X-Git-Url: http://nsz.repo.hu/git/?a=blobdiff_plain;f=parser.c;h=1c627a320172b04939a374ce38b275c98e74c63c;hb=92f3ee4c4e6f2709baa59bbf3ef1aae1d711f855;hp=eabb57e667cab3add04ed0532f50bfc12f739bd8;hpb=e3055b9a25f3d712c685e6e3f32b3d3ba064a400;p=cparser diff --git a/parser.c b/parser.c index eabb57e..f63c266 100644 --- a/parser.c +++ b/parser.c @@ -19,28 +19,48 @@ #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 namespc; +} stack_entry_t; + +typedef struct declaration_specifiers_t declaration_specifiers_t; +struct declaration_specifiers_t { + source_position_t source_position; + unsigned char storage_class; + bool is_inline; + type_t *type; }; -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 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; -static type_t *type_size_t = NULL; +typedef declaration_t* (*parsed_declaration_func) (declaration_t *declaration); + +static token_t token; +static token_t lookahead_buffer[MAX_LOOKAHEAD]; +static int lookahead_bufpos; +static stack_entry_t *environment_stack = NULL; +static stack_entry_t *label_stack = NULL; +static context_t *global_context = NULL; +static context_t *context = NULL; +static declaration_t *last_declaration = NULL; +static declaration_t *current_function = NULL; +static struct obstack temp_obst; +static bool found_error; + +static type_t *type_int = NULL; +static type_t *type_long_double = NULL; +static type_t *type_double = NULL; +static type_t *type_float = NULL; +static type_t *type_char = NULL; +static type_t *type_string = NULL; +static type_t *type_void = NULL; +static type_t *type_void_ptr = NULL; +static type_t *type_valist = NULL; + +type_t *type_size_t = NULL; +type_t *type_ptrdiff_t = NULL; +type_t *type_wchar_t = NULL; +type_t *type_wchar_t_ptr = NULL; static statement_t *parse_compound_statement(void); static statement_t *parse_statement(void); @@ -49,6 +69,13 @@ static expression_t *parse_sub_expression(unsigned precedence); static expression_t *parse_expression(void); static type_t *parse_typename(void); +static void parse_compound_type_entries(void); +static declaration_t *parse_declarator( + const declaration_specifiers_t *specifiers, bool may_be_abstract); +static declaration_t *record_declaration(declaration_t *declaration); + +static void semantic_comparison(binary_expression_t *expression); + #define STORAGE_CLASSES \ case T_typedef: \ case T_extern: \ @@ -65,33 +92,30 @@ static type_t *parse_typename(void); #ifdef PROVIDE_COMPLEX #define COMPLEX_SPECIFIERS \ case T__Complex: -#else -#define COMPLEX_SPECIFIERS -#endif - -#ifdef PROVIDE_IMAGINARY #define IMAGINARY_SPECIFIERS \ case T__Imaginary: #else +#define COMPLEX_SPECIFIERS #define IMAGINARY_SPECIFIERS #endif -#define TYPE_SPECIFIERS \ - case T_void: \ - case T_char: \ - case T_short: \ - case T_int: \ - case T_long: \ - case T_float: \ - case T_double: \ - case T_signed: \ - case T_unsigned: \ - case T__Bool: \ - case T_struct: \ - case T_union: \ - case T_enum: \ - case T___typeof__: \ - COMPLEX_SPECIFIERS \ +#define TYPE_SPECIFIERS \ + case T_void: \ + case T_char: \ + case T_short: \ + case T_int: \ + case T_long: \ + case T_float: \ + case T_double: \ + case T_signed: \ + case T_unsigned: \ + case T__Bool: \ + case T_struct: \ + case T_union: \ + case T_enum: \ + case T___typeof__: \ + case T___builtin_va_list: \ + COMPLEX_SPECIFIERS \ IMAGINARY_SPECIFIERS #define DECLARATION_START \ @@ -103,28 +127,160 @@ static type_t *parse_typename(void); TYPE_QUALIFIERS \ TYPE_SPECIFIERS -static inline void *allocate_ast_zero(size_t size) +static void *allocate_ast_zero(size_t size) { void *res = allocate_ast(size); memset(res, 0, size); return res; } -static inline void *allocate_type_zero(size_t size) +static size_t get_statement_struct_size(statement_type_t type) +{ + static const size_t sizes[] = { + [STATEMENT_COMPOUND] = sizeof(compound_statement_t), + [STATEMENT_RETURN] = sizeof(return_statement_t), + [STATEMENT_DECLARATION] = sizeof(declaration_statement_t), + [STATEMENT_IF] = sizeof(if_statement_t), + [STATEMENT_SWITCH] = sizeof(switch_statement_t), + [STATEMENT_EXPRESSION] = sizeof(expression_statement_t), + [STATEMENT_CONTINUE] = sizeof(statement_base_t), + [STATEMENT_BREAK] = sizeof(statement_base_t), + [STATEMENT_GOTO] = sizeof(goto_statement_t), + [STATEMENT_LABEL] = sizeof(label_statement_t), + [STATEMENT_CASE_LABEL] = sizeof(case_label_statement_t), + [STATEMENT_WHILE] = sizeof(while_statement_t), + [STATEMENT_DO_WHILE] = sizeof(do_while_statement_t), + [STATEMENT_FOR] = sizeof(for_statement_t), + [STATEMENT_ASM] = sizeof(asm_statement_t) + }; + assert(type <= sizeof(sizes) / sizeof(sizes[0])); + assert(sizes[type] != 0); + return sizes[type]; +} + +static statement_t *allocate_statement_zero(statement_type_t type) +{ + size_t size = get_statement_struct_size(type); + statement_t *res = allocate_ast_zero(size); + + res->base.type = type; + return res; +} + + +static size_t get_expression_struct_size(expression_type_t type) +{ + static const size_t sizes[] = { + [EXPR_INVALID] = sizeof(expression_base_t), + [EXPR_REFERENCE] = sizeof(reference_expression_t), + [EXPR_CONST] = sizeof(const_expression_t), + [EXPR_STRING_LITERAL] = sizeof(string_literal_expression_t), + [EXPR_WIDE_STRING_LITERAL] = sizeof(wide_string_literal_expression_t), + [EXPR_CALL] = sizeof(call_expression_t), + [EXPR_UNARY_FIRST] = sizeof(unary_expression_t), + [EXPR_BINARY_FIRST] = sizeof(binary_expression_t), + [EXPR_CONDITIONAL] = sizeof(conditional_expression_t), + [EXPR_SELECT] = sizeof(select_expression_t), + [EXPR_ARRAY_ACCESS] = sizeof(array_access_expression_t), + [EXPR_SIZEOF] = sizeof(sizeof_expression_t), + [EXPR_CLASSIFY_TYPE] = sizeof(classify_type_expression_t), + [EXPR_FUNCTION] = sizeof(string_literal_expression_t), + [EXPR_PRETTY_FUNCTION] = sizeof(string_literal_expression_t), + [EXPR_BUILTIN_SYMBOL] = sizeof(builtin_symbol_expression_t), + [EXPR_OFFSETOF] = sizeof(offsetof_expression_t), + [EXPR_VA_START] = sizeof(va_start_expression_t), + [EXPR_VA_ARG] = sizeof(va_arg_expression_t), + [EXPR_STATEMENT] = sizeof(statement_expression_t), + }; + if(type >= EXPR_UNARY_FIRST && type <= EXPR_UNARY_LAST) { + return sizes[EXPR_UNARY_FIRST]; + } + if(type >= EXPR_BINARY_FIRST && type <= EXPR_BINARY_LAST) { + return sizes[EXPR_BINARY_FIRST]; + } + assert(type <= sizeof(sizes) / sizeof(sizes[0])); + assert(sizes[type] != 0); + return sizes[type]; +} + +static expression_t *allocate_expression_zero(expression_type_t type) +{ + size_t size = get_expression_struct_size(type); + expression_t *res = allocate_ast_zero(size); + + res->base.type = type; + return res; +} + +static size_t get_type_struct_size(type_type_t type) +{ + static const size_t sizes[] = { + [TYPE_ATOMIC] = sizeof(atomic_type_t), + [TYPE_COMPOUND_STRUCT] = sizeof(compound_type_t), + [TYPE_COMPOUND_UNION] = sizeof(compound_type_t), + [TYPE_ENUM] = sizeof(enum_type_t), + [TYPE_FUNCTION] = sizeof(function_type_t), + [TYPE_POINTER] = sizeof(pointer_type_t), + [TYPE_ARRAY] = sizeof(array_type_t), + [TYPE_BUILTIN] = sizeof(builtin_type_t), + [TYPE_TYPEDEF] = sizeof(typedef_type_t), + [TYPE_TYPEOF] = sizeof(typeof_type_t), + }; + assert(sizeof(sizes) / sizeof(sizes[0]) == (int) TYPE_TYPEOF + 1); + assert(type <= TYPE_TYPEOF); + assert(sizes[type] != 0); + return sizes[type]; +} + +static type_t *allocate_type_zero(type_type_t type) { - void *res = obstack_alloc(type_obst, size); + size_t size = get_type_struct_size(type); + type_t *res = obstack_alloc(type_obst, size); memset(res, 0, size); + + res->base.type = type; return res; } +static size_t get_initializer_size(initializer_type_t type) +{ + static const size_t sizes[] = { + [INITIALIZER_VALUE] = sizeof(initializer_value_t), + [INITIALIZER_STRING] = sizeof(initializer_string_t), + [INITIALIZER_WIDE_STRING] = sizeof(initializer_wide_string_t), + [INITIALIZER_LIST] = sizeof(initializer_list_t) + }; + assert(type < sizeof(sizes) / sizeof(*sizes)); + assert(sizes[type] != 0); + return sizes[type]; +} + +static initializer_t *allocate_initializer(initializer_type_t type) +{ + initializer_t *result = allocate_ast_zero(get_initializer_size(type)); + result->type = type; + + return result; +} + +static void free_type(void *type) +{ + obstack_free(type_obst, type); +} + /** * returns the top element of the environment stack */ -static inline size_t environment_top(void) +static size_t environment_top(void) { return ARR_LEN(environment_stack); } +static size_t label_top(void) +{ + return ARR_LEN(label_stack); +} + static inline void next_token(void) @@ -145,41 +301,38 @@ static inline const token_t *look_ahead(int num) { assert(num > 0 && num <= MAX_LOOKAHEAD); int pos = (lookahead_bufpos+num-1) % MAX_LOOKAHEAD; - return & lookahead_buffer[pos]; + return &lookahead_buffer[pos]; } -static inline void eat(token_type_t type) -{ - assert(token.type == type); - next_token(); -} +#define eat(token_type) do { assert(token.type == token_type); next_token(); } while(0) -void error(void) +static void error(void) { + found_error = true; #ifdef ABORT_ON_ERROR abort(); #endif } -void parser_print_prefix_pos(const source_position_t source_position) +static 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); + fprintf(stderr, "%u", source_position.linenr); fputs(": ", stderr); } -void parser_print_error_prefix_pos(const source_position_t source_position) +static 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) +static void parser_print_error_prefix(void) { - parser_print_prefix_pos(token.source_position); - error(); + parser_print_error_prefix_pos(token.source_position); } static void parse_error(const char *message) @@ -188,13 +341,30 @@ static void parse_error(const char *message) fprintf(stderr, "parse error: %s\n", message); } -__attribute__((unused)) -static void parse_warning(const char *message) +static void parser_print_warning_prefix_pos( + const source_position_t source_position) +{ + parser_print_prefix_pos(source_position); + fputs("warning: ", stderr); +} + +static void parser_print_warning_prefix(void) +{ + parser_print_warning_prefix_pos(token.source_position); +} + +static void parse_warning_pos(const source_position_t source_position, + const char *const message) { - parser_print_prefix_pos(token.source_position); + parser_print_prefix_pos(source_position); fprintf(stderr, "warning: %s\n", message); } +static void parse_warning(const char *message) +{ + parse_warning_pos(token.source_position, message); +} + static void parse_error_expected(const char *message, ...) { va_list args; @@ -224,6 +394,33 @@ static void parse_error_expected(const char *message, ...) fprintf(stderr, "\n"); } +static void print_type_quoted(type_t *type) +{ + fputc('\'', stderr); + print_type(type); + fputc('\'', stderr); +} + +static void type_error(const char *msg, const source_position_t source_position, + type_t *type) +{ + parser_print_error_prefix_pos(source_position); + fprintf(stderr, "%s, but found type ", msg); + print_type_quoted(type); + fputc('\n', stderr); +} + +static void type_error_incompatible(const char *msg, + 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); + print_type_quoted(type1); + fprintf(stderr, " - "); + print_type_quoted(type2); + fprintf(stderr, ")\n"); +} + static void eat_block(void) { if(token.type == '{') @@ -257,7 +454,7 @@ static void eat_statement(void) eat(';'); } -static void eat_brace(void) +static void eat_paren(void) { if(token.type == '(') next_token(); @@ -269,7 +466,7 @@ static void eat_brace(void) return; } if(token.type == '(') { - eat_brace(); + eat_paren(); continue; } if(token.type == '{') { @@ -289,6 +486,14 @@ static void eat_brace(void) } \ next_token(); +#define expect_block(expected) \ + if(UNLIKELY(token.type != (expected))) { \ + parse_error_expected(NULL, (expected), 0); \ + eat_block(); \ + return NULL; \ + } \ + next_token(); + #define expect_void(expected) \ if(UNLIKELY(token.type != (expected))) { \ parse_error_expected(NULL, (expected), 0); \ @@ -301,361 +506,475 @@ 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->next != NULL) { + last_declaration = last_declaration->next; } } - - last_declaration = declaration; } /** * called when we find a 2nd declarator for an identifier we already have a * declarator for */ -static bool is_compatible_declaration (declaration_t *declaration, +static bool is_compatible_declaration(declaration_t *declaration, declaration_t *previous) { - /* TODO: not correct yet */ - return declaration->type == previous->type; + /* happens for K&R style function parameters */ + if(previous->type == NULL) { + previous->type = declaration->type; + return true; + } + + type_t *type1 = skip_typeref(declaration->type); + type_t *type2 = skip_typeref(previous->type); + + return types_compatible(type1, type2); +} + +static declaration_t *get_declaration(symbol_t *symbol, namespace_t namespc) +{ + declaration_t *declaration = symbol->declaration; + for( ; declaration != NULL; declaration = declaration->symbol_next) { + if(declaration->namespc == namespc) + return declaration; + } + + return NULL; +} + +static const char *get_namespace_prefix(namespace_t namespc) +{ + switch(namespc) { + case NAMESPACE_NORMAL: + return ""; + case NAMESPACE_UNION: + return "union "; + case NAMESPACE_STRUCT: + return "struct "; + case NAMESPACE_ENUM: + return "enum "; + case NAMESPACE_LABEL: + return "label "; + } + 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 *stack_push(stack_entry_t **stack_ptr, + declaration_t *declaration, + context_t *parent_context) { - symbol_t *symbol = declaration->symbol; - assert(declaration != symbol->declaration); - 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); - 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); + symbol_t *symbol = declaration->symbol; + namespace_t namespc = (namespace_t)declaration->namespc; + + /* a declaration should be only pushed once */ + declaration->parent_context = parent_context; + + declaration_t *previous_declaration = get_declaration(symbol, namespc); + 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(namespc), symbol->string); + print_type_quoted(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_quoted(previous_declaration->type); + fputc('\n', stderr); + } else { + unsigned old_storage_class = previous_declaration->storage_class; + unsigned new_storage_class = declaration->storage_class; + if (current_function == NULL) { + if (old_storage_class != STORAGE_CLASS_STATIC && + new_storage_class == STORAGE_CLASS_STATIC) { + parser_print_error_prefix_pos(declaration->source_position); + fprintf(stderr, + "static declaration of '%s' follows non-static declaration\n", + symbol->string); + parser_print_error_prefix_pos(previous_declaration->source_position); + fprintf(stderr, "previous declaration of '%s' was here\n", + symbol->string); + } else { + if (old_storage_class == STORAGE_CLASS_EXTERN) { + if (new_storage_class == STORAGE_CLASS_NONE) { + previous_declaration->storage_class = STORAGE_CLASS_NONE; + } + } else { + parser_print_warning_prefix_pos(declaration->source_position); + fprintf(stderr, "redundant declaration for '%s'\n", + symbol->string); + parser_print_warning_prefix_pos(previous_declaration->source_position); + fprintf(stderr, "previous declaration of '%s' was here\n", + symbol->string); + } + } + } else { + if (old_storage_class == STORAGE_CLASS_EXTERN && + new_storage_class == STORAGE_CLASS_EXTERN) { + parser_print_warning_prefix_pos(declaration->source_position); + fprintf(stderr, "redundant extern declaration for '%s'\n", + symbol->string); + parser_print_warning_prefix_pos(previous_declaration->source_position); + fprintf(stderr, "previous declaration of '%s' was here\n", + symbol->string); + } else { + parser_print_error_prefix_pos(declaration->source_position); + if (old_storage_class == new_storage_class) { + fprintf(stderr, "redeclaration of '%s'\n", symbol->string); + } else { + fprintf(stderr, "redeclaration of '%s' with different linkage\n", symbol->string); + } + parser_print_error_prefix_pos(previous_declaration->source_position); + fprintf(stderr, "previous declaration of '%s' was here\n", + symbol->string); + } } - return previous_declaration; } + return previous_declaration; } - environment_entry_t *entry - = obstack_alloc(&environment_obstack, sizeof(entry[0])); - memset(entry, 0, sizeof(entry[0])); - - int top = ARR_LEN(environment_stack); - ARR_RESIZE(environment_stack, top + 1); - environment_stack[top] = entry; + /* remember old declaration */ + stack_entry_t entry; + entry.symbol = symbol; + entry.old_declaration = symbol->declaration; + entry.namespc = (unsigned short) namespc; + ARR_APP1(stack_entry_t, *stack_ptr, 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_last = NULL; + declaration_t *iter = symbol->declaration; + for( ; iter != NULL; iter_last = iter, iter = iter->symbol_next) { + /* replace an entry? */ + if(iter->namespc == namespc) { + if(iter_last == NULL) { + symbol->declaration = declaration; + } else { + iter_last->symbol_next = declaration; + } + declaration->symbol_next = iter->symbol_next; + break; + } + } + if(iter == NULL) { + assert(iter_last->symbol_next == NULL); + iter_last->symbol_next = declaration; + } + } return declaration; } +static declaration_t *environment_push(declaration_t *declaration) +{ + assert(declaration->source_position.input_name != NULL); + return stack_push(&environment_stack, declaration, context); +} + +static declaration_t *label_push(declaration_t *declaration) +{ + return stack_push(&label_stack, declaration, ¤t_function->context); +} + /** * 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 stack_pop_to(stack_entry_t **stack_ptr, size_t new_top) { - environment_entry_t *entry = NULL; - size_t top = ARR_LEN(environment_stack); - size_t i; + stack_entry_t *stack = *stack_ptr; + size_t top = ARR_LEN(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]; + for(i = top; i > new_top; --i) { + stack_entry_t *entry = &stack[i - 1]; - symbol_t *symbol = entry->symbol; + declaration_t *old_declaration = entry->old_declaration; + symbol_t *symbol = entry->symbol; + namespace_t namespc = (namespace_t)entry->namespc; - symbol->declaration = entry->old_declaration; - symbol->context = entry->old_context; + /* replace/remove declaration */ + declaration_t *declaration = symbol->declaration; + assert(declaration != NULL); + if(declaration->namespc == namespc) { + if(old_declaration == NULL) { + symbol->declaration = declaration->symbol_next; + } else { + symbol->declaration = old_declaration; + } + } else { + declaration_t *iter_last = declaration; + declaration_t *iter = declaration->symbol_next; + for( ; iter != NULL; iter_last = iter, iter = iter->symbol_next) { + /* replace an entry? */ + if(iter->namespc == namespc) { + assert(iter_last != NULL); + iter_last->symbol_next = old_declaration; + old_declaration->symbol_next = iter->symbol_next; + break; + } + } + assert(iter != NULL); + } + } - --i; - } while(i != new_top); - obstack_free(&environment_obstack, entry); + ARR_SHRINKLEN(*stack_ptr, (int) new_top); +} - ARR_SHRINKLEN(environment_stack, (int) new_top); +static void environment_pop_to(size_t new_top) +{ + stack_pop_to(&environment_stack, new_top); } +static void label_pop_to(size_t new_top) +{ + stack_pop_to(&label_stack, new_top); +} -static expression_t *parse_constant_expression(void) +static int get_rank(const type_t *type) { - /* start parsing at precedence 7 (conditional expression) */ - return parse_sub_expression(7); + assert(!is_typeref(type)); + /* The C-standard allows promoting to int or unsigned int (see § 7.2.2 + * and esp. footnote 108). However we can't fold constants (yet), so we + * can't decide wether unsigned int is possible, while int always works. + * (unsigned int would be preferable when possible... for stuff like + * struct { enum { ... } bla : 4; } ) */ + if(type->type == TYPE_ENUM) + return ATOMIC_TYPE_INT; + + assert(type->type == TYPE_ATOMIC); + const atomic_type_t *atomic_type = &type->atomic; + atomic_type_type_t atype = atomic_type->atype; + return atype; } -static expression_t *parse_assignment_expression(void) +static type_t *promote_integer(type_t *type) { - /* start parsing at precedence 2 (assignment expression) */ - return parse_sub_expression(2); + if(get_rank(type) < ATOMIC_TYPE_INT) + type = type_int; + + return type; } -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 *record_declaration(declaration_t *declaration); +static expression_t *create_cast_expression(expression_t *expression, + type_t *dest_type) +{ + expression_t *cast = allocate_expression_zero(EXPR_UNARY_CAST_IMPLICIT); -typedef struct declaration_specifiers_t declaration_specifiers_t; -struct declaration_specifiers_t { - storage_class_t storage_class; - type_t *type; -}; + cast->unary.value = expression; + cast->base.datatype = dest_type; -static compound_type_t *find_compound_type(compound_type_t *types, - const symbol_t *symbol) + return cast; +} + +static bool is_null_pointer_constant(const expression_t *expression) { - compound_type_t *type = types; - for( ; type != NULL; type = type->next) { - if(type->symbol == symbol) - return type; + /* skip void* cast */ + if(expression->type == EXPR_UNARY_CAST + || expression->type == EXPR_UNARY_CAST_IMPLICIT) { + expression = expression->unary.value; } - return NULL; + /* TODO: not correct yet, should be any constant integer expression + * which evaluates to 0 */ + if (expression->type != EXPR_CONST) + return false; + + type_t *const type = skip_typeref(expression->base.datatype); + if (!is_type_integer(type)) + return false; + + return expression->conste.v.int_value == 0; } -static type_t *parse_compound_type_specifier(bool is_struct) +static expression_t *create_implicit_cast(expression_t *expression, + type_t *dest_type) { - if(is_struct) { - eat(T_struct); - } else { - eat(T_union); - } + type_t *source_type = expression->base.datatype; - symbol_t *symbol = NULL; - compound_type_t *compound_type = NULL; + if(source_type == NULL) + return expression; - if(token.type == T_IDENTIFIER) { - symbol = token.v.symbol; - next_token(); + source_type = skip_typeref(source_type); + dest_type = skip_typeref(dest_type); - if(context != NULL) { - if(is_struct) { - compound_type = find_compound_type(context->structs, symbol); - } else { - compound_type = find_compound_type(context->unions, symbol); + if(source_type == dest_type) + return expression; + + switch (dest_type->type) { + case TYPE_ENUM: + /* TODO warning for implicitly converting to enum */ + case TYPE_ATOMIC: + if (source_type->type != TYPE_ATOMIC && + source_type->type != TYPE_ENUM) { + panic("casting of non-atomic types not implemented yet"); } - } - } else if(token.type != '{') { - if(is_struct) { - parse_error_expected("problem while parsing struct type specifier", - T_IDENTIFIER, '{', 0); - } else { - parse_error_expected("problem while parsing union type specifier", - T_IDENTIFIER, '{', 0); - } - return NULL; - } + if(is_type_floating(dest_type) && !is_type_scalar(source_type)) { + type_error_incompatible("can't cast types", + expression->base.source_position, source_type, + dest_type); + return expression; + } - if(compound_type == NULL) { - compound_type = allocate_type_zero(sizeof(compound_type[0])); + return create_cast_expression(expression, dest_type); - 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; - } + case TYPE_POINTER: + switch (source_type->type) { + case TYPE_ATOMIC: + if (is_null_pointer_constant(expression)) { + return create_cast_expression(expression, dest_type); + } + break; - 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; + case TYPE_POINTER: + if (pointers_compatible(source_type, dest_type)) { + return create_cast_expression(expression, dest_type); + } + break; - int top = environment_top(); - context_t *last_context = context; - set_context(&compound_type->context); + case TYPE_ARRAY: { + array_type_t *array_type = &source_type->array; + pointer_type_t *pointer_type = &dest_type->pointer; + if (types_compatible(array_type->element_type, + pointer_type->points_to)) { + return create_cast_expression(expression, dest_type); + } + break; + } - parse_compound_type_entries(); + default: + panic("casting of non-atomic types not implemented yet"); + } - assert(context == &compound_type->context); - set_context(last_context); - environment_pop_to(top); - } + type_error_incompatible("can't implicitly cast types", + expression->base.source_position, source_type, dest_type); + return expression; - return (type_t*) compound_type; + default: + panic("casting of non-atomic types not implemented yet"); + } } -static void parse_enum_entries(void) +/** Implements the rules from § 6.5.16.1 */ +static void semantic_assign(type_t *orig_type_left, expression_t **right, + const char *context) { - eat('{'); + type_t *orig_type_right = (*right)->base.datatype; - if(token.type == '}') { - next_token(); - parse_error("empty enum not allowed"); + if(orig_type_right == NULL) + return; + + type_t *const type_left = skip_typeref(orig_type_left); + type_t *const type_right = skip_typeref(orig_type_right); + + if ((is_type_arithmetic(type_left) && is_type_arithmetic(type_right)) || + (is_type_pointer(type_left) && is_null_pointer_constant(*right)) || + (is_type_atomic(type_left, ATOMIC_TYPE_BOOL) + && is_type_pointer(type_right))) { + *right = create_implicit_cast(*right, type_left); return; } - do { - declaration_t *entry = allocate_ast_zero(sizeof(entry[0])); + if (is_type_pointer(type_left) && is_type_pointer(type_right)) { + pointer_type_t *pointer_type_left = &type_left->pointer; + pointer_type_t *pointer_type_right = &type_right->pointer; + type_t *points_to_left = pointer_type_left->points_to; + type_t *points_to_right = pointer_type_right->points_to; - if(token.type != T_IDENTIFIER) { - parse_error_expected("problem while parsing enum entry", - T_IDENTIFIER, 0); - eat_block(); + points_to_left = skip_typeref(points_to_left); + points_to_right = skip_typeref(points_to_right); + + /* the left type has all qualifiers from the right type */ + unsigned missing_qualifiers + = points_to_right->base.qualifiers & ~points_to_left->base.qualifiers; + if(missing_qualifiers != 0) { + parser_print_error_prefix(); + fprintf(stderr, "destination type "); + print_type_quoted(type_left); + fprintf(stderr, " in %s from type ", context); + print_type_quoted(type_right); + fprintf(stderr, " lacks qualifiers '"); + print_type_qualifiers(missing_qualifiers); + fprintf(stderr, "' in pointed-to type\n"); return; } - entry->storage_class = STORAGE_CLASS_ENUM_ENTRY; - entry->symbol = token.v.symbol; - entry->source_position = token.source_position; - next_token(); - if(token.type == '=') { - next_token(); - entry->initializer = parse_constant_expression(); + points_to_left = get_unqualified_type(points_to_left); + points_to_right = get_unqualified_type(points_to_right); + + if(!is_type_atomic(points_to_left, ATOMIC_TYPE_VOID) + && !is_type_atomic(points_to_right, ATOMIC_TYPE_VOID) + && !types_compatible(points_to_left, points_to_right)) { + goto incompatible_assign_types; } - record_declaration(entry); + *right = create_implicit_cast(*right, type_left); + return; + } - if(token.type != ',') - break; - next_token(); - } while(token.type != '}'); + if (is_type_compound(type_left) + && types_compatible(type_left, type_right)) { + *right = create_implicit_cast(*right, type_left); + return; + } - expect_void('}'); +incompatible_assign_types: + /* TODO: improve error message */ + parser_print_error_prefix(); + fprintf(stderr, "incompatible types in %s\n", context); + parser_print_error_prefix(); + print_type_quoted(orig_type_left); + fputs(" <- ", stderr); + print_type_quoted(orig_type_right); + fputs("\n", stderr); } -static enum_type_t *find_enum_type(enum_type_t *types, const symbol_t *symbol) +static expression_t *parse_constant_expression(void) { - enum_type_t *type = types; - for( ; type != NULL; type = type->next) { - if(type->symbol == symbol) - return type; - } + /* start parsing at precedence 7 (conditional expression) */ + return parse_sub_expression(7); +} - return NULL; +static expression_t *parse_assignment_expression(void) +{ + /* start parsing at precedence 2 (assignment expression) */ + return parse_sub_expression(2); } -static type_t *parse_enum_specifier(void) +static type_t *make_global_typedef(const char *name, type_t *type) { - eat(T_enum); + symbol_t *symbol = symbol_table_insert(name); - symbol_t *symbol = NULL; - enum_type_t *enum_type = NULL; - - if(token.type == T_IDENTIFIER) { - symbol = token.v.symbol; - next_token(); - - if(context != NULL) { - enum_type = find_enum_type(context->enums, symbol); - } - } else if(token.type != '{') { - parse_error_expected("problem while parsing enum type specifier", - T_IDENTIFIER, '{', 0); - return 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(token.type == '{') { - if(enum_type->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; - - 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; - } - - return (type_t*) enum_type; -} - -static type_t *parse_typeof(void) -{ - eat(T___typeof__); - - type_t *result; - - expect('('); - - declaration_t *declaration; - expression_t *expression; - -restart: - switch(token.type) { - case T___extension__: - /* this can be a prefix to a typename or an expression */ - /* we simply eat it now. */ - do { - next_token(); - } while(token.type == T___extension__); - goto restart; - - case T_IDENTIFIER: - declaration = token.v.symbol->declaration; - if(declaration != NULL - && declaration->storage_class == STORAGE_CLASS_TYPEDEF) { - result = parse_typename(); - break; - } - expression = parse_expression(); - result = expression->datatype; - break; - - TYPENAME_START - result = parse_typename(); - break; + declaration_t *declaration = allocate_ast_zero(sizeof(declaration[0])); + declaration->namespc = NAMESPACE_NORMAL; + declaration->storage_class = STORAGE_CLASS_TYPEDEF; + declaration->type = type; + declaration->symbol = symbol; + declaration->source_position = builtin_source_position; - default: - expression = parse_expression(); - result = expression->datatype; - break; - } + record_declaration(declaration); - expect(')'); + type_t *typedef_type = allocate_type_zero(TYPE_TYPEDEF); + typedef_type->typedeft.declaration = declaration; - return result; + return typedef_type; } static const char *parse_string_literals(void) @@ -677,7 +996,7 @@ static void parse_attributes(void) { while(true) { switch(token.type) { - case T___attribute__: + case T___attribute__: { next_token(); expect_void('('); @@ -700,13 +1019,14 @@ static void parse_attributes(void) } } 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(); + eat_paren(); break; } else { parse_string_literals(); @@ -722,176 +1042,775 @@ attributes_finished: ; } -typedef enum { - SPECIFIER_SIGNED = 1 << 0, - SPECIFIER_UNSIGNED = 1 << 1, - SPECIFIER_LONG = 1 << 2, - SPECIFIER_INT = 1 << 3, - SPECIFIER_DOUBLE = 1 << 4, - SPECIFIER_CHAR = 1 << 5, - SPECIFIER_SHORT = 1 << 6, - SPECIFIER_LONG_LONG = 1 << 7, - SPECIFIER_FLOAT = 1 << 8, - SPECIFIER_BOOL = 1 << 9, - SPECIFIER_VOID = 1 << 10, -#ifdef PROVIDE_COMPLEX - SPECIFIER_COMPLEX = 1 << 11, -#endif -#ifdef PROVIDE_IMAGINARY - SPECIFIER_IMAGINARY = 1 << 12, +#if 0 +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("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; + } +} #endif -} specifiers_t; -static type_t *create_builtin_type(symbol_t *symbol) +static initializer_t *initializer_from_string(array_type_t *type, + const char *string) { - builtin_type_t *type = allocate_type_zero(sizeof(type[0])); - type->type.type = TYPE_BUILTIN; - type->symbol = symbol; + /* TODO: check len vs. size of array type */ + (void) type; - type_t *result = typehash_insert((type_t*) type); - if(result != (type_t*) type) { - obstack_free(type_obst, type); - } + initializer_t *initializer = allocate_initializer(INITIALIZER_STRING); + initializer->string.string = string; - return result; + return initializer; } -static void parse_declaration_specifiers(declaration_specifiers_t *specifiers) +static initializer_t *initializer_from_wide_string(array_type_t *const type, + wide_string_t *const string) { - declaration_t *declaration; - type_t *type = NULL; - unsigned type_qualifiers = 0; - unsigned type_specifiers = 0; - int newtype = 0; + /* TODO: check len vs. size of array type */ + (void) type; - while(true) { - switch(token.type) { + initializer_t *const initializer = + allocate_initializer(INITIALIZER_WIDE_STRING); + initializer->wide_string.string = *string; - /* storage class */ -#define MATCH_STORAGE_CLASS(token, class) \ - case token: \ - if(specifiers->storage_class != STORAGE_CLASS_NONE) { \ - parse_error("multiple storage classes in declaration " \ - "specifiers"); \ - } \ - specifiers->storage_class = class; \ - next_token(); \ - break; + return initializer; +} - MATCH_STORAGE_CLASS(T_typedef, STORAGE_CLASS_TYPEDEF) - MATCH_STORAGE_CLASS(T_extern, STORAGE_CLASS_EXTERN) - MATCH_STORAGE_CLASS(T_static, STORAGE_CLASS_STATIC) - MATCH_STORAGE_CLASS(T_auto, STORAGE_CLASS_AUTO) - MATCH_STORAGE_CLASS(T_register, STORAGE_CLASS_REGISTER) +static initializer_t *initializer_from_expression(type_t *type, + expression_t *expression) +{ + /* TODO check that expression is a constant expression */ + + /* § 6.7.8.14/15 char array may be initialized by string literals */ + type_t *const expr_type = expression->base.datatype; + if (is_type_array(type) && expr_type->type == TYPE_POINTER) { + array_type_t *const array_type = &type->array; + type_t *const element_type = skip_typeref(array_type->element_type); + + if (element_type->type == TYPE_ATOMIC) { + switch (expression->type) { + case EXPR_STRING_LITERAL: + if (element_type->atomic.atype == ATOMIC_TYPE_CHAR) { + return initializer_from_string(array_type, + expression->string.value); + } + + case EXPR_WIDE_STRING_LITERAL: { + type_t *bare_wchar_type = skip_typeref(type_wchar_t); + if (get_unqualified_type(element_type) == bare_wchar_type) { + return initializer_from_wide_string(array_type, + &expression->wide_string.value); + } + } - /* type qualifiers */ -#define MATCH_TYPE_QUALIFIER(token, qualifier) \ - case token: \ - type_qualifiers |= qualifier; \ - next_token(); \ - break; + default: break; + } + } + } - MATCH_TYPE_QUALIFIER(T_const, TYPE_QUALIFIER_CONST); - MATCH_TYPE_QUALIFIER(T_restrict, TYPE_QUALIFIER_RESTRICT); - MATCH_TYPE_QUALIFIER(T_volatile, TYPE_QUALIFIER_VOLATILE); - MATCH_TYPE_QUALIFIER(T_inline, TYPE_QUALIFIER_INLINE); + type_t *expression_type = skip_typeref(expression->base.datatype); + if(is_type_scalar(type) || types_compatible(type, expression_type)) { + semantic_assign(type, &expression, "initializer"); - case T___extension__: - /* TODO */ - next_token(); - break; + initializer_t *result = allocate_initializer(INITIALIZER_VALUE); + result->value.value = expression; - /* type specifiers */ -#define MATCH_SPECIFIER(token, specifier, name) \ - case token: \ - next_token(); \ - if(type_specifiers & specifier) { \ - parse_error("multiple " name " type specifiers given"); \ - } else { \ - type_specifiers |= specifier; \ - } \ - break; + return result; + } - MATCH_SPECIFIER(T_void, SPECIFIER_VOID, "void") - MATCH_SPECIFIER(T_char, SPECIFIER_CHAR, "char") - MATCH_SPECIFIER(T_short, SPECIFIER_SHORT, "short") - MATCH_SPECIFIER(T_int, SPECIFIER_INT, "int") - MATCH_SPECIFIER(T_float, SPECIFIER_FLOAT, "float") - MATCH_SPECIFIER(T_double, SPECIFIER_DOUBLE, "double") - MATCH_SPECIFIER(T_signed, SPECIFIER_SIGNED, "signed") - MATCH_SPECIFIER(T_unsigned, SPECIFIER_UNSIGNED, "unsigned") - MATCH_SPECIFIER(T__Bool, SPECIFIER_BOOL, "_Bool") -#ifdef PROVIDE_COMPLEX - MATCH_SPECIFIER(T__Complex, SPECIFIER_COMPLEX, "_Complex") -#endif -#ifdef PROVIDE_IMAGINARY - MATCH_SPECIFIER(T__Imaginary, SPECIFIER_IMAGINARY, "_Imaginary") -#endif - case T_long: + return NULL; +} + +static initializer_t *parse_sub_initializer(type_t *type, + expression_t *expression, + type_t *expression_type); + +static initializer_t *parse_sub_initializer_elem(type_t *type) +{ + if(token.type == '{') { + return parse_sub_initializer(type, NULL, NULL); + } + + expression_t *expression = parse_assignment_expression(); + type_t *expression_type = skip_typeref(expression->base.datatype); + + return parse_sub_initializer(type, expression, expression_type); +} + +static bool had_initializer_brace_warning; + +static initializer_t *parse_sub_initializer(type_t *type, + expression_t *expression, + type_t *expression_type) +{ + if(is_type_scalar(type)) { + /* there might be extra {} hierarchies */ + if(token.type == '{') { next_token(); - if(type_specifiers & SPECIFIER_LONG_LONG) { - parse_error("multiple type specifiers given"); - } else if(type_specifiers & SPECIFIER_LONG) { - type_specifiers |= SPECIFIER_LONG_LONG; - } else { - type_specifiers |= SPECIFIER_LONG; + if(!had_initializer_brace_warning) { + parse_warning("braces around scalar initializer"); + had_initializer_brace_warning = true; } - break; + initializer_t *result = parse_sub_initializer(type, NULL, NULL); + if(token.type == ',') { + next_token(); + /* TODO: warn about excessive elements */ + } + expect_block('}'); + return result; + } - /* TODO: if type != NULL for the following rules issue an error */ - case T_struct: - type = parse_compound_type_specifier(true); - break; - case T_union: - type = parse_compound_type_specifier(false); - break; - case T_enum: - type = parse_enum_specifier(); - break; - case T___typeof__: - type = parse_typeof(); - break; - case T___builtin_va_list: - type = create_builtin_type(token.v.symbol); - next_token(); - break; + if(expression == NULL) { + expression = parse_assignment_expression(); + } + return initializer_from_expression(type, expression); + } - case T___attribute__: - /* TODO */ - parse_attributes(); - break; + /* does the expression match the currently looked at object to initalize */ + if(expression != NULL) { + initializer_t *result = initializer_from_expression(type, expression); + if(result != NULL) + return result; + } - case T_IDENTIFIER: - declaration = token.v.symbol->declaration; - if(declaration == NULL || - declaration->storage_class != STORAGE_CLASS_TYPEDEF) { - goto finish_specifiers; + bool read_paren = false; + if(token.type == '{') { + next_token(); + read_paren = true; + } + + /* descend into subtype */ + initializer_t *result = NULL; + initializer_t **elems; + if(is_type_array(type)) { + array_type_t *array_type = &type->array; + type_t *element_type = array_type->element_type; + element_type = skip_typeref(element_type); + + initializer_t *sub; + had_initializer_brace_warning = false; + if(expression == NULL) { + sub = parse_sub_initializer_elem(element_type); + } else { + sub = parse_sub_initializer(element_type, expression, + expression_type); + } + + /* didn't match the subtypes -> try the parent type */ + if(sub == NULL) { + assert(!read_paren); + return NULL; + } + + elems = NEW_ARR_F(initializer_t*, 0); + ARR_APP1(initializer_t*, elems, sub); + + while(true) { + if(token.type == '}') + break; + expect_block(','); + if(token.type == '}') + break; + + sub = parse_sub_initializer_elem(element_type); + if(sub == NULL) { + /* TODO error, do nicer cleanup */ + parse_error("member initializer didn't match"); + DEL_ARR_F(elems); + return NULL; } + ARR_APP1(initializer_t*, elems, sub); + } + } else { + assert(is_type_compound(type)); + compound_type_t *compound_type = &type->compound; + context_t *context = &compound_type->declaration->context; - type = declaration->type; - assert(type != NULL); - next_token(); - break; + declaration_t *first = context->declarations; + if(first == NULL) + return NULL; + type_t *first_type = first->type; + first_type = skip_typeref(first_type); - /* function specifier */ - default: - goto finish_specifiers; + initializer_t *sub; + had_initializer_brace_warning = false; + if(expression == NULL) { + sub = parse_sub_initializer_elem(first_type); + } else { + sub = parse_sub_initializer(first_type, expression,expression_type); } - } -finish_specifiers: + /* didn't match the subtypes -> try our parent type */ + if(sub == NULL) { + assert(!read_paren); + return NULL; + } - if(type == NULL) { - atomic_type_type_t atomic_type; + elems = NEW_ARR_F(initializer_t*, 0); + ARR_APP1(initializer_t*, elems, sub); - /* match valid basic types */ - switch(type_specifiers) { - case SPECIFIER_VOID: - atomic_type = ATOMIC_TYPE_VOID; - break; - case SPECIFIER_CHAR: - atomic_type = ATOMIC_TYPE_CHAR; - break; + declaration_t *iter = first->next; + for( ; iter != NULL; iter = iter->next) { + if(iter->symbol == NULL) + continue; + if(iter->namespc != NAMESPACE_NORMAL) + continue; + + if(token.type == '}') + break; + expect_block(','); + if(token.type == '}') + break; + + type_t *iter_type = iter->type; + iter_type = skip_typeref(iter_type); + + sub = parse_sub_initializer_elem(iter_type); + if(sub == NULL) { + /* TODO error, do nicer cleanup*/ + parse_error("member initializer didn't match"); + DEL_ARR_F(elems); + return NULL; + } + ARR_APP1(initializer_t*, elems, sub); + } + } + + int len = ARR_LEN(elems); + size_t elems_size = sizeof(initializer_t*) * len; + + initializer_list_t *init = allocate_ast_zero(sizeof(init[0]) + elems_size); + + init->initializer.type = INITIALIZER_LIST; + init->len = len; + memcpy(init->initializers, elems, elems_size); + DEL_ARR_F(elems); + + result = (initializer_t*) init; + + if(read_paren) { + if(token.type == ',') + next_token(); + expect('}'); + } + return result; +} + +static initializer_t *parse_initializer(type_t *type) +{ + initializer_t *result; + + type = skip_typeref(type); + + if(token.type != '{') { + expression_t *expression = parse_assignment_expression(); + initializer_t *initializer = initializer_from_expression(type, expression); + if(initializer == NULL) { + parser_print_error_prefix(); + fprintf(stderr, "initializer expression '"); + print_expression(expression); + fprintf(stderr, "', type "); + print_type_quoted(expression->base.datatype); + fprintf(stderr, " is incompatible with type "); + print_type_quoted(type); + fprintf(stderr, "\n"); + } + return initializer; + } + + if(is_type_scalar(type)) { + /* § 6.7.8.11 */ + eat('{'); + + expression_t *expression = parse_assignment_expression(); + result = initializer_from_expression(type, expression); + + if(token.type == ',') + next_token(); + + expect('}'); + return result; + } else { + result = parse_sub_initializer(type, NULL, NULL); + } + + return result; +} + + + +static declaration_t *parse_compound_type_specifier(bool is_struct) +{ + if(is_struct) { + eat(T_struct); + } else { + eat(T_union); + } + + symbol_t *symbol = NULL; + declaration_t *declaration = NULL; + + if (token.type == T___attribute__) { + /* TODO */ + parse_attributes(); + } + + if(token.type == T_IDENTIFIER) { + symbol = token.v.symbol; + next_token(); + + if(is_struct) { + declaration = get_declaration(symbol, NAMESPACE_STRUCT); + } else { + declaration = get_declaration(symbol, NAMESPACE_UNION); + } + } else if(token.type != '{') { + if(is_struct) { + parse_error_expected("while parsing struct type specifier", + T_IDENTIFIER, '{', 0); + } else { + parse_error_expected("while parsing union type specifier", + T_IDENTIFIER, '{', 0); + } + + return NULL; + } + + if(declaration == NULL) { + declaration = allocate_ast_zero(sizeof(declaration[0])); + + if(is_struct) { + declaration->namespc = NAMESPACE_STRUCT; + } else { + declaration->namespc = NAMESPACE_UNION; + } + declaration->source_position = token.source_position; + declaration->symbol = symbol; + record_declaration(declaration); + } + + if(token.type == '{') { + 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); + declaration->context.declarations = NULL; + } + declaration->init.is_defined = true; + + int top = environment_top(); + context_t *last_context = context; + set_context(&declaration->context); + + parse_compound_type_entries(); + parse_attributes(); + + assert(context == &declaration->context); + set_context(last_context); + environment_pop_to(top); + } + + return declaration; +} + +static void parse_enum_entries(enum_type_t *const enum_type) +{ + eat('{'); + + if(token.type == '}') { + next_token(); + parse_error("empty enum not allowed"); + return; + } + + do { + declaration_t *entry = allocate_ast_zero(sizeof(entry[0])); + + if(token.type != T_IDENTIFIER) { + parse_error_expected("while parsing enum entry", T_IDENTIFIER, 0); + eat_block(); + return; + } + entry->storage_class = STORAGE_CLASS_ENUM_ENTRY; + entry->type = (type_t*) enum_type; + entry->symbol = token.v.symbol; + entry->source_position = token.source_position; + next_token(); + + if(token.type == '=') { + next_token(); + entry->init.enum_value = parse_constant_expression(); + + /* TODO semantic */ + } + + record_declaration(entry); + + if(token.type != ',') + break; + next_token(); + } while(token.type != '}'); + + expect_void('}'); +} + +static type_t *parse_enum_specifier(void) +{ + eat(T_enum); + + declaration_t *declaration; + symbol_t *symbol; + + if(token.type == T_IDENTIFIER) { + symbol = token.v.symbol; + next_token(); + + declaration = get_declaration(symbol, NAMESPACE_ENUM); + } else if(token.type != '{') { + parse_error_expected("while parsing enum type specifier", + T_IDENTIFIER, '{', 0); + return NULL; + } else { + declaration = NULL; + symbol = NULL; + } + + if(declaration == NULL) { + declaration = allocate_ast_zero(sizeof(declaration[0])); + + declaration->namespc = NAMESPACE_ENUM; + declaration->source_position = token.source_position; + declaration->symbol = symbol; + } + + type_t *const type = allocate_type_zero(TYPE_ENUM); + type->enumt.declaration = declaration; + + if(token.type == '{') { + if(declaration->init.is_defined) { + parser_print_error_prefix(); + fprintf(stderr, "multiple definitions of enum %s\n", + symbol->string); + } + record_declaration(declaration); + declaration->init.is_defined = 1; + + parse_enum_entries(&type->enumt); + parse_attributes(); + } + + return type; +} + +/** + * if a symbol is a typedef to another type, return true + */ +static bool is_typedef_symbol(symbol_t *symbol) +{ + const declaration_t *const declaration = + get_declaration(symbol, NAMESPACE_NORMAL); + return + declaration != NULL && + declaration->storage_class == STORAGE_CLASS_TYPEDEF; +} + +static type_t *parse_typeof(void) +{ + eat(T___typeof__); + + type_t *type; + + expect('('); + + expression_t *expression = NULL; + +restart: + switch(token.type) { + case T___extension__: + /* this can be a prefix to a typename or an expression */ + /* we simply eat it now. */ + do { + next_token(); + } while(token.type == T___extension__); + goto restart; + + case T_IDENTIFIER: + if(is_typedef_symbol(token.v.symbol)) { + type = parse_typename(); + } else { + expression = parse_expression(); + type = expression->base.datatype; + } + break; + + TYPENAME_START + type = parse_typename(); + break; + + default: + expression = parse_expression(); + type = expression->base.datatype; + break; + } + + expect(')'); + + type_t *typeof_type = allocate_type_zero(TYPE_TYPEOF); + typeof_type->typeoft.expression = expression; + typeof_type->typeoft.typeof_type = type; + + return typeof_type; +} + +typedef enum { + SPECIFIER_SIGNED = 1 << 0, + SPECIFIER_UNSIGNED = 1 << 1, + SPECIFIER_LONG = 1 << 2, + SPECIFIER_INT = 1 << 3, + SPECIFIER_DOUBLE = 1 << 4, + SPECIFIER_CHAR = 1 << 5, + SPECIFIER_SHORT = 1 << 6, + SPECIFIER_LONG_LONG = 1 << 7, + SPECIFIER_FLOAT = 1 << 8, + SPECIFIER_BOOL = 1 << 9, + SPECIFIER_VOID = 1 << 10, +#ifdef PROVIDE_COMPLEX + SPECIFIER_COMPLEX = 1 << 11, + SPECIFIER_IMAGINARY = 1 << 12, +#endif +} specifiers_t; + +static type_t *create_builtin_type(symbol_t *const symbol, + type_t *const real_type) +{ + type_t *type = allocate_type_zero(TYPE_BUILTIN); + type->builtin.symbol = symbol; + type->builtin.real_type = real_type; + + type_t *result = typehash_insert(type); + if (type != result) { + free_type(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; + + type_t *type = allocate_type_zero(TYPE_TYPEDEF); + type->typedeft.declaration = declaration; + + return type; +} + +static void parse_declaration_specifiers(declaration_specifiers_t *specifiers) +{ + type_t *type = NULL; + unsigned type_qualifiers = 0; + unsigned type_specifiers = 0; + int newtype = 0; + + specifiers->source_position = token.source_position; + + while(true) { + switch(token.type) { + + /* storage class */ +#define MATCH_STORAGE_CLASS(token, class) \ + case token: \ + if(specifiers->storage_class != STORAGE_CLASS_NONE) { \ + parse_error("multiple storage classes in declaration " \ + "specifiers"); \ + } \ + specifiers->storage_class = class; \ + next_token(); \ + break; + + MATCH_STORAGE_CLASS(T_typedef, STORAGE_CLASS_TYPEDEF) + MATCH_STORAGE_CLASS(T_extern, STORAGE_CLASS_EXTERN) + MATCH_STORAGE_CLASS(T_static, STORAGE_CLASS_STATIC) + MATCH_STORAGE_CLASS(T_auto, STORAGE_CLASS_AUTO) + MATCH_STORAGE_CLASS(T_register, STORAGE_CLASS_REGISTER) + + case T___thread: + switch (specifiers->storage_class) { + case STORAGE_CLASS_NONE: + specifiers->storage_class = STORAGE_CLASS_THREAD; + break; + + case STORAGE_CLASS_EXTERN: + specifiers->storage_class = STORAGE_CLASS_THREAD_EXTERN; + break; + + case STORAGE_CLASS_STATIC: + specifiers->storage_class = STORAGE_CLASS_THREAD_STATIC; + break; + + default: + parse_error("multiple storage classes in declaration specifiers"); + break; + } + next_token(); + break; + + /* type qualifiers */ +#define MATCH_TYPE_QUALIFIER(token, qualifier) \ + case token: \ + type_qualifiers |= qualifier; \ + next_token(); \ + break; + + MATCH_TYPE_QUALIFIER(T_const, TYPE_QUALIFIER_CONST); + MATCH_TYPE_QUALIFIER(T_restrict, TYPE_QUALIFIER_RESTRICT); + MATCH_TYPE_QUALIFIER(T_volatile, TYPE_QUALIFIER_VOLATILE); + + case T___extension__: + /* TODO */ + next_token(); + break; + + /* type specifiers */ +#define MATCH_SPECIFIER(token, specifier, name) \ + case token: \ + next_token(); \ + if(type_specifiers & specifier) { \ + parse_error("multiple " name " type specifiers given"); \ + } else { \ + type_specifiers |= specifier; \ + } \ + break; + + MATCH_SPECIFIER(T_void, SPECIFIER_VOID, "void") + MATCH_SPECIFIER(T_char, SPECIFIER_CHAR, "char") + MATCH_SPECIFIER(T_short, SPECIFIER_SHORT, "short") + MATCH_SPECIFIER(T_int, SPECIFIER_INT, "int") + MATCH_SPECIFIER(T_float, SPECIFIER_FLOAT, "float") + MATCH_SPECIFIER(T_double, SPECIFIER_DOUBLE, "double") + MATCH_SPECIFIER(T_signed, SPECIFIER_SIGNED, "signed") + MATCH_SPECIFIER(T_unsigned, SPECIFIER_UNSIGNED, "unsigned") + MATCH_SPECIFIER(T__Bool, SPECIFIER_BOOL, "_Bool") +#ifdef PROVIDE_COMPLEX + MATCH_SPECIFIER(T__Complex, SPECIFIER_COMPLEX, "_Complex") + MATCH_SPECIFIER(T__Imaginary, SPECIFIER_IMAGINARY, "_Imaginary") +#endif + case T_inline: + next_token(); + specifiers->is_inline = true; + break; + + case T_long: + next_token(); + if(type_specifiers & SPECIFIER_LONG_LONG) { + parse_error("multiple type specifiers given"); + } else if(type_specifiers & SPECIFIER_LONG) { + type_specifiers |= SPECIFIER_LONG_LONG; + } else { + type_specifiers |= SPECIFIER_LONG; + } + break; + + /* TODO: if type != NULL for the following rules should issue + * an error */ + case T_struct: { + type = allocate_type_zero(TYPE_COMPOUND_STRUCT); + + type->compound.declaration = parse_compound_type_specifier(true); + break; + } + case T_union: { + type = allocate_type_zero(TYPE_COMPOUND_STRUCT); + + type->compound.declaration = parse_compound_type_specifier(false); + break; + } + case T_enum: + type = parse_enum_specifier(); + break; + case T___typeof__: + type = parse_typeof(); + break; + case T___builtin_va_list: + type = duplicate_type(type_valist); + next_token(); + break; + + case T___attribute__: + /* TODO */ + parse_attributes(); + break; + + case T_IDENTIFIER: { + type_t *typedef_type = get_typedef_type(token.v.symbol); + + if(typedef_type == NULL) + goto finish_specifiers; + + next_token(); + type = typedef_type; + break; + } + + /* function specifier */ + default: + goto finish_specifiers; + } + } + +finish_specifiers: + + if(type == NULL) { + atomic_type_type_t atomic_type; + + /* match valid basic types */ + switch(type_specifiers) { + case SPECIFIER_VOID: + atomic_type = ATOMIC_TYPE_VOID; + break; + case SPECIFIER_CHAR: + atomic_type = ATOMIC_TYPE_CHAR; + break; case SPECIFIER_SIGNED | SPECIFIER_CHAR: atomic_type = ATOMIC_TYPE_SCHAR; break; @@ -934,1371 +1853,2555 @@ finish_specifiers: | SPECIFIER_INT: atomic_type = ATOMIC_TYPE_LONGLONG; break; - case SPECIFIER_UNSIGNED | SPECIFIER_LONG | SPECIFIER_LONG_LONG: - case SPECIFIER_UNSIGNED | SPECIFIER_LONG | SPECIFIER_LONG_LONG - | SPECIFIER_INT: - atomic_type = ATOMIC_TYPE_ULONGLONG; + case SPECIFIER_UNSIGNED | SPECIFIER_LONG | SPECIFIER_LONG_LONG: + case SPECIFIER_UNSIGNED | SPECIFIER_LONG | SPECIFIER_LONG_LONG + | SPECIFIER_INT: + atomic_type = ATOMIC_TYPE_ULONGLONG; + break; + case SPECIFIER_FLOAT: + atomic_type = ATOMIC_TYPE_FLOAT; + break; + case SPECIFIER_DOUBLE: + atomic_type = ATOMIC_TYPE_DOUBLE; + break; + case SPECIFIER_LONG | SPECIFIER_DOUBLE: + atomic_type = ATOMIC_TYPE_LONG_DOUBLE; + break; + case SPECIFIER_BOOL: + atomic_type = ATOMIC_TYPE_BOOL; + break; +#ifdef PROVIDE_COMPLEX + case SPECIFIER_FLOAT | SPECIFIER_COMPLEX: + atomic_type = ATOMIC_TYPE_FLOAT_COMPLEX; + break; + case SPECIFIER_DOUBLE | SPECIFIER_COMPLEX: + atomic_type = ATOMIC_TYPE_DOUBLE_COMPLEX; + break; + case SPECIFIER_LONG | SPECIFIER_DOUBLE | SPECIFIER_COMPLEX: + atomic_type = ATOMIC_TYPE_LONG_DOUBLE_COMPLEX; + break; + case SPECIFIER_FLOAT | SPECIFIER_IMAGINARY: + atomic_type = ATOMIC_TYPE_FLOAT_IMAGINARY; + break; + case SPECIFIER_DOUBLE | SPECIFIER_IMAGINARY: + atomic_type = ATOMIC_TYPE_DOUBLE_IMAGINARY; + break; + case SPECIFIER_LONG | SPECIFIER_DOUBLE | SPECIFIER_IMAGINARY: + atomic_type = ATOMIC_TYPE_LONG_DOUBLE_IMAGINARY; + break; +#endif + 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"); + } else if(type_specifiers & (SPECIFIER_SIGNED | SPECIFIER_UNSIGNED)) { + parse_error("only integer types can be signed or unsigned"); + } else { + parse_error("multiple datatypes in declaration"); + } + atomic_type = ATOMIC_TYPE_INVALID; + } + + type = allocate_type_zero(TYPE_ATOMIC); + type->atomic.atype = atomic_type; + newtype = 1; + } else { + if(type_specifiers != 0) { + parse_error("multiple datatypes in declaration"); + } + } + + type->base.qualifiers = type_qualifiers; + + type_t *result = typehash_insert(type); + if(newtype && result != type) { + free_type(type); + } + + specifiers->type = result; +} + +static type_qualifiers_t parse_type_qualifiers(void) +{ + type_qualifiers_t type_qualifiers = TYPE_QUALIFIER_NONE; + + while(true) { + switch(token.type) { + /* type qualifiers */ + MATCH_TYPE_QUALIFIER(T_const, TYPE_QUALIFIER_CONST); + MATCH_TYPE_QUALIFIER(T_restrict, TYPE_QUALIFIER_RESTRICT); + MATCH_TYPE_QUALIFIER(T_volatile, TYPE_QUALIFIER_VOLATILE); + + default: + return type_qualifiers; + } + } +} + +static declaration_t *parse_identifier_list(void) +{ + declaration_t *declarations = NULL; + declaration_t *last_declaration = NULL; + do { + declaration_t *declaration = allocate_ast_zero(sizeof(declaration[0])); + + declaration->source_position = token.source_position; + declaration->symbol = token.v.symbol; + next_token(); + + if(last_declaration != NULL) { + last_declaration->next = declaration; + } else { + declarations = declaration; + } + last_declaration = declaration; + + if(token.type != ',') + break; + next_token(); + } while(token.type == T_IDENTIFIER); + + return declarations; +} + +static void semantic_parameter(declaration_t *declaration) +{ + /* TODO: improve error messages */ + + if(declaration->storage_class == STORAGE_CLASS_TYPEDEF) { + parse_error("typedef not allowed in parameter list"); + } else if(declaration->storage_class != STORAGE_CLASS_NONE + && declaration->storage_class != STORAGE_CLASS_REGISTER) { + parse_error("parameter may only have none or register storage class"); + } + + type_t *orig_type = declaration->type; + if(orig_type == NULL) + return; + type_t *type = skip_typeref(orig_type); + + /* Array as last part of a paramter type is just syntactic sugar. Turn it + * into a pointer. § 6.7.5.3 (7) */ + if (is_type_array(type)) { + const array_type_t *arr_type = &type->array; + type_t *element_type = arr_type->element_type; + + type = make_pointer_type(element_type, type->base.qualifiers); + + declaration->type = type; + } + + if(is_type_incomplete(type)) { + parser_print_error_prefix(); + fprintf(stderr, "incomplete type ("); + print_type_quoted(orig_type); + fprintf(stderr, ") not allowed for parameter '%s'\n", + declaration->symbol->string); + } +} + +static declaration_t *parse_parameter(void) +{ + declaration_specifiers_t specifiers; + memset(&specifiers, 0, sizeof(specifiers)); + + parse_declaration_specifiers(&specifiers); + + declaration_t *declaration = parse_declarator(&specifiers, true); + + semantic_parameter(declaration); + + return declaration; +} + +static declaration_t *parse_parameters(function_type_t *type) +{ + if(token.type == T_IDENTIFIER) { + symbol_t *symbol = token.v.symbol; + if(!is_typedef_symbol(symbol)) { + type->kr_style_parameters = true; + return parse_identifier_list(); + } + } + + if(token.type == ')') { + type->unspecified_parameters = 1; + return NULL; + } + if(token.type == T_void && look_ahead(1)->type == ')') { + next_token(); + return 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) { + case T_DOTDOTDOT: + next_token(); + type->variadic = 1; + return declarations; + + case T_IDENTIFIER: + case T___extension__: + DECLARATION_START + declaration = parse_parameter(); + + parameter = obstack_alloc(type_obst, sizeof(parameter[0])); + memset(parameter, 0, sizeof(parameter[0])); + parameter->type = declaration->type; + + if(last_parameter != NULL) { + last_declaration->next = declaration; + last_parameter->next = parameter; + } else { + type->parameters = parameter; + declarations = declaration; + } + last_parameter = parameter; + last_declaration = declaration; break; - case SPECIFIER_FLOAT: - atomic_type = ATOMIC_TYPE_FLOAT; + + default: + return declarations; + } + if(token.type != ',') + return declarations; + next_token(); + } +} + +typedef enum { + CONSTRUCT_INVALID, + CONSTRUCT_POINTER, + CONSTRUCT_FUNCTION, + CONSTRUCT_ARRAY +} construct_type_type_t; + +typedef struct construct_type_t construct_type_t; +struct construct_type_t { + construct_type_type_t type; + construct_type_t *next; +}; + +typedef struct parsed_pointer_t parsed_pointer_t; +struct parsed_pointer_t { + construct_type_t construct_type; + type_qualifiers_t type_qualifiers; +}; + +typedef struct construct_function_type_t construct_function_type_t; +struct construct_function_type_t { + construct_type_t construct_type; + type_t *function_type; +}; + +typedef struct parsed_array_t parsed_array_t; +struct parsed_array_t { + construct_type_t construct_type; + type_qualifiers_t type_qualifiers; + bool is_static; + bool is_variable; + expression_t *size; +}; + +typedef struct construct_base_type_t construct_base_type_t; +struct construct_base_type_t { + construct_type_t construct_type; + type_t *type; +}; + +static construct_type_t *parse_pointer_declarator(void) +{ + eat('*'); + + parsed_pointer_t *pointer = obstack_alloc(&temp_obst, sizeof(pointer[0])); + memset(pointer, 0, sizeof(pointer[0])); + pointer->construct_type.type = CONSTRUCT_POINTER; + pointer->type_qualifiers = parse_type_qualifiers(); + + return (construct_type_t*) pointer; +} + +static construct_type_t *parse_array_declarator(void) +{ + eat('['); + + parsed_array_t *array = obstack_alloc(&temp_obst, sizeof(array[0])); + memset(array, 0, sizeof(array[0])); + array->construct_type.type = CONSTRUCT_ARRAY; + + if(token.type == T_static) { + array->is_static = true; + next_token(); + } + + type_qualifiers_t type_qualifiers = parse_type_qualifiers(); + if(type_qualifiers != 0) { + if(token.type == T_static) { + array->is_static = true; + next_token(); + } + } + array->type_qualifiers = type_qualifiers; + + if(token.type == '*' && look_ahead(1)->type == ']') { + array->is_variable = true; + next_token(); + } else if(token.type != ']') { + array->size = parse_assignment_expression(); + } + + expect(']'); + + return (construct_type_t*) array; +} + +static construct_type_t *parse_function_declarator(declaration_t *declaration) +{ + eat('('); + + type_t *type = allocate_type_zero(TYPE_FUNCTION); + + declaration_t *parameters = parse_parameters(&type->function); + if(declaration != NULL) { + declaration->context.declarations = parameters; + } + + 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_function_type; +} + +static construct_type_t *parse_inner_declarator(declaration_t *declaration, + bool may_be_abstract) +{ + /* construct a single linked list of construct_type_t's which describe + * how to construct the final declarator type */ + construct_type_t *first = NULL; + construct_type_t *last = NULL; + + /* pointers */ + while(token.type == '*') { + construct_type_t *type = parse_pointer_declarator(); + + if(last == NULL) { + first = type; + last = type; + } else { + last->next = type; + last = type; + } + } + + /* TODO: find out if this is correct */ + parse_attributes(); + + construct_type_t *inner_types = NULL; + + switch(token.type) { + case T_IDENTIFIER: + if(declaration == NULL) { + parse_error("no identifier expected in typename"); + } else { + declaration->symbol = token.v.symbol; + declaration->source_position = token.source_position; + } + next_token(); + break; + case '(': + next_token(); + inner_types = parse_inner_declarator(declaration, may_be_abstract); + expect(')'); + break; + default: + if(may_be_abstract) break; - case SPECIFIER_DOUBLE: - atomic_type = ATOMIC_TYPE_DOUBLE; + parse_error_expected("while parsing declarator", T_IDENTIFIER, '(', 0); + /* avoid a loop in the outermost scope, because eat_statement doesn't + * eat '}' */ + if(token.type == '}' && current_function == NULL) { + next_token(); + } else { + eat_statement(); + } + return NULL; + } + + construct_type_t *p = last; + + while(true) { + construct_type_t *type; + switch(token.type) { + case '(': + type = parse_function_declarator(declaration); break; - case SPECIFIER_LONG | SPECIFIER_DOUBLE: - atomic_type = ATOMIC_TYPE_LONG_DOUBLE; + case '[': + type = parse_array_declarator(); break; - case SPECIFIER_BOOL: - atomic_type = ATOMIC_TYPE_BOOL; + default: + goto declarator_finished; + } + + /* insert in the middle of the list (behind p) */ + if(p != NULL) { + type->next = p->next; + p->next = type; + } else { + type->next = first; + first = type; + } + if(last == p) { + last = type; + } + } + +declarator_finished: + parse_attributes(); + + /* append inner_types at the end of the list, we don't to set last anymore + * as it's not needed anymore */ + if(last == NULL) { + assert(first == NULL); + first = inner_types; + } else { + last->next = inner_types; + } + + return first; +} + +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) { + switch(iter->type) { + case CONSTRUCT_INVALID: + panic("invalid type construction found"); + case CONSTRUCT_FUNCTION: { + construct_function_type_t *construct_function_type + = (construct_function_type_t*) iter; + + type_t *function_type = construct_function_type->function_type; + + function_type->function.return_type = type; + + type = function_type; break; -#ifdef PROVIDE_COMPLEX - case SPECIFIER_FLOAT | SPECIFIER_COMPLEX: - atomic_type = ATOMIC_TYPE_FLOAT_COMPLEX; + } + + case CONSTRUCT_POINTER: { + parsed_pointer_t *parsed_pointer = (parsed_pointer_t*) iter; + type_t *pointer_type = allocate_type_zero(TYPE_POINTER); + pointer_type->pointer.points_to = type; + pointer_type->base.qualifiers = parsed_pointer->type_qualifiers; + + type = pointer_type; break; - case SPECIFIER_DOUBLE | SPECIFIER_COMPLEX: - atomic_type = ATOMIC_TYPE_DOUBLE_COMPLEX; + } + + case CONSTRUCT_ARRAY: { + parsed_array_t *parsed_array = (parsed_array_t*) iter; + type_t *array_type = allocate_type_zero(TYPE_ARRAY); + + array_type->base.qualifiers = parsed_array->type_qualifiers; + array_type->array.element_type = type; + array_type->array.is_static = parsed_array->is_static; + array_type->array.is_variable = parsed_array->is_variable; + array_type->array.size = parsed_array->size; + + type = array_type; break; - case SPECIFIER_LONG | SPECIFIER_DOUBLE | SPECIFIER_COMPLEX: - atomic_type = ATOMIC_TYPE_LONG_DOUBLE_COMPLEX; + } + } + + type_t *hashed_type = typehash_insert(type); + if(hashed_type != type) { + /* the function type was constructed earlier freeing it here will + * destroy other types... */ + if(iter->type != CONSTRUCT_FUNCTION) { + free_type(type); + } + type = hashed_type; + } + } + + return type; +} + +static declaration_t *parse_declarator( + const declaration_specifiers_t *specifiers, bool may_be_abstract) +{ + type_t *type = specifiers->type; + declaration_t *declaration = allocate_ast_zero(sizeof(declaration[0])); + declaration->storage_class = specifiers->storage_class; + declaration->is_inline = specifiers->is_inline; + + construct_type_t *construct_type + = parse_inner_declarator(declaration, may_be_abstract); + declaration->type = construct_declarator_type(construct_type, type); + + 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); + + type_t *result = construct_declarator_type(construct_type, base_type); + if(construct_type != NULL) { + obstack_free(&temp_obst, construct_type); + } + + return result; +} + +static declaration_t *record_declaration(declaration_t *declaration) +{ + assert(declaration->parent_context == NULL); + assert(context != NULL); + + symbol_t *symbol = declaration->symbol; + if(symbol != NULL) { + declaration_t *alias = environment_push(declaration); + if(alias != declaration) + return alias; + } else { + declaration->parent_context = context; + } + + if(last_declaration != NULL) { + last_declaration->next = declaration; + } else { + context->declarations = declaration; + } + last_declaration = declaration; + + return declaration; +} + +static void parser_error_multiple_definition(declaration_t *declaration, + const source_position_t source_position) +{ + parser_print_error_prefix_pos(source_position); + fprintf(stderr, "multiple definition of symbol '%s'\n", + declaration->symbol->string); + parser_print_error_prefix_pos(declaration->source_position); + fprintf(stderr, "this is the location of the previous definition.\n"); +} + +static bool is_declaration_specifier(const token_t *token, + bool only_type_specifiers) +{ + switch(token->type) { + TYPE_SPECIFIERS + return true; + case T_IDENTIFIER: + return is_typedef_symbol(token->v.symbol); + + case T___extension__: + STORAGE_CLASSES + TYPE_QUALIFIERS + return !only_type_specifiers; + + default: + return false; + } +} + +static void parse_init_declarator_rest(declaration_t *declaration) +{ + eat('='); + + type_t *orig_type = declaration->type; + type_t *type = NULL; + if(orig_type != NULL) + type = skip_typeref(orig_type); + + if(declaration->init.initializer != NULL) { + parser_error_multiple_definition(declaration, token.source_position); + } + + initializer_t *initializer = parse_initializer(type); + + /* § 6.7.5 (22) array intializers for arrays with unknown size determine + * the array type size */ + if(type != NULL && is_type_array(type) && initializer != NULL) { + array_type_t *array_type = &type->array; + + if(array_type->size == NULL) { + expression_t *cnst = allocate_expression_zero(EXPR_CONST); + + cnst->base.datatype = type_size_t; + + switch (initializer->type) { + case INITIALIZER_LIST: { + initializer_list_t *const list = &initializer->list; + cnst->conste.v.int_value = list->len; + break; + } + + case INITIALIZER_STRING: { + initializer_string_t *const string = &initializer->string; + cnst->conste.v.int_value = strlen(string->string) + 1; + break; + } + + case INITIALIZER_WIDE_STRING: { + initializer_wide_string_t *const string = &initializer->wide_string; + cnst->conste.v.int_value = string->string.size; + break; + } + + default: + panic("invalid initializer type"); + } + + array_type->size = cnst; + } + } + + if(type != NULL && is_type_function(type)) { + parser_print_error_prefix_pos(declaration->source_position); + fprintf(stderr, "initializers not allowed for function types at " + "declator '%s' (type ", declaration->symbol->string); + print_type_quoted(orig_type); + fprintf(stderr, ")\n"); + } else { + declaration->init.initializer = initializer; + } +} + +/* parse rest of a declaration without any declarator */ +static void parse_anonymous_declaration_rest( + const declaration_specifiers_t *specifiers, + parsed_declaration_func finished_declaration) +{ + eat(';'); + + declaration_t *declaration = allocate_ast_zero(sizeof(declaration[0])); + + declaration->type = specifiers->type; + declaration->storage_class = specifiers->storage_class; + declaration->source_position = specifiers->source_position; + + if (declaration->storage_class != STORAGE_CLASS_NONE) { + parse_warning_pos(declaration->source_position, + "useless storage class in empty declaration"); + } + + type_t *type = declaration->type; + switch (type->type) { + case TYPE_COMPOUND_STRUCT: + case TYPE_COMPOUND_UNION: { + const compound_type_t *compound_type = &type->compound; + if (compound_type->declaration->symbol == NULL) { + parse_warning_pos(declaration->source_position, + "unnamed struct/union that defines no instances"); + } break; -#endif -#ifdef PROVIDE_IMAGINARY - case SPECIFIER_FLOAT | SPECIFIER_IMAGINARY: - atomic_type = ATOMIC_TYPE_FLOAT_IMAGINARY; + } + + case TYPE_ENUM: break; - case SPECIFIER_DOUBLE | SPECIFIER_IMAGINARY: - atomic_type = ATOMIC_TYPE_DOUBLE_IMAGINARY; + + default: + parse_warning_pos(declaration->source_position, + "empty declaration"); break; - case SPECIFIER_LONG | SPECIFIER_DOUBLE | SPECIFIER_IMAGINARY: - atomic_type = ATOMIC_TYPE_LONG_DOUBLE_IMAGINARY; + } + + finished_declaration(declaration); +} + +static void parse_declaration_rest(declaration_t *ndeclaration, + const declaration_specifiers_t *specifiers, + parsed_declaration_func finished_declaration) +{ + while(true) { + declaration_t *declaration = finished_declaration(ndeclaration); + + type_t *orig_type = declaration->type; + type_t *type = skip_typeref(orig_type); + + if(type->type != TYPE_FUNCTION && declaration->is_inline) { + parser_print_warning_prefix_pos(declaration->source_position); + fprintf(stderr, "variable '%s' declared 'inline'\n", + declaration->symbol->string); + } + + if(token.type == '=') { + parse_init_declarator_rest(declaration); + } + + if(token.type != ',') break; -#endif - 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"); - } else if(type_specifiers & (SPECIFIER_SIGNED | SPECIFIER_UNSIGNED)) { - parse_error("only integer types can be signed or unsigned"); - } else { - parse_error("multiple datatypes in declaration"); - } - atomic_type = ATOMIC_TYPE_INVALID; + eat(','); + + ndeclaration = parse_declarator(specifiers, false); + } + expect_void(';'); +} + +static declaration_t *finished_kr_declaration(declaration_t *declaration) +{ + /* TODO: check that it was actually a parameter that gets a type */ + + /* we should have a declaration for the parameter in the current + * scope */ + return record_declaration(declaration); +} + +static void parse_declaration(parsed_declaration_func finished_declaration) +{ + declaration_specifiers_t specifiers; + memset(&specifiers, 0, sizeof(specifiers)); + parse_declaration_specifiers(&specifiers); + + if(token.type == ';') { + parse_anonymous_declaration_rest(&specifiers, finished_declaration); + } else { + declaration_t *declaration = parse_declarator(&specifiers, false); + parse_declaration_rest(declaration, &specifiers, finished_declaration); + } +} + +static void parse_kr_declaration_list(declaration_t *declaration) +{ + type_t *type = skip_typeref(declaration->type); + if(!is_type_function(type)) + return; + + if(!type->function.kr_style_parameters) + return; + + /* push function parameters */ + int top = environment_top(); + context_t *last_context = context; + set_context(&declaration->context); + + declaration_t *parameter = declaration->context.declarations; + for( ; parameter != NULL; parameter = parameter->next) { + environment_push(parameter); + } + + /* parse declaration list */ + while(is_declaration_specifier(&token, false)) { + parse_declaration(finished_kr_declaration); + } + + /* pop function parameters */ + assert(context == &declaration->context); + set_context(last_context); + environment_pop_to(top); + + /* update function type */ + type_t *new_type = duplicate_type(type); + new_type->function.kr_style_parameters = false; + + function_parameter_t *parameters = NULL; + function_parameter_t *last_parameter = NULL; + + declaration_t *parameter_declaration = declaration->context.declarations; + for( ; parameter_declaration != NULL; + parameter_declaration = parameter_declaration->next) { + type_t *parameter_type = parameter_declaration->type; + if(parameter_type == NULL) { +#ifdef STRICT_C99 + parser_print_error_prefix(); + fprintf(stderr, "no type specified for function parameter '%s'\n", + parameter_declaration->symbol->string); +#else + parser_print_warning_prefix(); + fprintf(stderr, "no type specified for function parameter '%s', " + "using int\n", parameter_declaration->symbol->string); + parameter_type = type_int; + parameter_declaration->type = parameter_type; +#endif } - atomic_type_t *atype = allocate_type_zero(sizeof(atype[0])); - atype->type.type = TYPE_ATOMIC; - atype->atype = atomic_type; - newtype = 1; + semantic_parameter(parameter_declaration); + parameter_type = parameter_declaration->type; - type = (type_t*) atype; - } else { - if(type_specifiers != 0) { - parse_error("multiple datatypes in declaration"); + function_parameter_t *function_parameter + = obstack_alloc(type_obst, sizeof(function_parameter[0])); + memset(function_parameter, 0, sizeof(function_parameter[0])); + + function_parameter->type = parameter_type; + if(last_parameter != NULL) { + last_parameter->next = function_parameter; + } else { + parameters = function_parameter; } + last_parameter = function_parameter; } + new_type->function.parameters = parameters; - type->qualifiers = type_qualifiers; - - type_t *result = typehash_insert(type); - if(newtype && result != (type_t*) type) { - obstack_free(type_obst, type); + type = typehash_insert(new_type); + if(type != new_type) { + obstack_free(type_obst, new_type); } - specifiers->type = result; + declaration->type = type; } -static type_qualifier_t parse_type_qualifiers(void) +static void parse_external_declaration(void) { - type_qualifier_t type_qualifiers = 0; + /* function-definitions and declarations both start with declaration + * specifiers */ + declaration_specifiers_t specifiers; + memset(&specifiers, 0, sizeof(specifiers)); + parse_declaration_specifiers(&specifiers); - while(true) { - switch(token.type) { - /* type qualifiers */ - MATCH_TYPE_QUALIFIER(T_const, TYPE_QUALIFIER_CONST); - MATCH_TYPE_QUALIFIER(T_restrict, TYPE_QUALIFIER_RESTRICT); - MATCH_TYPE_QUALIFIER(T_volatile, TYPE_QUALIFIER_VOLATILE); - MATCH_TYPE_QUALIFIER(T_inline, TYPE_QUALIFIER_INLINE); + /* must be a declaration */ + if(token.type == ';') { + parse_anonymous_declaration_rest(&specifiers, record_declaration); + return; + } - default: - return type_qualifiers; + /* declarator is common to both function-definitions and declarations */ + declaration_t *ndeclaration = parse_declarator(&specifiers, false); + + /* must be a declaration */ + if(token.type == ',' || token.type == '=' || token.type == ';') { + parse_declaration_rest(ndeclaration, &specifiers, record_declaration); + return; + } + + /* must be a function definition */ + parse_kr_declaration_list(ndeclaration); + + if(token.type != '{') { + parse_error_expected("while parsing function definition", '{', 0); + eat_statement(); + return; + } + + type_t *type = ndeclaration->type; + if(type == NULL) { + eat_block(); + return; + } + + /* note that we don't skip typerefs: the standard doesn't allow them here + * (so we can't use is_type_function here) */ + if(type->type != TYPE_FUNCTION) { + parser_print_error_prefix(); + fprintf(stderr, "declarator '"); + print_type_ext(type, ndeclaration->symbol, NULL); + fprintf(stderr, "' has a body but is not a function type.\n"); + eat_block(); + return; + } + + /* § 6.7.5.3 (14) a function definition with () means no + * parameters (and not unspecified parameters) */ + if(type->function.unspecified_parameters) { + type_t *duplicate = duplicate_type(type); + duplicate->function.unspecified_parameters = false; + + type = typehash_insert(duplicate); + if(type != duplicate) { + obstack_free(type_obst, duplicate); } + ndeclaration->type = type; + } + + declaration_t *declaration = record_declaration(ndeclaration); + if(ndeclaration != declaration) { + memcpy(&declaration->context, &ndeclaration->context, + sizeof(declaration->context)); + } + type = skip_typeref(declaration->type); + + /* push function parameters and switch context */ + int top = environment_top(); + context_t *last_context = context; + set_context(&declaration->context); + + declaration_t *parameter = declaration->context.declarations; + for( ; parameter != NULL; parameter = parameter->next) { + environment_push(parameter); + } + + if(declaration->init.statement != NULL) { + parser_error_multiple_definition(declaration, token.source_position); + eat_block(); + goto end_of_parse_external_declaration; + } else { + /* parse function body */ + int label_stack_top = label_top(); + declaration_t *old_current_function = current_function; + current_function = declaration; + + declaration->init.statement = parse_compound_statement(); + + assert(current_function == declaration); + current_function = old_current_function; + label_pop_to(label_stack_top); } + +end_of_parse_external_declaration: + assert(context == &declaration->context); + set_context(last_context); + environment_pop_to(top); } -static void parse_identifier_list(void) +static void parse_struct_declarators(const declaration_specifiers_t *specifiers) { - while(true) { - if(token.type != T_IDENTIFIER) { - parse_error_expected("problem while parsing parameter identifier " - "list", T_IDENTIFIER, 0); - return; + while(1) { + if(token.type == ':') { + next_token(); + parse_constant_expression(); + /* TODO (bitfields) */ + } else { + declaration_t *declaration = parse_declarator(specifiers, true); + + /* TODO: check constraints for struct declarations */ + /* TODO: check for doubled fields */ + record_declaration(declaration); + + if(token.type == ':') { + next_token(); + parse_constant_expression(); + /* TODO (bitfields) */ + } } - next_token(); + if(token.type != ',') break; next_token(); } + expect_void(';'); } -static declaration_t *parse_parameter(void) +static void parse_compound_type_entries(void) +{ + eat('{'); + + while(token.type != '}' && token.type != T_EOF) { + declaration_specifiers_t specifiers; + memset(&specifiers, 0, sizeof(specifiers)); + parse_declaration_specifiers(&specifiers); + + parse_struct_declarators(&specifiers); + } + if(token.type == T_EOF) { + parse_error("EOF while parsing struct"); + } + next_token(); +} + +static type_t *parse_typename(void) { declaration_specifiers_t specifiers; memset(&specifiers, 0, sizeof(specifiers)); - parse_declaration_specifiers(&specifiers); + if(specifiers.storage_class != STORAGE_CLASS_NONE) { + /* TODO: improve error message, user does probably not know what a + * storage class is... + */ + parse_error("typename may not have a storage class"); + } - declaration_t *declaration = allocate_ast_zero(sizeof(declaration[0])); - parse_declarator(declaration, specifiers.storage_class, - specifiers.type, 1); + type_t *result = parse_abstract_declarator(specifiers.type); - return declaration; + return result; } -static declaration_t *parse_parameters(method_type_t *type) + + + +typedef expression_t* (*parse_expression_function) (unsigned precedence); +typedef expression_t* (*parse_expression_infix_function) (unsigned precedence, + expression_t *left); + +typedef struct expression_parser_function_t expression_parser_function_t; +struct expression_parser_function_t { + unsigned precedence; + parse_expression_function parser; + unsigned infix_precedence; + parse_expression_infix_function infix_parser; +}; + +expression_parser_function_t expression_parsers[T_LAST_TOKEN]; + +static expression_t *create_invalid_expression(void) { - 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 NULL; - } - } + expression_t *expression = allocate_expression_zero(EXPR_INVALID); + expression->base.source_position = token.source_position; + return expression; +} - if(token.type == ')') { - type->unspecified_parameters = 1; - return NULL; - } - if(token.type == T_void && look_ahead(1)->type == ')') { - next_token(); - return NULL; +static expression_t *expected_expression_error(void) +{ + parser_print_error_prefix(); + fprintf(stderr, "expected expression, got token "); + print_token(stderr, &token); + fprintf(stderr, "\n"); + + next_token(); + + return create_invalid_expression(); +} + +static expression_t *parse_string_const(void) +{ + expression_t *cnst = allocate_expression_zero(EXPR_STRING_LITERAL); + cnst->base.datatype = type_string; + cnst->string.value = parse_string_literals(); + + return cnst; +} + +static expression_t *parse_wide_string_const(void) +{ + expression_t *const cnst = allocate_expression_zero(EXPR_WIDE_STRING_LITERAL); + cnst->base.datatype = type_wchar_t_ptr; + cnst->wide_string.value = token.v.wide_string; /* TODO concatenate */ + next_token(); + return cnst; +} + +static expression_t *parse_int_const(void) +{ + expression_t *cnst = allocate_expression_zero(EXPR_CONST); + cnst->base.datatype = token.datatype; + cnst->conste.v.int_value = token.v.intvalue; + + next_token(); + + return cnst; +} + +static expression_t *parse_float_const(void) +{ + expression_t *cnst = allocate_expression_zero(EXPR_CONST); + cnst->base.datatype = token.datatype; + cnst->conste.v.float_value = token.v.floatvalue; + + next_token(); + + return cnst; +} + +static declaration_t *create_implicit_function(symbol_t *symbol, + const source_position_t source_position) +{ + type_t *ntype = allocate_type_zero(TYPE_FUNCTION); + ntype->function.return_type = type_int; + ntype->function.unspecified_parameters = true; + + type_t *type = typehash_insert(ntype); + if(type != ntype) { + free_type(ntype); } - declaration_t *declarations = NULL; - declaration_t *declaration; - declaration_t *last_declaration = NULL; - method_parameter_t *parameter; - method_parameter_t *last_parameter = NULL; + declaration_t *declaration = allocate_ast_zero(sizeof(declaration[0])); - while(true) { - switch(token.type) { - case T_DOTDOTDOT: - next_token(); - type->variadic = 1; - return declarations; + declaration->storage_class = STORAGE_CLASS_EXTERN; + declaration->type = type; + declaration->symbol = symbol; + declaration->source_position = source_position; - case T_IDENTIFIER: - case T___extension__: - DECLARATION_START - declaration = parse_parameter(); + /* 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); - parameter = allocate_type_zero(sizeof(parameter[0])); - parameter->type = declaration->type; + context_t *last_context = context; + context = global_context; - if(last_parameter != NULL) { - last_declaration->next = declaration; - last_parameter->next = parameter; - } else { - type->parameters = parameter; - declarations = declaration; - } - last_parameter = parameter; - last_declaration = declaration; - break; + environment_push(declaration); + declaration->next = context->declarations; + context->declarations = declaration; - default: - return declarations; - } - if(token.type != ',') - return declarations; - next_token(); + context = last_context; + + return declaration; +} + +static type_t *make_function_1_type(type_t *return_type, type_t *argument_type) +{ + function_parameter_t *parameter + = obstack_alloc(type_obst, sizeof(parameter[0])); + memset(parameter, 0, sizeof(parameter[0])); + parameter->type = argument_type; + + type_t *type = allocate_type_zero(TYPE_FUNCTION); + type->function.return_type = return_type; + type->function.parameters = parameter; + + type_t *result = typehash_insert(type); + if(result != type) { + free_type(type); } + + return result; } -typedef enum { - CONSTRUCT_POINTER, - CONSTRUCT_METHOD, - CONSTRUCT_ARRAY -} construct_type_type_t; +static type_t *get_builtin_symbol_type(symbol_t *symbol) +{ + switch(symbol->ID) { + case T___builtin_alloca: + return make_function_1_type(type_void_ptr, type_size_t); + case T___builtin_nan: + return make_function_1_type(type_double, type_string); + case T___builtin_nanf: + return make_function_1_type(type_float, type_string); + case T___builtin_nand: + return make_function_1_type(type_long_double, type_string); + case T___builtin_va_end: + return make_function_1_type(type_void, type_valist); + default: + panic("not implemented builtin symbol found"); + } +} -typedef struct construct_type_t construct_type_t; -struct construct_type_t { - construct_type_type_t type; - construct_type_t *next; -}; +/** + * performs automatic type cast as described in § 6.3.2.1 + */ +static type_t *automatic_type_conversion(type_t *orig_type) +{ + if(orig_type == NULL) + return NULL; -typedef struct parsed_pointer_t parsed_pointer_t; -struct parsed_pointer_t { - construct_type_t construct_type; - type_qualifier_t type_qualifiers; -}; + type_t *type = skip_typeref(orig_type); + if(is_type_array(type)) { + array_type_t *array_type = &type->array; + type_t *element_type = array_type->element_type; + unsigned qualifiers = array_type->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; -}; + return make_pointer_type(element_type, qualifiers); + } -typedef struct parsed_array_t parsed_array_t; -struct parsed_array_t { - construct_type_t construct_type; - type_qualifier_t type_qualifiers; - bool is_static; - bool is_variable; - expression_t *size; -}; + if(is_type_function(type)) { + return make_pointer_type(orig_type, TYPE_QUALIFIER_NONE); + } -typedef struct construct_base_type_t construct_base_type_t; -struct construct_base_type_t { - construct_type_t construct_type; - type_t *type; -}; + return orig_type; +} -static construct_type_t *parse_pointer_declarator(void) +/** + * reverts the automatic casts of array to pointer types and function + * to function-pointer types as defined § 6.3.2.1 + */ +type_t *revert_automatic_type_conversion(const expression_t *expression) { - eat('*'); + if(expression->base.datatype == NULL) + return NULL; - parsed_pointer_t *pointer = obstack_alloc(&temp_obst, sizeof(pointer[0])); - memset(pointer, 0, sizeof(pointer[0])); - pointer->type_qualifiers = parse_type_qualifiers(); + switch(expression->type) { + case EXPR_REFERENCE: { + const reference_expression_t *ref = &expression->reference; + return ref->declaration->type; + } + case EXPR_SELECT: { + const select_expression_t *select = &expression->select; + return select->compound_entry->type; + } + case EXPR_UNARY_DEREFERENCE: { + expression_t *value = expression->unary.value; + type_t *type = skip_typeref(value->base.datatype); + pointer_type_t *pointer_type = &type->pointer; - return (construct_type_t*) pointer; + return pointer_type->points_to; + } + case EXPR_BUILTIN_SYMBOL: { + const builtin_symbol_expression_t *builtin + = &expression->builtin_symbol; + return get_builtin_symbol_type(builtin->symbol); + } + case EXPR_ARRAY_ACCESS: { + const array_access_expression_t *array_access + = &expression->array_access; + const expression_t *array_ref = array_access->array_ref; + type_t *type_left = skip_typeref(array_ref->base.datatype); + assert(is_type_pointer(type_left)); + pointer_type_t *pointer_type = &type_left->pointer; + return pointer_type->points_to; + } + + default: + break; + } + + return expression->base.datatype; } -static construct_type_t *parse_array_declarator(void) +static expression_t *parse_reference(void) { - eat('['); + expression_t *expression = allocate_expression_zero(EXPR_REFERENCE); - parsed_array_t *array = obstack_alloc(&temp_obst, sizeof(array[0])); - memset(array, 0, sizeof(array[0])); + reference_expression_t *ref = &expression->reference; + ref->symbol = token.v.symbol; - if(token.type == T_static) { - array->is_static = true; - next_token(); - } + declaration_t *declaration = get_declaration(ref->symbol, NAMESPACE_NORMAL); - type_qualifier_t type_qualifiers = parse_type_qualifiers(); - if(type_qualifiers != 0) { - if(token.type == T_static) { - array->is_static = true; - next_token(); + source_position_t source_position = token.source_position; + next_token(); + + if(declaration == NULL) { +#ifndef STRICT_C99 + /* an implicitly 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; } } - array->type_qualifiers = type_qualifiers; - if(token.type == '*' && look_ahead(1)->type == ']') { - array->is_variable = true; - next_token(); - } else if(token.type != ']') { - array->size = parse_assignment_expression(); - } + type_t *type = declaration->type; + /* we always do the auto-type conversions; the & and sizeof parser contains + * code to revert this! */ + type = automatic_type_conversion(type); - expect(']'); + ref->declaration = declaration; + ref->expression.datatype = type; - return (construct_type_t*) array; + return expression; } -static construct_type_t *parse_method_declarator(declaration_t *declaration) +static void check_cast_allowed(expression_t *expression, type_t *dest_type) { - eat('('); + (void) expression; + (void) dest_type; + /* TODO check if explicit cast is allowed and issue warnings/errors */ +} - method_type_t *method_type - = allocate_type_zero(sizeof(method_type[0])); - method_type->type.type = TYPE_METHOD; +static expression_t *parse_cast(void) +{ + expression_t *cast = allocate_expression_zero(EXPR_UNARY_CAST); - declaration_t *parameters = parse_parameters(method_type); - if(declaration != NULL) { - declaration->context.declarations = parameters; - } + cast->base.source_position = token.source_position; - 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; + type_t *type = parse_typename(); expect(')'); + expression_t *value = parse_sub_expression(20); + + check_cast_allowed(value, type); + + cast->base.datatype = type; + cast->unary.value = value; - return (construct_type_t*) construct_method_type; + return cast; } -static construct_type_t *parse_inner_declarator(declaration_t *declaration, - int may_be_abstract) +static expression_t *parse_statement_expression(void) { - construct_type_t *result = NULL; - construct_type_t *last = NULL; + expression_t *expression = allocate_expression_zero(EXPR_STATEMENT); - while(token.type == '*') { - construct_type_t *type = parse_pointer_declarator(); - if(last != NULL) { - last->next = type; - } else { - result = type; - } - last = type; + statement_t *statement = parse_compound_statement(); + expression->statement.statement = statement; + if(statement == NULL) { + expect(')'); + return NULL; } - /* TODO: find out if this is correct */ - parse_attributes(); + assert(statement->type == STATEMENT_COMPOUND); + compound_statement_t *compound_statement = &statement->compound; - construct_type_t *inner_types = NULL; + /* find last statement and use it's type */ + const statement_t *last_statement = NULL; + const statement_t *iter = compound_statement->statements; + for( ; iter != NULL; iter = iter->base.next) { + last_statement = iter; + } - switch(token.type) { - case T_IDENTIFIER: - if(declaration == NULL) { - parse_error("no identifier expected in typename"); - } else { - declaration->symbol = token.v.symbol; - declaration->source_position = token.source_position; - } - next_token(); - break; - case '(': - next_token(); - inner_types = parse_inner_declarator(declaration, may_be_abstract); - expect(')'); - break; - default: - if(may_be_abstract) - break; - parse_error_expected("problem while parsing declarator", T_IDENTIFIER, - '(', 0); + if(last_statement->type == STATEMENT_EXPRESSION) { + const expression_statement_t *expression_statement + = &last_statement->expression; + expression->base.datatype + = expression_statement->expression->base.datatype; + } else { + expression->base.datatype = type_void; } - while(true) { - construct_type_t *type; - switch(token.type) { - case '(': - type = parse_method_declarator(declaration); - break; - case '[': - type = parse_array_declarator(); - break; - default: - goto declarator_finished; - } + expect(')'); - if(last != NULL) { - last->next = type; - } else { - result = type; - } - last = type; - } + return expression; +} -declarator_finished: - parse_attributes(); +static expression_t *parse_brace_expression(void) +{ + eat('('); - if(inner_types != NULL) { - if(last != NULL) { - last->next = inner_types; - } else { - result = inner_types; + switch(token.type) { + case '{': + /* gcc extension: a stement expression */ + return parse_statement_expression(); + + TYPE_QUALIFIERS + TYPE_SPECIFIERS + return parse_cast(); + case T_IDENTIFIER: + if(is_typedef_symbol(token.v.symbol)) { + return parse_cast(); } - last = inner_types; } + expression_t *result = parse_expression(); + expect(')'); + return result; } -static type_t *construct_declarator_type(construct_type_t *construct_list, - type_t *type) +static expression_t *parse_function_keyword(void) { - 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; + next_token(); + /* TODO */ - switch(iter->type) { - case CONSTRUCT_METHOD: - construct_method_type = (construct_method_type_t*) iter; - method_type = construct_method_type->method_type; + if (current_function == NULL) { + parse_error("'__func__' used outside of a function"); + } - method_type->result_type = type; - type = (type_t*) method_type; - break; + string_literal_expression_t *expression + = allocate_ast_zero(sizeof(expression[0])); - case CONSTRUCT_POINTER: - parsed_pointer = (parsed_pointer_t*) iter; - pointer_type = allocate_type_zero(sizeof(pointer_type[0])); + expression->expression.type = EXPR_FUNCTION; + expression->expression.datatype = type_string; + expression->value = "TODO: FUNCTION"; - pointer_type->type.type = TYPE_POINTER; - pointer_type->points_to = type; - pointer_type->type.qualifiers = parsed_pointer->type_qualifiers; - type = (type_t*) pointer_type; - break; + return (expression_t*) expression; +} - case CONSTRUCT_ARRAY: - parsed_array = (parsed_array_t*) iter; - array_type = allocate_type_zero(sizeof(array_type[0])); - - array_type->type.type = TYPE_ARRAY; - array_type->element_type = type; - array_type->type.qualifiers = parsed_array->type_qualifiers; - array_type->is_static = parsed_array->is_static; - array_type->is_variable = parsed_array->is_variable; - array_type->size = parsed_array->size; - type = (type_t*) array_type; - break; - } +static expression_t *parse_pretty_function_keyword(void) +{ + eat(T___PRETTY_FUNCTION__); + /* TODO */ - type_t *hashed_type = typehash_insert((type_t*) type); - if(hashed_type != type) { - obstack_free(type_obst, type); - type = hashed_type; - } - } + string_literal_expression_t *expression + = allocate_ast_zero(sizeof(expression[0])); - return type; + expression->expression.type = EXPR_PRETTY_FUNCTION; + expression->expression.datatype = type_string; + expression->value = "TODO: PRETTY FUNCTION"; + + return (expression_t*) expression; } -static void parse_declarator(declaration_t *declaration, - storage_class_t storage_class, type_t *type, - int may_be_abstract) +static designator_t *parse_designator(void) { - construct_type_t *construct_type - = parse_inner_declarator(declaration, may_be_abstract); + designator_t *result = allocate_ast_zero(sizeof(result[0])); - declaration->type = construct_declarator_type(construct_type, type); - declaration->storage_class = storage_class; - if(construct_type != NULL) { - obstack_free(&temp_obst, construct_type); + if(token.type != T_IDENTIFIER) { + parse_error_expected("while parsing member designator", + T_IDENTIFIER, 0); + eat_paren(); + return NULL; } -} + result->symbol = token.v.symbol; + next_token(); -static type_t *parse_abstract_declarator(type_t *base_type) -{ - construct_type_t *construct_type - = parse_inner_declarator(NULL, 1); + designator_t *last_designator = result; + while(true) { + if(token.type == '.') { + next_token(); + if(token.type != T_IDENTIFIER) { + parse_error_expected("while parsing member designator", + T_IDENTIFIER, 0); + eat_paren(); + return NULL; + } + designator_t *designator = allocate_ast_zero(sizeof(result[0])); + designator->symbol = token.v.symbol; + next_token(); - if(construct_type == NULL) - return NULL; + last_designator->next = designator; + last_designator = designator; + continue; + } + if(token.type == '[') { + next_token(); + designator_t *designator = allocate_ast_zero(sizeof(result[0])); + designator->array_access = parse_expression(); + if(designator->array_access == NULL) { + eat_paren(); + return NULL; + } + expect(']'); - type_t *result = construct_declarator_type(construct_type, base_type); - obstack_free(&temp_obst, construct_type); + last_designator->next = designator; + last_designator = designator; + continue; + } + break; + } return result; } -static declaration_t *record_declaration(declaration_t *declaration) +static expression_t *parse_offsetof(void) { - if(context == NULL) - return declaration; + eat(T___builtin_offsetof); - symbol_t *symbol = declaration->symbol; - if(symbol != NULL) { - declaration_t *alias = environment_push(declaration, context); - if(alias != declaration) - return alias; - } + expression_t *expression = allocate_expression_zero(EXPR_OFFSETOF); + expression->base.datatype = type_size_t; - if(last_declaration != NULL) { - last_declaration->next = declaration; - } else { - context->declarations = declaration; - } - last_declaration = declaration; + expect('('); + expression->offsetofe.type = parse_typename(); + expect(','); + expression->offsetofe.designator = parse_designator(); + expect(')'); - return declaration; + return expression; } -static void parser_error_multiple_definition(declaration_t *previous, - declaration_t *declaration) +static expression_t *parse_va_start(void) { - 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"); - error(); + eat(T___builtin_va_start); + + expression_t *expression = allocate_expression_zero(EXPR_VA_START); + + expect('('); + expression->va_starte.ap = parse_assignment_expression(); + expect(','); + expression_t *const expr = parse_assignment_expression(); + if (expr->type == EXPR_REFERENCE) { + declaration_t *const decl = expr->reference.declaration; + if (decl->parent_context == ¤t_function->context && + decl->next == NULL) { + expression->va_starte.parameter = decl; + expect(')'); + return expression; + } + } + parser_print_error_prefix_pos(expr->base.source_position); + fprintf(stderr, "second argument of 'va_start' must be last parameter " + "of the current function\n"); + + return create_invalid_expression(); } -static void parse_init_declarators(const declaration_specifiers_t *specifiers) +static expression_t *parse_va_arg(void) { - while(true) { - declaration_t *ndeclaration - = allocate_ast_zero(sizeof(ndeclaration[0])); + eat(T___builtin_va_arg); - parse_declarator(ndeclaration, specifiers->storage_class, - specifiers->type, 0); - declaration_t *declaration = record_declaration(ndeclaration); - if(token.type == '=') { - next_token(); + expression_t *expression = allocate_expression_zero(EXPR_VA_ARG); - /* TODO: check that this is an allowed type (esp. no method type) */ + expect('('); + expression->va_arge.ap = parse_assignment_expression(); + expect(','); + expression->base.datatype = parse_typename(); + expect(')'); - if(declaration->initializer != NULL) { - parser_error_multiple_definition(declaration, ndeclaration); - } + return expression; +} - if(token.type == '{') { - // TODO - expect_void('}'); - } else { - declaration->initializer = parse_assignment_expression(); - } - } else if(token.type == '{') { - if(declaration->type->type != TYPE_METHOD) { - parser_print_error_prefix(); - fprintf(stderr, "Declarator "); - print_type_ext(declaration->type, declaration->symbol, NULL); - fprintf(stderr, " is not a method type.\n"); - } +static expression_t *parse_builtin_symbol(void) +{ + expression_t *expression = allocate_expression_zero(EXPR_BUILTIN_SYMBOL); - if(declaration->initializer != NULL) { - parser_error_multiple_definition(declaration, ndeclaration); - } - if(ndeclaration != declaration) { - memcpy(&declaration->context, &ndeclaration->context, - sizeof(declaration->context)); - } + symbol_t *symbol = token.v.symbol; - int top = environment_top(); - context_t *last_context = context; - set_context(&declaration->context); + expression->builtin_symbol.symbol = symbol; + next_token(); - /* push function parameters */ - declaration_t *parameter = declaration->context.declarations; - for( ; parameter != NULL; parameter = parameter->next) { - environment_push(parameter, context); - } + type_t *type = get_builtin_symbol_type(symbol); + type = automatic_type_conversion(type); - statement_t *statement = parse_compound_statement(); + expression->base.datatype = type; + return expression; +} - assert(context == &declaration->context); - set_context(last_context); - environment_pop_to(top); +static expression_t *parse_compare_builtin(void) +{ + expression_t *expression; - declaration->statement = statement; - return; - } + switch(token.type) { + case T___builtin_isgreater: + expression = allocate_expression_zero(EXPR_BINARY_ISGREATER); + break; + case T___builtin_isgreaterequal: + expression = allocate_expression_zero(EXPR_BINARY_ISGREATEREQUAL); + break; + case T___builtin_isless: + expression = allocate_expression_zero(EXPR_BINARY_ISLESS); + break; + case T___builtin_islessequal: + expression = allocate_expression_zero(EXPR_BINARY_ISLESSEQUAL); + break; + case T___builtin_islessgreater: + expression = allocate_expression_zero(EXPR_BINARY_ISLESSGREATER); + break; + case T___builtin_isunordered: + expression = allocate_expression_zero(EXPR_BINARY_ISUNORDERED); + break; + default: + panic("invalid compare builtin found"); + break; + } + next_token(); - if(token.type != ',') - break; - next_token(); + expect('('); + expression->binary.left = parse_assignment_expression(); + expect(','); + expression->binary.right = parse_assignment_expression(); + expect(')'); + + type_t *orig_type_left = expression->binary.left->base.datatype; + type_t *orig_type_right = expression->binary.right->base.datatype; + if(orig_type_left == NULL || orig_type_right == NULL) + return expression; + + type_t *type_left = skip_typeref(orig_type_left); + type_t *type_right = skip_typeref(orig_type_right); + if(!is_type_floating(type_left) && !is_type_floating(type_right)) { + type_error_incompatible("invalid operands in comparison", + token.source_position, type_left, type_right); + } else { + semantic_comparison(&expression->binary); } - expect_void(';'); + + return expression; } -static void parse_struct_declarators(const declaration_specifiers_t *specifiers) +static expression_t *parse_primary_expression(void) { - while(1) { - if(token.type == ':') { - next_token(); - parse_constant_expression(); - /* TODO (bitfields) */ - } else { - declaration_t *declaration - = allocate_ast_zero(sizeof(declaration[0])); - parse_declarator(declaration, specifiers->storage_class, - specifiers->type, 1); + switch(token.type) { + case T_INTEGER: + return parse_int_const(); + case T_FLOATINGPOINT: + return parse_float_const(); + case T_STRING_LITERAL: /* TODO merge */ + return parse_string_const(); + case T_WIDE_STRING_LITERAL: + return parse_wide_string_const(); + case T_IDENTIFIER: + return parse_reference(); + case T___FUNCTION__: + case T___func__: + return parse_function_keyword(); + case T___PRETTY_FUNCTION__: + return parse_pretty_function_keyword(); + case T___builtin_offsetof: + return parse_offsetof(); + case T___builtin_va_start: + return parse_va_start(); + case T___builtin_va_arg: + return parse_va_arg(); + case T___builtin_nanf: + case T___builtin_alloca: + case T___builtin_expect: + case T___builtin_va_end: + return parse_builtin_symbol(); + case T___builtin_isgreater: + case T___builtin_isgreaterequal: + case T___builtin_isless: + case T___builtin_islessequal: + case T___builtin_islessgreater: + case T___builtin_isunordered: + return parse_compare_builtin(); - /* TODO: check for doubled fields */ - record_declaration(declaration); + case '(': + return parse_brace_expression(); + } - if(token.type == ':') { - next_token(); - parse_constant_expression(); - /* TODO (bitfields) */ - } - } + parser_print_error_prefix(); + fprintf(stderr, "unexpected token "); + print_token(stderr, &token); + fprintf(stderr, "\n"); + eat_statement(); - if(token.type != ',') - break; - next_token(); - } - expect_void(';'); + return create_invalid_expression(); } -static void parse_compound_type_entries(void) +static expression_t *parse_array_expression(unsigned precedence, + expression_t *left) { - eat('{'); + (void) precedence; - 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); + eat('['); - parse_struct_declarators(&specifiers); + expression_t *inside = parse_expression(); + + array_access_expression_t *array_access + = allocate_ast_zero(sizeof(array_access[0])); + + array_access->expression.type = EXPR_ARRAY_ACCESS; + + type_t *type_left = left->base.datatype; + type_t *type_inside = inside->base.datatype; + type_t *return_type = NULL; + + if(type_left != NULL && type_inside != NULL) { + type_left = skip_typeref(type_left); + type_inside = skip_typeref(type_inside); + + if(is_type_pointer(type_left)) { + pointer_type_t *pointer = &type_left->pointer; + return_type = pointer->points_to; + array_access->array_ref = left; + array_access->index = inside; + } else if(is_type_pointer(type_inside)) { + pointer_type_t *pointer = &type_inside->pointer; + return_type = pointer->points_to; + array_access->array_ref = inside; + array_access->index = left; + array_access->flipped = true; + } else { + parser_print_error_prefix(); + fprintf(stderr, "array access on object with non-pointer types "); + print_type_quoted(type_left); + fprintf(stderr, ", "); + print_type_quoted(type_inside); + fprintf(stderr, "\n"); + } + } else { + array_access->array_ref = left; + array_access->index = inside; } - if(token.type == T_EOF) { - parse_error("unexpected error while parsing struct"); + + if(token.type != ']') { + parse_error_expected("Problem while parsing array access", ']', 0); + return (expression_t*) array_access; } next_token(); + + return_type = automatic_type_conversion(return_type); + array_access->expression.datatype = return_type; + + return (expression_t*) array_access; } -static void parse_declaration(void) +static expression_t *parse_sizeof(unsigned precedence) { - declaration_specifiers_t specifiers; - memset(&specifiers, 0, sizeof(specifiers)); - parse_declaration_specifiers(&specifiers); + eat(T_sizeof); - if(token.type == ';') { + sizeof_expression_t *sizeof_expression + = allocate_ast_zero(sizeof(sizeof_expression[0])); + sizeof_expression->expression.type = EXPR_SIZEOF; + sizeof_expression->expression.datatype = type_size_t; + + if(token.type == '(' && is_declaration_specifier(look_ahead(1), true)) { next_token(); - return; - } - parse_init_declarators(&specifiers); -} + sizeof_expression->type = parse_typename(); + expect(')'); + } else { + expression_t *expression = parse_sub_expression(precedence); + expression->base.datatype = revert_automatic_type_conversion(expression); -static type_t *parse_typename(void) -{ - declaration_specifiers_t specifiers; - memset(&specifiers, 0, sizeof(specifiers)); - parse_declaration_specifiers(&specifiers); - if(specifiers.storage_class != STORAGE_CLASS_NONE) { - /* TODO: improve error message, user does probably not know what a - * storage class is... - */ - parse_error("typename may not have a storage class"); + sizeof_expression->type = expression->base.datatype; + sizeof_expression->size_expression = expression; } - type_t *result = parse_abstract_declarator(specifiers.type); - - return result; + return (expression_t*) sizeof_expression; } +static expression_t *parse_select_expression(unsigned precedence, + expression_t *compound) +{ + (void) precedence; + assert(token.type == '.' || token.type == T_MINUSGREATER); + bool is_pointer = (token.type == T_MINUSGREATER); + next_token(); + expression_t *select = allocate_expression_zero(EXPR_SELECT); + select->select.compound = compound; -typedef expression_t* (*parse_expression_function) (unsigned precedence); -typedef expression_t* (*parse_expression_infix_function) (unsigned precedence, - expression_t *left); + if(token.type != T_IDENTIFIER) { + parse_error_expected("while parsing select", T_IDENTIFIER, 0); + return select; + } + symbol_t *symbol = token.v.symbol; + select->select.symbol = symbol; + next_token(); -typedef struct expression_parser_function_t expression_parser_function_t; -struct expression_parser_function_t { - unsigned precedence; - parse_expression_function parser; - unsigned infix_precedence; - parse_expression_infix_function infix_parser; -}; + type_t *orig_type = compound->base.datatype; + if(orig_type == NULL) + return create_invalid_expression(); -expression_parser_function_t expression_parsers[T_LAST_TOKEN]; + type_t *type = skip_typeref(orig_type); -static expression_t *expected_expression_error(void) -{ - parser_print_error_prefix(); - fprintf(stderr, "expected expression, got token "); - print_token(stderr, & token); - fprintf(stderr, "\n"); + type_t *type_left = type; + if(is_pointer) { + if(type->type != TYPE_POINTER) { + parser_print_error_prefix(); + fprintf(stderr, "left hand side of '->' is not a pointer, but "); + print_type_quoted(orig_type); + fputc('\n', stderr); + return create_invalid_expression(); + } + pointer_type_t *pointer_type = &type->pointer; + type_left = pointer_type->points_to; + } + type_left = skip_typeref(type_left); - expression_t *expression = allocate_ast_zero(sizeof(expression[0])); - expression->type = EXPR_INVALID; - next_token(); + if(type_left->type != TYPE_COMPOUND_STRUCT + && type_left->type != TYPE_COMPOUND_UNION) { + parser_print_error_prefix(); + fprintf(stderr, "request for member '%s' in something not a struct or " + "union, but ", symbol->string); + print_type_quoted(type_left); + fputc('\n', stderr); + return create_invalid_expression(); + } - return expression; -} + compound_type_t *compound_type = &type_left->compound; + declaration_t *declaration = compound_type->declaration; -static expression_t *parse_string_const(void) -{ - string_literal_t *cnst = allocate_ast_zero(sizeof(cnst[0])); + if(!declaration->init.is_defined) { + parser_print_error_prefix(); + fprintf(stderr, "request for member '%s' of incomplete type ", + symbol->string); + print_type_quoted(type_left); + fputc('\n', stderr); + return create_invalid_expression(); + } + + declaration_t *iter = declaration->context.declarations; + for( ; iter != NULL; iter = iter->next) { + if(iter->symbol == symbol) { + break; + } + } + if(iter == NULL) { + parser_print_error_prefix(); + print_type_quoted(type_left); + fprintf(stderr, " has no member named '%s'\n", symbol->string); + return create_invalid_expression(); + } - cnst->expression.type = EXPR_STRING_LITERAL; - cnst->expression.datatype = type_string; - cnst->value = parse_string_literals(); + /* we always do the auto-type conversions; the & and sizeof parser contains + * code to revert this! */ + type_t *expression_type = automatic_type_conversion(iter->type); - return (expression_t*) cnst; + select->select.compound_entry = iter; + select->base.datatype = expression_type; + return select; } -static expression_t *parse_int_const(void) +static expression_t *parse_call_expression(unsigned precedence, + expression_t *expression) { - const_t *cnst = allocate_ast_zero(sizeof(cnst[0])); + (void) precedence; + expression_t *result = allocate_expression_zero(EXPR_CALL); - cnst->expression.type = EXPR_CONST; - cnst->expression.datatype = type_int; - cnst->v.int_value = token.v.intvalue; + call_expression_t *call = &result->call; + call->function = expression; - next_token(); + function_type_t *function_type = NULL; + type_t *orig_type = expression->base.datatype; + if(orig_type != NULL) { + type_t *type = skip_typeref(orig_type); - return (expression_t*) cnst; -} + if(is_type_pointer(type)) { + pointer_type_t *pointer_type = &type->pointer; -static expression_t *parse_float_const(void) -{ - const_t *cnst = allocate_ast_zero(sizeof(cnst[0])); + type = skip_typeref(pointer_type->points_to); - cnst->expression.type = EXPR_CONST; - cnst->expression.datatype = type_double; - cnst->v.float_value = token.v.floatvalue; + if (is_type_function(type)) { + function_type = &type->function; + call->expression.datatype = function_type->return_type; + } + } + if(function_type == NULL) { + parser_print_error_prefix(); + fputs("called object '", stderr); + print_expression(expression); + fputs("' (type ", stderr); + print_type_quoted(orig_type); + fputs(") is not a pointer to a function\n", stderr); + + function_type = NULL; + call->expression.datatype = NULL; + } + } - next_token(); + /* parse arguments */ + eat('('); - return (expression_t*) cnst; -} + if(token.type != ')') { + call_argument_t *last_argument = NULL; -static declaration_t *create_implicit_function(symbol_t *symbol, - const source_position_t source_position) -{ - method_type_t *method_type = allocate_type_zero(sizeof(method_type)); + while(true) { + call_argument_t *argument = allocate_ast_zero(sizeof(argument[0])); - method_type->type.type = TYPE_METHOD; - method_type->result_type = type_int; - method_type->unspecified_parameters = true; + argument->expression = parse_assignment_expression(); + if(last_argument == NULL) { + call->arguments = argument; + } else { + last_argument->next = argument; + } + last_argument = argument; - type_t *type = typehash_insert((type_t*) method_type); - if(type != (type_t*) method_type) { - obstack_free(type_obst, method_type); + if(token.type != ',') + break; + next_token(); + } } + expect(')'); - declaration_t *declaration = allocate_ast_zero(sizeof(declaration[0])); + if(function_type != NULL) { + function_parameter_t *parameter = function_type->parameters; + call_argument_t *argument = call->arguments; + for( ; parameter != NULL && argument != NULL; + parameter = parameter->next, argument = argument->next) { + type_t *expected_type = parameter->type; + /* TODO report context in error messages */ + argument->expression = create_implicit_cast(argument->expression, + expected_type); + } + /* too few parameters */ + if(parameter != NULL) { + parser_print_error_prefix(); + fprintf(stderr, "too few arguments to function '"); + print_expression(expression); + fprintf(stderr, "'\n"); + } else if(argument != NULL) { + /* too many parameters */ + if(!function_type->variadic + && !function_type->unspecified_parameters) { + parser_print_error_prefix(); + fprintf(stderr, "too many arguments to function '"); + print_expression(expression); + fprintf(stderr, "'\n"); + } else { + /* do default promotion */ + for( ; argument != NULL; argument = argument->next) { + type_t *type = argument->expression->base.datatype; + + if(type == NULL) + continue; + + type = skip_typeref(type); + if(is_type_integer(type)) { + type = promote_integer(type); + } else if(type == type_float) { + type = type_double; + } + + argument->expression + = create_implicit_cast(argument->expression, type); + } + } + } + } - declaration->storage_class = STORAGE_CLASS_EXTERN; - declaration->type = type; - declaration->symbol = symbol; - declaration->source_position = source_position; + return result; +} - /* we have to violate the environment push/pop rules here and assign - * the declaration directly to its symbol and append it to the - * global context */ - assert(symbol->declaration == NULL); - symbol->declaration = declaration; - declaration->next = global_context->declarations; - global_context->declarations = declaration; +static type_t *semantic_arithmetic(type_t *type_left, type_t *type_right); + +static bool same_compound_type(const type_t *type1, const type_t *type2) +{ + if(!is_type_compound(type1)) + return false; + if(type1->type != type2->type) + return false; + + const compound_type_t *compound1 = &type1->compound; + const compound_type_t *compound2 = &type2->compound; - return declaration; + return compound1->declaration == compound2->declaration; } -static expression_t *parse_reference(void) +static expression_t *parse_conditional_expression(unsigned precedence, + expression_t *expression) { - reference_expression_t *ref = allocate_ast_zero(sizeof(ref[0])); - - ref->expression.type = EXPR_REFERENCE; - ref->symbol = token.v.symbol; + eat('?'); - declaration_t *declaration = ref->symbol->declaration; - source_position_t source_position = token.source_position; - next_token(); + expression_t *result = allocate_expression_zero(EXPR_CONDITIONAL); - 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); + conditional_expression_t *conditional = &result->conditional; + conditional->condition = expression; - 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; + /* 6.5.15.2 */ + type_t *condition_type_orig = expression->base.datatype; + if(condition_type_orig != NULL) { + type_t *condition_type = skip_typeref(condition_type_orig); + if(condition_type != NULL && !is_type_scalar(condition_type)) { + type_error("expected a scalar type in conditional condition", + expression->base.source_position, condition_type_orig); } } - ref->declaration = declaration; - ref->expression.datatype = declaration->type; + expression_t *true_expression = parse_expression(); + expect(':'); + expression_t *false_expression = parse_sub_expression(precedence); + + conditional->true_expression = true_expression; + conditional->false_expression = false_expression; + + type_t *orig_true_type = true_expression->base.datatype; + type_t *orig_false_type = false_expression->base.datatype; + if(orig_true_type == NULL || orig_false_type == NULL) + return result; + + type_t *true_type = skip_typeref(orig_true_type); + type_t *false_type = skip_typeref(orig_false_type); + + /* 6.5.15.3 */ + type_t *result_type = NULL; + if (is_type_arithmetic(true_type) && is_type_arithmetic(false_type)) { + result_type = semantic_arithmetic(true_type, false_type); + + true_expression = create_implicit_cast(true_expression, result_type); + false_expression = create_implicit_cast(false_expression, result_type); + + conditional->true_expression = true_expression; + conditional->false_expression = false_expression; + conditional->expression.datatype = result_type; + } else if (same_compound_type(true_type, false_type) + || (is_type_atomic(true_type, ATOMIC_TYPE_VOID) && + is_type_atomic(false_type, ATOMIC_TYPE_VOID))) { + /* just take 1 of the 2 types */ + result_type = true_type; + } else if (is_type_pointer(true_type) && is_type_pointer(false_type) + && pointers_compatible(true_type, false_type)) { + /* ok */ + result_type = true_type; + } else { + /* TODO */ + type_error_incompatible("while parsing conditional", + expression->base.source_position, true_type, + false_type); + } - return (expression_t*) ref; + conditional->expression.datatype = result_type; + return result; } -static void check_cast_allowed(expression_t *expression, type_t *dest_type) +static expression_t *parse_extension(unsigned precedence) { - (void) expression; - (void) dest_type; - /* TODO check if cast is allowed and issue warnings/errors */ + eat(T___extension__); + + /* TODO enable extensions */ + + return parse_sub_expression(precedence); } -static expression_t *parse_cast(void) +static expression_t *parse_builtin_classify_type(const unsigned precedence) { - 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; + eat(T___builtin_classify_type); - type_t *type = parse_typename(); + expression_t *result = allocate_expression_zero(EXPR_CLASSIFY_TYPE); + result->base.datatype = type_int; + expect('('); + expression_t *expression = parse_sub_expression(precedence); expect(')'); - expression_t *value = parse_sub_expression(20); - - check_cast_allowed(value, type); - - cast->expression.datatype = type; - cast->value = value; + result->classify_type.type_expression = expression; - return (expression_t*) cast; + return result; } -static expression_t *parse_statement_expression(void) +static void semantic_incdec(unary_expression_t *expression) { - statement_expression_t *expression - = allocate_ast_zero(sizeof(expression[0])); - expression->expression.type = EXPR_STATEMENT; - expression->statement = parse_compound_statement(); - - /* find last statement and use it's type */ - const statement_t *last_statement = NULL; - const statement_t *statement = expression->statement; - for( ; statement != NULL; statement = statement->next) { - last_statement = statement; - } + type_t *orig_type = expression->value->base.datatype; + if(orig_type == NULL) + return; - if(last_statement->type == STATEMENT_EXPRESSION) { - const expression_statement_t *expression_statement = - (const expression_statement_t*) last_statement; - expression->expression.datatype - = expression_statement->expression->datatype; - } else { - expression->expression.datatype = type_void; + type_t *type = skip_typeref(orig_type); + if(!is_type_arithmetic(type) && type->type != TYPE_POINTER) { + /* TODO: improve error message */ + parser_print_error_prefix(); + fprintf(stderr, "operation needs an arithmetic or pointer type\n"); + return; } - expect(')'); - - return (expression_t*) expression; + expression->expression.datatype = orig_type; } -static expression_t *parse_brace_expression(void) +static void semantic_unexpr_arithmetic(unary_expression_t *expression) { - eat('('); - - declaration_t *declaration; - switch(token.type) { - case '{': - /* gcc extension: a stement expression */ - return parse_statement_expression(); + type_t *orig_type = expression->value->base.datatype; + if(orig_type == NULL) + return; - 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(); - } + type_t *type = skip_typeref(orig_type); + if(!is_type_arithmetic(type)) { + /* TODO: improve error message */ + parser_print_error_prefix(); + fprintf(stderr, "operation needs an arithmetic type\n"); + return; } - expression_t *result = parse_expression(); - expect(')'); - - return result; + expression->expression.datatype = orig_type; } -static expression_t *parse_function_keyword(void) +static void semantic_unexpr_scalar(unary_expression_t *expression) { - eat(T___FUNCTION__); - /* TODO */ + type_t *orig_type = expression->value->base.datatype; + if(orig_type == NULL) + return; - string_literal_t *expression = allocate_ast_zero(sizeof(expression[0])); - expression->expression.type = EXPR_FUNCTION; - expression->expression.datatype = type_string; - expression->value = "TODO: FUNCTION"; + type_t *type = skip_typeref(orig_type); + if (!is_type_scalar(type)) { + parse_error("operand of ! must be of scalar type\n"); + return; + } - return (expression_t*) expression; + expression->expression.datatype = orig_type; } -static expression_t *parse_pretty_function_keyword(void) +static void semantic_unexpr_integer(unary_expression_t *expression) { - eat(T___PRETTY_FUNCTION__); - /* TODO */ + type_t *orig_type = expression->value->base.datatype; + if(orig_type == NULL) + return; - string_literal_t *expression = allocate_ast_zero(sizeof(expression[0])); - expression->expression.type = EXPR_PRETTY_FUNCTION; - expression->expression.datatype = type_string; - expression->value = "TODO: PRETTY FUNCTION"; + type_t *type = skip_typeref(orig_type); + if (!is_type_integer(type)) { + parse_error("operand of ~ must be of integer type\n"); + return; + } - return (expression_t*) expression; + expression->expression.datatype = orig_type; } -static member_designator_t *parse_member_designators(void) +static void semantic_dereference(unary_expression_t *expression) { - member_designator_t *result = allocate_ast_zero(sizeof(result[0])); + type_t *orig_type = expression->value->base.datatype; + if(orig_type == NULL) + return; - if(token.type != T_IDENTIFIER) { - parse_error_expected("problem while parsing member designator", - T_IDENTIFIER, 0); - eat_brace(); - return NULL; + type_t *type = skip_typeref(orig_type); + if(!is_type_pointer(type)) { + parser_print_error_prefix(); + fputs("Unary '*' needs pointer or arrray type, but type ", stderr); + print_type_quoted(orig_type); + fputs(" given.\n", stderr); + return; } - result->symbol = token.v.symbol; - next_token(); - member_designator_t *last_designator = result; - while(true) { - if(token.type == '.') { - next_token(); - if(token.type != T_IDENTIFIER) { - parse_error_expected("problem while parsing member designator", - T_IDENTIFIER, 0); - eat_brace(); - return NULL; - } - member_designator_t *designator - = allocate_ast_zero(sizeof(result[0])); - designator->symbol = token.v.symbol; - next_token(); + pointer_type_t *pointer_type = &type->pointer; + type_t *result_type = pointer_type->points_to; - last_designator->next = designator; - last_designator = designator; - continue; - } - if(token.type == '[') { - next_token(); - member_designator_t *designator - = allocate_ast_zero(sizeof(result[0])); - designator->array_access = parse_expression(); - if(designator->array_access == NULL) { - eat_brace(); - return NULL; - } - expect(']'); + result_type = automatic_type_conversion(result_type); + expression->expression.datatype = result_type; +} - last_designator->next = designator; - last_designator = designator; - continue; +static void semantic_take_addr(unary_expression_t *expression) +{ + expression_t *value = expression->value; + value->base.datatype = revert_automatic_type_conversion(value); + + type_t *orig_type = value->base.datatype; + if(orig_type == NULL) + return; + + if(value->type == EXPR_REFERENCE) { + reference_expression_t *reference = (reference_expression_t*) value; + declaration_t *declaration = reference->declaration; + if(declaration != NULL) { + declaration->address_taken = 1; } - break; } - return result; + expression->expression.datatype = make_pointer_type(orig_type, TYPE_QUALIFIER_NONE); } -static expression_t *parse_offsetof(void) +#define CREATE_UNARY_EXPRESSION_PARSER(token_type, unexpression_type, sfunc) \ +static expression_t *parse_##unexpression_type(unsigned precedence) \ +{ \ + eat(token_type); \ + \ + expression_t *unary_expression \ + = allocate_expression_zero(unexpression_type); \ + unary_expression->unary.value = parse_sub_expression(precedence); \ + \ + sfunc(&unary_expression->unary); \ + \ + return unary_expression; \ +} + +CREATE_UNARY_EXPRESSION_PARSER('-', EXPR_UNARY_NEGATE, + semantic_unexpr_arithmetic) +CREATE_UNARY_EXPRESSION_PARSER('+', EXPR_UNARY_PLUS, + semantic_unexpr_arithmetic) +CREATE_UNARY_EXPRESSION_PARSER('!', EXPR_UNARY_NOT, + semantic_unexpr_scalar) +CREATE_UNARY_EXPRESSION_PARSER('*', EXPR_UNARY_DEREFERENCE, + semantic_dereference) +CREATE_UNARY_EXPRESSION_PARSER('&', EXPR_UNARY_TAKE_ADDRESS, + semantic_take_addr) +CREATE_UNARY_EXPRESSION_PARSER('~', EXPR_UNARY_BITWISE_NEGATE, + semantic_unexpr_integer) +CREATE_UNARY_EXPRESSION_PARSER(T_PLUSPLUS, EXPR_UNARY_PREFIX_INCREMENT, + semantic_incdec) +CREATE_UNARY_EXPRESSION_PARSER(T_MINUSMINUS, EXPR_UNARY_PREFIX_DECREMENT, + semantic_incdec) + +#define CREATE_UNARY_POSTFIX_EXPRESSION_PARSER(token_type, unexpression_type, \ + sfunc) \ +static expression_t *parse_##unexpression_type(unsigned precedence, \ + expression_t *left) \ +{ \ + (void) precedence; \ + eat(token_type); \ + \ + expression_t *unary_expression \ + = allocate_expression_zero(unexpression_type); \ + unary_expression->unary.value = left; \ + \ + sfunc(&unary_expression->unary); \ + \ + return unary_expression; \ +} + +CREATE_UNARY_POSTFIX_EXPRESSION_PARSER(T_PLUSPLUS, + EXPR_UNARY_POSTFIX_INCREMENT, + semantic_incdec) +CREATE_UNARY_POSTFIX_EXPRESSION_PARSER(T_MINUSMINUS, + EXPR_UNARY_POSTFIX_DECREMENT, + semantic_incdec) + +static type_t *semantic_arithmetic(type_t *type_left, type_t *type_right) { - eat(T___builtin_offsetof); + /* TODO: handle complex + imaginary types */ + + /* § 6.3.1.8 Usual arithmetic conversions */ + if(type_left == type_long_double || type_right == type_long_double) { + return type_long_double; + } else if(type_left == type_double || type_right == type_double) { + return type_double; + } else if(type_left == type_float || type_right == type_float) { + return type_float; + } - offsetof_expression_t *expression - = allocate_ast_zero(sizeof(expression[0])); - expression->expression.type = EXPR_OFFSETOF; - expression->expression.datatype = type_size_t; + type_right = promote_integer(type_right); + type_left = promote_integer(type_left); - expect('('); - expression->type = parse_typename(); - expect(','); - expression->member_designators = parse_member_designators(); - expect(')'); + if(type_left == type_right) + return type_left; - return (expression_t*) expression; + bool signed_left = is_type_signed(type_left); + bool signed_right = is_type_signed(type_right); + int rank_left = get_rank(type_left); + int rank_right = get_rank(type_right); + if(rank_left < rank_right) { + if(signed_left == signed_right || !signed_right) { + return type_right; + } else { + return type_left; + } + } else { + if(signed_left == signed_right || !signed_left) { + return type_left; + } else { + return type_right; + } + } } -static expression_t *parse_builtin_symbol(void) +static void semantic_binexpr_arithmetic(binary_expression_t *expression) { - builtin_symbol_expression_t *expression - = allocate_ast_zero(sizeof(expression[0])); - expression->expression.type = EXPR_BUILTIN_SYMBOL; + expression_t *left = expression->left; + expression_t *right = expression->right; + type_t *orig_type_left = left->base.datatype; + type_t *orig_type_right = right->base.datatype; - /* TODO: set datatype */ + if(orig_type_left == NULL || orig_type_right == NULL) + return; - expression->symbol = token.v.symbol; + type_t *type_left = skip_typeref(orig_type_left); + type_t *type_right = skip_typeref(orig_type_right); - next_token(); + if(!is_type_arithmetic(type_left) || !is_type_arithmetic(type_right)) { + /* TODO: improve error message */ + parser_print_error_prefix(); + fprintf(stderr, "operation needs arithmetic types\n"); + return; + } - return (expression_t*) expression; + type_t *arithmetic_type = semantic_arithmetic(type_left, type_right); + expression->left = create_implicit_cast(left, arithmetic_type); + expression->right = create_implicit_cast(right, arithmetic_type); + expression->expression.datatype = arithmetic_type; } -static expression_t *parse_primary_expression(void) +static void semantic_shift_op(binary_expression_t *expression) { - 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: - return parse_reference(); - case T___FUNCTION__: - return parse_function_keyword(); - case T___PRETTY_FUNCTION__: - return parse_pretty_function_keyword(); - case T___builtin_offsetof: - return parse_offsetof(); - case T___builtin_expect: - case T___builtin_va_start: - case T___builtin_va_arg: - case T___builtin_va_end: - return parse_builtin_symbol(); + expression_t *left = expression->left; + expression_t *right = expression->right; + type_t *orig_type_left = left->base.datatype; + type_t *orig_type_right = right->base.datatype; - case '(': - return parse_brace_expression(); - } + if(orig_type_left == NULL || orig_type_right == NULL) + return; - parser_print_error_prefix(); - fprintf(stderr, "unexpected token "); - print_token(stderr, &token); - fprintf(stderr, "\n"); - eat_statement(); + type_t *type_left = skip_typeref(orig_type_left); + type_t *type_right = skip_typeref(orig_type_right); - expression_t *expression = allocate_ast_zero(sizeof(expression[0])); - expression->type = EXPR_INVALID; - expression->datatype = type_void; + if(!is_type_integer(type_left) || !is_type_integer(type_right)) { + /* TODO: improve error message */ + parser_print_error_prefix(); + fprintf(stderr, "operation needs integer types\n"); + return; + } - return expression; + type_left = promote_integer(type_left); + type_right = promote_integer(type_right); + + expression->left = create_implicit_cast(left, type_left); + expression->right = create_implicit_cast(right, type_right); + expression->expression.datatype = type_left; } -static expression_t *parse_array_expression(unsigned precedence, - expression_t *array_ref) +static void semantic_add(binary_expression_t *expression) { - (void) precedence; - - eat('['); - - array_access_expression_t *array_access - = allocate_ast_zero(sizeof(array_access[0])); + expression_t *left = expression->left; + expression_t *right = expression->right; + type_t *orig_type_left = left->base.datatype; + type_t *orig_type_right = right->base.datatype; - array_access->expression.type = EXPR_ARRAY_ACCESS; - array_access->array_ref = array_ref; - array_access->index = parse_expression(); + if(orig_type_left == NULL || orig_type_right == NULL) + return; - type_t *array_type = array_ref->datatype; - if(array_type != NULL) { - if(array_type->type == TYPE_POINTER) { - pointer_type_t *pointer = (pointer_type_t*) array_type; - array_access->expression.datatype = pointer->points_to; - } else { - parser_print_error_prefix(); - fprintf(stderr, "array access on object with non-pointer type "); - print_type(array_type); - fprintf(stderr, "\n"); - } - } + type_t *type_left = skip_typeref(orig_type_left); + type_t *type_right = skip_typeref(orig_type_right); - if(token.type != ']') { - parse_error_expected("Problem while parsing array access", ']', 0); - return (expression_t*) array_access; + /* § 5.6.5 */ + if(is_type_arithmetic(type_left) && is_type_arithmetic(type_right)) { + type_t *arithmetic_type = semantic_arithmetic(type_left, type_right); + expression->left = create_implicit_cast(left, arithmetic_type); + expression->right = create_implicit_cast(right, arithmetic_type); + expression->expression.datatype = arithmetic_type; + return; + } else if(is_type_pointer(type_left) && is_type_integer(type_right)) { + expression->expression.datatype = type_left; + } else if(is_type_pointer(type_right) && is_type_integer(type_left)) { + expression->expression.datatype = type_right; + } else { + parser_print_error_prefix(); + fprintf(stderr, "invalid operands to binary + ("); + print_type_quoted(orig_type_left); + fprintf(stderr, ", "); + print_type_quoted(orig_type_right); + fprintf(stderr, ")\n"); } - next_token(); - - return (expression_t*) array_access; } -static bool is_declaration_specifier(const token_t *token, - bool only_type_specifiers) +static void semantic_sub(binary_expression_t *expression) { - declaration_t *declaration; + expression_t *left = expression->left; + expression_t *right = expression->right; + type_t *orig_type_left = left->base.datatype; + type_t *orig_type_right = right->base.datatype; - 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; - STORAGE_CLASSES - TYPE_QUALIFIERS - if(only_type_specifiers) - return 0; - return 1; + if(orig_type_left == NULL || orig_type_right == NULL) + return; - default: - return 0; + type_t *type_left = skip_typeref(orig_type_left); + type_t *type_right = skip_typeref(orig_type_right); + + /* § 5.6.5 */ + if(is_type_arithmetic(type_left) && is_type_arithmetic(type_right)) { + type_t *arithmetic_type = semantic_arithmetic(type_left, type_right); + expression->left = create_implicit_cast(left, arithmetic_type); + expression->right = create_implicit_cast(right, arithmetic_type); + expression->expression.datatype = arithmetic_type; + return; + } else if(is_type_pointer(type_left) && is_type_integer(type_right)) { + expression->expression.datatype = type_left; + } else if(is_type_pointer(type_left) && is_type_pointer(type_right)) { + if(!pointers_compatible(type_left, type_right)) { + parser_print_error_prefix(); + fprintf(stderr, "pointers to incompatible objects to binary - ("); + print_type_quoted(orig_type_left); + fprintf(stderr, ", "); + print_type_quoted(orig_type_right); + fprintf(stderr, ")\n"); + } else { + expression->expression.datatype = type_ptrdiff_t; + } + } else { + parser_print_error_prefix(); + fprintf(stderr, "invalid operands to binary - ("); + print_type_quoted(orig_type_left); + fprintf(stderr, ", "); + print_type_quoted(orig_type_right); + fprintf(stderr, ")\n"); } } -static expression_t *parse_sizeof(unsigned precedence) +static void semantic_comparison(binary_expression_t *expression) { - eat(T_sizeof); + expression_t *left = expression->left; + expression_t *right = expression->right; + type_t *orig_type_left = left->base.datatype; + type_t *orig_type_right = right->base.datatype; - sizeof_expression_t *sizeof_expression - = allocate_ast_zero(sizeof(sizeof_expression[0])); - sizeof_expression->expression.type = EXPR_SIZEOF; - sizeof_expression->expression.datatype = type_size_t; + if(orig_type_left == NULL || orig_type_right == NULL) + return; - if(token.type == '(' && is_declaration_specifier(look_ahead(1), true)) { - next_token(); - sizeof_expression->type = parse_typename(); - expect(')'); + type_t *type_left = skip_typeref(orig_type_left); + type_t *type_right = skip_typeref(orig_type_right); + + /* TODO non-arithmetic types */ + if(is_type_arithmetic(type_left) && is_type_arithmetic(type_right)) { + type_t *arithmetic_type = semantic_arithmetic(type_left, type_right); + expression->left = create_implicit_cast(left, arithmetic_type); + expression->right = create_implicit_cast(right, arithmetic_type); + expression->expression.datatype = arithmetic_type; + } else if (is_type_pointer(type_left) && is_type_pointer(type_right)) { + /* TODO check compatibility */ + } else if (is_type_pointer(type_left)) { + expression->right = create_implicit_cast(right, type_left); + } else if (is_type_pointer(type_right)) { + expression->left = create_implicit_cast(left, type_right); } else { - expression_t *expression = parse_sub_expression(precedence); - sizeof_expression->type = expression->datatype; - sizeof_expression->size_expression = expression; + type_error_incompatible("invalid operands in comparison", + token.source_position, type_left, type_right); } - - return (expression_t*) sizeof_expression; + expression->expression.datatype = type_int; } -static expression_t *parse_select_expression(unsigned precedence, - expression_t *compound) +static void semantic_arithmetic_assign(binary_expression_t *expression) { - (void) precedence; - - assert(token.type == '.' || token.type == T_MINUSGREATER); - next_token(); - - select_expression_t *select = allocate_ast_zero(sizeof(select[0])); + expression_t *left = expression->left; + expression_t *right = expression->right; + type_t *orig_type_left = left->base.datatype; + type_t *orig_type_right = right->base.datatype; - select->expression.type = EXPR_SELECT; - select->compound = compound; + if(orig_type_left == NULL || orig_type_right == NULL) + return; - /* TODO: datatype */ + type_t *type_left = skip_typeref(orig_type_left); + type_t *type_right = skip_typeref(orig_type_right); - if(token.type != T_IDENTIFIER) { - parse_error_expected("Problem while parsing select", T_IDENTIFIER, 0); - return (expression_t*) select; + if(!is_type_arithmetic(type_left) || !is_type_arithmetic(type_right)) { + /* TODO: improve error message */ + parser_print_error_prefix(); + fprintf(stderr, "operation needs arithmetic types\n"); + return; } - select->symbol = token.v.symbol; - next_token(); - return (expression_t*) select; + /* combined instructions are tricky. We can't create an implicit cast on + * the left side, because we need the uncasted form for the store. + * The ast2firm pass has to know that left_type must be right_type + * for the arithmeitc operation and create a cast by itself */ + type_t *arithmetic_type = semantic_arithmetic(type_left, type_right); + expression->right = create_implicit_cast(right, arithmetic_type); + expression->expression.datatype = type_left; } -static expression_t *parse_call_expression(unsigned precedence, - expression_t *expression) +static void semantic_arithmetic_addsubb_assign(binary_expression_t *expression) { - (void) precedence; - call_expression_t *call = allocate_ast_zero(sizeof(call[0])); - - call->expression.type = EXPR_CALL; - call->method = expression; - - /* parse arguments */ - eat('('); + expression_t *left = expression->left; + expression_t *right = expression->right; + type_t *orig_type_left = left->base.datatype; + type_t *orig_type_right = right->base.datatype; - if(token.type != ')') { - call_argument_t *last_argument = NULL; + if(orig_type_left == NULL || orig_type_right == NULL) + return; - while(true) { - call_argument_t *argument = allocate_ast_zero(sizeof(argument[0])); + type_t *type_left = skip_typeref(orig_type_left); + type_t *type_right = skip_typeref(orig_type_right); + + if (is_type_arithmetic(type_left) && is_type_arithmetic(type_right)) { + /* combined instructions are tricky. We can't create an implicit cast on + * the left side, because we need the uncasted form for the store. + * The ast2firm pass has to know that left_type must be right_type + * for the arithmeitc operation and create a cast by itself */ + type_t *const arithmetic_type = semantic_arithmetic(type_left, type_right); + expression->right = create_implicit_cast(right, arithmetic_type); + expression->expression.datatype = type_left; + } else if (is_type_pointer(type_left) && is_type_integer(type_right)) { + expression->expression.datatype = type_left; + } else { + parser_print_error_prefix(); + fputs("Incompatible types ", stderr); + print_type_quoted(orig_type_left); + fputs(" and ", stderr); + print_type_quoted(orig_type_right); + fputs(" in assignment\n", stderr); + return; + } +} - argument->expression = parse_assignment_expression(); - if(last_argument == NULL) { - call->arguments = argument; - } else { - last_argument->next = argument; - } - last_argument = argument; +static void semantic_logical_op(binary_expression_t *expression) +{ + expression_t *left = expression->left; + expression_t *right = expression->right; + type_t *orig_type_left = left->base.datatype; + type_t *orig_type_right = right->base.datatype; - if(token.type != ',') - break; - next_token(); - } - } - expect(')'); + if(orig_type_left == NULL || orig_type_right == NULL) + return; - type_t *type = expression->datatype; - if(type != NULL) { - /* we can call pointer to function */ - if(type->type == TYPE_POINTER) { - pointer_type_t *pointer = (pointer_type_t*) type; - type = pointer->points_to; - } + type_t *type_left = skip_typeref(orig_type_left); + type_t *type_right = skip_typeref(orig_type_right); - if(type == NULL || type->type != TYPE_METHOD) { - parser_print_error_prefix(); - fprintf(stderr, "expected a method 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; - } + if (!is_type_scalar(type_left) || !is_type_scalar(type_right)) { + /* TODO: improve error message */ + parser_print_error_prefix(); + fprintf(stderr, "operation needs scalar types\n"); + return; } - return (expression_t*) call; -} - -static void type_error(const char *msg, const source_position_t source_position, - const type_t *type) -{ - parser_print_error_prefix_pos(source_position); - fprintf(stderr, "%s, but found type ", msg); - print_type(type); - fputc('\n', stderr); - error(); + expression->expression.datatype = type_int; } -static void type_error_incompatible(const char *msg, - const source_position_t source_position, const type_t *type1, - const type_t *type2) +static bool has_const_fields(type_t *type) { - parser_print_error_prefix_pos(source_position); - fprintf(stderr, "%s, incompatible types: ", msg); - print_type(type1); - fprintf(stderr, " - "); - print_type(type2); - fprintf(stderr, ")\n"); - error(); + (void) type; + /* TODO */ + return false; } -static type_t *get_type_after_conversion(const type_t *type1, - const type_t *type2) +static void semantic_binexpr_assign(binary_expression_t *expression) { - /* TODO... */ - (void) type2; - return (type_t*) type1; -} + expression_t *left = expression->left; + type_t *orig_type_left = left->base.datatype; -static expression_t *parse_conditional_expression(unsigned precedence, - expression_t *expression) -{ - eat('?'); + if(orig_type_left == NULL) + return; - conditional_expression_t *conditional - = allocate_ast_zero(sizeof(conditional[0])); - conditional->expression.type = EXPR_CONDITIONAL; - conditional->condition = expression; + type_t *type_left = revert_automatic_type_conversion(left); + type_left = skip_typeref(orig_type_left); - /* 6.5.15.2 */ - type_t *condition_type = conditional->condition->datatype; - if(condition_type != NULL) { - if(!is_type_scalar(condition_type)) { - type_error("expected a scalar type", expression->source_position, - condition_type); - } + /* must be a modifiable lvalue */ + if (is_type_array(type_left)) { + parser_print_error_prefix(); + fprintf(stderr, "Cannot assign to arrays ('"); + print_expression(left); + fprintf(stderr, "')\n"); + return; } - - conditional->true_expression = parse_expression(); - expect(':'); - conditional->false_expression = parse_sub_expression(precedence); - - type_t *true_type = conditional->true_expression->datatype; - if(true_type == NULL) - return (expression_t*) conditional; - type_t *false_type = conditional->false_expression->datatype; - if(false_type == NULL) - return (expression_t*) conditional; - - /* 6.4.15.3 */ - if(true_type == false_type) { - conditional->expression.datatype = true_type; - } else if(is_type_arithmetic(true_type) && is_type_arithmetic(false_type)) { - type_t *result = get_type_after_conversion(true_type, false_type); - /* TODO: create implicit convs if necessary */ - conditional->expression.datatype = result; - } else if(true_type->type == TYPE_POINTER && - false_type->type == TYPE_POINTER && - true /* TODO compatible points_to types */) { - /* TODO */ - } else if(/* (is_null_ptr_const(true_type) && false_type->type == TYPE_POINTER) - || (is_null_ptr_const(false_type) && - true_type->type == TYPE_POINTER) TODO*/ false) { - /* TODO */ - } else if(/* 1 is pointer to object type, other is void* */ false) { - /* TODO */ - } else { - type_error_incompatible("problem while parsing conditional", - expression->source_position, true_type, - false_type); + if(type_left->base.qualifiers & TYPE_QUALIFIER_CONST) { + parser_print_error_prefix(); + fprintf(stderr, "assignment to readonly location '"); + print_expression(left); + fprintf(stderr, "' (type "); + print_type_quoted(orig_type_left); + fprintf(stderr, ")\n"); + return; + } + if(is_type_incomplete(type_left)) { + parser_print_error_prefix(); + fprintf(stderr, "left-hand side of assignment '"); + print_expression(left); + fprintf(stderr, "' has incomplete type "); + print_type_quoted(orig_type_left); + fprintf(stderr, "\n"); + return; + } + if(is_type_compound(type_left) && has_const_fields(type_left)) { + parser_print_error_prefix(); + fprintf(stderr, "can't assign to '"); + print_expression(left); + fprintf(stderr, "' because compound type "); + print_type_quoted(orig_type_left); + fprintf(stderr, " has readonly fields\n"); + return; } - return (expression_t*) conditional; + semantic_assign(orig_type_left, &expression->right, "assignment"); + + expression->expression.datatype = orig_type_left; } -static expression_t *parse_extension(unsigned precedence) +static void semantic_comma(binary_expression_t *expression) { - eat(T___extension__); - - /* TODO enable extensions */ - - return parse_sub_expression(precedence); + expression->expression.datatype = expression->right->base.datatype; } -#define CREATE_UNARY_EXPRESSION_PARSER(token_type, unexpression_type) \ -static \ -expression_t *parse_##unexpression_type(unsigned precedence) \ +#define CREATE_BINEXPR_PARSER(token_type, binexpression_type, sfunc, lr) \ +static expression_t *parse_##binexpression_type(unsigned precedence, \ + expression_t *left) \ { \ eat(token_type); \ \ - unary_expression_t *unary_expression \ - = allocate_ast_zero(sizeof(unary_expression[0])); \ - unary_expression->expression.type = EXPR_UNARY; \ - unary_expression->type = unexpression_type; \ - unary_expression->value = parse_sub_expression(precedence); \ + expression_t *right = parse_sub_expression(precedence + lr); \ \ - return (expression_t*) unary_expression; \ -} - -CREATE_UNARY_EXPRESSION_PARSER('-', UNEXPR_NEGATE) -CREATE_UNARY_EXPRESSION_PARSER('+', UNEXPR_PLUS) -CREATE_UNARY_EXPRESSION_PARSER('!', UNEXPR_NOT) -CREATE_UNARY_EXPRESSION_PARSER('*', UNEXPR_DEREFERENCE) -CREATE_UNARY_EXPRESSION_PARSER('&', UNEXPR_TAKE_ADDRESS) -CREATE_UNARY_EXPRESSION_PARSER('~', UNEXPR_BITWISE_NEGATE) -CREATE_UNARY_EXPRESSION_PARSER(T_PLUSPLUS, UNEXPR_PREFIX_INCREMENT) -CREATE_UNARY_EXPRESSION_PARSER(T_MINUSMINUS, UNEXPR_PREFIX_DECREMENT) - -#define CREATE_UNARY_POSTFIX_EXPRESSION_PARSER(token_type, unexpression_type) \ -static \ -expression_t *parse_##unexpression_type(unsigned precedence, \ - expression_t *left) \ -{ \ - (void) precedence; \ - eat(token_type); \ - \ - unary_expression_t *unary_expression \ - = allocate_ast_zero(sizeof(unary_expression[0])); \ - unary_expression->expression.type = EXPR_UNARY; \ - unary_expression->type = unexpression_type; \ - unary_expression->value = left; \ - \ - return (expression_t*) unary_expression; \ -} - -CREATE_UNARY_POSTFIX_EXPRESSION_PARSER(T_PLUSPLUS, UNEXPR_POSTFIX_INCREMENT) -CREATE_UNARY_POSTFIX_EXPRESSION_PARSER(T_MINUSMINUS, UNEXPR_POSTFIX_DECREMENT) - -#define CREATE_BINEXPR_PARSER(token_type, binexpression_type) \ -static \ -expression_t *parse_##binexpression_type(unsigned precedence, \ - expression_t *left) \ -{ \ - eat(token_type); \ - \ - expression_t *right = parse_sub_expression(precedence); \ - \ - binary_expression_t *binexpr \ - = allocate_ast_zero(sizeof(binexpr[0])); \ - binexpr->expression.type = EXPR_BINARY; \ - binexpr->type = binexpression_type; \ - binexpr->left = left; \ - binexpr->right = right; \ - \ - return (expression_t*) binexpr; \ -} - -CREATE_BINEXPR_PARSER(',', BINEXPR_COMMA) -CREATE_BINEXPR_PARSER('*', BINEXPR_MUL) -CREATE_BINEXPR_PARSER('/', BINEXPR_DIV) -CREATE_BINEXPR_PARSER('%', BINEXPR_MOD) -CREATE_BINEXPR_PARSER('+', BINEXPR_ADD) -CREATE_BINEXPR_PARSER('-', BINEXPR_SUB) -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_EXCLAMATIONMARKEQUAL, BINEXPR_NOTEQUAL) -CREATE_BINEXPR_PARSER(T_LESSEQUAL, BINEXPR_LESSEQUAL) -CREATE_BINEXPR_PARSER(T_GREATEREQUAL, BINEXPR_GREATEREQUAL) -CREATE_BINEXPR_PARSER('&', BINEXPR_BITWISE_AND) -CREATE_BINEXPR_PARSER('|', BINEXPR_BITWISE_OR) -CREATE_BINEXPR_PARSER('^', BINEXPR_BITWISE_XOR) -CREATE_BINEXPR_PARSER(T_ANDAND, BINEXPR_LOGICAL_AND) -CREATE_BINEXPR_PARSER(T_PIPEPIPE, BINEXPR_LOGICAL_OR) -CREATE_BINEXPR_PARSER(T_LESSLESS, BINEXPR_SHIFTLEFT) -CREATE_BINEXPR_PARSER(T_GREATERGREATER, BINEXPR_SHIFTRIGHT) -CREATE_BINEXPR_PARSER(T_PLUSEQUAL, BINEXPR_ADD_ASSIGN) -CREATE_BINEXPR_PARSER(T_MINUSEQUAL, BINEXPR_SUB_ASSIGN) -CREATE_BINEXPR_PARSER(T_ASTERISKEQUAL, BINEXPR_MUL_ASSIGN) -CREATE_BINEXPR_PARSER(T_SLASHEQUAL, BINEXPR_DIV_ASSIGN) -CREATE_BINEXPR_PARSER(T_PERCENTEQUAL, BINEXPR_MOD_ASSIGN) -CREATE_BINEXPR_PARSER(T_LESSLESSEQUAL, BINEXPR_SHIFTLEFT_ASSIGN) -CREATE_BINEXPR_PARSER(T_GREATERGREATEREQUAL, BINEXPR_SHIFTRIGHT_ASSIGN) -CREATE_BINEXPR_PARSER(T_ANDEQUAL, BINEXPR_BITWISE_AND_ASSIGN) -CREATE_BINEXPR_PARSER(T_PIPEEQUAL, BINEXPR_BITWISE_OR_ASSIGN) -CREATE_BINEXPR_PARSER(T_CARETEQUAL, BINEXPR_BITWISE_XOR_ASSIGN) + expression_t *binexpr = allocate_expression_zero(binexpression_type); \ + binexpr->binary.left = left; \ + binexpr->binary.right = right; \ + sfunc(&binexpr->binary); \ + \ + return binexpr; \ +} + +CREATE_BINEXPR_PARSER(',', EXPR_BINARY_COMMA, semantic_comma, 1) +CREATE_BINEXPR_PARSER('*', EXPR_BINARY_MUL, semantic_binexpr_arithmetic, 1) +CREATE_BINEXPR_PARSER('/', EXPR_BINARY_DIV, semantic_binexpr_arithmetic, 1) +CREATE_BINEXPR_PARSER('%', EXPR_BINARY_MOD, semantic_binexpr_arithmetic, 1) +CREATE_BINEXPR_PARSER('+', EXPR_BINARY_ADD, semantic_add, 1) +CREATE_BINEXPR_PARSER('-', EXPR_BINARY_SUB, semantic_sub, 1) +CREATE_BINEXPR_PARSER('<', EXPR_BINARY_LESS, semantic_comparison, 1) +CREATE_BINEXPR_PARSER('>', EXPR_BINARY_GREATER, semantic_comparison, 1) +CREATE_BINEXPR_PARSER('=', EXPR_BINARY_ASSIGN, semantic_binexpr_assign, 0) + +CREATE_BINEXPR_PARSER(T_EQUALEQUAL, EXPR_BINARY_EQUAL, + semantic_comparison, 1) +CREATE_BINEXPR_PARSER(T_EXCLAMATIONMARKEQUAL, EXPR_BINARY_NOTEQUAL, + semantic_comparison, 1) +CREATE_BINEXPR_PARSER(T_LESSEQUAL, EXPR_BINARY_LESSEQUAL, + semantic_comparison, 1) +CREATE_BINEXPR_PARSER(T_GREATEREQUAL, EXPR_BINARY_GREATEREQUAL, + semantic_comparison, 1) + +CREATE_BINEXPR_PARSER('&', EXPR_BINARY_BITWISE_AND, + semantic_binexpr_arithmetic, 1) +CREATE_BINEXPR_PARSER('|', EXPR_BINARY_BITWISE_OR, + semantic_binexpr_arithmetic, 1) +CREATE_BINEXPR_PARSER('^', EXPR_BINARY_BITWISE_XOR, + semantic_binexpr_arithmetic, 1) +CREATE_BINEXPR_PARSER(T_ANDAND, EXPR_BINARY_LOGICAL_AND, + semantic_logical_op, 1) +CREATE_BINEXPR_PARSER(T_PIPEPIPE, EXPR_BINARY_LOGICAL_OR, + semantic_logical_op, 1) +CREATE_BINEXPR_PARSER(T_LESSLESS, EXPR_BINARY_SHIFTLEFT, + semantic_shift_op, 1) +CREATE_BINEXPR_PARSER(T_GREATERGREATER, EXPR_BINARY_SHIFTRIGHT, + semantic_shift_op, 1) +CREATE_BINEXPR_PARSER(T_PLUSEQUAL, EXPR_BINARY_ADD_ASSIGN, + semantic_arithmetic_addsubb_assign, 0) +CREATE_BINEXPR_PARSER(T_MINUSEQUAL, EXPR_BINARY_SUB_ASSIGN, + semantic_arithmetic_addsubb_assign, 0) +CREATE_BINEXPR_PARSER(T_ASTERISKEQUAL, EXPR_BINARY_MUL_ASSIGN, + semantic_arithmetic_assign, 0) +CREATE_BINEXPR_PARSER(T_SLASHEQUAL, EXPR_BINARY_DIV_ASSIGN, + semantic_arithmetic_assign, 0) +CREATE_BINEXPR_PARSER(T_PERCENTEQUAL, EXPR_BINARY_MOD_ASSIGN, + semantic_arithmetic_assign, 0) +CREATE_BINEXPR_PARSER(T_LESSLESSEQUAL, EXPR_BINARY_SHIFTLEFT_ASSIGN, + semantic_arithmetic_assign, 0) +CREATE_BINEXPR_PARSER(T_GREATERGREATEREQUAL, EXPR_BINARY_SHIFTRIGHT_ASSIGN, + semantic_arithmetic_assign, 0) +CREATE_BINEXPR_PARSER(T_ANDEQUAL, EXPR_BINARY_BITWISE_AND_ASSIGN, + semantic_arithmetic_assign, 0) +CREATE_BINEXPR_PARSER(T_PIPEEQUAL, EXPR_BINARY_BITWISE_OR_ASSIGN, + semantic_arithmetic_assign, 0) +CREATE_BINEXPR_PARSER(T_CARETEQUAL, EXPR_BINARY_BITWISE_XOR_ASSIGN, + semantic_arithmetic_assign, 0) static expression_t *parse_sub_expression(unsigned precedence) { @@ -2317,7 +4420,7 @@ static expression_t *parse_sub_expression(unsigned precedence) left = parse_primary_expression(); } assert(left != NULL); - left->source_position = source_position; + left->base.source_position = source_position; while(true) { if(token.type < 0) { @@ -2333,8 +4436,8 @@ static expression_t *parse_sub_expression(unsigned precedence) left = parser->infix_parser(parser->infix_precedence, left); assert(left != NULL); - assert(left->type != EXPR_INVALID); - left->source_position = source_position; + assert(left->type != EXPR_UNKNOWN); + left->base.source_position = source_position; } return left; @@ -2347,14 +4450,14 @@ static expression_t *parse_expression(void) -void register_expression_parser(parse_expression_function parser, - int token_type, unsigned precedence) +static void register_expression_parser(parse_expression_function parser, + int token_type, unsigned precedence) { expression_parser_function_t *entry = &expression_parsers[token_type]; if(entry->parser != NULL) { fprintf(stderr, "for token "); - print_token_type(stderr, token_type); + print_token_type(stderr, (token_type_t) token_type); fprintf(stderr, "\n"); panic("trying to register multiple expression parsers for a token"); } @@ -2362,14 +4465,14 @@ void register_expression_parser(parse_expression_function parser, entry->precedence = precedence; } -void register_expression_infix_parser(parse_expression_infix_function parser, - int token_type, unsigned precedence) +static void register_infix_parser(parse_expression_infix_function parser, + int token_type, unsigned precedence) { expression_parser_function_t *entry = &expression_parsers[token_type]; if(entry->infix_parser != NULL) { fprintf(stderr, "for token "); - print_token_type(stderr, token_type); + print_token_type(stderr, (token_type_t) token_type); fprintf(stderr, "\n"); panic("trying to register multiple infix expression parsers for a " "token"); @@ -2382,108 +4485,260 @@ static void init_expression_parsers(void) { memset(&expression_parsers, 0, sizeof(expression_parsers)); - register_expression_infix_parser(parse_BINEXPR_MUL, '*', 16); - register_expression_infix_parser(parse_BINEXPR_DIV, '/', 16); - register_expression_infix_parser(parse_BINEXPR_MOD, '%', 16); - register_expression_infix_parser(parse_BINEXPR_SHIFTLEFT, T_LESSLESS, 16); - register_expression_infix_parser(parse_BINEXPR_SHIFTRIGHT, - T_GREATERGREATER, 16); - register_expression_infix_parser(parse_BINEXPR_ADD, '+', 15); - register_expression_infix_parser(parse_BINEXPR_SUB, '-', 15); - register_expression_infix_parser(parse_BINEXPR_LESS, '<', 14); - register_expression_infix_parser(parse_BINEXPR_GREATER, '>', 14); - register_expression_infix_parser(parse_BINEXPR_LESSEQUAL, T_LESSEQUAL, 14); - register_expression_infix_parser(parse_BINEXPR_GREATEREQUAL, - T_GREATEREQUAL, 14); - register_expression_infix_parser(parse_BINEXPR_EQUAL, T_EQUALEQUAL, 13); - register_expression_infix_parser(parse_BINEXPR_NOTEQUAL, - T_EXCLAMATIONMARKEQUAL, 13); - register_expression_infix_parser(parse_BINEXPR_BITWISE_AND, '&', 12); - register_expression_infix_parser(parse_BINEXPR_BITWISE_XOR, '^', 11); - register_expression_infix_parser(parse_BINEXPR_BITWISE_OR, '|', 10); - register_expression_infix_parser(parse_BINEXPR_LOGICAL_AND, T_ANDAND, 9); - register_expression_infix_parser(parse_BINEXPR_LOGICAL_OR, T_PIPEPIPE, 8); - register_expression_infix_parser(parse_conditional_expression, '?', 7); - register_expression_infix_parser(parse_BINEXPR_ASSIGN, '=', 2); - register_expression_infix_parser(parse_BINEXPR_ADD_ASSIGN, T_PLUSEQUAL, 2); - register_expression_infix_parser(parse_BINEXPR_SUB_ASSIGN, T_MINUSEQUAL, 2); - register_expression_infix_parser(parse_BINEXPR_MUL_ASSIGN, - T_ASTERISKEQUAL, 2); - register_expression_infix_parser(parse_BINEXPR_DIV_ASSIGN, T_SLASHEQUAL, 2); - register_expression_infix_parser(parse_BINEXPR_MOD_ASSIGN, - T_PERCENTEQUAL, 2); - register_expression_infix_parser(parse_BINEXPR_SHIFTLEFT_ASSIGN, + register_infix_parser(parse_array_expression, '[', 30); + register_infix_parser(parse_call_expression, '(', 30); + register_infix_parser(parse_select_expression, '.', 30); + register_infix_parser(parse_select_expression, T_MINUSGREATER, 30); + register_infix_parser(parse_EXPR_UNARY_POSTFIX_INCREMENT, + T_PLUSPLUS, 30); + register_infix_parser(parse_EXPR_UNARY_POSTFIX_DECREMENT, + T_MINUSMINUS, 30); + + register_infix_parser(parse_EXPR_BINARY_MUL, '*', 16); + register_infix_parser(parse_EXPR_BINARY_DIV, '/', 16); + register_infix_parser(parse_EXPR_BINARY_MOD, '%', 16); + register_infix_parser(parse_EXPR_BINARY_SHIFTLEFT, T_LESSLESS, 16); + register_infix_parser(parse_EXPR_BINARY_SHIFTRIGHT, T_GREATERGREATER, 16); + register_infix_parser(parse_EXPR_BINARY_ADD, '+', 15); + register_infix_parser(parse_EXPR_BINARY_SUB, '-', 15); + register_infix_parser(parse_EXPR_BINARY_LESS, '<', 14); + register_infix_parser(parse_EXPR_BINARY_GREATER, '>', 14); + register_infix_parser(parse_EXPR_BINARY_LESSEQUAL, T_LESSEQUAL, 14); + register_infix_parser(parse_EXPR_BINARY_GREATEREQUAL, T_GREATEREQUAL, 14); + register_infix_parser(parse_EXPR_BINARY_EQUAL, T_EQUALEQUAL, 13); + register_infix_parser(parse_EXPR_BINARY_NOTEQUAL, + T_EXCLAMATIONMARKEQUAL, 13); + register_infix_parser(parse_EXPR_BINARY_BITWISE_AND, '&', 12); + register_infix_parser(parse_EXPR_BINARY_BITWISE_XOR, '^', 11); + register_infix_parser(parse_EXPR_BINARY_BITWISE_OR, '|', 10); + register_infix_parser(parse_EXPR_BINARY_LOGICAL_AND, T_ANDAND, 9); + register_infix_parser(parse_EXPR_BINARY_LOGICAL_OR, T_PIPEPIPE, 8); + register_infix_parser(parse_conditional_expression, '?', 7); + register_infix_parser(parse_EXPR_BINARY_ASSIGN, '=', 2); + register_infix_parser(parse_EXPR_BINARY_ADD_ASSIGN, T_PLUSEQUAL, 2); + register_infix_parser(parse_EXPR_BINARY_SUB_ASSIGN, T_MINUSEQUAL, 2); + register_infix_parser(parse_EXPR_BINARY_MUL_ASSIGN, T_ASTERISKEQUAL, 2); + register_infix_parser(parse_EXPR_BINARY_DIV_ASSIGN, T_SLASHEQUAL, 2); + register_infix_parser(parse_EXPR_BINARY_MOD_ASSIGN, T_PERCENTEQUAL, 2); + register_infix_parser(parse_EXPR_BINARY_SHIFTLEFT_ASSIGN, T_LESSLESSEQUAL, 2); - register_expression_infix_parser(parse_BINEXPR_SHIFTRIGHT_ASSIGN, + register_infix_parser(parse_EXPR_BINARY_SHIFTRIGHT_ASSIGN, T_GREATERGREATEREQUAL, 2); - register_expression_infix_parser(parse_BINEXPR_BITWISE_AND_ASSIGN, + register_infix_parser(parse_EXPR_BINARY_BITWISE_AND_ASSIGN, T_ANDEQUAL, 2); - register_expression_infix_parser(parse_BINEXPR_BITWISE_OR_ASSIGN, + register_infix_parser(parse_EXPR_BINARY_BITWISE_OR_ASSIGN, T_PIPEEQUAL, 2); - register_expression_infix_parser(parse_BINEXPR_BITWISE_XOR_ASSIGN, + register_infix_parser(parse_EXPR_BINARY_BITWISE_XOR_ASSIGN, T_CARETEQUAL, 2); - register_expression_infix_parser(parse_BINEXPR_COMMA, ',', 1); - - register_expression_infix_parser(parse_array_expression, '[', 30); - register_expression_infix_parser(parse_call_expression, '(', 30); - register_expression_infix_parser(parse_select_expression, '.', 30); - register_expression_infix_parser(parse_select_expression, - T_MINUSGREATER, 30); - register_expression_infix_parser(parse_UNEXPR_POSTFIX_INCREMENT, - T_PLUSPLUS, 30); - register_expression_infix_parser(parse_UNEXPR_POSTFIX_DECREMENT, - T_MINUSMINUS, 30); - - register_expression_parser(parse_UNEXPR_NEGATE, '-', 25); - register_expression_parser(parse_UNEXPR_PLUS, '+', 25); - register_expression_parser(parse_UNEXPR_NOT, '!', 25); - register_expression_parser(parse_UNEXPR_BITWISE_NEGATE, '~', 25); - register_expression_parser(parse_UNEXPR_DEREFERENCE, '*', 25); - register_expression_parser(parse_UNEXPR_TAKE_ADDRESS, '&', 25); - register_expression_parser(parse_UNEXPR_PREFIX_INCREMENT, T_PLUSPLUS, 25); - register_expression_parser(parse_UNEXPR_PREFIX_DECREMENT, T_MINUSMINUS, 25); + register_infix_parser(parse_EXPR_BINARY_COMMA, ',', 1); + + register_expression_parser(parse_EXPR_UNARY_NEGATE, '-', 25); + register_expression_parser(parse_EXPR_UNARY_PLUS, '+', 25); + register_expression_parser(parse_EXPR_UNARY_NOT, '!', 25); + register_expression_parser(parse_EXPR_UNARY_BITWISE_NEGATE, '~', 25); + register_expression_parser(parse_EXPR_UNARY_DEREFERENCE, '*', 25); + register_expression_parser(parse_EXPR_UNARY_TAKE_ADDRESS, '&', 25); + register_expression_parser(parse_EXPR_UNARY_PREFIX_INCREMENT, + T_PLUSPLUS, 25); + register_expression_parser(parse_EXPR_UNARY_PREFIX_DECREMENT, + T_MINUSMINUS, 25); register_expression_parser(parse_sizeof, T_sizeof, 25); register_expression_parser(parse_extension, T___extension__, 25); + register_expression_parser(parse_builtin_classify_type, + T___builtin_classify_type, 25); +} + +static asm_constraint_t *parse_asm_constraints(void) +{ + asm_constraint_t *result = NULL; + asm_constraint_t *last = NULL; + + while(token.type == T_STRING_LITERAL || token.type == '[') { + asm_constraint_t *constraint = allocate_ast_zero(sizeof(constraint[0])); + memset(constraint, 0, sizeof(constraint[0])); + + if(token.type == '[') { + eat('['); + if(token.type != T_IDENTIFIER) { + parse_error_expected("while parsing asm constraint", + T_IDENTIFIER, 0); + return NULL; + } + constraint->symbol = token.v.symbol; + + expect(']'); + } + + constraint->constraints = parse_string_literals(); + expect('('); + constraint->expression = parse_expression(); + expect(')'); + + if(last != NULL) { + last->next = constraint; + } else { + result = constraint; + } + last = constraint; + + if(token.type != ',') + break; + eat(','); + } + + return result; +} + +static asm_clobber_t *parse_asm_clobbers(void) +{ + asm_clobber_t *result = NULL; + asm_clobber_t *last = NULL; + + while(token.type == T_STRING_LITERAL) { + asm_clobber_t *clobber = allocate_ast_zero(sizeof(clobber[0])); + clobber->clobber = parse_string_literals(); + + if(last != NULL) { + last->next = clobber; + } else { + result = clobber; + } + last = clobber; + + if(token.type != ',') + break; + eat(','); + } + + return result; } +static statement_t *parse_asm_statement(void) +{ + eat(T_asm); + + statement_t *statement = allocate_statement_zero(STATEMENT_ASM); + statement->base.source_position = token.source_position; + + asm_statement_t *asm_statement = &statement->asms; + + if(token.type == T_volatile) { + next_token(); + asm_statement->is_volatile = true; + } + + expect('('); + asm_statement->asm_text = parse_string_literals(); + + if(token.type != ':') + goto end_of_asm; + eat(':'); + + asm_statement->inputs = parse_asm_constraints(); + if(token.type != ':') + goto end_of_asm; + eat(':'); + + asm_statement->outputs = parse_asm_constraints(); + if(token.type != ':') + goto end_of_asm; + eat(':'); + + asm_statement->clobbers = parse_asm_clobbers(); + +end_of_asm: + expect(')'); + expect(';'); + return statement; +} static statement_t *parse_case_statement(void) { eat(T_case); - case_label_statement_t *label = allocate_ast_zero(sizeof(label[0])); - label->statement.type = STATEMENT_CASE_LABEL; - label->statement.source_position = token.source_position; - label->expression = parse_expression(); + statement_t *statement = allocate_statement_zero(STATEMENT_CASE_LABEL); + + statement->base.source_position = token.source_position; + statement->case_label.expression = parse_expression(); expect(':'); - label->statement.next = parse_statement(); + statement->case_label.label_statement = parse_statement(); - return (statement_t*) label; + return statement; } static statement_t *parse_default_statement(void) { eat(T_default); - case_label_statement_t *label = allocate_ast_zero(sizeof(label[0])); - label->statement.type = STATEMENT_CASE_LABEL; - label->statement.source_position = token.source_position; + statement_t *statement = allocate_statement_zero(STATEMENT_CASE_LABEL); + + statement->base.source_position = token.source_position; expect(':'); - label->statement.next = parse_statement(); + statement->label.label_statement = parse_statement(); + + return statement; +} - return (statement_t*) label; +static declaration_t *get_label(symbol_t *symbol) +{ + declaration_t *candidate = get_declaration(symbol, NAMESPACE_LABEL); + assert(current_function != NULL); + /* if we found a label in the same function, then we already created the + * declaration */ + if(candidate != NULL + && candidate->parent_context == ¤t_function->context) { + return candidate; + } + + /* otherwise we need to create a new one */ + declaration_t *declaration = allocate_ast_zero(sizeof(declaration[0])); + declaration->namespc = NAMESPACE_LABEL; + declaration->symbol = symbol; + + label_push(declaration); + + return declaration; } static statement_t *parse_label_statement(void) { - eat(T_IDENTIFIER); + assert(token.type == T_IDENTIFIER); + symbol_t *symbol = token.v.symbol; + next_token(); + + declaration_t *label = get_label(symbol); + + /* if source position is already set then the label is defined twice, + * otherwise it was just mentioned in a goto so far */ + if(label->source_position.input_name != NULL) { + parser_print_error_prefix(); + fprintf(stderr, "duplicate label '%s'\n", symbol->string); + parser_print_error_prefix_pos(label->source_position); + fprintf(stderr, "previous definition of '%s' was here\n", + symbol->string); + } else { + label->source_position = token.source_position; + } + + label_statement_t *label_statement = allocate_ast_zero(sizeof(label[0])); + + label_statement->statement.type = STATEMENT_LABEL; + label_statement->statement.source_position = token.source_position; + label_statement->label = label; + expect(':'); - parse_statement(); - return NULL; + if(token.type == '}') { + parse_error("label at end of compound statement"); + return (statement_t*) label_statement; + } else { + label_statement->label_statement = parse_statement(); + } + + return (statement_t*) label_statement; } static statement_t *parse_if(void) @@ -2573,7 +4828,7 @@ static statement_t *parse_for(void) if(token.type != ';') { if(is_declaration_specifier(&token, false)) { - parse_declaration(); + parse_declaration(record_declaration); } else { statement->initialisation = parse_expression(); expect(';'); @@ -2602,10 +4857,27 @@ static statement_t *parse_for(void) static statement_t *parse_goto(void) { eat(T_goto); - expect(T_IDENTIFIER); + + if(token.type != T_IDENTIFIER) { + parse_error_expected("while parsing goto", T_IDENTIFIER, 0); + eat_statement(); + return NULL; + } + symbol_t *symbol = token.v.symbol; + next_token(); + + declaration_t *label = get_label(symbol); + + goto_statement_t *statement = allocate_ast_zero(sizeof(statement[0])); + + statement->statement.type = STATEMENT_GOTO; + statement->statement.source_position = token.source_position; + + statement->label = label; + expect(';'); - return NULL; + return (statement_t*) statement; } static statement_t *parse_continue(void) @@ -2613,9 +4885,9 @@ static statement_t *parse_continue(void) eat(T_continue); expect(';'); - statement_t *statement = allocate_ast_zero(sizeof(statement[0])); - statement->source_position = token.source_position; - statement->type = STATEMENT_CONTINUE; + statement_t *statement = allocate_ast_zero(sizeof(statement[0])); + statement->type = STATEMENT_CONTINUE; + statement->base.source_position = token.source_position; return statement; } @@ -2625,9 +4897,9 @@ static statement_t *parse_break(void) eat(T_break); expect(';'); - statement_t *statement = allocate_ast_zero(sizeof(statement[0])); - statement->source_position = token.source_position; - statement->type = STATEMENT_BREAK; + statement_t *statement = allocate_ast_zero(sizeof(statement[0])); + statement->type = STATEMENT_BREAK; + statement->base.source_position = token.source_position; return statement; } @@ -2640,63 +4912,88 @@ static statement_t *parse_return(void) statement->statement.type = STATEMENT_RETURN; statement->statement.source_position = token.source_position; + + assert(is_type_function(current_function->type)); + function_type_t *function_type = ¤t_function->type->function; + type_t *return_type = function_type->return_type; + + expression_t *return_value = NULL; if(token.type != ';') { - statement->return_value = parse_expression(); + return_value = parse_expression(); } expect(';'); + if(return_type == NULL) + return (statement_t*) statement; + if(return_value != NULL && return_value->base.datatype == NULL) + return (statement_t*) statement; + + return_type = skip_typeref(return_type); + + if(return_value != NULL) { + type_t *return_value_type = skip_typeref(return_value->base.datatype); + + if(is_type_atomic(return_type, ATOMIC_TYPE_VOID) + && !is_type_atomic(return_value_type, ATOMIC_TYPE_VOID)) { + parse_warning("'return' with a value, in function returning void"); + return_value = NULL; + } else { + if(return_type != NULL) { + semantic_assign(return_type, &return_value, "'return'"); + } + } + } else { + if(!is_type_atomic(return_type, ATOMIC_TYPE_VOID)) { + parse_warning("'return' without value, in function returning " + "non-void"); + } + } + statement->return_value = return_value; + return (statement_t*) statement; } static statement_t *parse_declaration_statement(void) { - declaration_t *before = last_declaration; - - declaration_statement_t *statement - = allocate_ast_zero(sizeof(statement[0])); - statement->statement.type = STATEMENT_DECLARATION; - statement->statement.source_position = token.source_position; + statement_t *statement = allocate_statement_zero(STATEMENT_DECLARATION); - declaration_specifiers_t specifiers; - memset(&specifiers, 0, sizeof(specifiers)); - parse_declaration_specifiers(&specifiers); + statement->base.source_position = token.source_position; - if(token.type == ';') { - eat(';'); - } else { - parse_init_declarators(&specifiers); - } + declaration_t *before = last_declaration; + parse_declaration(record_declaration); if(before == NULL) { - statement->declarations_begin = context->declarations; + statement->declaration.declarations_begin = context->declarations; } else { - statement->declarations_begin = before->next; + statement->declaration.declarations_begin = before->next; } - statement->declarations_end = last_declaration; + statement->declaration.declarations_end = last_declaration; - return (statement_t*) statement; + return statement; } static statement_t *parse_expression_statement(void) { - expression_statement_t *statement = allocate_ast_zero(sizeof(statement[0])); - statement->statement.type = STATEMENT_EXPRESSION; - statement->statement.source_position = token.source_position; + statement_t *statement = allocate_statement_zero(STATEMENT_EXPRESSION); - statement->expression = parse_expression(); + statement->base.source_position = token.source_position; + statement->expression.expression = parse_expression(); expect(';'); - return (statement_t*) statement; + return statement; } static statement_t *parse_statement(void) { - declaration_t *declaration; statement_t *statement = NULL; /* declaration or statement */ switch(token.type) { + case T_asm: + statement = parse_asm_statement(); + break; + case T_case: statement = parse_case_statement(); break; @@ -2756,9 +5053,7 @@ 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; } @@ -2784,20 +5079,21 @@ static statement_t *parse_statement(void) break; } - assert(statement == NULL || statement->source_position.input_name != NULL); + assert(statement == NULL + || statement->base.source_position.input_name != NULL); return statement; } static statement_t *parse_compound_statement(void) { - eat('{'); - compound_statement_t *compound_statement = allocate_ast_zero(sizeof(compound_statement[0])); compound_statement->statement.type = STATEMENT_COMPOUND; compound_statement->statement.source_position = token.source_position; + eat('{'); + int top = environment_top(); context_t *last_context = context; set_context(&compound_statement->context); @@ -2810,26 +5106,41 @@ static statement_t *parse_compound_statement(void) continue; if(last_statement != NULL) { - last_statement->next = statement; + last_statement->base.next = statement; } else { compound_statement->statements = statement; } - while(statement->next != NULL) - statement = statement->next; + while(statement->base.next != NULL) + statement = statement->base.next; last_statement = statement; } + if(token.type != '}') { + parser_print_error_prefix_pos( + compound_statement->statement.source_position); + fprintf(stderr, "end of file while looking for closing '}'\n"); + } + next_token(); + assert(context == &compound_statement->context); set_context(last_context); environment_pop_to(top); - next_token(); - return (statement_t*) compound_statement; } +static void initialize_builtins(void) +{ + type_wchar_t = make_global_typedef("__WCHAR_TYPE__", type_int); + type_wchar_t_ptr = make_pointer_type(type_wchar_t, TYPE_QUALIFIER_NONE); + type_size_t = make_global_typedef("__SIZE_TYPE__", + make_atomic_type(ATOMIC_TYPE_ULONG, TYPE_QUALIFIER_NONE)); + type_ptrdiff_t = make_global_typedef("__PTRDIFF_TYPE__", + make_atomic_type(ATOMIC_TYPE_LONG, TYPE_QUALIFIER_NONE)); +} + static translation_unit_t *parse_translation_unit(void) { translation_unit_t *unit = allocate_ast_zero(sizeof(unit[0])); @@ -2840,8 +5151,10 @@ static translation_unit_t *parse_translation_unit(void) assert(context == NULL); set_context(&unit->context); + initialize_builtins(); + while(token.type != T_EOF) { - parse_declaration(); + parse_external_declaration(); } assert(context == &unit->context); @@ -2856,10 +5169,12 @@ static translation_unit_t *parse_translation_unit(void) 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); + label_stack = NEW_ARR_F(stack_entry_t, 0); + found_error = false; type_set_output(stderr); + ast_set_output(stderr); lookahead_bufpos = 0; for(int i = 0; i < MAX_LOOKAHEAD + 2; ++i) { @@ -2868,7 +5183,10 @@ translation_unit_t *parse(void) translation_unit_t *unit = parse_translation_unit(); DEL_ARR_F(environment_stack); - obstack_free(&environment_obstack, NULL); + DEL_ARR_F(label_stack); + + if(found_error) + return NULL; return unit; } @@ -2878,12 +5196,19 @@ void init_parser(void) init_expression_parsers(); 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); - type_string = make_pointer_type(type_const_char, 0); + type_int = make_atomic_type(ATOMIC_TYPE_INT, TYPE_QUALIFIER_NONE); + type_long_double = make_atomic_type(ATOMIC_TYPE_LONG_DOUBLE, + TYPE_QUALIFIER_NONE); + type_double = make_atomic_type(ATOMIC_TYPE_DOUBLE, + TYPE_QUALIFIER_NONE); + type_float = make_atomic_type(ATOMIC_TYPE_FLOAT, TYPE_QUALIFIER_NONE); + type_char = make_atomic_type(ATOMIC_TYPE_CHAR, TYPE_QUALIFIER_NONE); + type_void = make_atomic_type(ATOMIC_TYPE_VOID, TYPE_QUALIFIER_NONE); + type_void_ptr = make_pointer_type(type_void, TYPE_QUALIFIER_NONE); + type_string = make_pointer_type(type_char, TYPE_QUALIFIER_NONE); + + symbol_t *const va_list_sym = symbol_table_insert("__builtin_va_list"); + type_valist = create_builtin_type(va_list_sym, type_void_ptr); } void exit_parser(void)