstate->wrap<typeof_operation> ();
292 | TYPEOF '(' type ')'
294 pstate->push_new<type_operation> ($3);
296 | DECLTYPE '(' exp ')'
298 pstate->wrap<decltype_operation> ();
302 /* Expressions, including the comma operator. */
305 { pstate->wrap2<comma_operation> (); }
308 /* Expressions, not including the comma operator. */
309 exp : '*' exp %prec UNARY
310 { pstate->wrap<unop_ind_operation> (); }
313 exp : '&' exp %prec UNARY
314 { pstate->wrap<unop_addr_operation> (); }
317 exp : '-' exp %prec UNARY
318 { pstate->wrap<unary_neg_operation> (); }
321 exp : '+' exp %prec UNARY
322 { pstate->wrap<unary_plus_operation> (); }
325 exp : '!' exp %prec UNARY
327 if (pstate->language ()->la_language
329 pstate->wrap<opencl_not_operation> ();
331 pstate->wrap<unary_logical_not_operation> ();
335 exp : '~' exp %prec UNARY
336 { pstate->wrap<unary_complement_operation> (); }
339 exp : INCREMENT exp %prec UNARY
340 { pstate->wrap<preinc_operation> (); }
343 exp : DECREMENT exp %prec UNARY
344 { pstate->wrap<predec_operation> (); }
347 exp : exp INCREMENT %prec UNARY
348 { pstate->wrap<postinc_operation> (); }
351 exp : exp DECREMENT %prec UNARY
352 { pstate->wrap<postdec_operation> (); }
355 exp : TYPEID '(' exp ')' %prec UNARY
356 { pstate->wrap<typeid_operation> (); }
359 exp : TYPEID '(' type_exp ')' %prec UNARY
360 { pstate->wrap<typeid_operation> (); }
363 exp : SIZEOF exp %prec UNARY
364 { pstate->wrap<unop_sizeof_operation> (); }
367 exp : ALIGNOF '(' type_exp ')' %prec UNARY
368 { pstate->wrap<unop_alignof_operation> (); }
371 exp : exp ARROW field_name
373 pstate->push_new<structop_ptr_operation>
374 (pstate->pop (), copy_name ($3));
378 exp : exp ARROW field_name COMPLETE
380 structop_base_operation *op
381 = new structop_ptr_operation (pstate->pop (),
383 pstate->mark_struct_expression (op);
384 pstate->push (operation_up (op));
388 exp : exp ARROW COMPLETE
390 structop_base_operation *op
391 = new structop_ptr_operation (pstate->pop (), "");
392 pstate->mark_struct_expression (op);
393 pstate->push (operation_up (op));
397 exp : exp ARROW '~' name
399 pstate->push_new<structop_ptr_operation>
400 (pstate->pop (), "~" + copy_name ($4));
404 exp : exp ARROW '~' name COMPLETE
406 structop_base_operation *op
407 = new structop_ptr_operation (pstate->pop (),
408 "~" + copy_name ($4));
409 pstate->mark_struct_expression (op);
410 pstate->push (operation_up (op));
414 exp : exp ARROW qualified_name
415 { /* exp->type::name becomes exp->*(&type::name) */
416 /* Note: this doesn't work if name is a
417 static member! FIXME */
418 pstate->wrap<unop_addr_operation> ();
419 pstate->wrap2<structop_mptr_operation> (); }
422 exp : exp ARROW_STAR exp
423 { pstate->wrap2<structop_mptr_operation> (); }
426 exp : exp '.' field_name
428 if (pstate->language ()->la_language
430 pstate->push_new<opencl_structop_operation>
431 (pstate->pop (), copy_name ($3));
433 pstate->push_new<structop_operation>
434 (pstate->pop (), copy_name ($3));
438 exp : exp '.' field_name COMPLETE
440 structop_base_operation *op
441 = new structop_operation (pstate->pop (),
443 pstate->mark_struct_expression (op);
444 pstate->push (operation_up (op));
448 exp : exp '.' COMPLETE
450 structop_base_operation *op
451 = new structop_operation (pstate->pop (), "");
452 pstate->mark_struct_expression (op);
453 pstate->push (operation_up (op));
457 exp : exp '.' '~' name
459 pstate->push_new<structop_operation>
460 (pstate->pop (), "~" + copy_name ($4));
464 exp : exp '.' '~' name COMPLETE
466 structop_base_operation *op
467 = new structop_operation (pstate->pop (),
468 "~" + copy_name ($4));
469 pstate->mark_struct_expression (op);
470 pstate->push (operation_up (op));
474 exp : exp '.' qualified_name
475 { /* exp.type::name becomes exp.*(&type::name) */
476 /* Note: this doesn't work if name is a
477 static member! FIXME */
478 pstate->wrap<unop_addr_operation> ();
479 pstate->wrap2<structop_member_operation> (); }
482 exp : exp DOT_STAR exp
483 { pstate->wrap2<structop_member_operation> (); }
486 exp : exp '[' exp1 ']'
487 { pstate->wrap2<subscript_operation> (); }
490 exp : exp OBJC_LBRAC exp1 ']'
491 { pstate->wrap2<subscript_operation> (); }
495 * The rules below parse ObjC message calls of the form:
496 * '[' target selector {':' argument}* ']'
499 exp : OBJC_LBRAC TYPENAME
503 std::string copy = copy_name ($2.stoken);
504 theclass = lookup_objc_class (pstate->gdbarch (),
507 error (_("%s is not an ObjC Class"),
509 pstate->push_new<long_const_operation>
510 (parse_type (pstate)->builtin_int,
515 { end_msglist (pstate); }
518 exp : OBJC_LBRAC CLASSNAME
520 pstate->push_new<long_const_operation>
521 (parse_type (pstate)->builtin_int,
522 (LONGEST) $2.theclass);
526 { end_msglist (pstate); }
532 { end_msglist (pstate); }
536 { add_msglist(&$1, 0); }
544 msgarg : name ':' exp
545 { add_msglist(&$1, 1); }
546 | ':' exp /* Unnamed arg. */
547 { add_msglist(0, 1); }
548 | ',' exp /* Variable number of args. */
549 { add_msglist(0, 0); }
553 /* This is to save the value of arglist_len
554 being accumulated by an outer function call. */
555 { pstate->start_arglist (); }
556 arglist ')' %prec ARROW
558 std::vector<operation_up> args
559 = pstate->pop_vector (pstate->end_arglist ());
560 pstate->push_new<funcall_operation>
561 (pstate->pop (), std::move (args));
565 /* This is here to disambiguate with the production for
566 "func()::static_var" further below, which uses
567 function_method_void. */
568 exp : exp '(' ')' %prec ARROW
570 pstate->push_new<funcall_operation>
571 (pstate->pop (), std::vector<operation_up> ());
576 exp : UNKNOWN_CPP_NAME '('
578 /* This could potentially be a an argument defined
579 lookup function (Koenig). */
580 /* This is to save the value of arglist_len
581 being accumulated by an outer function call. */
582 pstate->start_arglist ();
584 arglist ')' %prec ARROW
586 std::vector<operation_up> args
587 = pstate->pop_vector (pstate->end_arglist ());
588 pstate->push_new<adl_func_operation>
589 (copy_name ($1.stoken),
590 pstate->expression_context_block,
596 { pstate->start_arglist (); }
603 { pstate->arglist_len = 1; }
606 arglist : arglist ',' exp %prec ABOVE_COMMA
607 { pstate->arglist_len++; }
610 function_method: exp '(' parameter_typelist ')' const_or_volatile
612 std::vector<struct type *> *type_list = $3;
613 /* Save the const/volatile qualifiers as
614 recorded by the const_or_volatile
615 production's actions. */
616 type_instance_flags flags
617 = (cpstate->type_stack
618 .follow_type_instance_flags ());
619 pstate->push_new<type_instance_operation>
620 (flags, std::move (*type_list),
625 function_method_void: exp '(' ')' const_or_volatile
627 type_instance_flags flags
628 = (cpstate->type_stack
629 .follow_type_instance_flags ());
630 pstate->push_new<type_instance_operation>
631 (flags, std::vector<type *> (), pstate->pop ());
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 pstate->push_new<func_static_var_operation>
649 (pstate->pop (), copy_name ($3));
654 { $$ = pstate->end_arglist () - 1; }
656 exp : lcurly arglist rcurly %prec ARROW
658 std::vector<operation_up> args
659 = pstate->pop_vector ($3 + 1);
660 pstate->push_new<array_operation> (0, $3,
665 exp : lcurly type_exp rcurly exp %prec UNARY
666 { pstate->wrap2<unop_memval_type_operation> (); }
669 exp : '(' type_exp ')' exp %prec UNARY
671 if (pstate->language ()->la_language
673 pstate->wrap2<opencl_cast_type_operation> ();
675 pstate->wrap2<unop_cast_type_operation> ();
683 /* Binary operators in order of decreasing precedence. */
686 { pstate->wrap2<repeat_operation> (); }
690 { pstate->wrap2<mul_operation> (); }
694 { pstate->wrap2<div_operation> (); }
698 { pstate->wrap2<rem_operation> (); }
702 { pstate->wrap2<add_operation> (); }
706 { pstate->wrap2<sub_operation> (); }
710 { pstate->wrap2<lsh_operation> (); }
714 { pstate->wrap2<rsh_operation> (); }
719 if (pstate->language ()->la_language
721 pstate->wrap2<opencl_equal_operation> ();
723 pstate->wrap2<equal_operation> ();
727 exp : exp NOTEQUAL exp
729 if (pstate->language ()->la_language
731 pstate->wrap2<opencl_notequal_operation> ();
733 pstate->wrap2<notequal_operation> ();
739 if (pstate->language ()->la_language
741 pstate->wrap2<opencl_leq_operation> ();
743 pstate->wrap2<leq_operation> ();
749 if (pstate->language ()->la_language
751 pstate->wrap2<opencl_geq_operation> ();
753 pstate->wrap2<geq_operation> ();
759 if (pstate->language ()->la_language
761 pstate->wrap2<opencl_less_operation> ();
763 pstate->wrap2<less_operation> ();
769 if (pstate->language ()->la_language
771 pstate->wrap2<opencl_gtr_operation> ();
773 pstate->wrap2<gtr_operation> ();
778 { pstate->wrap2<bitwise_and_operation> (); }
782 { pstate->wrap2<bitwise_xor_operation> (); }
786 { pstate->wrap2<bitwise_ior_operation> (); }
791 if (pstate->language ()->la_language
794 operation_up rhs = pstate->pop ();
795 operation_up lhs = pstate->pop ();
796 pstate->push_new<opencl_logical_binop_operation>
797 (BINOP_LOGICAL_AND, std::move (lhs),
801 pstate->wrap2<logical_and_operation> ();
807 if (pstate->language ()->la_language
810 operation_up rhs = pstate->pop ();
811 operation_up lhs = pstate->pop ();
812 pstate->push_new<opencl_logical_binop_operation>
813 (BINOP_LOGICAL_OR, std::move (lhs),
817 pstate->wrap2<logical_or_operation> ();
821 exp : exp '?' exp ':' exp %prec '?'
823 operation_up last = pstate->pop ();
824 operation_up mid = pstate->pop ();
825 operation_up first = pstate->pop ();
826 if (pstate->language ()->la_language
828 pstate->push_new<opencl_ternop_cond_operation>
829 (std::move (first), std::move (mid),
832 pstate->push_new<ternop_cond_operation>
833 (std::move (first), std::move (mid),
840 if (pstate->language ()->la_language
842 pstate->wrap2<opencl_assign_operation> ();
844 pstate->wrap2<assign_operation> ();
848 exp : exp ASSIGN_MODIFY exp
850 operation_up rhs = pstate->pop ();
851 operation_up lhs = pstate->pop ();
852 pstate->push_new<assign_modify_operation>
853 ($2, std::move (lhs), std::move (rhs));
859 pstate->push_new<long_const_operation>
867 = (make_operation<long_const_operation>
868 (TYPE_TARGET_TYPE ($1.type), 0));
870 = (make_operation<long_const_operation>
871 (TYPE_TARGET_TYPE ($1.type), $1.val));
872 pstate->push_new<complex_operation>
873 (std::move (real), std::move (imag), $1.type);
879 struct stoken_vector vec;
882 pstate->push_c_string ($1.type, &vec);
888 parse_number (pstate, $1.stoken.ptr,
889 $1.stoken.length, 0, &val);
890 pstate->push_new<long_const_operation>
891 (val.typed_val_int.type,
892 val.typed_val_int.val);
900 std::copy (std::begin ($1.val), std::end ($1.val),
902 pstate->push_new<float_const_operation> ($1.type, data);
908 struct type *underlying
909 = TYPE_TARGET_TYPE ($1.type);
912 target_float_from_host_double (val.data (),
915 = (make_operation<float_const_operation>
918 std::copy (std::begin ($1.val), std::end ($1.val),
921 = (make_operation<float_const_operation>
924 pstate->push_new<complex_operation>
925 (std::move (real), std::move (imag),
933 exp : DOLLAR_VARIABLE
935 pstate->push_dollar ($1);
939 exp : SELECTOR '(' name ')'
941 pstate->push_new<objc_selector_operation>
946 exp : SIZEOF '(' type ')' %prec UNARY
947 { struct type *type = $3;
948 struct type *int_type
949 = lookup_signed_typename (pstate->language (),
951 type = check_typedef (type);
953 /* $5.3.3/2 of the C++ Standard (n3290 draft)
954 says of sizeof: "When applied to a reference
955 or a reference type, the result is the size of
956 the referenced type." */
957 if (TYPE_IS_REFERENCE (type))
958 type = check_typedef (TYPE_TARGET_TYPE (type));
959 pstate->push_new<long_const_operation>
960 (int_type, TYPE_LENGTH (type));
964 exp : REINTERPRET_CAST '<' type_exp '>' '(' exp ')' %prec UNARY
965 { pstate->wrap2<reinterpret_cast_operation> (); }
968 exp : STATIC_CAST '<' type_exp '>' '(' exp ')' %prec UNARY
969 { pstate->wrap2<unop_cast_type_operation> (); }
972 exp : DYNAMIC_CAST '<' type_exp '>' '(' exp ')' %prec UNARY
973 { pstate->wrap2<dynamic_cast_operation> (); }
976 exp : CONST_CAST '<' type_exp '>' '(' exp ')' %prec UNARY
977 { /* We could do more error checking here, but
978 it doesn't seem worthwhile. */
979 pstate->wrap2<unop_cast_type_operation> (); }
985 /* We copy the string here, and not in the
986 lexer, to guarantee that we do not leak a
987 string. Note that we follow the
988 NUL-termination convention of the
990 struct typed_stoken *vec = XNEW (struct typed_stoken);
995 vec->length = $1.length;
996 vec->ptr = (char *) malloc ($1.length + 1);
997 memcpy (vec->ptr, $1.ptr, $1.length + 1);
1002 /* Note that we NUL-terminate here, but just
1006 $$.tokens = XRESIZEVEC (struct typed_stoken,
1009 p = (char *) malloc ($2.length + 1);
1010 memcpy (p, $2.ptr, $2.length + 1);
1012 $$.tokens[$$.len - 1].type = $2.type;
1013 $$.tokens[$$.len - 1].length = $2.length;
1014 $$.tokens[$$.len - 1].ptr = p;
1021 c_string_type type = C_STRING;
1023 for (i = 0; i < $1.len; ++i)
1025 switch ($1.tokens[i].type)
1032 if (type != C_STRING
1033 && type != $1.tokens[i].type)
1034 error (_("Undefined string concatenation."));
1035 type = (enum c_string_type_values) $1.tokens[i].type;
1038 /* internal error */
1039 internal_error (__FILE__, __LINE__,
1040 "unrecognized type in string concatenation");
1044 pstate->push_c_string (type, &$1);
1045 for (i = 0; i < $1.len; ++i)
1046 free ($1.tokens[i].ptr);
1051 exp : NSSTRING /* ObjC NextStep NSString constant
1052 * of the form '@' '"' string '"'.
1055 pstate->push_new<objc_nsstring_operation>
1062 { pstate->push_new<long_const_operation>
1063 (parse_type (pstate)->builtin_bool, 1);
1068 { pstate->push_new<long_const_operation>
1069 (parse_type (pstate)->builtin_bool, 0);
1078 $$ = SYMBOL_BLOCK_VALUE ($1.sym.symbol);
1080 error (_("No file or function \"%s\"."),
1081 copy_name ($1.stoken).c_str ());
1089 block : block COLONCOLON name
1091 std::string copy = copy_name ($3);
1093 = lookup_symbol (copy.c_str (), $1,
1094 VAR_DOMAIN, NULL).symbol;
1096 if (!tem || SYMBOL_CLASS (tem) != LOC_BLOCK)
1097 error (_("No function \"%s\" in specified context."),
1099 $$ = SYMBOL_BLOCK_VALUE (tem); }
1102 variable: name_not_typename ENTRY
1103 { struct symbol *sym = $1.sym.symbol;
1105 if (sym == NULL || !SYMBOL_IS_ARGUMENT (sym)
1106 || !symbol_read_needs_frame (sym))
1107 error (_("@entry can be used only for function "
1108 "parameters, not for \"%s\""),
1109 copy_name ($1.stoken).c_str ());
1111 pstate->push_new<var_entry_value_operation> (sym);
1115 variable: block COLONCOLON name
1117 std::string copy = copy_name ($3);
1118 struct block_symbol sym
1119 = lookup_symbol (copy.c_str (), $1,
1122 if (sym.symbol == 0)
1123 error (_("No symbol \"%s\" in specified context."),
1125 if (symbol_read_needs_frame (sym.symbol))
1126 pstate->block_tracker->update (sym);
1128 pstate->push_new<var_value_operation> (sym);
1132 qualified_name: TYPENAME COLONCOLON name
1134 struct type *type = $1.type;
1135 type = check_typedef (type);
1136 if (!type_aggregate_p (type))
1137 error (_("`%s' is not defined as an aggregate type."),
1138 TYPE_SAFE_NAME (type));
1140 pstate->push_new<scope_operation> (type,
1143 | TYPENAME COLONCOLON '~' name
1145 struct type *type = $1.type;
1147 type = check_typedef (type);
1148 if (!type_aggregate_p (type))
1149 error (_("`%s' is not defined as an aggregate type."),
1150 TYPE_SAFE_NAME (type));
1151 std::string name = "~" + std::string ($4.ptr,
1154 /* Check for valid destructor name. */
1155 destructor_name_p (name.c_str (), $1.type);
1156 pstate->push_new<scope_operation> (type,
1159 | TYPENAME COLONCOLON name COLONCOLON name
1161 std::string copy = copy_name ($3);
1162 error (_("No type \"%s\" within class "
1163 "or namespace \"%s\"."),
1164 copy.c_str (), TYPE_SAFE_NAME ($1.type));
1168 variable: qualified_name
1169 | COLONCOLON name_not_typename
1171 std::string name = copy_name ($2.stoken);
1172 struct block_symbol sym
1173 = lookup_symbol (name.c_str (),
1174 (const struct block *) NULL,
1176 pstate->push_symbol (name.c_str (), sym);
1180 variable: name_not_typename
1181 { struct block_symbol sym = $1.sym;
1185 if (symbol_read_needs_frame (sym.symbol))
1186 pstate->block_tracker->update (sym);
1188 /* If we found a function, see if it's
1189 an ifunc resolver that has the same
1190 address as the ifunc symbol itself.
1191 If so, prefer the ifunc symbol. */
1193 bound_minimal_symbol resolver
1194 = find_gnu_ifunc (sym.symbol);
1195 if (resolver.minsym != NULL)
1196 pstate->push_new<var_msym_value_operation>
1199 pstate->push_new<var_value_operation> (sym);
1201 else if ($1.is_a_field_of_this)
1203 /* C++: it hangs off of `this'. Must
1204 not inadvertently convert from a method call
1206 pstate->block_tracker->update (sym);
1208 = make_operation<op_this_operation> ();
1209 pstate->push_new<structop_ptr_operation>
1210 (std::move (thisop), copy_name ($1.stoken));
1214 std::string arg = copy_name ($1.stoken);
1216 bound_minimal_symbol msymbol
1217 = lookup_bound_minimal_symbol (arg.c_str ());
1218 if (msymbol.minsym == NULL)
1220 if (!have_full_symbols () && !have_partial_symbols ())
1221 error (_("No symbol table is loaded. Use the \"file\" command."));
1223 error (_("No symbol \"%s\" in current context."),
1227 /* This minsym might be an alias for
1228 another function. See if we can find
1229 the debug symbol for the target, and
1230 if so, use it instead, since it has
1231 return type / prototype info. This
1232 is important for example for "p
1233 *__errno_location()". */
1234 symbol *alias_target
1235 = ((msymbol.minsym->type != mst_text_gnu_ifunc
1236 && msymbol.minsym->type != mst_data_gnu_ifunc)
1237 ? find_function_alias_target (msymbol)
1239 if (alias_target != NULL)
1241 block_symbol bsym { alias_target,
1242 SYMBOL_BLOCK_VALUE (alias_target) };
1243 pstate->push_new<var_value_operation> (bsym);
1246 pstate->push_new<var_msym_value_operation>
1252 const_or_volatile: const_or_volatile_noopt
1258 { cpstate->type_stack.insert (tp_const); }
1260 { cpstate->type_stack.insert (tp_volatile); }
1262 { cpstate->type_stack.insert (tp_atomic); }
1264 { cpstate->type_stack.insert (tp_restrict); }
1267 cpstate->type_stack.insert (pstate,
1268 copy_name ($2.stoken).c_str ());
1270 | '@' UNKNOWN_CPP_NAME
1272 cpstate->type_stack.insert (pstate,
1273 copy_name ($2.stoken).c_str ());
1277 qualifier_seq_noopt:
1279 | qualifier_seq_noopt single_qualifier
1289 { cpstate->type_stack.insert (tp_pointer); }
1292 { cpstate->type_stack.insert (tp_pointer); }
1295 { cpstate->type_stack.insert (tp_reference); }
1297 { cpstate->type_stack.insert (tp_reference); }
1299 { cpstate->type_stack.insert (tp_rvalue_reference); }
1300 | ANDAND ptr_operator
1301 { cpstate->type_stack.insert (tp_rvalue_reference); }
1304 ptr_operator_ts: ptr_operator
1306 $$ = cpstate->type_stack.create ();
1307 cpstate->type_stacks.emplace_back ($$);
1311 abs_decl: ptr_operator_ts direct_abs_decl
1312 { $$ = $2->append ($1); }
1317 direct_abs_decl: '(' abs_decl ')'
1319 | direct_abs_decl array_mod
1321 cpstate->type_stack.push ($1);
1322 cpstate->type_stack.push ($2);
1323 cpstate->type_stack.push (tp_array);
1324 $$ = cpstate->type_stack.create ();
1325 cpstate->type_stacks.emplace_back ($$);
1329 cpstate->type_stack.push ($1);
1330 cpstate->type_stack.push (tp_array);
1331 $$ = cpstate->type_stack.create ();
1332 cpstate->type_stacks.emplace_back ($$);
1335 | direct_abs_decl func_mod
1337 cpstate->type_stack.push ($1);
1338 cpstate->type_stack.push ($2);
1339 $$ = cpstate->type_stack.create ();
1340 cpstate->type_stacks.emplace_back ($$);
1344 cpstate->type_stack.push ($1);
1345 $$ = cpstate->type_stack.create ();
1346 cpstate->type_stacks.emplace_back ($$);
1356 | OBJC_LBRAC INT ']'
1362 $$ = new std::vector<struct type *>;
1363 cpstate->type_lists.emplace_back ($$);
1365 | '(' parameter_typelist ')'
1369 /* We used to try to recognize pointer to member types here, but
1370 that didn't work (shift/reduce conflicts meant that these rules never
1371 got executed). The problem is that
1372 int (foo::bar::baz::bizzle)
1373 is a function type but
1374 int (foo::bar::baz::bizzle::*)
1375 is a pointer to member type. Stroustrup loses again! */
1380 /* A helper production that recognizes scalar types that can validly
1381 be used with _Complex. */
1385 { $$ = lookup_signed_typename (pstate->language (),
1388 { $$ = lookup_signed_typename (pstate->language (),
1391 { $$ = lookup_signed_typename (pstate->language (),
1394 { $$ = lookup_signed_typename (pstate->language (),
1396 | LONG SIGNED_KEYWORD INT_KEYWORD
1397 { $$ = lookup_signed_typename (pstate->language (),
1399 | LONG SIGNED_KEYWORD
1400 { $$ = lookup_signed_typename (pstate->language (),
1402 | SIGNED_KEYWORD LONG INT_KEYWORD
1403 { $$ = lookup_signed_typename (pstate->language (),
1405 | UNSIGNED LONG INT_KEYWORD
1406 { $$ = lookup_unsigned_typename (pstate->language (),
1408 | LONG UNSIGNED INT_KEYWORD
1409 { $$ = lookup_unsigned_typename (pstate->language (),
1412 { $$ = lookup_unsigned_typename (pstate->language (),
1415 { $$ = lookup_signed_typename (pstate->language (),
1417 | LONG LONG INT_KEYWORD
1418 { $$ = lookup_signed_typename (pstate->language (),
1420 | LONG LONG SIGNED_KEYWORD INT_KEYWORD
1421 { $$ = lookup_signed_typename (pstate->language (),
1423 | LONG LONG SIGNED_KEYWORD
1424 { $$ = lookup_signed_typename (pstate->language (),
1426 | SIGNED_KEYWORD LONG LONG
1427 { $$ = lookup_signed_typename (pstate->language (),
1429 | SIGNED_KEYWORD LONG LONG INT_KEYWORD
1430 { $$ = lookup_signed_typename (pstate->language (),
1432 | UNSIGNED LONG LONG
1433 { $$ = lookup_unsigned_typename (pstate->language (),
1435 | UNSIGNED LONG LONG INT_KEYWORD
1436 { $$ = lookup_unsigned_typename (pstate->language (),
1438 | LONG LONG UNSIGNED
1439 { $$ = lookup_unsigned_typename (pstate->language (),
1441 | LONG LONG UNSIGNED INT_KEYWORD
1442 { $$ = lookup_unsigned_typename (pstate->language (),
1445 { $$ = lookup_signed_typename (pstate->language (),
1447 | SHORT SIGNED_KEYWORD INT_KEYWORD
1448 { $$ = lookup_signed_typename (pstate->language (),
1450 | SHORT SIGNED_KEYWORD
1451 { $$ = lookup_signed_typename (pstate->language (),
1453 | UNSIGNED SHORT INT_KEYWORD
1454 { $$ = lookup_unsigned_typename (pstate->language (),
1457 { $$ = lookup_unsigned_typename (pstate->language (),
1459 | SHORT UNSIGNED INT_KEYWORD
1460 { $$ = lookup_unsigned_typename (pstate->language (),
1463 { $$ = lookup_typename (pstate->language (),
1468 { $$ = lookup_typename (pstate->language (),
1472 | LONG DOUBLE_KEYWORD
1473 { $$ = lookup_typename (pstate->language (),
1477 | UNSIGNED type_name
1478 { $$ = lookup_unsigned_typename (pstate->language (),
1479 $2.type->name ()); }
1481 { $$ = lookup_unsigned_typename (pstate->language (),
1483 | SIGNED_KEYWORD type_name
1484 { $$ = lookup_signed_typename (pstate->language (),
1485 $2.type->name ()); }
1487 { $$ = lookup_signed_typename (pstate->language (),
1491 /* Implements (approximately): (type-qualifier)* type-specifier.
1493 When type-specifier is only ever a single word, like 'float' then these
1494 arrive as pre-built TYPENAME tokens thanks to the classify_name
1495 function. However, when a type-specifier can contain multiple words,
1496 for example 'double' can appear as just 'double' or 'long double', and
1497 similarly 'long' can appear as just 'long' or in 'long double', then
1498 these type-specifiers are parsed into their own tokens in the function
1499 lex_one_token and the ident_tokens array. These separate tokens are all
1506 | COMPLEX scalar_type
1508 $$ = init_complex_type (nullptr, $2);
1512 = lookup_struct (copy_name ($2).c_str (),
1513 pstate->expression_context_block);
1517 pstate->mark_completion_tag (TYPE_CODE_STRUCT,
1521 | STRUCT name COMPLETE
1523 pstate->mark_completion_tag (TYPE_CODE_STRUCT,
1528 { $$ = lookup_struct
1529 (copy_name ($2).c_str (),
1530 pstate->expression_context_block);
1534 pstate->mark_completion_tag (TYPE_CODE_STRUCT,
1538 | CLASS name COMPLETE
1540 pstate->mark_completion_tag (TYPE_CODE_STRUCT,
1546 = lookup_union (copy_name ($2).c_str (),
1547 pstate->expression_context_block);
1551 pstate->mark_completion_tag (TYPE_CODE_UNION,
1555 | UNION name COMPLETE
1557 pstate->mark_completion_tag (TYPE_CODE_UNION,
1562 { $$ = lookup_enum (copy_name ($2).c_str (),
1563 pstate->expression_context_block);
1567 pstate->mark_completion_tag (TYPE_CODE_ENUM, "", 0);
1570 | ENUM name COMPLETE
1572 pstate->mark_completion_tag (TYPE_CODE_ENUM, $2.ptr,
1576 /* It appears that this rule for templates is never
1577 reduced; template recognition happens by lookahead
1578 in the token processing code in yylex. */
1579 | TEMPLATE name '<' type '>'
1580 { $$ = lookup_template_type
1581 (copy_name($2).c_str (), $4,
1582 pstate->expression_context_block);
1584 | qualifier_seq_noopt typebase
1585 { $$ = cpstate->type_stack.follow_types ($2); }
1586 | typebase qualifier_seq_noopt
1587 { $$ = cpstate->type_stack.follow_types ($1); }
1593 $$.stoken.ptr = "int";
1594 $$.stoken.length = 3;
1595 $$.type = lookup_signed_typename (pstate->language (),
1600 $$.stoken.ptr = "long";
1601 $$.stoken.length = 4;
1602 $$.type = lookup_signed_typename (pstate->language (),
1607 $$.stoken.ptr = "short";
1608 $$.stoken.length = 5;
1609 $$.type = lookup_signed_typename (pstate->language (),
1616 { check_parameter_typelist ($1); }
1617 | nonempty_typelist ',' DOTDOTDOT
1619 $1->push_back (NULL);
1620 check_parameter_typelist ($1);
1628 std::vector<struct type *> *typelist
1629 = new std::vector<struct type *>;
1630 cpstate->type_lists.emplace_back (typelist);
1632 typelist->push_back ($1);
1635 | nonempty_typelist ',' type
1645 cpstate->type_stack.push ($2);
1646 $$ = cpstate->type_stack.follow_types ($1);
1650 conversion_type_id: typebase conversion_declarator
1651 { $$ = cpstate->type_stack.follow_types ($1); }
1654 conversion_declarator: /* Nothing. */
1655 | ptr_operator conversion_declarator
1658 const_and_volatile: CONST_KEYWORD VOLATILE_KEYWORD
1659 | VOLATILE_KEYWORD CONST_KEYWORD
1662 const_or_volatile_noopt: const_and_volatile
1663 { cpstate->type_stack.insert (tp_const);
1664 cpstate->type_stack.insert (tp_volatile);
1667 { cpstate->type_stack.insert (tp_const); }
1669 { cpstate->type_stack.insert (tp_volatile); }
1673 { $$ = operator_stoken (" new"); }
1675 { $$ = operator_stoken (" delete"); }
1676 | OPERATOR NEW '[' ']'
1677 { $$ = operator_stoken (" new[]"); }
1678 | OPERATOR DELETE '[' ']'
1679 { $$ = operator_stoken (" delete[]"); }
1680 | OPERATOR NEW OBJC_LBRAC ']'
1681 { $$ = operator_stoken (" new[]"); }
1682 | OPERATOR DELETE OBJC_LBRAC ']'
1683 { $$ = operator_stoken (" delete[]"); }
1685 { $$ = operator_stoken ("+"); }
1687 { $$ = operator_stoken ("-"); }
1689 { $$ = operator_stoken ("*"); }
1691 { $$ = operator_stoken ("/"); }
1693 { $$ = operator_stoken ("%"); }
1695 { $$ = operator_stoken ("^"); }
1697 { $$ = operator_stoken ("&"); }
1699 { $$ = operator_stoken ("|"); }
1701 { $$ = operator_stoken ("~"); }
1703 { $$ = operator_stoken ("!"); }
1705 { $$ = operator_stoken ("="); }
1707 { $$ = operator_stoken ("<"); }
1709 { $$ = operator_stoken (">"); }
1710 | OPERATOR ASSIGN_MODIFY
1711 { const char *op = " unknown";
1735 case BINOP_BITWISE_IOR:
1738 case BINOP_BITWISE_AND:
1741 case BINOP_BITWISE_XOR:
1748 $$ = operator_stoken (op);
1751 { $$ = operator_stoken ("<<"); }
1753 { $$ = operator_stoken (">>"); }
1755 { $$ = operator_stoken ("=="); }
1757 { $$ = operator_stoken ("!="); }
1759 { $$ = operator_stoken ("<="); }
1761 { $$ = operator_stoken (">="); }
1763 { $$ = operator_stoken ("&&"); }
1765 { $$ = operator_stoken ("||"); }
1766 | OPERATOR INCREMENT
1767 { $$ = operator_stoken ("++"); }
1768 | OPERATOR DECREMENT
1769 { $$ = operator_stoken ("--"); }
1771 { $$ = operator_stoken (","); }
1772 | OPERATOR ARROW_STAR
1773 { $$ = operator_stoken ("->*"); }
1775 { $$ = operator_stoken ("->"); }
1777 { $$ = operator_stoken ("()"); }
1779 { $$ = operator_stoken ("[]"); }
1780 | OPERATOR OBJC_LBRAC ']'
1781 { $$ = operator_stoken ("[]"); }
1782 | OPERATOR conversion_type_id
1785 c_print_type ($2, NULL, &buf, -1, 0,
1786 &type_print_raw_options);
1787 std::string name = std::move (buf.string ());
1789 /* This also needs canonicalization. */
1790 gdb::unique_xmalloc_ptr<char> canon
1791 = cp_canonicalize_string (name.c_str ());
1792 if (canon != nullptr)
1793 name = canon.get ();
1794 $$ = operator_stoken ((" " + name).c_str ());
1798 /* This rule exists in order to allow some tokens that would not normally
1799 match the 'name' rule to appear as fields within a struct. The example
1800 that initially motivated this was the RISC-V target which models the
1801 floating point registers as a union with fields called 'float' and
1805 | DOUBLE_KEYWORD { $$ = typename_stoken ("double"); }
1806 | FLOAT_KEYWORD { $$ = typename_stoken ("float"); }
1807 | INT_KEYWORD { $$ = typename_stoken ("int"); }
1808 | LONG { $$ = typename_stoken ("long"); }
1809 | SHORT { $$ = typename_stoken ("short"); }
1810 | SIGNED_KEYWORD { $$ = typename_stoken ("signed"); }
1811 | UNSIGNED { $$ = typename_stoken ("unsigned"); }
1814 name : NAME { $$ = $1.stoken; }
1815 | BLOCKNAME { $$ = $1.stoken; }
1816 | TYPENAME { $$ = $1.stoken; }
1817 | NAME_OR_INT { $$ = $1.stoken; }
1818 | UNKNOWN_CPP_NAME { $$ = $1.stoken; }
1822 name_not_typename : NAME
1824 /* These would be useful if name_not_typename was useful, but it is just
1825 a fake for "variable", so these cause reduce/reduce conflicts because
1826 the parser can't tell whether NAME_OR_INT is a name_not_typename (=variable,
1827 =exp) or just an exp. If name_not_typename was ever used in an lvalue
1828 context where only a name could occur, this might be useful.
1833 struct field_of_this_result is_a_field_of_this;
1837 = lookup_symbol ($1.ptr,
1838 pstate->expression_context_block,
1840 &is_a_field_of_this);
1841 $$.is_a_field_of_this
1842 = is_a_field_of_this.type != NULL;
1849 /* Returns a stoken of the operator name given by OP (which does not
1850 include the string "operator"). */
1852 static struct stoken
1853 operator_stoken (const char *op)
1855 struct stoken st = { NULL, 0 };
1858 st.length = CP_OPERATOR_LEN + strlen (op);
1859 buf = (char *) malloc (st.length + 1);
1860 strcpy (buf, CP_OPERATOR_STR);
1864 /* The toplevel (c_parse) will free the memory allocated here. */
1865 cpstate->strings.emplace_back (buf);
1869 /* Returns a stoken of the type named TYPE. */
1871 static struct stoken
1872 typename_stoken (const char *type)
1874 struct stoken st = { type, 0 };
1875 st.length = strlen (type);
1879 /* Return true if the type is aggregate-like. */
1882 type_aggregate_p (struct type *type)
1884 return (type->code () == TYPE_CODE_STRUCT
1885 || type->code () == TYPE_CODE_UNION
1886 || type->code () == TYPE_CODE_NAMESPACE
1887 || (type->code () == TYPE_CODE_ENUM
1888 && type->is_declared_class ()));
1891 /* Validate a parameter typelist. */
1894 check_parameter_typelist (std::vector<struct type *> *params)
1899 for (ix = 0; ix < params->size (); ++ix)
1901 type = (*params)[ix];
1902 if (type != NULL && check_typedef (type)->code () == TYPE_CODE_VOID)
1906 if (params->size () == 1)
1911 error (_("parameter types following 'void'"));
1914 error (_("'void' invalid as parameter type"));
1919 /* Take care of parsing a number (anything that starts with a digit).
1920 Set yylval and return the token type; update lexptr.
1921 LEN is the number of characters in it. */
1923 /*** Needs some error checking for the float case ***/
1926 parse_number (struct parser_state *par_state,
1927 const char *buf, int len, int parsed_float, YYSTYPE *putithere)
1935 int base = input_radix;
1938 /* Number of "L" suffixes encountered. */
1941 /* Imaginary number. */
1942 bool imaginary_p = false;
1944 /* We have found a "L" or "U" (or "i") suffix. */
1945 int found_suffix = 0;
1948 struct type *signed_type;
1949 struct type *unsigned_type;
1952 p = (char *) alloca (len);
1953 memcpy (p, buf, len);
1957 if (len >= 1 && p[len - 1] == 'i')
1963 /* Handle suffixes for decimal floating-point: "df", "dd" or "dl". */
1964 if (len >= 2 && p[len - 2] == 'd' && p[len - 1] == 'f')
1966 putithere->typed_val_float.type
1967 = parse_type (par_state)->builtin_decfloat;
1970 else if (len >= 2 && p[len - 2] == 'd' && p[len - 1] == 'd')
1972 putithere->typed_val_float.type
1973 = parse_type (par_state)->builtin_decdouble;
1976 else if (len >= 2 && p[len - 2] == 'd' && p[len - 1] == 'l')
1978 putithere->typed_val_float.type
1979 = parse_type (par_state)->builtin_declong;
1982 /* Handle suffixes: 'f' for float, 'l' for long double. */
1983 else if (len >= 1 && TOLOWER (p[len - 1]) == 'f')
1985 putithere->typed_val_float.type
1986 = parse_type (par_state)->builtin_float;
1989 else if (len >= 1 && TOLOWER (p[len - 1]) == 'l')
1991 putithere->typed_val_float.type
1992 = parse_type (par_state)->builtin_long_double;
1995 /* Default type for floating-point literals is double. */
1998 putithere->typed_val_float.type
1999 = parse_type (par_state)->builtin_double;
2002 if (!parse_float (p, len,
2003 putithere->typed_val_float.type,
2004 putithere->typed_val_float.val))
2008 putithere->typed_val_float.type
2009 = init_complex_type (nullptr, putithere->typed_val_float.type);
2011 return imaginary_p ? COMPLEX_FLOAT : FLOAT;
2014 /* Handle base-switching prefixes 0x, 0t, 0d, 0 */
2015 if (p[0] == '0' && len > 1)
2058 if (c >= 'A' && c <= 'Z')
2060 if (c != 'l' && c != 'u' && c != 'i')
2062 if (c >= '0' && c <= '9')
2070 if (base > 10 && c >= 'a' && c <= 'f')
2074 n += i = c - 'a' + 10;
2092 return ERROR; /* Char not a digit */
2095 return ERROR; /* Invalid digit in this base */
2097 /* Portably test for overflow (only works for nonzero values, so make
2098 a second check for zero). FIXME: Can't we just make n and prevn
2099 unsigned and avoid this? */
2100 if (c != 'l' && c != 'u' && c != 'i' && (prevn >= n) && n != 0)
2101 unsigned_p = 1; /* Try something unsigned */
2103 /* Portably test for unsigned overflow.
2104 FIXME: This check is wrong; for example it doesn't find overflow
2105 on 0x123456789 when LONGEST is 32 bits. */
2106 if (c != 'l' && c != 'u' && c != 'i' && n != 0)
2108 if (unsigned_p && prevn >= n)
2109 error (_("Numeric constant too large."));
2114 /* An integer constant is an int, a long, or a long long. An L
2115 suffix forces it to be long; an LL suffix forces it to be long
2116 long. If not forced to a larger size, it gets the first type of
2117 the above that it fits in. To figure out whether it fits, we
2118 shift it right and see whether anything remains. Note that we
2119 can't shift sizeof (LONGEST) * HOST_CHAR_BIT bits or more in one
2120 operation, because many compilers will warn about such a shift
2121 (which always produces a zero result). Sometimes gdbarch_int_bit
2122 or gdbarch_long_bit will be that big, sometimes not. To deal with
2123 the case where it is we just always shift the value more than
2124 once, with fewer bits each time. */
2128 && (un >> (gdbarch_int_bit (par_state->gdbarch ()) - 2)) == 0)
2131 = ((ULONGEST)1) << (gdbarch_int_bit (par_state->gdbarch ()) - 1);
2133 /* A large decimal (not hex or octal) constant (between INT_MAX
2134 and UINT_MAX) is a long or unsigned long, according to ANSI,
2135 never an unsigned int, but this code treats it as unsigned
2136 int. This probably should be fixed. GCC gives a warning on
2139 unsigned_type = parse_type (par_state)->builtin_unsigned_int;
2140 signed_type = parse_type (par_state)->builtin_int;
2142 else if (long_p <= 1
2143 && (un >> (gdbarch_long_bit (par_state->gdbarch ()) - 2)) == 0)
2146 = ((ULONGEST)1) << (gdbarch_long_bit (par_state->gdbarch ()) - 1);
2147 unsigned_type = parse_type (par_state)->builtin_unsigned_long;
2148 signed_type = parse_type (par_state)->builtin_long;
2153 if (sizeof (ULONGEST) * HOST_CHAR_BIT
2154 < gdbarch_long_long_bit (par_state->gdbarch ()))
2155 /* A long long does not fit in a LONGEST. */
2156 shift = (sizeof (ULONGEST) * HOST_CHAR_BIT - 1);
2158 shift = (gdbarch_long_long_bit (par_state->gdbarch ()) - 1);
2159 high_bit = (ULONGEST) 1 << shift;
2160 unsigned_type = parse_type (par_state)->builtin_unsigned_long_long;
2161 signed_type = parse_type (par_state)->builtin_long_long;
2164 putithere->typed_val_int.val = n;
2166 /* If the high bit of the worked out type is set then this number
2167 has to be unsigned. */
2169 if (unsigned_p || (n & high_bit))
2171 putithere->typed_val_int.type = unsigned_type;
2175 putithere->typed_val_int.type = signed_type;
2179 putithere->typed_val_int.type
2180 = init_complex_type (nullptr, putithere->typed_val_int.type);
2182 return imaginary_p ? COMPLEX_INT : INT;
2185 /* Temporary obstack used for holding strings. */
2186 static struct obstack tempbuf;
2187 static int tempbuf_init;
2189 /* Parse a C escape sequence. The initial backslash of the sequence
2190 is at (*PTR)[-1]. *PTR will be updated to point to just after the
2191 last character of the sequence. If OUTPUT is not NULL, the
2192 translated form of the escape sequence will be written there. If
2193 OUTPUT is NULL, no output is written and the call will only affect
2194 *PTR. If an escape sequence is expressed in target bytes, then the
2195 entire sequence will simply be copied to OUTPUT. Return 1 if any
2196 character was emitted, 0 otherwise. */
2199 c_parse_escape (const char **ptr, struct obstack *output)
2201 const char *tokptr = *ptr;
2204 /* Some escape sequences undergo character set conversion. Those we
2208 /* Hex escapes do not undergo character set conversion, so keep
2209 the escape sequence for later. */
2212 obstack_grow_str (output, "\\x");
2214 if (!ISXDIGIT (*tokptr))
2215 error (_("\\x escape without a following hex digit"));
2216 while (ISXDIGIT (*tokptr))
2219 obstack_1grow (output, *tokptr);
2224 /* Octal escapes do not undergo character set conversion, so
2225 keep the escape sequence for later. */
2237 obstack_grow_str (output, "\\");
2239 i < 3 && ISDIGIT (*tokptr) && *tokptr != '8' && *tokptr != '9';
2243 obstack_1grow (output, *tokptr);
2249 /* We handle UCNs later. We could handle them here, but that
2250 would mean a spurious error in the case where the UCN could
2251 be converted to the target charset but not the host
2257 int i, len = c == 'U' ? 8 : 4;
2260 obstack_1grow (output, '\\');
2261 obstack_1grow (output, *tokptr);
2264 if (!ISXDIGIT (*tokptr))
2265 error (_("\\%c escape without a following hex digit"), c);
2266 for (i = 0; i < len && ISXDIGIT (*tokptr); ++i)
2269 obstack_1grow (output, *tokptr);
2275 /* We must pass backslash through so that it does not
2276 cause quoting during the second expansion. */
2279 obstack_grow_str (output, "\\\\");
2283 /* Escapes which undergo conversion. */
2286 obstack_1grow (output, '\a');
2291 obstack_1grow (output, '\b');
2296 obstack_1grow (output, '\f');
2301 obstack_1grow (output, '\n');
2306 obstack_1grow (output, '\r');
2311 obstack_1grow (output, '\t');
2316 obstack_1grow (output, '\v');
2320 /* GCC extension. */
2323 obstack_1grow (output, HOST_ESCAPE_CHAR);
2327 /* Backslash-newline expands to nothing at all. */
2333 /* A few escapes just expand to the character itself. */
2337 /* GCC extensions. */
2342 /* Unrecognized escapes turn into the character itself. */
2345 obstack_1grow (output, *tokptr);
2353 /* Parse a string or character literal from TOKPTR. The string or
2354 character may be wide or unicode. *OUTPTR is set to just after the
2355 end of the literal in the input string. The resulting token is
2356 stored in VALUE. This returns a token value, either STRING or
2357 CHAR, depending on what was parsed. *HOST_CHARS is set to the
2358 number of host characters in the literal. */
2361 parse_string_or_char (const char *tokptr, const char **outptr,
2362 struct typed_stoken *value, int *host_chars)
2368 /* Build the gdb internal form of the input string in tempbuf. Note
2369 that the buffer is null byte terminated *only* for the
2370 convenience of debugging gdb itself and printing the buffer
2371 contents when the buffer contains no embedded nulls. Gdb does
2372 not depend upon the buffer being null byte terminated, it uses
2373 the length string instead. This allows gdb to handle C strings
2374 (as well as strings in other languages) with embedded null
2380 obstack_free (&tempbuf, NULL);
2381 obstack_init (&tempbuf);
2383 /* Record the string type. */
2386 type = C_WIDE_STRING;
2389 else if (*tokptr == 'u')
2394 else if (*tokptr == 'U')
2399 else if (*tokptr == '@')
2401 /* An Objective C string. */
2409 /* Skip the quote. */
2423 *host_chars += c_parse_escape (&tokptr, &tempbuf);
2425 else if (c == quote)
2429 obstack_1grow (&tempbuf, c);
2431 /* FIXME: this does the wrong thing with multi-byte host
2432 characters. We could use mbrlen here, but that would
2433 make "set host-charset" a bit less useful. */
2438 if (*tokptr != quote)
2441 error (_("Unterminated string in expression."));
2443 error (_("Unmatched single quote."));
2448 value->ptr = (char *) obstack_base (&tempbuf);
2449 value->length = obstack_object_size (&tempbuf);
2453 return quote == '"' ? (is_objc ? NSSTRING : STRING) : CHAR;
2456 /* This is used to associate some attributes with a token. */
2460 /* If this bit is set, the token is C++-only. */
2464 /* If this bit is set, the token is C-only. */
2468 /* If this bit is set, the token is conditional: if there is a
2469 symbol of the same name, then the token is a symbol; otherwise,
2470 the token is a keyword. */
2474 DEF_ENUM_FLAGS_TYPE (enum token_flag, token_flags);
2480 enum exp_opcode opcode;
2484 static const struct token tokentab3[] =
2486 {">>=", ASSIGN_MODIFY, BINOP_RSH, 0},
2487 {"<<=", ASSIGN_MODIFY, BINOP_LSH, 0},
2488 {"->*", ARROW_STAR, OP_NULL, FLAG_CXX},
2489 {"...", DOTDOTDOT, OP_NULL, 0}
2492 static const struct token tokentab2[] =
2494 {"+=", ASSIGN_MODIFY, BINOP_ADD, 0},
2495 {"-=", ASSIGN_MODIFY, BINOP_SUB, 0},
2496 {"*=", ASSIGN_MODIFY, BINOP_MUL, 0},
2497 {"/=", ASSIGN_MODIFY, BINOP_DIV, 0},
2498 {"%=", ASSIGN_MODIFY, BINOP_REM, 0},
2499 {"|=", ASSIGN_MODIFY, BINOP_BITWISE_IOR, 0},
2500 {"&=", ASSIGN_MODIFY, BINOP_BITWISE_AND, 0},
2501 {"^=", ASSIGN_MODIFY, BINOP_BITWISE_XOR, 0},
2502 {"++", INCREMENT, OP_NULL, 0},
2503 {"--", DECREMENT, OP_NULL, 0},
2504 {"->", ARROW, OP_NULL, 0},
2505 {"&&", ANDAND, OP_NULL, 0},
2506 {"||", OROR, OP_NULL, 0},
2507 /* "::" is *not* only C++: gdb overrides its meaning in several
2508 different ways, e.g., 'filename'::func, function::variable. */
2509 {"::", COLONCOLON, OP_NULL, 0},
2510 {"<<", LSH, OP_NULL, 0},
2511 {">>", RSH, OP_NULL, 0},
2512 {"==", EQUAL, OP_NULL, 0},
2513 {"!=", NOTEQUAL, OP_NULL, 0},
2514 {"<=", LEQ, OP_NULL, 0},
2515 {">=", GEQ, OP_NULL, 0},
2516 {".*", DOT_STAR, OP_NULL, FLAG_CXX}
2519 /* Identifier-like tokens. Only type-specifiers than can appear in
2520 multi-word type names (for example 'double' can appear in 'long
2521 double') need to be listed here. type-specifiers that are only ever
2522 single word (like 'char') are handled by the classify_name function. */
2523 static const struct token ident_tokens[] =
2525 {"unsigned", UNSIGNED, OP_NULL, 0},
2526 {"template", TEMPLATE, OP_NULL, FLAG_CXX},
2527 {"volatile", VOLATILE_KEYWORD, OP_NULL, 0},
2528 {"struct", STRUCT, OP_NULL, 0},
2529 {"signed", SIGNED_KEYWORD, OP_NULL, 0},
2530 {"sizeof", SIZEOF, OP_NULL, 0},
2531 {"_Alignof", ALIGNOF, OP_NULL, 0},
2532 {"alignof", ALIGNOF, OP_NULL, FLAG_CXX},
2533 {"double", DOUBLE_KEYWORD, OP_NULL, 0},
2534 {"float", FLOAT_KEYWORD, OP_NULL, 0},
2535 {"false", FALSEKEYWORD, OP_NULL, FLAG_CXX},
2536 {"class", CLASS, OP_NULL, FLAG_CXX},
2537 {"union", UNION, OP_NULL, 0},
2538 {"short", SHORT, OP_NULL, 0},
2539 {"const", CONST_KEYWORD, OP_NULL, 0},
2540 {"restrict", RESTRICT, OP_NULL, FLAG_C | FLAG_SHADOW},
2541 {"__restrict__", RESTRICT, OP_NULL, 0},
2542 {"__restrict", RESTRICT, OP_NULL, 0},
2543 {"_Atomic", ATOMIC, OP_NULL, 0},
2544 {"enum", ENUM, OP_NULL, 0},
2545 {"long", LONG, OP_NULL, 0},
2546 {"_Complex", COMPLEX, OP_NULL, 0},
2547 {"__complex__", COMPLEX, OP_NULL, 0},
2549 {"true", TRUEKEYWORD, OP_NULL, FLAG_CXX},
2550 {"int", INT_KEYWORD, OP_NULL, 0},
2551 {"new", NEW, OP_NULL, FLAG_CXX},
2552 {"delete", DELETE, OP_NULL, FLAG_CXX},
2553 {"operator", OPERATOR, OP_NULL, FLAG_CXX},
2555 {"and", ANDAND, OP_NULL, FLAG_CXX},
2556 {"and_eq", ASSIGN_MODIFY, BINOP_BITWISE_AND, FLAG_CXX},
2557 {"bitand", '&', OP_NULL, FLAG_CXX},
2558 {"bitor", '|', OP_NULL, FLAG_CXX},
2559 {"compl", '~', OP_NULL, FLAG_CXX},
2560 {"not", '!', OP_NULL, FLAG_CXX},
2561 {"not_eq", NOTEQUAL, OP_NULL, FLAG_CXX},
2562 {"or", OROR, OP_NULL, FLAG_CXX},
2563 {"or_eq", ASSIGN_MODIFY, BINOP_BITWISE_IOR, FLAG_CXX},
2564 {"xor", '^', OP_NULL, FLAG_CXX},
2565 {"xor_eq", ASSIGN_MODIFY, BINOP_BITWISE_XOR, FLAG_CXX},
2567 {"const_cast", CONST_CAST, OP_NULL, FLAG_CXX },
2568 {"dynamic_cast", DYNAMIC_CAST, OP_NULL, FLAG_CXX },
2569 {"static_cast", STATIC_CAST, OP_NULL, FLAG_CXX },
2570 {"reinterpret_cast", REINTERPRET_CAST, OP_NULL, FLAG_CXX },
2572 {"__typeof__", TYPEOF, OP_TYPEOF, 0 },
2573 {"__typeof", TYPEOF, OP_TYPEOF, 0 },
2574 {"typeof", TYPEOF, OP_TYPEOF, FLAG_SHADOW },
2575 {"__decltype", DECLTYPE, OP_DECLTYPE, FLAG_CXX },
2576 {"decltype", DECLTYPE, OP_DECLTYPE, FLAG_CXX | FLAG_SHADOW },
2578 {"typeid", TYPEID, OP_TYPEID, FLAG_CXX}
2583 scan_macro_expansion (const char *expansion)
2585 /* We'd better not be trying to push the stack twice. */
2586 gdb_assert (! cpstate->macro_original_text);
2588 /* Copy to the obstack. */
2589 const char *copy = obstack_strdup (&cpstate->expansion_obstack, expansion);
2591 /* Save the old lexptr value, so we can return to it when we're done
2592 parsing the expanded text. */
2593 cpstate->macro_original_text = pstate->lexptr;
2594 pstate->lexptr = copy;
2598 scanning_macro_expansion (void)
2600 return cpstate->macro_original_text != 0;
2604 finished_macro_expansion (void)
2606 /* There'd better be something to pop back to. */
2607 gdb_assert (cpstate->macro_original_text);
2609 /* Pop back to the original text. */
2610 pstate->lexptr = cpstate->macro_original_text;
2611 cpstate->macro_original_text = 0;
2614 /* Return true iff the token represents a C++ cast operator. */
2617 is_cast_operator (const char *token, int len)
2619 return (! strncmp (token, "dynamic_cast", len)
2620 || ! strncmp (token, "static_cast", len)
2621 || ! strncmp (token, "reinterpret_cast", len)
2622 || ! strncmp (token, "const_cast", len));
2625 /* The scope used for macro expansion. */
2626 static struct macro_scope *expression_macro_scope;
2628 /* This is set if a NAME token appeared at the very end of the input
2629 string, with no whitespace separating the name from the EOF. This
2630 is used only when parsing to do field name completion. */
2631 static int saw_name_at_eof;
2633 /* This is set if the previously-returned token was a structure
2634 operator -- either '.' or ARROW. */
2635 static bool last_was_structop;
2637 /* Depth of parentheses. */
2638 static int paren_depth;
2640 /* Read one token, getting characters through lexptr. */
2643 lex_one_token (struct parser_state *par_state, bool *is_quoted_name)
2648 const char *tokstart;
2649 bool saw_structop = last_was_structop;
2651 last_was_structop = false;
2652 *is_quoted_name = false;
2656 /* Check if this is a macro invocation that we need to expand. */
2657 if (! scanning_macro_expansion ())
2659 gdb::unique_xmalloc_ptr<char> expanded
2660 = macro_expand_next (&pstate->lexptr, *expression_macro_scope);
2662 if (expanded != nullptr)
2663 scan_macro_expansion (expanded.get ());
2666 pstate->prev_lexptr = pstate->lexptr;
2668 tokstart = pstate->lexptr;
2669 /* See if it is a special token of length 3. */
2670 for (i = 0; i < sizeof tokentab3 / sizeof tokentab3[0]; i++)
2671 if (strncmp (tokstart, tokentab3[i].oper, 3) == 0)
2673 if ((tokentab3[i].flags & FLAG_CXX) != 0
2674 && par_state->language ()->la_language != language_cplus)
2676 gdb_assert ((tokentab3[i].flags & FLAG_C) == 0);
2678 pstate->lexptr += 3;
2679 yylval.opcode = tokentab3[i].opcode;
2680 return tokentab3[i].token;
2683 /* See if it is a special token of length 2. */
2684 for (i = 0; i < sizeof tokentab2 / sizeof tokentab2[0]; i++)
2685 if (strncmp (tokstart, tokentab2[i].oper, 2) == 0)
2687 if ((tokentab2[i].flags & FLAG_CXX) != 0
2688 && par_state->language ()->la_language != language_cplus)
2690 gdb_assert ((tokentab2[i].flags & FLAG_C) == 0);
2692 pstate->lexptr += 2;
2693 yylval.opcode = tokentab2[i].opcode;
2694 if (tokentab2[i].token == ARROW)
2695 last_was_structop = 1;
2696 return tokentab2[i].token;
2699 switch (c = *tokstart)
2702 /* If we were just scanning the result of a macro expansion,
2703 then we need to resume scanning the original text.
2704 If we're parsing for field name completion, and the previous
2705 token allows such completion, return a COMPLETE token.
2706 Otherwise, we were already scanning the original text, and
2707 we're really done. */
2708 if (scanning_macro_expansion ())
2710 finished_macro_expansion ();
2713 else if (saw_name_at_eof)
2715 saw_name_at_eof = 0;
2718 else if (par_state->parse_completion && saw_structop)
2733 if (par_state->language ()->la_language == language_objc
2740 if (paren_depth == 0)
2747 if (pstate->comma_terminates
2749 && ! scanning_macro_expansion ())
2755 /* Might be a floating point number. */
2756 if (pstate->lexptr[1] < '0' || pstate->lexptr[1] > '9')
2758 last_was_structop = true;
2759 goto symbol; /* Nope, must be a symbol. */
2774 /* It's a number. */
2775 int got_dot = 0, got_e = 0, got_p = 0, toktype;
2776 const char *p = tokstart;
2777 int hex = input_radix > 10;
2779 if (c == '0' && (p[1] == 'x' || p[1] == 'X'))
2784 else if (c == '0' && (p[1]=='t' || p[1]=='T' || p[1]=='d' || p[1]=='D'))
2792 /* This test includes !hex because 'e' is a valid hex digit
2793 and thus does not indicate a floating point number when
2794 the radix is hex. */
2795 if (!hex && !got_e && !got_p && (*p == 'e' || *p == 'E'))
2796 got_dot = got_e = 1;
2797 else if (!got_e && !got_p && (*p == 'p' || *p == 'P'))
2798 got_dot = got_p = 1;
2799 /* This test does not include !hex, because a '.' always indicates
2800 a decimal floating point number regardless of the radix. */
2801 else if (!got_dot && *p == '.')
2803 else if (((got_e && (p[-1] == 'e' || p[-1] == 'E'))
2804 || (got_p && (p[-1] == 'p' || p[-1] == 'P')))
2805 && (*p == '-' || *p == '+'))
2806 /* This is the sign of the exponent, not the end of the
2809 /* We will take any letters or digits. parse_number will
2810 complain if past the radix, or if L or U are not final. */
2811 else if ((*p < '0' || *p > '9')
2812 && ((*p < 'a' || *p > 'z')
2813 && (*p < 'A' || *p > 'Z')))
2816 toktype = parse_number (par_state, tokstart, p - tokstart,
2817 got_dot | got_e | got_p, &yylval);
2818 if (toktype == ERROR)
2820 char *err_copy = (char *) alloca (p - tokstart + 1);
2822 memcpy (err_copy, tokstart, p - tokstart);
2823 err_copy[p - tokstart] = 0;
2824 error (_("Invalid number \"%s\"."), err_copy);
2832 const char *p = &tokstart[1];
2834 if (par_state->language ()->la_language == language_objc)
2836 size_t len = strlen ("selector");
2838 if (strncmp (p, "selector", len) == 0
2839 && (p[len] == '\0' || ISSPACE (p[len])))
2841 pstate->lexptr = p + len;
2848 while (ISSPACE (*p))
2850 size_t len = strlen ("entry");
2851 if (strncmp (p, "entry", len) == 0 && !c_ident_is_alnum (p[len])
2854 pstate->lexptr = &p[len];
2883 if (tokstart[1] != '"' && tokstart[1] != '\'')
2892 int result = parse_string_or_char (tokstart, &pstate->lexptr,
2893 &yylval.tsval, &host_len);
2897 error (_("Empty character constant."));
2898 else if (host_len > 2 && c == '\'')
2901 namelen = pstate->lexptr - tokstart - 1;
2902 *is_quoted_name = true;
2906 else if (host_len > 1)
2907 error (_("Invalid character constant."));
2913 if (!(c == '_' || c == '$' || c_ident_is_alpha (c)))
2914 /* We must have come across a bad character (e.g. ';'). */
2915 error (_("Invalid character '%c' in expression."), c);
2917 /* It's a name. See how long it is. */
2919 for (c = tokstart[namelen];
2920 (c == '_' || c == '$' || c_ident_is_alnum (c) || c == '<');)
2922 /* Template parameter lists are part of the name.
2923 FIXME: This mishandles `print $a<4&&$a>3'. */
2927 if (! is_cast_operator (tokstart, namelen))
2929 /* Scan ahead to get rest of the template specification. Note
2930 that we look ahead only when the '<' adjoins non-whitespace
2931 characters; for comparison expressions, e.g. "a < b > c",
2932 there must be spaces before the '<', etc. */
2933 const char *p = find_template_name_end (tokstart + namelen);
2936 namelen = p - tokstart;
2940 c = tokstart[++namelen];
2943 /* The token "if" terminates the expression and is NOT removed from
2944 the input stream. It doesn't count if it appears in the
2945 expansion of a macro. */
2947 && tokstart[0] == 'i'
2948 && tokstart[1] == 'f'
2949 && ! scanning_macro_expansion ())
2954 /* For the same reason (breakpoint conditions), "thread N"
2955 terminates the expression. "thread" could be an identifier, but
2956 an identifier is never followed by a number without intervening
2957 punctuation. "task" is similar. Handle abbreviations of these,
2958 similarly to breakpoint.c:find_condition_and_thread. */
2960 && (strncmp (tokstart, "thread", namelen) == 0
2961 || strncmp (tokstart, "task", namelen) == 0)
2962 && (tokstart[namelen] == ' ' || tokstart[namelen] == '\t')
2963 && ! scanning_macro_expansion ())
2965 const char *p = tokstart + namelen + 1;
2967 while (*p == ' ' || *p == '\t')
2969 if (*p >= '0' && *p <= '9')
2973 pstate->lexptr += namelen;
2977 yylval.sval.ptr = tokstart;
2978 yylval.sval.length = namelen;
2980 /* Catch specific keywords. */
2981 std::string copy = copy_name (yylval.sval);
2982 for (i = 0; i < sizeof ident_tokens / sizeof ident_tokens[0]; i++)
2983 if (copy == ident_tokens[i].oper)
2985 if ((ident_tokens[i].flags & FLAG_CXX) != 0
2986 && par_state->language ()->la_language != language_cplus)
2988 if ((ident_tokens[i].flags & FLAG_C) != 0
2989 && par_state->language ()->la_language != language_c
2990 && par_state->language ()->la_language != language_objc)
2993 if ((ident_tokens[i].flags & FLAG_SHADOW) != 0)
2995 struct field_of_this_result is_a_field_of_this;
2997 if (lookup_symbol (copy.c_str (),
2998 pstate->expression_context_block,
3000 (par_state->language ()->la_language
3001 == language_cplus ? &is_a_field_of_this
3005 /* The keyword is shadowed. */
3010 /* It is ok to always set this, even though we don't always
3011 strictly need to. */
3012 yylval.opcode = ident_tokens[i].opcode;
3013 return ident_tokens[i].token;
3016 if (*tokstart == '$')
3017 return DOLLAR_VARIABLE;
3019 if (pstate->parse_completion && *pstate->lexptr == '\0')
3020 saw_name_at_eof = 1;
3022 yylval.ssym.stoken = yylval.sval;
3023 yylval.ssym.sym.symbol = NULL;
3024 yylval.ssym.sym.block = NULL;
3025 yylval.ssym.is_a_field_of_this = 0;
3029 /* An object of this type is pushed on a FIFO by the "outer" lexer. */
3030 struct token_and_value
3036 /* A FIFO of tokens that have been read but not yet returned to the
3038 static std::vector<token_and_value> token_fifo;
3040 /* Non-zero if the lexer should return tokens from the FIFO. */
3043 /* Temporary storage for c_lex; this holds symbol names as they are
3045 static auto_obstack name_obstack;
3047 /* Classify a NAME token. The contents of the token are in `yylval'.
3048 Updates yylval and returns the new token type. BLOCK is the block
3049 in which lookups start; this can be NULL to mean the global scope.
3050 IS_QUOTED_NAME is non-zero if the name token was originally quoted
3051 in single quotes. IS_AFTER_STRUCTOP is true if this name follows
3052 a structure operator -- either '.' or ARROW */
3055 classify_name (struct parser_state *par_state, const struct block *block,
3056 bool is_quoted_name, bool is_after_structop)
3058 struct block_symbol bsym;
3059 struct field_of_this_result is_a_field_of_this;
3061 std::string copy = copy_name (yylval.sval);
3063 /* Initialize this in case we *don't* use it in this call; that way
3064 we can refer to it unconditionally below. */
3065 memset (&is_a_field_of_this, 0, sizeof (is_a_field_of_this));
3067 bsym = lookup_symbol (copy.c_str (), block, VAR_DOMAIN,
3068 par_state->language ()->name_of_this ()
3069 ? &is_a_field_of_this : NULL);
3071 if (bsym.symbol && SYMBOL_CLASS (bsym.symbol) == LOC_BLOCK)
3073 yylval.ssym.sym = bsym;
3074 yylval.ssym.is_a_field_of_this = is_a_field_of_this.type != NULL;
3077 else if (!bsym.symbol)
3079 /* If we found a field of 'this', we might have erroneously
3080 found a constructor where we wanted a type name. Handle this
3081 case by noticing that we found a constructor and then look up
3082 the type tag instead. */
3083 if (is_a_field_of_this.type != NULL
3084 && is_a_field_of_this.fn_field != NULL
3085 && TYPE_FN_FIELD_CONSTRUCTOR (is_a_field_of_this.fn_field->fn_fields,
3088 struct field_of_this_result inner_is_a_field_of_this;
3090 bsym = lookup_symbol (copy.c_str (), block, STRUCT_DOMAIN,
3091 &inner_is_a_field_of_this);
3092 if (bsym.symbol != NULL)
3094 yylval.tsym.type = SYMBOL_TYPE (bsym.symbol);
3099 /* If we found a field on the "this" object, or we are looking
3100 up a field on a struct, then we want to prefer it over a
3101 filename. However, if the name was quoted, then it is better
3102 to check for a filename or a block, since this is the only
3103 way the user has of requiring the extension to be used. */
3104 if ((is_a_field_of_this.type == NULL && !is_after_structop)
3107 /* See if it's a file name. */
3108 struct symtab *symtab;
3110 symtab = lookup_symtab (copy.c_str ());
3113 yylval.bval = BLOCKVECTOR_BLOCK (SYMTAB_BLOCKVECTOR (symtab),
3120 if (bsym.symbol && SYMBOL_CLASS (bsym.symbol) == LOC_TYPEDEF)
3122 yylval.tsym.type = SYMBOL_TYPE (bsym.symbol);
3126 /* See if it's an ObjC classname. */
3127 if (par_state->language ()->la_language == language_objc && !bsym.symbol)
3129 CORE_ADDR Class = lookup_objc_class (par_state->gdbarch (),
3135 yylval.theclass.theclass = Class;
3136 sym = lookup_struct_typedef (copy.c_str (),
3137 par_state->expression_context_block, 1);
3139 yylval.theclass.type = SYMBOL_TYPE (sym);
3144 /* Input names that aren't symbols but ARE valid hex numbers, when
3145 the input radix permits them, can be names or numbers depending
3146 on the parse. Note we support radixes > 16 here. */
3148 && ((copy[0] >= 'a' && copy[0] < 'a' + input_radix - 10)
3149 || (copy[0] >= 'A' && copy[0] < 'A' + input_radix - 10)))
3151 YYSTYPE newlval; /* Its value is ignored. */
3152 int hextype = parse_number (par_state, copy.c_str (), yylval.sval.length,
3157 yylval.ssym.sym = bsym;
3158 yylval.ssym.is_a_field_of_this = is_a_field_of_this.type != NULL;
3163 /* Any other kind of symbol */
3164 yylval.ssym.sym = bsym;
3165 yylval.ssym.is_a_field_of_this = is_a_field_of_this.type != NULL;
3167 if (bsym.symbol == NULL
3168 && par_state->language ()->la_language == language_cplus
3169 && is_a_field_of_this.type == NULL
3170 && lookup_minimal_symbol (copy.c_str (), NULL, NULL).minsym == NULL)
3171 return UNKNOWN_CPP_NAME;
3176 /* Like classify_name, but used by the inner loop of the lexer, when a
3177 name might have already been seen. CONTEXT is the context type, or
3178 NULL if this is the first component of a name. */
3181 classify_inner_name (struct parser_state *par_state,
3182 const struct block *block, struct type *context)
3186 if (context == NULL)
3187 return classify_name (par_state, block, false, false);
3189 type = check_typedef (context);
3190 if (!type_aggregate_p (type))
3193 std::string copy = copy_name (yylval.ssym.stoken);
3194 /* N.B. We assume the symbol can only be in VAR_DOMAIN. */
3195 yylval.ssym.sym = cp_lookup_nested_symbol (type, copy.c_str (), block,
3198 /* If no symbol was found, search for a matching base class named
3199 COPY. This will allow users to enter qualified names of class members
3200 relative to the `this' pointer. */
3201 if (yylval.ssym.sym.symbol == NULL)
3203 struct type *base_type = cp_find_type_baseclass_by_name (type,
3206 if (base_type != NULL)
3208 yylval.tsym.type = base_type;
3215 switch (SYMBOL_CLASS (yylval.ssym.sym.symbol))
3219 /* cp_lookup_nested_symbol might have accidentally found a constructor
3220 named COPY when we really wanted a base class of the same name.
3221 Double-check this case by looking for a base class. */
3223 struct type *base_type
3224 = cp_find_type_baseclass_by_name (type, copy.c_str ());
3226 if (base_type != NULL)
3228 yylval.tsym.type = base_type;
3235 yylval.tsym.type = SYMBOL_TYPE (yylval.ssym.sym.symbol);
3241 internal_error (__FILE__, __LINE__, _("not reached"));
3244 /* The outer level of a two-level lexer. This calls the inner lexer
3245 to return tokens. It then either returns these tokens, or
3246 aggregates them into a larger token. This lets us work around a
3247 problem in our parsing approach, where the parser could not
3248 distinguish between qualified names and qualified types at the
3251 This approach is still not ideal, because it mishandles template
3252 types. See the comment in lex_one_token for an example. However,
3253 this is still an improvement over the earlier approach, and will
3254 suffice until we move to better parsing technology. */
3259 token_and_value current;
3260 int first_was_coloncolon, last_was_coloncolon;
3261 struct type *context_type = NULL;
3262 int last_to_examine, next_to_examine, checkpoint;
3263 const struct block *search_block;
3264 bool is_quoted_name, last_lex_was_structop;
3266 if (popping && !token_fifo.empty ())
3270 last_lex_was_structop = last_was_structop;
3272 /* Read the first token and decide what to do. Most of the
3273 subsequent code is C++-only; but also depends on seeing a "::" or
3275 current.token = lex_one_token (pstate, &is_quoted_name);
3276 if (current.token == NAME)
3277 current.token = classify_name (pstate, pstate->expression_context_block,
3278 is_quoted_name, last_lex_was_structop);
3279 if (pstate->language ()->la_language != language_cplus
3280 || (current.token != TYPENAME && current.token != COLONCOLON
3281 && current.token != FILENAME))
3282 return current.token;
3284 /* Read any sequence of alternating "::" and name-like tokens into
3286 current.value = yylval;
3287 token_fifo.push_back (current);
3288 last_was_coloncolon = current.token == COLONCOLON;
3293 /* We ignore quoted names other than the very first one.
3294 Subsequent ones do not have any special meaning. */
3295 current.token = lex_one_token (pstate, &ignore);
3296 current.value = yylval;
3297 token_fifo.push_back (current);
3299 if ((last_was_coloncolon && current.token != NAME)
3300 || (!last_was_coloncolon && current.token != COLONCOLON))
3302 last_was_coloncolon = !last_was_coloncolon;
3306 /* We always read one extra token, so compute the number of tokens
3307 to examine accordingly. */
3308 last_to_examine = token_fifo.size () - 2;
3309 next_to_examine = 0;
3311 current = token_fifo[next_to_examine];
3314 name_obstack.clear ();
3316 if (current.token == FILENAME)
3317 search_block = current.value.bval;
3318 else if (current.token == COLONCOLON)
3319 search_block = NULL;
3322 gdb_assert (current.token == TYPENAME);
3323 search_block = pstate->expression_context_block;
3324 obstack_grow (&name_obstack, current.value.sval.ptr,
3325 current.value.sval.length);
3326 context_type = current.value.tsym.type;
3330 first_was_coloncolon = current.token == COLONCOLON;
3331 last_was_coloncolon = first_was_coloncolon;
3333 while (next_to_examine <= last_to_examine)
3335 token_and_value next;
3337 next = token_fifo[next_to_examine];
3340 if (next.token == NAME && last_was_coloncolon)
3344 yylval = next.value;
3345 classification = classify_inner_name (pstate, search_block,
3347 /* We keep going until we either run out of names, or until
3348 we have a qualified name which is not a type. */
3349 if (classification != TYPENAME && classification != NAME)
3352 /* Accept up to this token. */
3353 checkpoint = next_to_examine;
3355 /* Update the partial name we are constructing. */
3356 if (context_type != NULL)
3358 /* We don't want to put a leading "::" into the name. */
3359 obstack_grow_str (&name_obstack, "::");
3361 obstack_grow (&name_obstack, next.value.sval.ptr,
3362 next.value.sval.length);
3364 yylval.sval.ptr = (const char *) obstack_base (&name_obstack);
3365 yylval.sval.length = obstack_object_size (&name_obstack);
3366 current.value = yylval;
3367 current.token = classification;
3369 last_was_coloncolon = 0;
3371 if (classification == NAME)
3374 context_type = yylval.tsym.type;
3376 else if (next.token == COLONCOLON && !last_was_coloncolon)
3377 last_was_coloncolon = 1;
3380 /* We've reached the end of the name. */
3385 /* If we have a replacement token, install it as the first token in
3386 the FIFO, and delete the other constituent tokens. */
3389 current.value.sval.ptr
3390 = obstack_strndup (&cpstate->expansion_obstack,
3391 current.value.sval.ptr,
3392 current.value.sval.length);
3394 token_fifo[0] = current;
3396 token_fifo.erase (token_fifo.begin () + 1,
3397 token_fifo.begin () + checkpoint);
3401 current = token_fifo[0];
3402 token_fifo.erase (token_fifo.begin ());
3403 yylval = current.value;
3404 return current.token;
3408 c_parse (struct parser_state *par_state)
3410 /* Setting up the parser state. */
3411 scoped_restore pstate_restore = make_scoped_restore (&pstate);
3412 gdb_assert (par_state != NULL);
3415 c_parse_state cstate;
3416 scoped_restore cstate_restore = make_scoped_restore (&cpstate, &cstate);
3418 gdb::unique_xmalloc_ptr<struct macro_scope> macro_scope;
3420 if (par_state->expression_context_block)
3422 = sal_macro_scope (find_pc_line (par_state->expression_context_pc, 0));
3424 macro_scope = default_macro_scope ();
3426 macro_scope = user_macro_scope ();
3428 scoped_restore restore_macro_scope
3429 = make_scoped_restore (&expression_macro_scope, macro_scope.get ());
3431 scoped_restore restore_yydebug = make_scoped_restore (&yydebug,
3434 /* Initialize some state used by the lexer. */
3435 last_was_structop = false;
3436 saw_name_at_eof = 0;
3439 token_fifo.clear ();
3441 name_obstack.clear ();
3443 int result = yyparse ();
3445 pstate->set_operation (pstate->pop ());
3451 /* This is called via the YYPRINT macro when parser debugging is
3452 enabled. It prints a token's value. */
3455 c_print_token (FILE *file, int type, YYSTYPE value)
3460 parser_fprintf (file, "typed_val_int<%s, %s>",
3461 TYPE_SAFE_NAME (value.typed_val_int.type),
3462 pulongest (value.typed_val_int.val));
3468 char *copy = (char *) alloca (value.tsval.length + 1);
3470 memcpy (copy, value.tsval.ptr, value.tsval.length);
3471 copy[value.tsval.length] = '\0';
3473 parser_fprintf (file, "tsval<type=%d, %s>", value.tsval.type, copy);
3478 case DOLLAR_VARIABLE:
3479 parser_fprintf (file, "sval<%s>", copy_name (value.sval).c_str ());
3483 parser_fprintf (file, "tsym<type=%s, name=%s>",
3484 TYPE_SAFE_NAME (value.tsym.type),
3485 copy_name (value.tsym.stoken).c_str ());
3489 case UNKNOWN_CPP_NAME:
3492 parser_fprintf (file, "ssym<name=%s, sym=%s, field_of_this=%d>",
3493 copy_name (value.ssym.stoken).c_str (),
3494 (value.ssym.sym.symbol == NULL
3495 ? "(null)" : value.ssym.sym.symbol->print_name ()),
3496 value.ssym.is_a_field_of_this);
3500 parser_fprintf (file, "bval<%s>", host_address_to_string (value.bval));
3508 yyerror (const char *msg)
3510 if (pstate->prev_lexptr)
3511 pstate->lexptr = pstate->prev_lexptr;
3513 error (_("A %s in expression, near `%s'."), msg, pstate->lexptr);