X-Git-Url: http://nsz.repo.hu/git/?a=blobdiff_plain;f=parser.c;h=44b5dc669062fbc74a9eafde620363e11dee7293;hb=087130404a59dcdd95aa0d7ac7feec2793e9ce87;hp=c1f162b139ba1744e5ec02d6c11aa3da7ec82a5f;hpb=d5a5e90eca7d95d0efa66a6e8dd5ca0ed031321b;p=cparser diff --git a/parser.c b/parser.c index c1f162b..44b5dc6 100644 --- a/parser.c +++ b/parser.c @@ -17,23 +17,25 @@ //#define PRINT_TOKENS //#define ABORT_ON_ERROR #define MAX_LOOKAHEAD 2 +//#define STRICT_C99 -struct environment_entry_t { - symbol_t *symbol; +typedef struct { declaration_t *old_declaration; - const void *old_context; -}; + symbol_t *symbol; + unsigned short namespace; +} stack_entry_t; static token_t token; static token_t lookahead_buffer[MAX_LOOKAHEAD]; static int lookahead_bufpos; -static struct obstack environment_obstack; -static environment_entry_t **environment_stack = NULL; +static stack_entry_t *environment_stack = NULL; +static context_t *global_context = NULL; static context_t *context = NULL; static declaration_t *last_declaration = NULL; static struct obstack temp_obst; static type_t *type_int = NULL; +static type_t *type_double = NULL; static type_t *type_const_char = NULL; static type_t *type_string = NULL; static type_t *type_void = NULL; @@ -114,6 +116,11 @@ static inline void *allocate_type_zero(size_t size) return res; } +static inline void free_type(void *type) +{ + obstack_free(type_obst, type); +} + /** * returns the top element of the environment stack */ @@ -145,26 +152,34 @@ static inline const token_t *look_ahead(int num) return & lookahead_buffer[pos]; } -static inline void eat(token_type_t type) +#define eat(token_type) do { assert(token.type == token_type); next_token(); } while(0) + +void error(void) { - assert(token.type == type); - next_token(); +#ifdef ABORT_ON_ERROR + abort(); +#endif } -void parser_print_error_prefix_pos(const source_position_t source_position) +void parser_print_prefix_pos(const source_position_t source_position) { fputs(source_position.input_name, stderr); fputc(':', stderr); fprintf(stderr, "%d", source_position.linenr); - fputs(": error: ", stderr); -#ifdef ABORT_ON_ERROR - abort(); -#endif + fputs(": ", stderr); +} + +void parser_print_error_prefix_pos(const source_position_t source_position) +{ + parser_print_prefix_pos(source_position); + fputs("error: ", stderr); + error(); } void parser_print_error_prefix(void) { - parser_print_error_prefix_pos(token.source_position); + parser_print_prefix_pos(token.source_position); + error(); } static void parse_error(const char *message) @@ -173,6 +188,13 @@ static void parse_error(const char *message) fprintf(stderr, "parse error: %s\n", message); } +__attribute__((unused)) +static void parse_warning(const char *message) +{ + parser_print_prefix_pos(token.source_position); + fprintf(stderr, "warning: %s\n", message); +} + static void parse_error_expected(const char *message, ...) { va_list args; @@ -243,6 +265,13 @@ static void eat_brace(void) while(token.type != ')') { if(token.type == T_EOF) return; + if(token.type == ')' || token.type == ';' || token.type == '}') { + return; + } + if(token.type == '(') { + eat_brace(); + continue; + } if(token.type == '{') { eat_block(); continue; @@ -272,16 +301,12 @@ static void set_context(context_t *new_context) { context = new_context; - declaration_t *declaration = new_context->declarations; - if(declaration != NULL) { - while(true) { - if(declaration->next == NULL) - break; - declaration = declaration->next; + last_declaration = new_context->declarations; + if(last_declaration != NULL) { + while(last_declaration->context_next != NULL) { + last_declaration = last_declaration->context_next; } } - - last_declaration = declaration; } /** @@ -295,50 +320,93 @@ static bool is_compatible_declaration (declaration_t *declaration, return declaration->type == previous->type; } +static declaration_t *get_declaration(symbol_t *symbol, namespace_t namespace) +{ + declaration_t *declaration = symbol->declaration; + for( ; declaration != NULL; declaration = declaration->next) { + if(declaration->namespace == namespace) + return declaration; + } + + return NULL; +} + +static const char *get_namespace_prefix(namespace_t namespace) +{ + switch(namespace) { + case NAMESPACE_NORMAL: + return ""; + case NAMESPACE_UNION: + return "union "; + case NAMESPACE_STRUCT: + return "struct "; + case NAMESPACE_ENUM: + return "enum "; + } + panic("invalid namespace found"); +} + /** * pushs an environment_entry on the environment stack and links the * corresponding symbol to the new entry */ -static inline declaration_t *environment_push(declaration_t *declaration, - const void *context) +static declaration_t *environment_push(declaration_t *declaration) { - symbol_t *symbol = declaration->symbol; - assert(declaration != symbol->declaration); + symbol_t *symbol = declaration->symbol; + namespace_t namespace = declaration->namespace; assert(declaration->source_position.input_name != NULL); - if(symbol->context == context) { - declaration_t *previous_declaration = symbol->declaration; - if(symbol->declaration != NULL) { - 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); - 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); - fputc('\n', stderr); - } - return previous_declaration; + /* a declaration should be only pushed once */ + assert(declaration->parent_context == NULL); + declaration->parent_context = context; + + declaration_t *previous_declaration = get_declaration(symbol, namespace); + assert(declaration != previous_declaration); + if(previous_declaration != NULL + && previous_declaration->parent_context == context) { + if(!is_compatible_declaration(declaration, previous_declaration)) { + parser_print_error_prefix_pos(declaration->source_position); + fprintf(stderr, "definition of symbol %s%s with type ", + get_namespace_prefix(namespace), symbol->string); + error(); + print_type(declaration->type); + 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); + fputc('\n', stderr); } + return previous_declaration; } - environment_entry_t *entry - = obstack_alloc(&environment_obstack, sizeof(entry[0])); - memset(entry, 0, sizeof(entry[0])); + /* remember old declaration */ + stack_entry_t entry; + entry.symbol = symbol; + entry.old_declaration = symbol->declaration; + entry.namespace = namespace; + ARR_APP1(environment_stack, entry); - int top = ARR_LEN(environment_stack); - ARR_RESIZE(environment_stack, top + 1); - environment_stack[top] = entry; - - entry->old_declaration = symbol->declaration; - entry->old_context = symbol->context; - entry->symbol = symbol; - symbol->declaration = declaration; - symbol->context = context; + /* replace/add declaration into declaration list of the symbol */ + if(symbol->declaration == NULL) { + symbol->declaration = declaration; + } else { + declaration_t *iter = symbol->declaration; + for( ; iter != NULL; iter = iter->next) { + declaration_t *next = iter->next; + if(next == NULL) { + iter->next = declaration; + assert(declaration->next == NULL); + break; + } + if(next->namespace == namespace) { + iter->next = declaration; + declaration->next = next->next; + break; + } + } + } return declaration; } @@ -346,28 +414,44 @@ static inline declaration_t *environment_push(declaration_t *declaration, /** * pops symbols from the environment stack until @p new_top is the top element */ -static inline void environment_pop_to(size_t new_top) +static void environment_pop_to(size_t new_top) { - environment_entry_t *entry = NULL; size_t top = ARR_LEN(environment_stack); size_t i; + assert(new_top <= top); if(new_top == top) return; - assert(new_top < top); - i = top; - do { - entry = environment_stack[i - 1]; - - symbol_t *symbol = entry->symbol; + for(i = top; i > new_top; --i) { + stack_entry_t *entry = & environment_stack[i - 1]; - symbol->declaration = entry->old_declaration; - symbol->context = entry->old_context; + declaration_t *old_declaration = entry->old_declaration; + symbol_t *symbol = entry->symbol; + namespace_t namespace = entry->namespace; - --i; - } while(i != new_top); - obstack_free(&environment_obstack, entry); + /* replace/remove declaration */ + declaration_t *declaration = symbol->declaration; + assert(declaration != NULL); + if(declaration->namespace == namespace) { + if(old_declaration == NULL) { + symbol->declaration = declaration->next; + } else { + symbol->declaration = old_declaration; + assert(old_declaration->next == declaration->next); + } + } else { + for( ; declaration != NULL; declaration = declaration->next) { + declaration_t *next = declaration->next; + if(next->namespace == namespace) { + declaration->next = old_declaration; + assert(old_declaration->next == next->next); + break; + } + } + assert(declaration != NULL); + } + } ARR_SHRINKLEN(environment_stack, (int) new_top); } @@ -387,9 +471,8 @@ static expression_t *parse_assignment_expression(void) } 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 declaration_t *parse_declarator(storage_class_t storage_class, + type_t *type, int may_be_abstract); static declaration_t *record_declaration(declaration_t *declaration); typedef struct declaration_specifiers_t declaration_specifiers_t; @@ -398,19 +481,170 @@ struct declaration_specifiers_t { type_t *type; }; -static compound_type_t *find_compound_type(compound_type_t *types, - const symbol_t *symbol) +static const char *parse_string_literals(void) { - compound_type_t *type = types; - for( ; type != NULL; type = type->next) { - if(type->symbol == symbol) - return type; + 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 NULL; + return result; } -static type_t *parse_compound_type_specifier(bool is_struct) +static void parse_attributes(void) +{ + while(true) { + 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: + ; +} + +static designator_t *parse_designation(void) +{ + if(token.type != '[' && token.type != '.') + return NULL; + + designator_t *result = NULL; + designator_t *last = NULL; + + while(1) { + designator_t *designator; + switch(token.type) { + case '[': + designator = allocate_ast_zero(sizeof(designator[0])); + next_token(); + designator->array_access = parse_constant_expression(); + expect(']'); + break; + case '.': + designator = allocate_ast_zero(sizeof(designator[0])); + next_token(); + if(token.type != T_IDENTIFIER) { + parse_error_expected("problem while parsing designator", + T_IDENTIFIER, 0); + return NULL; + } + designator->symbol = token.v.symbol; + next_token(); + break; + default: + expect('='); + return result; + } + + assert(designator != NULL); + if(last != NULL) { + last->next = designator; + } else { + result = designator; + } + last = designator; + } +} + +static initializer_t *parse_initializer_list(void); + +static initializer_t *parse_initializer(void) +{ + designator_t *designator = parse_designation(); + + initializer_t *result; + if(token.type == '{') { + result = parse_initializer_list(); + } else { + result = allocate_ast_zero(sizeof(result[0])); + result->type = INITIALIZER_VALUE; + result->v.value = parse_assignment_expression(); + } + result->designator = designator; + + return result; +} + +static initializer_t *parse_initializer_list(void) +{ + eat('{'); + + initializer_t *result = allocate_ast_zero(sizeof(result[0])); + result->type = INITIALIZER_LIST; + + initializer_t *last = NULL; + while(1) { + initializer_t *initializer = parse_initializer(); + if(last != NULL) { + last->next = initializer; + } else { + result->v.list = initializer; + } + last = initializer; + + if(token.type == '}') + break; + + if(token.type != ',') { + parse_error_expected("problem while parsing initializer list", + ',', '}', 0); + eat_block(); + return result; + } + eat(','); + + if(token.type == '}') + break; + } + + expect('}'); + + return result; +} + +static declaration_t *parse_compound_type_specifier(bool is_struct) { if(is_struct) { eat(T_struct); @@ -418,19 +652,17 @@ static type_t *parse_compound_type_specifier(bool is_struct) eat(T_union); } - symbol_t *symbol = NULL; - compound_type_t *compound_type = NULL; + symbol_t *symbol = NULL; + declaration_t *declaration = NULL; if(token.type == T_IDENTIFIER) { symbol = token.v.symbol; next_token(); - if(context != NULL) { - if(is_struct) { - compound_type = find_compound_type(context->structs, symbol); - } else { - compound_type = find_compound_type(context->unions, symbol); - } + if(is_struct) { + declaration = get_declaration(symbol, NAMESPACE_STRUCT); + } else { + declaration = get_declaration(symbol, NAMESPACE_UNION); } } else if(token.type != '{') { if(is_struct) { @@ -444,39 +676,42 @@ static type_t *parse_compound_type_specifier(bool is_struct) return NULL; } - if(compound_type == NULL) { - compound_type = allocate_type_zero(sizeof(compound_type[0])); + if(declaration == NULL) { + declaration = allocate_type_zero(sizeof(declaration[0])); if(is_struct) { - compound_type->type.type = TYPE_COMPOUND_STRUCT; + declaration->namespace = NAMESPACE_STRUCT; } else { - compound_type->type.type = TYPE_COMPOUND_UNION; + declaration->namespace = NAMESPACE_UNION; } - compound_type->source_position = token.source_position; - compound_type->symbol = symbol; + declaration->source_position = token.source_position; + declaration->symbol = symbol; } if(token.type == '{') { - if(compound_type->defined) { + if(declaration->init.is_defined) { + assert(symbol != NULL); parser_print_error_prefix(); fprintf(stderr, "multiple definition of %s %s\n", is_struct ? "struct" : "union", symbol->string); - compound_type->context.declarations = NULL; + declaration->context.declarations = NULL; } - compound_type->defined = 1; + record_declaration(declaration); + declaration->init.is_defined = true; int top = environment_top(); context_t *last_context = context; - set_context(&compound_type->context); + set_context(& declaration->context); parse_compound_type_entries(); + parse_attributes(); - assert(context == &compound_type->context); + assert(context == & declaration->context); set_context(last_context); environment_pop_to(top); } - return (type_t*) compound_type; + return declaration; } static void parse_enum_entries(void) @@ -505,7 +740,7 @@ static void parse_enum_entries(void) if(token.type == '=') { next_token(); - entry->initializer = parse_constant_expression(); + entry->init.initializer = parse_initializer(); } record_declaration(entry); @@ -518,79 +753,73 @@ static void parse_enum_entries(void) expect_void('}'); } -static enum_type_t *find_enum_type(enum_type_t *types, const symbol_t *symbol) -{ - enum_type_t *type = types; - for( ; type != NULL; type = type->next) { - if(type->symbol == symbol) - return type; - } - - return NULL; -} - -static type_t *parse_enum_specifier(void) +static declaration_t *parse_enum_specifier(void) { eat(T_enum); - symbol_t *symbol = NULL; - enum_type_t *enum_type = NULL; + declaration_t *declaration; + symbol_t *symbol; if(token.type == T_IDENTIFIER) { symbol = token.v.symbol; next_token(); - if(context != NULL) { - enum_type = find_enum_type(context->enums, symbol); - } + declaration = get_declaration(symbol, NAMESPACE_ENUM); } else if(token.type != '{') { parse_error_expected("problem while parsing enum type specifier", T_IDENTIFIER, '{', 0); return NULL; + } else { + declaration = NULL; + symbol = NULL; } - if(enum_type == NULL) { - enum_type = allocate_type_zero(sizeof(enum_type[0])); - enum_type->type.type = TYPE_ENUM; - enum_type->source_position = token.source_position; - enum_type->symbol = symbol; + if(declaration == NULL) { + declaration = allocate_type_zero(sizeof(declaration[0])); + + declaration->namespace = NAMESPACE_ENUM; + declaration->source_position = token.source_position; + declaration->symbol = symbol; } if(token.type == '{') { - if(enum_type->defined) { + if(declaration->init.is_defined) { parser_print_error_prefix(); fprintf(stderr, "multiple definitions of enum %s\n", symbol->string); - enum_type->entries_begin = NULL; - enum_type->entries_end = NULL; } - enum_type->defined = 1; - - declaration_t *before = last_declaration; + record_declaration(declaration); + declaration->init.is_defined = 1; parse_enum_entries(); - - if(before == NULL) { - enum_type->entries_begin = context->declarations; - } else { - enum_type->entries_begin = before->next; - } - enum_type->entries_end = last_declaration; + parse_attributes(); } - return (type_t*) enum_type; + return declaration; +} + +/** + * if a symbol is a typedef to another type, return true + */ +static bool is_typedef_symbol(symbol_t *symbol) +{ + declaration_t *declaration = get_declaration(symbol, NAMESPACE_NORMAL); + if(declaration == NULL + || declaration->storage_class != STORAGE_CLASS_TYPEDEF) + return false; + + return true; } static type_t *parse_typeof(void) { eat(T___typeof__); - type_t *result; + type_t *type; expect('('); - declaration_t *declaration; - expression_t *expression; + expression_t *expression = NULL; restart: switch(token.type) { @@ -603,93 +832,32 @@ restart: goto restart; case T_IDENTIFIER: - declaration = token.v.symbol->declaration; - if(declaration != NULL - && declaration->storage_class == STORAGE_CLASS_TYPEDEF) { - result = parse_typename(); - break; + if(is_typedef_symbol(token.v.symbol)) { + type = parse_typename(); + } else { + expression = parse_expression(); + type = expression->datatype; } - expression = parse_expression(); - result = expression->datatype; break; TYPENAME_START - result = parse_typename(); + type = parse_typename(); break; default: expression = parse_expression(); - result = expression->datatype; + type = expression->datatype; break; } expect(')'); - return result; -} - -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(true) { - 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; - } - } + typeof_type_t *typeof = allocate_type_zero(sizeof(typeof[0])); + typeof->type.type = TYPE_TYPEOF; + typeof->expression = expression; + typeof->typeof_type = type; -attributes_finished: - ; + return (type_t*) typeof; } typedef enum { @@ -718,17 +886,25 @@ static type_t *create_builtin_type(symbol_t *symbol) type->type.type = TYPE_BUILTIN; type->symbol = symbol; - type_t *result = typehash_insert((type_t*) type); - if(result != (type_t*) type) { - obstack_free(type_obst, type); - } + return (type_t*) type; +} - return result; +static type_t *get_typedef_type(symbol_t *symbol) +{ + declaration_t *declaration = get_declaration(symbol, NAMESPACE_NORMAL); + if(declaration == NULL + || declaration->storage_class != STORAGE_CLASS_TYPEDEF) + return NULL; + + typedef_type_t *typedef_type = allocate_type_zero(sizeof(typedef_type[0])); + typedef_type->type.type = TYPE_TYPEDEF; + typedef_type->declaration = declaration; + + return (type_t*) typedef_type; } static void parse_declaration_specifiers(declaration_specifiers_t *specifiers) { - declaration_t *declaration; type_t *type = NULL; unsigned type_qualifiers = 0; unsigned type_specifiers = 0; @@ -808,16 +984,34 @@ static void parse_declaration_specifiers(declaration_specifiers_t *specifiers) } break; - /* TODO: if type != NULL for the following rules issue an error */ - case T_struct: - type = parse_compound_type_specifier(true); + /* TODO: if type != NULL for the following rules should issue + * an error */ + case T_struct: { + compound_type_t *compound_type + = allocate_type_zero(sizeof(compound_type[0])); + compound_type->type.type = TYPE_COMPOUND_STRUCT; + compound_type->declaration = parse_compound_type_specifier(true); + + type = (type_t*) compound_type; break; - case T_union: - type = parse_compound_type_specifier(false); + } + case T_union: { + compound_type_t *compound_type + = allocate_type_zero(sizeof(compound_type[0])); + compound_type->type.type = TYPE_COMPOUND_UNION; + compound_type->declaration = parse_compound_type_specifier(false); + + type = (type_t*) compound_type; break; - case T_enum: - type = parse_enum_specifier(); + } + case T_enum: { + enum_type_t *enum_type = allocate_type_zero(sizeof(enum_type[0])); + enum_type->type.type = TYPE_ENUM; + enum_type->declaration = parse_enum_specifier(); + + type = (type_t*) enum_type; break; + } case T___typeof__: type = parse_typeof(); break; @@ -831,17 +1025,16 @@ static void parse_declaration_specifiers(declaration_specifiers_t *specifiers) parse_attributes(); break; - case T_IDENTIFIER: - declaration = token.v.symbol->declaration; - if(declaration == NULL || - declaration->storage_class != STORAGE_CLASS_TYPEDEF) { + case T_IDENTIFIER: { + type_t *typedef_type = get_typedef_type(token.v.symbol); + + if(typedef_type == NULL) goto finish_specifiers; - } - type = declaration->type; - assert(type != NULL); next_token(); + type = typedef_type; break; + } /* function specifier */ default: @@ -946,7 +1139,13 @@ finish_specifiers: default: /* invalid specifier combination, give an error message */ if(type_specifiers == 0) { +#ifndef STRICT_C99 + parse_warning("no type specifiers in declaration (using int)"); + atomic_type = ATOMIC_TYPE_INT; + break; +#else parse_error("no type specifiers given in declaration"); +#endif } else if((type_specifiers & SPECIFIER_SIGNED) && (type_specifiers & SPECIFIER_UNSIGNED)) { parse_error("signed and unsigned specifiers gives"); @@ -974,7 +1173,7 @@ finish_specifiers: type_t *result = typehash_insert(type); if(newtype && result != (type_t*) type) { - obstack_free(type_obst, type); + free_type(type); } specifiers->type = result; @@ -1020,20 +1219,22 @@ static declaration_t *parse_parameter(void) parse_declaration_specifiers(&specifiers); - declaration_t *declaration = allocate_ast_zero(sizeof(declaration[0])); - parse_declarator(declaration, specifiers.storage_class, - specifiers.type, 1); + declaration_t *declaration = parse_declarator(specifiers.storage_class, + specifiers.type, 1); + + /* TODO check declaration constraints for parameters */ + if(declaration->storage_class == STORAGE_CLASS_TYPEDEF) { + parse_error("typedef not allowed in parameter list"); + } return declaration; } -static declaration_t *parse_parameters(method_type_t *type) +static declaration_t *parse_parameters(function_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) { + symbol_t *symbol = token.v.symbol; + if(!is_typedef_symbol(symbol)) { /* TODO */ parse_identifier_list(); return NULL; @@ -1049,11 +1250,11 @@ static declaration_t *parse_parameters(method_type_t *type) return NULL; } - declaration_t *declarations = NULL; - declaration_t *declaration; - declaration_t *last_declaration = NULL; - method_parameter_t *parameter; - method_parameter_t *last_parameter = NULL; + declaration_t *declarations = NULL; + declaration_t *declaration; + declaration_t *last_declaration = NULL; + function_parameter_t *parameter; + function_parameter_t *last_parameter = NULL; while(true) { switch(token.type) { @@ -1071,8 +1272,8 @@ static declaration_t *parse_parameters(method_type_t *type) parameter->type = declaration->type; if(last_parameter != NULL) { - last_declaration->next = declaration; - last_parameter->next = parameter; + last_declaration->context_next = declaration; + last_parameter->next = parameter; } else { type->parameters = parameter; declarations = declaration; @@ -1092,7 +1293,7 @@ static declaration_t *parse_parameters(method_type_t *type) typedef enum { CONSTRUCT_POINTER, - CONSTRUCT_METHOD, + CONSTRUCT_FUNCTION, CONSTRUCT_ARRAY } construct_type_type_t; @@ -1108,10 +1309,10 @@ struct parsed_pointer_t { type_qualifier_t type_qualifiers; }; -typedef struct construct_method_type_t construct_method_type_t; -struct construct_method_type_t { - construct_type_t construct_type; - method_type_t *method_type; +typedef struct construct_function_type_t construct_function_type_t; +struct construct_function_type_t { + construct_type_t construct_type; + function_type_t *function_type; }; typedef struct parsed_array_t parsed_array_t; @@ -1173,28 +1374,27 @@ static construct_type_t *parse_array_declarator(void) return (construct_type_t*) array; } -static construct_type_t *parse_method_declarator(declaration_t *declaration) +static construct_type_t *parse_function_declarator(declaration_t *declaration) { eat('('); - method_type_t *method_type - = allocate_type_zero(sizeof(method_type[0])); - method_type->type.type = TYPE_METHOD; + function_type_t *type = allocate_type_zero(sizeof(type[0])); + type->type.type = TYPE_FUNCTION; - declaration_t *parameters = parse_parameters(method_type); + declaration_t *parameters = parse_parameters(type); if(declaration != NULL) { declaration->context.declarations = parameters; } - construct_method_type_t *construct_method_type = - obstack_alloc(&temp_obst, sizeof(construct_method_type[0])); - memset(construct_method_type, 0, sizeof(construct_method_type[0])); - construct_method_type->construct_type.type = CONSTRUCT_METHOD; - construct_method_type->method_type = method_type; + construct_function_type_t *construct_function_type = + obstack_alloc(&temp_obst, sizeof(construct_function_type[0])); + memset(construct_function_type, 0, sizeof(construct_function_type[0])); + construct_function_type->construct_type.type = CONSTRUCT_FUNCTION; + construct_function_type->function_type = type; expect(')'); - return (construct_type_t*) construct_method_type; + return (construct_type_t*) construct_function_type; } static construct_type_t *parse_inner_declarator(declaration_t *declaration, @@ -1244,7 +1444,7 @@ static construct_type_t *parse_inner_declarator(declaration_t *declaration, construct_type_t *type; switch(token.type) { case '(': - type = parse_method_declarator(declaration); + type = parse_function_declarator(declaration); break; case '[': type = parse_array_declarator(); @@ -1276,47 +1476,25 @@ declarator_finished: return result; } -#if 0 -static type_t *make_pointers(type_t *type, parsed_pointer_t *pointer) -{ - for( ; pointer != NULL; pointer = pointer->next) { - pointer_type_t *pointer_type - = allocate_type_zero(sizeof(pointer_type[0])); - pointer_type->type.type = TYPE_POINTER; - pointer_type->points_to = type; - pointer_type->type.qualifiers = pointer->type_qualifiers; - - type_t *result = typehash_insert((type_t*) pointer_type); - if(result != (type_t*) pointer_type) { - obstack_free(type_obst, pointer_type); - } - - type = result; - } - - return type; -} -#endif - static type_t *construct_declarator_type(construct_type_t *construct_list, type_t *type) { construct_type_t *iter = construct_list; for( ; iter != NULL; iter = iter->next) { - parsed_pointer_t *parsed_pointer; - parsed_array_t *parsed_array; - construct_method_type_t *construct_method_type; - method_type_t *method_type; - pointer_type_t *pointer_type; - array_type_t *array_type; + parsed_pointer_t *parsed_pointer; + parsed_array_t *parsed_array; + construct_function_type_t *construct_function_type; + function_type_t *function_type; + pointer_type_t *pointer_type; + array_type_t *array_type; switch(iter->type) { - case CONSTRUCT_METHOD: - construct_method_type = (construct_method_type_t*) iter; - method_type = construct_method_type->method_type; + case CONSTRUCT_FUNCTION: + construct_function_type = (construct_function_type_t*) iter; + function_type = construct_function_type->function_type; - method_type->result_type = type; - type = (type_t*) method_type; + function_type->result_type = type; + type = (type_t*) function_type; break; case CONSTRUCT_POINTER: @@ -1345,7 +1523,7 @@ static type_t *construct_declarator_type(construct_type_t *construct_list, type_t *hashed_type = typehash_insert((type_t*) type); if(hashed_type != type) { - obstack_free(type_obst, type); + free_type(type); type = hashed_type; } } @@ -1353,50 +1531,52 @@ static type_t *construct_declarator_type(construct_type_t *construct_list, return type; } -static void parse_declarator(declaration_t *declaration, - storage_class_t storage_class, type_t *type, - int may_be_abstract) +static declaration_t *parse_declarator(storage_class_t storage_class, + type_t *type, int may_be_abstract) { + declaration_t *declaration = allocate_ast_zero(sizeof(declaration[0])); + declaration->storage_class = storage_class; + construct_type_t *construct_type = parse_inner_declarator(declaration, may_be_abstract); + declaration->type = construct_declarator_type(construct_type, type); - declaration->type = construct_declarator_type(construct_type, type); - declaration->storage_class = storage_class; if(construct_type != NULL) { obstack_free(&temp_obst, construct_type); } + + return declaration; } static type_t *parse_abstract_declarator(type_t *base_type) { - construct_type_t *construct_type - = parse_inner_declarator(NULL, 1); - - if(construct_type == NULL) - return NULL; + construct_type_t *construct_type = parse_inner_declarator(NULL, 1); type_t *result = construct_declarator_type(construct_type, base_type); - obstack_free(&temp_obst, construct_type); + if(construct_type != NULL) { + obstack_free(&temp_obst, construct_type); + } return result; } static declaration_t *record_declaration(declaration_t *declaration) { - if(context == NULL) - return declaration; + assert(context != NULL); symbol_t *symbol = declaration->symbol; if(symbol != NULL) { - declaration_t *alias = environment_push(declaration, context); + declaration_t *alias = environment_push(declaration); if(alias != declaration) return alias; + } else { + declaration->parent_context = context; } if(last_declaration != NULL) { - last_declaration->next = declaration; + last_declaration->context_next = declaration; } else { - context->declarations = declaration; + context->declarations = declaration; } last_declaration = declaration; @@ -1412,41 +1592,37 @@ static void parser_error_multiple_definition(declaration_t *previous, parser_print_error_prefix_pos(previous->source_position); fprintf(stderr, "this is the location of the previous " "definition.\n"); + error(); } static void parse_init_declarators(const declaration_specifiers_t *specifiers) { while(true) { declaration_t *ndeclaration - = allocate_ast_zero(sizeof(ndeclaration[0])); + = parse_declarator(specifiers->storage_class, specifiers->type, 0); - parse_declarator(ndeclaration, specifiers->storage_class, - specifiers->type, 0); declaration_t *declaration = record_declaration(ndeclaration); if(token.type == '=') { next_token(); - /* TODO: check that this is an allowed type (esp. no method type) */ + /* TODO: check that this is an allowed type (no function type) */ - if(declaration->initializer != NULL) { + if(declaration->init.initializer != NULL) { parser_error_multiple_definition(declaration, ndeclaration); } - if(token.type == '{') { - // TODO - expect_void('}'); - } else { - declaration->initializer = parse_assignment_expression(); - } + ndeclaration->init.initializer = parse_initializer(); } else if(token.type == '{') { - if(declaration->type->type != TYPE_METHOD) { + if(declaration->type->type != TYPE_FUNCTION) { parser_print_error_prefix(); fprintf(stderr, "Declarator "); print_type_ext(declaration->type, declaration->symbol, NULL); - fprintf(stderr, " is not a method type.\n"); + fprintf(stderr, " has a body but is not a function type.\n"); + eat_block(); + continue; } - if(declaration->initializer != NULL) { + if(declaration->init.statement != NULL) { parser_error_multiple_definition(declaration, ndeclaration); } if(ndeclaration != declaration) { @@ -1460,8 +1636,8 @@ static void parse_init_declarators(const declaration_specifiers_t *specifiers) /* push function parameters */ declaration_t *parameter = declaration->context.declarations; - for( ; parameter != NULL; parameter = parameter->next) { - environment_push(parameter, context); + for( ; parameter != NULL; parameter = parameter->context_next) { + environment_push(parameter); } statement_t *statement = parse_compound_statement(); @@ -1470,7 +1646,7 @@ static void parse_init_declarators(const declaration_specifiers_t *specifiers) set_context(last_context); environment_pop_to(top); - declaration->statement = statement; + declaration->init.statement = statement; return; } @@ -1490,10 +1666,10 @@ static void parse_struct_declarators(const declaration_specifiers_t *specifiers) /* TODO (bitfields) */ } else { declaration_t *declaration - = allocate_ast_zero(sizeof(declaration[0])); - parse_declarator(declaration, specifiers->storage_class, - specifiers->type, 1); + = parse_declarator(specifiers->storage_class, + specifiers->type, 1); + /* TODO: check constraints for struct declarations */ /* TODO: check for doubled fields */ record_declaration(declaration); @@ -1518,8 +1694,6 @@ static void parse_compound_type_entries(void) while(token.type != '}' && token.type != T_EOF) { declaration_specifiers_t specifiers; memset(&specifiers, 0, sizeof(specifiers)); - /* TODO not correct as this allows storage class stuff... but only - * specifiers and qualifiers sould be allowed here */ parse_declaration_specifiers(&specifiers); parse_struct_declarators(&specifiers); @@ -1532,12 +1706,21 @@ static void parse_compound_type_entries(void) static void parse_declaration(void) { + source_position_t source_position = token.source_position; + declaration_specifiers_t specifiers; memset(&specifiers, 0, sizeof(specifiers)); parse_declaration_specifiers(&specifiers); if(token.type == ';') { next_token(); + + declaration_t *declaration = allocate_ast_zero(sizeof(declaration[0])); + + declaration->type = specifiers.type; + declaration->storage_class = specifiers.storage_class; + declaration->source_position = source_position; + record_declaration(declaration); return; } parse_init_declarators(&specifiers); @@ -1608,13 +1791,64 @@ static expression_t *parse_int_const(void) cnst->expression.type = EXPR_CONST; cnst->expression.datatype = type_int; - cnst->value = token.v.intvalue; + cnst->v.int_value = token.v.intvalue; next_token(); return (expression_t*) cnst; } +static expression_t *parse_float_const(void) +{ + const_t *cnst = allocate_ast_zero(sizeof(cnst[0])); + + cnst->expression.type = EXPR_CONST; + cnst->expression.datatype = type_double; + cnst->v.float_value = token.v.floatvalue; + + next_token(); + + return (expression_t*) cnst; +} + +static declaration_t *create_implicit_function(symbol_t *symbol, + const source_position_t source_position) +{ + function_type_t *function_type = allocate_type_zero(sizeof(function_type)); + + function_type->type.type = TYPE_FUNCTION; + function_type->result_type = type_int; + function_type->unspecified_parameters = true; + + type_t *type = typehash_insert((type_t*) function_type); + if(type != (type_t*) function_type) { + free_type(function_type); + } + + declaration_t *declaration = allocate_ast_zero(sizeof(declaration[0])); + + declaration->storage_class = STORAGE_CLASS_EXTERN; + declaration->type = type; + declaration->symbol = symbol; + declaration->source_position = source_position; + + /* prepend the implicit definition to the global context + * this is safe since the symbol wasn't declared as anything else yet + */ + assert(symbol->declaration == NULL); + + context_t *last_context = context; + context = global_context; + + environment_push(declaration); + declaration->context_next = context->declarations; + context->declarations = declaration; + + context = last_context; + + return declaration; +} + static expression_t *parse_reference(void) { reference_expression_t *ref = allocate_ast_zero(sizeof(ref[0])); @@ -1622,15 +1856,33 @@ static expression_t *parse_reference(void) 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; - ref->expression.datatype = ref->declaration->type; + declaration_t *declaration = get_declaration(ref->symbol, NAMESPACE_NORMAL); + source_position_t source_position = token.source_position; next_token(); + if(declaration == NULL) { +#ifndef STRICT_C99 + /* an implicitely defined function */ + if(token.type == '(') { + parser_print_prefix_pos(token.source_position); + fprintf(stderr, "warning: implicit declaration of function '%s'\n", + ref->symbol->string); + + declaration = create_implicit_function(ref->symbol, + source_position); + } else +#endif + { + parser_print_error_prefix(); + fprintf(stderr, "unknown symbol '%s' found.\n", ref->symbol->string); + return (expression_t*) ref; + } + } + + ref->declaration = declaration; + ref->expression.datatype = declaration->type; + return (expression_t*) ref; } @@ -1694,7 +1946,6 @@ static expression_t *parse_brace_expression(void) { eat('('); - declaration_t *declaration; switch(token.type) { case '{': /* gcc extension: a stement expression */ @@ -1704,9 +1955,7 @@ static expression_t *parse_brace_expression(void) TYPE_SPECIFIERS return parse_cast(); case T_IDENTIFIER: - declaration = token.v.symbol->declaration; - if(declaration != NULL && - (declaration->storage_class == STORAGE_CLASS_TYPEDEF)) { + if(is_typedef_symbol(token.v.symbol)) { return parse_cast(); } } @@ -1743,9 +1992,9 @@ static expression_t *parse_pretty_function_keyword(void) return (expression_t*) expression; } -static member_designator_t *parse_member_designators(void) +static designator_t *parse_designator(void) { - member_designator_t *result = allocate_ast_zero(sizeof(result[0])); + designator_t *result = allocate_ast_zero(sizeof(result[0])); if(token.type != T_IDENTIFIER) { parse_error_expected("problem while parsing member designator", @@ -1756,7 +2005,7 @@ static member_designator_t *parse_member_designators(void) result->symbol = token.v.symbol; next_token(); - member_designator_t *last_designator = result; + designator_t *last_designator = result; while(true) { if(token.type == '.') { next_token(); @@ -1766,9 +2015,8 @@ static member_designator_t *parse_member_designators(void) eat_brace(); return NULL; } - member_designator_t *designator - = allocate_ast_zero(sizeof(result[0])); - designator->symbol = token.v.symbol; + designator_t *designator = allocate_ast_zero(sizeof(result[0])); + designator->symbol = token.v.symbol; next_token(); last_designator->next = designator; @@ -1777,8 +2025,7 @@ static member_designator_t *parse_member_designators(void) } if(token.type == '[') { next_token(); - member_designator_t *designator - = allocate_ast_zero(sizeof(result[0])); + designator_t *designator = allocate_ast_zero(sizeof(result[0])); designator->array_access = parse_expression(); if(designator->array_access == NULL) { eat_brace(); @@ -1808,7 +2055,23 @@ static expression_t *parse_offsetof(void) expect('('); expression->type = parse_typename(); expect(','); - expression->member_designators = parse_member_designators(); + expression->designator = parse_designator(); + expect(')'); + + return (expression_t*) expression; +} + +static expression_t *parse_va_arg(void) +{ + eat(T___builtin_va_arg); + + va_arg_expression_t *expression = allocate_ast_zero(sizeof(expression[0])); + expression->expression.type = EXPR_VA_ARG; + + expect('('); + expression->arg = parse_assignment_expression(); + expect(','); + expression->expression.datatype = parse_typename(); expect(')'); return (expression_t*) expression; @@ -1834,6 +2097,8 @@ static expression_t *parse_primary_expression(void) switch(token.type) { case T_INTEGER: return parse_int_const(); + case T_FLOATINGPOINT: + return parse_float_const(); case T_STRING_LITERAL: return parse_string_const(); case T_IDENTIFIER: @@ -1844,9 +2109,10 @@ static expression_t *parse_primary_expression(void) return parse_pretty_function_keyword(); case T___builtin_offsetof: return parse_offsetof(); + case T___builtin_va_arg: + return parse_va_arg(); case T___builtin_expect: case T___builtin_va_start: - case T___builtin_va_arg: case T___builtin_va_end: return parse_builtin_symbol(); @@ -1859,7 +2125,12 @@ static expression_t *parse_primary_expression(void) print_token(stderr, &token); fprintf(stderr, "\n"); eat_statement(); - return NULL; + + expression_t *expression = allocate_ast_zero(sizeof(expression[0])); + expression->type = EXPR_INVALID; + expression->datatype = type_void; + + return expression; } static expression_t *parse_array_expression(unsigned precedence, @@ -1891,7 +2162,7 @@ static expression_t *parse_array_expression(unsigned precedence, if(token.type != ']') { parse_error_expected("Problem while parsing array access", ']', 0); - return NULL; + return (expression_t*) array_access; } next_token(); @@ -1901,18 +2172,11 @@ static expression_t *parse_array_expression(unsigned precedence, static bool is_declaration_specifier(const token_t *token, bool only_type_specifiers) { - 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; + return is_typedef_symbol(token->v.symbol); STORAGE_CLASSES TYPE_QUALIFIERS if(only_type_specifiers) @@ -1963,7 +2227,7 @@ static expression_t *parse_select_expression(unsigned precedence, if(token.type != T_IDENTIFIER) { parse_error_expected("Problem while parsing select", T_IDENTIFIER, 0); - return NULL; + return (expression_t*) select; } select->symbol = token.v.symbol; next_token(); @@ -1978,7 +2242,7 @@ static expression_t *parse_call_expression(unsigned precedence, call_expression_t *call = allocate_ast_zero(sizeof(call[0])); call->expression.type = EXPR_CALL; - call->method = expression; + call->function = expression; /* parse arguments */ eat('('); @@ -2012,14 +2276,15 @@ static expression_t *parse_call_expression(unsigned precedence, type = pointer->points_to; } - if(type == NULL || type->type != TYPE_METHOD) { + if(type == NULL || type->type != TYPE_FUNCTION) { parser_print_error_prefix(); - fprintf(stderr, "expected a method type for call but found type "); + fprintf(stderr, "expected a function type for call but found " + "type "); print_type(expression->datatype); fprintf(stderr, "\n"); } else { - method_type_t *method_type = (method_type_t*) type; - call->expression.datatype = method_type->result_type; + function_type_t *function_type = (function_type_t*) type; + call->expression.datatype = function_type->result_type; } } @@ -2027,17 +2292,17 @@ static expression_t *parse_call_expression(unsigned precedence, } static void type_error(const char *msg, const source_position_t source_position, - const type_t *type) + type_t *type) { parser_print_error_prefix_pos(source_position); fprintf(stderr, "%s, but found type ", msg); print_type(type); fputc('\n', stderr); + error(); } static void type_error_incompatible(const char *msg, - const source_position_t source_position, const type_t *type1, - const type_t *type2) + const source_position_t source_position, type_t *type1, type_t *type2) { parser_print_error_prefix_pos(source_position); fprintf(stderr, "%s, incompatible types: ", msg); @@ -2045,6 +2310,7 @@ static void type_error_incompatible(const char *msg, fprintf(stderr, " - "); print_type(type2); fprintf(stderr, ")\n"); + error(); } static type_t *get_type_after_conversion(const type_t *type1, @@ -2085,7 +2351,7 @@ static expression_t *parse_conditional_expression(unsigned precedence, if(false_type == NULL) return (expression_t*) conditional; - /* 6.4.15.3 */ + /* 6.5.15.3 */ if(true_type == false_type) { conditional->expression.datatype = true_type; } else if(is_type_arithmetic(true_type) && is_type_arithmetic(false_type)) { @@ -2231,7 +2497,6 @@ static expression_t *parse_sub_expression(unsigned precedence) left = parse_primary_expression(); } assert(left != NULL); - assert(left->type != EXPR_INVALID); left->source_position = source_position; while(true) { @@ -2585,7 +2850,7 @@ static statement_t *parse_declaration_statement(void) if(before == NULL) { statement->declarations_begin = context->declarations; } else { - statement->declarations_begin = before->next; + statement->declarations_begin = before->context_next; } statement->declarations_end = last_declaration; @@ -2607,7 +2872,6 @@ static statement_t *parse_expression_statement(void) static statement_t *parse_statement(void) { - declaration_t *declaration; statement_t *statement = NULL; /* declaration or statement */ @@ -2671,9 +2935,7 @@ static statement_t *parse_statement(void) break; } - declaration = token.v.symbol->declaration; - if(declaration != NULL && - declaration->storage_class == STORAGE_CLASS_TYPEDEF) { + if(is_typedef_symbol(token.v.symbol)) { statement = parse_declaration_statement(); break; } @@ -2749,6 +3011,9 @@ static translation_unit_t *parse_translation_unit(void) { translation_unit_t *unit = allocate_ast_zero(sizeof(unit[0])); + assert(global_context == NULL); + global_context = &unit->context; + assert(context == NULL); set_context(&unit->context); @@ -2760,13 +3025,15 @@ static translation_unit_t *parse_translation_unit(void) context = NULL; last_declaration = NULL; + assert(global_context == &unit->context); + global_context = NULL; + return unit; } translation_unit_t *parse(void) { - obstack_init(&environment_obstack); - environment_stack = NEW_ARR_F(environment_entry_t*, 0); + environment_stack = NEW_ARR_F(stack_entry_t, 0); type_set_output(stderr); @@ -2777,7 +3044,6 @@ translation_unit_t *parse(void) translation_unit_t *unit = parse_translation_unit(); DEL_ARR_F(environment_stack); - obstack_free(&environment_obstack, NULL); return unit; } @@ -2788,6 +3054,7 @@ void init_parser(void) obstack_init(&temp_obst); type_int = make_atomic_type(ATOMIC_TYPE_INT, 0); + type_double = make_atomic_type(ATOMIC_TYPE_DOUBLE, 0); type_size_t = make_atomic_type(ATOMIC_TYPE_UINT, 0); type_const_char = make_atomic_type(ATOMIC_TYPE_CHAR, TYPE_QUALIFIER_CONST); type_void = make_atomic_type(ATOMIC_TYPE_VOID, 0);