2003-02-18 Elena Zannoni <ezannoni@redhat.com>
[deliverable/binutils-gdb.git] / gdb / f-exp.y
CommitLineData
c906108c 1/* YACC parser for Fortran expressions, for GDB.
b6ba6518
KB
2 Copyright 1986, 1989, 1990, 1991, 1993, 1994, 1995, 1996, 2000, 2001
3 Free Software Foundation, Inc.
4fcf66da 4
c906108c
SS
5 Contributed by Motorola. Adapted from the C parser by Farooq Butt
6 (fmbutt@engage.sps.mot.com).
7
8This file is part of GDB.
9
10This program is free software; you can redistribute it and/or modify
11it under the terms of the GNU General Public License as published by
12the Free Software Foundation; either version 2 of the License, or
13(at your option) any later version.
14
15This program is distributed in the hope that it will be useful,
16but WITHOUT ANY WARRANTY; without even the implied warranty of
17MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
18GNU General Public License for more details.
19
20You should have received a copy of the GNU General Public License
21along with this program; if not, write to the Free Software
22Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA. */
23
24/* This was blantantly ripped off the C expression parser, please
25 be aware of that as you look at its basic structure -FMB */
26
27/* Parse a F77 expression from text in a string,
28 and return the result as a struct expression pointer.
29 That structure contains arithmetic operations in reverse polish,
30 with constants represented by operations that are followed by special data.
31 See expression.h for the details of the format.
32 What is important here is that it can be built up sequentially
33 during the process of parsing; the lower levels of the tree always
34 come first in the result.
35
36 Note that malloc's and realloc's in this file are transformed to
37 xmalloc and xrealloc respectively by the same sed command in the
38 makefile that remaps any other malloc/realloc inserted by the parser
39 generator. Doing this with #defines and trying to control the interaction
40 with include files (<malloc.h> and <stdlib.h> for example) just became
41 too messy, particularly when such includes can be inserted at random
42 times by the parser generator. */
43
44%{
45
46#include "defs.h"
47#include "gdb_string.h"
48#include "expression.h"
49#include "value.h"
50#include "parser-defs.h"
51#include "language.h"
52#include "f-lang.h"
53#include "bfd.h" /* Required by objfiles.h. */
54#include "symfile.h" /* Required by objfiles.h. */
55#include "objfiles.h" /* For have_full_symbols and have_partial_symbols */
0f6e1ba6 56#include <ctype.h>
c906108c
SS
57
58/* Remap normal yacc parser interface names (yyparse, yylex, yyerror, etc),
59 as well as gratuitiously global symbol names, so we can have multiple
60 yacc generated parsers in gdb. Note that these are only the variables
61 produced by yacc. If other parser generators (bison, byacc, etc) produce
62 additional global names that conflict at link time, then those parser
63 generators need to be fixed instead of adding those names to this list. */
64
65#define yymaxdepth f_maxdepth
66#define yyparse f_parse
67#define yylex f_lex
68#define yyerror f_error
69#define yylval f_lval
70#define yychar f_char
71#define yydebug f_debug
72#define yypact f_pact
73#define yyr1 f_r1
74#define yyr2 f_r2
75#define yydef f_def
76#define yychk f_chk
77#define yypgo f_pgo
78#define yyact f_act
79#define yyexca f_exca
80#define yyerrflag f_errflag
81#define yynerrs f_nerrs
82#define yyps f_ps
83#define yypv f_pv
84#define yys f_s
85#define yy_yys f_yys
86#define yystate f_state
87#define yytmp f_tmp
88#define yyv f_v
89#define yy_yyv f_yyv
90#define yyval f_val
91#define yylloc f_lloc
92#define yyreds f_reds /* With YYDEBUG defined */
93#define yytoks f_toks /* With YYDEBUG defined */
06891d83
JT
94#define yyname f_name /* With YYDEBUG defined */
95#define yyrule f_rule /* With YYDEBUG defined */
c906108c
SS
96#define yylhs f_yylhs
97#define yylen f_yylen
98#define yydefred f_yydefred
99#define yydgoto f_yydgoto
100#define yysindex f_yysindex
101#define yyrindex f_yyrindex
102#define yygindex f_yygindex
103#define yytable f_yytable
104#define yycheck f_yycheck
105
106#ifndef YYDEBUG
f461f5cf 107#define YYDEBUG 1 /* Default to yydebug support */
c906108c
SS
108#endif
109
f461f5cf
PM
110#define YYFPRINTF parser_fprintf
111
a14ed312 112int yyparse (void);
c906108c 113
a14ed312 114static int yylex (void);
c906108c 115
a14ed312 116void yyerror (char *);
c906108c 117
a14ed312 118static void growbuf_by_size (int);
c906108c 119
a14ed312 120static int match_string_literal (void);
c906108c
SS
121
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;
135 DOUBLEST dval;
136 struct symbol *sym;
137 struct type *tval;
138 struct stoken sval;
139 struct ttype tsym;
140 struct symtoken ssym;
141 int voidval;
142 struct block *bval;
143 enum exp_opcode opcode;
144 struct internalvar *ivar;
145
146 struct type **tvec;
147 int *ivec;
148 }
149
150%{
151/* YYSTYPE gets defined by %union */
a14ed312 152static int parse_number (char *, int, int, YYSTYPE *);
c906108c
SS
153%}
154
155%type <voidval> exp type_exp start variable
156%type <tval> type typebase
157%type <tvec> nonempty_typelist
158/* %type <bval> block */
159
160/* Fancy type parsing. */
161%type <voidval> func_mod direct_abs_decl abs_decl
162%type <tval> ptype
163
164%token <typed_val> INT
165%token <dval> FLOAT
166
167/* Both NAME and TYPENAME tokens represent symbols in the input,
168 and both convey their data as strings.
169 But a TYPENAME is a string that happens to be defined as a typedef
170 or builtin type name (such as int or char)
171 and a NAME is any other symbol.
172 Contexts where this distinction is not important can use the
173 nonterminal "name", which matches either NAME or TYPENAME. */
174
175%token <sval> STRING_LITERAL
176%token <lval> BOOLEAN_LITERAL
177%token <ssym> NAME
178%token <tsym> TYPENAME
179%type <sval> name
180%type <ssym> name_not_typename
181%type <tsym> typename
182
183/* A NAME_OR_INT is a symbol which is not known in the symbol table,
184 but which would parse as a valid number in the current input radix.
185 E.g. "c" when input_radix==16. Depending on the parse, it will be
186 turned into a name or into a number. */
187
188%token <ssym> NAME_OR_INT
189
190%token SIZEOF
191%token ERROR
192
193/* Special type cases, put in to allow the parser to distinguish different
194 legal basetypes. */
195%token INT_KEYWORD INT_S2_KEYWORD LOGICAL_S1_KEYWORD LOGICAL_S2_KEYWORD
196%token LOGICAL_KEYWORD REAL_KEYWORD REAL_S8_KEYWORD REAL_S16_KEYWORD
197%token COMPLEX_S8_KEYWORD COMPLEX_S16_KEYWORD COMPLEX_S32_KEYWORD
198%token BOOL_AND BOOL_OR BOOL_NOT
199%token <lval> CHARACTER
200
201%token <voidval> VARIABLE
202
203%token <opcode> ASSIGN_MODIFY
204
205%left ','
206%left ABOVE_COMMA
207%right '=' ASSIGN_MODIFY
208%right '?'
209%left BOOL_OR
210%right BOOL_NOT
211%left BOOL_AND
212%left '|'
213%left '^'
214%left '&'
215%left EQUAL NOTEQUAL
216%left LESSTHAN GREATERTHAN LEQ GEQ
217%left LSH RSH
218%left '@'
219%left '+' '-'
220%left '*' '/' '%'
221%right UNARY
222%right '('
223
224\f
225%%
226
227start : exp
228 | type_exp
229 ;
230
231type_exp: type
232 { write_exp_elt_opcode(OP_TYPE);
233 write_exp_elt_type($1);
234 write_exp_elt_opcode(OP_TYPE); }
235 ;
236
237exp : '(' exp ')'
238 { }
239 ;
240
241/* Expressions, not including the comma operator. */
242exp : '*' exp %prec UNARY
243 { write_exp_elt_opcode (UNOP_IND); }
ef944135 244 ;
c906108c
SS
245
246exp : '&' exp %prec UNARY
247 { write_exp_elt_opcode (UNOP_ADDR); }
ef944135 248 ;
c906108c
SS
249
250exp : '-' exp %prec UNARY
251 { write_exp_elt_opcode (UNOP_NEG); }
252 ;
253
254exp : BOOL_NOT exp %prec UNARY
255 { write_exp_elt_opcode (UNOP_LOGICAL_NOT); }
256 ;
257
258exp : '~' exp %prec UNARY
259 { write_exp_elt_opcode (UNOP_COMPLEMENT); }
260 ;
261
262exp : SIZEOF exp %prec UNARY
263 { write_exp_elt_opcode (UNOP_SIZEOF); }
264 ;
265
266/* No more explicit array operators, we treat everything in F77 as
267 a function call. The disambiguation as to whether we are
268 doing a subscript operation or a function call is done
269 later in eval.c. */
270
271exp : exp '('
272 { start_arglist (); }
273 arglist ')'
274 { write_exp_elt_opcode (OP_F77_UNDETERMINED_ARGLIST);
275 write_exp_elt_longcst ((LONGEST) end_arglist ());
276 write_exp_elt_opcode (OP_F77_UNDETERMINED_ARGLIST); }
277 ;
278
279arglist :
280 ;
281
282arglist : exp
283 { arglist_len = 1; }
284 ;
285
286arglist : substring
287 { arglist_len = 2;}
ef944135 288 ;
c906108c
SS
289
290arglist : arglist ',' exp %prec ABOVE_COMMA
291 { arglist_len++; }
292 ;
293
294substring: exp ':' exp %prec ABOVE_COMMA
295 { }
296 ;
297
298
299complexnum: exp ',' exp
300 { }
301 ;
302
303exp : '(' complexnum ')'
304 { write_exp_elt_opcode(OP_COMPLEX); }
305 ;
306
307exp : '(' type ')' exp %prec UNARY
308 { write_exp_elt_opcode (UNOP_CAST);
309 write_exp_elt_type ($2);
310 write_exp_elt_opcode (UNOP_CAST); }
311 ;
312
313/* Binary operators in order of decreasing precedence. */
314
315exp : exp '@' exp
316 { write_exp_elt_opcode (BINOP_REPEAT); }
317 ;
318
319exp : exp '*' exp
320 { write_exp_elt_opcode (BINOP_MUL); }
321 ;
322
323exp : exp '/' exp
324 { write_exp_elt_opcode (BINOP_DIV); }
325 ;
326
327exp : exp '%' exp
328 { write_exp_elt_opcode (BINOP_REM); }
329 ;
330
331exp : exp '+' exp
332 { write_exp_elt_opcode (BINOP_ADD); }
333 ;
334
335exp : exp '-' exp
336 { write_exp_elt_opcode (BINOP_SUB); }
337 ;
338
339exp : exp LSH exp
340 { write_exp_elt_opcode (BINOP_LSH); }
341 ;
342
343exp : exp RSH exp
344 { write_exp_elt_opcode (BINOP_RSH); }
345 ;
346
347exp : exp EQUAL exp
348 { write_exp_elt_opcode (BINOP_EQUAL); }
349 ;
350
351exp : exp NOTEQUAL exp
352 { write_exp_elt_opcode (BINOP_NOTEQUAL); }
353 ;
354
355exp : exp LEQ exp
356 { write_exp_elt_opcode (BINOP_LEQ); }
357 ;
358
359exp : exp GEQ exp
360 { write_exp_elt_opcode (BINOP_GEQ); }
361 ;
362
363exp : exp LESSTHAN exp
364 { write_exp_elt_opcode (BINOP_LESS); }
365 ;
366
367exp : exp GREATERTHAN exp
368 { write_exp_elt_opcode (BINOP_GTR); }
369 ;
370
371exp : exp '&' exp
372 { write_exp_elt_opcode (BINOP_BITWISE_AND); }
373 ;
374
375exp : exp '^' exp
376 { write_exp_elt_opcode (BINOP_BITWISE_XOR); }
377 ;
378
379exp : exp '|' exp
380 { write_exp_elt_opcode (BINOP_BITWISE_IOR); }
381 ;
382
383exp : exp BOOL_AND exp
384 { write_exp_elt_opcode (BINOP_LOGICAL_AND); }
385 ;
386
387
388exp : exp BOOL_OR exp
389 { write_exp_elt_opcode (BINOP_LOGICAL_OR); }
390 ;
391
392exp : exp '=' exp
393 { write_exp_elt_opcode (BINOP_ASSIGN); }
394 ;
395
396exp : exp ASSIGN_MODIFY exp
397 { write_exp_elt_opcode (BINOP_ASSIGN_MODIFY);
398 write_exp_elt_opcode ($2);
399 write_exp_elt_opcode (BINOP_ASSIGN_MODIFY); }
400 ;
401
402exp : INT
403 { write_exp_elt_opcode (OP_LONG);
404 write_exp_elt_type ($1.type);
405 write_exp_elt_longcst ((LONGEST)($1.val));
406 write_exp_elt_opcode (OP_LONG); }
407 ;
408
409exp : NAME_OR_INT
410 { YYSTYPE val;
411 parse_number ($1.stoken.ptr, $1.stoken.length, 0, &val);
412 write_exp_elt_opcode (OP_LONG);
413 write_exp_elt_type (val.typed_val.type);
414 write_exp_elt_longcst ((LONGEST)val.typed_val.val);
415 write_exp_elt_opcode (OP_LONG); }
416 ;
417
418exp : FLOAT
419 { write_exp_elt_opcode (OP_DOUBLE);
420 write_exp_elt_type (builtin_type_f_real_s8);
421 write_exp_elt_dblcst ($1);
422 write_exp_elt_opcode (OP_DOUBLE); }
423 ;
424
425exp : variable
426 ;
427
428exp : VARIABLE
429 ;
430
431exp : SIZEOF '(' type ')' %prec UNARY
432 { write_exp_elt_opcode (OP_LONG);
433 write_exp_elt_type (builtin_type_f_integer);
434 CHECK_TYPEDEF ($3);
435 write_exp_elt_longcst ((LONGEST) TYPE_LENGTH ($3));
436 write_exp_elt_opcode (OP_LONG); }
437 ;
438
439exp : BOOLEAN_LITERAL
440 { write_exp_elt_opcode (OP_BOOL);
441 write_exp_elt_longcst ((LONGEST) $1);
442 write_exp_elt_opcode (OP_BOOL);
443 }
444 ;
445
446exp : STRING_LITERAL
447 {
448 write_exp_elt_opcode (OP_STRING);
449 write_exp_string ($1);
450 write_exp_elt_opcode (OP_STRING);
451 }
452 ;
453
454variable: name_not_typename
455 { struct symbol *sym = $1.sym;
456
457 if (sym)
458 {
459 if (symbol_read_needs_frame (sym))
460 {
461 if (innermost_block == 0 ||
462 contained_in (block_found,
463 innermost_block))
464 innermost_block = block_found;
465 }
466 write_exp_elt_opcode (OP_VAR_VALUE);
467 /* We want to use the selected frame, not
468 another more inner frame which happens to
469 be in the same block. */
470 write_exp_elt_block (NULL);
471 write_exp_elt_sym (sym);
472 write_exp_elt_opcode (OP_VAR_VALUE);
473 break;
474 }
475 else
476 {
477 struct minimal_symbol *msymbol;
478 register char *arg = copy_name ($1.stoken);
479
480 msymbol =
481 lookup_minimal_symbol (arg, NULL, NULL);
482 if (msymbol != NULL)
483 {
484 write_exp_msymbol (msymbol,
485 lookup_function_type (builtin_type_int),
486 builtin_type_int);
487 }
488 else if (!have_full_symbols () && !have_partial_symbols ())
489 error ("No symbol table is loaded. Use the \"file\" command.");
490 else
491 error ("No symbol \"%s\" in current context.",
492 copy_name ($1.stoken));
493 }
494 }
495 ;
496
497
498type : ptype
499 ;
500
501ptype : typebase
502 | typebase abs_decl
503 {
504 /* This is where the interesting stuff happens. */
505 int done = 0;
506 int array_size;
507 struct type *follow_type = $1;
508 struct type *range_type;
509
510 while (!done)
511 switch (pop_type ())
512 {
513 case tp_end:
514 done = 1;
515 break;
516 case tp_pointer:
517 follow_type = lookup_pointer_type (follow_type);
518 break;
519 case tp_reference:
520 follow_type = lookup_reference_type (follow_type);
521 break;
522 case tp_array:
523 array_size = pop_type_int ();
524 if (array_size != -1)
525 {
526 range_type =
527 create_range_type ((struct type *) NULL,
528 builtin_type_f_integer, 0,
529 array_size - 1);
530 follow_type =
531 create_array_type ((struct type *) NULL,
532 follow_type, range_type);
533 }
534 else
535 follow_type = lookup_pointer_type (follow_type);
536 break;
537 case tp_function:
538 follow_type = lookup_function_type (follow_type);
539 break;
540 }
541 $$ = follow_type;
542 }
543 ;
544
545abs_decl: '*'
546 { push_type (tp_pointer); $$ = 0; }
547 | '*' abs_decl
548 { push_type (tp_pointer); $$ = $2; }
549 | '&'
550 { push_type (tp_reference); $$ = 0; }
551 | '&' abs_decl
552 { push_type (tp_reference); $$ = $2; }
553 | direct_abs_decl
554 ;
555
556direct_abs_decl: '(' abs_decl ')'
557 { $$ = $2; }
558 | direct_abs_decl func_mod
559 { push_type (tp_function); }
560 | func_mod
561 { push_type (tp_function); }
562 ;
563
564func_mod: '(' ')'
565 { $$ = 0; }
566 | '(' nonempty_typelist ')'
8dbb1c65 567 { free ($2); $$ = 0; }
c906108c
SS
568 ;
569
570typebase /* Implements (approximately): (type-qualifier)* type-specifier */
571 : TYPENAME
572 { $$ = $1.type; }
573 | INT_KEYWORD
574 { $$ = builtin_type_f_integer; }
575 | INT_S2_KEYWORD
576 { $$ = builtin_type_f_integer_s2; }
577 | CHARACTER
578 { $$ = builtin_type_f_character; }
579 | LOGICAL_KEYWORD
580 { $$ = builtin_type_f_logical;}
581 | LOGICAL_S2_KEYWORD
582 { $$ = builtin_type_f_logical_s2;}
583 | LOGICAL_S1_KEYWORD
584 { $$ = builtin_type_f_logical_s1;}
585 | REAL_KEYWORD
586 { $$ = builtin_type_f_real;}
587 | REAL_S8_KEYWORD
588 { $$ = builtin_type_f_real_s8;}
589 | REAL_S16_KEYWORD
590 { $$ = builtin_type_f_real_s16;}
591 | COMPLEX_S8_KEYWORD
592 { $$ = builtin_type_f_complex_s8;}
593 | COMPLEX_S16_KEYWORD
594 { $$ = builtin_type_f_complex_s16;}
595 | COMPLEX_S32_KEYWORD
596 { $$ = builtin_type_f_complex_s32;}
597 ;
598
599typename: TYPENAME
600 ;
601
602nonempty_typelist
603 : type
604 { $$ = (struct type **) malloc (sizeof (struct type *) * 2);
605 $<ivec>$[0] = 1; /* Number of types in vector */
606 $$[1] = $1;
607 }
608 | nonempty_typelist ',' type
609 { int len = sizeof (struct type *) * (++($<ivec>1[0]) + 1);
610 $$ = (struct type **) realloc ((char *) $1, len);
611 $$[$<ivec>$[0]] = $3;
612 }
613 ;
614
615name : NAME
616 { $$ = $1.stoken; }
617 | TYPENAME
618 { $$ = $1.stoken; }
619 | NAME_OR_INT
620 { $$ = $1.stoken; }
621 ;
622
623name_not_typename : NAME
624/* These would be useful if name_not_typename was useful, but it is just
625 a fake for "variable", so these cause reduce/reduce conflicts because
626 the parser can't tell whether NAME_OR_INT is a name_not_typename (=variable,
627 =exp) or just an exp. If name_not_typename was ever used in an lvalue
628 context where only a name could occur, this might be useful.
629 | NAME_OR_INT
630 */
631 ;
632
633%%
634
635/* Take care of parsing a number (anything that starts with a digit).
636 Set yylval and return the token type; update lexptr.
637 LEN is the number of characters in it. */
638
639/*** Needs some error checking for the float case ***/
640
641static int
642parse_number (p, len, parsed_float, putithere)
643 register char *p;
644 register int len;
645 int parsed_float;
646 YYSTYPE *putithere;
647{
648 register LONGEST n = 0;
649 register LONGEST prevn = 0;
c906108c
SS
650 register int c;
651 register int base = input_radix;
652 int unsigned_p = 0;
653 int long_p = 0;
654 ULONGEST high_bit;
655 struct type *signed_type;
656 struct type *unsigned_type;
657
658 if (parsed_float)
659 {
660 /* It's a float since it contains a point or an exponent. */
661 /* [dD] is not understood as an exponent by atof, change it to 'e'. */
662 char *tmp, *tmp2;
663
4fcf66da 664 tmp = xstrdup (p);
c906108c
SS
665 for (tmp2 = tmp; *tmp2; ++tmp2)
666 if (*tmp2 == 'd' || *tmp2 == 'D')
667 *tmp2 = 'e';
668 putithere->dval = atof (tmp);
669 free (tmp);
670 return FLOAT;
671 }
672
673 /* Handle base-switching prefixes 0x, 0t, 0d, 0 */
674 if (p[0] == '0')
675 switch (p[1])
676 {
677 case 'x':
678 case 'X':
679 if (len >= 3)
680 {
681 p += 2;
682 base = 16;
683 len -= 2;
684 }
685 break;
686
687 case 't':
688 case 'T':
689 case 'd':
690 case 'D':
691 if (len >= 3)
692 {
693 p += 2;
694 base = 10;
695 len -= 2;
696 }
697 break;
698
699 default:
700 base = 8;
701 break;
702 }
703
704 while (len-- > 0)
705 {
706 c = *p++;
0f6e1ba6
AC
707 if (isupper (c))
708 c = tolower (c);
709 if (len == 0 && c == 'l')
710 long_p = 1;
711 else if (len == 0 && c == 'u')
712 unsigned_p = 1;
c906108c
SS
713 else
714 {
0f6e1ba6
AC
715 int i;
716 if (c >= '0' && c <= '9')
717 i = c - '0';
718 else if (c >= 'a' && c <= 'f')
719 i = c - 'a' + 10;
c906108c
SS
720 else
721 return ERROR; /* Char not a digit */
0f6e1ba6
AC
722 if (i >= base)
723 return ERROR; /* Invalid digit in this base */
724 n *= base;
725 n += i;
c906108c 726 }
c906108c
SS
727 /* Portably test for overflow (only works for nonzero values, so make
728 a second check for zero). */
729 if ((prevn >= n) && n != 0)
730 unsigned_p=1; /* Try something unsigned */
731 /* If range checking enabled, portably test for unsigned overflow. */
732 if (RANGE_CHECK && n != 0)
733 {
734 if ((unsigned_p && (unsigned)prevn >= (unsigned)n))
735 range_error("Overflow on numeric constant.");
736 }
737 prevn = n;
738 }
739
740 /* If the number is too big to be an int, or it's got an l suffix
741 then it's a long. Work out if this has to be a long by
742 shifting right and and seeing if anything remains, and the
743 target int size is different to the target long size.
744
745 In the expression below, we could have tested
746 (n >> TARGET_INT_BIT)
747 to see if it was zero,
748 but too many compilers warn about that, when ints and longs
749 are the same size. So we shift it twice, with fewer bits
750 each time, for the same result. */
751
752 if ((TARGET_INT_BIT != TARGET_LONG_BIT
753 && ((n >> 2) >> (TARGET_INT_BIT-2))) /* Avoid shift warning */
754 || long_p)
755 {
756 high_bit = ((ULONGEST)1) << (TARGET_LONG_BIT-1);
757 unsigned_type = builtin_type_unsigned_long;
758 signed_type = builtin_type_long;
759 }
760 else
761 {
762 high_bit = ((ULONGEST)1) << (TARGET_INT_BIT-1);
763 unsigned_type = builtin_type_unsigned_int;
764 signed_type = builtin_type_int;
765 }
766
767 putithere->typed_val.val = n;
768
769 /* If the high bit of the worked out type is set then this number
770 has to be unsigned. */
771
772 if (unsigned_p || (n & high_bit))
773 putithere->typed_val.type = unsigned_type;
774 else
775 putithere->typed_val.type = signed_type;
776
777 return INT;
778}
779
780struct token
781{
782 char *operator;
783 int token;
784 enum exp_opcode opcode;
785};
786
787static const struct token dot_ops[] =
788{
789 { ".and.", BOOL_AND, BINOP_END },
790 { ".AND.", BOOL_AND, BINOP_END },
791 { ".or.", BOOL_OR, BINOP_END },
792 { ".OR.", BOOL_OR, BINOP_END },
793 { ".not.", BOOL_NOT, BINOP_END },
794 { ".NOT.", BOOL_NOT, BINOP_END },
795 { ".eq.", EQUAL, BINOP_END },
796 { ".EQ.", EQUAL, BINOP_END },
797 { ".eqv.", EQUAL, BINOP_END },
798 { ".NEQV.", NOTEQUAL, BINOP_END },
799 { ".neqv.", NOTEQUAL, BINOP_END },
800 { ".EQV.", EQUAL, BINOP_END },
801 { ".ne.", NOTEQUAL, BINOP_END },
802 { ".NE.", NOTEQUAL, BINOP_END },
803 { ".le.", LEQ, BINOP_END },
804 { ".LE.", LEQ, BINOP_END },
805 { ".ge.", GEQ, BINOP_END },
806 { ".GE.", GEQ, BINOP_END },
807 { ".gt.", GREATERTHAN, BINOP_END },
808 { ".GT.", GREATERTHAN, BINOP_END },
809 { ".lt.", LESSTHAN, BINOP_END },
810 { ".LT.", LESSTHAN, BINOP_END },
811 { NULL, 0, 0 }
812};
813
814struct f77_boolean_val
815{
816 char *name;
817 int value;
818};
819
820static const struct f77_boolean_val boolean_values[] =
821{
822 { ".true.", 1 },
823 { ".TRUE.", 1 },
824 { ".false.", 0 },
825 { ".FALSE.", 0 },
826 { NULL, 0 }
827};
828
829static const struct token f77_keywords[] =
830{
831 { "complex_16", COMPLEX_S16_KEYWORD, BINOP_END },
832 { "complex_32", COMPLEX_S32_KEYWORD, BINOP_END },
833 { "character", CHARACTER, BINOP_END },
834 { "integer_2", INT_S2_KEYWORD, BINOP_END },
835 { "logical_1", LOGICAL_S1_KEYWORD, BINOP_END },
836 { "logical_2", LOGICAL_S2_KEYWORD, BINOP_END },
837 { "complex_8", COMPLEX_S8_KEYWORD, BINOP_END },
838 { "integer", INT_KEYWORD, BINOP_END },
839 { "logical", LOGICAL_KEYWORD, BINOP_END },
840 { "real_16", REAL_S16_KEYWORD, BINOP_END },
841 { "complex", COMPLEX_S8_KEYWORD, BINOP_END },
842 { "sizeof", SIZEOF, BINOP_END },
843 { "real_8", REAL_S8_KEYWORD, BINOP_END },
844 { "real", REAL_KEYWORD, BINOP_END },
845 { NULL, 0, 0 }
846};
847
848/* Implementation of a dynamically expandable buffer for processing input
849 characters acquired through lexptr and building a value to return in
850 yylval. Ripped off from ch-exp.y */
851
852static char *tempbuf; /* Current buffer contents */
853static int tempbufsize; /* Size of allocated buffer */
854static int tempbufindex; /* Current index into buffer */
855
856#define GROWBY_MIN_SIZE 64 /* Minimum amount to grow buffer by */
857
858#define CHECKBUF(size) \
859 do { \
860 if (tempbufindex + (size) >= tempbufsize) \
861 { \
862 growbuf_by_size (size); \
863 } \
864 } while (0);
865
866
867/* Grow the static temp buffer if necessary, including allocating the first one
868 on demand. */
869
870static void
871growbuf_by_size (count)
872 int count;
873{
874 int growby;
875
876 growby = max (count, GROWBY_MIN_SIZE);
877 tempbufsize += growby;
878 if (tempbuf == NULL)
879 tempbuf = (char *) malloc (tempbufsize);
880 else
881 tempbuf = (char *) realloc (tempbuf, tempbufsize);
882}
883
884/* Blatantly ripped off from ch-exp.y. This routine recognizes F77
885 string-literals.
886
887 Recognize a string literal. A string literal is a nonzero sequence
888 of characters enclosed in matching single quotes, except that
889 a single character inside single quotes is a character literal, which
890 we reject as a string literal. To embed the terminator character inside
891 a string, it is simply doubled (I.E. 'this''is''one''string') */
892
893static int
894match_string_literal ()
895{
896 char *tokptr = lexptr;
897
898 for (tempbufindex = 0, tokptr++; *tokptr != '\0'; tokptr++)
899 {
900 CHECKBUF (1);
901 if (*tokptr == *lexptr)
902 {
903 if (*(tokptr + 1) == *lexptr)
904 tokptr++;
905 else
906 break;
907 }
908 tempbuf[tempbufindex++] = *tokptr;
909 }
910 if (*tokptr == '\0' /* no terminator */
911 || tempbufindex == 0) /* no string */
912 return 0;
913 else
914 {
915 tempbuf[tempbufindex] = '\0';
916 yylval.sval.ptr = tempbuf;
917 yylval.sval.length = tempbufindex;
918 lexptr = ++tokptr;
919 return STRING_LITERAL;
920 }
921}
922
923/* Read one token, getting characters through lexptr. */
924
925static int
926yylex ()
927{
928 int c;
929 int namelen;
930 unsigned int i,token;
931 char *tokstart;
932
933 retry:
065432a8
PM
934
935 prev_lexptr = lexptr;
936
c906108c
SS
937 tokstart = lexptr;
938
939 /* First of all, let us make sure we are not dealing with the
940 special tokens .true. and .false. which evaluate to 1 and 0. */
941
942 if (*lexptr == '.')
943 {
944 for (i = 0; boolean_values[i].name != NULL; i++)
945 {
946 if STREQN (tokstart, boolean_values[i].name,
947 strlen (boolean_values[i].name))
948 {
949 lexptr += strlen (boolean_values[i].name);
950 yylval.lval = boolean_values[i].value;
951 return BOOLEAN_LITERAL;
952 }
953 }
954 }
955
956 /* See if it is a special .foo. operator */
957
958 for (i = 0; dot_ops[i].operator != NULL; i++)
959 if (STREQN (tokstart, dot_ops[i].operator, strlen (dot_ops[i].operator)))
960 {
961 lexptr += strlen (dot_ops[i].operator);
962 yylval.opcode = dot_ops[i].opcode;
963 return dot_ops[i].token;
964 }
965
966 switch (c = *tokstart)
967 {
968 case 0:
969 return 0;
970
971 case ' ':
972 case '\t':
973 case '\n':
974 lexptr++;
975 goto retry;
976
977 case '\'':
978 token = match_string_literal ();
979 if (token != 0)
980 return (token);
981 break;
982
983 case '(':
984 paren_depth++;
985 lexptr++;
986 return c;
987
988 case ')':
989 if (paren_depth == 0)
990 return 0;
991 paren_depth--;
992 lexptr++;
993 return c;
994
995 case ',':
996 if (comma_terminates && paren_depth == 0)
997 return 0;
998 lexptr++;
999 return c;
1000
1001 case '.':
1002 /* Might be a floating point number. */
1003 if (lexptr[1] < '0' || lexptr[1] > '9')
1004 goto symbol; /* Nope, must be a symbol. */
1005 /* FALL THRU into number case. */
1006
1007 case '0':
1008 case '1':
1009 case '2':
1010 case '3':
1011 case '4':
1012 case '5':
1013 case '6':
1014 case '7':
1015 case '8':
1016 case '9':
1017 {
1018 /* It's a number. */
1019 int got_dot = 0, got_e = 0, got_d = 0, toktype;
1020 register char *p = tokstart;
1021 int hex = input_radix > 10;
1022
1023 if (c == '0' && (p[1] == 'x' || p[1] == 'X'))
1024 {
1025 p += 2;
1026 hex = 1;
1027 }
1028 else if (c == '0' && (p[1]=='t' || p[1]=='T' || p[1]=='d' || p[1]=='D'))
1029 {
1030 p += 2;
1031 hex = 0;
1032 }
1033
1034 for (;; ++p)
1035 {
1036 if (!hex && !got_e && (*p == 'e' || *p == 'E'))
1037 got_dot = got_e = 1;
1038 else if (!hex && !got_d && (*p == 'd' || *p == 'D'))
1039 got_dot = got_d = 1;
1040 else if (!hex && !got_dot && *p == '.')
1041 got_dot = 1;
1042 else if (((got_e && (p[-1] == 'e' || p[-1] == 'E'))
1043 || (got_d && (p[-1] == 'd' || p[-1] == 'D')))
1044 && (*p == '-' || *p == '+'))
1045 /* This is the sign of the exponent, not the end of the
1046 number. */
1047 continue;
1048 /* We will take any letters or digits. parse_number will
1049 complain if past the radix, or if L or U are not final. */
1050 else if ((*p < '0' || *p > '9')
1051 && ((*p < 'a' || *p > 'z')
1052 && (*p < 'A' || *p > 'Z')))
1053 break;
1054 }
1055 toktype = parse_number (tokstart, p - tokstart, got_dot|got_e|got_d,
1056 &yylval);
1057 if (toktype == ERROR)
1058 {
1059 char *err_copy = (char *) alloca (p - tokstart + 1);
1060
1061 memcpy (err_copy, tokstart, p - tokstart);
1062 err_copy[p - tokstart] = 0;
1063 error ("Invalid number \"%s\".", err_copy);
1064 }
1065 lexptr = p;
1066 return toktype;
1067 }
1068
1069 case '+':
1070 case '-':
1071 case '*':
1072 case '/':
1073 case '%':
1074 case '|':
1075 case '&':
1076 case '^':
1077 case '~':
1078 case '!':
1079 case '@':
1080 case '<':
1081 case '>':
1082 case '[':
1083 case ']':
1084 case '?':
1085 case ':':
1086 case '=':
1087 case '{':
1088 case '}':
1089 symbol:
1090 lexptr++;
1091 return c;
1092 }
1093
1094 if (!(c == '_' || c == '$'
1095 || (c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z')))
1096 /* We must have come across a bad character (e.g. ';'). */
1097 error ("Invalid character '%c' in expression.", c);
1098
1099 namelen = 0;
1100 for (c = tokstart[namelen];
1101 (c == '_' || c == '$' || (c >= '0' && c <= '9')
1102 || (c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z'));
1103 c = tokstart[++namelen]);
1104
1105 /* The token "if" terminates the expression and is NOT
1106 removed from the input stream. */
1107
1108 if (namelen == 2 && tokstart[0] == 'i' && tokstart[1] == 'f')
1109 return 0;
1110
1111 lexptr += namelen;
1112
1113 /* Catch specific keywords. */
1114
1115 for (i = 0; f77_keywords[i].operator != NULL; i++)
1116 if (STREQN(tokstart, f77_keywords[i].operator,
1117 strlen(f77_keywords[i].operator)))
1118 {
1119 /* lexptr += strlen(f77_keywords[i].operator); */
1120 yylval.opcode = f77_keywords[i].opcode;
1121 return f77_keywords[i].token;
1122 }
1123
1124 yylval.sval.ptr = tokstart;
1125 yylval.sval.length = namelen;
1126
1127 if (*tokstart == '$')
1128 {
1129 write_dollar_variable (yylval.sval);
1130 return VARIABLE;
1131 }
1132
1133 /* Use token-type TYPENAME for symbols that happen to be defined
1134 currently as names of types; NAME for other symbols.
1135 The caller is not constrained to care about the distinction. */
1136 {
1137 char *tmp = copy_name (yylval.sval);
1138 struct symbol *sym;
1139 int is_a_field_of_this = 0;
1140 int hextype;
1141
1142 sym = lookup_symbol (tmp, expression_context_block,
1143 VAR_NAMESPACE,
1144 current_language->la_language == language_cplus
1145 ? &is_a_field_of_this : NULL,
1146 NULL);
1147 if (sym && SYMBOL_CLASS (sym) == LOC_TYPEDEF)
1148 {
1149 yylval.tsym.type = SYMBOL_TYPE (sym);
1150 return TYPENAME;
1151 }
1152 if ((yylval.tsym.type = lookup_primitive_typename (tmp)) != 0)
1153 return TYPENAME;
1154
1155 /* Input names that aren't symbols but ARE valid hex numbers,
1156 when the input radix permits them, can be names or numbers
1157 depending on the parse. Note we support radixes > 16 here. */
1158 if (!sym
1159 && ((tokstart[0] >= 'a' && tokstart[0] < 'a' + input_radix - 10)
1160 || (tokstart[0] >= 'A' && tokstart[0] < 'A' + input_radix - 10)))
1161 {
1162 YYSTYPE newlval; /* Its value is ignored. */
1163 hextype = parse_number (tokstart, namelen, 0, &newlval);
1164 if (hextype == INT)
1165 {
1166 yylval.ssym.sym = sym;
1167 yylval.ssym.is_a_field_of_this = is_a_field_of_this;
1168 return NAME_OR_INT;
1169 }
1170 }
1171
1172 /* Any other kind of symbol */
1173 yylval.ssym.sym = sym;
1174 yylval.ssym.is_a_field_of_this = is_a_field_of_this;
1175 return NAME;
1176 }
1177}
1178
1179void
1180yyerror (msg)
1181 char *msg;
1182{
065432a8
PM
1183 if (prev_lexptr)
1184 lexptr = prev_lexptr;
1185
c906108c
SS
1186 error ("A %s in expression, near `%s'.", (msg ? msg : "error"), lexptr);
1187}
This page took 0.306852 seconds and 4 git commands to generate.