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, 2008, 2009, 2010
4 Free Software Foundation, Inc.
6 This file is part of GDB.
8 This program is free software; you can redistribute it and/or modify
9 it under the terms of the GNU General Public License as published by
10 the Free Software Foundation; either version 3 of the License, or
11 (at your option) any later version.
13 This program is distributed in the hope that it will be useful,
14 but WITHOUT ANY WARRANTY; without even the implied warranty of
15 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
16 GNU General Public License for more details.
18 You should have received a copy of the GNU General Public License
19 along with this program. If not, see <http://www.gnu.org/licenses/>. */
21 /* Parse a C expression from text in a string,
22 and return the result as a struct expression pointer.
23 That structure contains arithmetic operations in reverse polish,
24 with constants represented by operations that are followed by special data.
25 See expression.h for the details of the format.
26 What is important here is that it can be built up sequentially
27 during the process of parsing; the lower levels of the tree always
28 come first in the result.
30 Note that malloc's and realloc's in this file are transformed to
31 xmalloc and xrealloc respectively by the same sed command in the
32 makefile that remaps any other malloc/realloc inserted by the parser
33 generator. Doing this with #defines and trying to control the interaction
34 with include files (<malloc.h> and <stdlib.h> for example) just became
35 too messy, particularly when such includes can be inserted at random
36 times by the parser generator. */
41 #include "gdb_string.h"
43 #include "expression.h"
45 #include "parser-defs.h"
48 #include "bfd.h" /* Required by objfiles.h. */
49 #include "symfile.h" /* Required by objfiles.h. */
50 #include "objfiles.h" /* For have_full_symbols and have_partial_symbols */
53 #include "cp-support.h"
55 #include "gdb_assert.h"
56 #include "macroscope.h"
58 #define parse_type builtin_type (parse_gdbarch)
60 /* Remap normal yacc parser interface names (yyparse, yylex, yyerror, etc),
61 as well as gratuitiously global symbol names, so we can have multiple
62 yacc generated parsers in gdb. Note that these are only the variables
63 produced by yacc. If other parser generators (bison, byacc, etc) produce
64 additional global names that conflict at link time, then those parser
65 generators need to be fixed instead of adding those names to this list. */
67 #define yymaxdepth c_maxdepth
68 #define yyparse c_parse_internal
70 #define yyerror c_error
73 #define yydebug c_debug
82 #define yyerrflag c_errflag
83 #define yynerrs c_nerrs
88 #define yystate c_state
94 #define yyreds c_reds /* With YYDEBUG defined */
95 #define yytoks c_toks /* With YYDEBUG defined */
96 #define yyname c_name /* With YYDEBUG defined */
97 #define yyrule c_rule /* With YYDEBUG defined */
100 #define yydefred c_yydefred
101 #define yydgoto c_yydgoto
102 #define yysindex c_yysindex
103 #define yyrindex c_yyrindex
104 #define yygindex c_yygindex
105 #define yytable c_yytable
106 #define yycheck c_yycheck
109 #define YYDEBUG 1 /* Default to yydebug support */
112 #define YYFPRINTF parser_fprintf
116 static int yylex (void);
118 void yyerror (char *);
122 /* Although the yacc "value" of an expression is not used,
123 since the result is stored in the structure being created,
124 other node types do have values. */
140 } typed_val_decfloat;
144 struct typed_stoken tsval;
146 struct symtoken ssym;
149 enum exp_opcode opcode;
150 struct internalvar *ivar;
152 struct stoken_vector svec;
158 /* YYSTYPE gets defined by %union */
159 static int parse_number (char *, int, int, YYSTYPE *);
160 static struct stoken operator_stoken (const char *);
163 %type <voidval> exp exp1 type_exp start variable qualified_name lcurly
165 %type <tval> type typebase
166 %type <tvec> nonempty_typelist
167 /* %type <bval> block */
169 /* Fancy type parsing. */
170 %type <voidval> func_mod direct_abs_decl abs_decl
172 %type <lval> array_mod
174 %token <typed_val_int> INT
175 %token <typed_val_float> FLOAT
176 %token <typed_val_decfloat> DECFLOAT
178 /* Both NAME and TYPENAME tokens represent symbols in the input,
179 and both convey their data as strings.
180 But a TYPENAME is a string that happens to be defined as a typedef
181 or builtin type name (such as int or char)
182 and a NAME is any other symbol.
183 Contexts where this distinction is not important can use the
184 nonterminal "name", which matches either NAME or TYPENAME. */
186 %token <tsval> STRING
188 %token <ssym> NAME /* BLOCKNAME defined below to give it higher precedence. */
189 %token <ssym> UNKNOWN_CPP_NAME
190 %token <voidval> COMPLETE
191 %token <tsym> TYPENAME
193 %type <svec> string_exp
194 %type <ssym> name_not_typename
195 %type <tsym> typename
197 /* A NAME_OR_INT is a symbol which is not known in the symbol table,
198 but which would parse as a valid number in the current input radix.
199 E.g. "c" when input_radix==16. Depending on the parse, it will be
200 turned into a name or into a number. */
202 %token <ssym> NAME_OR_INT
205 %token STRUCT CLASS UNION ENUM SIZEOF UNSIGNED COLONCOLON
209 %type <sval> operator
210 %token REINTERPRET_CAST DYNAMIC_CAST STATIC_CAST CONST_CAST
212 /* Special type cases, put in to allow the parser to distinguish different
214 %token SIGNED_KEYWORD LONG SHORT INT_KEYWORD CONST_KEYWORD VOLATILE_KEYWORD DOUBLE_KEYWORD
216 %token <sval> VARIABLE
218 %token <opcode> ASSIGN_MODIFY
227 %right '=' ASSIGN_MODIFY
235 %left '<' '>' LEQ GEQ
240 %right UNARY INCREMENT DECREMENT
241 %right ARROW ARROW_STAR '.' DOT_STAR '[' '('
242 %token <ssym> BLOCKNAME
243 %token <bval> FILENAME
255 { write_exp_elt_opcode(OP_TYPE);
256 write_exp_elt_type($1);
257 write_exp_elt_opcode(OP_TYPE);}
260 /* Expressions, including the comma operator. */
263 { write_exp_elt_opcode (BINOP_COMMA); }
266 /* Expressions, not including the comma operator. */
267 exp : '*' exp %prec UNARY
268 { write_exp_elt_opcode (UNOP_IND); }
271 exp : '&' exp %prec UNARY
272 { write_exp_elt_opcode (UNOP_ADDR); }
275 exp : '-' exp %prec UNARY
276 { write_exp_elt_opcode (UNOP_NEG); }
279 exp : '+' exp %prec UNARY
280 { write_exp_elt_opcode (UNOP_PLUS); }
283 exp : '!' exp %prec UNARY
284 { write_exp_elt_opcode (UNOP_LOGICAL_NOT); }
287 exp : '~' exp %prec UNARY
288 { write_exp_elt_opcode (UNOP_COMPLEMENT); }
291 exp : INCREMENT exp %prec UNARY
292 { write_exp_elt_opcode (UNOP_PREINCREMENT); }
295 exp : DECREMENT exp %prec UNARY
296 { write_exp_elt_opcode (UNOP_PREDECREMENT); }
299 exp : exp INCREMENT %prec UNARY
300 { write_exp_elt_opcode (UNOP_POSTINCREMENT); }
303 exp : exp DECREMENT %prec UNARY
304 { write_exp_elt_opcode (UNOP_POSTDECREMENT); }
307 exp : SIZEOF exp %prec UNARY
308 { write_exp_elt_opcode (UNOP_SIZEOF); }
312 { write_exp_elt_opcode (STRUCTOP_PTR);
313 write_exp_string ($3);
314 write_exp_elt_opcode (STRUCTOP_PTR); }
317 exp : exp ARROW name COMPLETE
318 { mark_struct_expression ();
319 write_exp_elt_opcode (STRUCTOP_PTR);
320 write_exp_string ($3);
321 write_exp_elt_opcode (STRUCTOP_PTR); }
324 exp : exp ARROW COMPLETE
326 mark_struct_expression ();
327 write_exp_elt_opcode (STRUCTOP_PTR);
330 write_exp_string (s);
331 write_exp_elt_opcode (STRUCTOP_PTR); }
334 exp : exp ARROW qualified_name
335 { /* exp->type::name becomes exp->*(&type::name) */
336 /* Note: this doesn't work if name is a
337 static member! FIXME */
338 write_exp_elt_opcode (UNOP_ADDR);
339 write_exp_elt_opcode (STRUCTOP_MPTR); }
342 exp : exp ARROW_STAR exp
343 { write_exp_elt_opcode (STRUCTOP_MPTR); }
347 { write_exp_elt_opcode (STRUCTOP_STRUCT);
348 write_exp_string ($3);
349 write_exp_elt_opcode (STRUCTOP_STRUCT); }
352 exp : exp '.' name COMPLETE
353 { mark_struct_expression ();
354 write_exp_elt_opcode (STRUCTOP_STRUCT);
355 write_exp_string ($3);
356 write_exp_elt_opcode (STRUCTOP_STRUCT); }
359 exp : exp '.' COMPLETE
361 mark_struct_expression ();
362 write_exp_elt_opcode (STRUCTOP_STRUCT);
365 write_exp_string (s);
366 write_exp_elt_opcode (STRUCTOP_STRUCT); }
369 exp : exp '.' qualified_name
370 { /* exp.type::name becomes exp.*(&type::name) */
371 /* Note: this doesn't work if name is a
372 static member! FIXME */
373 write_exp_elt_opcode (UNOP_ADDR);
374 write_exp_elt_opcode (STRUCTOP_MEMBER); }
377 exp : exp DOT_STAR exp
378 { write_exp_elt_opcode (STRUCTOP_MEMBER); }
381 exp : exp '[' exp1 ']'
382 { write_exp_elt_opcode (BINOP_SUBSCRIPT); }
386 /* This is to save the value of arglist_len
387 being accumulated by an outer function call. */
388 { start_arglist (); }
389 arglist ')' %prec ARROW
390 { write_exp_elt_opcode (OP_FUNCALL);
391 write_exp_elt_longcst ((LONGEST) end_arglist ());
392 write_exp_elt_opcode (OP_FUNCALL); }
395 exp : UNKNOWN_CPP_NAME '('
397 /* This could potentially be a an argument defined
398 lookup function (Koenig). */
399 write_exp_elt_opcode (OP_ADL_FUNC);
400 write_exp_elt_block (expression_context_block);
401 write_exp_elt_sym (NULL); /* Placeholder. */
402 write_exp_string ($1.stoken);
403 write_exp_elt_opcode (OP_ADL_FUNC);
405 /* This is to save the value of arglist_len
406 being accumulated by an outer function call. */
410 arglist ')' %prec ARROW
412 write_exp_elt_opcode (OP_FUNCALL);
413 write_exp_elt_longcst ((LONGEST) end_arglist ());
414 write_exp_elt_opcode (OP_FUNCALL);
419 { start_arglist (); }
429 arglist : arglist ',' exp %prec ABOVE_COMMA
433 exp : exp '(' nonempty_typelist ')' const_or_volatile
435 write_exp_elt_opcode (TYPE_INSTANCE);
436 write_exp_elt_longcst ((LONGEST) $<ivec>3[0]);
437 for (i = 0; i < $<ivec>3[0]; ++i)
438 write_exp_elt_type ($<tvec>3[i + 1]);
439 write_exp_elt_longcst((LONGEST) $<ivec>3[0]);
440 write_exp_elt_opcode (TYPE_INSTANCE);
446 { $$ = end_arglist () - 1; }
448 exp : lcurly arglist rcurly %prec ARROW
449 { write_exp_elt_opcode (OP_ARRAY);
450 write_exp_elt_longcst ((LONGEST) 0);
451 write_exp_elt_longcst ((LONGEST) $3);
452 write_exp_elt_opcode (OP_ARRAY); }
455 exp : lcurly type rcurly exp %prec UNARY
456 { write_exp_elt_opcode (UNOP_MEMVAL);
457 write_exp_elt_type ($2);
458 write_exp_elt_opcode (UNOP_MEMVAL); }
461 exp : '(' type ')' exp %prec UNARY
462 { write_exp_elt_opcode (UNOP_CAST);
463 write_exp_elt_type ($2);
464 write_exp_elt_opcode (UNOP_CAST); }
471 /* Binary operators in order of decreasing precedence. */
474 { write_exp_elt_opcode (BINOP_REPEAT); }
478 { write_exp_elt_opcode (BINOP_MUL); }
482 { write_exp_elt_opcode (BINOP_DIV); }
486 { write_exp_elt_opcode (BINOP_REM); }
490 { write_exp_elt_opcode (BINOP_ADD); }
494 { write_exp_elt_opcode (BINOP_SUB); }
498 { write_exp_elt_opcode (BINOP_LSH); }
502 { write_exp_elt_opcode (BINOP_RSH); }
506 { write_exp_elt_opcode (BINOP_EQUAL); }
509 exp : exp NOTEQUAL exp
510 { write_exp_elt_opcode (BINOP_NOTEQUAL); }
514 { write_exp_elt_opcode (BINOP_LEQ); }
518 { write_exp_elt_opcode (BINOP_GEQ); }
522 { write_exp_elt_opcode (BINOP_LESS); }
526 { write_exp_elt_opcode (BINOP_GTR); }
530 { write_exp_elt_opcode (BINOP_BITWISE_AND); }
534 { write_exp_elt_opcode (BINOP_BITWISE_XOR); }
538 { write_exp_elt_opcode (BINOP_BITWISE_IOR); }
542 { write_exp_elt_opcode (BINOP_LOGICAL_AND); }
546 { write_exp_elt_opcode (BINOP_LOGICAL_OR); }
549 exp : exp '?' exp ':' exp %prec '?'
550 { write_exp_elt_opcode (TERNOP_COND); }
554 { write_exp_elt_opcode (BINOP_ASSIGN); }
557 exp : exp ASSIGN_MODIFY exp
558 { write_exp_elt_opcode (BINOP_ASSIGN_MODIFY);
559 write_exp_elt_opcode ($2);
560 write_exp_elt_opcode (BINOP_ASSIGN_MODIFY); }
564 { write_exp_elt_opcode (OP_LONG);
565 write_exp_elt_type ($1.type);
566 write_exp_elt_longcst ((LONGEST)($1.val));
567 write_exp_elt_opcode (OP_LONG); }
572 struct stoken_vector vec;
575 write_exp_string_vector ($1.type, &vec);
581 parse_number ($1.stoken.ptr, $1.stoken.length, 0, &val);
582 write_exp_elt_opcode (OP_LONG);
583 write_exp_elt_type (val.typed_val_int.type);
584 write_exp_elt_longcst ((LONGEST)val.typed_val_int.val);
585 write_exp_elt_opcode (OP_LONG);
591 { write_exp_elt_opcode (OP_DOUBLE);
592 write_exp_elt_type ($1.type);
593 write_exp_elt_dblcst ($1.dval);
594 write_exp_elt_opcode (OP_DOUBLE); }
598 { write_exp_elt_opcode (OP_DECFLOAT);
599 write_exp_elt_type ($1.type);
600 write_exp_elt_decfloatcst ($1.val);
601 write_exp_elt_opcode (OP_DECFLOAT); }
609 write_dollar_variable ($1);
613 exp : SIZEOF '(' type ')' %prec UNARY
614 { write_exp_elt_opcode (OP_LONG);
615 write_exp_elt_type (parse_type->builtin_int);
617 write_exp_elt_longcst ((LONGEST) TYPE_LENGTH ($3));
618 write_exp_elt_opcode (OP_LONG); }
621 exp : REINTERPRET_CAST '<' type '>' '(' exp ')' %prec UNARY
622 { write_exp_elt_opcode (UNOP_REINTERPRET_CAST);
623 write_exp_elt_type ($3);
624 write_exp_elt_opcode (UNOP_REINTERPRET_CAST); }
627 exp : STATIC_CAST '<' type '>' '(' exp ')' %prec UNARY
628 { write_exp_elt_opcode (UNOP_CAST);
629 write_exp_elt_type ($3);
630 write_exp_elt_opcode (UNOP_CAST); }
633 exp : DYNAMIC_CAST '<' type '>' '(' exp ')' %prec UNARY
634 { write_exp_elt_opcode (UNOP_DYNAMIC_CAST);
635 write_exp_elt_type ($3);
636 write_exp_elt_opcode (UNOP_DYNAMIC_CAST); }
639 exp : CONST_CAST '<' type '>' '(' exp ')' %prec UNARY
640 { /* We could do more error checking here, but
641 it doesn't seem worthwhile. */
642 write_exp_elt_opcode (UNOP_CAST);
643 write_exp_elt_type ($3);
644 write_exp_elt_opcode (UNOP_CAST); }
650 /* We copy the string here, and not in the
651 lexer, to guarantee that we do not leak a
652 string. Note that we follow the
653 NUL-termination convention of the
655 struct typed_stoken *vec = XNEW (struct typed_stoken);
660 vec->length = $1.length;
661 vec->ptr = malloc ($1.length + 1);
662 memcpy (vec->ptr, $1.ptr, $1.length + 1);
667 /* Note that we NUL-terminate here, but just
671 $$.tokens = realloc ($$.tokens,
672 $$.len * sizeof (struct typed_stoken));
674 p = malloc ($2.length + 1);
675 memcpy (p, $2.ptr, $2.length + 1);
677 $$.tokens[$$.len - 1].type = $2.type;
678 $$.tokens[$$.len - 1].length = $2.length;
679 $$.tokens[$$.len - 1].ptr = p;
686 enum c_string_type type = C_STRING;
688 for (i = 0; i < $1.len; ++i)
690 switch ($1.tokens[i].type)
698 && type != $1.tokens[i].type)
699 error ("Undefined string concatenation.");
700 type = $1.tokens[i].type;
704 internal_error (__FILE__, __LINE__,
705 "unrecognized type in string concatenation");
709 write_exp_string_vector (type, &$1);
710 for (i = 0; i < $1.len; ++i)
711 free ($1.tokens[i].ptr);
718 { write_exp_elt_opcode (OP_LONG);
719 write_exp_elt_type (parse_type->builtin_bool);
720 write_exp_elt_longcst ((LONGEST) 1);
721 write_exp_elt_opcode (OP_LONG); }
725 { write_exp_elt_opcode (OP_LONG);
726 write_exp_elt_type (parse_type->builtin_bool);
727 write_exp_elt_longcst ((LONGEST) 0);
728 write_exp_elt_opcode (OP_LONG); }
736 $$ = SYMBOL_BLOCK_VALUE ($1.sym);
738 error ("No file or function \"%s\".",
739 copy_name ($1.stoken));
747 block : block COLONCOLON name
749 = lookup_symbol (copy_name ($3), $1,
750 VAR_DOMAIN, (int *) NULL);
751 if (!tem || SYMBOL_CLASS (tem) != LOC_BLOCK)
752 error ("No function \"%s\" in specified context.",
754 $$ = SYMBOL_BLOCK_VALUE (tem); }
757 variable: block COLONCOLON name
758 { struct symbol *sym;
759 sym = lookup_symbol (copy_name ($3), $1,
760 VAR_DOMAIN, (int *) NULL);
762 error ("No symbol \"%s\" in specified context.",
765 write_exp_elt_opcode (OP_VAR_VALUE);
766 /* block_found is set by lookup_symbol. */
767 write_exp_elt_block (block_found);
768 write_exp_elt_sym (sym);
769 write_exp_elt_opcode (OP_VAR_VALUE); }
772 qualified_name: TYPENAME COLONCOLON name
774 struct type *type = $1.type;
775 CHECK_TYPEDEF (type);
776 if (TYPE_CODE (type) != TYPE_CODE_STRUCT
777 && TYPE_CODE (type) != TYPE_CODE_UNION
778 && TYPE_CODE (type) != TYPE_CODE_NAMESPACE)
779 error ("`%s' is not defined as an aggregate type.",
782 write_exp_elt_opcode (OP_SCOPE);
783 write_exp_elt_type (type);
784 write_exp_string ($3);
785 write_exp_elt_opcode (OP_SCOPE);
787 | TYPENAME COLONCOLON '~' name
789 struct type *type = $1.type;
790 struct stoken tmp_token;
791 CHECK_TYPEDEF (type);
792 if (TYPE_CODE (type) != TYPE_CODE_STRUCT
793 && TYPE_CODE (type) != TYPE_CODE_UNION
794 && TYPE_CODE (type) != TYPE_CODE_NAMESPACE)
795 error ("`%s' is not defined as an aggregate type.",
798 tmp_token.ptr = (char*) alloca ($4.length + 2);
799 tmp_token.length = $4.length + 1;
800 tmp_token.ptr[0] = '~';
801 memcpy (tmp_token.ptr+1, $4.ptr, $4.length);
802 tmp_token.ptr[tmp_token.length] = 0;
804 /* Check for valid destructor name. */
805 destructor_name_p (tmp_token.ptr, type);
806 write_exp_elt_opcode (OP_SCOPE);
807 write_exp_elt_type (type);
808 write_exp_string (tmp_token);
809 write_exp_elt_opcode (OP_SCOPE);
811 | TYPENAME COLONCOLON name COLONCOLON name
813 char *copy = copy_name ($3);
814 error (_("No type \"%s\" within class "
815 "or namespace \"%s\"."),
816 copy, TYPE_NAME ($1.type));
820 variable: qualified_name
821 | COLONCOLON name_not_typename
823 char *name = copy_name ($2.stoken);
825 struct minimal_symbol *msymbol;
828 lookup_symbol (name, (const struct block *) NULL,
829 VAR_DOMAIN, (int *) NULL);
832 write_exp_elt_opcode (OP_VAR_VALUE);
833 write_exp_elt_block (NULL);
834 write_exp_elt_sym (sym);
835 write_exp_elt_opcode (OP_VAR_VALUE);
839 msymbol = lookup_minimal_symbol (name, NULL, NULL);
841 write_exp_msymbol (msymbol);
842 else if (!have_full_symbols () && !have_partial_symbols ())
843 error ("No symbol table is loaded. Use the \"file\" command.");
845 error ("No symbol \"%s\" in current context.", name);
849 variable: name_not_typename
850 { struct symbol *sym = $1.sym;
854 if (symbol_read_needs_frame (sym))
856 if (innermost_block == 0
857 || contained_in (block_found,
859 innermost_block = block_found;
862 write_exp_elt_opcode (OP_VAR_VALUE);
863 /* We want to use the selected frame, not
864 another more inner frame which happens to
865 be in the same block. */
866 write_exp_elt_block (NULL);
867 write_exp_elt_sym (sym);
868 write_exp_elt_opcode (OP_VAR_VALUE);
870 else if ($1.is_a_field_of_this)
872 /* C++: it hangs off of `this'. Must
873 not inadvertently convert from a method call
875 if (innermost_block == 0
876 || contained_in (block_found,
878 innermost_block = block_found;
879 write_exp_elt_opcode (OP_THIS);
880 write_exp_elt_opcode (OP_THIS);
881 write_exp_elt_opcode (STRUCTOP_PTR);
882 write_exp_string ($1.stoken);
883 write_exp_elt_opcode (STRUCTOP_PTR);
887 struct minimal_symbol *msymbol;
888 char *arg = copy_name ($1.stoken);
891 lookup_minimal_symbol (arg, NULL, NULL);
893 write_exp_msymbol (msymbol);
894 else if (!have_full_symbols () && !have_partial_symbols ())
895 error ("No symbol table is loaded. Use the \"file\" command.");
897 error ("No symbol \"%s\" in current context.",
898 copy_name ($1.stoken));
903 space_identifier : '@' NAME
904 { push_type_address_space (copy_name ($2.stoken));
905 push_type (tp_space_identifier);
909 const_or_volatile: const_or_volatile_noopt
913 cv_with_space_id : const_or_volatile space_identifier const_or_volatile
916 const_or_volatile_or_space_identifier_noopt: cv_with_space_id
917 | const_or_volatile_noopt
920 const_or_volatile_or_space_identifier:
921 const_or_volatile_or_space_identifier_noopt
926 { push_type (tp_pointer); $$ = 0; }
928 { push_type (tp_pointer); $$ = $2; }
930 { push_type (tp_reference); $$ = 0; }
932 { push_type (tp_reference); $$ = $2; }
936 direct_abs_decl: '(' abs_decl ')'
938 | direct_abs_decl array_mod
941 push_type (tp_array);
946 push_type (tp_array);
950 | direct_abs_decl func_mod
951 { push_type (tp_function); }
953 { push_type (tp_function); }
964 | '(' nonempty_typelist ')'
965 { free ($2); $$ = 0; }
968 /* We used to try to recognize pointer to member types here, but
969 that didn't work (shift/reduce conflicts meant that these rules never
970 got executed). The problem is that
971 int (foo::bar::baz::bizzle)
972 is a function type but
973 int (foo::bar::baz::bizzle::*)
974 is a pointer to member type. Stroustrup loses again! */
979 typebase /* Implements (approximately): (type-qualifier)* type-specifier */
983 { $$ = parse_type->builtin_int; }
985 { $$ = parse_type->builtin_long; }
987 { $$ = parse_type->builtin_short; }
989 { $$ = parse_type->builtin_long; }
990 | LONG SIGNED_KEYWORD INT_KEYWORD
991 { $$ = parse_type->builtin_long; }
992 | LONG SIGNED_KEYWORD
993 { $$ = parse_type->builtin_long; }
994 | SIGNED_KEYWORD LONG INT_KEYWORD
995 { $$ = parse_type->builtin_long; }
996 | UNSIGNED LONG INT_KEYWORD
997 { $$ = parse_type->builtin_unsigned_long; }
998 | LONG UNSIGNED INT_KEYWORD
999 { $$ = parse_type->builtin_unsigned_long; }
1001 { $$ = parse_type->builtin_unsigned_long; }
1003 { $$ = parse_type->builtin_long_long; }
1004 | LONG LONG INT_KEYWORD
1005 { $$ = parse_type->builtin_long_long; }
1006 | LONG LONG SIGNED_KEYWORD INT_KEYWORD
1007 { $$ = parse_type->builtin_long_long; }
1008 | LONG LONG SIGNED_KEYWORD
1009 { $$ = parse_type->builtin_long_long; }
1010 | SIGNED_KEYWORD LONG LONG
1011 { $$ = parse_type->builtin_long_long; }
1012 | SIGNED_KEYWORD LONG LONG INT_KEYWORD
1013 { $$ = parse_type->builtin_long_long; }
1014 | UNSIGNED LONG LONG
1015 { $$ = parse_type->builtin_unsigned_long_long; }
1016 | UNSIGNED LONG LONG INT_KEYWORD
1017 { $$ = parse_type->builtin_unsigned_long_long; }
1018 | LONG LONG UNSIGNED
1019 { $$ = parse_type->builtin_unsigned_long_long; }
1020 | LONG LONG UNSIGNED INT_KEYWORD
1021 { $$ = parse_type->builtin_unsigned_long_long; }
1023 { $$ = parse_type->builtin_short; }
1024 | SHORT SIGNED_KEYWORD INT_KEYWORD
1025 { $$ = parse_type->builtin_short; }
1026 | SHORT SIGNED_KEYWORD
1027 { $$ = parse_type->builtin_short; }
1028 | UNSIGNED SHORT INT_KEYWORD
1029 { $$ = parse_type->builtin_unsigned_short; }
1031 { $$ = parse_type->builtin_unsigned_short; }
1032 | SHORT UNSIGNED INT_KEYWORD
1033 { $$ = parse_type->builtin_unsigned_short; }
1035 { $$ = parse_type->builtin_double; }
1036 | LONG DOUBLE_KEYWORD
1037 { $$ = parse_type->builtin_long_double; }
1039 { $$ = lookup_struct (copy_name ($2),
1040 expression_context_block); }
1042 { $$ = lookup_struct (copy_name ($2),
1043 expression_context_block); }
1045 { $$ = lookup_union (copy_name ($2),
1046 expression_context_block); }
1048 { $$ = lookup_enum (copy_name ($2),
1049 expression_context_block); }
1051 { $$ = lookup_unsigned_typename (parse_language,
1053 TYPE_NAME($2.type)); }
1055 { $$ = parse_type->builtin_unsigned_int; }
1056 | SIGNED_KEYWORD typename
1057 { $$ = lookup_signed_typename (parse_language,
1059 TYPE_NAME($2.type)); }
1061 { $$ = parse_type->builtin_int; }
1062 /* It appears that this rule for templates is never
1063 reduced; template recognition happens by lookahead
1064 in the token processing code in yylex. */
1065 | TEMPLATE name '<' type '>'
1066 { $$ = lookup_template_type(copy_name($2), $4,
1067 expression_context_block);
1069 | const_or_volatile_or_space_identifier_noopt typebase
1070 { $$ = follow_types ($2); }
1071 | typebase const_or_volatile_or_space_identifier_noopt
1072 { $$ = follow_types ($1); }
1078 $$.stoken.ptr = "int";
1079 $$.stoken.length = 3;
1080 $$.type = parse_type->builtin_int;
1084 $$.stoken.ptr = "long";
1085 $$.stoken.length = 4;
1086 $$.type = parse_type->builtin_long;
1090 $$.stoken.ptr = "short";
1091 $$.stoken.length = 5;
1092 $$.type = parse_type->builtin_short;
1098 { $$ = (struct type **) malloc (sizeof (struct type *) * 2);
1099 $<ivec>$[0] = 1; /* Number of types in vector */
1102 | nonempty_typelist ',' type
1103 { int len = sizeof (struct type *) * (++($<ivec>1[0]) + 1);
1104 $$ = (struct type **) realloc ((char *) $1, len);
1105 $$[$<ivec>$[0]] = $3;
1110 | ptype const_or_volatile_or_space_identifier abs_decl const_or_volatile_or_space_identifier
1111 { $$ = follow_types ($1); }
1114 const_and_volatile: CONST_KEYWORD VOLATILE_KEYWORD
1115 | VOLATILE_KEYWORD CONST_KEYWORD
1118 const_or_volatile_noopt: const_and_volatile
1119 { push_type (tp_const);
1120 push_type (tp_volatile);
1123 { push_type (tp_const); }
1125 { push_type (tp_volatile); }
1128 operator: OPERATOR NEW
1129 { $$ = operator_stoken (" new"); }
1131 { $$ = operator_stoken (" delete"); }
1132 | OPERATOR NEW '[' ']'
1133 { $$ = operator_stoken (" new[]"); }
1134 | OPERATOR DELETE '[' ']'
1135 { $$ = operator_stoken (" delete[]"); }
1137 { $$ = operator_stoken ("+"); }
1139 { $$ = operator_stoken ("-"); }
1141 { $$ = operator_stoken ("*"); }
1143 { $$ = operator_stoken ("/"); }
1145 { $$ = operator_stoken ("%"); }
1147 { $$ = operator_stoken ("^"); }
1149 { $$ = operator_stoken ("&"); }
1151 { $$ = operator_stoken ("|"); }
1153 { $$ = operator_stoken ("~"); }
1155 { $$ = operator_stoken ("!"); }
1157 { $$ = operator_stoken ("="); }
1159 { $$ = operator_stoken ("<"); }
1161 { $$ = operator_stoken (">"); }
1162 | OPERATOR ASSIGN_MODIFY
1163 { const char *op = "unknown";
1187 case BINOP_BITWISE_IOR:
1190 case BINOP_BITWISE_AND:
1193 case BINOP_BITWISE_XOR:
1200 $$ = operator_stoken (op);
1203 { $$ = operator_stoken ("<<"); }
1205 { $$ = operator_stoken (">>"); }
1207 { $$ = operator_stoken ("=="); }
1209 { $$ = operator_stoken ("!="); }
1211 { $$ = operator_stoken ("<="); }
1213 { $$ = operator_stoken (">="); }
1215 { $$ = operator_stoken ("&&"); }
1217 { $$ = operator_stoken ("||"); }
1218 | OPERATOR INCREMENT
1219 { $$ = operator_stoken ("++"); }
1220 | OPERATOR DECREMENT
1221 { $$ = operator_stoken ("--"); }
1223 { $$ = operator_stoken (","); }
1224 | OPERATOR ARROW_STAR
1225 { $$ = operator_stoken ("->*"); }
1227 { $$ = operator_stoken ("->"); }
1229 { $$ = operator_stoken ("()"); }
1231 { $$ = operator_stoken ("[]"); }
1235 struct ui_file *buf = mem_fileopen ();
1237 c_print_type ($2, NULL, buf, -1, 0);
1238 name = ui_file_xstrdup (buf, &length);
1239 ui_file_delete (buf);
1240 $$ = operator_stoken (name);
1247 name : NAME { $$ = $1.stoken; }
1248 | BLOCKNAME { $$ = $1.stoken; }
1249 | TYPENAME { $$ = $1.stoken; }
1250 | NAME_OR_INT { $$ = $1.stoken; }
1251 | UNKNOWN_CPP_NAME { $$ = $1.stoken; }
1252 | operator { $$ = $1; }
1255 name_not_typename : NAME
1257 /* These would be useful if name_not_typename was useful, but it is just
1258 a fake for "variable", so these cause reduce/reduce conflicts because
1259 the parser can't tell whether NAME_OR_INT is a name_not_typename (=variable,
1260 =exp) or just an exp. If name_not_typename was ever used in an lvalue
1261 context where only a name could occur, this might be useful.
1267 $$.sym = lookup_symbol ($1.ptr,
1268 expression_context_block,
1270 &$$.is_a_field_of_this);
1277 /* Returns a stoken of the operator name given by OP (which does not
1278 include the string "operator"). */
1279 static struct stoken
1280 operator_stoken (const char *op)
1282 static const char *operator_string = "operator";
1283 struct stoken st = { NULL, 0 };
1284 st.length = strlen (operator_string) + strlen (op);
1285 st.ptr = malloc (st.length + 1);
1286 strcpy (st.ptr, operator_string);
1287 strcat (st.ptr, op);
1289 /* The toplevel (c_parse) will free the memory allocated here. */
1290 make_cleanup (free, st.ptr);
1294 /* Take care of parsing a number (anything that starts with a digit).
1295 Set yylval and return the token type; update lexptr.
1296 LEN is the number of characters in it. */
1298 /*** Needs some error checking for the float case ***/
1301 parse_number (char *p, int len, int parsed_float, YYSTYPE *putithere)
1303 /* FIXME: Shouldn't these be unsigned? We don't deal with negative values
1304 here, and we do kind of silly things like cast to unsigned. */
1311 int base = input_radix;
1314 /* Number of "L" suffixes encountered. */
1317 /* We have found a "L" or "U" suffix. */
1318 int found_suffix = 0;
1321 struct type *signed_type;
1322 struct type *unsigned_type;
1329 /* If it ends at "df", "dd" or "dl", take it as type of decimal floating
1330 point. Return DECFLOAT. */
1332 if (len >= 2 && p[len - 2] == 'd' && p[len - 1] == 'f')
1335 putithere->typed_val_decfloat.type
1336 = parse_type->builtin_decfloat;
1337 decimal_from_string (putithere->typed_val_decfloat.val, 4,
1338 gdbarch_byte_order (parse_gdbarch), p);
1343 if (len >= 2 && p[len - 2] == 'd' && p[len - 1] == 'd')
1346 putithere->typed_val_decfloat.type
1347 = parse_type->builtin_decdouble;
1348 decimal_from_string (putithere->typed_val_decfloat.val, 8,
1349 gdbarch_byte_order (parse_gdbarch), p);
1354 if (len >= 2 && p[len - 2] == 'd' && p[len - 1] == 'l')
1357 putithere->typed_val_decfloat.type
1358 = parse_type->builtin_declong;
1359 decimal_from_string (putithere->typed_val_decfloat.val, 16,
1360 gdbarch_byte_order (parse_gdbarch), p);
1365 if (! parse_c_float (parse_gdbarch, p, len,
1366 &putithere->typed_val_float.dval,
1367 &putithere->typed_val_float.type))
1372 /* Handle base-switching prefixes 0x, 0t, 0d, 0 */
1416 if (c >= 'A' && c <= 'Z')
1418 if (c != 'l' && c != 'u')
1420 if (c >= '0' && c <= '9')
1428 if (base > 10 && c >= 'a' && c <= 'f')
1432 n += i = c - 'a' + 10;
1445 return ERROR; /* Char not a digit */
1448 return ERROR; /* Invalid digit in this base */
1450 /* Portably test for overflow (only works for nonzero values, so make
1451 a second check for zero). FIXME: Can't we just make n and prevn
1452 unsigned and avoid this? */
1453 if (c != 'l' && c != 'u' && (prevn >= n) && n != 0)
1454 unsigned_p = 1; /* Try something unsigned */
1456 /* Portably test for unsigned overflow.
1457 FIXME: This check is wrong; for example it doesn't find overflow
1458 on 0x123456789 when LONGEST is 32 bits. */
1459 if (c != 'l' && c != 'u' && n != 0)
1461 if ((unsigned_p && (ULONGEST) prevn >= (ULONGEST) n))
1462 error ("Numeric constant too large.");
1467 /* An integer constant is an int, a long, or a long long. An L
1468 suffix forces it to be long; an LL suffix forces it to be long
1469 long. If not forced to a larger size, it gets the first type of
1470 the above that it fits in. To figure out whether it fits, we
1471 shift it right and see whether anything remains. Note that we
1472 can't shift sizeof (LONGEST) * HOST_CHAR_BIT bits or more in one
1473 operation, because many compilers will warn about such a shift
1474 (which always produces a zero result). Sometimes gdbarch_int_bit
1475 or gdbarch_long_bit will be that big, sometimes not. To deal with
1476 the case where it is we just always shift the value more than
1477 once, with fewer bits each time. */
1479 un = (ULONGEST)n >> 2;
1481 && (un >> (gdbarch_int_bit (parse_gdbarch) - 2)) == 0)
1483 high_bit = ((ULONGEST)1) << (gdbarch_int_bit (parse_gdbarch) - 1);
1485 /* A large decimal (not hex or octal) constant (between INT_MAX
1486 and UINT_MAX) is a long or unsigned long, according to ANSI,
1487 never an unsigned int, but this code treats it as unsigned
1488 int. This probably should be fixed. GCC gives a warning on
1491 unsigned_type = parse_type->builtin_unsigned_int;
1492 signed_type = parse_type->builtin_int;
1494 else if (long_p <= 1
1495 && (un >> (gdbarch_long_bit (parse_gdbarch) - 2)) == 0)
1497 high_bit = ((ULONGEST)1) << (gdbarch_long_bit (parse_gdbarch) - 1);
1498 unsigned_type = parse_type->builtin_unsigned_long;
1499 signed_type = parse_type->builtin_long;
1504 if (sizeof (ULONGEST) * HOST_CHAR_BIT
1505 < gdbarch_long_long_bit (parse_gdbarch))
1506 /* A long long does not fit in a LONGEST. */
1507 shift = (sizeof (ULONGEST) * HOST_CHAR_BIT - 1);
1509 shift = (gdbarch_long_long_bit (parse_gdbarch) - 1);
1510 high_bit = (ULONGEST) 1 << shift;
1511 unsigned_type = parse_type->builtin_unsigned_long_long;
1512 signed_type = parse_type->builtin_long_long;
1515 putithere->typed_val_int.val = n;
1517 /* If the high bit of the worked out type is set then this number
1518 has to be unsigned. */
1520 if (unsigned_p || (n & high_bit))
1522 putithere->typed_val_int.type = unsigned_type;
1526 putithere->typed_val_int.type = signed_type;
1532 /* Temporary obstack used for holding strings. */
1533 static struct obstack tempbuf;
1534 static int tempbuf_init;
1536 /* Parse a C escape sequence. The initial backslash of the sequence
1537 is at (*PTR)[-1]. *PTR will be updated to point to just after the
1538 last character of the sequence. If OUTPUT is not NULL, the
1539 translated form of the escape sequence will be written there. If
1540 OUTPUT is NULL, no output is written and the call will only affect
1541 *PTR. If an escape sequence is expressed in target bytes, then the
1542 entire sequence will simply be copied to OUTPUT. Return 1 if any
1543 character was emitted, 0 otherwise. */
1546 c_parse_escape (char **ptr, struct obstack *output)
1548 char *tokptr = *ptr;
1551 /* Some escape sequences undergo character set conversion. Those we
1555 /* Hex escapes do not undergo character set conversion, so keep
1556 the escape sequence for later. */
1559 obstack_grow_str (output, "\\x");
1561 if (!isxdigit (*tokptr))
1562 error (_("\\x escape without a following hex digit"));
1563 while (isxdigit (*tokptr))
1566 obstack_1grow (output, *tokptr);
1571 /* Octal escapes do not undergo character set conversion, so
1572 keep the escape sequence for later. */
1584 obstack_grow_str (output, "\\");
1586 i < 3 && isdigit (*tokptr) && *tokptr != '8' && *tokptr != '9';
1590 obstack_1grow (output, *tokptr);
1596 /* We handle UCNs later. We could handle them here, but that
1597 would mean a spurious error in the case where the UCN could
1598 be converted to the target charset but not the host
1604 int i, len = c == 'U' ? 8 : 4;
1607 obstack_1grow (output, '\\');
1608 obstack_1grow (output, *tokptr);
1611 if (!isxdigit (*tokptr))
1612 error (_("\\%c escape without a following hex digit"), c);
1613 for (i = 0; i < len && isxdigit (*tokptr); ++i)
1616 obstack_1grow (output, *tokptr);
1622 /* We must pass backslash through so that it does not
1623 cause quoting during the second expansion. */
1626 obstack_grow_str (output, "\\\\");
1630 /* Escapes which undergo conversion. */
1633 obstack_1grow (output, '\a');
1638 obstack_1grow (output, '\b');
1643 obstack_1grow (output, '\f');
1648 obstack_1grow (output, '\n');
1653 obstack_1grow (output, '\r');
1658 obstack_1grow (output, '\t');
1663 obstack_1grow (output, '\v');
1667 /* GCC extension. */
1670 obstack_1grow (output, HOST_ESCAPE_CHAR);
1674 /* Backslash-newline expands to nothing at all. */
1680 /* A few escapes just expand to the character itself. */
1684 /* GCC extensions. */
1689 /* Unrecognized escapes turn into the character itself. */
1692 obstack_1grow (output, *tokptr);
1700 /* Parse a string or character literal from TOKPTR. The string or
1701 character may be wide or unicode. *OUTPTR is set to just after the
1702 end of the literal in the input string. The resulting token is
1703 stored in VALUE. This returns a token value, either STRING or
1704 CHAR, depending on what was parsed. *HOST_CHARS is set to the
1705 number of host characters in the literal. */
1707 parse_string_or_char (char *tokptr, char **outptr, struct typed_stoken *value,
1711 enum c_string_type type;
1713 /* Build the gdb internal form of the input string in tempbuf. Note
1714 that the buffer is null byte terminated *only* for the
1715 convenience of debugging gdb itself and printing the buffer
1716 contents when the buffer contains no embedded nulls. Gdb does
1717 not depend upon the buffer being null byte terminated, it uses
1718 the length string instead. This allows gdb to handle C strings
1719 (as well as strings in other languages) with embedded null
1725 obstack_free (&tempbuf, NULL);
1726 obstack_init (&tempbuf);
1728 /* Record the string type. */
1731 type = C_WIDE_STRING;
1734 else if (*tokptr == 'u')
1739 else if (*tokptr == 'U')
1747 /* Skip the quote. */
1761 *host_chars += c_parse_escape (&tokptr, &tempbuf);
1763 else if (c == quote)
1767 obstack_1grow (&tempbuf, c);
1769 /* FIXME: this does the wrong thing with multi-byte host
1770 characters. We could use mbrlen here, but that would
1771 make "set host-charset" a bit less useful. */
1776 if (*tokptr != quote)
1779 error ("Unterminated string in expression.");
1781 error ("Unmatched single quote.");
1786 value->ptr = obstack_base (&tempbuf);
1787 value->length = obstack_object_size (&tempbuf);
1791 return quote == '"' ? STRING : CHAR;
1798 enum exp_opcode opcode;
1802 static const struct token tokentab3[] =
1804 {">>=", ASSIGN_MODIFY, BINOP_RSH, 0},
1805 {"<<=", ASSIGN_MODIFY, BINOP_LSH, 0},
1806 {"->*", ARROW_STAR, BINOP_END, 1}
1809 static const struct token tokentab2[] =
1811 {"+=", ASSIGN_MODIFY, BINOP_ADD, 0},
1812 {"-=", ASSIGN_MODIFY, BINOP_SUB, 0},
1813 {"*=", ASSIGN_MODIFY, BINOP_MUL, 0},
1814 {"/=", ASSIGN_MODIFY, BINOP_DIV, 0},
1815 {"%=", ASSIGN_MODIFY, BINOP_REM, 0},
1816 {"|=", ASSIGN_MODIFY, BINOP_BITWISE_IOR, 0},
1817 {"&=", ASSIGN_MODIFY, BINOP_BITWISE_AND, 0},
1818 {"^=", ASSIGN_MODIFY, BINOP_BITWISE_XOR, 0},
1819 {"++", INCREMENT, BINOP_END, 0},
1820 {"--", DECREMENT, BINOP_END, 0},
1821 {"->", ARROW, BINOP_END, 0},
1822 {"&&", ANDAND, BINOP_END, 0},
1823 {"||", OROR, BINOP_END, 0},
1824 /* "::" is *not* only C++: gdb overrides its meaning in several
1825 different ways, e.g., 'filename'::func, function::variable. */
1826 {"::", COLONCOLON, BINOP_END, 0},
1827 {"<<", LSH, BINOP_END, 0},
1828 {">>", RSH, BINOP_END, 0},
1829 {"==", EQUAL, BINOP_END, 0},
1830 {"!=", NOTEQUAL, BINOP_END, 0},
1831 {"<=", LEQ, BINOP_END, 0},
1832 {">=", GEQ, BINOP_END, 0},
1833 {".*", DOT_STAR, BINOP_END, 1}
1836 /* Identifier-like tokens. */
1837 static const struct token ident_tokens[] =
1839 {"unsigned", UNSIGNED, OP_NULL, 0},
1840 {"template", TEMPLATE, OP_NULL, 1},
1841 {"volatile", VOLATILE_KEYWORD, OP_NULL, 0},
1842 {"struct", STRUCT, OP_NULL, 0},
1843 {"signed", SIGNED_KEYWORD, OP_NULL, 0},
1844 {"sizeof", SIZEOF, OP_NULL, 0},
1845 {"double", DOUBLE_KEYWORD, OP_NULL, 0},
1846 {"false", FALSEKEYWORD, OP_NULL, 1},
1847 {"class", CLASS, OP_NULL, 1},
1848 {"union", UNION, OP_NULL, 0},
1849 {"short", SHORT, OP_NULL, 0},
1850 {"const", CONST_KEYWORD, OP_NULL, 0},
1851 {"enum", ENUM, OP_NULL, 0},
1852 {"long", LONG, OP_NULL, 0},
1853 {"true", TRUEKEYWORD, OP_NULL, 1},
1854 {"int", INT_KEYWORD, OP_NULL, 0},
1855 {"new", NEW, OP_NULL, 1},
1856 {"delete", DELETE, OP_NULL, 1},
1857 {"operator", OPERATOR, OP_NULL, 1},
1859 {"and", ANDAND, BINOP_END, 1},
1860 {"and_eq", ASSIGN_MODIFY, BINOP_BITWISE_AND, 1},
1861 {"bitand", '&', OP_NULL, 1},
1862 {"bitor", '|', OP_NULL, 1},
1863 {"compl", '~', OP_NULL, 1},
1864 {"not", '!', OP_NULL, 1},
1865 {"not_eq", NOTEQUAL, BINOP_END, 1},
1866 {"or", OROR, BINOP_END, 1},
1867 {"or_eq", ASSIGN_MODIFY, BINOP_BITWISE_IOR, 1},
1868 {"xor", '^', OP_NULL, 1},
1869 {"xor_eq", ASSIGN_MODIFY, BINOP_BITWISE_XOR, 1},
1871 {"const_cast", CONST_CAST, OP_NULL, 1 },
1872 {"dynamic_cast", DYNAMIC_CAST, OP_NULL, 1 },
1873 {"static_cast", STATIC_CAST, OP_NULL, 1 },
1874 {"reinterpret_cast", REINTERPRET_CAST, OP_NULL, 1 }
1877 /* When we find that lexptr (the global var defined in parse.c) is
1878 pointing at a macro invocation, we expand the invocation, and call
1879 scan_macro_expansion to save the old lexptr here and point lexptr
1880 into the expanded text. When we reach the end of that, we call
1881 end_macro_expansion to pop back to the value we saved here. The
1882 macro expansion code promises to return only fully-expanded text,
1883 so we don't need to "push" more than one level.
1885 This is disgusting, of course. It would be cleaner to do all macro
1886 expansion beforehand, and then hand that to lexptr. But we don't
1887 really know where the expression ends. Remember, in a command like
1889 (gdb) break *ADDRESS if CONDITION
1891 we evaluate ADDRESS in the scope of the current frame, but we
1892 evaluate CONDITION in the scope of the breakpoint's location. So
1893 it's simply wrong to try to macro-expand the whole thing at once. */
1894 static char *macro_original_text;
1896 /* We save all intermediate macro expansions on this obstack for the
1897 duration of a single parse. The expansion text may sometimes have
1898 to live past the end of the expansion, due to yacc lookahead.
1899 Rather than try to be clever about saving the data for a single
1900 token, we simply keep it all and delete it after parsing has
1902 static struct obstack expansion_obstack;
1905 scan_macro_expansion (char *expansion)
1909 /* We'd better not be trying to push the stack twice. */
1910 gdb_assert (! macro_original_text);
1912 /* Copy to the obstack, and then free the intermediate
1914 copy = obstack_copy0 (&expansion_obstack, expansion, strlen (expansion));
1917 /* Save the old lexptr value, so we can return to it when we're done
1918 parsing the expanded text. */
1919 macro_original_text = lexptr;
1925 scanning_macro_expansion (void)
1927 return macro_original_text != 0;
1932 finished_macro_expansion (void)
1934 /* There'd better be something to pop back to. */
1935 gdb_assert (macro_original_text);
1937 /* Pop back to the original text. */
1938 lexptr = macro_original_text;
1939 macro_original_text = 0;
1944 scan_macro_cleanup (void *dummy)
1946 if (macro_original_text)
1947 finished_macro_expansion ();
1949 obstack_free (&expansion_obstack, NULL);
1952 /* Return true iff the token represents a C++ cast operator. */
1955 is_cast_operator (const char *token, int len)
1957 return (! strncmp (token, "dynamic_cast", len)
1958 || ! strncmp (token, "static_cast", len)
1959 || ! strncmp (token, "reinterpret_cast", len)
1960 || ! strncmp (token, "const_cast", len));
1963 /* The scope used for macro expansion. */
1964 static struct macro_scope *expression_macro_scope;
1966 /* This is set if a NAME token appeared at the very end of the input
1967 string, with no whitespace separating the name from the EOF. This
1968 is used only when parsing to do field name completion. */
1969 static int saw_name_at_eof;
1971 /* This is set if the previously-returned token was a structure
1972 operator -- either '.' or ARROW. This is used only when parsing to
1973 do field name completion. */
1974 static int last_was_structop;
1976 /* Read one token, getting characters through lexptr. */
1979 lex_one_token (void)
1985 int saw_structop = last_was_structop;
1988 last_was_structop = 0;
1992 /* Check if this is a macro invocation that we need to expand. */
1993 if (! scanning_macro_expansion ())
1995 char *expanded = macro_expand_next (&lexptr,
1996 standard_macro_lookup,
1997 expression_macro_scope);
2000 scan_macro_expansion (expanded);
2003 prev_lexptr = lexptr;
2006 /* See if it is a special token of length 3. */
2007 for (i = 0; i < sizeof tokentab3 / sizeof tokentab3[0]; i++)
2008 if (strncmp (tokstart, tokentab3[i].operator, 3) == 0)
2010 if (tokentab3[i].cxx_only
2011 && parse_language->la_language != language_cplus)
2015 yylval.opcode = tokentab3[i].opcode;
2016 return tokentab3[i].token;
2019 /* See if it is a special token of length 2. */
2020 for (i = 0; i < sizeof tokentab2 / sizeof tokentab2[0]; i++)
2021 if (strncmp (tokstart, tokentab2[i].operator, 2) == 0)
2023 if (tokentab2[i].cxx_only
2024 && parse_language->la_language != language_cplus)
2028 yylval.opcode = tokentab2[i].opcode;
2029 if (in_parse_field && tokentab2[i].token == ARROW)
2030 last_was_structop = 1;
2031 return tokentab2[i].token;
2034 switch (c = *tokstart)
2037 /* If we were just scanning the result of a macro expansion,
2038 then we need to resume scanning the original text.
2039 If we're parsing for field name completion, and the previous
2040 token allows such completion, return a COMPLETE token.
2041 Otherwise, we were already scanning the original text, and
2042 we're really done. */
2043 if (scanning_macro_expansion ())
2045 finished_macro_expansion ();
2048 else if (saw_name_at_eof)
2050 saw_name_at_eof = 0;
2053 else if (saw_structop)
2072 if (paren_depth == 0)
2079 if (comma_terminates
2081 && ! scanning_macro_expansion ())
2087 /* Might be a floating point number. */
2088 if (lexptr[1] < '0' || lexptr[1] > '9')
2091 last_was_structop = 1;
2092 goto symbol; /* Nope, must be a symbol. */
2094 /* FALL THRU into number case. */
2107 /* It's a number. */
2108 int got_dot = 0, got_e = 0, toktype;
2110 int hex = input_radix > 10;
2112 if (c == '0' && (p[1] == 'x' || p[1] == 'X'))
2117 else if (c == '0' && (p[1]=='t' || p[1]=='T' || p[1]=='d' || p[1]=='D'))
2125 /* This test includes !hex because 'e' is a valid hex digit
2126 and thus does not indicate a floating point number when
2127 the radix is hex. */
2128 if (!hex && !got_e && (*p == 'e' || *p == 'E'))
2129 got_dot = got_e = 1;
2130 /* This test does not include !hex, because a '.' always indicates
2131 a decimal floating point number regardless of the radix. */
2132 else if (!got_dot && *p == '.')
2134 else if (got_e && (p[-1] == 'e' || p[-1] == 'E')
2135 && (*p == '-' || *p == '+'))
2136 /* This is the sign of the exponent, not the end of the
2139 /* We will take any letters or digits. parse_number will
2140 complain if past the radix, or if L or U are not final. */
2141 else if ((*p < '0' || *p > '9')
2142 && ((*p < 'a' || *p > 'z')
2143 && (*p < 'A' || *p > 'Z')))
2146 toktype = parse_number (tokstart, p - tokstart, got_dot|got_e, &yylval);
2147 if (toktype == ERROR)
2149 char *err_copy = (char *) alloca (p - tokstart + 1);
2151 memcpy (err_copy, tokstart, p - tokstart);
2152 err_copy[p - tokstart] = 0;
2153 error ("Invalid number \"%s\".", err_copy);
2184 if (tokstart[1] != '"' && tokstart[1] != '\'')
2191 int result = parse_string_or_char (tokstart, &lexptr, &yylval.tsval,
2196 error ("Empty character constant.");
2197 else if (host_len > 2 && c == '\'')
2200 namelen = lexptr - tokstart - 1;
2203 else if (host_len > 1)
2204 error ("Invalid character constant.");
2210 if (!(c == '_' || c == '$'
2211 || (c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z')))
2212 /* We must have come across a bad character (e.g. ';'). */
2213 error ("Invalid character '%c' in expression.", c);
2215 /* It's a name. See how long it is. */
2217 for (c = tokstart[namelen];
2218 (c == '_' || c == '$' || (c >= '0' && c <= '9')
2219 || (c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z') || c == '<');)
2221 /* Template parameter lists are part of the name.
2222 FIXME: This mishandles `print $a<4&&$a>3'. */
2226 if (! is_cast_operator (tokstart, namelen))
2228 /* Scan ahead to get rest of the template specification. Note
2229 that we look ahead only when the '<' adjoins non-whitespace
2230 characters; for comparison expressions, e.g. "a < b > c",
2231 there must be spaces before the '<', etc. */
2233 char * p = find_template_name_end (tokstart + namelen);
2235 namelen = p - tokstart;
2239 c = tokstart[++namelen];
2242 /* The token "if" terminates the expression and is NOT removed from
2243 the input stream. It doesn't count if it appears in the
2244 expansion of a macro. */
2246 && tokstart[0] == 'i'
2247 && tokstart[1] == 'f'
2248 && ! scanning_macro_expansion ())
2253 /* For the same reason (breakpoint conditions), "thread N"
2254 terminates the expression. "thread" could be an identifier, but
2255 an identifier is never followed by a number without intervening
2256 punctuation. "task" is similar. Handle abbreviations of these,
2257 similarly to breakpoint.c:find_condition_and_thread. */
2259 && (strncmp (tokstart, "thread", namelen) == 0
2260 || strncmp (tokstart, "task", namelen) == 0)
2261 && (tokstart[namelen] == ' ' || tokstart[namelen] == '\t')
2262 && ! scanning_macro_expansion ())
2264 char *p = tokstart + namelen + 1;
2265 while (*p == ' ' || *p == '\t')
2267 if (*p >= '0' && *p <= '9')
2275 yylval.sval.ptr = tokstart;
2276 yylval.sval.length = namelen;
2278 /* Catch specific keywords. */
2279 copy = copy_name (yylval.sval);
2280 for (i = 0; i < sizeof ident_tokens / sizeof ident_tokens[0]; i++)
2281 if (strcmp (copy, ident_tokens[i].operator) == 0)
2283 if (ident_tokens[i].cxx_only
2284 && parse_language->la_language != language_cplus)
2287 /* It is ok to always set this, even though we don't always
2288 strictly need to. */
2289 yylval.opcode = ident_tokens[i].opcode;
2290 return ident_tokens[i].token;
2293 if (*tokstart == '$')
2296 if (in_parse_field && *lexptr == '\0')
2297 saw_name_at_eof = 1;
2301 /* An object of this type is pushed on a FIFO by the "outer" lexer. */
2308 DEF_VEC_O (token_and_value);
2310 /* A FIFO of tokens that have been read but not yet returned to the
2312 static VEC (token_and_value) *token_fifo;
2314 /* Non-zero if the lexer should return tokens from the FIFO. */
2317 /* Temporary storage for c_lex; this holds symbol names as they are
2319 static struct obstack name_obstack;
2321 /* Classify a NAME token. The contents of the token are in `yylval'.
2322 Updates yylval and returns the new token type. BLOCK is the block
2323 in which lookups start; this can be NULL to mean the global
2326 classify_name (struct block *block)
2330 int is_a_field_of_this = 0;
2332 copy = copy_name (yylval.sval);
2334 sym = lookup_symbol (copy, block, VAR_DOMAIN,
2335 parse_language->la_language == language_cplus
2336 ? &is_a_field_of_this : (int *) NULL);
2338 if (sym && SYMBOL_CLASS (sym) == LOC_BLOCK)
2340 yylval.ssym.sym = sym;
2341 yylval.ssym.is_a_field_of_this = is_a_field_of_this;
2346 /* See if it's a file name. */
2347 struct symtab *symtab;
2349 symtab = lookup_symtab (copy);
2352 yylval.bval = BLOCKVECTOR_BLOCK (BLOCKVECTOR (symtab), STATIC_BLOCK);
2357 if (sym && SYMBOL_CLASS (sym) == LOC_TYPEDEF)
2359 yylval.tsym.type = SYMBOL_TYPE (sym);
2364 = language_lookup_primitive_type_by_name (parse_language,
2365 parse_gdbarch, copy);
2366 if (yylval.tsym.type != NULL)
2369 /* Input names that aren't symbols but ARE valid hex numbers, when
2370 the input radix permits them, can be names or numbers depending
2371 on the parse. Note we support radixes > 16 here. */
2373 && ((copy[0] >= 'a' && copy[0] < 'a' + input_radix - 10)
2374 || (copy[0] >= 'A' && copy[0] < 'A' + input_radix - 10)))
2376 YYSTYPE newlval; /* Its value is ignored. */
2377 int hextype = parse_number (copy, yylval.sval.length, 0, &newlval);
2380 yylval.ssym.sym = sym;
2381 yylval.ssym.is_a_field_of_this = is_a_field_of_this;
2386 /* Any other kind of symbol */
2387 yylval.ssym.sym = sym;
2388 yylval.ssym.is_a_field_of_this = is_a_field_of_this;
2391 && parse_language->la_language == language_cplus
2392 && !lookup_minimal_symbol (copy, NULL, NULL))
2393 return UNKNOWN_CPP_NAME;
2398 /* Like classify_name, but used by the inner loop of the lexer, when a
2399 name might have already been seen. FIRST_NAME is true if the token
2400 in `yylval' is the first component of a name, false otherwise. If
2401 this function returns NAME, it might not have updated `yylval'.
2402 This is ok because the caller only cares about TYPENAME. */
2404 classify_inner_name (struct block *block, int first_name)
2406 struct type *type, *new_type;
2410 return classify_name (block);
2412 type = check_typedef (yylval.tsym.type);
2413 if (TYPE_CODE (type) != TYPE_CODE_STRUCT
2414 && TYPE_CODE (type) != TYPE_CODE_UNION
2415 && TYPE_CODE (type) != TYPE_CODE_NAMESPACE)
2416 /* We know the caller won't expect us to update yylval. */
2419 copy = copy_name (yylval.tsym.stoken);
2420 new_type = cp_lookup_nested_type (type, copy, block);
2422 if (new_type == NULL)
2423 /* We know the caller won't expect us to update yylval. */
2426 yylval.tsym.type = new_type;
2430 /* The outer level of a two-level lexer. This calls the inner lexer
2431 to return tokens. It then either returns these tokens, or
2432 aggregates them into a larger token. This lets us work around a
2433 problem in our parsing approach, where the parser could not
2434 distinguish between qualified names and qualified types at the
2437 This approach is still not ideal, because it mishandles template
2438 types. See the comment in lex_one_token for an example. However,
2439 this is still an improvement over the earlier approach, and will
2440 suffice until we move to better parsing technology. */
2444 token_and_value current;
2445 int first_was_coloncolon, last_was_coloncolon, first_iter;
2447 if (popping && !VEC_empty (token_and_value, token_fifo))
2449 token_and_value tv = *VEC_index (token_and_value, token_fifo, 0);
2450 VEC_ordered_remove (token_and_value, token_fifo, 0);
2456 current.token = lex_one_token ();
2457 if (current.token == NAME)
2458 current.token = classify_name (expression_context_block);
2459 if (parse_language->la_language != language_cplus
2460 || (current.token != TYPENAME && current.token != COLONCOLON))
2461 return current.token;
2463 first_was_coloncolon = current.token == COLONCOLON;
2464 last_was_coloncolon = first_was_coloncolon;
2465 obstack_free (&name_obstack, obstack_base (&name_obstack));
2466 if (!last_was_coloncolon)
2467 obstack_grow (&name_obstack, yylval.sval.ptr, yylval.sval.length);
2468 current.value = yylval;
2472 token_and_value next;
2474 next.token = lex_one_token ();
2475 next.value = yylval;
2477 if (next.token == NAME && last_was_coloncolon)
2481 classification = classify_inner_name (first_was_coloncolon
2483 : expression_context_block,
2485 /* We keep going until we either run out of names, or until
2486 we have a qualified name which is not a type. */
2487 if (classification != TYPENAME)
2489 /* Push the final component and leave the loop. */
2490 VEC_safe_push (token_and_value, token_fifo, &next);
2494 /* Update the partial name we are constructing. */
2497 /* We don't want to put a leading "::" into the name. */
2498 obstack_grow_str (&name_obstack, "::");
2500 obstack_grow (&name_obstack, next.value.sval.ptr,
2501 next.value.sval.length);
2503 yylval.sval.ptr = obstack_base (&name_obstack);
2504 yylval.sval.length = obstack_object_size (&name_obstack);
2505 current.value = yylval;
2506 current.token = classification;
2508 last_was_coloncolon = 0;
2510 else if (next.token == COLONCOLON && !last_was_coloncolon)
2511 last_was_coloncolon = 1;
2514 /* We've reached the end of the name. */
2515 VEC_safe_push (token_and_value, token_fifo, &next);
2524 /* If we ended with a "::", insert it too. */
2525 if (last_was_coloncolon)
2528 memset (&cc, 0, sizeof (token_and_value));
2529 if (first_was_coloncolon && first_iter)
2534 cc.token = COLONCOLON;
2535 VEC_safe_insert (token_and_value, token_fifo, 0, &cc);
2538 yylval = current.value;
2539 yylval.sval.ptr = obstack_copy0 (&expansion_obstack,
2541 yylval.sval.length);
2542 return current.token;
2549 struct cleanup *back_to = make_cleanup (free_current_contents,
2550 &expression_macro_scope);
2552 /* Set up the scope for macro expansion. */
2553 expression_macro_scope = NULL;
2555 if (expression_context_block)
2556 expression_macro_scope
2557 = sal_macro_scope (find_pc_line (expression_context_pc, 0));
2559 expression_macro_scope = default_macro_scope ();
2560 if (! expression_macro_scope)
2561 expression_macro_scope = user_macro_scope ();
2563 /* Initialize macro expansion code. */
2564 obstack_init (&expansion_obstack);
2565 gdb_assert (! macro_original_text);
2566 make_cleanup (scan_macro_cleanup, 0);
2568 make_cleanup_restore_integer (&yydebug);
2569 yydebug = parser_debug;
2571 /* Initialize some state used by the lexer. */
2572 last_was_structop = 0;
2573 saw_name_at_eof = 0;
2575 VEC_free (token_and_value, token_fifo);
2577 obstack_init (&name_obstack);
2578 make_cleanup_obstack_free (&name_obstack);
2580 result = yyparse ();
2581 do_cleanups (back_to);
2590 lexptr = prev_lexptr;
2592 error ("A %s in expression, near `%s'.", (msg ? msg : "error"), lexptr);