2 /* YACC parser for Fortran expressions, for GDB.
3 Copyright (C) 1986-2015 Free Software Foundation, Inc.
5 Contributed by Motorola. Adapted from the C parser by Farooq Butt
6 (fmbutt@engage.sps.mot.com).
8 This file is part of GDB.
10 This program is free software; you can redistribute it and/or modify
11 it under the terms of the GNU General Public License as published by
12 the Free Software Foundation; either version 3 of the License, or
13 (at your option) any later version.
15 This program is distributed in the hope that it will be useful,
16 but WITHOUT ANY WARRANTY; without even the implied warranty of
17 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
18 GNU General Public License for more details.
20 You should have received a copy of the GNU General Public License
21 along with this program. If not, see <http://www.gnu.org/licenses/>. */
23 /* This was blantantly ripped off the C expression parser, please
24 be aware of that as you look at its basic structure -FMB */
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.
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. */
46 #include "expression.h"
48 #include "parser-defs.h"
51 #include "bfd.h" /* Required by objfiles.h. */
52 #include "symfile.h" /* Required by objfiles.h. */
53 #include "objfiles.h" /* For have_full_symbols and have_partial_symbols */
57 #define parse_type(ps) builtin_type (parse_gdbarch (ps))
58 #define parse_f_type(ps) builtin_f_type (parse_gdbarch (ps))
60 /* Remap normal yacc parser interface names (yyparse, yylex, yyerror, etc),
61 as well as gratuitiously global symbol names, so we can have multiple
62 yacc generated parsers in gdb. Note that these are only the variables
63 produced by yacc. If other parser generators (bison, byacc, etc) produce
64 additional global names that conflict at link time, then those parser
65 generators need to be fixed instead of adding those names to this list. */
67 #define yymaxdepth f_maxdepth
68 #define yyparse f_parse_internal
70 #define yyerror f_error
73 #define yydebug f_debug
82 #define yyerrflag f_errflag
83 #define yynerrs f_nerrs
88 #define yystate f_state
94 #define yyreds f_reds /* With YYDEBUG defined */
95 #define yytoks f_toks /* With YYDEBUG defined */
96 #define yyname f_name /* With YYDEBUG defined */
97 #define yyrule f_rule /* With YYDEBUG defined */
100 #define yydefred f_yydefred
101 #define yydgoto f_yydgoto
102 #define yysindex f_yysindex
103 #define yyrindex f_yyrindex
104 #define yygindex f_yygindex
105 #define yytable f_yytable
106 #define yycheck f_yycheck
108 #define yysslim f_yysslim
109 #define yyssp f_yyssp
110 #define yystacksize f_yystacksize
112 #define yyvsp f_yyvsp
115 #define YYDEBUG 1 /* Default to yydebug support */
118 #define YYFPRINTF parser_fprintf
120 /* The state of the parser, used internally when we are parsing the
123 static struct parser_state *pstate = NULL;
127 static int yylex (void);
129 void yyerror (char *);
131 static void growbuf_by_size (int);
133 static int match_string_literal (void);
137 /* Although the yacc "value" of an expression is not used,
138 since the result is stored in the structure being created,
139 other node types do have values. */
153 struct symtoken ssym;
156 enum exp_opcode opcode;
157 struct internalvar *ivar;
164 /* YYSTYPE gets defined by %union */
165 static int parse_number (struct parser_state *, const char *, int,
169 %type <voidval> exp type_exp start variable
170 %type <tval> type typebase
171 %type <tvec> nonempty_typelist
172 /* %type <bval> block */
174 /* Fancy type parsing. */
175 %type <voidval> func_mod direct_abs_decl abs_decl
178 %token <typed_val> INT
181 /* Both NAME and TYPENAME tokens represent symbols in the input,
182 and both convey their data as strings.
183 But a TYPENAME is a string that happens to be defined as a typedef
184 or builtin type name (such as int or char)
185 and a NAME is any other symbol.
186 Contexts where this distinction is not important can use the
187 nonterminal "name", which matches either NAME or TYPENAME. */
189 %token <sval> STRING_LITERAL
190 %token <lval> BOOLEAN_LITERAL
192 %token <tsym> TYPENAME
194 %type <ssym> name_not_typename
196 /* A NAME_OR_INT is a symbol which is not known in the symbol table,
197 but which would parse as a valid number in the current input radix.
198 E.g. "c" when input_radix==16. Depending on the parse, it will be
199 turned into a name or into a number. */
201 %token <ssym> NAME_OR_INT
206 /* Special type cases, put in to allow the parser to distinguish different
208 %token INT_KEYWORD INT_S2_KEYWORD LOGICAL_S1_KEYWORD LOGICAL_S2_KEYWORD
209 %token LOGICAL_S8_KEYWORD
210 %token LOGICAL_KEYWORD REAL_KEYWORD REAL_S8_KEYWORD REAL_S16_KEYWORD
211 %token COMPLEX_S8_KEYWORD COMPLEX_S16_KEYWORD COMPLEX_S32_KEYWORD
212 %token BOOL_AND BOOL_OR BOOL_NOT
213 %token <lval> CHARACTER
215 %token <voidval> VARIABLE
217 %token <opcode> ASSIGN_MODIFY
221 %right '=' ASSIGN_MODIFY
230 %left LESSTHAN GREATERTHAN LEQ GEQ
248 { write_exp_elt_opcode (pstate, OP_TYPE);
249 write_exp_elt_type (pstate, $1);
250 write_exp_elt_opcode (pstate, OP_TYPE); }
257 /* Expressions, not including the comma operator. */
258 exp : '*' exp %prec UNARY
259 { write_exp_elt_opcode (pstate, UNOP_IND); }
262 exp : '&' exp %prec UNARY
263 { write_exp_elt_opcode (pstate, UNOP_ADDR); }
266 exp : '-' exp %prec UNARY
267 { write_exp_elt_opcode (pstate, UNOP_NEG); }
270 exp : BOOL_NOT exp %prec UNARY
271 { write_exp_elt_opcode (pstate, UNOP_LOGICAL_NOT); }
274 exp : '~' exp %prec UNARY
275 { write_exp_elt_opcode (pstate, UNOP_COMPLEMENT); }
278 exp : SIZEOF exp %prec UNARY
279 { write_exp_elt_opcode (pstate, UNOP_SIZEOF); }
282 /* No more explicit array operators, we treat everything in F77 as
283 a function call. The disambiguation as to whether we are
284 doing a subscript operation or a function call is done
288 { start_arglist (); }
290 { write_exp_elt_opcode (pstate,
291 OP_F77_UNDETERMINED_ARGLIST);
292 write_exp_elt_longcst (pstate,
293 (LONGEST) end_arglist ());
294 write_exp_elt_opcode (pstate,
295 OP_F77_UNDETERMINED_ARGLIST); }
309 arglist : arglist ',' exp %prec ABOVE_COMMA
313 /* There are four sorts of subrange types in F90. */
315 subrange: exp ':' exp %prec ABOVE_COMMA
316 { write_exp_elt_opcode (pstate, OP_F90_RANGE);
317 write_exp_elt_longcst (pstate, NONE_BOUND_DEFAULT);
318 write_exp_elt_opcode (pstate, OP_F90_RANGE); }
321 subrange: exp ':' %prec ABOVE_COMMA
322 { write_exp_elt_opcode (pstate, OP_F90_RANGE);
323 write_exp_elt_longcst (pstate, HIGH_BOUND_DEFAULT);
324 write_exp_elt_opcode (pstate, OP_F90_RANGE); }
327 subrange: ':' exp %prec ABOVE_COMMA
328 { write_exp_elt_opcode (pstate, OP_F90_RANGE);
329 write_exp_elt_longcst (pstate, LOW_BOUND_DEFAULT);
330 write_exp_elt_opcode (pstate, OP_F90_RANGE); }
333 subrange: ':' %prec ABOVE_COMMA
334 { write_exp_elt_opcode (pstate, OP_F90_RANGE);
335 write_exp_elt_longcst (pstate, BOTH_BOUND_DEFAULT);
336 write_exp_elt_opcode (pstate, OP_F90_RANGE); }
339 complexnum: exp ',' exp
343 exp : '(' complexnum ')'
344 { write_exp_elt_opcode (pstate, OP_COMPLEX);
345 write_exp_elt_type (pstate,
346 parse_f_type (pstate)
347 ->builtin_complex_s16);
348 write_exp_elt_opcode (pstate, OP_COMPLEX); }
351 exp : '(' type ')' exp %prec UNARY
352 { write_exp_elt_opcode (pstate, UNOP_CAST);
353 write_exp_elt_type (pstate, $2);
354 write_exp_elt_opcode (pstate, UNOP_CAST); }
358 { write_exp_elt_opcode (pstate, STRUCTOP_STRUCT);
359 write_exp_string (pstate, $3);
360 write_exp_elt_opcode (pstate, STRUCTOP_STRUCT); }
363 /* Binary operators in order of decreasing precedence. */
366 { write_exp_elt_opcode (pstate, BINOP_REPEAT); }
369 exp : exp STARSTAR exp
370 { write_exp_elt_opcode (pstate, BINOP_EXP); }
374 { write_exp_elt_opcode (pstate, BINOP_MUL); }
378 { write_exp_elt_opcode (pstate, BINOP_DIV); }
382 { write_exp_elt_opcode (pstate, BINOP_ADD); }
386 { write_exp_elt_opcode (pstate, BINOP_SUB); }
390 { write_exp_elt_opcode (pstate, BINOP_LSH); }
394 { write_exp_elt_opcode (pstate, BINOP_RSH); }
398 { write_exp_elt_opcode (pstate, BINOP_EQUAL); }
401 exp : exp NOTEQUAL exp
402 { write_exp_elt_opcode (pstate, BINOP_NOTEQUAL); }
406 { write_exp_elt_opcode (pstate, BINOP_LEQ); }
410 { write_exp_elt_opcode (pstate, BINOP_GEQ); }
413 exp : exp LESSTHAN exp
414 { write_exp_elt_opcode (pstate, BINOP_LESS); }
417 exp : exp GREATERTHAN exp
418 { write_exp_elt_opcode (pstate, BINOP_GTR); }
422 { write_exp_elt_opcode (pstate, BINOP_BITWISE_AND); }
426 { write_exp_elt_opcode (pstate, BINOP_BITWISE_XOR); }
430 { write_exp_elt_opcode (pstate, BINOP_BITWISE_IOR); }
433 exp : exp BOOL_AND exp
434 { write_exp_elt_opcode (pstate, BINOP_LOGICAL_AND); }
438 exp : exp BOOL_OR exp
439 { write_exp_elt_opcode (pstate, BINOP_LOGICAL_OR); }
443 { write_exp_elt_opcode (pstate, BINOP_ASSIGN); }
446 exp : exp ASSIGN_MODIFY exp
447 { write_exp_elt_opcode (pstate, BINOP_ASSIGN_MODIFY);
448 write_exp_elt_opcode (pstate, $2);
449 write_exp_elt_opcode (pstate, BINOP_ASSIGN_MODIFY); }
453 { write_exp_elt_opcode (pstate, OP_LONG);
454 write_exp_elt_type (pstate, $1.type);
455 write_exp_elt_longcst (pstate, (LONGEST) ($1.val));
456 write_exp_elt_opcode (pstate, OP_LONG); }
461 parse_number (pstate, $1.stoken.ptr,
462 $1.stoken.length, 0, &val);
463 write_exp_elt_opcode (pstate, OP_LONG);
464 write_exp_elt_type (pstate, val.typed_val.type);
465 write_exp_elt_longcst (pstate,
466 (LONGEST)val.typed_val.val);
467 write_exp_elt_opcode (pstate, OP_LONG); }
471 { write_exp_elt_opcode (pstate, OP_DOUBLE);
472 write_exp_elt_type (pstate,
473 parse_f_type (pstate)
475 write_exp_elt_dblcst (pstate, $1);
476 write_exp_elt_opcode (pstate, OP_DOUBLE); }
485 exp : SIZEOF '(' type ')' %prec UNARY
486 { write_exp_elt_opcode (pstate, OP_LONG);
487 write_exp_elt_type (pstate,
488 parse_f_type (pstate)
490 $3 = check_typedef ($3);
491 write_exp_elt_longcst (pstate,
492 (LONGEST) TYPE_LENGTH ($3));
493 write_exp_elt_opcode (pstate, OP_LONG); }
496 exp : BOOLEAN_LITERAL
497 { write_exp_elt_opcode (pstate, OP_BOOL);
498 write_exp_elt_longcst (pstate, (LONGEST) $1);
499 write_exp_elt_opcode (pstate, OP_BOOL);
505 write_exp_elt_opcode (pstate, OP_STRING);
506 write_exp_string (pstate, $1);
507 write_exp_elt_opcode (pstate, OP_STRING);
511 variable: name_not_typename
512 { struct block_symbol sym = $1.sym;
516 if (symbol_read_needs_frame (sym.symbol))
518 if (innermost_block == 0
519 || contained_in (sym.block,
521 innermost_block = sym.block;
523 write_exp_elt_opcode (pstate, OP_VAR_VALUE);
524 write_exp_elt_block (pstate, sym.block);
525 write_exp_elt_sym (pstate, sym.symbol);
526 write_exp_elt_opcode (pstate, OP_VAR_VALUE);
531 struct bound_minimal_symbol msymbol;
532 char *arg = copy_name ($1.stoken);
535 lookup_bound_minimal_symbol (arg);
536 if (msymbol.minsym != NULL)
537 write_exp_msymbol (pstate, msymbol);
538 else if (!have_full_symbols () && !have_partial_symbols ())
539 error (_("No symbol table is loaded. Use the \"file\" command."));
541 error (_("No symbol \"%s\" in current context."),
542 copy_name ($1.stoken));
554 /* This is where the interesting stuff happens. */
557 struct type *follow_type = $1;
558 struct type *range_type;
567 follow_type = lookup_pointer_type (follow_type);
570 follow_type = lookup_reference_type (follow_type);
573 array_size = pop_type_int ();
574 if (array_size != -1)
577 create_static_range_type ((struct type *) NULL,
578 parse_f_type (pstate)
582 create_array_type ((struct type *) NULL,
583 follow_type, range_type);
586 follow_type = lookup_pointer_type (follow_type);
589 follow_type = lookup_function_type (follow_type);
597 { push_type (tp_pointer); $$ = 0; }
599 { push_type (tp_pointer); $$ = $2; }
601 { push_type (tp_reference); $$ = 0; }
603 { push_type (tp_reference); $$ = $2; }
607 direct_abs_decl: '(' abs_decl ')'
609 | direct_abs_decl func_mod
610 { push_type (tp_function); }
612 { push_type (tp_function); }
617 | '(' nonempty_typelist ')'
618 { free ($2); $$ = 0; }
621 typebase /* Implements (approximately): (type-qualifier)* type-specifier */
625 { $$ = parse_f_type (pstate)->builtin_integer; }
627 { $$ = parse_f_type (pstate)->builtin_integer_s2; }
629 { $$ = parse_f_type (pstate)->builtin_character; }
631 { $$ = parse_f_type (pstate)->builtin_logical_s8; }
633 { $$ = parse_f_type (pstate)->builtin_logical; }
635 { $$ = parse_f_type (pstate)->builtin_logical_s2; }
637 { $$ = parse_f_type (pstate)->builtin_logical_s1; }
639 { $$ = parse_f_type (pstate)->builtin_real; }
641 { $$ = parse_f_type (pstate)->builtin_real_s8; }
643 { $$ = parse_f_type (pstate)->builtin_real_s16; }
645 { $$ = parse_f_type (pstate)->builtin_complex_s8; }
646 | COMPLEX_S16_KEYWORD
647 { $$ = parse_f_type (pstate)->builtin_complex_s16; }
648 | COMPLEX_S32_KEYWORD
649 { $$ = parse_f_type (pstate)->builtin_complex_s32; }
654 { $$ = (struct type **) malloc (sizeof (struct type *) * 2);
655 $<ivec>$[0] = 1; /* Number of types in vector */
658 | nonempty_typelist ',' type
659 { int len = sizeof (struct type *) * (++($<ivec>1[0]) + 1);
660 $$ = (struct type **) realloc ((char *) $1, len);
661 $$[$<ivec>$[0]] = $3;
669 name_not_typename : NAME
670 /* These would be useful if name_not_typename was useful, but it is just
671 a fake for "variable", so these cause reduce/reduce conflicts because
672 the parser can't tell whether NAME_OR_INT is a name_not_typename (=variable,
673 =exp) or just an exp. If name_not_typename was ever used in an lvalue
674 context where only a name could occur, this might be useful.
681 /* Take care of parsing a number (anything that starts with a digit).
682 Set yylval and return the token type; update lexptr.
683 LEN is the number of characters in it. */
685 /*** Needs some error checking for the float case ***/
688 parse_number (struct parser_state *par_state,
689 const char *p, int len, int parsed_float, YYSTYPE *putithere)
694 int base = input_radix;
698 struct type *signed_type;
699 struct type *unsigned_type;
703 /* It's a float since it contains a point or an exponent. */
704 /* [dD] is not understood as an exponent by atof, change it to 'e'. */
708 for (tmp2 = tmp; *tmp2; ++tmp2)
709 if (*tmp2 == 'd' || *tmp2 == 'D')
711 putithere->dval = atof (tmp);
716 /* Handle base-switching prefixes 0x, 0t, 0d, 0 */
752 if (len == 0 && c == 'l')
754 else if (len == 0 && c == 'u')
759 if (c >= '0' && c <= '9')
761 else if (c >= 'a' && c <= 'f')
764 return ERROR; /* Char not a digit */
766 return ERROR; /* Invalid digit in this base */
770 /* Portably test for overflow (only works for nonzero values, so make
771 a second check for zero). */
772 if ((prevn >= n) && n != 0)
773 unsigned_p=1; /* Try something unsigned */
774 /* If range checking enabled, portably test for unsigned overflow. */
775 if (RANGE_CHECK && n != 0)
777 if ((unsigned_p && (unsigned)prevn >= (unsigned)n))
778 range_error (_("Overflow on numeric constant."));
783 /* If the number is too big to be an int, or it's got an l suffix
784 then it's a long. Work out if this has to be a long by
785 shifting right and seeing if anything remains, and the
786 target int size is different to the target long size.
788 In the expression below, we could have tested
789 (n >> gdbarch_int_bit (parse_gdbarch))
790 to see if it was zero,
791 but too many compilers warn about that, when ints and longs
792 are the same size. So we shift it twice, with fewer bits
793 each time, for the same result. */
795 if ((gdbarch_int_bit (parse_gdbarch (par_state))
796 != gdbarch_long_bit (parse_gdbarch (par_state))
798 >> (gdbarch_int_bit (parse_gdbarch (par_state))-2))) /* Avoid
802 high_bit = ((ULONGEST)1)
803 << (gdbarch_long_bit (parse_gdbarch (par_state))-1);
804 unsigned_type = parse_type (par_state)->builtin_unsigned_long;
805 signed_type = parse_type (par_state)->builtin_long;
810 ((ULONGEST)1) << (gdbarch_int_bit (parse_gdbarch (par_state)) - 1);
811 unsigned_type = parse_type (par_state)->builtin_unsigned_int;
812 signed_type = parse_type (par_state)->builtin_int;
815 putithere->typed_val.val = n;
817 /* If the high bit of the worked out type is set then this number
818 has to be unsigned. */
820 if (unsigned_p || (n & high_bit))
821 putithere->typed_val.type = unsigned_type;
823 putithere->typed_val.type = signed_type;
832 enum exp_opcode opcode;
835 static const struct token dot_ops[] =
837 { ".and.", BOOL_AND, BINOP_END },
838 { ".AND.", BOOL_AND, BINOP_END },
839 { ".or.", BOOL_OR, BINOP_END },
840 { ".OR.", BOOL_OR, BINOP_END },
841 { ".not.", BOOL_NOT, BINOP_END },
842 { ".NOT.", BOOL_NOT, BINOP_END },
843 { ".eq.", EQUAL, BINOP_END },
844 { ".EQ.", EQUAL, BINOP_END },
845 { ".eqv.", EQUAL, BINOP_END },
846 { ".NEQV.", NOTEQUAL, BINOP_END },
847 { ".neqv.", NOTEQUAL, BINOP_END },
848 { ".EQV.", EQUAL, BINOP_END },
849 { ".ne.", NOTEQUAL, BINOP_END },
850 { ".NE.", NOTEQUAL, BINOP_END },
851 { ".le.", LEQ, BINOP_END },
852 { ".LE.", LEQ, BINOP_END },
853 { ".ge.", GEQ, BINOP_END },
854 { ".GE.", GEQ, BINOP_END },
855 { ".gt.", GREATERTHAN, BINOP_END },
856 { ".GT.", GREATERTHAN, BINOP_END },
857 { ".lt.", LESSTHAN, BINOP_END },
858 { ".LT.", LESSTHAN, BINOP_END },
859 { NULL, 0, BINOP_END }
862 struct f77_boolean_val
868 static const struct f77_boolean_val boolean_values[] =
877 static const struct token f77_keywords[] =
879 { "complex_16", COMPLEX_S16_KEYWORD, BINOP_END },
880 { "complex_32", COMPLEX_S32_KEYWORD, BINOP_END },
881 { "character", CHARACTER, BINOP_END },
882 { "integer_2", INT_S2_KEYWORD, BINOP_END },
883 { "logical_1", LOGICAL_S1_KEYWORD, BINOP_END },
884 { "logical_2", LOGICAL_S2_KEYWORD, BINOP_END },
885 { "logical_8", LOGICAL_S8_KEYWORD, BINOP_END },
886 { "complex_8", COMPLEX_S8_KEYWORD, BINOP_END },
887 { "integer", INT_KEYWORD, BINOP_END },
888 { "logical", LOGICAL_KEYWORD, BINOP_END },
889 { "real_16", REAL_S16_KEYWORD, BINOP_END },
890 { "complex", COMPLEX_S8_KEYWORD, BINOP_END },
891 { "sizeof", SIZEOF, BINOP_END },
892 { "real_8", REAL_S8_KEYWORD, BINOP_END },
893 { "real", REAL_KEYWORD, BINOP_END },
894 { NULL, 0, BINOP_END }
897 /* Implementation of a dynamically expandable buffer for processing input
898 characters acquired through lexptr and building a value to return in
899 yylval. Ripped off from ch-exp.y */
901 static char *tempbuf; /* Current buffer contents */
902 static int tempbufsize; /* Size of allocated buffer */
903 static int tempbufindex; /* Current index into buffer */
905 #define GROWBY_MIN_SIZE 64 /* Minimum amount to grow buffer by */
907 #define CHECKBUF(size) \
909 if (tempbufindex + (size) >= tempbufsize) \
911 growbuf_by_size (size); \
916 /* Grow the static temp buffer if necessary, including allocating the
917 first one on demand. */
920 growbuf_by_size (int count)
924 growby = max (count, GROWBY_MIN_SIZE);
925 tempbufsize += growby;
927 tempbuf = (char *) malloc (tempbufsize);
929 tempbuf = (char *) realloc (tempbuf, tempbufsize);
932 /* Blatantly ripped off from ch-exp.y. This routine recognizes F77
935 Recognize a string literal. A string literal is a nonzero sequence
936 of characters enclosed in matching single quotes, except that
937 a single character inside single quotes is a character literal, which
938 we reject as a string literal. To embed the terminator character inside
939 a string, it is simply doubled (I.E. 'this''is''one''string') */
942 match_string_literal (void)
944 const char *tokptr = lexptr;
946 for (tempbufindex = 0, tokptr++; *tokptr != '\0'; tokptr++)
949 if (*tokptr == *lexptr)
951 if (*(tokptr + 1) == *lexptr)
956 tempbuf[tempbufindex++] = *tokptr;
958 if (*tokptr == '\0' /* no terminator */
959 || tempbufindex == 0) /* no string */
963 tempbuf[tempbufindex] = '\0';
964 yylval.sval.ptr = tempbuf;
965 yylval.sval.length = tempbufindex;
967 return STRING_LITERAL;
971 /* Read one token, getting characters through lexptr. */
978 unsigned int i,token;
979 const char *tokstart;
983 prev_lexptr = lexptr;
987 /* First of all, let us make sure we are not dealing with the
988 special tokens .true. and .false. which evaluate to 1 and 0. */
992 for (i = 0; boolean_values[i].name != NULL; i++)
994 if (strncmp (tokstart, boolean_values[i].name,
995 strlen (boolean_values[i].name)) == 0)
997 lexptr += strlen (boolean_values[i].name);
998 yylval.lval = boolean_values[i].value;
999 return BOOLEAN_LITERAL;
1004 /* See if it is a special .foo. operator. */
1006 for (i = 0; dot_ops[i].oper != NULL; i++)
1007 if (strncmp (tokstart, dot_ops[i].oper,
1008 strlen (dot_ops[i].oper)) == 0)
1010 lexptr += strlen (dot_ops[i].oper);
1011 yylval.opcode = dot_ops[i].opcode;
1012 return dot_ops[i].token;
1015 /* See if it is an exponentiation operator. */
1017 if (strncmp (tokstart, "**", 2) == 0)
1020 yylval.opcode = BINOP_EXP;
1024 switch (c = *tokstart)
1036 token = match_string_literal ();
1047 if (paren_depth == 0)
1054 if (comma_terminates && paren_depth == 0)
1060 /* Might be a floating point number. */
1061 if (lexptr[1] < '0' || lexptr[1] > '9')
1062 goto symbol; /* Nope, must be a symbol. */
1063 /* FALL THRU into number case. */
1076 /* It's a number. */
1077 int got_dot = 0, got_e = 0, got_d = 0, toktype;
1078 const char *p = tokstart;
1079 int hex = input_radix > 10;
1081 if (c == '0' && (p[1] == 'x' || p[1] == 'X'))
1086 else if (c == '0' && (p[1]=='t' || p[1]=='T'
1087 || p[1]=='d' || p[1]=='D'))
1095 if (!hex && !got_e && (*p == 'e' || *p == 'E'))
1096 got_dot = got_e = 1;
1097 else if (!hex && !got_d && (*p == 'd' || *p == 'D'))
1098 got_dot = got_d = 1;
1099 else if (!hex && !got_dot && *p == '.')
1101 else if (((got_e && (p[-1] == 'e' || p[-1] == 'E'))
1102 || (got_d && (p[-1] == 'd' || p[-1] == 'D')))
1103 && (*p == '-' || *p == '+'))
1104 /* This is the sign of the exponent, not the end of the
1107 /* We will take any letters or digits. parse_number will
1108 complain if past the radix, or if L or U are not final. */
1109 else if ((*p < '0' || *p > '9')
1110 && ((*p < 'a' || *p > 'z')
1111 && (*p < 'A' || *p > 'Z')))
1114 toktype = parse_number (pstate, tokstart, p - tokstart,
1115 got_dot|got_e|got_d,
1117 if (toktype == ERROR)
1119 char *err_copy = (char *) alloca (p - tokstart + 1);
1121 memcpy (err_copy, tokstart, p - tokstart);
1122 err_copy[p - tokstart] = 0;
1123 error (_("Invalid number \"%s\"."), err_copy);
1154 if (!(c == '_' || c == '$' || c ==':'
1155 || (c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z')))
1156 /* We must have come across a bad character (e.g. ';'). */
1157 error (_("Invalid character '%c' in expression."), c);
1160 for (c = tokstart[namelen];
1161 (c == '_' || c == '$' || c == ':' || (c >= '0' && c <= '9')
1162 || (c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z'));
1163 c = tokstart[++namelen]);
1165 /* The token "if" terminates the expression and is NOT
1166 removed from the input stream. */
1168 if (namelen == 2 && tokstart[0] == 'i' && tokstart[1] == 'f')
1173 /* Catch specific keywords. */
1175 for (i = 0; f77_keywords[i].oper != NULL; i++)
1176 if (strlen (f77_keywords[i].oper) == namelen
1177 && strncmp (tokstart, f77_keywords[i].oper, namelen) == 0)
1179 /* lexptr += strlen(f77_keywords[i].operator); */
1180 yylval.opcode = f77_keywords[i].opcode;
1181 return f77_keywords[i].token;
1184 yylval.sval.ptr = tokstart;
1185 yylval.sval.length = namelen;
1187 if (*tokstart == '$')
1189 write_dollar_variable (pstate, yylval.sval);
1193 /* Use token-type TYPENAME for symbols that happen to be defined
1194 currently as names of types; NAME for other symbols.
1195 The caller is not constrained to care about the distinction. */
1197 char *tmp = copy_name (yylval.sval);
1198 struct block_symbol result;
1199 struct field_of_this_result is_a_field_of_this;
1200 enum domain_enum_tag lookup_domains[] =
1209 for (i = 0; i < ARRAY_SIZE (lookup_domains); ++i)
1211 /* Initialize this in case we *don't* use it in this call; that
1212 way we can refer to it unconditionally below. */
1213 memset (&is_a_field_of_this, 0, sizeof (is_a_field_of_this));
1215 result = lookup_symbol (tmp, expression_context_block,
1217 parse_language (pstate)->la_language
1219 ? &is_a_field_of_this : NULL);
1220 if (result.symbol && SYMBOL_CLASS (result.symbol) == LOC_TYPEDEF)
1222 yylval.tsym.type = SYMBOL_TYPE (result.symbol);
1231 = language_lookup_primitive_type (parse_language (pstate),
1232 parse_gdbarch (pstate), tmp);
1233 if (yylval.tsym.type != NULL)
1236 /* Input names that aren't symbols but ARE valid hex numbers,
1237 when the input radix permits them, can be names or numbers
1238 depending on the parse. Note we support radixes > 16 here. */
1240 && ((tokstart[0] >= 'a' && tokstart[0] < 'a' + input_radix - 10)
1241 || (tokstart[0] >= 'A' && tokstart[0] < 'A' + input_radix - 10)))
1243 YYSTYPE newlval; /* Its value is ignored. */
1244 hextype = parse_number (pstate, tokstart, namelen, 0, &newlval);
1247 yylval.ssym.sym = result;
1248 yylval.ssym.is_a_field_of_this = is_a_field_of_this.type != NULL;
1253 /* Any other kind of symbol */
1254 yylval.ssym.sym = result;
1255 yylval.ssym.is_a_field_of_this = is_a_field_of_this.type != NULL;
1261 f_parse (struct parser_state *par_state)
1264 struct cleanup *c = make_cleanup_clear_parser_state (&pstate);
1266 /* Setting up the parser state. */
1267 gdb_assert (par_state != NULL);
1270 result = yyparse ();
1279 lexptr = prev_lexptr;
1281 error (_("A %s in expression, near `%s'."), (msg ? msg : "error"), lexptr);