]> git.ipfire.org Git - thirdparty/binutils-gdb.git/blob - gdb/p-exp.y
* gdbtypes.h (struct main_type): Change type of name,tag_name,
[thirdparty/binutils-gdb.git] / gdb / p-exp.y
1 /* YACC parser for Pascal expressions, for GDB.
2 Copyright (C) 2000, 2006-2012 Free Software Foundation, Inc.
3
4 This file is part of GDB.
5
6 This program is free software; you can redistribute it and/or modify
7 it under the terms of the GNU General Public License as published by
8 the Free Software Foundation; either version 3 of the License, or
9 (at your option) any later version.
10
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.
15
16 You should have received a copy of the GNU General Public License
17 along with this program. If not, see <http://www.gnu.org/licenses/>. */
18
19 /* This file is derived from c-exp.y */
20
21 /* Parse a Pascal 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.
29
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. */
37
38 /* Known bugs or limitations:
39 - pascal string operations are not supported at all.
40 - there are some problems with boolean types.
41 - Pascal type hexadecimal constants are not supported
42 because they conflict with the internal variables format.
43 Probably also lots of other problems, less well defined PM. */
44 %{
45
46 #include "defs.h"
47 #include "gdb_string.h"
48 #include <ctype.h>
49 #include "expression.h"
50 #include "value.h"
51 #include "parser-defs.h"
52 #include "language.h"
53 #include "p-lang.h"
54 #include "bfd.h" /* Required by objfiles.h. */
55 #include "symfile.h" /* Required by objfiles.h. */
56 #include "objfiles.h" /* For have_full_symbols and have_partial_symbols. */
57 #include "block.h"
58
59 #define parse_type builtin_type (parse_gdbarch)
60
61 /* Remap normal yacc parser interface names (yyparse, yylex, yyerror, etc),
62 as well as gratuitiously global symbol names, so we can have multiple
63 yacc generated parsers in gdb. Note that these are only the variables
64 produced by yacc. If other parser generators (bison, byacc, etc) produce
65 additional global names that conflict at link time, then those parser
66 generators need to be fixed instead of adding those names to this list. */
67
68 #define yymaxdepth pascal_maxdepth
69 #define yyparse pascal_parse
70 #define yylex pascal_lex
71 #define yyerror pascal_error
72 #define yylval pascal_lval
73 #define yychar pascal_char
74 #define yydebug pascal_debug
75 #define yypact pascal_pact
76 #define yyr1 pascal_r1
77 #define yyr2 pascal_r2
78 #define yydef pascal_def
79 #define yychk pascal_chk
80 #define yypgo pascal_pgo
81 #define yyact pascal_act
82 #define yyexca pascal_exca
83 #define yyerrflag pascal_errflag
84 #define yynerrs pascal_nerrs
85 #define yyps pascal_ps
86 #define yypv pascal_pv
87 #define yys pascal_s
88 #define yy_yys pascal_yys
89 #define yystate pascal_state
90 #define yytmp pascal_tmp
91 #define yyv pascal_v
92 #define yy_yyv pascal_yyv
93 #define yyval pascal_val
94 #define yylloc pascal_lloc
95 #define yyreds pascal_reds /* With YYDEBUG defined */
96 #define yytoks pascal_toks /* With YYDEBUG defined */
97 #define yyname pascal_name /* With YYDEBUG defined */
98 #define yyrule pascal_rule /* With YYDEBUG defined */
99 #define yylhs pascal_yylhs
100 #define yylen pascal_yylen
101 #define yydefred pascal_yydefred
102 #define yydgoto pascal_yydgoto
103 #define yysindex pascal_yysindex
104 #define yyrindex pascal_yyrindex
105 #define yygindex pascal_yygindex
106 #define yytable pascal_yytable
107 #define yycheck pascal_yycheck
108
109 #ifndef YYDEBUG
110 #define YYDEBUG 1 /* Default to yydebug support */
111 #endif
112
113 #define YYFPRINTF parser_fprintf
114
115 int yyparse (void);
116
117 static int yylex (void);
118
119 void yyerror (char *);
120
121 static char * uptok (char *, int);
122 %}
123
124 /* Although the yacc "value" of an expression is not used,
125 since the result is stored in the structure being created,
126 other node types do have values. */
127
128 %union
129 {
130 LONGEST lval;
131 struct {
132 LONGEST val;
133 struct type *type;
134 } typed_val_int;
135 struct {
136 DOUBLEST dval;
137 struct type *type;
138 } typed_val_float;
139 struct symbol *sym;
140 struct type *tval;
141 struct stoken sval;
142 struct ttype tsym;
143 struct symtoken ssym;
144 int voidval;
145 struct block *bval;
146 enum exp_opcode opcode;
147 struct internalvar *ivar;
148
149 struct type **tvec;
150 int *ivec;
151 }
152
153 %{
154 /* YYSTYPE gets defined by %union */
155 static int parse_number (char *, int, int, YYSTYPE *);
156
157 static struct type *current_type;
158 static struct internalvar *intvar;
159 static int leftdiv_is_integer;
160 static void push_current_type (void);
161 static void pop_current_type (void);
162 static int search_field;
163 %}
164
165 %type <voidval> exp exp1 type_exp start normal_start variable qualified_name
166 %type <tval> type typebase
167 /* %type <bval> block */
168
169 /* Fancy type parsing. */
170 %type <tval> ptype
171
172 %token <typed_val_int> INT
173 %token <typed_val_float> FLOAT
174
175 /* Both NAME and TYPENAME tokens represent symbols in the input,
176 and both convey their data as strings.
177 But a TYPENAME is a string that happens to be defined as a typedef
178 or builtin type name (such as int or char)
179 and a NAME is any other symbol.
180 Contexts where this distinction is not important can use the
181 nonterminal "name", which matches either NAME or TYPENAME. */
182
183 %token <sval> STRING
184 %token <sval> FIELDNAME
185 %token <voidval> COMPLETE
186 %token <ssym> NAME /* BLOCKNAME defined below to give it higher precedence. */
187 %token <tsym> TYPENAME
188 %type <sval> name
189 %type <ssym> name_not_typename
190
191 /* A NAME_OR_INT is a symbol which is not known in the symbol table,
192 but which would parse as a valid number in the current input radix.
193 E.g. "c" when input_radix==16. Depending on the parse, it will be
194 turned into a name or into a number. */
195
196 %token <ssym> NAME_OR_INT
197
198 %token STRUCT CLASS SIZEOF COLONCOLON
199 %token ERROR
200
201 /* Special type cases, put in to allow the parser to distinguish different
202 legal basetypes. */
203
204 %token <voidval> VARIABLE
205
206
207 /* Object pascal */
208 %token THIS
209 %token <lval> TRUEKEYWORD FALSEKEYWORD
210
211 %left ','
212 %left ABOVE_COMMA
213 %right ASSIGN
214 %left NOT
215 %left OR
216 %left XOR
217 %left ANDAND
218 %left '=' NOTEQUAL
219 %left '<' '>' LEQ GEQ
220 %left LSH RSH DIV MOD
221 %left '@'
222 %left '+' '-'
223 %left '*' '/'
224 %right UNARY INCREMENT DECREMENT
225 %right ARROW '.' '[' '('
226 %left '^'
227 %token <ssym> BLOCKNAME
228 %type <bval> block
229 %left COLONCOLON
230
231 \f
232 %%
233
234 start : { current_type = NULL;
235 intvar = NULL;
236 search_field = 0;
237 leftdiv_is_integer = 0;
238 }
239 normal_start {}
240 ;
241
242 normal_start :
243 exp1
244 | type_exp
245 ;
246
247 type_exp: type
248 { write_exp_elt_opcode(OP_TYPE);
249 write_exp_elt_type($1);
250 write_exp_elt_opcode(OP_TYPE);
251 current_type = $1; } ;
252
253 /* Expressions, including the comma operator. */
254 exp1 : exp
255 | exp1 ',' exp
256 { write_exp_elt_opcode (BINOP_COMMA); }
257 ;
258
259 /* Expressions, not including the comma operator. */
260 exp : exp '^' %prec UNARY
261 { write_exp_elt_opcode (UNOP_IND);
262 if (current_type)
263 current_type = TYPE_TARGET_TYPE (current_type); }
264 ;
265
266 exp : '@' exp %prec UNARY
267 { write_exp_elt_opcode (UNOP_ADDR);
268 if (current_type)
269 current_type = TYPE_POINTER_TYPE (current_type); }
270 ;
271
272 exp : '-' exp %prec UNARY
273 { write_exp_elt_opcode (UNOP_NEG); }
274 ;
275
276 exp : NOT exp %prec UNARY
277 { write_exp_elt_opcode (UNOP_LOGICAL_NOT); }
278 ;
279
280 exp : INCREMENT '(' exp ')' %prec UNARY
281 { write_exp_elt_opcode (UNOP_PREINCREMENT); }
282 ;
283
284 exp : DECREMENT '(' exp ')' %prec UNARY
285 { write_exp_elt_opcode (UNOP_PREDECREMENT); }
286 ;
287
288
289 field_exp : exp '.' %prec UNARY
290 { search_field = 1; }
291 ;
292
293 exp : field_exp FIELDNAME
294 { write_exp_elt_opcode (STRUCTOP_STRUCT);
295 write_exp_string ($2);
296 write_exp_elt_opcode (STRUCTOP_STRUCT);
297 search_field = 0;
298 if (current_type)
299 {
300 while (TYPE_CODE (current_type)
301 == TYPE_CODE_PTR)
302 current_type =
303 TYPE_TARGET_TYPE (current_type);
304 current_type = lookup_struct_elt_type (
305 current_type, $2.ptr, 0);
306 }
307 }
308 ;
309
310
311 exp : field_exp name
312 { mark_struct_expression ();
313 write_exp_elt_opcode (STRUCTOP_STRUCT);
314 write_exp_string ($2);
315 write_exp_elt_opcode (STRUCTOP_STRUCT);
316 search_field = 0;
317 if (current_type)
318 {
319 while (TYPE_CODE (current_type)
320 == TYPE_CODE_PTR)
321 current_type =
322 TYPE_TARGET_TYPE (current_type);
323 current_type = lookup_struct_elt_type (
324 current_type, $2.ptr, 0);
325 }
326 }
327 ;
328
329 exp : field_exp COMPLETE
330 { struct stoken s;
331 mark_struct_expression ();
332 write_exp_elt_opcode (STRUCTOP_STRUCT);
333 s.ptr = "";
334 s.length = 0;
335 write_exp_string (s);
336 write_exp_elt_opcode (STRUCTOP_STRUCT); }
337 ;
338
339 exp : exp '['
340 /* We need to save the current_type value. */
341 { const char *arrayname;
342 int arrayfieldindex;
343 arrayfieldindex = is_pascal_string_type (
344 current_type, NULL, NULL,
345 NULL, NULL, &arrayname);
346 if (arrayfieldindex)
347 {
348 struct stoken stringsval;
349 stringsval.ptr = alloca (strlen (arrayname) + 1);
350 stringsval.length = strlen (arrayname);
351 strcpy (stringsval.ptr, arrayname);
352 current_type = TYPE_FIELD_TYPE (current_type,
353 arrayfieldindex - 1);
354 write_exp_elt_opcode (STRUCTOP_STRUCT);
355 write_exp_string (stringsval);
356 write_exp_elt_opcode (STRUCTOP_STRUCT);
357 }
358 push_current_type (); }
359 exp1 ']'
360 { pop_current_type ();
361 write_exp_elt_opcode (BINOP_SUBSCRIPT);
362 if (current_type)
363 current_type = TYPE_TARGET_TYPE (current_type); }
364 ;
365
366 exp : exp '('
367 /* This is to save the value of arglist_len
368 being accumulated by an outer function call. */
369 { push_current_type ();
370 start_arglist (); }
371 arglist ')' %prec ARROW
372 { write_exp_elt_opcode (OP_FUNCALL);
373 write_exp_elt_longcst ((LONGEST) end_arglist ());
374 write_exp_elt_opcode (OP_FUNCALL);
375 pop_current_type ();
376 if (current_type)
377 current_type = TYPE_TARGET_TYPE (current_type);
378 }
379 ;
380
381 arglist :
382 | exp
383 { arglist_len = 1; }
384 | arglist ',' exp %prec ABOVE_COMMA
385 { arglist_len++; }
386 ;
387
388 exp : type '(' exp ')' %prec UNARY
389 { if (current_type)
390 {
391 /* Allow automatic dereference of classes. */
392 if ((TYPE_CODE (current_type) == TYPE_CODE_PTR)
393 && (TYPE_CODE (TYPE_TARGET_TYPE (current_type)) == TYPE_CODE_CLASS)
394 && (TYPE_CODE ($1) == TYPE_CODE_CLASS))
395 write_exp_elt_opcode (UNOP_IND);
396 }
397 write_exp_elt_opcode (UNOP_CAST);
398 write_exp_elt_type ($1);
399 write_exp_elt_opcode (UNOP_CAST);
400 current_type = $1; }
401 ;
402
403 exp : '(' exp1 ')'
404 { }
405 ;
406
407 /* Binary operators in order of decreasing precedence. */
408
409 exp : exp '*' exp
410 { write_exp_elt_opcode (BINOP_MUL); }
411 ;
412
413 exp : exp '/' {
414 if (current_type && is_integral_type (current_type))
415 leftdiv_is_integer = 1;
416 }
417 exp
418 {
419 if (leftdiv_is_integer && current_type
420 && is_integral_type (current_type))
421 {
422 write_exp_elt_opcode (UNOP_CAST);
423 write_exp_elt_type (parse_type->builtin_long_double);
424 current_type = parse_type->builtin_long_double;
425 write_exp_elt_opcode (UNOP_CAST);
426 leftdiv_is_integer = 0;
427 }
428
429 write_exp_elt_opcode (BINOP_DIV);
430 }
431 ;
432
433 exp : exp DIV exp
434 { write_exp_elt_opcode (BINOP_INTDIV); }
435 ;
436
437 exp : exp MOD exp
438 { write_exp_elt_opcode (BINOP_REM); }
439 ;
440
441 exp : exp '+' exp
442 { write_exp_elt_opcode (BINOP_ADD); }
443 ;
444
445 exp : exp '-' exp
446 { write_exp_elt_opcode (BINOP_SUB); }
447 ;
448
449 exp : exp LSH exp
450 { write_exp_elt_opcode (BINOP_LSH); }
451 ;
452
453 exp : exp RSH exp
454 { write_exp_elt_opcode (BINOP_RSH); }
455 ;
456
457 exp : exp '=' exp
458 { write_exp_elt_opcode (BINOP_EQUAL);
459 current_type = parse_type->builtin_bool;
460 }
461 ;
462
463 exp : exp NOTEQUAL exp
464 { write_exp_elt_opcode (BINOP_NOTEQUAL);
465 current_type = parse_type->builtin_bool;
466 }
467 ;
468
469 exp : exp LEQ exp
470 { write_exp_elt_opcode (BINOP_LEQ);
471 current_type = parse_type->builtin_bool;
472 }
473 ;
474
475 exp : exp GEQ exp
476 { write_exp_elt_opcode (BINOP_GEQ);
477 current_type = parse_type->builtin_bool;
478 }
479 ;
480
481 exp : exp '<' exp
482 { write_exp_elt_opcode (BINOP_LESS);
483 current_type = parse_type->builtin_bool;
484 }
485 ;
486
487 exp : exp '>' exp
488 { write_exp_elt_opcode (BINOP_GTR);
489 current_type = parse_type->builtin_bool;
490 }
491 ;
492
493 exp : exp ANDAND exp
494 { write_exp_elt_opcode (BINOP_BITWISE_AND); }
495 ;
496
497 exp : exp XOR exp
498 { write_exp_elt_opcode (BINOP_BITWISE_XOR); }
499 ;
500
501 exp : exp OR exp
502 { write_exp_elt_opcode (BINOP_BITWISE_IOR); }
503 ;
504
505 exp : exp ASSIGN exp
506 { write_exp_elt_opcode (BINOP_ASSIGN); }
507 ;
508
509 exp : TRUEKEYWORD
510 { write_exp_elt_opcode (OP_BOOL);
511 write_exp_elt_longcst ((LONGEST) $1);
512 current_type = parse_type->builtin_bool;
513 write_exp_elt_opcode (OP_BOOL); }
514 ;
515
516 exp : FALSEKEYWORD
517 { write_exp_elt_opcode (OP_BOOL);
518 write_exp_elt_longcst ((LONGEST) $1);
519 current_type = parse_type->builtin_bool;
520 write_exp_elt_opcode (OP_BOOL); }
521 ;
522
523 exp : INT
524 { write_exp_elt_opcode (OP_LONG);
525 write_exp_elt_type ($1.type);
526 current_type = $1.type;
527 write_exp_elt_longcst ((LONGEST)($1.val));
528 write_exp_elt_opcode (OP_LONG); }
529 ;
530
531 exp : NAME_OR_INT
532 { YYSTYPE val;
533 parse_number ($1.stoken.ptr,
534 $1.stoken.length, 0, &val);
535 write_exp_elt_opcode (OP_LONG);
536 write_exp_elt_type (val.typed_val_int.type);
537 current_type = val.typed_val_int.type;
538 write_exp_elt_longcst ((LONGEST)
539 val.typed_val_int.val);
540 write_exp_elt_opcode (OP_LONG);
541 }
542 ;
543
544
545 exp : FLOAT
546 { write_exp_elt_opcode (OP_DOUBLE);
547 write_exp_elt_type ($1.type);
548 current_type = $1.type;
549 write_exp_elt_dblcst ($1.dval);
550 write_exp_elt_opcode (OP_DOUBLE); }
551 ;
552
553 exp : variable
554 ;
555
556 exp : VARIABLE
557 /* Already written by write_dollar_variable.
558 Handle current_type. */
559 { if (intvar) {
560 struct value * val, * mark;
561
562 mark = value_mark ();
563 val = value_of_internalvar (parse_gdbarch,
564 intvar);
565 current_type = value_type (val);
566 value_release_to_mark (mark);
567 }
568 }
569 ;
570
571 exp : SIZEOF '(' type ')' %prec UNARY
572 { write_exp_elt_opcode (OP_LONG);
573 write_exp_elt_type (parse_type->builtin_int);
574 CHECK_TYPEDEF ($3);
575 write_exp_elt_longcst ((LONGEST) TYPE_LENGTH ($3));
576 write_exp_elt_opcode (OP_LONG); }
577 ;
578
579 exp : SIZEOF '(' exp ')' %prec UNARY
580 { write_exp_elt_opcode (UNOP_SIZEOF); }
581
582 exp : STRING
583 { /* C strings are converted into array constants with
584 an explicit null byte added at the end. Thus
585 the array upper bound is the string length.
586 There is no such thing in C as a completely empty
587 string. */
588 char *sp = $1.ptr; int count = $1.length;
589 while (count-- > 0)
590 {
591 write_exp_elt_opcode (OP_LONG);
592 write_exp_elt_type (parse_type->builtin_char);
593 write_exp_elt_longcst ((LONGEST)(*sp++));
594 write_exp_elt_opcode (OP_LONG);
595 }
596 write_exp_elt_opcode (OP_LONG);
597 write_exp_elt_type (parse_type->builtin_char);
598 write_exp_elt_longcst ((LONGEST)'\0');
599 write_exp_elt_opcode (OP_LONG);
600 write_exp_elt_opcode (OP_ARRAY);
601 write_exp_elt_longcst ((LONGEST) 0);
602 write_exp_elt_longcst ((LONGEST) ($1.length));
603 write_exp_elt_opcode (OP_ARRAY); }
604 ;
605
606 /* Object pascal */
607 exp : THIS
608 {
609 struct value * this_val;
610 struct type * this_type;
611 write_exp_elt_opcode (OP_THIS);
612 write_exp_elt_opcode (OP_THIS);
613 /* We need type of this. */
614 this_val = value_of_this_silent (parse_language);
615 if (this_val)
616 this_type = value_type (this_val);
617 else
618 this_type = NULL;
619 if (this_type)
620 {
621 if (TYPE_CODE (this_type) == TYPE_CODE_PTR)
622 {
623 this_type = TYPE_TARGET_TYPE (this_type);
624 write_exp_elt_opcode (UNOP_IND);
625 }
626 }
627
628 current_type = this_type;
629 }
630 ;
631
632 /* end of object pascal. */
633
634 block : BLOCKNAME
635 {
636 if ($1.sym != 0)
637 $$ = SYMBOL_BLOCK_VALUE ($1.sym);
638 else
639 {
640 struct symtab *tem =
641 lookup_symtab (copy_name ($1.stoken));
642 if (tem)
643 $$ = BLOCKVECTOR_BLOCK (BLOCKVECTOR (tem),
644 STATIC_BLOCK);
645 else
646 error (_("No file or function \"%s\"."),
647 copy_name ($1.stoken));
648 }
649 }
650 ;
651
652 block : block COLONCOLON name
653 { struct symbol *tem
654 = lookup_symbol (copy_name ($3), $1,
655 VAR_DOMAIN, (int *) NULL);
656 if (!tem || SYMBOL_CLASS (tem) != LOC_BLOCK)
657 error (_("No function \"%s\" in specified context."),
658 copy_name ($3));
659 $$ = SYMBOL_BLOCK_VALUE (tem); }
660 ;
661
662 variable: block COLONCOLON name
663 { struct symbol *sym;
664 sym = lookup_symbol (copy_name ($3), $1,
665 VAR_DOMAIN, (int *) NULL);
666 if (sym == 0)
667 error (_("No symbol \"%s\" in specified context."),
668 copy_name ($3));
669
670 write_exp_elt_opcode (OP_VAR_VALUE);
671 /* block_found is set by lookup_symbol. */
672 write_exp_elt_block (block_found);
673 write_exp_elt_sym (sym);
674 write_exp_elt_opcode (OP_VAR_VALUE); }
675 ;
676
677 qualified_name: typebase COLONCOLON name
678 {
679 struct type *type = $1;
680 if (TYPE_CODE (type) != TYPE_CODE_STRUCT
681 && TYPE_CODE (type) != TYPE_CODE_UNION)
682 error (_("`%s' is not defined as an aggregate type."),
683 TYPE_NAME (type));
684
685 write_exp_elt_opcode (OP_SCOPE);
686 write_exp_elt_type (type);
687 write_exp_string ($3);
688 write_exp_elt_opcode (OP_SCOPE);
689 }
690 ;
691
692 variable: qualified_name
693 | COLONCOLON name
694 {
695 char *name = copy_name ($2);
696 struct symbol *sym;
697 struct minimal_symbol *msymbol;
698
699 sym =
700 lookup_symbol (name, (const struct block *) NULL,
701 VAR_DOMAIN, (int *) NULL);
702 if (sym)
703 {
704 write_exp_elt_opcode (OP_VAR_VALUE);
705 write_exp_elt_block (NULL);
706 write_exp_elt_sym (sym);
707 write_exp_elt_opcode (OP_VAR_VALUE);
708 break;
709 }
710
711 msymbol = lookup_minimal_symbol (name, NULL, NULL);
712 if (msymbol != NULL)
713 write_exp_msymbol (msymbol);
714 else if (!have_full_symbols ()
715 && !have_partial_symbols ())
716 error (_("No symbol table is loaded. "
717 "Use the \"file\" command."));
718 else
719 error (_("No symbol \"%s\" in current context."),
720 name);
721 }
722 ;
723
724 variable: name_not_typename
725 { struct symbol *sym = $1.sym;
726
727 if (sym)
728 {
729 if (symbol_read_needs_frame (sym))
730 {
731 if (innermost_block == 0
732 || contained_in (block_found,
733 innermost_block))
734 innermost_block = block_found;
735 }
736
737 write_exp_elt_opcode (OP_VAR_VALUE);
738 /* We want to use the selected frame, not
739 another more inner frame which happens to
740 be in the same block. */
741 write_exp_elt_block (NULL);
742 write_exp_elt_sym (sym);
743 write_exp_elt_opcode (OP_VAR_VALUE);
744 current_type = sym->type; }
745 else if ($1.is_a_field_of_this)
746 {
747 struct value * this_val;
748 struct type * this_type;
749 /* Object pascal: it hangs off of `this'. Must
750 not inadvertently convert from a method call
751 to data ref. */
752 if (innermost_block == 0
753 || contained_in (block_found,
754 innermost_block))
755 innermost_block = block_found;
756 write_exp_elt_opcode (OP_THIS);
757 write_exp_elt_opcode (OP_THIS);
758 write_exp_elt_opcode (STRUCTOP_PTR);
759 write_exp_string ($1.stoken);
760 write_exp_elt_opcode (STRUCTOP_PTR);
761 /* We need type of this. */
762 this_val = value_of_this_silent (parse_language);
763 if (this_val)
764 this_type = value_type (this_val);
765 else
766 this_type = NULL;
767 if (this_type)
768 current_type = lookup_struct_elt_type (
769 this_type,
770 copy_name ($1.stoken), 0);
771 else
772 current_type = NULL;
773 }
774 else
775 {
776 struct minimal_symbol *msymbol;
777 char *arg = copy_name ($1.stoken);
778
779 msymbol =
780 lookup_minimal_symbol (arg, NULL, NULL);
781 if (msymbol != NULL)
782 write_exp_msymbol (msymbol);
783 else if (!have_full_symbols ()
784 && !have_partial_symbols ())
785 error (_("No symbol table is loaded. "
786 "Use the \"file\" command."));
787 else
788 error (_("No symbol \"%s\" in current context."),
789 copy_name ($1.stoken));
790 }
791 }
792 ;
793
794
795 ptype : typebase
796 ;
797
798 /* We used to try to recognize more pointer to member types here, but
799 that didn't work (shift/reduce conflicts meant that these rules never
800 got executed). The problem is that
801 int (foo::bar::baz::bizzle)
802 is a function type but
803 int (foo::bar::baz::bizzle::*)
804 is a pointer to member type. Stroustrup loses again! */
805
806 type : ptype
807 ;
808
809 typebase /* Implements (approximately): (type-qualifier)* type-specifier */
810 : '^' typebase
811 { $$ = lookup_pointer_type ($2); }
812 | TYPENAME
813 { $$ = $1.type; }
814 | STRUCT name
815 { $$ = lookup_struct (copy_name ($2),
816 expression_context_block); }
817 | CLASS name
818 { $$ = lookup_struct (copy_name ($2),
819 expression_context_block); }
820 /* "const" and "volatile" are curently ignored. A type qualifier
821 after the type is handled in the ptype rule. I think these could
822 be too. */
823 ;
824
825 name : NAME { $$ = $1.stoken; }
826 | BLOCKNAME { $$ = $1.stoken; }
827 | TYPENAME { $$ = $1.stoken; }
828 | NAME_OR_INT { $$ = $1.stoken; }
829 ;
830
831 name_not_typename : NAME
832 | BLOCKNAME
833 /* These would be useful if name_not_typename was useful, but it is just
834 a fake for "variable", so these cause reduce/reduce conflicts because
835 the parser can't tell whether NAME_OR_INT is a name_not_typename (=variable,
836 =exp) or just an exp. If name_not_typename was ever used in an lvalue
837 context where only a name could occur, this might be useful.
838 | NAME_OR_INT
839 */
840 ;
841
842 %%
843
844 /* Take care of parsing a number (anything that starts with a digit).
845 Set yylval and return the token type; update lexptr.
846 LEN is the number of characters in it. */
847
848 /*** Needs some error checking for the float case ***/
849
850 static int
851 parse_number (char *p, int len, int parsed_float, YYSTYPE *putithere)
852 {
853 /* FIXME: Shouldn't these be unsigned? We don't deal with negative values
854 here, and we do kind of silly things like cast to unsigned. */
855 LONGEST n = 0;
856 LONGEST prevn = 0;
857 ULONGEST un;
858
859 int i = 0;
860 int c;
861 int base = input_radix;
862 int unsigned_p = 0;
863
864 /* Number of "L" suffixes encountered. */
865 int long_p = 0;
866
867 /* We have found a "L" or "U" suffix. */
868 int found_suffix = 0;
869
870 ULONGEST high_bit;
871 struct type *signed_type;
872 struct type *unsigned_type;
873
874 if (parsed_float)
875 {
876 if (! parse_c_float (parse_gdbarch, p, len,
877 &putithere->typed_val_float.dval,
878 &putithere->typed_val_float.type))
879 return ERROR;
880 return FLOAT;
881 }
882
883 /* Handle base-switching prefixes 0x, 0t, 0d, 0. */
884 if (p[0] == '0')
885 switch (p[1])
886 {
887 case 'x':
888 case 'X':
889 if (len >= 3)
890 {
891 p += 2;
892 base = 16;
893 len -= 2;
894 }
895 break;
896
897 case 't':
898 case 'T':
899 case 'd':
900 case 'D':
901 if (len >= 3)
902 {
903 p += 2;
904 base = 10;
905 len -= 2;
906 }
907 break;
908
909 default:
910 base = 8;
911 break;
912 }
913
914 while (len-- > 0)
915 {
916 c = *p++;
917 if (c >= 'A' && c <= 'Z')
918 c += 'a' - 'A';
919 if (c != 'l' && c != 'u')
920 n *= base;
921 if (c >= '0' && c <= '9')
922 {
923 if (found_suffix)
924 return ERROR;
925 n += i = c - '0';
926 }
927 else
928 {
929 if (base > 10 && c >= 'a' && c <= 'f')
930 {
931 if (found_suffix)
932 return ERROR;
933 n += i = c - 'a' + 10;
934 }
935 else if (c == 'l')
936 {
937 ++long_p;
938 found_suffix = 1;
939 }
940 else if (c == 'u')
941 {
942 unsigned_p = 1;
943 found_suffix = 1;
944 }
945 else
946 return ERROR; /* Char not a digit */
947 }
948 if (i >= base)
949 return ERROR; /* Invalid digit in this base. */
950
951 /* Portably test for overflow (only works for nonzero values, so make
952 a second check for zero). FIXME: Can't we just make n and prevn
953 unsigned and avoid this? */
954 if (c != 'l' && c != 'u' && (prevn >= n) && n != 0)
955 unsigned_p = 1; /* Try something unsigned. */
956
957 /* Portably test for unsigned overflow.
958 FIXME: This check is wrong; for example it doesn't find overflow
959 on 0x123456789 when LONGEST is 32 bits. */
960 if (c != 'l' && c != 'u' && n != 0)
961 {
962 if ((unsigned_p && (ULONGEST) prevn >= (ULONGEST) n))
963 error (_("Numeric constant too large."));
964 }
965 prevn = n;
966 }
967
968 /* An integer constant is an int, a long, or a long long. An L
969 suffix forces it to be long; an LL suffix forces it to be long
970 long. If not forced to a larger size, it gets the first type of
971 the above that it fits in. To figure out whether it fits, we
972 shift it right and see whether anything remains. Note that we
973 can't shift sizeof (LONGEST) * HOST_CHAR_BIT bits or more in one
974 operation, because many compilers will warn about such a shift
975 (which always produces a zero result). Sometimes gdbarch_int_bit
976 or gdbarch_long_bit will be that big, sometimes not. To deal with
977 the case where it is we just always shift the value more than
978 once, with fewer bits each time. */
979
980 un = (ULONGEST)n >> 2;
981 if (long_p == 0
982 && (un >> (gdbarch_int_bit (parse_gdbarch) - 2)) == 0)
983 {
984 high_bit = ((ULONGEST)1) << (gdbarch_int_bit (parse_gdbarch) - 1);
985
986 /* A large decimal (not hex or octal) constant (between INT_MAX
987 and UINT_MAX) is a long or unsigned long, according to ANSI,
988 never an unsigned int, but this code treats it as unsigned
989 int. This probably should be fixed. GCC gives a warning on
990 such constants. */
991
992 unsigned_type = parse_type->builtin_unsigned_int;
993 signed_type = parse_type->builtin_int;
994 }
995 else if (long_p <= 1
996 && (un >> (gdbarch_long_bit (parse_gdbarch) - 2)) == 0)
997 {
998 high_bit = ((ULONGEST)1) << (gdbarch_long_bit (parse_gdbarch) - 1);
999 unsigned_type = parse_type->builtin_unsigned_long;
1000 signed_type = parse_type->builtin_long;
1001 }
1002 else
1003 {
1004 int shift;
1005 if (sizeof (ULONGEST) * HOST_CHAR_BIT
1006 < gdbarch_long_long_bit (parse_gdbarch))
1007 /* A long long does not fit in a LONGEST. */
1008 shift = (sizeof (ULONGEST) * HOST_CHAR_BIT - 1);
1009 else
1010 shift = (gdbarch_long_long_bit (parse_gdbarch) - 1);
1011 high_bit = (ULONGEST) 1 << shift;
1012 unsigned_type = parse_type->builtin_unsigned_long_long;
1013 signed_type = parse_type->builtin_long_long;
1014 }
1015
1016 putithere->typed_val_int.val = n;
1017
1018 /* If the high bit of the worked out type is set then this number
1019 has to be unsigned. */
1020
1021 if (unsigned_p || (n & high_bit))
1022 {
1023 putithere->typed_val_int.type = unsigned_type;
1024 }
1025 else
1026 {
1027 putithere->typed_val_int.type = signed_type;
1028 }
1029
1030 return INT;
1031 }
1032
1033
1034 struct type_push
1035 {
1036 struct type *stored;
1037 struct type_push *next;
1038 };
1039
1040 static struct type_push *tp_top = NULL;
1041
1042 static void
1043 push_current_type (void)
1044 {
1045 struct type_push *tpnew;
1046 tpnew = (struct type_push *) malloc (sizeof (struct type_push));
1047 tpnew->next = tp_top;
1048 tpnew->stored = current_type;
1049 current_type = NULL;
1050 tp_top = tpnew;
1051 }
1052
1053 static void
1054 pop_current_type (void)
1055 {
1056 struct type_push *tp = tp_top;
1057 if (tp)
1058 {
1059 current_type = tp->stored;
1060 tp_top = tp->next;
1061 free (tp);
1062 }
1063 }
1064
1065 struct token
1066 {
1067 char *operator;
1068 int token;
1069 enum exp_opcode opcode;
1070 };
1071
1072 static const struct token tokentab3[] =
1073 {
1074 {"shr", RSH, BINOP_END},
1075 {"shl", LSH, BINOP_END},
1076 {"and", ANDAND, BINOP_END},
1077 {"div", DIV, BINOP_END},
1078 {"not", NOT, BINOP_END},
1079 {"mod", MOD, BINOP_END},
1080 {"inc", INCREMENT, BINOP_END},
1081 {"dec", DECREMENT, BINOP_END},
1082 {"xor", XOR, BINOP_END}
1083 };
1084
1085 static const struct token tokentab2[] =
1086 {
1087 {"or", OR, BINOP_END},
1088 {"<>", NOTEQUAL, BINOP_END},
1089 {"<=", LEQ, BINOP_END},
1090 {">=", GEQ, BINOP_END},
1091 {":=", ASSIGN, BINOP_END},
1092 {"::", COLONCOLON, BINOP_END} };
1093
1094 /* Allocate uppercased var: */
1095 /* make an uppercased copy of tokstart. */
1096 static char *
1097 uptok (char *tokstart, int namelen)
1098 {
1099 int i;
1100 char *uptokstart = (char *)malloc(namelen+1);
1101 for (i = 0;i <= namelen;i++)
1102 {
1103 if ((tokstart[i]>='a' && tokstart[i]<='z'))
1104 uptokstart[i] = tokstart[i]-('a'-'A');
1105 else
1106 uptokstart[i] = tokstart[i];
1107 }
1108 uptokstart[namelen]='\0';
1109 return uptokstart;
1110 }
1111
1112 /* This is set if the previously-returned token was a structure
1113 operator '.'. This is used only when parsing to
1114 do field name completion. */
1115 static int last_was_structop;
1116
1117 /* Read one token, getting characters through lexptr. */
1118
1119 static int
1120 yylex (void)
1121 {
1122 int c;
1123 int namelen;
1124 unsigned int i;
1125 char *tokstart;
1126 char *uptokstart;
1127 char *tokptr;
1128 int explen, tempbufindex;
1129 static char *tempbuf;
1130 static int tempbufsize;
1131 int saw_structop = last_was_structop;
1132
1133 last_was_structop = 0;
1134 retry:
1135
1136 prev_lexptr = lexptr;
1137
1138 tokstart = lexptr;
1139 explen = strlen (lexptr);
1140 /* See if it is a special token of length 3. */
1141 if (explen > 2)
1142 for (i = 0; i < sizeof (tokentab3) / sizeof (tokentab3[0]); i++)
1143 if (strncasecmp (tokstart, tokentab3[i].operator, 3) == 0
1144 && (!isalpha (tokentab3[i].operator[0]) || explen == 3
1145 || (!isalpha (tokstart[3])
1146 && !isdigit (tokstart[3]) && tokstart[3] != '_')))
1147 {
1148 lexptr += 3;
1149 yylval.opcode = tokentab3[i].opcode;
1150 return tokentab3[i].token;
1151 }
1152
1153 /* See if it is a special token of length 2. */
1154 if (explen > 1)
1155 for (i = 0; i < sizeof (tokentab2) / sizeof (tokentab2[0]); i++)
1156 if (strncasecmp (tokstart, tokentab2[i].operator, 2) == 0
1157 && (!isalpha (tokentab2[i].operator[0]) || explen == 2
1158 || (!isalpha (tokstart[2])
1159 && !isdigit (tokstart[2]) && tokstart[2] != '_')))
1160 {
1161 lexptr += 2;
1162 yylval.opcode = tokentab2[i].opcode;
1163 return tokentab2[i].token;
1164 }
1165
1166 switch (c = *tokstart)
1167 {
1168 case 0:
1169 if (saw_structop && search_field)
1170 return COMPLETE;
1171 else
1172 return 0;
1173
1174 case ' ':
1175 case '\t':
1176 case '\n':
1177 lexptr++;
1178 goto retry;
1179
1180 case '\'':
1181 /* We either have a character constant ('0' or '\177' for example)
1182 or we have a quoted symbol reference ('foo(int,int)' in object pascal
1183 for example). */
1184 lexptr++;
1185 c = *lexptr++;
1186 if (c == '\\')
1187 c = parse_escape (parse_gdbarch, &lexptr);
1188 else if (c == '\'')
1189 error (_("Empty character constant."));
1190
1191 yylval.typed_val_int.val = c;
1192 yylval.typed_val_int.type = parse_type->builtin_char;
1193
1194 c = *lexptr++;
1195 if (c != '\'')
1196 {
1197 namelen = skip_quoted (tokstart) - tokstart;
1198 if (namelen > 2)
1199 {
1200 lexptr = tokstart + namelen;
1201 if (lexptr[-1] != '\'')
1202 error (_("Unmatched single quote."));
1203 namelen -= 2;
1204 tokstart++;
1205 uptokstart = uptok(tokstart,namelen);
1206 goto tryname;
1207 }
1208 error (_("Invalid character constant."));
1209 }
1210 return INT;
1211
1212 case '(':
1213 paren_depth++;
1214 lexptr++;
1215 return c;
1216
1217 case ')':
1218 if (paren_depth == 0)
1219 return 0;
1220 paren_depth--;
1221 lexptr++;
1222 return c;
1223
1224 case ',':
1225 if (comma_terminates && paren_depth == 0)
1226 return 0;
1227 lexptr++;
1228 return c;
1229
1230 case '.':
1231 /* Might be a floating point number. */
1232 if (lexptr[1] < '0' || lexptr[1] > '9')
1233 {
1234 if (in_parse_field)
1235 last_was_structop = 1;
1236 goto symbol; /* Nope, must be a symbol. */
1237 }
1238
1239 /* FALL THRU into number case. */
1240
1241 case '0':
1242 case '1':
1243 case '2':
1244 case '3':
1245 case '4':
1246 case '5':
1247 case '6':
1248 case '7':
1249 case '8':
1250 case '9':
1251 {
1252 /* It's a number. */
1253 int got_dot = 0, got_e = 0, toktype;
1254 char *p = tokstart;
1255 int hex = input_radix > 10;
1256
1257 if (c == '0' && (p[1] == 'x' || p[1] == 'X'))
1258 {
1259 p += 2;
1260 hex = 1;
1261 }
1262 else if (c == '0' && (p[1]=='t' || p[1]=='T'
1263 || p[1]=='d' || p[1]=='D'))
1264 {
1265 p += 2;
1266 hex = 0;
1267 }
1268
1269 for (;; ++p)
1270 {
1271 /* This test includes !hex because 'e' is a valid hex digit
1272 and thus does not indicate a floating point number when
1273 the radix is hex. */
1274 if (!hex && !got_e && (*p == 'e' || *p == 'E'))
1275 got_dot = got_e = 1;
1276 /* This test does not include !hex, because a '.' always indicates
1277 a decimal floating point number regardless of the radix. */
1278 else if (!got_dot && *p == '.')
1279 got_dot = 1;
1280 else if (got_e && (p[-1] == 'e' || p[-1] == 'E')
1281 && (*p == '-' || *p == '+'))
1282 /* This is the sign of the exponent, not the end of the
1283 number. */
1284 continue;
1285 /* We will take any letters or digits. parse_number will
1286 complain if past the radix, or if L or U are not final. */
1287 else if ((*p < '0' || *p > '9')
1288 && ((*p < 'a' || *p > 'z')
1289 && (*p < 'A' || *p > 'Z')))
1290 break;
1291 }
1292 toktype = parse_number (tokstart,
1293 p - tokstart, got_dot | got_e, &yylval);
1294 if (toktype == ERROR)
1295 {
1296 char *err_copy = (char *) alloca (p - tokstart + 1);
1297
1298 memcpy (err_copy, tokstart, p - tokstart);
1299 err_copy[p - tokstart] = 0;
1300 error (_("Invalid number \"%s\"."), err_copy);
1301 }
1302 lexptr = p;
1303 return toktype;
1304 }
1305
1306 case '+':
1307 case '-':
1308 case '*':
1309 case '/':
1310 case '|':
1311 case '&':
1312 case '^':
1313 case '~':
1314 case '!':
1315 case '@':
1316 case '<':
1317 case '>':
1318 case '[':
1319 case ']':
1320 case '?':
1321 case ':':
1322 case '=':
1323 case '{':
1324 case '}':
1325 symbol:
1326 lexptr++;
1327 return c;
1328
1329 case '"':
1330
1331 /* Build the gdb internal form of the input string in tempbuf,
1332 translating any standard C escape forms seen. Note that the
1333 buffer is null byte terminated *only* for the convenience of
1334 debugging gdb itself and printing the buffer contents when
1335 the buffer contains no embedded nulls. Gdb does not depend
1336 upon the buffer being null byte terminated, it uses the length
1337 string instead. This allows gdb to handle C strings (as well
1338 as strings in other languages) with embedded null bytes. */
1339
1340 tokptr = ++tokstart;
1341 tempbufindex = 0;
1342
1343 do {
1344 /* Grow the static temp buffer if necessary, including allocating
1345 the first one on demand. */
1346 if (tempbufindex + 1 >= tempbufsize)
1347 {
1348 tempbuf = (char *) realloc (tempbuf, tempbufsize += 64);
1349 }
1350
1351 switch (*tokptr)
1352 {
1353 case '\0':
1354 case '"':
1355 /* Do nothing, loop will terminate. */
1356 break;
1357 case '\\':
1358 tokptr++;
1359 c = parse_escape (parse_gdbarch, &tokptr);
1360 if (c == -1)
1361 {
1362 continue;
1363 }
1364 tempbuf[tempbufindex++] = c;
1365 break;
1366 default:
1367 tempbuf[tempbufindex++] = *tokptr++;
1368 break;
1369 }
1370 } while ((*tokptr != '"') && (*tokptr != '\0'));
1371 if (*tokptr++ != '"')
1372 {
1373 error (_("Unterminated string in expression."));
1374 }
1375 tempbuf[tempbufindex] = '\0'; /* See note above. */
1376 yylval.sval.ptr = tempbuf;
1377 yylval.sval.length = tempbufindex;
1378 lexptr = tokptr;
1379 return (STRING);
1380 }
1381
1382 if (!(c == '_' || c == '$'
1383 || (c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z')))
1384 /* We must have come across a bad character (e.g. ';'). */
1385 error (_("Invalid character '%c' in expression."), c);
1386
1387 /* It's a name. See how long it is. */
1388 namelen = 0;
1389 for (c = tokstart[namelen];
1390 (c == '_' || c == '$' || (c >= '0' && c <= '9')
1391 || (c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z') || c == '<');)
1392 {
1393 /* Template parameter lists are part of the name.
1394 FIXME: This mishandles `print $a<4&&$a>3'. */
1395 if (c == '<')
1396 {
1397 int i = namelen;
1398 int nesting_level = 1;
1399 while (tokstart[++i])
1400 {
1401 if (tokstart[i] == '<')
1402 nesting_level++;
1403 else if (tokstart[i] == '>')
1404 {
1405 if (--nesting_level == 0)
1406 break;
1407 }
1408 }
1409 if (tokstart[i] == '>')
1410 namelen = i;
1411 else
1412 break;
1413 }
1414
1415 /* do NOT uppercase internals because of registers !!! */
1416 c = tokstart[++namelen];
1417 }
1418
1419 uptokstart = uptok(tokstart,namelen);
1420
1421 /* The token "if" terminates the expression and is NOT
1422 removed from the input stream. */
1423 if (namelen == 2 && uptokstart[0] == 'I' && uptokstart[1] == 'F')
1424 {
1425 free (uptokstart);
1426 return 0;
1427 }
1428
1429 lexptr += namelen;
1430
1431 tryname:
1432
1433 /* Catch specific keywords. Should be done with a data structure. */
1434 switch (namelen)
1435 {
1436 case 6:
1437 if (strcmp (uptokstart, "OBJECT") == 0)
1438 {
1439 free (uptokstart);
1440 return CLASS;
1441 }
1442 if (strcmp (uptokstart, "RECORD") == 0)
1443 {
1444 free (uptokstart);
1445 return STRUCT;
1446 }
1447 if (strcmp (uptokstart, "SIZEOF") == 0)
1448 {
1449 free (uptokstart);
1450 return SIZEOF;
1451 }
1452 break;
1453 case 5:
1454 if (strcmp (uptokstart, "CLASS") == 0)
1455 {
1456 free (uptokstart);
1457 return CLASS;
1458 }
1459 if (strcmp (uptokstart, "FALSE") == 0)
1460 {
1461 yylval.lval = 0;
1462 free (uptokstart);
1463 return FALSEKEYWORD;
1464 }
1465 break;
1466 case 4:
1467 if (strcmp (uptokstart, "TRUE") == 0)
1468 {
1469 yylval.lval = 1;
1470 free (uptokstart);
1471 return TRUEKEYWORD;
1472 }
1473 if (strcmp (uptokstart, "SELF") == 0)
1474 {
1475 /* Here we search for 'this' like
1476 inserted in FPC stabs debug info. */
1477 static const char this_name[] = "this";
1478
1479 if (lookup_symbol (this_name, expression_context_block,
1480 VAR_DOMAIN, (int *) NULL))
1481 {
1482 free (uptokstart);
1483 return THIS;
1484 }
1485 }
1486 break;
1487 default:
1488 break;
1489 }
1490
1491 yylval.sval.ptr = tokstart;
1492 yylval.sval.length = namelen;
1493
1494 if (*tokstart == '$')
1495 {
1496 char c;
1497 /* $ is the normal prefix for pascal hexadecimal values
1498 but this conflicts with the GDB use for debugger variables
1499 so in expression to enter hexadecimal values
1500 we still need to use C syntax with 0xff */
1501 write_dollar_variable (yylval.sval);
1502 c = tokstart[namelen];
1503 tokstart[namelen] = 0;
1504 intvar = lookup_only_internalvar (++tokstart);
1505 --tokstart;
1506 tokstart[namelen] = c;
1507 free (uptokstart);
1508 return VARIABLE;
1509 }
1510
1511 /* Use token-type BLOCKNAME for symbols that happen to be defined as
1512 functions or symtabs. If this is not so, then ...
1513 Use token-type TYPENAME for symbols that happen to be defined
1514 currently as names of types; NAME for other symbols.
1515 The caller is not constrained to care about the distinction. */
1516 {
1517 char *tmp = copy_name (yylval.sval);
1518 struct symbol *sym;
1519 int is_a_field_of_this = 0;
1520 int is_a_field = 0;
1521 int hextype;
1522
1523
1524 if (search_field && current_type)
1525 is_a_field = (lookup_struct_elt_type (current_type, tmp, 1) != NULL);
1526 if (is_a_field || in_parse_field)
1527 sym = NULL;
1528 else
1529 sym = lookup_symbol (tmp, expression_context_block,
1530 VAR_DOMAIN, &is_a_field_of_this);
1531 /* second chance uppercased (as Free Pascal does). */
1532 if (!sym && !is_a_field_of_this && !is_a_field)
1533 {
1534 for (i = 0; i <= namelen; i++)
1535 {
1536 if ((tmp[i] >= 'a' && tmp[i] <= 'z'))
1537 tmp[i] -= ('a'-'A');
1538 }
1539 if (search_field && current_type)
1540 is_a_field = (lookup_struct_elt_type (current_type, tmp, 1) != NULL);
1541 if (is_a_field || in_parse_field)
1542 sym = NULL;
1543 else
1544 sym = lookup_symbol (tmp, expression_context_block,
1545 VAR_DOMAIN, &is_a_field_of_this);
1546 if (sym || is_a_field_of_this || is_a_field)
1547 for (i = 0; i <= namelen; i++)
1548 {
1549 if ((tokstart[i] >= 'a' && tokstart[i] <= 'z'))
1550 tokstart[i] -= ('a'-'A');
1551 }
1552 }
1553 /* Third chance Capitalized (as GPC does). */
1554 if (!sym && !is_a_field_of_this && !is_a_field)
1555 {
1556 for (i = 0; i <= namelen; i++)
1557 {
1558 if (i == 0)
1559 {
1560 if ((tmp[i] >= 'a' && tmp[i] <= 'z'))
1561 tmp[i] -= ('a'-'A');
1562 }
1563 else
1564 if ((tmp[i] >= 'A' && tmp[i] <= 'Z'))
1565 tmp[i] -= ('A'-'a');
1566 }
1567 if (search_field && current_type)
1568 is_a_field = (lookup_struct_elt_type (current_type, tmp, 1) != NULL);
1569 if (is_a_field || in_parse_field)
1570 sym = NULL;
1571 else
1572 sym = lookup_symbol (tmp, expression_context_block,
1573 VAR_DOMAIN, &is_a_field_of_this);
1574 if (sym || is_a_field_of_this || is_a_field)
1575 for (i = 0; i <= namelen; i++)
1576 {
1577 if (i == 0)
1578 {
1579 if ((tokstart[i] >= 'a' && tokstart[i] <= 'z'))
1580 tokstart[i] -= ('a'-'A');
1581 }
1582 else
1583 if ((tokstart[i] >= 'A' && tokstart[i] <= 'Z'))
1584 tokstart[i] -= ('A'-'a');
1585 }
1586 }
1587
1588 if (is_a_field)
1589 {
1590 tempbuf = (char *) realloc (tempbuf, namelen + 1);
1591 strncpy (tempbuf, tokstart, namelen); tempbuf [namelen] = 0;
1592 yylval.sval.ptr = tempbuf;
1593 yylval.sval.length = namelen;
1594 free (uptokstart);
1595 return FIELDNAME;
1596 }
1597 /* Call lookup_symtab, not lookup_partial_symtab, in case there are
1598 no psymtabs (coff, xcoff, or some future change to blow away the
1599 psymtabs once once symbols are read). */
1600 if ((sym && SYMBOL_CLASS (sym) == LOC_BLOCK)
1601 || lookup_symtab (tmp))
1602 {
1603 yylval.ssym.sym = sym;
1604 yylval.ssym.is_a_field_of_this = is_a_field_of_this;
1605 free (uptokstart);
1606 return BLOCKNAME;
1607 }
1608 if (sym && SYMBOL_CLASS (sym) == LOC_TYPEDEF)
1609 {
1610 #if 1
1611 /* Despite the following flaw, we need to keep this code enabled.
1612 Because we can get called from check_stub_method, if we don't
1613 handle nested types then it screws many operations in any
1614 program which uses nested types. */
1615 /* In "A::x", if x is a member function of A and there happens
1616 to be a type (nested or not, since the stabs don't make that
1617 distinction) named x, then this code incorrectly thinks we
1618 are dealing with nested types rather than a member function. */
1619
1620 char *p;
1621 char *namestart;
1622 struct symbol *best_sym;
1623
1624 /* Look ahead to detect nested types. This probably should be
1625 done in the grammar, but trying seemed to introduce a lot
1626 of shift/reduce and reduce/reduce conflicts. It's possible
1627 that it could be done, though. Or perhaps a non-grammar, but
1628 less ad hoc, approach would work well. */
1629
1630 /* Since we do not currently have any way of distinguishing
1631 a nested type from a non-nested one (the stabs don't tell
1632 us whether a type is nested), we just ignore the
1633 containing type. */
1634
1635 p = lexptr;
1636 best_sym = sym;
1637 while (1)
1638 {
1639 /* Skip whitespace. */
1640 while (*p == ' ' || *p == '\t' || *p == '\n')
1641 ++p;
1642 if (*p == ':' && p[1] == ':')
1643 {
1644 /* Skip the `::'. */
1645 p += 2;
1646 /* Skip whitespace. */
1647 while (*p == ' ' || *p == '\t' || *p == '\n')
1648 ++p;
1649 namestart = p;
1650 while (*p == '_' || *p == '$' || (*p >= '0' && *p <= '9')
1651 || (*p >= 'a' && *p <= 'z')
1652 || (*p >= 'A' && *p <= 'Z'))
1653 ++p;
1654 if (p != namestart)
1655 {
1656 struct symbol *cur_sym;
1657 /* As big as the whole rest of the expression, which is
1658 at least big enough. */
1659 char *ncopy = alloca (strlen (tmp)+strlen (namestart)+3);
1660 char *tmp1;
1661
1662 tmp1 = ncopy;
1663 memcpy (tmp1, tmp, strlen (tmp));
1664 tmp1 += strlen (tmp);
1665 memcpy (tmp1, "::", 2);
1666 tmp1 += 2;
1667 memcpy (tmp1, namestart, p - namestart);
1668 tmp1[p - namestart] = '\0';
1669 cur_sym = lookup_symbol (ncopy, expression_context_block,
1670 VAR_DOMAIN, (int *) NULL);
1671 if (cur_sym)
1672 {
1673 if (SYMBOL_CLASS (cur_sym) == LOC_TYPEDEF)
1674 {
1675 best_sym = cur_sym;
1676 lexptr = p;
1677 }
1678 else
1679 break;
1680 }
1681 else
1682 break;
1683 }
1684 else
1685 break;
1686 }
1687 else
1688 break;
1689 }
1690
1691 yylval.tsym.type = SYMBOL_TYPE (best_sym);
1692 #else /* not 0 */
1693 yylval.tsym.type = SYMBOL_TYPE (sym);
1694 #endif /* not 0 */
1695 free (uptokstart);
1696 return TYPENAME;
1697 }
1698 yylval.tsym.type
1699 = language_lookup_primitive_type_by_name (parse_language,
1700 parse_gdbarch, tmp);
1701 if (yylval.tsym.type != NULL)
1702 {
1703 free (uptokstart);
1704 return TYPENAME;
1705 }
1706
1707 /* Input names that aren't symbols but ARE valid hex numbers,
1708 when the input radix permits them, can be names or numbers
1709 depending on the parse. Note we support radixes > 16 here. */
1710 if (!sym
1711 && ((tokstart[0] >= 'a' && tokstart[0] < 'a' + input_radix - 10)
1712 || (tokstart[0] >= 'A' && tokstart[0] < 'A' + input_radix - 10)))
1713 {
1714 YYSTYPE newlval; /* Its value is ignored. */
1715 hextype = parse_number (tokstart, namelen, 0, &newlval);
1716 if (hextype == INT)
1717 {
1718 yylval.ssym.sym = sym;
1719 yylval.ssym.is_a_field_of_this = is_a_field_of_this;
1720 free (uptokstart);
1721 return NAME_OR_INT;
1722 }
1723 }
1724
1725 free(uptokstart);
1726 /* Any other kind of symbol. */
1727 yylval.ssym.sym = sym;
1728 yylval.ssym.is_a_field_of_this = is_a_field_of_this;
1729 return NAME;
1730 }
1731 }
1732
1733 void
1734 yyerror (char *msg)
1735 {
1736 if (prev_lexptr)
1737 lexptr = prev_lexptr;
1738
1739 error (_("A %s in expression, near `%s'."), (msg ? msg : "error"), lexptr);
1740 }