1 /* YACC parser for C expressions, for GDB.
2 Copyright (C) 1986, 1989, 1990, 1991 Free Software Foundation, Inc.
4 This file is part of GDB.
6 This program is free software; you can redistribute it and/or modify
7 it under the terms of the GNU General Public License as published by
8 the Free Software Foundation; either version 2 of the License, or
9 (at your option) any later version.
11 This program is distributed in the hope that it will be useful,
12 but WITHOUT ANY WARRANTY; without even the implied warranty of
13 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
14 GNU General Public License for more details.
16 You should have received a copy of the GNU General Public License
17 along with this program; if not, write to the Free Software
18 Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. */
20 /* Parse a C expression from text in a string,
21 and return the result as a struct expression pointer.
22 That structure contains arithmetic operations in reverse polish,
23 with constants represented by operations that are followed by special data.
24 See expression.h for the details of the format.
25 What is important here is that it can be built up sequentially
26 during the process of parsing; the lower levels of the tree always
27 come first in the result.
29 Note that malloc's and realloc's in this file are transformed to
30 xmalloc and xrealloc respectively by the same sed command in the
31 makefile that remaps any other malloc/realloc inserted by the parser
32 generator. Doing this with #defines and trying to control the interaction
33 with include files (<malloc.h> and <stdlib.h> for example) just became
34 too messy, particularly when such includes can be inserted at random
35 times by the parser generator. */
40 #include "expression.h"
42 #include "parser-defs.h"
45 #include "bfd.h" /* Required by objfiles.h. */
46 #include "symfile.h" /* Required by objfiles.h. */
47 #include "objfiles.h" /* For have_full_symbols and have_partial_symbols */
49 /* Remap normal yacc parser interface names (yyparse, yylex, yyerror, etc),
50 as well as gratuitiously global symbol names, so we can have multiple
51 yacc generated parsers in gdb. Note that these are only the variables
52 produced by yacc. If other parser generators (bison, byacc, etc) produce
53 additional global names that conflict at link time, then those parser
54 generators need to be fixed instead of adding those names to this list. */
56 #define yymaxdepth c_maxdepth
57 #define yyparse c_parse
59 #define yyerror c_error
62 #define yydebug c_debug
71 #define yyerrflag c_errflag
72 #define yynerrs c_nerrs
77 #define yystate c_state
83 #define yyreds c_reds /* With YYDEBUG defined */
84 #define yytoks c_toks /* With YYDEBUG defined */
87 #define YYDEBUG 0 /* Default to no yydebug support */
91 yyparse PARAMS ((void));
94 yylex PARAMS ((void));
97 yyerror PARAMS ((char *));
101 /* Although the yacc "value" of an expression is not used,
102 since the result is stored in the structure being created,
103 other node types do have values. */
117 struct symtoken ssym;
120 enum exp_opcode opcode;
121 struct internalvar *ivar;
128 /* YYSTYPE gets defined by %union */
130 parse_number PARAMS ((char *, int, int, YYSTYPE *));
133 %type <voidval> exp exp1 type_exp start variable qualified_name lcurly
135 %type <tval> type typebase
136 %type <tvec> nonempty_typelist
137 /* %type <bval> block */
139 /* Fancy type parsing. */
140 %type <voidval> func_mod direct_abs_decl abs_decl
142 %type <lval> array_mod
144 %token <typed_val> INT
147 /* Both NAME and TYPENAME tokens represent symbols in the input,
148 and both convey their data as strings.
149 But a TYPENAME is a string that happens to be defined as a typedef
150 or builtin type name (such as int or char)
151 and a NAME is any other symbol.
152 Contexts where this distinction is not important can use the
153 nonterminal "name", which matches either NAME or TYPENAME. */
156 %token <ssym> NAME /* BLOCKNAME defined below to give it higher precedence. */
157 %token <tsym> TYPENAME
159 %type <ssym> name_not_typename
160 %type <tsym> typename
162 /* A NAME_OR_INT is a symbol which is not known in the symbol table,
163 but which would parse as a valid number in the current input radix.
164 E.g. "c" when input_radix==16. Depending on the parse, it will be
165 turned into a name or into a number. */
167 %token <ssym> NAME_OR_INT
169 %token STRUCT CLASS UNION ENUM SIZEOF UNSIGNED COLONCOLON
173 /* Special type cases, put in to allow the parser to distinguish different
175 %token SIGNED_KEYWORD LONG SHORT INT_KEYWORD CONST_KEYWORD VOLATILE_KEYWORD
176 %token <lval> LAST REGNAME
178 %token <ivar> VARIABLE
180 %token <opcode> ASSIGN_MODIFY
187 %right '=' ASSIGN_MODIFY
195 %left '<' '>' LEQ GEQ
200 %right UNARY INCREMENT DECREMENT
201 %right ARROW '.' '[' '('
202 %token <ssym> BLOCKNAME
214 { write_exp_elt_opcode(OP_TYPE);
215 write_exp_elt_type($1);
216 write_exp_elt_opcode(OP_TYPE);}
219 /* Expressions, including the comma operator. */
222 { write_exp_elt_opcode (BINOP_COMMA); }
225 /* Expressions, not including the comma operator. */
226 exp : '*' exp %prec UNARY
227 { write_exp_elt_opcode (UNOP_IND); }
229 exp : '&' exp %prec UNARY
230 { write_exp_elt_opcode (UNOP_ADDR); }
232 exp : '-' exp %prec UNARY
233 { write_exp_elt_opcode (UNOP_NEG); }
236 exp : '!' exp %prec UNARY
237 { write_exp_elt_opcode (UNOP_LOGICAL_NOT); }
240 exp : '~' exp %prec UNARY
241 { write_exp_elt_opcode (UNOP_COMPLEMENT); }
244 exp : INCREMENT exp %prec UNARY
245 { write_exp_elt_opcode (UNOP_PREINCREMENT); }
248 exp : DECREMENT exp %prec UNARY
249 { write_exp_elt_opcode (UNOP_PREDECREMENT); }
252 exp : exp INCREMENT %prec UNARY
253 { write_exp_elt_opcode (UNOP_POSTINCREMENT); }
256 exp : exp DECREMENT %prec UNARY
257 { write_exp_elt_opcode (UNOP_POSTDECREMENT); }
260 exp : SIZEOF exp %prec UNARY
261 { write_exp_elt_opcode (UNOP_SIZEOF); }
265 { write_exp_elt_opcode (STRUCTOP_PTR);
266 write_exp_string ($3);
267 write_exp_elt_opcode (STRUCTOP_PTR); }
270 exp : exp ARROW qualified_name
271 { /* exp->type::name becomes exp->*(&type::name) */
272 /* Note: this doesn't work if name is a
273 static member! FIXME */
274 write_exp_elt_opcode (UNOP_ADDR);
275 write_exp_elt_opcode (STRUCTOP_MPTR); }
277 exp : exp ARROW '*' exp
278 { write_exp_elt_opcode (STRUCTOP_MPTR); }
282 { write_exp_elt_opcode (STRUCTOP_STRUCT);
283 write_exp_string ($3);
284 write_exp_elt_opcode (STRUCTOP_STRUCT); }
287 exp : exp '.' qualified_name
288 { /* exp.type::name becomes exp.*(&type::name) */
289 /* Note: this doesn't work if name is a
290 static member! FIXME */
291 write_exp_elt_opcode (UNOP_ADDR);
292 write_exp_elt_opcode (STRUCTOP_MEMBER); }
295 exp : exp '.' '*' exp
296 { write_exp_elt_opcode (STRUCTOP_MEMBER); }
299 exp : exp '[' exp1 ']'
300 { write_exp_elt_opcode (BINOP_SUBSCRIPT); }
304 /* This is to save the value of arglist_len
305 being accumulated by an outer function call. */
306 { start_arglist (); }
307 arglist ')' %prec ARROW
308 { write_exp_elt_opcode (OP_FUNCALL);
309 write_exp_elt_longcst ((LONGEST) end_arglist ());
310 write_exp_elt_opcode (OP_FUNCALL); }
314 { start_arglist (); }
324 arglist : arglist ',' exp %prec ABOVE_COMMA
329 { $$ = end_arglist () - 1; }
331 exp : lcurly arglist rcurly %prec ARROW
332 { write_exp_elt_opcode (OP_ARRAY);
333 write_exp_elt_longcst ((LONGEST) 0);
334 write_exp_elt_longcst ((LONGEST) $3);
335 write_exp_elt_opcode (OP_ARRAY); }
338 exp : lcurly type rcurly exp %prec UNARY
339 { write_exp_elt_opcode (UNOP_MEMVAL);
340 write_exp_elt_type ($2);
341 write_exp_elt_opcode (UNOP_MEMVAL); }
344 exp : '(' type ')' exp %prec UNARY
345 { write_exp_elt_opcode (UNOP_CAST);
346 write_exp_elt_type ($2);
347 write_exp_elt_opcode (UNOP_CAST); }
354 /* Binary operators in order of decreasing precedence. */
357 { write_exp_elt_opcode (BINOP_REPEAT); }
361 { write_exp_elt_opcode (BINOP_MUL); }
365 { write_exp_elt_opcode (BINOP_DIV); }
369 { write_exp_elt_opcode (BINOP_REM); }
373 { write_exp_elt_opcode (BINOP_ADD); }
377 { write_exp_elt_opcode (BINOP_SUB); }
381 { write_exp_elt_opcode (BINOP_LSH); }
385 { write_exp_elt_opcode (BINOP_RSH); }
389 { write_exp_elt_opcode (BINOP_EQUAL); }
392 exp : exp NOTEQUAL exp
393 { write_exp_elt_opcode (BINOP_NOTEQUAL); }
397 { write_exp_elt_opcode (BINOP_LEQ); }
401 { write_exp_elt_opcode (BINOP_GEQ); }
405 { write_exp_elt_opcode (BINOP_LESS); }
409 { write_exp_elt_opcode (BINOP_GTR); }
413 { write_exp_elt_opcode (BINOP_BITWISE_AND); }
417 { write_exp_elt_opcode (BINOP_BITWISE_XOR); }
421 { write_exp_elt_opcode (BINOP_BITWISE_IOR); }
425 { write_exp_elt_opcode (BINOP_LOGICAL_AND); }
429 { write_exp_elt_opcode (BINOP_LOGICAL_OR); }
432 exp : exp '?' exp ':' exp %prec '?'
433 { write_exp_elt_opcode (TERNOP_COND); }
437 { write_exp_elt_opcode (BINOP_ASSIGN); }
440 exp : exp ASSIGN_MODIFY exp
441 { write_exp_elt_opcode (BINOP_ASSIGN_MODIFY);
442 write_exp_elt_opcode ($2);
443 write_exp_elt_opcode (BINOP_ASSIGN_MODIFY); }
447 { write_exp_elt_opcode (OP_LONG);
448 write_exp_elt_type ($1.type);
449 write_exp_elt_longcst ((LONGEST)($1.val));
450 write_exp_elt_opcode (OP_LONG); }
455 parse_number ($1.stoken.ptr, $1.stoken.length, 0, &val);
456 write_exp_elt_opcode (OP_LONG);
457 write_exp_elt_type (val.typed_val.type);
458 write_exp_elt_longcst ((LONGEST)val.typed_val.val);
459 write_exp_elt_opcode (OP_LONG);
465 { write_exp_elt_opcode (OP_DOUBLE);
466 write_exp_elt_type (builtin_type_double);
467 write_exp_elt_dblcst ($1);
468 write_exp_elt_opcode (OP_DOUBLE); }
475 { write_exp_elt_opcode (OP_LAST);
476 write_exp_elt_longcst ((LONGEST) $1);
477 write_exp_elt_opcode (OP_LAST); }
481 { write_exp_elt_opcode (OP_REGISTER);
482 write_exp_elt_longcst ((LONGEST) $1);
483 write_exp_elt_opcode (OP_REGISTER); }
487 { write_exp_elt_opcode (OP_INTERNALVAR);
488 write_exp_elt_intern ($1);
489 write_exp_elt_opcode (OP_INTERNALVAR); }
492 exp : SIZEOF '(' type ')' %prec UNARY
493 { write_exp_elt_opcode (OP_LONG);
494 write_exp_elt_type (builtin_type_int);
495 write_exp_elt_longcst ((LONGEST) TYPE_LENGTH ($3));
496 write_exp_elt_opcode (OP_LONG); }
500 { /* C strings are converted into array constants with
501 an explicit null byte added at the end. Thus
502 the array upper bound is the string length.
503 There is no such thing in C as a completely empty
505 char *sp = $1.ptr; int count = $1.length;
508 write_exp_elt_opcode (OP_LONG);
509 write_exp_elt_type (builtin_type_char);
510 write_exp_elt_longcst ((LONGEST)(*sp++));
511 write_exp_elt_opcode (OP_LONG);
513 write_exp_elt_opcode (OP_LONG);
514 write_exp_elt_type (builtin_type_char);
515 write_exp_elt_longcst ((LONGEST)'\0');
516 write_exp_elt_opcode (OP_LONG);
517 write_exp_elt_opcode (OP_ARRAY);
518 write_exp_elt_longcst ((LONGEST) 0);
519 write_exp_elt_longcst ((LONGEST) ($1.length));
520 write_exp_elt_opcode (OP_ARRAY); }
525 { write_exp_elt_opcode (OP_THIS);
526 write_exp_elt_opcode (OP_THIS); }
534 $$ = SYMBOL_BLOCK_VALUE ($1.sym);
538 lookup_symtab (copy_name ($1.stoken));
540 $$ = BLOCKVECTOR_BLOCK
541 (BLOCKVECTOR (tem), STATIC_BLOCK);
543 error ("No file or function \"%s\".",
544 copy_name ($1.stoken));
549 block : block COLONCOLON name
551 = lookup_symbol (copy_name ($3), $1,
552 VAR_NAMESPACE, (int *) NULL,
553 (struct symtab **) NULL);
554 if (!tem || SYMBOL_CLASS (tem) != LOC_BLOCK)
555 error ("No function \"%s\" in specified context.",
557 $$ = SYMBOL_BLOCK_VALUE (tem); }
560 variable: block COLONCOLON name
561 { struct symbol *sym;
562 sym = lookup_symbol (copy_name ($3), $1,
563 VAR_NAMESPACE, (int *) NULL,
564 (struct symtab **) NULL);
566 error ("No symbol \"%s\" in specified context.",
569 write_exp_elt_opcode (OP_VAR_VALUE);
570 /* block_found is set by lookup_symbol. */
571 write_exp_elt_block (block_found);
572 write_exp_elt_sym (sym);
573 write_exp_elt_opcode (OP_VAR_VALUE); }
576 qualified_name: typebase COLONCOLON name
578 struct type *type = $1;
579 if (TYPE_CODE (type) != TYPE_CODE_STRUCT
580 && TYPE_CODE (type) != TYPE_CODE_UNION)
581 error ("`%s' is not defined as an aggregate type.",
584 write_exp_elt_opcode (OP_SCOPE);
585 write_exp_elt_type (type);
586 write_exp_string ($3);
587 write_exp_elt_opcode (OP_SCOPE);
589 | typebase COLONCOLON '~' name
591 struct type *type = $1;
592 struct stoken tmp_token;
593 if (TYPE_CODE (type) != TYPE_CODE_STRUCT
594 && TYPE_CODE (type) != TYPE_CODE_UNION)
595 error ("`%s' is not defined as an aggregate type.",
598 if (!STREQ (type_name_no_tag (type), $4.ptr))
599 error ("invalid destructor `%s::~%s'",
600 type_name_no_tag (type), $4.ptr);
602 tmp_token.ptr = (char*) alloca ($4.length + 2);
603 tmp_token.length = $4.length + 1;
604 tmp_token.ptr[0] = '~';
605 memcpy (tmp_token.ptr+1, $4.ptr, $4.length);
606 tmp_token.ptr[tmp_token.length] = 0;
607 write_exp_elt_opcode (OP_SCOPE);
608 write_exp_elt_type (type);
609 write_exp_string (tmp_token);
610 write_exp_elt_opcode (OP_SCOPE);
614 variable: qualified_name
617 char *name = copy_name ($2);
619 struct minimal_symbol *msymbol;
622 lookup_symbol (name, (const struct block *) NULL,
623 VAR_NAMESPACE, (int *) NULL,
624 (struct symtab **) NULL);
627 write_exp_elt_opcode (OP_VAR_VALUE);
628 write_exp_elt_block (NULL);
629 write_exp_elt_sym (sym);
630 write_exp_elt_opcode (OP_VAR_VALUE);
634 msymbol = lookup_minimal_symbol (name,
635 (struct objfile *) NULL);
638 write_exp_msymbol (msymbol,
639 lookup_function_type (builtin_type_int),
643 if (!have_full_symbols () && !have_partial_symbols ())
644 error ("No symbol table is loaded. Use the \"file\" command.");
646 error ("No symbol \"%s\" in current context.", name);
650 variable: name_not_typename
651 { struct symbol *sym = $1.sym;
655 if (symbol_read_needs_frame (sym))
657 if (innermost_block == 0 ||
658 contained_in (block_found,
660 innermost_block = block_found;
663 write_exp_elt_opcode (OP_VAR_VALUE);
664 /* We want to use the selected frame, not
665 another more inner frame which happens to
666 be in the same block. */
667 write_exp_elt_block (NULL);
668 write_exp_elt_sym (sym);
669 write_exp_elt_opcode (OP_VAR_VALUE);
671 else if ($1.is_a_field_of_this)
673 /* C++: it hangs off of `this'. Must
674 not inadvertently convert from a method call
676 if (innermost_block == 0 ||
677 contained_in (block_found, innermost_block))
678 innermost_block = block_found;
679 write_exp_elt_opcode (OP_THIS);
680 write_exp_elt_opcode (OP_THIS);
681 write_exp_elt_opcode (STRUCTOP_PTR);
682 write_exp_string ($1.stoken);
683 write_exp_elt_opcode (STRUCTOP_PTR);
687 struct minimal_symbol *msymbol;
688 register char *arg = copy_name ($1.stoken);
690 msymbol = lookup_minimal_symbol (arg,
691 (struct objfile *) NULL);
694 write_exp_msymbol (msymbol,
695 lookup_function_type (builtin_type_int),
698 else if (!have_full_symbols () && !have_partial_symbols ())
699 error ("No symbol table is loaded. Use the \"file\" command.");
701 error ("No symbol \"%s\" in current context.",
702 copy_name ($1.stoken));
709 /* "const" and "volatile" are curently ignored. A type qualifier
710 before the type is currently handled in the typebase rule.
711 The reason for recognizing these here (shift/reduce conflicts)
712 might be obsolete now that some pointer to member rules have
714 | typebase CONST_KEYWORD
715 | typebase VOLATILE_KEYWORD
717 { $$ = follow_types ($1); }
718 | typebase CONST_KEYWORD abs_decl
719 { $$ = follow_types ($1); }
720 | typebase VOLATILE_KEYWORD abs_decl
721 { $$ = follow_types ($1); }
725 { push_type (tp_pointer); $$ = 0; }
727 { push_type (tp_pointer); $$ = $2; }
729 { push_type (tp_reference); $$ = 0; }
731 { push_type (tp_reference); $$ = $2; }
735 direct_abs_decl: '(' abs_decl ')'
737 | direct_abs_decl array_mod
740 push_type (tp_array);
745 push_type (tp_array);
749 | direct_abs_decl func_mod
750 { push_type (tp_function); }
752 { push_type (tp_function); }
763 | '(' nonempty_typelist ')'
764 { free ((PTR)$2); $$ = 0; }
767 /* We used to try to recognize more pointer to member types here, but
768 that didn't work (shift/reduce conflicts meant that these rules never
769 got executed). The problem is that
770 int (foo::bar::baz::bizzle)
771 is a function type but
772 int (foo::bar::baz::bizzle::*)
773 is a pointer to member type. Stroustrup loses again! */
776 | typebase COLONCOLON '*'
777 { $$ = lookup_member_type (builtin_type_int, $1); }
780 typebase /* Implements (approximately): (type-qualifier)* type-specifier */
784 { $$ = builtin_type_int; }
786 { $$ = builtin_type_long; }
788 { $$ = builtin_type_short; }
790 { $$ = builtin_type_long; }
791 | UNSIGNED LONG INT_KEYWORD
792 { $$ = builtin_type_unsigned_long; }
794 { $$ = builtin_type_long_long; }
795 | LONG LONG INT_KEYWORD
796 { $$ = builtin_type_long_long; }
798 { $$ = builtin_type_unsigned_long_long; }
799 | UNSIGNED LONG LONG INT_KEYWORD
800 { $$ = builtin_type_unsigned_long_long; }
802 { $$ = builtin_type_short; }
803 | UNSIGNED SHORT INT_KEYWORD
804 { $$ = builtin_type_unsigned_short; }
806 { $$ = lookup_struct (copy_name ($2),
807 expression_context_block); }
809 { $$ = lookup_struct (copy_name ($2),
810 expression_context_block); }
812 { $$ = lookup_union (copy_name ($2),
813 expression_context_block); }
815 { $$ = lookup_enum (copy_name ($2),
816 expression_context_block); }
818 { $$ = lookup_unsigned_typename (TYPE_NAME($2.type)); }
820 { $$ = builtin_type_unsigned_int; }
821 | SIGNED_KEYWORD typename
822 { $$ = lookup_signed_typename (TYPE_NAME($2.type)); }
824 { $$ = builtin_type_int; }
825 | TEMPLATE name '<' type '>'
826 { $$ = lookup_template_type(copy_name($2), $4,
827 expression_context_block);
829 /* "const" and "volatile" are curently ignored. A type qualifier
830 after the type is handled in the ptype rule. I think these could
832 | CONST_KEYWORD typebase { $$ = $2; }
833 | VOLATILE_KEYWORD typebase { $$ = $2; }
839 $$.stoken.ptr = "int";
840 $$.stoken.length = 3;
841 $$.type = builtin_type_int;
845 $$.stoken.ptr = "long";
846 $$.stoken.length = 4;
847 $$.type = builtin_type_long;
851 $$.stoken.ptr = "short";
852 $$.stoken.length = 5;
853 $$.type = builtin_type_short;
859 { $$ = (struct type **) malloc (sizeof (struct type *) * 2);
860 $<ivec>$[0] = 1; /* Number of types in vector */
863 | nonempty_typelist ',' type
864 { int len = sizeof (struct type *) * (++($<ivec>1[0]) + 1);
865 $$ = (struct type **) realloc ((char *) $1, len);
866 $$[$<ivec>$[0]] = $3;
870 name : NAME { $$ = $1.stoken; }
871 | BLOCKNAME { $$ = $1.stoken; }
872 | TYPENAME { $$ = $1.stoken; }
873 | NAME_OR_INT { $$ = $1.stoken; }
876 name_not_typename : NAME
878 /* These would be useful if name_not_typename was useful, but it is just
879 a fake for "variable", so these cause reduce/reduce conflicts because
880 the parser can't tell whether NAME_OR_INT is a name_not_typename (=variable,
881 =exp) or just an exp. If name_not_typename was ever used in an lvalue
882 context where only a name could occur, this might be useful.
889 /* Take care of parsing a number (anything that starts with a digit).
890 Set yylval and return the token type; update lexptr.
891 LEN is the number of characters in it. */
893 /*** Needs some error checking for the float case ***/
896 parse_number (p, len, parsed_float, putithere)
902 register LONGEST n = 0;
903 register LONGEST prevn = 0;
906 register int base = input_radix;
909 unsigned LONGEST high_bit;
910 struct type *signed_type;
911 struct type *unsigned_type;
915 /* It's a float since it contains a point or an exponent. */
916 putithere->dval = atof (p);
920 /* Handle base-switching prefixes 0x, 0t, 0d, 0 */
954 if (c >= 'A' && c <= 'Z')
956 if (c != 'l' && c != 'u')
958 if (c >= '0' && c <= '9')
962 if (base > 10 && c >= 'a' && c <= 'f')
963 n += i = c - 'a' + 10;
964 else if (len == 0 && c == 'l')
966 else if (len == 0 && c == 'u')
969 return ERROR; /* Char not a digit */
972 return ERROR; /* Invalid digit in this base */
974 /* Portably test for overflow (only works for nonzero values, so make
975 a second check for zero). */
976 if((prevn >= n) && n != 0)
977 unsigned_p=1; /* Try something unsigned */
978 /* If range checking enabled, portably test for unsigned overflow. */
979 if(RANGE_CHECK && n!=0)
981 if((unsigned_p && (unsigned)prevn >= (unsigned)n))
982 range_error("Overflow on numeric constant.");
987 /* If the number is too big to be an int, or it's got an l suffix
988 then it's a long. Work out if this has to be a long by
989 shifting right and and seeing if anything remains, and the
990 target int size is different to the target long size.
992 In the expression below, we could have tested
993 (n >> TARGET_INT_BIT)
994 to see if it was zero,
995 but too many compilers warn about that, when ints and longs
996 are the same size. So we shift it twice, with fewer bits
997 each time, for the same result. */
999 if ( (TARGET_INT_BIT != TARGET_LONG_BIT
1000 && ((n >> 2) >> (TARGET_INT_BIT-2))) /* Avoid shift warning */
1003 high_bit = ((unsigned LONGEST)1) << (TARGET_LONG_BIT-1);
1004 unsigned_type = builtin_type_unsigned_long;
1005 signed_type = builtin_type_long;
1009 high_bit = ((unsigned LONGEST)1) << (TARGET_INT_BIT-1);
1010 unsigned_type = builtin_type_unsigned_int;
1011 signed_type = builtin_type_int;
1014 putithere->typed_val.val = n;
1016 /* If the high bit of the worked out type is set then this number
1017 has to be unsigned. */
1019 if (unsigned_p || (n & high_bit))
1021 putithere->typed_val.type = unsigned_type;
1025 putithere->typed_val.type = signed_type;
1035 enum exp_opcode opcode;
1038 static const struct token tokentab3[] =
1040 {">>=", ASSIGN_MODIFY, BINOP_RSH},
1041 {"<<=", ASSIGN_MODIFY, BINOP_LSH}
1044 static const struct token tokentab2[] =
1046 {"+=", ASSIGN_MODIFY, BINOP_ADD},
1047 {"-=", ASSIGN_MODIFY, BINOP_SUB},
1048 {"*=", ASSIGN_MODIFY, BINOP_MUL},
1049 {"/=", ASSIGN_MODIFY, BINOP_DIV},
1050 {"%=", ASSIGN_MODIFY, BINOP_REM},
1051 {"|=", ASSIGN_MODIFY, BINOP_BITWISE_IOR},
1052 {"&=", ASSIGN_MODIFY, BINOP_BITWISE_AND},
1053 {"^=", ASSIGN_MODIFY, BINOP_BITWISE_XOR},
1054 {"++", INCREMENT, BINOP_END},
1055 {"--", DECREMENT, BINOP_END},
1056 {"->", ARROW, BINOP_END},
1057 {"&&", ANDAND, BINOP_END},
1058 {"||", OROR, BINOP_END},
1059 {"::", COLONCOLON, BINOP_END},
1060 {"<<", LSH, BINOP_END},
1061 {">>", RSH, BINOP_END},
1062 {"==", EQUAL, BINOP_END},
1063 {"!=", NOTEQUAL, BINOP_END},
1064 {"<=", LEQ, BINOP_END},
1065 {">=", GEQ, BINOP_END}
1068 /* Read one token, getting characters through lexptr. */
1079 static char *tempbuf;
1080 static int tempbufsize;
1085 /* See if it is a special token of length 3. */
1086 for (i = 0; i < sizeof tokentab3 / sizeof tokentab3[0]; i++)
1087 if (STREQN (tokstart, tokentab3[i].operator, 3))
1090 yylval.opcode = tokentab3[i].opcode;
1091 return tokentab3[i].token;
1094 /* See if it is a special token of length 2. */
1095 for (i = 0; i < sizeof tokentab2 / sizeof tokentab2[0]; i++)
1096 if (STREQN (tokstart, tokentab2[i].operator, 2))
1099 yylval.opcode = tokentab2[i].opcode;
1100 return tokentab2[i].token;
1103 switch (c = *tokstart)
1115 /* We either have a character constant ('0' or '\177' for example)
1116 or we have a quoted symbol reference ('foo(int,int)' in C++
1121 c = parse_escape (&lexptr);
1123 yylval.typed_val.val = c;
1124 yylval.typed_val.type = builtin_type_char;
1129 namelen = skip_quoted (tokstart) - tokstart;
1132 lexptr = tokstart + namelen;
1133 if (lexptr[-1] != '\'')
1134 error ("Unmatched single quote.");
1139 error ("Invalid character constant.");
1149 if (paren_depth == 0)
1156 if (comma_terminates && paren_depth == 0)
1162 /* Might be a floating point number. */
1163 if (lexptr[1] < '0' || lexptr[1] > '9')
1164 goto symbol; /* Nope, must be a symbol. */
1165 /* FALL THRU into number case. */
1178 /* It's a number. */
1179 int got_dot = 0, got_e = 0, toktype;
1180 register char *p = tokstart;
1181 int hex = input_radix > 10;
1183 if (c == '0' && (p[1] == 'x' || p[1] == 'X'))
1188 else if (c == '0' && (p[1]=='t' || p[1]=='T' || p[1]=='d' || p[1]=='D'))
1196 /* This test includes !hex because 'e' is a valid hex digit
1197 and thus does not indicate a floating point number when
1198 the radix is hex. */
1199 if (!hex && !got_e && (*p == 'e' || *p == 'E'))
1200 got_dot = got_e = 1;
1201 /* This test does not include !hex, because a '.' always indicates
1202 a decimal floating point number regardless of the radix. */
1203 else if (!got_dot && *p == '.')
1205 else if (got_e && (p[-1] == 'e' || p[-1] == 'E')
1206 && (*p == '-' || *p == '+'))
1207 /* This is the sign of the exponent, not the end of the
1210 /* We will take any letters or digits. parse_number will
1211 complain if past the radix, or if L or U are not final. */
1212 else if ((*p < '0' || *p > '9')
1213 && ((*p < 'a' || *p > 'z')
1214 && (*p < 'A' || *p > 'Z')))
1217 toktype = parse_number (tokstart, p - tokstart, got_dot|got_e, &yylval);
1218 if (toktype == ERROR)
1220 char *err_copy = (char *) alloca (p - tokstart + 1);
1222 memcpy (err_copy, tokstart, p - tokstart);
1223 err_copy[p - tokstart] = 0;
1224 error ("Invalid number \"%s\".", err_copy);
1256 /* Build the gdb internal form of the input string in tempbuf,
1257 translating any standard C escape forms seen. Note that the
1258 buffer is null byte terminated *only* for the convenience of
1259 debugging gdb itself and printing the buffer contents when
1260 the buffer contains no embedded nulls. Gdb does not depend
1261 upon the buffer being null byte terminated, it uses the length
1262 string instead. This allows gdb to handle C strings (as well
1263 as strings in other languages) with embedded null bytes */
1265 tokptr = ++tokstart;
1269 /* Grow the static temp buffer if necessary, including allocating
1270 the first one on demand. */
1271 if (tempbufindex + 1 >= tempbufsize)
1273 tempbuf = (char *) realloc (tempbuf, tempbufsize += 64);
1279 /* Do nothing, loop will terminate. */
1283 c = parse_escape (&tokptr);
1288 tempbuf[tempbufindex++] = c;
1291 tempbuf[tempbufindex++] = *tokptr++;
1294 } while ((*tokptr != '"') && (*tokptr != '\0'));
1295 if (*tokptr++ != '"')
1297 error ("Unterminated string in expression.");
1299 tempbuf[tempbufindex] = '\0'; /* See note above */
1300 yylval.sval.ptr = tempbuf;
1301 yylval.sval.length = tempbufindex;
1306 if (!(c == '_' || c == '$'
1307 || (c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z')))
1308 /* We must have come across a bad character (e.g. ';'). */
1309 error ("Invalid character '%c' in expression.", c);
1311 /* It's a name. See how long it is. */
1313 for (c = tokstart[namelen];
1314 (c == '_' || c == '$' || (c >= '0' && c <= '9')
1315 || (c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z'));
1316 c = tokstart[++namelen])
1319 /* The token "if" terminates the expression and is NOT
1320 removed from the input stream. */
1321 if (namelen == 2 && tokstart[0] == 'i' && tokstart[1] == 'f')
1328 /* Handle the tokens $digits; also $ (short for $0) and $$ (short for $$1)
1329 and $$digits (equivalent to $<-digits> if you could type that).
1330 Make token type LAST, and put the number (the digits) in yylval. */
1333 if (*tokstart == '$')
1335 register int negate = 0;
1337 /* Double dollar means negate the number and add -1 as well.
1338 Thus $$ alone means -1. */
1339 if (namelen >= 2 && tokstart[1] == '$')
1346 /* Just dollars (one or two) */
1347 yylval.lval = - negate;
1350 /* Is the rest of the token digits? */
1351 for (; c < namelen; c++)
1352 if (!(tokstart[c] >= '0' && tokstart[c] <= '9'))
1356 yylval.lval = atoi (tokstart + 1 + negate);
1358 yylval.lval = - yylval.lval;
1363 /* Handle tokens that refer to machine registers:
1364 $ followed by a register name. */
1366 if (*tokstart == '$') {
1367 for (c = 0; c < NUM_REGS; c++)
1368 if (namelen - 1 == strlen (reg_names[c])
1369 && STREQN (tokstart + 1, reg_names[c], namelen - 1))
1374 for (c = 0; c < num_std_regs; c++)
1375 if (namelen - 1 == strlen (std_regs[c].name)
1376 && STREQN (tokstart + 1, std_regs[c].name, namelen - 1))
1378 yylval.lval = std_regs[c].regnum;
1382 /* Catch specific keywords. Should be done with a data structure. */
1386 if (STREQN (tokstart, "unsigned", 8))
1388 if (current_language->la_language == language_cplus
1389 && STREQN (tokstart, "template", 8))
1391 if (STREQN (tokstart, "volatile", 8))
1392 return VOLATILE_KEYWORD;
1395 if (STREQN (tokstart, "struct", 6))
1397 if (STREQN (tokstart, "signed", 6))
1398 return SIGNED_KEYWORD;
1399 if (STREQN (tokstart, "sizeof", 6))
1403 if (current_language->la_language == language_cplus
1404 && STREQN (tokstart, "class", 5))
1406 if (STREQN (tokstart, "union", 5))
1408 if (STREQN (tokstart, "short", 5))
1410 if (STREQN (tokstart, "const", 5))
1411 return CONST_KEYWORD;
1414 if (STREQN (tokstart, "enum", 4))
1416 if (STREQN (tokstart, "long", 4))
1418 if (current_language->la_language == language_cplus
1419 && STREQN (tokstart, "this", 4))
1421 static const char this_name[] =
1422 { CPLUS_MARKER, 't', 'h', 'i', 's', '\0' };
1424 if (lookup_symbol (this_name, expression_context_block,
1425 VAR_NAMESPACE, (int *) NULL,
1426 (struct symtab **) NULL))
1431 if (STREQN (tokstart, "int", 3))
1438 yylval.sval.ptr = tokstart;
1439 yylval.sval.length = namelen;
1441 /* Any other names starting in $ are debugger internal variables. */
1443 if (*tokstart == '$')
1445 yylval.ivar = lookup_internalvar (copy_name (yylval.sval) + 1);
1449 /* Use token-type BLOCKNAME for symbols that happen to be defined as
1450 functions or symtabs. If this is not so, then ...
1451 Use token-type TYPENAME for symbols that happen to be defined
1452 currently as names of types; NAME for other symbols.
1453 The caller is not constrained to care about the distinction. */
1455 char *tmp = copy_name (yylval.sval);
1457 int is_a_field_of_this = 0;
1460 sym = lookup_symbol (tmp, expression_context_block,
1462 current_language->la_language == language_cplus
1463 ? &is_a_field_of_this : (int *) NULL,
1464 (struct symtab **) NULL);
1465 /* Call lookup_symtab, not lookup_partial_symtab, in case there are
1466 no psymtabs (coff, xcoff, or some future change to blow away the
1467 psymtabs once once symbols are read). */
1468 if ((sym && SYMBOL_CLASS (sym) == LOC_BLOCK) ||
1469 lookup_symtab (tmp))
1471 yylval.ssym.sym = sym;
1472 yylval.ssym.is_a_field_of_this = is_a_field_of_this;
1475 if (sym && SYMBOL_CLASS (sym) == LOC_TYPEDEF)
1479 struct symbol *best_sym;
1481 /* Look ahead to detect nested types. This probably should be
1482 done in the grammar, but trying seemed to introduce a lot
1483 of shift/reduce and reduce/reduce conflicts. It's possible
1484 that it could be done, though. Or perhaps a non-grammar, but
1485 less ad hoc, approach would work well. */
1487 /* Since we do not currently have any way of distinguishing
1488 a nested type from a non-nested one (the stabs don't tell
1489 us whether a type is nested), we just ignore the
1496 /* Skip whitespace. */
1497 while (*p == ' ' || *p == '\t' || *p == '\n')
1499 if (*p == ':' && p[1] == ':')
1501 /* Skip the `::'. */
1503 /* Skip whitespace. */
1504 while (*p == ' ' || *p == '\t' || *p == '\n')
1507 while (*p == '_' || *p == '$' || (*p >= '0' && *p <= '9')
1508 || (*p >= 'a' && *p <= 'z')
1509 || (*p >= 'A' && *p <= 'Z'))
1513 struct symbol *cur_sym;
1514 /* As big as the whole rest of the expression, which is
1515 at least big enough. */
1516 char *tmp = alloca (strlen (namestart));
1518 memcpy (tmp, namestart, p - namestart);
1519 tmp[p - namestart] = '\0';
1520 cur_sym = lookup_symbol (tmp, expression_context_block,
1521 VAR_NAMESPACE, (int *) NULL,
1522 (struct symtab **) NULL);
1525 if (SYMBOL_CLASS (cur_sym) == LOC_TYPEDEF)
1543 yylval.tsym.type = SYMBOL_TYPE (best_sym);
1546 if ((yylval.tsym.type = lookup_primitive_typename (tmp)) != 0)
1549 /* Input names that aren't symbols but ARE valid hex numbers,
1550 when the input radix permits them, can be names or numbers
1551 depending on the parse. Note we support radixes > 16 here. */
1553 ((tokstart[0] >= 'a' && tokstart[0] < 'a' + input_radix - 10) ||
1554 (tokstart[0] >= 'A' && tokstart[0] < 'A' + input_radix - 10)))
1556 YYSTYPE newlval; /* Its value is ignored. */
1557 hextype = parse_number (tokstart, namelen, 0, &newlval);
1560 yylval.ssym.sym = sym;
1561 yylval.ssym.is_a_field_of_this = is_a_field_of_this;
1566 /* Any other kind of symbol */
1567 yylval.ssym.sym = sym;
1568 yylval.ssym.is_a_field_of_this = is_a_field_of_this;
1577 error (msg ? msg : "Invalid syntax in expression.");