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).
7 This file is part of GDB.
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.
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.
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., 675 Mass Ave, Cambridge, MA 02139, USA. */
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"
47 #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 */
55 /* Remap normal yacc parser interface names (yyparse, yylex, yyerror, etc),
56 as well as gratuitiously global symbol names, so we can have multiple
57 yacc generated parsers in gdb. Note that these are only the variables
58 produced by yacc. If other parser generators (bison, byacc, etc) produce
59 additional global names that conflict at link time, then those parser
60 generators need to be fixed instead of adding those names to this list. */
62 #define yymaxdepth f_maxdepth
63 #define yyparse f_parse
65 #define yyerror f_error
68 #define yydebug f_debug
77 #define yyerrflag f_errflag
78 #define yynerrs f_nerrs
83 #define yystate f_state
89 #define yyreds f_reds /* With YYDEBUG defined */
90 #define yytoks f_toks /* With YYDEBUG defined */
93 #define YYDEBUG 1 /* Default to no yydebug support */
96 int yyparse PARAMS ((void));
98 static int yylex PARAMS ((void));
100 void yyerror PARAMS ((char *));
104 /* Although the yacc "value" of an expression is not used,
105 since the result is stored in the structure being created,
106 other node types do have values. */
120 struct symtoken ssym;
123 enum exp_opcode opcode;
124 struct internalvar *ivar;
131 /* YYSTYPE gets defined by %union */
132 static int parse_number PARAMS ((char *, int, int, YYSTYPE *));
135 %type <voidval> exp type_exp start variable
136 %type <tval> type typebase
137 %type <tvec> nonempty_typelist
138 /* %type <bval> block */
140 /* Fancy type parsing. */
141 %type <voidval> func_mod direct_abs_decl abs_decl
144 %token <typed_val> INT
147 /* Both NAME and TYPENAME tokens represent symbols in the input,
148 and both convey their data as strings.
149 But a TYPENAME is a string that happens to be defined as a typedef
150 or builtin type name (such as int or char)
151 and a NAME is any other symbol.
152 Contexts where this distinction is not important can use the
153 nonterminal "name", which matches either NAME or TYPENAME. */
155 %token <sval> STRING_LITERAL
156 %token <lval> BOOLEAN_LITERAL
158 %token <tsym> TYPENAME
160 %type <ssym> name_not_typename
161 %type <tsym> typename
163 /* A NAME_OR_INT is a symbol which is not known in the symbol table,
164 but which would parse as a valid number in the current input radix.
165 E.g. "c" when input_radix==16. Depending on the parse, it will be
166 turned into a name or into a number. */
168 %token <ssym> NAME_OR_INT
173 /* Special type cases, put in to allow the parser to distinguish different
175 %token INT_KEYWORD INT_S2_KEYWORD LOGICAL_S1_KEYWORD LOGICAL_S2_KEYWORD
176 %token LOGICAL_KEYWORD REAL_KEYWORD REAL_S8_KEYWORD REAL_S16_KEYWORD
177 %token COMPLEX_S8_KEYWORD COMPLEX_S16_KEYWORD COMPLEX_S32_KEYWORD
178 %token BOOL_AND BOOL_OR BOOL_NOT
179 %token <lval> LAST REGNAME CHARACTER
181 %token <ivar> VARIABLE
183 %token <opcode> ASSIGN_MODIFY
187 %right '=' ASSIGN_MODIFY
196 %left LESSTHAN GREATERTHAN LEQ GEQ
212 { write_exp_elt_opcode(OP_TYPE);
213 write_exp_elt_type($1);
214 write_exp_elt_opcode(OP_TYPE); }
222 /* Expressions, not including the comma operator. */
223 exp : '*' exp %prec UNARY
224 { write_exp_elt_opcode (UNOP_IND); }
226 exp : '&' exp %prec UNARY
227 { write_exp_elt_opcode (UNOP_ADDR); }
229 exp : '-' exp %prec UNARY
230 { write_exp_elt_opcode (UNOP_NEG); }
233 exp : BOOL_NOT exp %prec UNARY
234 { write_exp_elt_opcode (UNOP_LOGICAL_NOT); }
237 exp : '~' exp %prec UNARY
238 { write_exp_elt_opcode (UNOP_COMPLEMENT); }
241 exp : SIZEOF exp %prec UNARY
242 { write_exp_elt_opcode (UNOP_SIZEOF); }
245 /* No more explicit array operators, we treat everything in F77 as
246 a function call. The disambiguation as to whether we are
247 doing a subscript operation or a function call is done
251 { start_arglist (); }
253 { write_exp_elt_opcode (OP_F77_UNDETERMINED_ARGLIST);
254 write_exp_elt_longcst ((LONGEST) end_arglist ());
255 write_exp_elt_opcode (OP_F77_UNDETERMINED_ARGLIST); }
268 arglist : arglist ',' exp %prec ABOVE_COMMA
272 substring: exp ':' exp %prec ABOVE_COMMA
277 complexnum: exp ',' exp
281 exp : '(' complexnum ')'
282 { write_exp_elt_opcode(OP_F77_LITERAL_COMPLEX); }
285 exp : '(' type ')' exp %prec UNARY
286 { write_exp_elt_opcode (UNOP_CAST);
287 write_exp_elt_type ($2);
288 write_exp_elt_opcode (UNOP_CAST); }
291 /* Binary operators in order of decreasing precedence. */
294 { write_exp_elt_opcode (BINOP_REPEAT); }
298 { write_exp_elt_opcode (BINOP_MUL); }
302 { write_exp_elt_opcode (BINOP_DIV); }
306 { write_exp_elt_opcode (BINOP_REM); }
310 { write_exp_elt_opcode (BINOP_ADD); }
314 { write_exp_elt_opcode (BINOP_SUB); }
318 { write_exp_elt_opcode (BINOP_LSH); }
322 { write_exp_elt_opcode (BINOP_RSH); }
326 { write_exp_elt_opcode (BINOP_EQUAL); }
329 exp : exp NOTEQUAL exp
330 { write_exp_elt_opcode (BINOP_NOTEQUAL); }
334 { write_exp_elt_opcode (BINOP_LEQ); }
338 { write_exp_elt_opcode (BINOP_GEQ); }
341 exp : exp LESSTHAN exp
342 { write_exp_elt_opcode (BINOP_LESS); }
345 exp : exp GREATERTHAN exp
346 { write_exp_elt_opcode (BINOP_GTR); }
350 { write_exp_elt_opcode (BINOP_BITWISE_AND); }
354 { write_exp_elt_opcode (BINOP_BITWISE_XOR); }
358 { write_exp_elt_opcode (BINOP_BITWISE_IOR); }
361 exp : exp BOOL_AND exp
362 { write_exp_elt_opcode (BINOP_LOGICAL_AND); }
366 exp : exp BOOL_OR exp
367 { write_exp_elt_opcode (BINOP_LOGICAL_OR); }
371 { write_exp_elt_opcode (BINOP_ASSIGN); }
374 exp : exp ASSIGN_MODIFY exp
375 { write_exp_elt_opcode (BINOP_ASSIGN_MODIFY);
376 write_exp_elt_opcode ($2);
377 write_exp_elt_opcode (BINOP_ASSIGN_MODIFY); }
381 { write_exp_elt_opcode (OP_LONG);
382 write_exp_elt_type ($1.type);
383 write_exp_elt_longcst ((LONGEST)($1.val));
384 write_exp_elt_opcode (OP_LONG); }
389 parse_number ($1.stoken.ptr, $1.stoken.length, 0, &val);
390 write_exp_elt_opcode (OP_LONG);
391 write_exp_elt_type (val.typed_val.type);
392 write_exp_elt_longcst ((LONGEST)val.typed_val.val);
393 write_exp_elt_opcode (OP_LONG);
398 { write_exp_elt_opcode (OP_DOUBLE);
399 write_exp_elt_type (builtin_type_f_real_s8);
400 write_exp_elt_dblcst ($1);
401 write_exp_elt_opcode (OP_DOUBLE); }
408 { write_exp_elt_opcode (OP_LAST);
409 write_exp_elt_longcst ((LONGEST) $1);
410 write_exp_elt_opcode (OP_LAST); }
414 { write_exp_elt_opcode (OP_REGISTER);
415 write_exp_elt_longcst ((LONGEST) $1);
416 write_exp_elt_opcode (OP_REGISTER); }
420 { write_exp_elt_opcode (OP_INTERNALVAR);
421 write_exp_elt_intern ($1);
422 write_exp_elt_opcode (OP_INTERNALVAR); }
425 exp : SIZEOF '(' type ')' %prec UNARY
426 { write_exp_elt_opcode (OP_LONG);
427 write_exp_elt_type (builtin_type_f_integer);
428 write_exp_elt_longcst ((LONGEST) TYPE_LENGTH ($3));
429 write_exp_elt_opcode (OP_LONG); }
432 exp : BOOLEAN_LITERAL
433 { write_exp_elt_opcode (OP_BOOL);
434 write_exp_elt_longcst ((LONGEST) $1);
435 write_exp_elt_opcode (OP_BOOL);
440 { /* In F77, we encounter string literals
441 basically in only one place:
442 when we are setting up manual parameter
443 lists to functions we call by hand or
444 when setting string vars to manual values.
445 These are character*N type variables.
446 They are treated specially behind the
447 scenes. Remember that the literal strings's
448 OPs are being emitted in reverse order, thus
449 we first have the elements and then
450 the array descriptor itself. */
451 char *sp = $1.ptr; int count = $1.length;
455 write_exp_elt_opcode (OP_LONG);
456 write_exp_elt_type (builtin_type_f_character);
457 write_exp_elt_longcst ((LONGEST)(*sp++));
458 write_exp_elt_opcode (OP_LONG);
460 write_exp_elt_opcode (OP_ARRAY);
461 write_exp_elt_longcst ((LONGEST) 1);
462 write_exp_elt_longcst ((LONGEST) ($1.length));
463 write_exp_elt_opcode (OP_ARRAY);
468 variable: name_not_typename
469 { struct symbol *sym = $1.sym;
473 if (symbol_read_needs_frame (sym))
475 if (innermost_block == 0 ||
476 contained_in (block_found,
478 innermost_block = block_found;
480 write_exp_elt_opcode (OP_VAR_VALUE);
481 write_exp_elt_sym (sym);
482 write_exp_elt_opcode (OP_VAR_VALUE);
487 struct minimal_symbol *msymbol;
488 register char *arg = copy_name ($1.stoken);
490 msymbol = lookup_minimal_symbol (arg, NULL);
493 write_exp_msymbol (msymbol,
494 lookup_function_type (builtin_type_int),
497 else if (!have_full_symbols () && !have_partial_symbols ())
498 error ("No symbol table is loaded. Use the \"file\" command.");
500 error ("No symbol \"%s\" in current context.",
501 copy_name ($1.stoken));
513 /* This is where the interesting stuff happens. */
516 struct type *follow_type = $1;
517 struct type *range_type;
526 follow_type = lookup_pointer_type (follow_type);
529 follow_type = lookup_reference_type (follow_type);
532 array_size = pop_type_int ();
533 if (array_size != -1)
536 create_range_type ((struct type *) NULL,
537 builtin_type_f_integer, 0,
540 create_array_type ((struct type *) NULL,
541 follow_type, range_type);
544 follow_type = lookup_pointer_type (follow_type);
547 follow_type = lookup_function_type (follow_type);
555 { push_type (tp_pointer); $$ = 0; }
557 { push_type (tp_pointer); $$ = $2; }
559 { push_type (tp_reference); $$ = 0; }
561 { push_type (tp_reference); $$ = $2; }
565 direct_abs_decl: '(' abs_decl ')'
567 | direct_abs_decl func_mod
568 { push_type (tp_function); }
570 { push_type (tp_function); }
575 | '(' nonempty_typelist ')'
576 { free ((PTR)$2); $$ = 0; }
579 typebase /* Implements (approximately): (type-qualifier)* type-specifier */
583 { $$ = builtin_type_f_integer; }
585 { $$ = builtin_type_f_integer_s2; }
587 { $$ = builtin_type_f_character; }
589 { $$ = builtin_type_f_logical;}
591 { $$ = builtin_type_f_logical_s2;}
593 { $$ = builtin_type_f_logical_s1;}
595 { $$ = builtin_type_f_real;}
597 { $$ = builtin_type_f_real_s8;}
599 { $$ = builtin_type_f_real_s16;}
601 { $$ = builtin_type_f_complex_s8;}
602 | COMPLEX_S16_KEYWORD
603 { $$ = builtin_type_f_complex_s16;}
604 | COMPLEX_S32_KEYWORD
605 { $$ = builtin_type_f_complex_s32;}
613 { $$ = (struct type **) malloc (sizeof (struct type *) * 2);
614 $<ivec>$[0] = 1; /* Number of types in vector */
617 | nonempty_typelist ',' type
618 { int len = sizeof (struct type *) * (++($<ivec>1[0]) + 1);
619 $$ = (struct type **) realloc ((char *) $1, len);
620 $$[$<ivec>$[0]] = $3;
632 name_not_typename : NAME
633 /* These would be useful if name_not_typename was useful, but it is just
634 a fake for "variable", so these cause reduce/reduce conflicts because
635 the parser can't tell whether NAME_OR_INT is a name_not_typename (=variable,
636 =exp) or just an exp. If name_not_typename was ever used in an lvalue
637 context where only a name could occur, this might be useful.
644 /* Take care of parsing a number (anything that starts with a digit).
645 Set yylval and return the token type; update lexptr.
646 LEN is the number of characters in it. */
648 /*** Needs some error checking for the float case ***/
651 parse_number (p, len, parsed_float, putithere)
657 register LONGEST n = 0;
658 register LONGEST prevn = 0;
661 register int base = input_radix;
664 unsigned LONGEST high_bit;
665 struct type *signed_type;
666 struct type *unsigned_type;
670 /* It's a float since it contains a point or an exponent. */
671 putithere->dval = atof (p);
675 /* Handle base-switching prefixes 0x, 0t, 0d, 0 */
709 if (c >= 'A' && c <= 'Z')
711 if (c != 'l' && c != 'u')
713 if (c >= '0' && c <= '9')
717 if (base > 10 && c >= 'a' && c <= 'f')
718 n += i = c - 'a' + 10;
719 else if (len == 0 && c == 'l')
721 else if (len == 0 && c == 'u')
724 return ERROR; /* Char not a digit */
727 return ERROR; /* Invalid digit in this base */
729 /* Portably test for overflow (only works for nonzero values, so make
730 a second check for zero). */
731 if ((prevn >= n) && n != 0)
732 unsigned_p=1; /* Try something unsigned */
733 /* If range checking enabled, portably test for unsigned overflow. */
734 if (RANGE_CHECK && n != 0)
736 if ((unsigned_p && (unsigned)prevn >= (unsigned)n))
737 range_error("Overflow on numeric constant.");
742 /* If the number is too big to be an int, or it's got an l suffix
743 then it's a long. Work out if this has to be a long by
744 shifting right and and seeing if anything remains, and the
745 target int size is different to the target long size.
747 In the expression below, we could have tested
748 (n >> TARGET_INT_BIT)
749 to see if it was zero,
750 but too many compilers warn about that, when ints and longs
751 are the same size. So we shift it twice, with fewer bits
752 each time, for the same result. */
754 if ((TARGET_INT_BIT != TARGET_LONG_BIT
755 && ((n >> 2) >> (TARGET_INT_BIT-2))) /* Avoid shift warning */
758 high_bit = ((unsigned LONGEST)1) << (TARGET_LONG_BIT-1);
759 unsigned_type = builtin_type_unsigned_long;
760 signed_type = builtin_type_long;
764 high_bit = ((unsigned LONGEST)1) << (TARGET_INT_BIT-1);
765 unsigned_type = builtin_type_unsigned_int;
766 signed_type = builtin_type_int;
769 putithere->typed_val.val = n;
771 /* If the high bit of the worked out type is set then this number
772 has to be unsigned. */
774 if (unsigned_p || (n & high_bit))
775 putithere->typed_val.type = unsigned_type;
777 putithere->typed_val.type = signed_type;
786 enum exp_opcode opcode;
789 static const struct token dot_ops[] =
791 { ".and.", BOOL_AND, BINOP_END },
792 { ".AND.", BOOL_AND, BINOP_END },
793 { ".or.", BOOL_OR, BINOP_END },
794 { ".OR.", BOOL_OR, BINOP_END },
795 { ".not.", BOOL_NOT, BINOP_END },
796 { ".NOT.", BOOL_NOT, BINOP_END },
797 { ".eq.", EQUAL, BINOP_END },
798 { ".EQ.", EQUAL, BINOP_END },
799 { ".eqv.", EQUAL, BINOP_END },
800 { ".NEQV.", NOTEQUAL, BINOP_END },
801 { ".neqv.", NOTEQUAL, BINOP_END },
802 { ".EQV.", EQUAL, BINOP_END },
803 { ".ne.", NOTEQUAL, BINOP_END },
804 { ".NE.", NOTEQUAL, BINOP_END },
805 { ".le.", LEQ, BINOP_END },
806 { ".LE.", LEQ, BINOP_END },
807 { ".ge.", GEQ, BINOP_END },
808 { ".GE.", GEQ, BINOP_END },
809 { ".gt.", GREATERTHAN, BINOP_END },
810 { ".GT.", GREATERTHAN, BINOP_END },
811 { ".lt.", LESSTHAN, BINOP_END },
812 { ".LT.", LESSTHAN, BINOP_END },
816 struct f77_boolean_val
822 static const struct f77_boolean_val boolean_values[] =
831 static const struct token f77_keywords[] =
833 { "complex_16", COMPLEX_S16_KEYWORD, BINOP_END },
834 { "complex_32", COMPLEX_S32_KEYWORD, BINOP_END },
835 { "character", CHARACTER, BINOP_END },
836 { "integer_2", INT_S2_KEYWORD, BINOP_END },
837 { "logical_1", LOGICAL_S1_KEYWORD, BINOP_END },
838 { "logical_2", LOGICAL_S2_KEYWORD, BINOP_END },
839 { "complex_8", COMPLEX_S8_KEYWORD, BINOP_END },
840 { "integer", INT_KEYWORD, BINOP_END },
841 { "logical", LOGICAL_KEYWORD, BINOP_END },
842 { "real_16", REAL_S16_KEYWORD, BINOP_END },
843 { "complex", COMPLEX_S8_KEYWORD, BINOP_END },
844 { "sizeof", SIZEOF, BINOP_END },
845 { "real_8", REAL_S8_KEYWORD, BINOP_END },
846 { "real", REAL_KEYWORD, BINOP_END },
850 /* Implementation of a dynamically expandable buffer for processing input
851 characters acquired through lexptr and building a value to return in
852 yylval. Ripped off from ch-exp.y */
854 static char *tempbuf; /* Current buffer contents */
855 static int tempbufsize; /* Size of allocated buffer */
856 static int tempbufindex; /* Current index into buffer */
858 #define GROWBY_MIN_SIZE 64 /* Minimum amount to grow buffer by */
860 #define CHECKBUF(size) \
862 if (tempbufindex + (size) >= tempbufsize) \
864 growbuf_by_size (size); \
869 /* Grow the static temp buffer if necessary, including allocating the first one
873 growbuf_by_size (count)
878 growby = max (count, GROWBY_MIN_SIZE);
879 tempbufsize += growby;
881 tempbuf = (char *) malloc (tempbufsize);
883 tempbuf = (char *) realloc (tempbuf, tempbufsize);
886 /* Blatantly ripped off from ch-exp.y. This routine recognizes F77
889 Recognize a string literal. A string literal is a nonzero sequence
890 of characters enclosed in matching single quotes, except that
891 a single character inside single quotes is a character literal, which
892 we reject as a string literal. To embed the terminator character inside
893 a string, it is simply doubled (I.E. 'this''is''one''string') */
896 match_string_literal ()
898 char *tokptr = lexptr;
900 for (tempbufindex = 0, tokptr++; *tokptr != '\0'; tokptr++)
903 if (*tokptr == *lexptr)
905 if (*(tokptr + 1) == *lexptr)
910 tempbuf[tempbufindex++] = *tokptr;
912 if (*tokptr == '\0' /* no terminator */
913 || tempbufindex == 0) /* no string */
917 tempbuf[tempbufindex] = '\0';
918 yylval.sval.ptr = tempbuf;
919 yylval.sval.length = tempbufindex;
921 return STRING_LITERAL;
925 /* Read one token, getting characters through lexptr. */
932 unsigned int i,token;
936 static char *tempbuf;
937 static int tempbufsize;
943 /* First of all, let us make sure we are not dealing with the
944 special tokens .true. and .false. which evaluate to 1 and 0. */
948 for (i=0;boolean_values[i].name != NULL;i++)
950 if STREQN(tokstart,boolean_values[i].name,
951 strlen(boolean_values[i].name))
953 lexptr += strlen(boolean_values[i].name);
954 yylval.lval = boolean_values[i].value;
955 return (BOOLEAN_LITERAL);
960 /* See if it is a special .foo. operator */
962 for (i = 0; dot_ops[i].operator != NULL; i++)
963 if (STREQN(tokstart, dot_ops[i].operator,
964 strlen(dot_ops[i].operator)))
966 lexptr += strlen(dot_ops[i].operator);
967 yylval.opcode = dot_ops[i].opcode;
968 return dot_ops[i].token;
971 switch (c = *tokstart)
983 token = match_string_literal ();
994 if (paren_depth == 0)
1001 if (comma_terminates && paren_depth == 0)
1007 /* Might be a floating point number. */
1008 if (lexptr[1] < '0' || lexptr[1] > '9')
1009 goto symbol; /* Nope, must be a symbol. */
1010 /* FALL THRU into number case. */
1023 /* It's a number. */
1024 int got_dot = 0, got_e = 0, got_d = 0, toktype;
1025 register char *p = tokstart;
1026 int hex = input_radix > 10;
1028 if (c == '0' && (p[1] == 'x' || p[1] == 'X'))
1033 else if (c == '0' && (p[1]=='t' || p[1]=='T' || p[1]=='d' || p[1]=='D'))
1041 if (!hex && !got_e && (*p == 'e' || *p == 'E'))
1042 got_dot = got_e = 1;
1043 else if (!hex && !got_e && (*p == 'd' || *p == 'D'))
1044 got_dot = got_d = 1;
1045 else if (!hex && !got_dot && *p == '.')
1047 else if ((got_e && (p[-1] == 'e' || p[-1] == 'E')
1048 || got_d && (p[-1] == 'd' || p[-1] == 'D'))
1049 && (*p == '-' || *p == '+'))
1050 /* This is the sign of the exponent, not the end of the
1053 /* We will take any letters or digits. parse_number will
1054 complain if past the radix, or if L or U are not final. */
1055 else if ((*p < '0' || *p > '9')
1056 && ((*p < 'a' || *p > 'z')
1057 && (*p < 'A' || *p > 'Z')))
1060 toktype = parse_number (tokstart, p - tokstart, got_dot|got_e|got_d,
1062 if (toktype == ERROR)
1064 char *err_copy = (char *) alloca (p - tokstart + 1);
1066 memcpy (err_copy, tokstart, p - tokstart);
1067 err_copy[p - tokstart] = 0;
1068 error ("Invalid number \"%s\".", err_copy);
1099 if (!(c == '_' || c == '$'
1100 || (c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z')))
1101 /* We must have come across a bad character (e.g. ';'). */
1102 error ("Invalid character '%c' in expression.", c);
1105 for (c = tokstart[namelen];
1106 (c == '_' || c == '$' || (c >= '0' && c <= '9')
1107 || (c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z'));
1108 c = tokstart[++namelen]);
1110 /* The token "if" terminates the expression and is NOT
1111 removed from the input stream. */
1113 if (namelen == 2 && tokstart[0] == 'i' && tokstart[1] == 'f')
1118 /* Handle the tokens $digits; also $ (short for $0) and $$ (short for $$1)
1119 and $$digits (equivalent to $<-digits> if you could type that).
1120 Make token type LAST, and put the number (the digits) in yylval. */
1123 if (*tokstart == '$')
1125 register int negate = 0;
1128 /* Double dollar means negate the number and add -1 as well.
1129 Thus $$ alone means -1. */
1130 if (namelen >= 2 && tokstart[1] == '$')
1137 /* Just dollars (one or two) */
1138 yylval.lval = - negate;
1141 /* Is the rest of the token digits? */
1142 for (; c < namelen; c++)
1143 if (!(tokstart[c] >= '0' && tokstart[c] <= '9'))
1147 yylval.lval = atoi (tokstart + 1 + negate);
1149 yylval.lval = - yylval.lval;
1154 /* Handle tokens that refer to machine registers:
1155 $ followed by a register name. */
1157 if (*tokstart == '$') {
1158 for (c = 0; c < NUM_REGS; c++)
1159 if (namelen - 1 == strlen (reg_names[c])
1160 && STREQN (tokstart + 1, reg_names[c], namelen - 1))
1165 for (c = 0; c < num_std_regs; c++)
1166 if (namelen - 1 == strlen (std_regs[c].name)
1167 && STREQN (tokstart + 1, std_regs[c].name, namelen - 1))
1169 yylval.lval = std_regs[c].regnum;
1173 /* Catch specific keywords. */
1175 for (i = 0; f77_keywords[i].operator != NULL; i++)
1176 if (STREQN(tokstart, f77_keywords[i].operator,
1177 strlen(f77_keywords[i].operator)))
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 /* Any other names starting in $ are debugger internal variables. */
1189 if (*tokstart == '$')
1191 yylval.ivar = lookup_internalvar (copy_name (yylval.sval) + 1);
1195 /* Use token-type TYPENAME for symbols that happen to be defined
1196 currently as names of types; NAME for other symbols.
1197 The caller is not constrained to care about the distinction. */
1199 char *tmp = copy_name (yylval.sval);
1201 int is_a_field_of_this = 0;
1204 sym = lookup_symbol (tmp, expression_context_block,
1206 current_language->la_language == language_cplus
1207 ? &is_a_field_of_this : NULL,
1209 if (sym && SYMBOL_CLASS (sym) == LOC_TYPEDEF)
1211 yylval.tsym.type = SYMBOL_TYPE (sym);
1214 if ((yylval.tsym.type = lookup_primitive_typename (tmp)) != 0)
1217 /* Input names that aren't symbols but ARE valid hex numbers,
1218 when the input radix permits them, can be names or numbers
1219 depending on the parse. Note we support radixes > 16 here. */
1221 && ((tokstart[0] >= 'a' && tokstart[0] < 'a' + input_radix - 10)
1222 || (tokstart[0] >= 'A' && tokstart[0] < 'A' + input_radix - 10)))
1224 YYSTYPE newlval; /* Its value is ignored. */
1225 hextype = parse_number (tokstart, namelen, 0, &newlval);
1228 yylval.ssym.sym = sym;
1229 yylval.ssym.is_a_field_of_this = is_a_field_of_this;
1234 /* Any other kind of symbol */
1235 yylval.ssym.sym = sym;
1236 yylval.ssym.is_a_field_of_this = is_a_field_of_this;
1245 error ("A %s in expression, near `%s'.", (msg ? msg : "error"), lexptr);