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