1 /* YACC parser for C expressions, for GDB.
2 Copyright (C) 1986, 1989, 1990, 1991, 1992, 1993, 1994, 1995, 1996, 1997,
3 1998, 1999, 2000, 2003, 2004, 2006, 2007 Free Software Foundation, Inc.
5 This file is part of GDB.
7 This program is free software; you can redistribute it and/or modify
8 it under the terms of the GNU General Public License as published by
9 the Free Software Foundation; either version 2 of the License, or
10 (at your option) any later version.
12 This program is distributed in the hope that it will be useful,
13 but WITHOUT ANY WARRANTY; without even the implied warranty of
14 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
15 GNU General Public License for more details.
17 You should have received a copy of the GNU General Public License
18 along with this program; if not, write to the Free Software
19 Foundation, Inc., 51 Franklin Street, Fifth Floor,
20 Boston, MA 02110-1301, USA. */
22 /* Parse a C expression from text in a string,
23 and return the result as a struct expression pointer.
24 That structure contains arithmetic operations in reverse polish,
25 with constants represented by operations that are followed by special data.
26 See expression.h for the details of the format.
27 What is important here is that it can be built up sequentially
28 during the process of parsing; the lower levels of the tree always
29 come first in the result.
31 Note that malloc's and realloc's in this file are transformed to
32 xmalloc and xrealloc respectively by the same sed command in the
33 makefile that remaps any other malloc/realloc inserted by the parser
34 generator. Doing this with #defines and trying to control the interaction
35 with include files (<malloc.h> and <stdlib.h> for example) just became
36 too messy, particularly when such includes can be inserted at random
37 times by the parser generator. */
42 #include "gdb_string.h"
44 #include "expression.h"
46 #include "parser-defs.h"
49 #include "bfd.h" /* Required by objfiles.h. */
50 #include "symfile.h" /* Required by objfiles.h. */
51 #include "objfiles.h" /* For have_full_symbols and have_partial_symbols */
54 #include "cp-support.h"
56 /* Remap normal yacc parser interface names (yyparse, yylex, yyerror, etc),
57 as well as gratuitiously global symbol names, so we can have multiple
58 yacc generated parsers in gdb. Note that these are only the variables
59 produced by yacc. If other parser generators (bison, byacc, etc) produce
60 additional global names that conflict at link time, then those parser
61 generators need to be fixed instead of adding those names to this list. */
63 #define yymaxdepth c_maxdepth
64 #define yyparse c_parse
66 #define yyerror c_error
69 #define yydebug c_debug
78 #define yyerrflag c_errflag
79 #define yynerrs c_nerrs
84 #define yystate c_state
90 #define yyreds c_reds /* With YYDEBUG defined */
91 #define yytoks c_toks /* With YYDEBUG defined */
92 #define yyname c_name /* With YYDEBUG defined */
93 #define yyrule c_rule /* With YYDEBUG defined */
96 #define yydefred c_yydefred
97 #define yydgoto c_yydgoto
98 #define yysindex c_yysindex
99 #define yyrindex c_yyrindex
100 #define yygindex c_yygindex
101 #define yytable c_yytable
102 #define yycheck c_yycheck
105 #define YYDEBUG 1 /* Default to yydebug support */
108 #define YYFPRINTF parser_fprintf
112 static int yylex (void);
114 void yyerror (char *);
118 /* Although the yacc "value" of an expression is not used,
119 since the result is stored in the structure being created,
120 other node types do have values. */
137 struct symtoken ssym;
140 enum exp_opcode opcode;
141 struct internalvar *ivar;
148 /* YYSTYPE gets defined by %union */
149 static int parse_number (char *, int, int, YYSTYPE *);
152 %type <voidval> exp exp1 type_exp start variable qualified_name lcurly
154 %type <tval> type typebase qualified_type
155 %type <tvec> nonempty_typelist
156 /* %type <bval> block */
158 /* Fancy type parsing. */
159 %type <voidval> func_mod direct_abs_decl abs_decl
161 %type <lval> array_mod
163 %token <typed_val_int> INT
164 %token <typed_val_float> FLOAT
166 /* Both NAME and TYPENAME tokens represent symbols in the input,
167 and both convey their data as strings.
168 But a TYPENAME is a string that happens to be defined as a typedef
169 or builtin type name (such as int or char)
170 and a NAME is any other symbol.
171 Contexts where this distinction is not important can use the
172 nonterminal "name", which matches either NAME or TYPENAME. */
175 %token <ssym> NAME /* BLOCKNAME defined below to give it higher precedence. */
176 %token <tsym> TYPENAME
178 %type <ssym> name_not_typename
179 %type <tsym> typename
181 /* A NAME_OR_INT is a symbol which is not known in the symbol table,
182 but which would parse as a valid number in the current input radix.
183 E.g. "c" when input_radix==16. Depending on the parse, it will be
184 turned into a name or into a number. */
186 %token <ssym> NAME_OR_INT
188 %token STRUCT CLASS UNION ENUM SIZEOF UNSIGNED COLONCOLON
192 /* Special type cases, put in to allow the parser to distinguish different
194 %token SIGNED_KEYWORD LONG SHORT INT_KEYWORD CONST_KEYWORD VOLATILE_KEYWORD DOUBLE_KEYWORD
196 %token <voidval> VARIABLE
198 %token <opcode> ASSIGN_MODIFY
207 %right '=' ASSIGN_MODIFY
215 %left '<' '>' LEQ GEQ
220 %right UNARY INCREMENT DECREMENT
221 %right ARROW '.' '[' '('
222 %token <ssym> BLOCKNAME
223 %token <bval> FILENAME
235 { write_exp_elt_opcode(OP_TYPE);
236 write_exp_elt_type($1);
237 write_exp_elt_opcode(OP_TYPE);}
240 /* Expressions, including the comma operator. */
243 { write_exp_elt_opcode (BINOP_COMMA); }
246 /* Expressions, not including the comma operator. */
247 exp : '*' exp %prec UNARY
248 { write_exp_elt_opcode (UNOP_IND); }
251 exp : '&' exp %prec UNARY
252 { write_exp_elt_opcode (UNOP_ADDR); }
255 exp : '-' exp %prec UNARY
256 { write_exp_elt_opcode (UNOP_NEG); }
259 exp : '+' exp %prec UNARY
260 { write_exp_elt_opcode (UNOP_PLUS); }
263 exp : '!' exp %prec UNARY
264 { write_exp_elt_opcode (UNOP_LOGICAL_NOT); }
267 exp : '~' exp %prec UNARY
268 { write_exp_elt_opcode (UNOP_COMPLEMENT); }
271 exp : INCREMENT exp %prec UNARY
272 { write_exp_elt_opcode (UNOP_PREINCREMENT); }
275 exp : DECREMENT exp %prec UNARY
276 { write_exp_elt_opcode (UNOP_PREDECREMENT); }
279 exp : exp INCREMENT %prec UNARY
280 { write_exp_elt_opcode (UNOP_POSTINCREMENT); }
283 exp : exp DECREMENT %prec UNARY
284 { write_exp_elt_opcode (UNOP_POSTDECREMENT); }
287 exp : SIZEOF exp %prec UNARY
288 { write_exp_elt_opcode (UNOP_SIZEOF); }
292 { write_exp_elt_opcode (STRUCTOP_PTR);
293 write_exp_string ($3);
294 write_exp_elt_opcode (STRUCTOP_PTR); }
297 exp : exp ARROW qualified_name
298 { /* exp->type::name becomes exp->*(&type::name) */
299 /* Note: this doesn't work if name is a
300 static member! FIXME */
301 write_exp_elt_opcode (UNOP_ADDR);
302 write_exp_elt_opcode (STRUCTOP_MPTR); }
305 exp : exp ARROW '*' exp
306 { write_exp_elt_opcode (STRUCTOP_MPTR); }
310 { write_exp_elt_opcode (STRUCTOP_STRUCT);
311 write_exp_string ($3);
312 write_exp_elt_opcode (STRUCTOP_STRUCT); }
315 exp : exp '.' qualified_name
316 { /* exp.type::name becomes exp.*(&type::name) */
317 /* Note: this doesn't work if name is a
318 static member! FIXME */
319 write_exp_elt_opcode (UNOP_ADDR);
320 write_exp_elt_opcode (STRUCTOP_MEMBER); }
323 exp : exp '.' '*' exp
324 { write_exp_elt_opcode (STRUCTOP_MEMBER); }
327 exp : exp '[' exp1 ']'
328 { write_exp_elt_opcode (BINOP_SUBSCRIPT); }
332 /* This is to save the value of arglist_len
333 being accumulated by an outer function call. */
334 { start_arglist (); }
335 arglist ')' %prec ARROW
336 { write_exp_elt_opcode (OP_FUNCALL);
337 write_exp_elt_longcst ((LONGEST) end_arglist ());
338 write_exp_elt_opcode (OP_FUNCALL); }
342 { start_arglist (); }
352 arglist : arglist ',' exp %prec ABOVE_COMMA
357 { $$ = end_arglist () - 1; }
359 exp : lcurly arglist rcurly %prec ARROW
360 { write_exp_elt_opcode (OP_ARRAY);
361 write_exp_elt_longcst ((LONGEST) 0);
362 write_exp_elt_longcst ((LONGEST) $3);
363 write_exp_elt_opcode (OP_ARRAY); }
366 exp : lcurly type rcurly exp %prec UNARY
367 { write_exp_elt_opcode (UNOP_MEMVAL);
368 write_exp_elt_type ($2);
369 write_exp_elt_opcode (UNOP_MEMVAL); }
372 exp : '(' type ')' exp %prec UNARY
373 { write_exp_elt_opcode (UNOP_CAST);
374 write_exp_elt_type ($2);
375 write_exp_elt_opcode (UNOP_CAST); }
382 /* Binary operators in order of decreasing precedence. */
385 { write_exp_elt_opcode (BINOP_REPEAT); }
389 { write_exp_elt_opcode (BINOP_MUL); }
393 { write_exp_elt_opcode (BINOP_DIV); }
397 { write_exp_elt_opcode (BINOP_REM); }
401 { write_exp_elt_opcode (BINOP_ADD); }
405 { write_exp_elt_opcode (BINOP_SUB); }
409 { write_exp_elt_opcode (BINOP_LSH); }
413 { write_exp_elt_opcode (BINOP_RSH); }
417 { write_exp_elt_opcode (BINOP_EQUAL); }
420 exp : exp NOTEQUAL exp
421 { write_exp_elt_opcode (BINOP_NOTEQUAL); }
425 { write_exp_elt_opcode (BINOP_LEQ); }
429 { write_exp_elt_opcode (BINOP_GEQ); }
433 { write_exp_elt_opcode (BINOP_LESS); }
437 { write_exp_elt_opcode (BINOP_GTR); }
441 { write_exp_elt_opcode (BINOP_BITWISE_AND); }
445 { write_exp_elt_opcode (BINOP_BITWISE_XOR); }
449 { write_exp_elt_opcode (BINOP_BITWISE_IOR); }
453 { write_exp_elt_opcode (BINOP_LOGICAL_AND); }
457 { write_exp_elt_opcode (BINOP_LOGICAL_OR); }
460 exp : exp '?' exp ':' exp %prec '?'
461 { write_exp_elt_opcode (TERNOP_COND); }
465 { write_exp_elt_opcode (BINOP_ASSIGN); }
468 exp : exp ASSIGN_MODIFY exp
469 { write_exp_elt_opcode (BINOP_ASSIGN_MODIFY);
470 write_exp_elt_opcode ($2);
471 write_exp_elt_opcode (BINOP_ASSIGN_MODIFY); }
475 { write_exp_elt_opcode (OP_LONG);
476 write_exp_elt_type ($1.type);
477 write_exp_elt_longcst ((LONGEST)($1.val));
478 write_exp_elt_opcode (OP_LONG); }
483 parse_number ($1.stoken.ptr, $1.stoken.length, 0, &val);
484 write_exp_elt_opcode (OP_LONG);
485 write_exp_elt_type (val.typed_val_int.type);
486 write_exp_elt_longcst ((LONGEST)val.typed_val_int.val);
487 write_exp_elt_opcode (OP_LONG);
493 { write_exp_elt_opcode (OP_DOUBLE);
494 write_exp_elt_type ($1.type);
495 write_exp_elt_dblcst ($1.dval);
496 write_exp_elt_opcode (OP_DOUBLE); }
503 /* Already written by write_dollar_variable. */
506 exp : SIZEOF '(' type ')' %prec UNARY
507 { write_exp_elt_opcode (OP_LONG);
508 write_exp_elt_type (builtin_type (current_gdbarch)->builtin_int);
510 write_exp_elt_longcst ((LONGEST) TYPE_LENGTH ($3));
511 write_exp_elt_opcode (OP_LONG); }
515 { /* C strings are converted into array constants with
516 an explicit null byte added at the end. Thus
517 the array upper bound is the string length.
518 There is no such thing in C as a completely empty
520 char *sp = $1.ptr; int count = $1.length;
523 write_exp_elt_opcode (OP_LONG);
524 write_exp_elt_type (builtin_type (current_gdbarch)->builtin_char);
525 write_exp_elt_longcst ((LONGEST)(*sp++));
526 write_exp_elt_opcode (OP_LONG);
528 write_exp_elt_opcode (OP_LONG);
529 write_exp_elt_type (builtin_type (current_gdbarch)->builtin_char);
530 write_exp_elt_longcst ((LONGEST)'\0');
531 write_exp_elt_opcode (OP_LONG);
532 write_exp_elt_opcode (OP_ARRAY);
533 write_exp_elt_longcst ((LONGEST) 0);
534 write_exp_elt_longcst ((LONGEST) ($1.length));
535 write_exp_elt_opcode (OP_ARRAY); }
540 { write_exp_elt_opcode (OP_LONG);
541 write_exp_elt_type (builtin_type (current_gdbarch)->builtin_bool);
542 write_exp_elt_longcst ((LONGEST) 1);
543 write_exp_elt_opcode (OP_LONG); }
547 { write_exp_elt_opcode (OP_LONG);
548 write_exp_elt_type (builtin_type (current_gdbarch)->builtin_bool);
549 write_exp_elt_longcst ((LONGEST) 0);
550 write_exp_elt_opcode (OP_LONG); }
558 $$ = SYMBOL_BLOCK_VALUE ($1.sym);
560 error ("No file or function \"%s\".",
561 copy_name ($1.stoken));
569 block : block COLONCOLON name
571 = lookup_symbol (copy_name ($3), $1,
572 VAR_DOMAIN, (int *) NULL,
573 (struct symtab **) NULL);
574 if (!tem || SYMBOL_CLASS (tem) != LOC_BLOCK)
575 error ("No function \"%s\" in specified context.",
577 $$ = SYMBOL_BLOCK_VALUE (tem); }
580 variable: block COLONCOLON name
581 { struct symbol *sym;
582 sym = lookup_symbol (copy_name ($3), $1,
583 VAR_DOMAIN, (int *) NULL,
584 (struct symtab **) NULL);
586 error ("No symbol \"%s\" in specified context.",
589 write_exp_elt_opcode (OP_VAR_VALUE);
590 /* block_found is set by lookup_symbol. */
591 write_exp_elt_block (block_found);
592 write_exp_elt_sym (sym);
593 write_exp_elt_opcode (OP_VAR_VALUE); }
596 qualified_name: typebase COLONCOLON name
598 struct type *type = $1;
599 if (TYPE_CODE (type) != TYPE_CODE_STRUCT
600 && TYPE_CODE (type) != TYPE_CODE_UNION
601 && TYPE_CODE (type) != TYPE_CODE_NAMESPACE)
602 error ("`%s' is not defined as an aggregate type.",
605 write_exp_elt_opcode (OP_SCOPE);
606 write_exp_elt_type (type);
607 write_exp_string ($3);
608 write_exp_elt_opcode (OP_SCOPE);
610 | typebase COLONCOLON '~' name
612 struct type *type = $1;
613 struct stoken tmp_token;
614 if (TYPE_CODE (type) != TYPE_CODE_STRUCT
615 && TYPE_CODE (type) != TYPE_CODE_UNION
616 && TYPE_CODE (type) != TYPE_CODE_NAMESPACE)
617 error ("`%s' is not defined as an aggregate type.",
620 tmp_token.ptr = (char*) alloca ($4.length + 2);
621 tmp_token.length = $4.length + 1;
622 tmp_token.ptr[0] = '~';
623 memcpy (tmp_token.ptr+1, $4.ptr, $4.length);
624 tmp_token.ptr[tmp_token.length] = 0;
626 /* Check for valid destructor name. */
627 destructor_name_p (tmp_token.ptr, type);
628 write_exp_elt_opcode (OP_SCOPE);
629 write_exp_elt_type (type);
630 write_exp_string (tmp_token);
631 write_exp_elt_opcode (OP_SCOPE);
635 variable: qualified_name
638 char *name = copy_name ($2);
640 struct minimal_symbol *msymbol;
643 lookup_symbol (name, (const struct block *) NULL,
644 VAR_DOMAIN, (int *) NULL,
645 (struct symtab **) NULL);
648 write_exp_elt_opcode (OP_VAR_VALUE);
649 write_exp_elt_block (NULL);
650 write_exp_elt_sym (sym);
651 write_exp_elt_opcode (OP_VAR_VALUE);
655 msymbol = lookup_minimal_symbol (name, NULL, NULL);
658 write_exp_msymbol (msymbol,
659 lookup_function_type (builtin_type (current_gdbarch)->builtin_int),
660 builtin_type (current_gdbarch)->builtin_int);
663 if (!have_full_symbols () && !have_partial_symbols ())
664 error ("No symbol table is loaded. Use the \"file\" command.");
666 error ("No symbol \"%s\" in current context.", name);
670 variable: name_not_typename
671 { struct symbol *sym = $1.sym;
675 if (symbol_read_needs_frame (sym))
677 if (innermost_block == 0 ||
678 contained_in (block_found,
680 innermost_block = block_found;
683 write_exp_elt_opcode (OP_VAR_VALUE);
684 /* We want to use the selected frame, not
685 another more inner frame which happens to
686 be in the same block. */
687 write_exp_elt_block (NULL);
688 write_exp_elt_sym (sym);
689 write_exp_elt_opcode (OP_VAR_VALUE);
691 else if ($1.is_a_field_of_this)
693 /* C++: it hangs off of `this'. Must
694 not inadvertently convert from a method call
696 if (innermost_block == 0 ||
697 contained_in (block_found, innermost_block))
698 innermost_block = block_found;
699 write_exp_elt_opcode (OP_THIS);
700 write_exp_elt_opcode (OP_THIS);
701 write_exp_elt_opcode (STRUCTOP_PTR);
702 write_exp_string ($1.stoken);
703 write_exp_elt_opcode (STRUCTOP_PTR);
707 struct minimal_symbol *msymbol;
708 char *arg = copy_name ($1.stoken);
711 lookup_minimal_symbol (arg, NULL, NULL);
714 write_exp_msymbol (msymbol,
715 lookup_function_type (builtin_type (current_gdbarch)->builtin_int),
716 builtin_type (current_gdbarch)->builtin_int);
718 else if (!have_full_symbols () && !have_partial_symbols ())
719 error ("No symbol table is loaded. Use the \"file\" command.");
721 error ("No symbol \"%s\" in current context.",
722 copy_name ($1.stoken));
727 space_identifier : '@' NAME
728 { push_type_address_space (copy_name ($2.stoken));
729 push_type (tp_space_identifier);
733 const_or_volatile: const_or_volatile_noopt
737 cv_with_space_id : const_or_volatile space_identifier const_or_volatile
740 const_or_volatile_or_space_identifier_noopt: cv_with_space_id
741 | const_or_volatile_noopt
744 const_or_volatile_or_space_identifier:
745 const_or_volatile_or_space_identifier_noopt
750 { push_type (tp_pointer); $$ = 0; }
752 { push_type (tp_pointer); $$ = $2; }
754 { push_type (tp_reference); $$ = 0; }
756 { push_type (tp_reference); $$ = $2; }
760 direct_abs_decl: '(' abs_decl ')'
762 | direct_abs_decl array_mod
765 push_type (tp_array);
770 push_type (tp_array);
774 | direct_abs_decl func_mod
775 { push_type (tp_function); }
777 { push_type (tp_function); }
788 | '(' nonempty_typelist ')'
789 { free ($2); $$ = 0; }
792 /* We used to try to recognize pointer to member types here, but
793 that didn't work (shift/reduce conflicts meant that these rules never
794 got executed). The problem is that
795 int (foo::bar::baz::bizzle)
796 is a function type but
797 int (foo::bar::baz::bizzle::*)
798 is a pointer to member type. Stroustrup loses again! */
803 typebase /* Implements (approximately): (type-qualifier)* type-specifier */
807 { $$ = builtin_type (current_gdbarch)->builtin_int; }
809 { $$ = builtin_type (current_gdbarch)->builtin_long; }
811 { $$ = builtin_type (current_gdbarch)->builtin_short; }
813 { $$ = builtin_type (current_gdbarch)->builtin_long; }
814 | LONG SIGNED_KEYWORD INT_KEYWORD
815 { $$ = builtin_type (current_gdbarch)->builtin_long; }
816 | LONG SIGNED_KEYWORD
817 { $$ = builtin_type (current_gdbarch)->builtin_long; }
818 | SIGNED_KEYWORD LONG INT_KEYWORD
819 { $$ = builtin_type (current_gdbarch)->builtin_long; }
820 | UNSIGNED LONG INT_KEYWORD
821 { $$ = builtin_type (current_gdbarch)->builtin_unsigned_long; }
822 | LONG UNSIGNED INT_KEYWORD
823 { $$ = builtin_type (current_gdbarch)->builtin_unsigned_long; }
825 { $$ = builtin_type (current_gdbarch)->builtin_unsigned_long; }
827 { $$ = builtin_type (current_gdbarch)->builtin_long_long; }
828 | LONG LONG INT_KEYWORD
829 { $$ = builtin_type (current_gdbarch)->builtin_long_long; }
830 | LONG LONG SIGNED_KEYWORD INT_KEYWORD
831 { $$ = builtin_type (current_gdbarch)->builtin_long_long; }
832 | LONG LONG SIGNED_KEYWORD
833 { $$ = builtin_type (current_gdbarch)->builtin_long_long; }
834 | SIGNED_KEYWORD LONG LONG
835 { $$ = builtin_type (current_gdbarch)->builtin_long_long; }
836 | SIGNED_KEYWORD LONG LONG INT_KEYWORD
837 { $$ = builtin_type (current_gdbarch)->builtin_long_long; }
839 { $$ = builtin_type (current_gdbarch)->builtin_unsigned_long_long; }
840 | UNSIGNED LONG LONG INT_KEYWORD
841 { $$ = builtin_type (current_gdbarch)->builtin_unsigned_long_long; }
843 { $$ = builtin_type (current_gdbarch)->builtin_unsigned_long_long; }
844 | LONG LONG UNSIGNED INT_KEYWORD
845 { $$ = builtin_type (current_gdbarch)->builtin_unsigned_long_long; }
847 { $$ = builtin_type (current_gdbarch)->builtin_short; }
848 | SHORT SIGNED_KEYWORD INT_KEYWORD
849 { $$ = builtin_type (current_gdbarch)->builtin_short; }
850 | SHORT SIGNED_KEYWORD
851 { $$ = builtin_type (current_gdbarch)->builtin_short; }
852 | UNSIGNED SHORT INT_KEYWORD
853 { $$ = builtin_type (current_gdbarch)->builtin_unsigned_short; }
855 { $$ = builtin_type (current_gdbarch)->builtin_unsigned_short; }
856 | SHORT UNSIGNED INT_KEYWORD
857 { $$ = builtin_type (current_gdbarch)->builtin_unsigned_short; }
859 { $$ = builtin_type (current_gdbarch)->builtin_double; }
860 | LONG DOUBLE_KEYWORD
861 { $$ = builtin_type (current_gdbarch)->builtin_long_double; }
863 { $$ = lookup_struct (copy_name ($2),
864 expression_context_block); }
866 { $$ = lookup_struct (copy_name ($2),
867 expression_context_block); }
869 { $$ = lookup_union (copy_name ($2),
870 expression_context_block); }
872 { $$ = lookup_enum (copy_name ($2),
873 expression_context_block); }
875 { $$ = lookup_unsigned_typename (TYPE_NAME($2.type)); }
877 { $$ = builtin_type (current_gdbarch)->builtin_unsigned_int; }
878 | SIGNED_KEYWORD typename
879 { $$ = lookup_signed_typename (TYPE_NAME($2.type)); }
881 { $$ = builtin_type (current_gdbarch)->builtin_int; }
882 /* It appears that this rule for templates is never
883 reduced; template recognition happens by lookahead
884 in the token processing code in yylex. */
885 | TEMPLATE name '<' type '>'
886 { $$ = lookup_template_type(copy_name($2), $4,
887 expression_context_block);
889 | const_or_volatile_or_space_identifier_noopt typebase
890 { $$ = follow_types ($2); }
891 | typebase const_or_volatile_or_space_identifier_noopt
892 { $$ = follow_types ($1); }
896 /* FIXME: carlton/2003-09-25: This next bit leads to lots of
897 reduce-reduce conflicts, because the parser doesn't know whether or
898 not to use qualified_name or qualified_type: the rules are
899 identical. If the parser is parsing 'A::B::x', then, when it sees
900 the second '::', it knows that the expression to the left of it has
901 to be a type, so it uses qualified_type. But if it is parsing just
902 'A::B', then it doesn't have any way of knowing which rule to use,
903 so there's a reduce-reduce conflict; it picks qualified_name, since
904 that occurs earlier in this file than qualified_type.
906 There's no good way to fix this with the grammar as it stands; as
907 far as I can tell, some of the problems arise from ambiguities that
908 GDB introduces ('start' can be either an expression or a type), but
909 some of it is inherent to the nature of C++ (you want to treat the
910 input "(FOO)" fairly differently depending on whether FOO is an
911 expression or a type, and if FOO is a complex expression, this can
912 be hard to determine at the right time). Fortunately, it works
913 pretty well in most cases. For example, if you do 'ptype A::B',
914 where A::B is a nested type, then the parser will mistakenly
915 misidentify it as an expression; but evaluate_subexp will get
916 called with 'noside' set to EVAL_AVOID_SIDE_EFFECTS, and everything
917 will work out anyways. But there are situations where the parser
918 will get confused: the most common one that I've run into is when
923 where the parser doesn't realize that A::B has to be a type until
924 it hits the first right paren, at which point it's too late. (The
925 workaround is to type "print *(('A::B' *) x)" instead.) (And
926 another solution is to fix our symbol-handling code so that the
927 user never wants to type something like that in the first place,
928 because we get all the types right without the user's help!)
930 Perhaps we could fix this by making the lexer smarter. Some of
931 this functionality used to be in the lexer, but in a way that
932 worked even less well than the current solution: that attempt
933 involved having the parser sometimes handle '::' and having the
934 lexer sometimes handle it, and without a clear division of
935 responsibility, it quickly degenerated into a big mess. Probably
936 the eventual correct solution will give more of a role to the lexer
937 (ideally via code that is shared between the lexer and
938 decode_line_1), but I'm not holding my breath waiting for somebody
939 to get around to cleaning this up... */
941 qualified_type: typebase COLONCOLON name
943 struct type *type = $1;
944 struct type *new_type;
945 char *ncopy = alloca ($3.length + 1);
947 memcpy (ncopy, $3.ptr, $3.length);
948 ncopy[$3.length] = '\0';
950 if (TYPE_CODE (type) != TYPE_CODE_STRUCT
951 && TYPE_CODE (type) != TYPE_CODE_UNION
952 && TYPE_CODE (type) != TYPE_CODE_NAMESPACE)
953 error ("`%s' is not defined as an aggregate type.",
956 new_type = cp_lookup_nested_type (type, ncopy,
957 expression_context_block);
958 if (new_type == NULL)
959 error ("No type \"%s\" within class or namespace \"%s\".",
960 ncopy, TYPE_NAME (type));
969 $$.stoken.ptr = "int";
970 $$.stoken.length = 3;
971 $$.type = builtin_type (current_gdbarch)->builtin_int;
975 $$.stoken.ptr = "long";
976 $$.stoken.length = 4;
977 $$.type = builtin_type (current_gdbarch)->builtin_long;
981 $$.stoken.ptr = "short";
982 $$.stoken.length = 5;
983 $$.type = builtin_type (current_gdbarch)->builtin_short;
989 { $$ = (struct type **) malloc (sizeof (struct type *) * 2);
990 $<ivec>$[0] = 1; /* Number of types in vector */
993 | nonempty_typelist ',' type
994 { int len = sizeof (struct type *) * (++($<ivec>1[0]) + 1);
995 $$ = (struct type **) realloc ((char *) $1, len);
996 $$[$<ivec>$[0]] = $3;
1001 | ptype const_or_volatile_or_space_identifier abs_decl const_or_volatile_or_space_identifier
1002 { $$ = follow_types ($1); }
1005 const_and_volatile: CONST_KEYWORD VOLATILE_KEYWORD
1006 | VOLATILE_KEYWORD CONST_KEYWORD
1009 const_or_volatile_noopt: const_and_volatile
1010 { push_type (tp_const);
1011 push_type (tp_volatile);
1014 { push_type (tp_const); }
1016 { push_type (tp_volatile); }
1019 name : NAME { $$ = $1.stoken; }
1020 | BLOCKNAME { $$ = $1.stoken; }
1021 | TYPENAME { $$ = $1.stoken; }
1022 | NAME_OR_INT { $$ = $1.stoken; }
1025 name_not_typename : NAME
1027 /* These would be useful if name_not_typename was useful, but it is just
1028 a fake for "variable", so these cause reduce/reduce conflicts because
1029 the parser can't tell whether NAME_OR_INT is a name_not_typename (=variable,
1030 =exp) or just an exp. If name_not_typename was ever used in an lvalue
1031 context where only a name could occur, this might be useful.
1038 /* Take care of parsing a number (anything that starts with a digit).
1039 Set yylval and return the token type; update lexptr.
1040 LEN is the number of characters in it. */
1042 /*** Needs some error checking for the float case ***/
1045 parse_number (p, len, parsed_float, putithere)
1051 /* FIXME: Shouldn't these be unsigned? We don't deal with negative values
1052 here, and we do kind of silly things like cast to unsigned. */
1059 int base = input_radix;
1062 /* Number of "L" suffixes encountered. */
1065 /* We have found a "L" or "U" suffix. */
1066 int found_suffix = 0;
1069 struct type *signed_type;
1070 struct type *unsigned_type;
1074 /* It's a float since it contains a point or an exponent. */
1075 char *s = malloc (len);
1076 int num = 0; /* number of tokens scanned by scanf */
1077 char saved_char = p[len];
1079 p[len] = 0; /* null-terminate the token */
1080 num = sscanf (p, DOUBLEST_SCAN_FORMAT "%s",
1081 &putithere->typed_val_float.dval, s);
1082 p[len] = saved_char; /* restore the input stream */
1085 putithere->typed_val_float.type =
1086 builtin_type (current_gdbarch)->builtin_double;
1090 /* See if it has any float suffix: 'f' for float, 'l' for long
1092 if (!strcasecmp (s, "f"))
1093 putithere->typed_val_float.type =
1094 builtin_type (current_gdbarch)->builtin_float;
1095 else if (!strcasecmp (s, "l"))
1096 putithere->typed_val_float.type =
1097 builtin_type (current_gdbarch)->builtin_long_double;
1105 /* Handle base-switching prefixes 0x, 0t, 0d, 0 */
1139 if (c >= 'A' && c <= 'Z')
1141 if (c != 'l' && c != 'u')
1143 if (c >= '0' && c <= '9')
1151 if (base > 10 && c >= 'a' && c <= 'f')
1155 n += i = c - 'a' + 10;
1168 return ERROR; /* Char not a digit */
1171 return ERROR; /* Invalid digit in this base */
1173 /* Portably test for overflow (only works for nonzero values, so make
1174 a second check for zero). FIXME: Can't we just make n and prevn
1175 unsigned and avoid this? */
1176 if (c != 'l' && c != 'u' && (prevn >= n) && n != 0)
1177 unsigned_p = 1; /* Try something unsigned */
1179 /* Portably test for unsigned overflow.
1180 FIXME: This check is wrong; for example it doesn't find overflow
1181 on 0x123456789 when LONGEST is 32 bits. */
1182 if (c != 'l' && c != 'u' && n != 0)
1184 if ((unsigned_p && (ULONGEST) prevn >= (ULONGEST) n))
1185 error ("Numeric constant too large.");
1190 /* An integer constant is an int, a long, or a long long. An L
1191 suffix forces it to be long; an LL suffix forces it to be long
1192 long. If not forced to a larger size, it gets the first type of
1193 the above that it fits in. To figure out whether it fits, we
1194 shift it right and see whether anything remains. Note that we
1195 can't shift sizeof (LONGEST) * HOST_CHAR_BIT bits or more in one
1196 operation, because many compilers will warn about such a shift
1197 (which always produces a zero result). Sometimes TARGET_INT_BIT
1198 or TARGET_LONG_BIT will be that big, sometimes not. To deal with
1199 the case where it is we just always shift the value more than
1200 once, with fewer bits each time. */
1202 un = (ULONGEST)n >> 2;
1204 && (un >> (TARGET_INT_BIT - 2)) == 0)
1206 high_bit = ((ULONGEST)1) << (TARGET_INT_BIT-1);
1208 /* A large decimal (not hex or octal) constant (between INT_MAX
1209 and UINT_MAX) is a long or unsigned long, according to ANSI,
1210 never an unsigned int, but this code treats it as unsigned
1211 int. This probably should be fixed. GCC gives a warning on
1214 unsigned_type = builtin_type (current_gdbarch)->builtin_unsigned_int;
1215 signed_type = builtin_type (current_gdbarch)->builtin_int;
1217 else if (long_p <= 1
1218 && (un >> (TARGET_LONG_BIT - 2)) == 0)
1220 high_bit = ((ULONGEST)1) << (TARGET_LONG_BIT-1);
1221 unsigned_type = builtin_type (current_gdbarch)->builtin_unsigned_long;
1222 signed_type = builtin_type (current_gdbarch)->builtin_long;
1227 if (sizeof (ULONGEST) * HOST_CHAR_BIT < TARGET_LONG_LONG_BIT)
1228 /* A long long does not fit in a LONGEST. */
1229 shift = (sizeof (ULONGEST) * HOST_CHAR_BIT - 1);
1231 shift = (TARGET_LONG_LONG_BIT - 1);
1232 high_bit = (ULONGEST) 1 << shift;
1233 unsigned_type = builtin_type (current_gdbarch)->builtin_unsigned_long_long;
1234 signed_type = builtin_type (current_gdbarch)->builtin_long_long;
1237 putithere->typed_val_int.val = n;
1239 /* If the high bit of the worked out type is set then this number
1240 has to be unsigned. */
1242 if (unsigned_p || (n & high_bit))
1244 putithere->typed_val_int.type = unsigned_type;
1248 putithere->typed_val_int.type = signed_type;
1258 enum exp_opcode opcode;
1261 static const struct token tokentab3[] =
1263 {">>=", ASSIGN_MODIFY, BINOP_RSH},
1264 {"<<=", ASSIGN_MODIFY, BINOP_LSH}
1267 static const struct token tokentab2[] =
1269 {"+=", ASSIGN_MODIFY, BINOP_ADD},
1270 {"-=", ASSIGN_MODIFY, BINOP_SUB},
1271 {"*=", ASSIGN_MODIFY, BINOP_MUL},
1272 {"/=", ASSIGN_MODIFY, BINOP_DIV},
1273 {"%=", ASSIGN_MODIFY, BINOP_REM},
1274 {"|=", ASSIGN_MODIFY, BINOP_BITWISE_IOR},
1275 {"&=", ASSIGN_MODIFY, BINOP_BITWISE_AND},
1276 {"^=", ASSIGN_MODIFY, BINOP_BITWISE_XOR},
1277 {"++", INCREMENT, BINOP_END},
1278 {"--", DECREMENT, BINOP_END},
1279 {"->", ARROW, BINOP_END},
1280 {"&&", ANDAND, BINOP_END},
1281 {"||", OROR, BINOP_END},
1282 {"::", COLONCOLON, BINOP_END},
1283 {"<<", LSH, BINOP_END},
1284 {">>", RSH, BINOP_END},
1285 {"==", EQUAL, BINOP_END},
1286 {"!=", NOTEQUAL, BINOP_END},
1287 {"<=", LEQ, BINOP_END},
1288 {">=", GEQ, BINOP_END}
1291 /* Read one token, getting characters through lexptr. */
1302 static char *tempbuf;
1303 static int tempbufsize;
1304 struct symbol * sym_class = NULL;
1305 char * token_string = NULL;
1306 int class_prefix = 0;
1311 /* Check if this is a macro invocation that we need to expand. */
1312 if (! scanning_macro_expansion ())
1314 char *expanded = macro_expand_next (&lexptr,
1315 expression_macro_lookup_func,
1316 expression_macro_lookup_baton);
1319 scan_macro_expansion (expanded);
1322 prev_lexptr = lexptr;
1326 /* See if it is a special token of length 3. */
1327 for (i = 0; i < sizeof tokentab3 / sizeof tokentab3[0]; i++)
1328 if (strncmp (tokstart, tokentab3[i].operator, 3) == 0)
1331 yylval.opcode = tokentab3[i].opcode;
1332 return tokentab3[i].token;
1335 /* See if it is a special token of length 2. */
1336 for (i = 0; i < sizeof tokentab2 / sizeof tokentab2[0]; i++)
1337 if (strncmp (tokstart, tokentab2[i].operator, 2) == 0)
1340 yylval.opcode = tokentab2[i].opcode;
1341 return tokentab2[i].token;
1344 switch (c = *tokstart)
1347 /* If we were just scanning the result of a macro expansion,
1348 then we need to resume scanning the original text.
1349 Otherwise, we were already scanning the original text, and
1350 we're really done. */
1351 if (scanning_macro_expansion ())
1353 finished_macro_expansion ();
1366 /* We either have a character constant ('0' or '\177' for example)
1367 or we have a quoted symbol reference ('foo(int,int)' in C++
1372 c = parse_escape (&lexptr);
1374 error ("Empty character constant.");
1375 else if (! host_char_to_target (c, &c))
1377 int toklen = lexptr - tokstart + 1;
1378 char *tok = alloca (toklen + 1);
1379 memcpy (tok, tokstart, toklen);
1381 error ("There is no character corresponding to %s in the target "
1382 "character set `%s'.", tok, target_charset ());
1385 yylval.typed_val_int.val = c;
1386 yylval.typed_val_int.type = builtin_type (current_gdbarch)->builtin_char;
1391 namelen = skip_quoted (tokstart) - tokstart;
1394 lexptr = tokstart + namelen;
1396 if (lexptr[-1] != '\'')
1397 error ("Unmatched single quote.");
1402 error ("Invalid character constant.");
1412 if (paren_depth == 0)
1419 if (comma_terminates
1421 && ! scanning_macro_expansion ())
1427 /* Might be a floating point number. */
1428 if (lexptr[1] < '0' || lexptr[1] > '9')
1429 goto symbol; /* Nope, must be a symbol. */
1430 /* FALL THRU into number case. */
1443 /* It's a number. */
1444 int got_dot = 0, got_e = 0, toktype;
1446 int hex = input_radix > 10;
1448 if (c == '0' && (p[1] == 'x' || p[1] == 'X'))
1453 else if (c == '0' && (p[1]=='t' || p[1]=='T' || p[1]=='d' || p[1]=='D'))
1461 /* This test includes !hex because 'e' is a valid hex digit
1462 and thus does not indicate a floating point number when
1463 the radix is hex. */
1464 if (!hex && !got_e && (*p == 'e' || *p == 'E'))
1465 got_dot = got_e = 1;
1466 /* This test does not include !hex, because a '.' always indicates
1467 a decimal floating point number regardless of the radix. */
1468 else if (!got_dot && *p == '.')
1470 else if (got_e && (p[-1] == 'e' || p[-1] == 'E')
1471 && (*p == '-' || *p == '+'))
1472 /* This is the sign of the exponent, not the end of the
1475 /* We will take any letters or digits. parse_number will
1476 complain if past the radix, or if L or U are not final. */
1477 else if ((*p < '0' || *p > '9')
1478 && ((*p < 'a' || *p > 'z')
1479 && (*p < 'A' || *p > 'Z')))
1482 toktype = parse_number (tokstart, p - tokstart, got_dot|got_e, &yylval);
1483 if (toktype == ERROR)
1485 char *err_copy = (char *) alloca (p - tokstart + 1);
1487 memcpy (err_copy, tokstart, p - tokstart);
1488 err_copy[p - tokstart] = 0;
1489 error ("Invalid number \"%s\".", err_copy);
1521 /* Build the gdb internal form of the input string in tempbuf,
1522 translating any standard C escape forms seen. Note that the
1523 buffer is null byte terminated *only* for the convenience of
1524 debugging gdb itself and printing the buffer contents when
1525 the buffer contains no embedded nulls. Gdb does not depend
1526 upon the buffer being null byte terminated, it uses the length
1527 string instead. This allows gdb to handle C strings (as well
1528 as strings in other languages) with embedded null bytes */
1530 tokptr = ++tokstart;
1534 char *char_start_pos = tokptr;
1536 /* Grow the static temp buffer if necessary, including allocating
1537 the first one on demand. */
1538 if (tempbufindex + 1 >= tempbufsize)
1540 tempbuf = (char *) realloc (tempbuf, tempbufsize += 64);
1546 /* Do nothing, loop will terminate. */
1550 c = parse_escape (&tokptr);
1555 tempbuf[tempbufindex++] = c;
1559 if (! host_char_to_target (c, &c))
1561 int len = tokptr - char_start_pos;
1562 char *copy = alloca (len + 1);
1563 memcpy (copy, char_start_pos, len);
1566 error ("There is no character corresponding to `%s' "
1567 "in the target character set `%s'.",
1568 copy, target_charset ());
1570 tempbuf[tempbufindex++] = c;
1573 } while ((*tokptr != '"') && (*tokptr != '\0'));
1574 if (*tokptr++ != '"')
1576 error ("Unterminated string in expression.");
1578 tempbuf[tempbufindex] = '\0'; /* See note above */
1579 yylval.sval.ptr = tempbuf;
1580 yylval.sval.length = tempbufindex;
1585 if (!(c == '_' || c == '$'
1586 || (c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z')))
1587 /* We must have come across a bad character (e.g. ';'). */
1588 error ("Invalid character '%c' in expression.", c);
1590 /* It's a name. See how long it is. */
1592 for (c = tokstart[namelen];
1593 (c == '_' || c == '$' || (c >= '0' && c <= '9')
1594 || (c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z') || c == '<');)
1596 /* Template parameter lists are part of the name.
1597 FIXME: This mishandles `print $a<4&&$a>3'. */
1601 /* Scan ahead to get rest of the template specification. Note
1602 that we look ahead only when the '<' adjoins non-whitespace
1603 characters; for comparison expressions, e.g. "a < b > c",
1604 there must be spaces before the '<', etc. */
1606 char * p = find_template_name_end (tokstart + namelen);
1608 namelen = p - tokstart;
1611 c = tokstart[++namelen];
1614 /* The token "if" terminates the expression and is NOT removed from
1615 the input stream. It doesn't count if it appears in the
1616 expansion of a macro. */
1618 && tokstart[0] == 'i'
1619 && tokstart[1] == 'f'
1620 && ! scanning_macro_expansion ())
1629 /* Catch specific keywords. Should be done with a data structure. */
1633 if (strncmp (tokstart, "unsigned", 8) == 0)
1635 if (current_language->la_language == language_cplus
1636 && strncmp (tokstart, "template", 8) == 0)
1638 if (strncmp (tokstart, "volatile", 8) == 0)
1639 return VOLATILE_KEYWORD;
1642 if (strncmp (tokstart, "struct", 6) == 0)
1644 if (strncmp (tokstart, "signed", 6) == 0)
1645 return SIGNED_KEYWORD;
1646 if (strncmp (tokstart, "sizeof", 6) == 0)
1648 if (strncmp (tokstart, "double", 6) == 0)
1649 return DOUBLE_KEYWORD;
1652 if (current_language->la_language == language_cplus)
1654 if (strncmp (tokstart, "false", 5) == 0)
1655 return FALSEKEYWORD;
1656 if (strncmp (tokstart, "class", 5) == 0)
1659 if (strncmp (tokstart, "union", 5) == 0)
1661 if (strncmp (tokstart, "short", 5) == 0)
1663 if (strncmp (tokstart, "const", 5) == 0)
1664 return CONST_KEYWORD;
1667 if (strncmp (tokstart, "enum", 4) == 0)
1669 if (strncmp (tokstart, "long", 4) == 0)
1671 if (current_language->la_language == language_cplus)
1673 if (strncmp (tokstart, "true", 4) == 0)
1678 if (strncmp (tokstart, "int", 3) == 0)
1685 yylval.sval.ptr = tokstart;
1686 yylval.sval.length = namelen;
1688 if (*tokstart == '$')
1690 write_dollar_variable (yylval.sval);
1694 /* Look ahead and see if we can consume more of the input
1695 string to get a reasonable class/namespace spec or a
1696 fully-qualified name. This is a kludge to get around the
1697 HP aCC compiler's generation of symbol names with embedded
1698 colons for namespace and nested classes. */
1700 /* NOTE: carlton/2003-09-24: I don't entirely understand the
1701 HP-specific code, either here or in linespec. Having said that,
1702 I suspect that we're actually moving towards their model: we want
1703 symbols whose names are fully qualified, which matches the
1704 description above. */
1707 /* Only do it if not inside single quotes */
1708 sym_class = parse_nested_classes_for_hpacc (yylval.sval.ptr, yylval.sval.length,
1709 &token_string, &class_prefix, &lexptr);
1712 /* Replace the current token with the bigger one we found */
1713 yylval.sval.ptr = token_string;
1714 yylval.sval.length = strlen (token_string);
1718 /* Use token-type BLOCKNAME for symbols that happen to be defined as
1719 functions or symtabs. If this is not so, then ...
1720 Use token-type TYPENAME for symbols that happen to be defined
1721 currently as names of types; NAME for other symbols.
1722 The caller is not constrained to care about the distinction. */
1724 char *tmp = copy_name (yylval.sval);
1726 int is_a_field_of_this = 0;
1729 sym = lookup_symbol (tmp, expression_context_block,
1731 current_language->la_language == language_cplus
1732 ? &is_a_field_of_this : (int *) NULL,
1733 (struct symtab **) NULL);
1734 /* Call lookup_symtab, not lookup_partial_symtab, in case there are
1735 no psymtabs (coff, xcoff, or some future change to blow away the
1736 psymtabs once once symbols are read). */
1737 if (sym && SYMBOL_CLASS (sym) == LOC_BLOCK)
1739 yylval.ssym.sym = sym;
1740 yylval.ssym.is_a_field_of_this = is_a_field_of_this;
1744 { /* See if it's a file name. */
1745 struct symtab *symtab;
1747 symtab = lookup_symtab (tmp);
1751 yylval.bval = BLOCKVECTOR_BLOCK (BLOCKVECTOR (symtab), STATIC_BLOCK);
1756 if (sym && SYMBOL_CLASS (sym) == LOC_TYPEDEF)
1758 /* NOTE: carlton/2003-09-25: There used to be code here to
1759 handle nested types. It didn't work very well. See the
1760 comment before qualified_type for more info. */
1761 yylval.tsym.type = SYMBOL_TYPE (sym);
1765 = language_lookup_primitive_type_by_name (current_language,
1766 current_gdbarch, tmp);
1767 if (yylval.tsym.type != NULL)
1770 /* Input names that aren't symbols but ARE valid hex numbers,
1771 when the input radix permits them, can be names or numbers
1772 depending on the parse. Note we support radixes > 16 here. */
1774 ((tokstart[0] >= 'a' && tokstart[0] < 'a' + input_radix - 10) ||
1775 (tokstart[0] >= 'A' && tokstart[0] < 'A' + input_radix - 10)))
1777 YYSTYPE newlval; /* Its value is ignored. */
1778 hextype = parse_number (tokstart, namelen, 0, &newlval);
1781 yylval.ssym.sym = sym;
1782 yylval.ssym.is_a_field_of_this = is_a_field_of_this;
1787 /* Any other kind of symbol */
1788 yylval.ssym.sym = sym;
1789 yylval.ssym.is_a_field_of_this = is_a_field_of_this;
1799 lexptr = prev_lexptr;
1801 error ("A %s in expression, near `%s'.", (msg ? msg : "error"), lexptr);