X-Git-Url: http://nsz.repo.hu/git/?a=blobdiff_plain;f=parser.c;h=31d692bcc780c62571579a6d82f5814e39e609d2;hb=3e3bf65b33ece40f474c91dac37cf8ea716a5b03;hp=58045f3f65789552a9b5c771e09adf1e66911962;hpb=bc2c469866488880921a9809d8dee3bd1333a25c;p=cparser diff --git a/parser.c b/parser.c index 58045f3..31d692b 100644 --- a/parser.c +++ b/parser.c @@ -85,7 +85,7 @@ void next_token(void) } static inline -const token_t *la(int num) +const token_t *look_ahead(int num) { assert(num > 0 && num <= MAX_LOOKAHEAD); int pos = (lookahead_bufpos+num-1) % MAX_LOOKAHEAD; @@ -152,21 +152,60 @@ void parse_error_expected(const char *message, ...) fprintf(stderr, "\n"); } -static -void eat_until(int token_type) +static void eat_block(void) { - while(token.type != token_type) { + if(token.type == '{') + next_token(); + + while(token.type != '}') { if(token.type == T_EOF) return; + if(token.type == '{') { + eat_block(); + continue; + } next_token(); } - next_token(); + eat('}'); +} + +static void eat_statement(void) +{ + while(token.type != ';') { + if(token.type == T_EOF) + return; + if(token.type == '}') + return; + if(token.type == '{') { + eat_block(); + continue; + } + next_token(); + } + eat(';'); +} + +static void eat_brace(void) +{ + if(token.type == '(') + next_token(); + + while(token.type != ')') { + if(token.type == T_EOF) + return; + if(token.type == '{') { + eat_block(); + continue; + } + next_token(); + } + eat(')'); } #define expect(expected) \ if(UNLIKELY(token.type != (expected))) { \ parse_error_expected(NULL, (expected), 0); \ - eat_until(';'); \ + eat_statement(); \ return NULL; \ } \ next_token(); @@ -174,7 +213,7 @@ void eat_until(int token_type) #define expect_void(expected) \ if(UNLIKELY(token.type != (expected))) { \ parse_error_expected(NULL, (expected), 0); \ - eat_until(';'); \ + eat_statement(); \ return; \ } \ next_token(); @@ -195,12 +234,23 @@ static void set_context(context_t *new_context) last_declaration = declaration; } +/** + * called when we find a 2nd declarator for an identifier we already have a + * declarator for + */ +static int is_compatible_declaration (declaration_t *declaration, + declaration_t *previous) +{ + /* TODO: not correct yet */ + return declaration->type == previous->type; +} + /** * pushs an environment_entry on the environment stack and links the * corresponding symbol to the new entry */ static inline -void environment_push(declaration_t *declaration, const void *context) +declaration_t *environment_push(declaration_t *declaration, const void *context) { environment_entry_t *entry = obstack_alloc(&environment_obstack, sizeof(entry[0])); @@ -216,13 +266,22 @@ void environment_push(declaration_t *declaration, const void *context) assert(declaration != symbol->declaration); if(symbol->context == context) { + declaration_t *previous_declaration = symbol->declaration; if(symbol->declaration != NULL) { - assert(symbol->declaration != NULL); - parser_print_error_prefix_pos(declaration->source_position); - fprintf(stderr, "multiple definitions for symbol '%s'.\n", - symbol->string); - parser_print_error_prefix_pos(symbol->declaration->source_position); - fprintf(stderr, "this is the location of the previous declaration.\n"); + if(!is_compatible_declaration(declaration, previous_declaration)) { + parser_print_error_prefix_pos(declaration->source_position); + fprintf(stderr, "definition of symbol '%s' with type ", + declaration->symbol->string); + print_type(declaration->type, NULL); + fputc('\n', stderr); + parser_print_error_prefix_pos( + previous_declaration->source_position); + fprintf(stderr, "is incompatible with previous declaration " + "of type "); + print_type(previous_declaration->type, NULL); + fputc('\n', stderr); + } + return previous_declaration; } } @@ -231,6 +290,8 @@ void environment_push(declaration_t *declaration, const void *context) entry->symbol = symbol; symbol->declaration = declaration; symbol->context = context; + + return declaration; } /** @@ -281,8 +342,7 @@ static void parse_compound_type_entries(void); static void parse_declarator(declaration_t *declaration, storage_class_t storage_class, type_t *type, int may_be_abstract); -static void maybe_push_declaration(declaration_t *declaration); -static void record_declaration(declaration_t *declaration); +static declaration_t *record_declaration(declaration_t *declaration); typedef struct declaration_specifiers_t declaration_specifiers_t; struct declaration_specifiers_t { @@ -290,100 +350,129 @@ struct declaration_specifiers_t { type_t *type; }; -static type_t *parse_struct_specifier(void) +static compound_type_t *find_compound_type(compound_type_t *types, + const symbol_t *symbol) { - eat(T_struct); + compound_type_t *type = types; + for( ; type != NULL; type = type->next) { + if(type->symbol == symbol) + return type; + } + + return NULL; +} - compound_type_t *struct_type = allocate_type_zero(sizeof(struct_type[0])); - struct_type->type.type = TYPE_COMPOUND_STRUCT; - struct_type->source_position = token.source_position; +static type_t *parse_compound_type_specifier(int is_struct) +{ + if(is_struct) { + eat(T_struct); + } else { + eat(T_union); + } - int top = environment_top(); - context_t *last_context = context; - set_context(&struct_type->context); + symbol_t *symbol = NULL; + compound_type_t *compound_type = NULL; if(token.type == T_IDENTIFIER) { + symbol = token.v.symbol; next_token(); - if(token.type == '{') { - parse_compound_type_entries(); + + if(context != NULL) { + if(is_struct) { + compound_type = find_compound_type(context->structs, symbol); + } else { + compound_type = find_compound_type(context->unions, symbol); + } + } + } else if(token.type != '{') { + if(is_struct) { + parse_error_expected("problem while parsing struct type specifiers", + T_IDENTIFIER, '{', 0); + } else { + parse_error_expected("problem while parsing union type specifiers", + T_IDENTIFIER, '{', 0); } - } else if(token.type == '{') { - parse_compound_type_entries(); - } else { - parse_error_expected("problem while parsing struct type specifiers", - T_IDENTIFIER, '{', 0); - struct_type = NULL; - } - assert(context == &struct_type->context); - set_context(last_context); - environment_pop_to(top); + return NULL; + } - return (type_t*) struct_type; -} + if(compound_type == NULL) { + compound_type = allocate_type_zero(sizeof(compound_type[0])); -static type_t *parse_union_specifier(void) -{ - eat(T_union); + if(is_struct) { + compound_type->type.type = TYPE_COMPOUND_STRUCT; + } else { + compound_type->type.type = TYPE_COMPOUND_UNION; + } + compound_type->source_position = token.source_position; + compound_type->symbol = symbol; + } - compound_type_t *union_type = allocate_type_zero(sizeof(union_type[0])); - union_type->type.type = TYPE_COMPOUND_UNION; - union_type->source_position = token.source_position; + if(token.type == '{') { + if(compound_type->defined) { + parser_print_error_prefix(); + fprintf(stderr, "multiple definition of %s %s\n", + is_struct ? "struct" : "union", symbol->string); + compound_type->context.declarations = NULL; + } + compound_type->defined = 1; - int top = environment_top(); - context_t *last_context = context; - set_context(&union_type->context); + int top = environment_top(); + context_t *last_context = context; + set_context(&compound_type->context); - if(token.type == T_IDENTIFIER) { - union_type->symbol = token.v.symbol; - next_token(); - if(token.type == '{') { - parse_compound_type_entries(); - } - } else if(token.type == '{') { parse_compound_type_entries(); - } else { - parse_error_expected("problem while parsing union type specifiers", - T_IDENTIFIER, '{'); - union_type = NULL; - } - assert(context == &union_type->context); - set_context(last_context); - environment_pop_to(top); + assert(context == &compound_type->context); + set_context(last_context); + environment_pop_to(top); + } - return (type_t*) union_type; + return (type_t*) compound_type; } -static void parse_enum_type_entries(void) +static enum_entry_t *parse_enum_type_entries(void) { eat('{'); if(token.type == '}') { next_token(); parse_error("empty enum not allowed"); - return; + return NULL; } + enum_entry_t *result = NULL; + enum_entry_t *last_entry = NULL; do { + enum_entry_t *entry = allocate_ast_zero(sizeof(entry[0])); if(token.type != T_IDENTIFIER) { parse_error_expected("problem while parsing enum entry", T_IDENTIFIER, 0); - eat_until('}'); - return; + eat_block(); + return result; } + entry->symbol = token.v.symbol; next_token(); if(token.type == '=') { - parse_constant_expression(); + next_token(); + entry->value = parse_constant_expression(); + } + + if(last_entry != NULL) { + last_entry->next = entry; + } else { + result = entry; } + last_entry = entry; if(token.type != ',') break; next_token(); } while(token.type != '}'); - expect_void('}'); + expect('}'); + return result; } static type_t *parse_enum_specifier(void) @@ -394,14 +483,18 @@ static type_t *parse_enum_specifier(void) enum_type->type.type = TYPE_ENUM; enum_type->source_position = token.source_position; + /* TODO: rewrite to the same style as struct/union above to handle + * type identities correctly + */ + if(token.type == T_IDENTIFIER) { enum_type->symbol = token.v.symbol; next_token(); if(token.type == '{') { - parse_enum_type_entries(); + enum_type->entries = parse_enum_type_entries(); } } else if(token.type == '{') { - parse_enum_type_entries(); + enum_type->entries = parse_enum_type_entries(); } else { parse_error_expected("problem while parsing enum type specifiers", T_IDENTIFIER, '{'); @@ -410,6 +503,72 @@ static type_t *parse_enum_specifier(void) return (type_t*) enum_type; } +static +const char *parse_string_literals(void) +{ + assert(token.type == T_STRING_LITERAL); + const char *result = token.v.string; + + next_token(); + + while(token.type == T_STRING_LITERAL) { + result = concat_strings(result, token.v.string); + next_token(); + } + + return result; +} + +static +void parse_attributes(void) +{ + while(1) { + switch(token.type) { + case T___attribute__: + next_token(); + + expect_void('('); + int depth = 1; + while(depth > 0) { + switch(token.type) { + case T_EOF: + parse_error("EOF while parsing attribute"); + break; + case '(': + next_token(); + depth++; + break; + case ')': + next_token(); + depth--; + break; + default: + next_token(); + } + } + break; + case T_asm: + next_token(); + expect_void('('); + if(token.type != T_STRING_LITERAL) { + parse_error_expected("while parsing assembler attribute", + T_STRING_LITERAL); + eat_brace(); + break; + } else { + parse_string_literals(); + } + expect_void(')'); + break; + default: + goto attributes_finished; + } + } + +attributes_finished: + ; +} + typedef enum { SPECIFIER_SIGNED = 1 << 0, SPECIFIER_UNSIGNED = 1 << 1, @@ -580,10 +739,10 @@ void parse_declaration_specifiers(declaration_specifiers_t *specifiers) /* TODO: if type != NULL for the following rules issue an error */ case T_struct: - type = parse_struct_specifier(); + type = parse_compound_type_specifier(1); break; case T_union: - type = parse_union_specifier(); + type = parse_compound_type_specifier(0); break; case T_enum: type = parse_enum_specifier(); @@ -593,6 +752,11 @@ void parse_declaration_specifiers(declaration_specifiers_t *specifiers) next_token(); break; + case T___attribute__: + /* TODO */ + parse_attributes(); + break; + case T_IDENTIFIER: declaration = token.v.symbol->declaration; if(declaration == NULL || @@ -843,127 +1007,66 @@ declaration_t *parse_parameter(void) } static -void parse_parameters(method_type_t *type) +declaration_t *parse_parameters(method_type_t *type) { if(token.type == T_IDENTIFIER) { symbol_t *symbol = token.v.symbol; declaration_t *declaration = symbol->declaration; if(declaration == NULL || declaration->storage_class != STORAGE_CLASS_TYPEDEF) { + /* TODO */ parse_identifier_list(); - return; + return NULL; } } if(token.type == ')') { type->unspecified_parameters = 1; - return; + return NULL; } - if(token.type == T_void && la(1)->type == ')') { + if(token.type == T_void && look_ahead(1)->type == ')') { next_token(); - return; + return NULL; } - declaration_t *declaration; - method_parameter_type_t *parameter_type; - method_parameter_type_t *last_parameter_type = NULL; + declaration_t *declarations = NULL; + declaration_t *declaration; + declaration_t *last_declaration = NULL; + method_parameter_t *parameter; + method_parameter_t *last_parameter = NULL; while(1) { switch(token.type) { case T_DOTDOTDOT: next_token(); type->variadic = 1; - return; + return declarations; case T_IDENTIFIER: DECLARATION_START declaration = parse_parameter(); - parameter_type = allocate_type_zero(sizeof(parameter_type[0])); - parameter_type->type = declaration->type; - parameter_type->symbol = declaration->symbol; + parameter = allocate_type_zero(sizeof(parameter[0])); + parameter->type = declaration->type; - if(last_parameter_type != NULL) { - last_parameter_type->next = parameter_type; + if(last_parameter != NULL) { + last_declaration->next = declaration; + last_parameter->next = parameter; } else { - type->parameter_types = parameter_type; + type->parameters = parameter; + declarations = declaration; } - last_parameter_type = parameter_type; + last_parameter = parameter; + last_declaration = declaration; break; default: - return; + return declarations; } if(token.type != ',') - return; - next_token(); - } -} - -static -const char *parse_string_literals(void) -{ - assert(token.type == T_STRING_LITERAL); - const char *result = token.v.string; - - next_token(); - - while(token.type == T_STRING_LITERAL) { - result = concat_strings(result, token.v.string); + return declarations; next_token(); } - - return result; -} - -static -void parse_attributes(void) -{ - while(1) { - switch(token.type) { - case T___attribute__: - next_token(); - - expect_void('('); - int depth = 1; - while(depth > 0) { - switch(token.type) { - case T_EOF: - parse_error("EOF while parsing attribute"); - break; - case '(': - next_token(); - depth++; - break; - case ')': - next_token(); - depth--; - break; - default: - next_token(); - } - } - break; - case T_asm: - next_token(); - expect_void('('); - if(token.type != T_STRING_LITERAL) { - parse_error_expected("while parsing assembler attribute", - T_STRING_LITERAL); - eat_until(')'); - break; - } else { - parse_string_literals(); - } - expect_void(')'); - break; - default: - goto attributes_finished; - } - } - -attributes_finished: - ; } typedef struct declarator_part declarator_part; @@ -1017,7 +1120,10 @@ declarator_part *parse_inner_declarator(declaration_t *declaration, = allocate_type_zero(sizeof(method_type[0])); method_type->type.type = TYPE_METHOD; - parse_parameters(method_type); + declaration_t *parameters = parse_parameters(method_type); + if(declaration != NULL) { + declaration->context.declarations = parameters; + } part->method_type = method_type; @@ -1039,7 +1145,7 @@ declarator_part *parse_inner_declarator(declaration_t *declaration, /* TODO */ - if(token.type == '*' && la(1)->type == ']') { + if(token.type == '*' && look_ahead(1)->type == ']') { next_token(); } else if(token.type != ']') { parse_assignment_expression(); @@ -1068,7 +1174,11 @@ type_t *construct_declarator_type(declarator_part *part, type_t *type) if(method_type != NULL) { method_type->result_type = type; - type = (type_t*) method_type; + type_t *result = typehash_insert((type_t*) method_type); + if(result != (type_t*) method_type) { + obstack_free(type_obst, method_type); + } + type = result; } part = part->inner; @@ -1096,53 +1206,102 @@ type_t *parse_abstract_declarator(type_t *base_type) { declarator_part *part = parse_inner_declarator(NULL, 1); + if(part == NULL) + return NULL; + type_t *result = construct_declarator_type(part, base_type); obstack_free(&temp_obst, part); return result; } -static void record_declaration(declaration_t *declaration) +static declaration_t *record_declaration(declaration_t *declaration) { + if(context == NULL) + return declaration; + + symbol_t *symbol = declaration->symbol; + if(symbol != NULL) { + declaration_t *alias = environment_push(declaration, context); + if(alias != declaration) + return alias; + } + if(last_declaration != NULL) { last_declaration->next = declaration; } else { - if(context != NULL) - context->declarations = declaration; + context->declarations = declaration; } last_declaration = declaration; + + return declaration; } static -void maybe_push_declaration(declaration_t *declaration) +void parser_error_multiple_definition(declaration_t *previous, + declaration_t *declaration) { - symbol_t *symbol = declaration->symbol; - - if(symbol != NULL) { - environment_push(declaration, context); - } + parser_print_error_prefix_pos(declaration->source_position); + fprintf(stderr, "multiple definition of symbol '%s'\n", + declaration->symbol->string); + parser_print_error_prefix_pos(previous->source_position); + fprintf(stderr, "this is the location of the previous " + "definition.\n"); } static void parse_init_declarators(const declaration_specifiers_t *specifiers) { while(1) { - declaration_t *declaration = allocate_ast_zero(sizeof(declaration[0])); + declaration_t *ndeclaration + = allocate_ast_zero(sizeof(ndeclaration[0])); - parse_declarator(declaration, specifiers->storage_class, + parse_declarator(ndeclaration, specifiers->storage_class, specifiers->type, 0); - maybe_push_declaration(declaration); - record_declaration(declaration); + declaration_t *declaration = record_declaration(ndeclaration); if(token.type == '=') { next_token(); + + /* TODO: check that this is an allowed type (esp. no method type) */ + + if(declaration->initializer != NULL) { + parser_error_multiple_definition(declaration, ndeclaration); + } + if(token.type == '{') { // TODO expect_void('}'); } else { - parse_assignment_expression(); + declaration->initializer = parse_assignment_expression(); } } else if(token.type == '{') { + if(declaration->type->type != TYPE_METHOD) { + parser_print_error_prefix(); + fprintf(stderr, "Declarator "); + print_type(declaration->type, declaration->symbol); + fprintf(stderr, " is not a method type.\n"); + } + + if(declaration->initializer != NULL) { + parser_error_multiple_definition(declaration, ndeclaration); + } + + int top = environment_top(); + context_t *last_context = context; + set_context(&declaration->context); + + /* push function parameters */ + declaration_t *parameter = declaration->context.declarations; + for( ; parameter != NULL; parameter = parameter->next) { + environment_push(parameter, context); + } + statement_t *statement = parse_compound_statement(); + + assert(context == &declaration->context); + set_context(last_context); + environment_pop_to(top); + declaration->statement = statement; return; } @@ -1166,8 +1325,9 @@ void parse_struct_declarators(const declaration_specifiers_t *specifiers) declaration_t *declaration = allocate_ast_zero(sizeof(declaration[0])); parse_declarator(declaration, specifiers->storage_class, - specifiers->type, 0); - maybe_push_declaration(declaration); + specifiers->type, 1); + + /* TODO: check for doubled fields */ record_declaration(declaration); if(token.type == ':') { @@ -1294,19 +1454,58 @@ expression_t *parse_reference(void) { reference_expression_t *ref = allocate_ast_zero(sizeof(ref[0])); - ref->expression.type = EXPR_REFERENCE; - ref->symbol = token.v.symbol; + ref->expression.type = EXPR_REFERENCE; + ref->symbol = token.v.symbol; + + if(ref->symbol->declaration == NULL) { + parser_print_error_prefix(); + fprintf(stderr, "unknown symbol '%s' found.\n", ref->symbol->string); + } + ref->declaration = ref->symbol->declaration; next_token(); return (expression_t*) ref; } +static +expression_t *parse_cast(void) +{ + unary_expression_t *cast = allocate_ast_zero(sizeof(cast[0])); + + cast->expression.type = EXPR_UNARY; + cast->type = UNEXPR_CAST; + cast->expression.source_position = token.source_position; + + type_t *type = parse_typename(); + + expect(')'); + expression_t *value = parse_sub_expression(20); + + cast->expression.datatype = type; + cast->value = value; + + return (expression_t*) cast; +} + static expression_t *parse_brace_expression(void) { eat('('); + declaration_t *declaration; + switch(token.type) { + TYPE_QUALIFIERS + TYPE_SPECIFIERS + return parse_cast(); + case T_IDENTIFIER: + declaration = token.v.symbol->declaration; + if(declaration != NULL && + (declaration->storage_class & STORAGE_CLASS_TYPEDEF)) { + return parse_cast(); + } + } + expression_t *result = parse_expression(); expect(')'); @@ -1327,7 +1526,11 @@ expression_t *parse_primary_expression(void) return parse_brace_expression(); } - /* TODO: error message */ + parser_print_error_prefix(); + fprintf(stderr, "unexpected token "); + print_token(stderr, &token); + fprintf(stderr, "\n"); + eat_statement(); return NULL; } @@ -1363,6 +1566,26 @@ type_t *get_expression_type(const expression_t *expression) return NULL; } +static +int is_type_specifier(const token_t *token) +{ + declaration_t *declaration; + + switch(token->type) { + TYPE_SPECIFIERS + return 1; + case T_IDENTIFIER: + declaration = token->v.symbol->declaration; + if(declaration == NULL) + return 0; + if(declaration->storage_class != STORAGE_CLASS_TYPEDEF) + return 0; + return 1; + default: + return 0; + } +} + static expression_t *parse_sizeof(unsigned precedence) { @@ -1372,7 +1595,7 @@ expression_t *parse_sizeof(unsigned precedence) = allocate_ast_zero(sizeof(sizeof_expression[0])); sizeof_expression->expression.type = EXPR_SIZEOF; - if(token.type == '(' /* && LA1 is type_specifier */) { + if(token.type == '(' && is_type_specifier(look_ahead(1))) { next_token(); sizeof_expression->type = parse_typename(); expect(')'); @@ -1534,7 +1757,7 @@ CREATE_BINEXPR_PARSER('<', BINEXPR_LESS) CREATE_BINEXPR_PARSER('>', BINEXPR_GREATER) CREATE_BINEXPR_PARSER('=', BINEXPR_ASSIGN) CREATE_BINEXPR_PARSER(T_EQUALEQUAL, BINEXPR_EQUAL) -CREATE_BINEXPR_PARSER(T_SLASHEQUAL, BINEXPR_NOTEQUAL) +CREATE_BINEXPR_PARSER(T_EXCLAMATIONMARKEQUAL, BINEXPR_NOTEQUAL) CREATE_BINEXPR_PARSER(T_LESSEQUAL, BINEXPR_LESSEQUAL) CREATE_BINEXPR_PARSER(T_GREATEREQUAL, BINEXPR_GREATEREQUAL) CREATE_BINEXPR_PARSER('&', BINEXPR_BITWISE_AND) @@ -1708,17 +1931,21 @@ static statement_t *parse_if(void) { eat(T_if); + + if_statement_t *statement = allocate_ast_zero(sizeof(statement[0])); + statement->statement.type = STATEMENT_IF; + expect('('); - parse_expression(); + statement->condition = parse_expression(); expect(')'); - parse_statement(); + statement->true_statement = parse_statement(); if(token.type == T_else) { next_token(); - parse_statement(); + statement->false_statement = parse_statement(); } - return NULL; + return (statement_t*) statement; } static @@ -1899,11 +2126,12 @@ statement_t *parse_statement(void) break; case ';': + next_token(); statement = NULL; break; case T_IDENTIFIER: - if(la(1)->type == ':') { + if(look_ahead(1)->type == ':') { statement = parse_label_statement(); break; } @@ -1921,6 +2149,10 @@ statement_t *parse_statement(void) DECLARATION_START statement = parse_declaration_statement(); break; + + default: + statement = parse_expression_statement(); + break; } return statement; @@ -1985,6 +2217,8 @@ translation_unit_t *parse(void) obstack_init(&environment_obstack); environment_stack = NEW_ARR_F(environment_entry_t*, 0); + type_set_output(stderr); + lookahead_bufpos = 0; for(int i = 0; i < MAX_LOOKAHEAD + 2; ++i) { next_token();