1 /* YACC parser for Fortran expressions, for GDB.
2 Copyright 1986, 1989, 1990, 1991, 1993, 1994, 1995, 1996, 2000, 2001
3 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 2 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, write to the Free Software
22 Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA. */
24 /* This was blantantly ripped off the C expression parser, please
25 be aware of that as you look at its basic structure -FMB */
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.
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. */
47 #include "gdb_string.h"
48 #include "expression.h"
50 #include "parser-defs.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 */
59 /* Remap normal yacc parser interface names (yyparse, yylex, yyerror, etc),
60 as well as gratuitiously global symbol names, so we can have multiple
61 yacc generated parsers in gdb. Note that these are only the variables
62 produced by yacc. If other parser generators (bison, byacc, etc) produce
63 additional global names that conflict at link time, then those parser
64 generators need to be fixed instead of adding those names to this list. */
66 #define yymaxdepth f_maxdepth
67 #define yyparse f_parse
69 #define yyerror f_error
72 #define yydebug f_debug
81 #define yyerrflag f_errflag
82 #define yynerrs f_nerrs
87 #define yystate f_state
93 #define yyreds f_reds /* With YYDEBUG defined */
94 #define yytoks f_toks /* With YYDEBUG defined */
95 #define yyname f_name /* With YYDEBUG defined */
96 #define yyrule f_rule /* With YYDEBUG defined */
99 #define yydefred f_yydefred
100 #define yydgoto f_yydgoto
101 #define yysindex f_yysindex
102 #define yyrindex f_yyrindex
103 #define yygindex f_yygindex
104 #define yytable f_yytable
105 #define yycheck f_yycheck
108 #define YYDEBUG 1 /* Default to yydebug support */
111 #define YYFPRINTF parser_fprintf
115 static int yylex (void);
117 void yyerror (char *);
119 static void growbuf_by_size (int);
121 static int match_string_literal (void);
125 /* Although the yacc "value" of an expression is not used,
126 since the result is stored in the structure being created,
127 other node types do have values. */
141 struct symtoken ssym;
144 enum exp_opcode opcode;
145 struct internalvar *ivar;
152 /* YYSTYPE gets defined by %union */
153 static int parse_number (char *, int, int, YYSTYPE *);
156 %type <voidval> exp type_exp start variable
157 %type <tval> type typebase
158 %type <tvec> nonempty_typelist
159 /* %type <bval> block */
161 /* Fancy type parsing. */
162 %type <voidval> func_mod direct_abs_decl abs_decl
165 %token <typed_val> INT
168 /* Both NAME and TYPENAME tokens represent symbols in the input,
169 and both convey their data as strings.
170 But a TYPENAME is a string that happens to be defined as a typedef
171 or builtin type name (such as int or char)
172 and a NAME is any other symbol.
173 Contexts where this distinction is not important can use the
174 nonterminal "name", which matches either NAME or TYPENAME. */
176 %token <sval> STRING_LITERAL
177 %token <lval> BOOLEAN_LITERAL
179 %token <tsym> TYPENAME
181 %type <ssym> name_not_typename
182 %type <tsym> typename
184 /* A NAME_OR_INT is a symbol which is not known in the symbol table,
185 but which would parse as a valid number in the current input radix.
186 E.g. "c" when input_radix==16. Depending on the parse, it will be
187 turned into a name or into a number. */
189 %token <ssym> NAME_OR_INT
194 /* Special type cases, put in to allow the parser to distinguish different
196 %token INT_KEYWORD INT_S2_KEYWORD LOGICAL_S1_KEYWORD LOGICAL_S2_KEYWORD
197 %token LOGICAL_KEYWORD REAL_KEYWORD REAL_S8_KEYWORD REAL_S16_KEYWORD
198 %token COMPLEX_S8_KEYWORD COMPLEX_S16_KEYWORD COMPLEX_S32_KEYWORD
199 %token BOOL_AND BOOL_OR BOOL_NOT
200 %token <lval> CHARACTER
202 %token <voidval> VARIABLE
204 %token <opcode> ASSIGN_MODIFY
208 %right '=' ASSIGN_MODIFY
217 %left LESSTHAN GREATERTHAN LEQ GEQ
233 { write_exp_elt_opcode(OP_TYPE);
234 write_exp_elt_type($1);
235 write_exp_elt_opcode(OP_TYPE); }
242 /* Expressions, not including the comma operator. */
243 exp : '*' exp %prec UNARY
244 { write_exp_elt_opcode (UNOP_IND); }
247 exp : '&' exp %prec UNARY
248 { write_exp_elt_opcode (UNOP_ADDR); }
251 exp : '-' exp %prec UNARY
252 { write_exp_elt_opcode (UNOP_NEG); }
255 exp : BOOL_NOT exp %prec UNARY
256 { write_exp_elt_opcode (UNOP_LOGICAL_NOT); }
259 exp : '~' exp %prec UNARY
260 { write_exp_elt_opcode (UNOP_COMPLEMENT); }
263 exp : SIZEOF exp %prec UNARY
264 { write_exp_elt_opcode (UNOP_SIZEOF); }
267 /* No more explicit array operators, we treat everything in F77 as
268 a function call. The disambiguation as to whether we are
269 doing a subscript operation or a function call is done
273 { start_arglist (); }
275 { write_exp_elt_opcode (OP_F77_UNDETERMINED_ARGLIST);
276 write_exp_elt_longcst ((LONGEST) end_arglist ());
277 write_exp_elt_opcode (OP_F77_UNDETERMINED_ARGLIST); }
291 arglist : arglist ',' exp %prec ABOVE_COMMA
295 substring: exp ':' exp %prec ABOVE_COMMA
300 complexnum: exp ',' exp
304 exp : '(' complexnum ')'
305 { write_exp_elt_opcode(OP_COMPLEX); }
308 exp : '(' type ')' exp %prec UNARY
309 { write_exp_elt_opcode (UNOP_CAST);
310 write_exp_elt_type ($2);
311 write_exp_elt_opcode (UNOP_CAST); }
314 /* Binary operators in order of decreasing precedence. */
317 { write_exp_elt_opcode (BINOP_REPEAT); }
321 { write_exp_elt_opcode (BINOP_MUL); }
325 { write_exp_elt_opcode (BINOP_DIV); }
329 { write_exp_elt_opcode (BINOP_REM); }
333 { write_exp_elt_opcode (BINOP_ADD); }
337 { write_exp_elt_opcode (BINOP_SUB); }
341 { write_exp_elt_opcode (BINOP_LSH); }
345 { write_exp_elt_opcode (BINOP_RSH); }
349 { write_exp_elt_opcode (BINOP_EQUAL); }
352 exp : exp NOTEQUAL exp
353 { write_exp_elt_opcode (BINOP_NOTEQUAL); }
357 { write_exp_elt_opcode (BINOP_LEQ); }
361 { write_exp_elt_opcode (BINOP_GEQ); }
364 exp : exp LESSTHAN exp
365 { write_exp_elt_opcode (BINOP_LESS); }
368 exp : exp GREATERTHAN exp
369 { write_exp_elt_opcode (BINOP_GTR); }
373 { write_exp_elt_opcode (BINOP_BITWISE_AND); }
377 { write_exp_elt_opcode (BINOP_BITWISE_XOR); }
381 { write_exp_elt_opcode (BINOP_BITWISE_IOR); }
384 exp : exp BOOL_AND exp
385 { write_exp_elt_opcode (BINOP_LOGICAL_AND); }
389 exp : exp BOOL_OR exp
390 { write_exp_elt_opcode (BINOP_LOGICAL_OR); }
394 { write_exp_elt_opcode (BINOP_ASSIGN); }
397 exp : exp ASSIGN_MODIFY exp
398 { write_exp_elt_opcode (BINOP_ASSIGN_MODIFY);
399 write_exp_elt_opcode ($2);
400 write_exp_elt_opcode (BINOP_ASSIGN_MODIFY); }
404 { write_exp_elt_opcode (OP_LONG);
405 write_exp_elt_type ($1.type);
406 write_exp_elt_longcst ((LONGEST)($1.val));
407 write_exp_elt_opcode (OP_LONG); }
412 parse_number ($1.stoken.ptr, $1.stoken.length, 0, &val);
413 write_exp_elt_opcode (OP_LONG);
414 write_exp_elt_type (val.typed_val.type);
415 write_exp_elt_longcst ((LONGEST)val.typed_val.val);
416 write_exp_elt_opcode (OP_LONG); }
420 { write_exp_elt_opcode (OP_DOUBLE);
421 write_exp_elt_type (builtin_type_f_real_s8);
422 write_exp_elt_dblcst ($1);
423 write_exp_elt_opcode (OP_DOUBLE); }
432 exp : SIZEOF '(' type ')' %prec UNARY
433 { write_exp_elt_opcode (OP_LONG);
434 write_exp_elt_type (builtin_type_f_integer);
436 write_exp_elt_longcst ((LONGEST) TYPE_LENGTH ($3));
437 write_exp_elt_opcode (OP_LONG); }
440 exp : BOOLEAN_LITERAL
441 { write_exp_elt_opcode (OP_BOOL);
442 write_exp_elt_longcst ((LONGEST) $1);
443 write_exp_elt_opcode (OP_BOOL);
449 write_exp_elt_opcode (OP_STRING);
450 write_exp_string ($1);
451 write_exp_elt_opcode (OP_STRING);
455 variable: name_not_typename
456 { struct symbol *sym = $1.sym;
460 if (symbol_read_needs_frame (sym))
462 if (innermost_block == 0 ||
463 contained_in (block_found,
465 innermost_block = block_found;
467 write_exp_elt_opcode (OP_VAR_VALUE);
468 /* We want to use the selected frame, not
469 another more inner frame which happens to
470 be in the same block. */
471 write_exp_elt_block (NULL);
472 write_exp_elt_sym (sym);
473 write_exp_elt_opcode (OP_VAR_VALUE);
478 struct minimal_symbol *msymbol;
479 char *arg = copy_name ($1.stoken);
482 lookup_minimal_symbol (arg, NULL, NULL);
485 write_exp_msymbol (msymbol,
486 lookup_function_type (builtin_type_int),
489 else if (!have_full_symbols () && !have_partial_symbols ())
490 error ("No symbol table is loaded. Use the \"file\" command.");
492 error ("No symbol \"%s\" in current context.",
493 copy_name ($1.stoken));
505 /* This is where the interesting stuff happens. */
508 struct type *follow_type = $1;
509 struct type *range_type;
518 follow_type = lookup_pointer_type (follow_type);
521 follow_type = lookup_reference_type (follow_type);
524 array_size = pop_type_int ();
525 if (array_size != -1)
528 create_range_type ((struct type *) NULL,
529 builtin_type_f_integer, 0,
532 create_array_type ((struct type *) NULL,
533 follow_type, range_type);
536 follow_type = lookup_pointer_type (follow_type);
539 follow_type = lookup_function_type (follow_type);
547 { push_type (tp_pointer); $$ = 0; }
549 { push_type (tp_pointer); $$ = $2; }
551 { push_type (tp_reference); $$ = 0; }
553 { push_type (tp_reference); $$ = $2; }
557 direct_abs_decl: '(' abs_decl ')'
559 | direct_abs_decl func_mod
560 { push_type (tp_function); }
562 { push_type (tp_function); }
567 | '(' nonempty_typelist ')'
568 { free ($2); $$ = 0; }
571 typebase /* Implements (approximately): (type-qualifier)* type-specifier */
575 { $$ = builtin_type_f_integer; }
577 { $$ = builtin_type_f_integer_s2; }
579 { $$ = builtin_type_f_character; }
581 { $$ = builtin_type_f_logical;}
583 { $$ = builtin_type_f_logical_s2;}
585 { $$ = builtin_type_f_logical_s1;}
587 { $$ = builtin_type_f_real;}
589 { $$ = builtin_type_f_real_s8;}
591 { $$ = builtin_type_f_real_s16;}
593 { $$ = builtin_type_f_complex_s8;}
594 | COMPLEX_S16_KEYWORD
595 { $$ = builtin_type_f_complex_s16;}
596 | COMPLEX_S32_KEYWORD
597 { $$ = builtin_type_f_complex_s32;}
605 { $$ = (struct type **) malloc (sizeof (struct type *) * 2);
606 $<ivec>$[0] = 1; /* Number of types in vector */
609 | nonempty_typelist ',' type
610 { int len = sizeof (struct type *) * (++($<ivec>1[0]) + 1);
611 $$ = (struct type **) realloc ((char *) $1, len);
612 $$[$<ivec>$[0]] = $3;
624 name_not_typename : NAME
625 /* These would be useful if name_not_typename was useful, but it is just
626 a fake for "variable", so these cause reduce/reduce conflicts because
627 the parser can't tell whether NAME_OR_INT is a name_not_typename (=variable,
628 =exp) or just an exp. If name_not_typename was ever used in an lvalue
629 context where only a name could occur, this might be useful.
636 /* Take care of parsing a number (anything that starts with a digit).
637 Set yylval and return the token type; update lexptr.
638 LEN is the number of characters in it. */
640 /*** Needs some error checking for the float case ***/
643 parse_number (p, len, parsed_float, putithere)
652 int base = input_radix;
656 struct type *signed_type;
657 struct type *unsigned_type;
661 /* It's a float since it contains a point or an exponent. */
662 /* [dD] is not understood as an exponent by atof, change it to 'e'. */
666 for (tmp2 = tmp; *tmp2; ++tmp2)
667 if (*tmp2 == 'd' || *tmp2 == 'D')
669 putithere->dval = atof (tmp);
674 /* Handle base-switching prefixes 0x, 0t, 0d, 0 */
710 if (len == 0 && c == 'l')
712 else if (len == 0 && c == 'u')
717 if (c >= '0' && c <= '9')
719 else if (c >= 'a' && c <= 'f')
722 return ERROR; /* Char not a digit */
724 return ERROR; /* Invalid digit in this base */
728 /* Portably test for overflow (only works for nonzero values, so make
729 a second check for zero). */
730 if ((prevn >= n) && n != 0)
731 unsigned_p=1; /* Try something unsigned */
732 /* If range checking enabled, portably test for unsigned overflow. */
733 if (RANGE_CHECK && n != 0)
735 if ((unsigned_p && (unsigned)prevn >= (unsigned)n))
736 range_error("Overflow on numeric constant.");
741 /* If the number is too big to be an int, or it's got an l suffix
742 then it's a long. Work out if this has to be a long by
743 shifting right and and seeing if anything remains, and the
744 target int size is different to the target long size.
746 In the expression below, we could have tested
747 (n >> TARGET_INT_BIT)
748 to see if it was zero,
749 but too many compilers warn about that, when ints and longs
750 are the same size. So we shift it twice, with fewer bits
751 each time, for the same result. */
753 if ((TARGET_INT_BIT != TARGET_LONG_BIT
754 && ((n >> 2) >> (TARGET_INT_BIT-2))) /* Avoid shift warning */
757 high_bit = ((ULONGEST)1) << (TARGET_LONG_BIT-1);
758 unsigned_type = builtin_type_unsigned_long;
759 signed_type = builtin_type_long;
763 high_bit = ((ULONGEST)1) << (TARGET_INT_BIT-1);
764 unsigned_type = builtin_type_unsigned_int;
765 signed_type = builtin_type_int;
768 putithere->typed_val.val = n;
770 /* If the high bit of the worked out type is set then this number
771 has to be unsigned. */
773 if (unsigned_p || (n & high_bit))
774 putithere->typed_val.type = unsigned_type;
776 putithere->typed_val.type = signed_type;
785 enum exp_opcode opcode;
788 static const struct token dot_ops[] =
790 { ".and.", BOOL_AND, BINOP_END },
791 { ".AND.", BOOL_AND, BINOP_END },
792 { ".or.", BOOL_OR, BINOP_END },
793 { ".OR.", BOOL_OR, BINOP_END },
794 { ".not.", BOOL_NOT, BINOP_END },
795 { ".NOT.", BOOL_NOT, BINOP_END },
796 { ".eq.", EQUAL, BINOP_END },
797 { ".EQ.", EQUAL, BINOP_END },
798 { ".eqv.", EQUAL, BINOP_END },
799 { ".NEQV.", NOTEQUAL, BINOP_END },
800 { ".neqv.", NOTEQUAL, BINOP_END },
801 { ".EQV.", EQUAL, BINOP_END },
802 { ".ne.", NOTEQUAL, BINOP_END },
803 { ".NE.", NOTEQUAL, BINOP_END },
804 { ".le.", LEQ, BINOP_END },
805 { ".LE.", LEQ, BINOP_END },
806 { ".ge.", GEQ, BINOP_END },
807 { ".GE.", GEQ, BINOP_END },
808 { ".gt.", GREATERTHAN, BINOP_END },
809 { ".GT.", GREATERTHAN, BINOP_END },
810 { ".lt.", LESSTHAN, BINOP_END },
811 { ".LT.", LESSTHAN, BINOP_END },
815 struct f77_boolean_val
821 static const struct f77_boolean_val boolean_values[] =
830 static const struct token f77_keywords[] =
832 { "complex_16", COMPLEX_S16_KEYWORD, BINOP_END },
833 { "complex_32", COMPLEX_S32_KEYWORD, BINOP_END },
834 { "character", CHARACTER, BINOP_END },
835 { "integer_2", INT_S2_KEYWORD, BINOP_END },
836 { "logical_1", LOGICAL_S1_KEYWORD, BINOP_END },
837 { "logical_2", LOGICAL_S2_KEYWORD, BINOP_END },
838 { "complex_8", COMPLEX_S8_KEYWORD, BINOP_END },
839 { "integer", INT_KEYWORD, BINOP_END },
840 { "logical", LOGICAL_KEYWORD, BINOP_END },
841 { "real_16", REAL_S16_KEYWORD, BINOP_END },
842 { "complex", COMPLEX_S8_KEYWORD, BINOP_END },
843 { "sizeof", SIZEOF, BINOP_END },
844 { "real_8", REAL_S8_KEYWORD, BINOP_END },
845 { "real", REAL_KEYWORD, BINOP_END },
849 /* Implementation of a dynamically expandable buffer for processing input
850 characters acquired through lexptr and building a value to return in
851 yylval. Ripped off from ch-exp.y */
853 static char *tempbuf; /* Current buffer contents */
854 static int tempbufsize; /* Size of allocated buffer */
855 static int tempbufindex; /* Current index into buffer */
857 #define GROWBY_MIN_SIZE 64 /* Minimum amount to grow buffer by */
859 #define CHECKBUF(size) \
861 if (tempbufindex + (size) >= tempbufsize) \
863 growbuf_by_size (size); \
868 /* Grow the static temp buffer if necessary, including allocating the first one
872 growbuf_by_size (count)
877 growby = max (count, GROWBY_MIN_SIZE);
878 tempbufsize += growby;
880 tempbuf = (char *) malloc (tempbufsize);
882 tempbuf = (char *) realloc (tempbuf, tempbufsize);
885 /* Blatantly ripped off from ch-exp.y. This routine recognizes F77
888 Recognize a string literal. A string literal is a nonzero sequence
889 of characters enclosed in matching single quotes, except that
890 a single character inside single quotes is a character literal, which
891 we reject as a string literal. To embed the terminator character inside
892 a string, it is simply doubled (I.E. 'this''is''one''string') */
895 match_string_literal ()
897 char *tokptr = lexptr;
899 for (tempbufindex = 0, tokptr++; *tokptr != '\0'; tokptr++)
902 if (*tokptr == *lexptr)
904 if (*(tokptr + 1) == *lexptr)
909 tempbuf[tempbufindex++] = *tokptr;
911 if (*tokptr == '\0' /* no terminator */
912 || tempbufindex == 0) /* no string */
916 tempbuf[tempbufindex] = '\0';
917 yylval.sval.ptr = tempbuf;
918 yylval.sval.length = tempbufindex;
920 return STRING_LITERAL;
924 /* Read one token, getting characters through lexptr. */
931 unsigned int i,token;
936 prev_lexptr = lexptr;
940 /* First of all, let us make sure we are not dealing with the
941 special tokens .true. and .false. which evaluate to 1 and 0. */
945 for (i = 0; boolean_values[i].name != NULL; i++)
947 if (strncmp (tokstart, boolean_values[i].name,
948 strlen (boolean_values[i].name)) == 0)
950 lexptr += strlen (boolean_values[i].name);
951 yylval.lval = boolean_values[i].value;
952 return BOOLEAN_LITERAL;
957 /* See if it is a special .foo. operator */
959 for (i = 0; dot_ops[i].operator != NULL; i++)
960 if (strncmp (tokstart, dot_ops[i].operator, strlen (dot_ops[i].operator)) == 0)
962 lexptr += strlen (dot_ops[i].operator);
963 yylval.opcode = dot_ops[i].opcode;
964 return dot_ops[i].token;
967 switch (c = *tokstart)
979 token = match_string_literal ();
990 if (paren_depth == 0)
997 if (comma_terminates && paren_depth == 0)
1003 /* Might be a floating point number. */
1004 if (lexptr[1] < '0' || lexptr[1] > '9')
1005 goto symbol; /* Nope, must be a symbol. */
1006 /* FALL THRU into number case. */
1019 /* It's a number. */
1020 int got_dot = 0, got_e = 0, got_d = 0, toktype;
1022 int hex = input_radix > 10;
1024 if (c == '0' && (p[1] == 'x' || p[1] == 'X'))
1029 else if (c == '0' && (p[1]=='t' || p[1]=='T' || p[1]=='d' || p[1]=='D'))
1037 if (!hex && !got_e && (*p == 'e' || *p == 'E'))
1038 got_dot = got_e = 1;
1039 else if (!hex && !got_d && (*p == 'd' || *p == 'D'))
1040 got_dot = got_d = 1;
1041 else if (!hex && !got_dot && *p == '.')
1043 else if (((got_e && (p[-1] == 'e' || p[-1] == 'E'))
1044 || (got_d && (p[-1] == 'd' || p[-1] == 'D')))
1045 && (*p == '-' || *p == '+'))
1046 /* This is the sign of the exponent, not the end of the
1049 /* We will take any letters or digits. parse_number will
1050 complain if past the radix, or if L or U are not final. */
1051 else if ((*p < '0' || *p > '9')
1052 && ((*p < 'a' || *p > 'z')
1053 && (*p < 'A' || *p > 'Z')))
1056 toktype = parse_number (tokstart, p - tokstart, got_dot|got_e|got_d,
1058 if (toktype == ERROR)
1060 char *err_copy = (char *) alloca (p - tokstart + 1);
1062 memcpy (err_copy, tokstart, p - tokstart);
1063 err_copy[p - tokstart] = 0;
1064 error ("Invalid number \"%s\".", err_copy);
1095 if (!(c == '_' || c == '$'
1096 || (c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z')))
1097 /* We must have come across a bad character (e.g. ';'). */
1098 error ("Invalid character '%c' in expression.", c);
1101 for (c = tokstart[namelen];
1102 (c == '_' || c == '$' || (c >= '0' && c <= '9')
1103 || (c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z'));
1104 c = tokstart[++namelen]);
1106 /* The token "if" terminates the expression and is NOT
1107 removed from the input stream. */
1109 if (namelen == 2 && tokstart[0] == 'i' && tokstart[1] == 'f')
1114 /* Catch specific keywords. */
1116 for (i = 0; f77_keywords[i].operator != NULL; i++)
1117 if (strncmp (tokstart, f77_keywords[i].operator,
1118 strlen(f77_keywords[i].operator)) == 0)
1120 /* lexptr += strlen(f77_keywords[i].operator); */
1121 yylval.opcode = f77_keywords[i].opcode;
1122 return f77_keywords[i].token;
1125 yylval.sval.ptr = tokstart;
1126 yylval.sval.length = namelen;
1128 if (*tokstart == '$')
1130 write_dollar_variable (yylval.sval);
1134 /* Use token-type TYPENAME for symbols that happen to be defined
1135 currently as names of types; NAME for other symbols.
1136 The caller is not constrained to care about the distinction. */
1138 char *tmp = copy_name (yylval.sval);
1140 int is_a_field_of_this = 0;
1143 sym = lookup_symbol (tmp, expression_context_block,
1145 current_language->la_language == language_cplus
1146 ? &is_a_field_of_this : NULL,
1148 if (sym && SYMBOL_CLASS (sym) == LOC_TYPEDEF)
1150 yylval.tsym.type = SYMBOL_TYPE (sym);
1154 = language_lookup_primitive_type_by_name (current_language,
1155 current_gdbarch, tmp);
1156 if (yylval.tsym.type != NULL)
1159 /* Input names that aren't symbols but ARE valid hex numbers,
1160 when the input radix permits them, can be names or numbers
1161 depending on the parse. Note we support radixes > 16 here. */
1163 && ((tokstart[0] >= 'a' && tokstart[0] < 'a' + input_radix - 10)
1164 || (tokstart[0] >= 'A' && tokstart[0] < 'A' + input_radix - 10)))
1166 YYSTYPE newlval; /* Its value is ignored. */
1167 hextype = parse_number (tokstart, namelen, 0, &newlval);
1170 yylval.ssym.sym = sym;
1171 yylval.ssym.is_a_field_of_this = is_a_field_of_this;
1176 /* Any other kind of symbol */
1177 yylval.ssym.sym = sym;
1178 yylval.ssym.is_a_field_of_this = is_a_field_of_this;
1188 lexptr = prev_lexptr;
1190 error ("A %s in expression, near `%s'.", (msg ? msg : "error"), lexptr);