* corelow.c, exec.c, inftarg.c, m3-nat.c, op50-rom.c, procfs.c,
[deliverable/binutils-gdb.git] / gdb / f-exp.y
CommitLineData
a91a6192
SS
1/* YACC parser for Fortran expressions, for GDB.
2 Copyright 1986, 1989, 1990, 1991, 1993, 1994
3 Free Software Foundation, Inc.
4 Contributed by Motorola. Adapted from the C parser by Farooq Butt
5 (fmbutt@engage.sps.mot.com).
6
7This file is part of GDB.
8
9This program is free software; you can redistribute it and/or modify
10it under the terms of the GNU General Public License as published by
11the Free Software Foundation; either version 2 of the License, or
12(at your option) any later version.
13
14This program is distributed in the hope that it will be useful,
15but WITHOUT ANY WARRANTY; without even the implied warranty of
16MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
17GNU General Public License for more details.
18
19You should have received a copy of the GNU General Public License
20along with this program; if not, write to the Free Software
21Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. */
22
23/* This was blantantly ripped off the C expression parser, please
24 be aware of that as you look at its basic structure -FMB */
25
26/* Parse a F77 expression from text in a string,
27 and return the result as a struct expression pointer.
28 That structure contains arithmetic operations in reverse polish,
29 with constants represented by operations that are followed by special data.
30 See expression.h for the details of the format.
31 What is important here is that it can be built up sequentially
32 during the process of parsing; the lower levels of the tree always
33 come first in the result.
34
35 Note that malloc's and realloc's in this file are transformed to
36 xmalloc and xrealloc respectively by the same sed command in the
37 makefile that remaps any other malloc/realloc inserted by the parser
38 generator. Doing this with #defines and trying to control the interaction
39 with include files (<malloc.h> and <stdlib.h> for example) just became
40 too messy, particularly when such includes can be inserted at random
41 times by the parser generator. */
42
43%{
44
45#include "defs.h"
22d7f91e 46#include <string.h>
a91a6192 47#include "expression.h"
a91a6192 48#include "value.h"
22d7f91e 49#include "parser-defs.h"
a91a6192
SS
50#include "language.h"
51#include "f-lang.h"
52#include "bfd.h" /* Required by objfiles.h. */
53#include "symfile.h" /* Required by objfiles.h. */
54#include "objfiles.h" /* For have_full_symbols and have_partial_symbols */
55
56/* Remap normal yacc parser interface names (yyparse, yylex, yyerror, etc),
57 as well as gratuitiously global symbol names, so we can have multiple
58 yacc generated parsers in gdb. Note that these are only the variables
59 produced by yacc. If other parser generators (bison, byacc, etc) produce
60 additional global names that conflict at link time, then those parser
61 generators need to be fixed instead of adding those names to this list. */
62
63#define yymaxdepth f_maxdepth
64#define yyparse f_parse
65#define yylex f_lex
66#define yyerror f_error
67#define yylval f_lval
68#define yychar f_char
69#define yydebug f_debug
70#define yypact f_pact
71#define yyr1 f_r1
72#define yyr2 f_r2
73#define yydef f_def
74#define yychk f_chk
75#define yypgo f_pgo
76#define yyact f_act
77#define yyexca f_exca
78#define yyerrflag f_errflag
79#define yynerrs f_nerrs
80#define yyps f_ps
81#define yypv f_pv
82#define yys f_s
83#define yy_yys f_yys
84#define yystate f_state
85#define yytmp f_tmp
86#define yyv f_v
87#define yy_yyv f_yyv
88#define yyval f_val
89#define yylloc f_lloc
90#define yyreds f_reds /* With YYDEBUG defined */
91#define yytoks f_toks /* With YYDEBUG defined */
92
93#ifndef YYDEBUG
94#define YYDEBUG 1 /* Default to no yydebug support */
95#endif
96
97int yyparse PARAMS ((void));
98
99static int yylex PARAMS ((void));
100
101void yyerror PARAMS ((char *));
102
103%}
104
105/* Although the yacc "value" of an expression is not used,
106 since the result is stored in the structure being created,
107 other node types do have values. */
108
109%union
110 {
111 LONGEST lval;
112 struct {
113 LONGEST val;
114 struct type *type;
115 } typed_val;
116 double dval;
117 struct symbol *sym;
118 struct type *tval;
119 struct stoken sval;
120 struct ttype tsym;
121 struct symtoken ssym;
122 int voidval;
123 struct block *bval;
124 enum exp_opcode opcode;
125 struct internalvar *ivar;
126
127 struct type **tvec;
128 int *ivec;
129 }
130
131%{
132/* YYSTYPE gets defined by %union */
133static int parse_number PARAMS ((char *, int, int, YYSTYPE *));
134%}
135
136%type <voidval> exp type_exp start variable
137%type <tval> type typebase
138%type <tvec> nonempty_typelist
139/* %type <bval> block */
140
141/* Fancy type parsing. */
142%type <voidval> func_mod direct_abs_decl abs_decl
143%type <tval> ptype
144
145%token <typed_val> INT
146%token <dval> FLOAT
147
148/* Both NAME and TYPENAME tokens represent symbols in the input,
149 and both convey their data as strings.
150 But a TYPENAME is a string that happens to be defined as a typedef
151 or builtin type name (such as int or char)
152 and a NAME is any other symbol.
153 Contexts where this distinction is not important can use the
154 nonterminal "name", which matches either NAME or TYPENAME. */
155
156%token <sval> STRING_LITERAL
157%token <lval> BOOLEAN_LITERAL
158%token <ssym> NAME
159%token <tsym> TYPENAME
160%type <sval> name
161%type <ssym> name_not_typename
162%type <tsym> typename
163
164/* A NAME_OR_INT is a symbol which is not known in the symbol table,
165 but which would parse as a valid number in the current input radix.
166 E.g. "c" when input_radix==16. Depending on the parse, it will be
167 turned into a name or into a number. */
168
169%token <ssym> NAME_OR_INT
170
171%token SIZEOF
172%token ERROR
173
174/* Special type cases, put in to allow the parser to distinguish different
175 legal basetypes. */
176%token INT_KEYWORD INT_S2_KEYWORD LOGICAL_S1_KEYWORD LOGICAL_S2_KEYWORD
177%token LOGICAL_KEYWORD REAL_KEYWORD REAL_S8_KEYWORD REAL_S16_KEYWORD
178%token COMPLEX_S8_KEYWORD COMPLEX_S16_KEYWORD COMPLEX_S32_KEYWORD
179%token BOOL_AND BOOL_OR BOOL_NOT
180%token <lval> LAST REGNAME CHARACTER
181
182%token <ivar> VARIABLE
183
184%token <opcode> ASSIGN_MODIFY
185
186%left ','
187%left ABOVE_COMMA
188%right '=' ASSIGN_MODIFY
189%right '?'
190%left BOOL_OR
191%right BOOL_NOT
192%left BOOL_AND
193%left '|'
194%left '^'
195%left '&'
196%left EQUAL NOTEQUAL
197%left LESSTHAN GREATERTHAN LEQ GEQ
198%left LSH RSH
199%left '@'
200%left '+' '-'
201%left '*' '/' '%'
202%right UNARY
203%right '('
204
205\f
206%%
207
208start : exp
209 | type_exp
210 ;
211
212type_exp: type
213 { write_exp_elt_opcode(OP_TYPE);
214 write_exp_elt_type($1);
215 write_exp_elt_opcode(OP_TYPE); }
216 ;
217
a91a6192
SS
218exp : '(' exp ')'
219 { }
220 ;
221
222/* Expressions, not including the comma operator. */
223exp : '*' exp %prec UNARY
224 { write_exp_elt_opcode (UNOP_IND); }
225
226exp : '&' exp %prec UNARY
227 { write_exp_elt_opcode (UNOP_ADDR); }
228
229exp : '-' exp %prec UNARY
230 { write_exp_elt_opcode (UNOP_NEG); }
231 ;
232
233exp : BOOL_NOT exp %prec UNARY
234 { write_exp_elt_opcode (UNOP_LOGICAL_NOT); }
235 ;
236
237exp : '~' exp %prec UNARY
238 { write_exp_elt_opcode (UNOP_COMPLEMENT); }
239 ;
240
241exp : SIZEOF exp %prec UNARY
242 { write_exp_elt_opcode (UNOP_SIZEOF); }
243 ;
244
245/* No more explicit array operators, we treat everything in F77 as
246 a function call. The disambiguation as to whether we are
247 doing a subscript operation or a function call is done
248 later in eval.c. */
249
250exp : exp '('
251 { start_arglist (); }
252 arglist ')'
253 { write_exp_elt_opcode (OP_F77_UNDETERMINED_ARGLIST);
254 write_exp_elt_longcst ((LONGEST) end_arglist ());
255 write_exp_elt_opcode (OP_F77_UNDETERMINED_ARGLIST); }
256 ;
257
258arglist :
259 ;
260
261arglist : exp
262 { arglist_len = 1; }
263 ;
264
265arglist : substring
266 { arglist_len = 2;}
267
268arglist : arglist ',' exp %prec ABOVE_COMMA
269 { arglist_len++; }
270 ;
271
272substring: exp ':' exp %prec ABOVE_COMMA
273 { }
274 ;
275
276
277complexnum: exp ',' exp
278 { }
279 ;
280
281exp : '(' complexnum ')'
282 { write_exp_elt_opcode(OP_F77_LITERAL_COMPLEX); }
283 ;
284
285exp : '(' type ')' exp %prec UNARY
286 { write_exp_elt_opcode (UNOP_CAST);
287 write_exp_elt_type ($2);
288 write_exp_elt_opcode (UNOP_CAST); }
289 ;
290
291/* Binary operators in order of decreasing precedence. */
292
293exp : exp '@' exp
294 { write_exp_elt_opcode (BINOP_REPEAT); }
295 ;
296
297exp : exp '*' exp
298 { write_exp_elt_opcode (BINOP_MUL); }
299 ;
300
301exp : exp '/' exp
302 { write_exp_elt_opcode (BINOP_DIV); }
303 ;
304
305exp : exp '%' exp
306 { write_exp_elt_opcode (BINOP_REM); }
307 ;
308
309exp : exp '+' exp
310 { write_exp_elt_opcode (BINOP_ADD); }
311 ;
312
313exp : exp '-' exp
314 { write_exp_elt_opcode (BINOP_SUB); }
315 ;
316
317exp : exp LSH exp
318 { write_exp_elt_opcode (BINOP_LSH); }
319 ;
320
321exp : exp RSH exp
322 { write_exp_elt_opcode (BINOP_RSH); }
323 ;
324
325exp : exp EQUAL exp
326 { write_exp_elt_opcode (BINOP_EQUAL); }
327 ;
328
329exp : exp NOTEQUAL exp
330 { write_exp_elt_opcode (BINOP_NOTEQUAL); }
331 ;
332
333exp : exp LEQ exp
334 { write_exp_elt_opcode (BINOP_LEQ); }
335 ;
336
337exp : exp GEQ exp
338 { write_exp_elt_opcode (BINOP_GEQ); }
339 ;
340
341exp : exp LESSTHAN exp
342 { write_exp_elt_opcode (BINOP_LESS); }
343 ;
344
345exp : exp GREATERTHAN exp
346 { write_exp_elt_opcode (BINOP_GTR); }
347 ;
348
349exp : exp '&' exp
350 { write_exp_elt_opcode (BINOP_BITWISE_AND); }
351 ;
352
353exp : exp '^' exp
354 { write_exp_elt_opcode (BINOP_BITWISE_XOR); }
355 ;
356
357exp : exp '|' exp
358 { write_exp_elt_opcode (BINOP_BITWISE_IOR); }
359 ;
360
361exp : exp BOOL_AND exp
362 { write_exp_elt_opcode (BINOP_LOGICAL_AND); }
363 ;
364
365
366exp : exp BOOL_OR exp
367 { write_exp_elt_opcode (BINOP_LOGICAL_OR); }
368 ;
369
370exp : exp '=' exp
371 { write_exp_elt_opcode (BINOP_ASSIGN); }
372 ;
373
374exp : exp ASSIGN_MODIFY exp
375 { write_exp_elt_opcode (BINOP_ASSIGN_MODIFY);
376 write_exp_elt_opcode ($2);
377 write_exp_elt_opcode (BINOP_ASSIGN_MODIFY); }
378 ;
379
380exp : INT
381 { write_exp_elt_opcode (OP_LONG);
382 write_exp_elt_type ($1.type);
383 write_exp_elt_longcst ((LONGEST)($1.val));
384 write_exp_elt_opcode (OP_LONG); }
385 ;
386
387exp : NAME_OR_INT
388 { YYSTYPE val;
389 parse_number ($1.stoken.ptr, $1.stoken.length, 0, &val);
390 write_exp_elt_opcode (OP_LONG);
391 write_exp_elt_type (val.typed_val.type);
392 write_exp_elt_longcst ((LONGEST)val.typed_val.val);
22d7f91e 393 write_exp_elt_opcode (OP_LONG); }
a91a6192
SS
394 ;
395
396exp : FLOAT
397 { write_exp_elt_opcode (OP_DOUBLE);
398 write_exp_elt_type (builtin_type_f_real_s8);
399 write_exp_elt_dblcst ($1);
400 write_exp_elt_opcode (OP_DOUBLE); }
401 ;
402
403exp : variable
404 ;
405
406exp : LAST
407 { write_exp_elt_opcode (OP_LAST);
408 write_exp_elt_longcst ((LONGEST) $1);
409 write_exp_elt_opcode (OP_LAST); }
410 ;
411
412exp : REGNAME
413 { write_exp_elt_opcode (OP_REGISTER);
414 write_exp_elt_longcst ((LONGEST) $1);
415 write_exp_elt_opcode (OP_REGISTER); }
416 ;
417
418exp : VARIABLE
419 { write_exp_elt_opcode (OP_INTERNALVAR);
420 write_exp_elt_intern ($1);
421 write_exp_elt_opcode (OP_INTERNALVAR); }
422 ;
423
424exp : SIZEOF '(' type ')' %prec UNARY
425 { write_exp_elt_opcode (OP_LONG);
426 write_exp_elt_type (builtin_type_f_integer);
427 write_exp_elt_longcst ((LONGEST) TYPE_LENGTH ($3));
428 write_exp_elt_opcode (OP_LONG); }
429 ;
430
431exp : BOOLEAN_LITERAL
432 { write_exp_elt_opcode (OP_BOOL);
433 write_exp_elt_longcst ((LONGEST) $1);
434 write_exp_elt_opcode (OP_BOOL);
435 }
436 ;
437
438exp : STRING_LITERAL
439 { /* In F77, we encounter string literals
440 basically in only one place:
441 when we are setting up manual parameter
442 lists to functions we call by hand or
443 when setting string vars to manual values.
444 These are character*N type variables.
445 They are treated specially behind the
446 scenes. Remember that the literal strings's
447 OPs are being emitted in reverse order, thus
448 we first have the elements and then
449 the array descriptor itself. */
450 char *sp = $1.ptr; int count = $1.length;
451
452 while (count-- > 0)
453 {
454 write_exp_elt_opcode (OP_LONG);
455 write_exp_elt_type (builtin_type_f_character);
456 write_exp_elt_longcst ((LONGEST)(*sp++));
457 write_exp_elt_opcode (OP_LONG);
458 }
459 write_exp_elt_opcode (OP_ARRAY);
460 write_exp_elt_longcst ((LONGEST) 1);
461 write_exp_elt_longcst ((LONGEST) ($1.length));
462 write_exp_elt_opcode (OP_ARRAY);
463 }
464
465 ;
466
467variable: name_not_typename
468 { struct symbol *sym = $1.sym;
469
470 if (sym)
471 {
472 if (symbol_read_needs_frame (sym))
473 {
474 if (innermost_block == 0 ||
475 contained_in (block_found,
476 innermost_block))
477 innermost_block = block_found;
478 }
479 write_exp_elt_opcode (OP_VAR_VALUE);
4c664b8d
PS
480 /* We want to use the selected frame, not
481 another more inner frame which happens to
482 be in the same block. */
483 write_exp_elt_block (NULL);
a91a6192
SS
484 write_exp_elt_sym (sym);
485 write_exp_elt_opcode (OP_VAR_VALUE);
486 break;
487 }
488 else
489 {
490 struct minimal_symbol *msymbol;
491 register char *arg = copy_name ($1.stoken);
492
493 msymbol = lookup_minimal_symbol (arg, NULL);
494 if (msymbol != NULL)
495 {
496 write_exp_msymbol (msymbol,
497 lookup_function_type (builtin_type_int),
498 builtin_type_int);
499 }
500 else if (!have_full_symbols () && !have_partial_symbols ())
501 error ("No symbol table is loaded. Use the \"file\" command.");
502 else
503 error ("No symbol \"%s\" in current context.",
504 copy_name ($1.stoken));
505 }
506 }
507 ;
508
509
510type : ptype
511 ;
512
513ptype : typebase
514 | typebase abs_decl
515 {
516 /* This is where the interesting stuff happens. */
517 int done = 0;
518 int array_size;
519 struct type *follow_type = $1;
520 struct type *range_type;
521
522 while (!done)
523 switch (pop_type ())
524 {
525 case tp_end:
526 done = 1;
527 break;
528 case tp_pointer:
529 follow_type = lookup_pointer_type (follow_type);
530 break;
531 case tp_reference:
532 follow_type = lookup_reference_type (follow_type);
533 break;
534 case tp_array:
535 array_size = pop_type_int ();
536 if (array_size != -1)
537 {
538 range_type =
539 create_range_type ((struct type *) NULL,
540 builtin_type_f_integer, 0,
541 array_size - 1);
542 follow_type =
543 create_array_type ((struct type *) NULL,
544 follow_type, range_type);
545 }
546 else
547 follow_type = lookup_pointer_type (follow_type);
548 break;
549 case tp_function:
550 follow_type = lookup_function_type (follow_type);
551 break;
552 }
553 $$ = follow_type;
554 }
555 ;
556
557abs_decl: '*'
558 { push_type (tp_pointer); $$ = 0; }
559 | '*' abs_decl
560 { push_type (tp_pointer); $$ = $2; }
561 | '&'
562 { push_type (tp_reference); $$ = 0; }
563 | '&' abs_decl
564 { push_type (tp_reference); $$ = $2; }
565 | direct_abs_decl
566 ;
567
568direct_abs_decl: '(' abs_decl ')'
569 { $$ = $2; }
570 | direct_abs_decl func_mod
571 { push_type (tp_function); }
572 | func_mod
573 { push_type (tp_function); }
574 ;
575
576func_mod: '(' ')'
577 { $$ = 0; }
578 | '(' nonempty_typelist ')'
579 { free ((PTR)$2); $$ = 0; }
580 ;
581
582typebase /* Implements (approximately): (type-qualifier)* type-specifier */
583 : TYPENAME
584 { $$ = $1.type; }
585 | INT_KEYWORD
586 { $$ = builtin_type_f_integer; }
587 | INT_S2_KEYWORD
588 { $$ = builtin_type_f_integer_s2; }
589 | CHARACTER
590 { $$ = builtin_type_f_character; }
591 | LOGICAL_KEYWORD
592 { $$ = builtin_type_f_logical;}
593 | LOGICAL_S2_KEYWORD
594 { $$ = builtin_type_f_logical_s2;}
595 | LOGICAL_S1_KEYWORD
596 { $$ = builtin_type_f_logical_s1;}
597 | REAL_KEYWORD
598 { $$ = builtin_type_f_real;}
599 | REAL_S8_KEYWORD
600 { $$ = builtin_type_f_real_s8;}
601 | REAL_S16_KEYWORD
602 { $$ = builtin_type_f_real_s16;}
603 | COMPLEX_S8_KEYWORD
604 { $$ = builtin_type_f_complex_s8;}
605 | COMPLEX_S16_KEYWORD
606 { $$ = builtin_type_f_complex_s16;}
607 | COMPLEX_S32_KEYWORD
608 { $$ = builtin_type_f_complex_s32;}
609 ;
610
611typename: TYPENAME
612 ;
613
614nonempty_typelist
615 : type
616 { $$ = (struct type **) malloc (sizeof (struct type *) * 2);
617 $<ivec>$[0] = 1; /* Number of types in vector */
618 $$[1] = $1;
619 }
620 | nonempty_typelist ',' type
621 { int len = sizeof (struct type *) * (++($<ivec>1[0]) + 1);
622 $$ = (struct type **) realloc ((char *) $1, len);
623 $$[$<ivec>$[0]] = $3;
624 }
625 ;
626
627name : NAME
628 { $$ = $1.stoken; }
629 | TYPENAME
630 { $$ = $1.stoken; }
631 | NAME_OR_INT
632 { $$ = $1.stoken; }
633 ;
634
635name_not_typename : NAME
636/* These would be useful if name_not_typename was useful, but it is just
637 a fake for "variable", so these cause reduce/reduce conflicts because
638 the parser can't tell whether NAME_OR_INT is a name_not_typename (=variable,
639 =exp) or just an exp. If name_not_typename was ever used in an lvalue
640 context where only a name could occur, this might be useful.
641 | NAME_OR_INT
642 */
643 ;
644
645%%
646
647/* Take care of parsing a number (anything that starts with a digit).
648 Set yylval and return the token type; update lexptr.
649 LEN is the number of characters in it. */
650
651/*** Needs some error checking for the float case ***/
652
653static int
654parse_number (p, len, parsed_float, putithere)
655 register char *p;
656 register int len;
657 int parsed_float;
658 YYSTYPE *putithere;
659{
660 register LONGEST n = 0;
661 register LONGEST prevn = 0;
662 register int i;
663 register int c;
664 register int base = input_radix;
665 int unsigned_p = 0;
666 int long_p = 0;
667 unsigned LONGEST high_bit;
668 struct type *signed_type;
669 struct type *unsigned_type;
670
671 if (parsed_float)
672 {
673 /* It's a float since it contains a point or an exponent. */
22d7f91e
SS
674 /* [dD] is not understood as an exponent by atof, change it to 'e'. */
675 char *tmp, *tmp2;
676
677 tmp = strsave (p);
678 for (tmp2 = tmp; *tmp2; ++tmp2)
679 if (*tmp2 == 'd' || *tmp2 == 'D')
680 *tmp2 = 'e';
681 putithere->dval = atof (tmp);
682 free (tmp);
a91a6192
SS
683 return FLOAT;
684 }
685
686 /* Handle base-switching prefixes 0x, 0t, 0d, 0 */
687 if (p[0] == '0')
688 switch (p[1])
689 {
690 case 'x':
691 case 'X':
692 if (len >= 3)
693 {
694 p += 2;
695 base = 16;
696 len -= 2;
697 }
698 break;
699
700 case 't':
701 case 'T':
702 case 'd':
703 case 'D':
704 if (len >= 3)
705 {
706 p += 2;
707 base = 10;
708 len -= 2;
709 }
710 break;
711
712 default:
713 base = 8;
714 break;
715 }
716
717 while (len-- > 0)
718 {
719 c = *p++;
720 if (c >= 'A' && c <= 'Z')
721 c += 'a' - 'A';
722 if (c != 'l' && c != 'u')
723 n *= base;
724 if (c >= '0' && c <= '9')
725 n += i = c - '0';
726 else
727 {
728 if (base > 10 && c >= 'a' && c <= 'f')
729 n += i = c - 'a' + 10;
730 else if (len == 0 && c == 'l')
731 long_p = 1;
732 else if (len == 0 && c == 'u')
733 unsigned_p = 1;
734 else
735 return ERROR; /* Char not a digit */
736 }
737 if (i >= base)
738 return ERROR; /* Invalid digit in this base */
739
740 /* Portably test for overflow (only works for nonzero values, so make
741 a second check for zero). */
742 if ((prevn >= n) && n != 0)
743 unsigned_p=1; /* Try something unsigned */
744 /* If range checking enabled, portably test for unsigned overflow. */
745 if (RANGE_CHECK && n != 0)
746 {
747 if ((unsigned_p && (unsigned)prevn >= (unsigned)n))
748 range_error("Overflow on numeric constant.");
749 }
750 prevn = n;
751 }
752
753 /* If the number is too big to be an int, or it's got an l suffix
754 then it's a long. Work out if this has to be a long by
755 shifting right and and seeing if anything remains, and the
756 target int size is different to the target long size.
757
758 In the expression below, we could have tested
759 (n >> TARGET_INT_BIT)
760 to see if it was zero,
761 but too many compilers warn about that, when ints and longs
762 are the same size. So we shift it twice, with fewer bits
763 each time, for the same result. */
764
765 if ((TARGET_INT_BIT != TARGET_LONG_BIT
766 && ((n >> 2) >> (TARGET_INT_BIT-2))) /* Avoid shift warning */
767 || long_p)
768 {
769 high_bit = ((unsigned LONGEST)1) << (TARGET_LONG_BIT-1);
770 unsigned_type = builtin_type_unsigned_long;
771 signed_type = builtin_type_long;
772 }
773 else
774 {
775 high_bit = ((unsigned LONGEST)1) << (TARGET_INT_BIT-1);
776 unsigned_type = builtin_type_unsigned_int;
777 signed_type = builtin_type_int;
778 }
779
780 putithere->typed_val.val = n;
781
782 /* If the high bit of the worked out type is set then this number
783 has to be unsigned. */
784
785 if (unsigned_p || (n & high_bit))
786 putithere->typed_val.type = unsigned_type;
787 else
788 putithere->typed_val.type = signed_type;
789
790 return INT;
791}
792
793struct token
794{
795 char *operator;
796 int token;
797 enum exp_opcode opcode;
798};
799
800static const struct token dot_ops[] =
801{
802 { ".and.", BOOL_AND, BINOP_END },
803 { ".AND.", BOOL_AND, BINOP_END },
804 { ".or.", BOOL_OR, BINOP_END },
805 { ".OR.", BOOL_OR, BINOP_END },
806 { ".not.", BOOL_NOT, BINOP_END },
807 { ".NOT.", BOOL_NOT, BINOP_END },
808 { ".eq.", EQUAL, BINOP_END },
809 { ".EQ.", EQUAL, BINOP_END },
810 { ".eqv.", EQUAL, BINOP_END },
811 { ".NEQV.", NOTEQUAL, BINOP_END },
812 { ".neqv.", NOTEQUAL, BINOP_END },
813 { ".EQV.", EQUAL, BINOP_END },
814 { ".ne.", NOTEQUAL, BINOP_END },
815 { ".NE.", NOTEQUAL, BINOP_END },
816 { ".le.", LEQ, BINOP_END },
817 { ".LE.", LEQ, BINOP_END },
818 { ".ge.", GEQ, BINOP_END },
819 { ".GE.", GEQ, BINOP_END },
820 { ".gt.", GREATERTHAN, BINOP_END },
821 { ".GT.", GREATERTHAN, BINOP_END },
822 { ".lt.", LESSTHAN, BINOP_END },
823 { ".LT.", LESSTHAN, BINOP_END },
824 { NULL, 0, 0 }
825};
826
827struct f77_boolean_val
828{
829 char *name;
830 int value;
831};
832
833static const struct f77_boolean_val boolean_values[] =
834{
835 { ".true.", 1 },
836 { ".TRUE.", 1 },
837 { ".false.", 0 },
838 { ".FALSE.", 0 },
839 { NULL, 0 }
840};
841
842static const struct token f77_keywords[] =
843{
844 { "complex_16", COMPLEX_S16_KEYWORD, BINOP_END },
845 { "complex_32", COMPLEX_S32_KEYWORD, BINOP_END },
846 { "character", CHARACTER, BINOP_END },
847 { "integer_2", INT_S2_KEYWORD, BINOP_END },
848 { "logical_1", LOGICAL_S1_KEYWORD, BINOP_END },
849 { "logical_2", LOGICAL_S2_KEYWORD, BINOP_END },
850 { "complex_8", COMPLEX_S8_KEYWORD, BINOP_END },
851 { "integer", INT_KEYWORD, BINOP_END },
852 { "logical", LOGICAL_KEYWORD, BINOP_END },
853 { "real_16", REAL_S16_KEYWORD, BINOP_END },
854 { "complex", COMPLEX_S8_KEYWORD, BINOP_END },
855 { "sizeof", SIZEOF, BINOP_END },
856 { "real_8", REAL_S8_KEYWORD, BINOP_END },
857 { "real", REAL_KEYWORD, BINOP_END },
858 { NULL, 0, 0 }
859};
860
861/* Implementation of a dynamically expandable buffer for processing input
862 characters acquired through lexptr and building a value to return in
863 yylval. Ripped off from ch-exp.y */
864
865static char *tempbuf; /* Current buffer contents */
866static int tempbufsize; /* Size of allocated buffer */
867static int tempbufindex; /* Current index into buffer */
868
869#define GROWBY_MIN_SIZE 64 /* Minimum amount to grow buffer by */
870
871#define CHECKBUF(size) \
872 do { \
873 if (tempbufindex + (size) >= tempbufsize) \
874 { \
875 growbuf_by_size (size); \
876 } \
877 } while (0);
878
879
880/* Grow the static temp buffer if necessary, including allocating the first one
881 on demand. */
882
883static void
884growbuf_by_size (count)
885 int count;
886{
887 int growby;
888
889 growby = max (count, GROWBY_MIN_SIZE);
890 tempbufsize += growby;
891 if (tempbuf == NULL)
892 tempbuf = (char *) malloc (tempbufsize);
893 else
894 tempbuf = (char *) realloc (tempbuf, tempbufsize);
895}
896
897/* Blatantly ripped off from ch-exp.y. This routine recognizes F77
898 string-literals.
899
900 Recognize a string literal. A string literal is a nonzero sequence
901 of characters enclosed in matching single quotes, except that
902 a single character inside single quotes is a character literal, which
903 we reject as a string literal. To embed the terminator character inside
904 a string, it is simply doubled (I.E. 'this''is''one''string') */
905
906static int
907match_string_literal ()
908{
909 char *tokptr = lexptr;
910
911 for (tempbufindex = 0, tokptr++; *tokptr != '\0'; tokptr++)
912 {
913 CHECKBUF (1);
914 if (*tokptr == *lexptr)
915 {
916 if (*(tokptr + 1) == *lexptr)
917 tokptr++;
918 else
919 break;
920 }
921 tempbuf[tempbufindex++] = *tokptr;
922 }
923 if (*tokptr == '\0' /* no terminator */
924 || tempbufindex == 0) /* no string */
925 return 0;
926 else
927 {
928 tempbuf[tempbufindex] = '\0';
929 yylval.sval.ptr = tempbuf;
930 yylval.sval.length = tempbufindex;
931 lexptr = ++tokptr;
932 return STRING_LITERAL;
933 }
934}
935
936/* Read one token, getting characters through lexptr. */
937
938static int
939yylex ()
940{
941 int c;
942 int namelen;
943 unsigned int i,token;
944 char *tokstart;
a91a6192
SS
945
946 retry:
947
948 tokstart = lexptr;
949
950 /* First of all, let us make sure we are not dealing with the
951 special tokens .true. and .false. which evaluate to 1 and 0. */
952
953 if (*lexptr == '.')
954 {
22d7f91e 955 for (i = 0; boolean_values[i].name != NULL; i++)
a91a6192 956 {
22d7f91e
SS
957 if STREQN (tokstart, boolean_values[i].name,
958 strlen (boolean_values[i].name))
a91a6192 959 {
22d7f91e 960 lexptr += strlen (boolean_values[i].name);
a91a6192 961 yylval.lval = boolean_values[i].value;
22d7f91e 962 return BOOLEAN_LITERAL;
a91a6192
SS
963 }
964 }
965 }
966
967 /* See if it is a special .foo. operator */
968
969 for (i = 0; dot_ops[i].operator != NULL; i++)
22d7f91e 970 if (STREQN (tokstart, dot_ops[i].operator, strlen (dot_ops[i].operator)))
a91a6192 971 {
22d7f91e 972 lexptr += strlen (dot_ops[i].operator);
a91a6192
SS
973 yylval.opcode = dot_ops[i].opcode;
974 return dot_ops[i].token;
975 }
976
977 switch (c = *tokstart)
978 {
979 case 0:
980 return 0;
981
982 case ' ':
983 case '\t':
984 case '\n':
985 lexptr++;
986 goto retry;
987
988 case '\'':
989 token = match_string_literal ();
990 if (token != 0)
991 return (token);
992 break;
993
994 case '(':
995 paren_depth++;
996 lexptr++;
997 return c;
998
999 case ')':
1000 if (paren_depth == 0)
1001 return 0;
1002 paren_depth--;
1003 lexptr++;
1004 return c;
1005
1006 case ',':
1007 if (comma_terminates && paren_depth == 0)
1008 return 0;
1009 lexptr++;
1010 return c;
1011
1012 case '.':
1013 /* Might be a floating point number. */
1014 if (lexptr[1] < '0' || lexptr[1] > '9')
1015 goto symbol; /* Nope, must be a symbol. */
1016 /* FALL THRU into number case. */
1017
1018 case '0':
1019 case '1':
1020 case '2':
1021 case '3':
1022 case '4':
1023 case '5':
1024 case '6':
1025 case '7':
1026 case '8':
1027 case '9':
1028 {
1029 /* It's a number. */
1030 int got_dot = 0, got_e = 0, got_d = 0, toktype;
1031 register char *p = tokstart;
1032 int hex = input_radix > 10;
1033
1034 if (c == '0' && (p[1] == 'x' || p[1] == 'X'))
1035 {
1036 p += 2;
1037 hex = 1;
1038 }
1039 else if (c == '0' && (p[1]=='t' || p[1]=='T' || p[1]=='d' || p[1]=='D'))
1040 {
1041 p += 2;
1042 hex = 0;
1043 }
1044
1045 for (;; ++p)
1046 {
1047 if (!hex && !got_e && (*p == 'e' || *p == 'E'))
1048 got_dot = got_e = 1;
22d7f91e 1049 else if (!hex && !got_d && (*p == 'd' || *p == 'D'))
a91a6192
SS
1050 got_dot = got_d = 1;
1051 else if (!hex && !got_dot && *p == '.')
1052 got_dot = 1;
22d7f91e
SS
1053 else if ((got_e && (p[-1] == 'e' || p[-1] == 'E'))
1054 || (got_d && (p[-1] == 'd' || p[-1] == 'D'))
a91a6192
SS
1055 && (*p == '-' || *p == '+'))
1056 /* This is the sign of the exponent, not the end of the
1057 number. */
1058 continue;
1059 /* We will take any letters or digits. parse_number will
1060 complain if past the radix, or if L or U are not final. */
1061 else if ((*p < '0' || *p > '9')
1062 && ((*p < 'a' || *p > 'z')
1063 && (*p < 'A' || *p > 'Z')))
1064 break;
1065 }
1066 toktype = parse_number (tokstart, p - tokstart, got_dot|got_e|got_d,
1067 &yylval);
1068 if (toktype == ERROR)
1069 {
1070 char *err_copy = (char *) alloca (p - tokstart + 1);
1071
1072 memcpy (err_copy, tokstart, p - tokstart);
1073 err_copy[p - tokstart] = 0;
1074 error ("Invalid number \"%s\".", err_copy);
1075 }
1076 lexptr = p;
1077 return toktype;
1078 }
1079
1080 case '+':
1081 case '-':
1082 case '*':
1083 case '/':
1084 case '%':
1085 case '|':
1086 case '&':
1087 case '^':
1088 case '~':
1089 case '!':
1090 case '@':
1091 case '<':
1092 case '>':
1093 case '[':
1094 case ']':
1095 case '?':
1096 case ':':
1097 case '=':
1098 case '{':
1099 case '}':
1100 symbol:
1101 lexptr++;
1102 return c;
1103 }
1104
1105 if (!(c == '_' || c == '$'
1106 || (c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z')))
1107 /* We must have come across a bad character (e.g. ';'). */
1108 error ("Invalid character '%c' in expression.", c);
1109
1110 namelen = 0;
1111 for (c = tokstart[namelen];
1112 (c == '_' || c == '$' || (c >= '0' && c <= '9')
1113 || (c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z'));
1114 c = tokstart[++namelen]);
1115
1116 /* The token "if" terminates the expression and is NOT
1117 removed from the input stream. */
1118
1119 if (namelen == 2 && tokstart[0] == 'i' && tokstart[1] == 'f')
1120 return 0;
1121
1122 lexptr += namelen;
1123
1124 /* Handle the tokens $digits; also $ (short for $0) and $$ (short for $$1)
1125 and $$digits (equivalent to $<-digits> if you could type that).
1126 Make token type LAST, and put the number (the digits) in yylval. */
1127
1128 tryname:
1129 if (*tokstart == '$')
1130 {
1131 register int negate = 0;
1132
1133 c = 1;
1134 /* Double dollar means negate the number and add -1 as well.
1135 Thus $$ alone means -1. */
1136 if (namelen >= 2 && tokstart[1] == '$')
1137 {
1138 negate = 1;
1139 c = 2;
1140 }
1141 if (c == namelen)
1142 {
1143 /* Just dollars (one or two) */
1144 yylval.lval = - negate;
1145 return LAST;
1146 }
1147 /* Is the rest of the token digits? */
1148 for (; c < namelen; c++)
1149 if (!(tokstart[c] >= '0' && tokstart[c] <= '9'))
1150 break;
1151 if (c == namelen)
1152 {
1153 yylval.lval = atoi (tokstart + 1 + negate);
1154 if (negate)
1155 yylval.lval = - yylval.lval;
1156 return LAST;
1157 }
1158 }
1159
1160 /* Handle tokens that refer to machine registers:
1161 $ followed by a register name. */
1162
1163 if (*tokstart == '$') {
1164 for (c = 0; c < NUM_REGS; c++)
1165 if (namelen - 1 == strlen (reg_names[c])
1166 && STREQN (tokstart + 1, reg_names[c], namelen - 1))
1167 {
1168 yylval.lval = c;
1169 return REGNAME;
1170 }
1171 for (c = 0; c < num_std_regs; c++)
1172 if (namelen - 1 == strlen (std_regs[c].name)
1173 && STREQN (tokstart + 1, std_regs[c].name, namelen - 1))
1174 {
1175 yylval.lval = std_regs[c].regnum;
1176 return REGNAME;
1177 }
1178 }
1179 /* Catch specific keywords. */
1180
1181 for (i = 0; f77_keywords[i].operator != NULL; i++)
1182 if (STREQN(tokstart, f77_keywords[i].operator,
1183 strlen(f77_keywords[i].operator)))
1184 {
1185 /* lexptr += strlen(f77_keywords[i].operator); */
1186 yylval.opcode = f77_keywords[i].opcode;
1187 return f77_keywords[i].token;
1188 }
1189
1190 yylval.sval.ptr = tokstart;
1191 yylval.sval.length = namelen;
1192
1193 /* Any other names starting in $ are debugger internal variables. */
1194
1195 if (*tokstart == '$')
1196 {
1197 yylval.ivar = lookup_internalvar (copy_name (yylval.sval) + 1);
1198 return VARIABLE;
1199 }
1200
1201 /* Use token-type TYPENAME for symbols that happen to be defined
1202 currently as names of types; NAME for other symbols.
1203 The caller is not constrained to care about the distinction. */
1204 {
1205 char *tmp = copy_name (yylval.sval);
1206 struct symbol *sym;
1207 int is_a_field_of_this = 0;
1208 int hextype;
1209
1210 sym = lookup_symbol (tmp, expression_context_block,
1211 VAR_NAMESPACE,
1212 current_language->la_language == language_cplus
1213 ? &is_a_field_of_this : NULL,
1214 NULL);
1215 if (sym && SYMBOL_CLASS (sym) == LOC_TYPEDEF)
1216 {
1217 yylval.tsym.type = SYMBOL_TYPE (sym);
1218 return TYPENAME;
1219 }
1220 if ((yylval.tsym.type = lookup_primitive_typename (tmp)) != 0)
1221 return TYPENAME;
1222
1223 /* Input names that aren't symbols but ARE valid hex numbers,
1224 when the input radix permits them, can be names or numbers
1225 depending on the parse. Note we support radixes > 16 here. */
1226 if (!sym
1227 && ((tokstart[0] >= 'a' && tokstart[0] < 'a' + input_radix - 10)
1228 || (tokstart[0] >= 'A' && tokstart[0] < 'A' + input_radix - 10)))
1229 {
1230 YYSTYPE newlval; /* Its value is ignored. */
1231 hextype = parse_number (tokstart, namelen, 0, &newlval);
1232 if (hextype == INT)
1233 {
1234 yylval.ssym.sym = sym;
1235 yylval.ssym.is_a_field_of_this = is_a_field_of_this;
1236 return NAME_OR_INT;
1237 }
1238 }
1239
1240 /* Any other kind of symbol */
1241 yylval.ssym.sym = sym;
1242 yylval.ssym.is_a_field_of_this = is_a_field_of_this;
1243 return NAME;
1244 }
1245}
1246
1247void
1248yyerror (msg)
1249 char *msg;
1250{
1251 error ("A %s in expression, near `%s'.", (msg ? msg : "error"), lexptr);
1252}
This page took 0.074522 seconds and 4 git commands to generate.