/* YACC grammar for Chill expressions, for GDB. Copyright (C) 1992 Free Software Foundation, Inc. This file is part of GDB. This program is free software; you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation; either version 2 of the License, or (at your option) any later version. This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. You should have received a copy of the GNU General Public License along with this program; if not, write to the Free Software Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. */ /* Parse a Chill expression from text in a string, and return the result as a struct expression pointer. That structure contains arithmetic operations in reverse polish, with constants represented by operations that are followed by special data. See expression.h for the details of the format. What is important here is that it can be built up sequentially during the process of parsing; the lower levels of the tree always come first in the result. Note that malloc's and realloc's in this file are transformed to xmalloc and xrealloc respectively by the same sed command in the makefile that remaps any other malloc/realloc inserted by the parser generator. Doing this with #defines and trying to control the interaction with include files ( and for example) just became too messy, particularly when such includes can be inserted at random times by the parser generator. Also note that the language accepted by this parser is more liberal than the one accepted by an actual Chill compiler. For example, the language rule that a simple name string can not be one of the reserved simple name strings is not enforced (e.g "case" is not treated as a reserved name). Another example is that Chill is a strongly typed language, and certain expressions that violate the type constraints may still be evaluated if gdb can do so in a meaningful manner, while such expressions would be rejected by the compiler. The reason for this more liberal behavior is the philosophy that the debugger is intended to be a tool that is used by the programmer when things go wrong, and as such, it should provide as few artificial barriers to it's use as possible. If it can do something meaningful, even something that violates language contraints that are enforced by the compiler, it should do so without complaint. */ %{ #include "defs.h" #include "expression.h" #include "language.h" #include "value.h" #include "parser-defs.h" #include "ch-lang.h" /* These MUST be included in any grammar file!!!! Please choose unique names! Note that this are a combined list of variables that can be produced by any one of bison, byacc, or yacc. */ #define yymaxdepth chill_maxdepth #define yyparse chill_parse #define yylex chill_lex #define yyerror chill_error #define yylval chill_lval #define yychar chill_char #define yydebug chill_debug #define yypact chill_pact #define yyr1 chill_r1 #define yyr2 chill_r2 #define yydef chill_def #define yychk chill_chk #define yypgo chill_pgo #define yyact chill_act #define yyexca chill_exca #define yyerrflag chill_errflag #define yynerrs chill_nerrs #define yyps chill_ps #define yypv chill_pv #define yys chill_s #define yy_yys chill_yys #define yystate chill_state #define yytmp chill_tmp #define yyv chill_v #define yy_yyv chill_yyv #define yyval chill_val #define yylloc chill_lloc #define yyss chill_yyss /* byacc */ #define yyssp chill_yysp /* byacc */ #define yyvs chill_yyvs /* byacc */ #define yyvsp chill_yyvsp /* byacc */ static int yylex PARAMS ((void)); void yyerror PARAMS ((char *)); int yyparse PARAMS ((void)); /* #define YYDEBUG 1 */ %} /* Although the yacc "value" of an expression is not used, since the result is stored in the structure being created, other node types do have values. */ %union { LONGEST lval; unsigned LONGEST ulval; struct { LONGEST val; struct type *type; } typed_val; double dval; struct symbol *sym; struct type *tval; struct stoken sval; struct ttype tsym; struct symtoken ssym; int voidval; struct block *bval; enum exp_opcode opcode; struct internalvar *ivar; struct type **tvec; int *ivec; } %{ static int parse_number PARAMS ((void)); %} %token FIXME %token INTEGER_LITERAL %token BOOLEAN_LITERAL %token CHARACTER_LITERAL %token GENERAL_PROCEDURE_NAME %token LOCATION_NAME %token SET_LITERAL %token EMPTINESS_LITERAL %token CHARACTER_STRING_LITERAL %token BIT_STRING_LITERAL %token STRING %token CONSTANT %token '.' %token ';' %token ':' %token CASE %token OF %token ESAC %token LOGIOR %token ORIF %token LOGXOR %token LOGAND %token ANDIF %token '=' %token NOTEQUAL %token '>' %token GTR %token '<' %token LEQ %token IN %token '+' %token '-' %token '*' %token '/' %token SLASH_SLASH %token MOD %token REM %token NOT %token POINTER %token RECEIVE %token SC %token '[' %token ']' %token '(' %token ')' %token UP %token IF %token THEN %token ELSE %token FI %token ELSIF %token ILLEGAL_TOKEN %type location %type access_name %type primitive_value %type location_contents %type value_name %type literal %type tuple %type value_string_element %type value_string_slice %type value_array_element %type value_array_slice %type value_structure_field %type expression_conversion %type value_procedure_call %type value_built_in_routine_call %type start_expression %type zero_adic_operator %type parenthesised_expression %type value %type undefined_value %type expression %type conditional_expression %type then_alternative %type else_alternative %type sub_expression %type value_case_alternative %type operand_0 %type operand_1 %type operand_2 %type operand_3 %type operand_4 %type operand_5 %type operand_6 %type integer_literal_expression %type synonym_name %type value_enumeration_name %type value_do_with_name %type value_receive_name %type string_primitive_value %type start_element %type left_element %type right_element %type slice_size %type array_primitive_value %type expression_list %type lower_element %type upper_element %type first_element %type structure_primitive_value %type field_name %type mode_name %type boolean_expression %type case_selector_list %type subexpression %type case_label_specification %type buffer_location %% /* Z.200, 5.3.1 */ value : expression { $$ = 0; /* FIXME */ } | undefined_value { $$ = 0; /* FIXME */ } ; undefined_value : FIXME { $$ = 0; /* FIXME */ } ; /* Z.200, 4.2.1 */ location : access_name { $$ = 0; /* FIXME */ } | FIXME { $$ = 0; /* FIXME */ } ; /* Z.200, 4.2.2 */ access_name : LOCATION_NAME { write_exp_elt_opcode (OP_VAR_VALUE); write_exp_elt_sym ($1.sym); write_exp_elt_opcode (OP_VAR_VALUE); } | FIXME { $$ = 0; /* FIXME */ } ; /* Z.200, 5.2.1 */ primitive_value : location_contents { $$ = 0; /* FIXME */ } | value_name { $$ = 0; /* FIXME */ } | literal { $$ = 0; /* FIXME */ } | tuple { $$ = 0; /* FIXME */ } | value_string_element { $$ = 0; /* FIXME */ } | value_string_slice { $$ = 0; /* FIXME */ } | value_array_element { $$ = 0; /* FIXME */ } | value_array_slice { $$ = 0; /* FIXME */ } | value_structure_field { $$ = 0; /* FIXME */ } | expression_conversion { $$ = 0; /* FIXME */ } | value_procedure_call { $$ = 0; /* FIXME */ } | value_built_in_routine_call { $$ = 0; /* FIXME */ } | start_expression { $$ = 0; /* FIXME */ } | zero_adic_operator { $$ = 0; /* FIXME */ } | parenthesised_expression { $$ = 0; /* FIXME */ } ; /* Z.200, 5.2.2 */ location_contents: location { $$ = 0; /* FIXME */ } ; /* Z.200, 5.2.3 */ value_name : synonym_name { $$ = 0; /* FIXME */ } | value_enumeration_name { $$ = 0; /* FIXME */ } | value_do_with_name { $$ = 0; /* FIXME */ } | value_receive_name { $$ = 0; /* FIXME */ } | GENERAL_PROCEDURE_NAME { write_exp_elt_opcode (OP_VAR_VALUE); write_exp_elt_sym ($1.sym); write_exp_elt_opcode (OP_VAR_VALUE); } ; /* Z.200, 5.2.4.1 */ literal : INTEGER_LITERAL { write_exp_elt_opcode (OP_LONG); write_exp_elt_type ($1.type); write_exp_elt_longcst ((LONGEST) ($1.val)); write_exp_elt_opcode (OP_LONG); } | BOOLEAN_LITERAL { write_exp_elt_opcode (OP_BOOL); write_exp_elt_longcst ((LONGEST) $1); write_exp_elt_opcode (OP_BOOL); } | CHARACTER_LITERAL { write_exp_elt_opcode (OP_LONG); write_exp_elt_type ($1.type); write_exp_elt_longcst ((LONGEST) ($1.val)); write_exp_elt_opcode (OP_LONG); } | SET_LITERAL { $$ = 0; /* FIXME */ } | EMPTINESS_LITERAL { $$ = 0; /* FIXME */ } | CHARACTER_STRING_LITERAL { $$ = 0; /* FIXME */ } | BIT_STRING_LITERAL { $$ = 0; /* FIXME */ } ; /* Z.200, 5.2.5 */ tuple : FIXME { $$ = 0; /* FIXME */ } ; /* Z.200, 5.2.6 */ value_string_element: string_primitive_value '(' start_element ')' { $$ = 0; /* FIXME */ } ; /* Z.200, 5.2.7 */ value_string_slice: string_primitive_value '(' left_element ':' right_element ')' { $$ = 0; /* FIXME */ } | string_primitive_value '(' start_element UP slice_size ')' { $$ = 0; /* FIXME */ } ; /* Z.200, 5.2.8 */ value_array_element: array_primitive_value '(' expression_list ')' { $$ = 0; /* FIXME */ } ; /* Z.200, 5.2.9 */ value_array_slice: array_primitive_value '(' lower_element ':' upper_element ')' { $$ = 0; /* FIXME */ } | array_primitive_value '(' first_element UP slice_size '(' { $$ = 0; /* FIXME */ } ; /* Z.200, 5.2.10 */ value_structure_field: structure_primitive_value '.' field_name { $$ = 0; /* FIXME */ } ; /* Z.200, 5.2.11 */ expression_conversion: mode_name '(' expression ')' { $$ = 0; /* FIXME */ } ; /* Z.200, 5.2.12 */ value_procedure_call: FIXME { $$ = 0; /* FIXME */ } ; /* Z.200, 5.2.13 */ value_built_in_routine_call: FIXME { $$ = 0; /* FIXME */ } ; /* Z.200, 5.2.14 */ start_expression: FIXME { $$ = 0; /* FIXME */ } /* Not in GNU-Chill */ ; /* Z.200, 5.2.15 */ zero_adic_operator: FIXME { $$ = 0; /* FIXME */ } ; /* Z.200, 5.2.16 */ parenthesised_expression: '(' expression ')' { $$ = 0; /* FIXME */ } ; /* Z.200, 5.3.2 */ expression : operand_0 { $$ = 0; /* FIXME */ } | conditional_expression { $$ = 0; /* FIXME */ } ; conditional_expression : IF boolean_expression then_alternative else_alternative FI { $$ = 0; /* FIXME */ } | CASE case_selector_list OF value_case_alternative '[' ELSE sub_expression ']' ESAC { $$ = 0; /* FIXME */ } ; then_alternative: THEN subexpression { $$ = 0; /* FIXME */ } ; else_alternative: ELSE subexpression { $$ = 0; /* FIXME */ } | ELSIF boolean_expression then_alternative else_alternative { $$ = 0; /* FIXME */ } ; sub_expression : expression { $$ = 0; /* FIXME */ } ; value_case_alternative: case_label_specification ':' sub_expression ';' { $$ = 0; /* FIXME */ } ; /* Z.200, 5.3.3 */ operand_0 : operand_1 { $$ = 0; /* FIXME */ } | operand_0 LOGIOR operand_1 { write_exp_elt_opcode (BINOP_BITWISE_IOR); } | operand_0 ORIF operand_1 { $$ = 0; /* FIXME */ } | operand_0 LOGXOR operand_1 { write_exp_elt_opcode (BINOP_BITWISE_XOR); } ; /* Z.200, 5.3.4 */ operand_1 : operand_2 { $$ = 0; /* FIXME */ } | operand_1 LOGAND operand_2 { write_exp_elt_opcode (BINOP_BITWISE_AND); } | operand_1 ANDIF operand_2 { $$ = 0; /* FIXME */ } ; /* Z.200, 5.3.5 */ operand_2 : operand_3 { $$ = 0; /* FIXME */ } | operand_2 '=' operand_3 { write_exp_elt_opcode (BINOP_EQUAL); } | operand_2 NOTEQUAL operand_3 { write_exp_elt_opcode (BINOP_NOTEQUAL); } | operand_2 '>' operand_3 { write_exp_elt_opcode (BINOP_GTR); } | operand_2 GTR operand_3 { write_exp_elt_opcode (BINOP_GEQ); } | operand_2 '<' operand_3 { write_exp_elt_opcode (BINOP_LESS); } | operand_2 LEQ operand_3 { write_exp_elt_opcode (BINOP_LEQ); } | operand_2 IN operand_3 { $$ = 0; /* FIXME */ } ; /* Z.200, 5.3.6 */ operand_3 : operand_4 { $$ = 0; /* FIXME */ } | operand_3 '+' operand_4 { write_exp_elt_opcode (BINOP_ADD); } | operand_3 '-' operand_4 { write_exp_elt_opcode (BINOP_SUB); } | operand_3 SLASH_SLASH operand_4 { $$ = 0; /* FIXME */ } ; /* Z.200, 5.3.7 */ operand_4 : operand_5 { $$ = 0; /* FIXME */ } | operand_4 '*' operand_5 { write_exp_elt_opcode (BINOP_MUL); } | operand_4 '/' operand_5 { write_exp_elt_opcode (BINOP_DIV); } | operand_4 MOD operand_5 { $$ = 0; /* FIXME */ } | operand_4 REM operand_5 { $$ = 0; /* FIXME */ } ; /* Z.200, 5.3.8 */ operand_5 : operand_6 { $$ = 0; /* FIXME */ } | '-' operand_6 { write_exp_elt_opcode (UNOP_NEG); } | NOT operand_6 { write_exp_elt_opcode (UNOP_LOGICAL_NOT); } | '(' integer_literal_expression ')' operand_6 { $$ = 0; /* FIXME */ } ; /* Z.200, 5.3.9 */ operand_6 : POINTER location { $$ = 0; /* FIXME */ } | RECEIVE buffer_location { $$ = 0; /* FIXME */ } | primitive_value { $$ = 0; /* FIXME */ } ; /* Z.200, 12.4.3 */ /* FIXME: For now we just accept only a single integer literal. */ integer_literal_expression: INTEGER_LITERAL { $$ = 0; } /* Things which still need productions... */ synonym_name : FIXME { $$ = 0; } value_enumeration_name : FIXME { $$ = 0; } value_do_with_name : FIXME { $$ = 0; } value_receive_name : FIXME { $$ = 0; } string_primitive_value : FIXME { $$ = 0; } start_element : FIXME { $$ = 0; } left_element : FIXME { $$ = 0; } right_element : FIXME { $$ = 0; } slice_size : FIXME { $$ = 0; } array_primitive_value : FIXME { $$ = 0; } expression_list : FIXME { $$ = 0; } lower_element : FIXME { $$ = 0; } upper_element : FIXME { $$ = 0; } first_element : FIXME { $$ = 0; } structure_primitive_value: FIXME { $$ = 0; } field_name : FIXME { $$ = 0; } mode_name : FIXME { $$ = 0; } boolean_expression : FIXME { $$ = 0; } case_selector_list : FIXME { $$ = 0; } subexpression : FIXME { $$ = 0; } case_label_specification: FIXME { $$ = 0; } buffer_location : FIXME { $$ = 0; } %% /* Try to consume a simple name string token. If successful, returns a pointer to a nullbyte terminated copy of the name that can be used in symbol table lookups. If not successful, returns NULL. */ static char * match_simple_name_string () { char *tokptr = lexptr; if (isalpha (*tokptr)) { do { tokptr++; } while (isalpha (*tokptr) || isdigit (*tokptr) || (*tokptr == '_')); yylval.sval.ptr = lexptr; yylval.sval.length = tokptr - lexptr; lexptr = tokptr; return (copy_name (yylval.sval)); } return (NULL); } /* Start looking for a value composed of valid digits as set by the base in use. Note that '_' characters are valid anywhere, in any quantity, and are simply ignored. Since we must find at least one valid digit, or reject this token as an integer literal, we keep track of how many digits we have encountered. */ static int decode_integer_value (base, tokptrptr, ivalptr) int base; char **tokptrptr; int *ivalptr; { char *tokptr = *tokptrptr; int temp; int digits = 0; while (*tokptr != '\0') { temp = tolower (*tokptr); tokptr++; switch (temp) { case '_': continue; case '0': case '1': case '2': case '3': case '4': case '5': case '6': case '7': case '8': case '9': temp -= '0'; break; case 'a': case 'b': case 'c': case 'd': case 'e': case 'f': temp -= 'a'; temp += 10; break; default: temp = base; break; } if (temp < base) { digits++; *ivalptr *= base; *ivalptr += temp; } else { /* Found something not in domain for current base. */ tokptr--; /* Unconsume what gave us indigestion. */ break; } } /* If we didn't find any digits, then we don't have a valid integer value, so reject the entire token. Otherwise, update the lexical scan pointer, and return non-zero for success. */ if (digits == 0) { return (0); } else { *tokptrptr = tokptr; return (1); } } static int decode_integer_literal (valptr, tokptrptr) int *valptr; char **tokptrptr; { char *tokptr = *tokptrptr; int base = 0; int ival = 0; int digits = 0; int temp; int explicit_base = 0; /* Look for an explicit base specifier, which is optional. */ switch (*tokptr) { case 'd': case 'D': explicit_base++; base = 10; tokptr++; break; case 'b': case 'B': explicit_base++; base = 2; tokptr++; break; case 'h': case 'H': explicit_base++; base = 16; tokptr++; break; case 'o': case 'O': explicit_base++; base = 8; tokptr++; break; default: base = 10; break; } /* If we found an explicit base ensure that the character after the explicit base is a single quote. */ if (explicit_base && (*tokptr++ != '\'')) { return (0); } /* Attempt to decode whatever follows as an integer value in the indicated base, updating the token pointer in the process and computing the value into ival. Also, if we have an explicit base, then the next character must not be a single quote, or we have a bitstring literal, so reject the entire token in this case. Otherwise, update the lexical scan pointer, and return non-zero for success. */ if (!decode_integer_value (base, &tokptr, &ival)) { return (0); } else if (explicit_base && (*tokptr == '\'')) { return (0); } else { *valptr = ival; *tokptrptr = tokptr; return (1); } } /* Recognize a character literal. A character literal is single character or a control sequence, enclosed in single quotes. A control sequence is a comma separated list of one or more integer literals, enclosed in parenthesis and introduced with a circumflex character. EX: 'a' '^(7)' '^(7,8)' As a GNU chill extension, the syntax C'xx' is also recognized as a character literal, where xx is a hex value for the character. Returns CHARACTER_LITERAL if a match is found. */ static int match_character_literal () { char *tokptr = lexptr; int ival = 0; if ((tolower (*tokptr) == 'c') && (*(tokptr + 1) == '\'')) { /* We have a GNU chill extension form, so skip the leading "C'", decode the hex value, and then ensure that we have a trailing single quote character. */ tokptr += 2; if (!decode_integer_value (16, &tokptr, &ival) || (*tokptr != '\'')) { return (0); } tokptr++; } else if (*tokptr == '\'') { tokptr++; /* Determine which form we have, either a control sequence or the single character form. */ if ((*tokptr == '^') && (*(tokptr + 1) == '(')) { /* Match and decode a control sequence. Return zero if we don't find a valid integer literal, or if the next unconsumed character after the integer literal is not the trailing ')'. FIXME: We currently don't handle the multiple integer literal form. */ tokptr += 2; if (!decode_integer_literal (&ival, &tokptr) || (*tokptr++ != ')')) { return (0); } } else { ival = *tokptr++; } /* The trailing quote has not yet been consumed. If we don't find it, then we have no match. */ if (*tokptr++ != '\'') { return (0); } } else { /* Not a character literal. */ return (0); } yylval.typed_val.val = ival; yylval.typed_val.type = builtin_type_chill_char; lexptr = tokptr; return (CHARACTER_LITERAL); } /* Recognize an integer literal, as specified in Z.200 sec 5.2.4.2. Note that according to 5.2.4.2, a single "_" is also a valid integer literal, however GNU-chill requires there to be at least one "digit" in any integer literal. */ static int match_integer_literal () { char *tokptr = lexptr; int ival; if (!decode_integer_literal (&ival, &tokptr)) { return (0); } else { yylval.typed_val.val = ival; yylval.typed_val.type = builtin_type_int; lexptr = tokptr; return (INTEGER_LITERAL); } } static void convert_float () { #if 0 extern double strtod (); double d; char tmp[256]; char *p = yytext, *p1 = tmp; char c; while (c = *p++) { switch (c) { case '_': break; case 'E': case 'd': case 'D': *p1++ = 'e'; break; default: *p1++ = c; break; } } *p1 = '\0'; d = strtod (tmp, &p1); if (*p1) { /* add error handling here */ ; } yylval.dval = d; #endif } /* Take care of parsing a number (anything that starts with a digit). Set yylval and return the token type; update lexptr. LEN is the number of characters in it. */ /*** Needs some error checking for the float case ***/ static int parse_number () { } struct token { char *operator; int token; }; const static struct token tokentab5[] = { { "ANDIF", ANDIF } }; const static struct token tokentab4[] = { { "ORIF", ORIF } }; const static struct token tokentab3[] = { { "NOT", NOT }, { "XOR", LOGXOR }, { "AND", LOGAND } }; const static struct token tokentab2[] = { { "//", SLASH_SLASH }, { "/=", NOTEQUAL }, { "<=", LEQ }, { ">=", GTR }, { "IN", IN }, { "OR", LOGIOR } }; /* Read one token, getting characters through lexptr. */ /* This is where we will check to make sure that the language and the operators used are compatible. */ static int yylex () { unsigned int i; int token; char *simplename; struct symbol *sym; /* Skip over any leading whitespace. */ while (isspace (*lexptr)) { lexptr++; } /* Look for special single character cases which can't be the first character of some other multicharacter token. */ switch (*lexptr) { case '\0': return (0); case '.': case '=': case ':': case ';': case '!': case '+': case '-': case '*': case '/': case '(': case ')': case '[': case ']': return (*lexptr++); } /* Look for characters which start a particular kind of multicharacter token, such as a character literal. */ switch (*lexptr) { case 'C': case 'c': case '\'': token = match_character_literal (); if (token != 0) { return (token); } break; } /* See if it is a special token of length 5. */ for (i = 0; i < sizeof (tokentab5) / sizeof (tokentab5[0]); i++) { if (strncmp (lexptr, tokentab5[i].operator, 5) == 0) { lexptr += 5; return (tokentab5[i].token); } } /* See if it is a special token of length 4. */ for (i = 0; i < sizeof (tokentab4) / sizeof (tokentab4[0]); i++) { if (strncmp (lexptr, tokentab4[i].operator, 4) == 0) { lexptr += 4; return (tokentab4[i].token); } } /* See if it is a special token of length 3. */ for (i = 0; i < sizeof (tokentab3) / sizeof (tokentab3[0]); i++) { if (strncmp (lexptr, tokentab3[i].operator, 3) == 0) { lexptr += 3; return (tokentab3[i].token); } } /* See if it is a special token of length 2. */ for (i = 0; i < sizeof (tokentab2) / sizeof (tokentab2[0]); i++) { if (strncmp (lexptr, tokentab2[i].operator, 2) == 0) { lexptr += 2; return (tokentab2[i].token); } } /* Look for single character cases which which could be the first character of some other multicharacter token, but aren't, or we would already have found it. */ switch (*lexptr) { case '/': case '<': case '>': return (*lexptr++); } /* Look for other special tokens. */ if (strncmp (lexptr, "TRUE", 4) == 0) /* FIXME: What about lowercase? */ { yylval.ulval = 1; lexptr += 4; return (BOOLEAN_LITERAL); } if (strncmp (lexptr, "FALSE", 5) == 0) /* FIXME: What about lowercase? */ { yylval.ulval = 0; lexptr += 5; return (BOOLEAN_LITERAL); } token = match_integer_literal (); if (token != 0) { return (token); } /* Try to match a simple name string, and if a match is found, then further classify what sort of name it is and return an appropriate token. Note that attempting to match a simple name string consumes the token from lexptr, so we can't back out if we later find that we can't classify what sort of name it is. */ simplename = match_simple_name_string (); if (simplename != NULL) { sym = lookup_symbol (simplename, expression_context_block, VAR_NAMESPACE, (int *) NULL, (struct symtab **) NULL); if (sym != NULL) { yylval.ssym.stoken.ptr = NULL; yylval.ssym.stoken.length = 0; yylval.ssym.sym = sym; yylval.ssym.is_a_field_of_this = 0; /* FIXME, C++'ism */ switch (SYMBOL_CLASS (sym)) { case LOC_BLOCK: /* Found a procedure name. */ return (GENERAL_PROCEDURE_NAME); case LOC_STATIC: /* Found a global or local static variable. */ return (LOCATION_NAME); } } else if (!have_full_symbols () && !have_partial_symbols ()) { error ("No symbol table is loaded. Use the \"file\" command."); } else { error ("No symbol \"%s\" in current context.", simplename); } } return (ILLEGAL_TOKEN); } void yyerror (msg) char *msg; /* unused */ { printf ("Parsing: %s\n", lexptr); if (yychar < 256) { error ("Invalid syntax in expression near character '%c'.", yychar); } else { error ("Invalid syntax in expression"); } }