X-Git-Url: http://nsz.repo.hu/git/?a=blobdiff_plain;f=parser.c;h=144468797ab63cd3b893f641ecb20b3985ea242a;hb=57569ce27c909aa9cda1d4e1b405c83ae0ef29cb;hp=879b620f642c25bb4077885943029e083ea5c663;hpb=4eade33d45008f90c8952a65a92cdda17d9598f9;p=cparser diff --git a/parser.c b/parser.c index 879b620..1444687 100644 --- a/parser.c +++ b/parser.c @@ -14,31 +14,40 @@ #include "adt/error.h" #include "adt/array.h" -#define PRINT_TOKENS +//#define PRINT_TOKENS //#define ABORT_ON_ERROR #define MAX_LOOKAHEAD 2 //#define STRICT_C99 -struct environment_entry_t { - symbol_t *symbol; +typedef struct { declaration_t *old_declaration; - const void *old_context; -}; - -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 *context = NULL; -static declaration_t *last_declaration = NULL; -static struct obstack temp_obst; - -static type_t *type_int = 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; + symbol_t *symbol; + unsigned short namespc; +} stack_entry_t; + +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_uint = NULL; +static type_t *type_long_double = NULL; +static type_t *type_double = NULL; +static type_t *type_float = NULL; +static type_t *type_const_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_size_t = NULL; +static type_t *type_ptrdiff_t = NULL; static statement_t *parse_compound_statement(void); static statement_t *parse_statement(void); @@ -63,14 +72,10 @@ 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 @@ -115,6 +120,11 @@ static inline void *allocate_type_zero(size_t size) return res; } +static inline void free_type(void *type) +{ + obstack_free(type_obst, type); +} + /** * returns the top element of the environment stack */ @@ -123,6 +133,11 @@ static inline size_t environment_top(void) return ARR_LEN(environment_stack); } +static inline size_t label_top(void) +{ + return ARR_LEN(label_stack); +} + static inline void next_token(void) @@ -146,20 +161,17 @@ static inline const token_t *look_ahead(int num) 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); @@ -167,17 +179,17 @@ void parser_print_prefix_pos(const source_position_t source_position) 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) @@ -186,13 +198,25 @@ 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 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; @@ -222,6 +246,33 @@ 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 == '{') @@ -287,6 +338,14 @@ 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); \ @@ -299,16 +358,12 @@ static void set_context(context_t *new_context) { context = new_context; - declaration_t *declaration = new_context->declarations; - if(declaration != NULL) { - while(true) { - if(declaration->next == NULL) - break; - declaration = declaration->next; + last_declaration = new_context->declarations; + if(last_declaration != NULL) { + while(last_declaration->next != NULL) { + last_declaration = last_declaration->next; } } - - last_declaration = declaration; } /** @@ -322,339 +377,413 @@ static bool is_compatible_declaration (declaration_t *declaration, return declaration->type == previous->type; } +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 */ + assert(declaration->parent_context == NULL); + 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 { + const storage_class_t old_storage = previous_declaration->storage_class; + const storage_class_t new_storage = declaration->storage_class; + if (current_function == NULL) { + if (old_storage != STORAGE_CLASS_STATIC && + new_storage == 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 == STORAGE_CLASS_EXTERN) { + if (new_storage == 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 == STORAGE_CLASS_EXTERN && + new_storage == 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 == new_storage) { + 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 = 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]; - - symbol_t *symbol = entry->symbol; + for(i = top; i > new_top; --i) { + stack_entry_t *entry = & stack[i - 1]; - symbol->declaration = entry->old_declaration; - symbol->context = entry->old_context; + declaration_t *old_declaration = entry->old_declaration; + symbol_t *symbol = entry->symbol; + namespace_t namespc = (namespace_t)entry->namespc; - --i; - } while(i != new_top); - obstack_free(&environment_obstack, entry); + /* 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); + } + } - ARR_SHRINKLEN(environment_stack, (int) new_top); + ARR_SHRINKLEN(*stack_ptr, (int) new_top); } - - -static expression_t *parse_constant_expression(void) +static void environment_pop_to(size_t new_top) { - /* start parsing at precedence 7 (conditional expression) */ - return parse_sub_expression(7); + stack_pop_to(&environment_stack, new_top); } -static expression_t *parse_assignment_expression(void) +static void label_pop_to(size_t new_top) { - /* start parsing at precedence 2 (assignment expression) */ - return parse_sub_expression(2); + stack_pop_to(&label_stack, new_top); } -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); -typedef struct declaration_specifiers_t declaration_specifiers_t; -struct declaration_specifiers_t { - storage_class_t storage_class; - type_t *type; -}; +static int get_rank(const type_t *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); + atomic_type_t *atomic_type = (atomic_type_t*) type; + atomic_type_type_t atype = atomic_type->atype; + return atype; +} -static compound_type_t *find_compound_type(compound_type_t *types, - const symbol_t *symbol) +static type_t *promote_integer(type_t *type) { - compound_type_t *type = types; - for( ; type != NULL; type = type->next) { - if(type->symbol == symbol) - return type; - } + if(get_rank(type) < ATOMIC_TYPE_INT) + type = type_int; - return NULL; + return type; } -static type_t *parse_compound_type_specifier(bool is_struct) +static expression_t *create_cast_expression(expression_t *expression, + type_t *dest_type) { - if(is_struct) { - eat(T_struct); - } else { - eat(T_union); - } - - symbol_t *symbol = NULL; - compound_type_t *compound_type = NULL; + unary_expression_t *cast = allocate_ast_zero(sizeof(cast[0])); - if(token.type == T_IDENTIFIER) { - symbol = token.v.symbol; - next_token(); + cast->expression.type = EXPR_UNARY; + cast->type = UNEXPR_CAST; + cast->value = expression; + cast->expression.datatype = 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); - } - } - } 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 (expression_t*) cast; +} - return NULL; - } +static bool is_null_expression(const expression_t *const expr) +{ + if (expr->type != EXPR_CONST) return false; - if(compound_type == NULL) { - compound_type = allocate_type_zero(sizeof(compound_type[0])); + type_t *const type = skip_typeref(expr->datatype); + if (!is_type_integer(type)) return false; - 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; - } + const const_t *const const_expr = (const const_t*)expr; + return const_expr->v.int_value == 0; +} - 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; +static expression_t *create_implicit_cast(expression_t *expression, + type_t *dest_type) +{ + type_t *source_type = expression->datatype; - int top = environment_top(); - context_t *last_context = context; - set_context(&compound_type->context); + if(source_type == NULL) + return expression; - parse_compound_type_entries(); + source_type = skip_typeref(source_type); + dest_type = skip_typeref(dest_type); - assert(context == &compound_type->context); - set_context(last_context); - environment_pop_to(top); - } + if(source_type == dest_type) + return expression; - return (type_t*) compound_type; -} + if(dest_type->type == TYPE_ATOMIC) { + if(source_type->type != TYPE_ATOMIC) + panic("casting of non-atomic types not implemented yet"); -static void parse_enum_entries(void) -{ - eat('{'); + if(is_type_floating(dest_type) && !is_type_scalar(source_type)) { + type_error_incompatible("can't cast types", + expression->source_position, + source_type, dest_type); + return expression; + } - if(token.type == '}') { - next_token(); - parse_error("empty enum not allowed"); - return; + return create_cast_expression(expression, dest_type); } + if(dest_type->type == TYPE_POINTER) { + pointer_type_t *pointer_type + = (pointer_type_t*) dest_type; + switch (source_type->type) { + case TYPE_ATOMIC: + if (is_null_expression(expression)) { + return create_cast_expression(expression, dest_type); + } + break; - do { - declaration_t *entry = allocate_ast_zero(sizeof(entry[0])); + case TYPE_POINTER: + if (pointers_compatible(source_type, dest_type)) { + return create_cast_expression(expression, dest_type); + } + break; - if(token.type != T_IDENTIFIER) { - parse_error_expected("problem while parsing enum entry", - T_IDENTIFIER, 0); - eat_block(); - return; - } - entry->storage_class = STORAGE_CLASS_ENUM_ENTRY; - entry->symbol = token.v.symbol; - entry->source_position = token.source_position; - next_token(); + case TYPE_ARRAY: { + array_type_t *const array_type = (array_type_t*) source_type; + if (types_compatible(array_type->element_type, + pointer_type->points_to)) { + return create_cast_expression(expression, dest_type); + } + break; + } - if(token.type == '=') { - next_token(); - entry->initializer = parse_constant_expression(); + default: + panic("casting of non-atomic types not implemented yet"); } - record_declaration(entry); - - if(token.type != ',') - break; - next_token(); - } while(token.type != '}'); - - expect_void('}'); -} - -static enum_type_t *find_enum_type(enum_type_t *types, const symbol_t *symbol) -{ - enum_type_t *type = types; - for( ; type != NULL; type = type->next) { - if(type->symbol == symbol) - return type; + type_error_incompatible("can't implicitly cast types", + expression->source_position, + source_type, dest_type); + return expression; } - return NULL; + panic("casting of non-atomic types not implemented yet"); } -static type_t *parse_enum_specifier(void) +static void semantic_assign(type_t *orig_type_left, expression_t **right, + const char *context) { - eat(T_enum); + type_t *orig_type_right = (*right)->datatype; - symbol_t *symbol = NULL; - enum_type_t *enum_type = NULL; + if(orig_type_right == NULL) + return; - if(token.type == T_IDENTIFIER) { - symbol = token.v.symbol; - next_token(); + type_t *const type_left = skip_typeref(orig_type_left); + type_t *const type_right = skip_typeref(orig_type_right); - 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 (type_left == type_right) { + return; } - 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 ((is_type_arithmetic(type_left) && is_type_arithmetic(type_right)) || + (type_left->type == TYPE_POINTER && is_null_expression(*right)) || + (type_left->type == TYPE_POINTER && type_right->type == TYPE_POINTER)) { + *right = create_implicit_cast(*right, type_left); + return; } - 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; + if (type_left->type == TYPE_POINTER) { + switch (type_right->type) { + case TYPE_FUNCTION: { + pointer_type_t *const ptr_type = (pointer_type_t*)type_left; + if (ptr_type->points_to == type_right) { + return; + } + break; + } - parse_enum_entries(); + case TYPE_ARRAY: { + pointer_type_t *const ptr_type = (pointer_type_t*)type_left; + array_type_t *const arr_type = (array_type_t*)type_right; + if (ptr_type->points_to == arr_type->element_type) { + return; + } + break; + } - if(before == NULL) { - enum_type->entries_begin = context->declarations; - } else { - enum_type->entries_begin = before->next; + default: break; } - enum_type->entries_end = last_declaration; } - return (type_t*) enum_type; + /* TODO: improve error message */ + parser_print_error_prefix(); + fprintf(stderr, "incompatible types in %s\n", context); + parser_print_error_prefix(); + print_type_quoted(type_left); + fputs(" <- ", stderr); + print_type_quoted(type_right); + fputs("\n", stderr); } -static type_t *parse_typeof(void) +static expression_t *parse_constant_expression(void) { - eat(T___typeof__); + /* start parsing at precedence 7 (conditional expression) */ + return parse_sub_expression(7); +} - type_t *result; +static expression_t *parse_assignment_expression(void) +{ + /* start parsing at precedence 2 (assignment expression) */ + return parse_sub_expression(2); +} - expect('('); +typedef struct declaration_specifiers_t declaration_specifiers_t; +struct declaration_specifiers_t { + storage_class_t storage_class; + bool is_inline; + type_t *type; +}; - 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; - - default: - expression = parse_expression(); - result = expression->datatype; - break; - } - - expect(')'); - - return result; -} +static void parse_compound_type_entries(void); +static declaration_t *parse_declarator( + const declaration_specifiers_t *specifiers, type_t *type, + bool may_be_abstract); +static declaration_t *record_declaration(declaration_t *declaration); static const char *parse_string_literals(void) { @@ -720,191 +849,729 @@ 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_expression(type_t *type, + expression_t *expression) { - builtin_type_t *type = allocate_type_zero(sizeof(type[0])); - type->type.type = TYPE_BUILTIN; - type->symbol = symbol; + initializer_value_t *result = allocate_ast_zero(sizeof(result[0])); - type_t *result = typehash_insert((type_t*) type); - if(result != (type_t*) type) { - obstack_free(type_obst, type); + /* TODO check that expression is a constant expression */ + + /* § 6.7.8.14/15 char array may be initialized by string literals */ + if(type->type == TYPE_ARRAY && expression->type == EXPR_STRING_LITERAL) { + array_type_t *array_type = (array_type_t*) type; + type_t *element_type = array_type->element_type; + + if(element_type->type == TYPE_ATOMIC) { + atomic_type_t *atomic_type = (atomic_type_t*) element_type; + atomic_type_type_t atype = atomic_type->atype; + + /* TODO handle wide strings */ + if(atype == ATOMIC_TYPE_CHAR + || atype == ATOMIC_TYPE_SCHAR + || atype == ATOMIC_TYPE_UCHAR) { + /* it's fine TODO: check for length of string array... */ + goto initializer_from_expression_finished; + } + } } - return result; + semantic_assign(type, &expression, "initializer"); + +initializer_from_expression_finished: + result->initializer.type = INITIALIZER_VALUE; + result->value = expression; + + return (initializer_t*) result; } -static void parse_declaration_specifiers(declaration_specifiers_t *specifiers) +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) { - declaration_t *declaration; - type_t *type = NULL; - unsigned type_qualifiers = 0; - unsigned type_specifiers = 0; - int newtype = 0; + if(token.type == '{') { + return parse_sub_initializer(type, NULL, NULL); + } - while(true) { - switch(token.type) { + expression_t *expression = parse_assignment_expression(); + type_t *expression_type = skip_typeref(expression->datatype); - /* 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 parse_sub_initializer(type, expression, expression_type); +} - 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 bool had_initializer_brace_warning; - /* type qualifiers */ -#define MATCH_TYPE_QUALIFIER(token, qualifier) \ - case token: \ - type_qualifiers |= qualifier; \ - next_token(); \ - break; +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(!had_initializer_brace_warning) { + parse_warning("braces around scalar initializer"); + had_initializer_brace_warning = true; + } + initializer_t *result = parse_sub_initializer(type, NULL, NULL); + if(token.type == ',') { + next_token(); + /* TODO: warn about excessive elements */ + } + expect_block('}'); + return result; + } - 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); + if(expression == NULL) { + expression = parse_assignment_expression(); + } + return initializer_from_expression(type, expression); + } - case T___extension__: - /* TODO */ - next_token(); - break; + /* TODO: ignore qualifiers, comparing pointers is probably + * not correct */ + if(expression != NULL && expression_type == type) { + initializer_value_t *result = allocate_ast_zero(sizeof(result[0])); + result->initializer.type = INITIALIZER_VALUE; - /* 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; + if(type != NULL) { + semantic_assign(type, &expression, "initializer"); + } + result->value = expression; - 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: - 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; + return (initializer_t*) result; + } + + bool read_paren = false; + if(token.type == '{') { + next_token(); + read_paren = true; + } + + /* descend into subtype */ + initializer_t *result = NULL; + initializer_t **elems; + if(type->type == TYPE_ARRAY) { + array_type_t *array_type = (array_type_t*) type; + 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(','); + + initializer_t *sub + = parse_sub_initializer(element_type, NULL, NULL); + if(sub == NULL) { + /* TODO error, do nicer cleanup */ + parse_error("member initializer didn't match"); + DEL_ARR_F(elems); + return NULL; } - break; + ARR_APP1(initializer_t*, elems, sub); + } + } else { + assert(type->type == TYPE_COMPOUND_STRUCT + || type->type == TYPE_COMPOUND_UNION); + compound_type_t *compound_type = (compound_type_t*) type; + context_t *context = & compound_type->declaration->context; - /* 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; + declaration_t *first = context->declarations; + if(first == NULL) + return NULL; + type_t *first_type = first->type; + first_type = skip_typeref(first_type); - case T___attribute__: - /* TODO */ - parse_attributes(); - break; + 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); + } - case T_IDENTIFIER: - declaration = token.v.symbol->declaration; - if(declaration == NULL || - declaration->storage_class != STORAGE_CLASS_TYPEDEF) { - goto finish_specifiers; + /* didn't match the subtypes -> try our parent type */ + if(sub == NULL) { + assert(!read_paren); + return NULL; + } + + elems = NEW_ARR_F(initializer_t*, 0); + ARR_APP1(initializer_t*, elems, sub); + + 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(','); + + type_t *iter_type = iter->type; + iter_type = skip_typeref(iter_type); + + initializer_t *sub = parse_sub_initializer(iter_type, NULL, NULL); + 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; - type = declaration->type; - assert(type != NULL); + 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(); - break; + expect('}'); + } + return result; +} - /* function specifier */ - default: - goto finish_specifiers; - } +static initializer_t *parse_initializer(type_t *type) +{ + initializer_t *result; + + type = skip_typeref(type); + + if(token.type != '{') { + expression_t *expression = parse_assignment_expression(); + return initializer_from_expression(type, expression); } -finish_specifiers: + if(is_type_scalar(type)) { + /* § 6.7.8.11 */ + eat('{'); - if(type == NULL) { - atomic_type_type_t atomic_type; + expression_t *expression = parse_assignment_expression(); + result = initializer_from_expression(type, expression); - /* 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; - case SPECIFIER_UNSIGNED | SPECIFIER_CHAR: - atomic_type = ATOMIC_TYPE_UCHAR; - break; - case SPECIFIER_SHORT: - case SPECIFIER_SIGNED | SPECIFIER_SHORT: - case SPECIFIER_SHORT | SPECIFIER_INT: - case SPECIFIER_SIGNED | SPECIFIER_SHORT | SPECIFIER_INT: - atomic_type = ATOMIC_TYPE_SHORT; - break; - case SPECIFIER_UNSIGNED | SPECIFIER_SHORT: - case SPECIFIER_UNSIGNED | SPECIFIER_SHORT | SPECIFIER_INT: - atomic_type = ATOMIC_TYPE_USHORT; + 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_type_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(void) +{ + 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->symbol = token.v.symbol; + entry->source_position = token.source_position; + next_token(); + + if(token.type == '=') { + next_token(); + entry->init.enum_value = parse_constant_expression(); + } + + record_declaration(entry); + + if(token.type != ',') + break; + next_token(); + } while(token.type != '}'); + + expect_void('}'); +} + +static declaration_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_type_zero(sizeof(declaration[0])); + + declaration->namespc = NAMESPACE_ENUM; + declaration->source_position = token.source_position; + declaration->symbol = symbol; + } + + 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(); + parse_attributes(); + } + + return declaration; +} + +/** + * 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->datatype; + } + break; + + TYPENAME_START + type = parse_typename(); + break; + + default: + expression = parse_expression(); + type = expression->datatype; + break; + } + + expect(')'); + + typeof_type_t *typeof = allocate_type_zero(sizeof(typeof[0])); + typeof->type.type = TYPE_TYPEOF; + typeof->expression = expression; + typeof->typeof_type = type; + + return (type_t*) typeof; +} + +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 *symbol) +{ + builtin_type_t *type = allocate_type_zero(sizeof(type[0])); + type->type.type = TYPE_BUILTIN; + type->symbol = symbol; + /* TODO... */ + type->real_type = type_int; + + return (type_t*) type; +} + +static type_t *get_typedef_type(symbol_t *symbol) +{ + declaration_t *declaration = get_declaration(symbol, NAMESPACE_NORMAL); + if(declaration == NULL + || declaration->storage_class != STORAGE_CLASS_TYPEDEF) + return NULL; + + typedef_type_t *typedef_type = allocate_type_zero(sizeof(typedef_type[0])); + typedef_type->type.type = TYPE_TYPEDEF; + typedef_type->declaration = declaration; + + return (type_t*) typedef_type; +} + +static void parse_declaration_specifiers(declaration_specifiers_t *specifiers) +{ + type_t *type = NULL; + unsigned type_qualifiers = 0; + unsigned type_specifiers = 0; + int newtype = 0; + + 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) + + /* 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: { + compound_type_t *compound_type + = allocate_type_zero(sizeof(compound_type[0])); + compound_type->type.type = TYPE_COMPOUND_STRUCT; + compound_type->declaration = parse_compound_type_specifier(true); + + type = (type_t*) compound_type; + break; + } + case T_union: { + compound_type_t *compound_type + = allocate_type_zero(sizeof(compound_type[0])); + compound_type->type.type = TYPE_COMPOUND_UNION; + compound_type->declaration = parse_compound_type_specifier(false); + + type = (type_t*) compound_type; + break; + } + case T_enum: { + enum_type_t *enum_type = allocate_type_zero(sizeof(enum_type[0])); + enum_type->type.type = TYPE_ENUM; + enum_type->declaration = parse_enum_specifier(); + + type = (type_t*) enum_type; + break; + } + case T___typeof__: + type = parse_typeof(); + break; + case T___builtin_va_list: + type = create_builtin_type(token.v.symbol); + 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; + case SPECIFIER_UNSIGNED | SPECIFIER_CHAR: + atomic_type = ATOMIC_TYPE_UCHAR; + break; + case SPECIFIER_SHORT: + case SPECIFIER_SIGNED | SPECIFIER_SHORT: + case SPECIFIER_SHORT | SPECIFIER_INT: + case SPECIFIER_SIGNED | SPECIFIER_SHORT | SPECIFIER_INT: + atomic_type = ATOMIC_TYPE_SHORT; + break; + case SPECIFIER_UNSIGNED | SPECIFIER_SHORT: + case SPECIFIER_UNSIGNED | SPECIFIER_SHORT | SPECIFIER_INT: + atomic_type = ATOMIC_TYPE_USHORT; break; case SPECIFIER_INT: case SPECIFIER_SIGNED: @@ -959,8 +1626,6 @@ finish_specifiers: case SPECIFIER_LONG | SPECIFIER_DOUBLE | SPECIFIER_COMPLEX: atomic_type = ATOMIC_TYPE_LONG_DOUBLE_COMPLEX; break; -#endif -#ifdef PROVIDE_IMAGINARY case SPECIFIER_FLOAT | SPECIFIER_IMAGINARY: atomic_type = ATOMIC_TYPE_FLOAT_IMAGINARY; break; @@ -1004,19 +1669,19 @@ finish_specifiers: } } - type->qualifiers = type_qualifiers; + type->qualifiers = (type_qualifier_t)type_qualifiers; type_t *result = typehash_insert(type); if(newtype && result != (type_t*) type) { - obstack_free(type_obst, type); + free_type(type); } specifiers->type = result; } -static type_qualifier_t parse_type_qualifiers(void) +static unsigned parse_type_qualifiers(void) { - type_qualifier_t type_qualifiers = 0; + unsigned type_qualifiers = TYPE_QUALIFIER_NONE; while(true) { switch(token.type) { @@ -1024,7 +1689,6 @@ static type_qualifier_t parse_type_qualifiers(void) 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); default: return type_qualifiers; @@ -1036,8 +1700,8 @@ static void parse_identifier_list(void) { while(true) { if(token.type != T_IDENTIFIER) { - parse_error_expected("problem while parsing parameter identifier " - "list", T_IDENTIFIER, 0); + parse_error_expected("while parsing parameter identifier list", + T_IDENTIFIER, 0); return; } next_token(); @@ -1054,21 +1718,32 @@ static declaration_t *parse_parameter(void) parse_declaration_specifiers(&specifiers); - declaration_t *declaration = allocate_ast_zero(sizeof(declaration[0])); - parse_declarator(declaration, specifiers.storage_class, - specifiers.type, 1); + declaration_t *declaration + = parse_declarator(&specifiers, specifiers.type, true); + + /* TODO check declaration constraints for parameters */ + if(declaration->storage_class == STORAGE_CLASS_TYPEDEF) { + parse_error("typedef not allowed in parameter list"); + } + + /* Array as last part of a paramter type is just syntactic sugar. Turn it + * into a pointer */ + if (declaration->type->type == TYPE_ARRAY) { + const array_type_t *const arr_type = + (const array_type_t*)declaration->type; + declaration->type = + make_pointer_type(arr_type->element_type, TYPE_QUALIFIER_NONE); + } return declaration; } -static declaration_t *parse_parameters(method_type_t *type) +static declaration_t *parse_parameters(function_type_t *type) { if(token.type == T_IDENTIFIER) { - symbol_t *symbol = token.v.symbol; - declaration_t *declaration = symbol->declaration; - if(declaration == NULL - || declaration->storage_class != STORAGE_CLASS_TYPEDEF) { - /* TODO */ + symbol_t *symbol = token.v.symbol; + if(!is_typedef_symbol(symbol)) { + /* TODO: K&R style C parameters */ parse_identifier_list(); return NULL; } @@ -1083,11 +1758,11 @@ static declaration_t *parse_parameters(method_type_t *type) return NULL; } - declaration_t *declarations = NULL; - declaration_t *declaration; - declaration_t *last_declaration = NULL; - method_parameter_t *parameter; - method_parameter_t *last_parameter = NULL; + declaration_t *declarations = NULL; + declaration_t *declaration; + declaration_t *last_declaration = NULL; + function_parameter_t *parameter; + function_parameter_t *last_parameter = NULL; while(true) { switch(token.type) { @@ -1125,8 +1800,9 @@ static declaration_t *parse_parameters(method_type_t *type) } typedef enum { + CONSTRUCT_INVALID, CONSTRUCT_POINTER, - CONSTRUCT_METHOD, + CONSTRUCT_FUNCTION, CONSTRUCT_ARRAY } construct_type_type_t; @@ -1142,10 +1818,10 @@ struct parsed_pointer_t { type_qualifier_t type_qualifiers; }; -typedef struct construct_method_type_t construct_method_type_t; -struct construct_method_type_t { - construct_type_t construct_type; - method_type_t *method_type; +typedef struct construct_function_type_t construct_function_type_t; +struct construct_function_type_t { + construct_type_t construct_type; + function_type_t *function_type; }; typedef struct parsed_array_t parsed_array_t; @@ -1169,7 +1845,8 @@ static construct_type_t *parse_pointer_declarator(void) parsed_pointer_t *pointer = obstack_alloc(&temp_obst, sizeof(pointer[0])); memset(pointer, 0, sizeof(pointer[0])); - pointer->type_qualifiers = parse_type_qualifiers(); + pointer->construct_type.type = CONSTRUCT_POINTER; + pointer->type_qualifiers = parse_type_qualifiers(); return (construct_type_t*) pointer; } @@ -1180,6 +1857,7 @@ static construct_type_t *parse_array_declarator(void) 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; @@ -1207,28 +1885,27 @@ static construct_type_t *parse_array_declarator(void) return (construct_type_t*) array; } -static construct_type_t *parse_method_declarator(declaration_t *declaration) +static construct_type_t *parse_function_declarator(declaration_t *declaration) { eat('('); - method_type_t *method_type - = allocate_type_zero(sizeof(method_type[0])); - method_type->type.type = TYPE_METHOD; + function_type_t *type = allocate_type_zero(sizeof(type[0])); + type->type.type = TYPE_FUNCTION; - declaration_t *parameters = parse_parameters(method_type); + declaration_t *parameters = parse_parameters(type); if(declaration != NULL) { declaration->context.declarations = parameters; } - construct_method_type_t *construct_method_type = - obstack_alloc(&temp_obst, sizeof(construct_method_type[0])); - memset(construct_method_type, 0, sizeof(construct_method_type[0])); - construct_method_type->construct_type.type = CONSTRUCT_METHOD; - construct_method_type->method_type = method_type; + construct_function_type_t *construct_function_type = + obstack_alloc(&temp_obst, sizeof(construct_function_type[0])); + memset(construct_function_type, 0, sizeof(construct_function_type[0])); + construct_function_type->construct_type.type = CONSTRUCT_FUNCTION; + construct_function_type->function_type = type; expect(')'); - return (construct_type_t*) construct_method_type; + return (construct_type_t*) construct_function_type; } static construct_type_t *parse_inner_declarator(declaration_t *declaration, @@ -1270,15 +1947,22 @@ static construct_type_t *parse_inner_declarator(declaration_t *declaration, default: if(may_be_abstract) break; - parse_error_expected("problem while parsing declarator", T_IDENTIFIER, - '(', 0); + 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; } while(true) { construct_type_t *type; switch(token.type) { case '(': - type = parse_method_declarator(declaration); + type = parse_function_declarator(declaration); break; case '[': type = parse_array_declarator(); @@ -1315,20 +1999,22 @@ static type_t *construct_declarator_type(construct_type_t *construct_list, { construct_type_t *iter = construct_list; for( ; iter != NULL; iter = iter->next) { - parsed_pointer_t *parsed_pointer; - parsed_array_t *parsed_array; - construct_method_type_t *construct_method_type; - method_type_t *method_type; - pointer_type_t *pointer_type; - array_type_t *array_type; + parsed_pointer_t *parsed_pointer; + parsed_array_t *parsed_array; + construct_function_type_t *construct_function_type; + function_type_t *function_type; + pointer_type_t *pointer_type; + array_type_t *array_type; switch(iter->type) { - case CONSTRUCT_METHOD: - construct_method_type = (construct_method_type_t*) iter; - method_type = construct_method_type->method_type; - - method_type->result_type = type; - type = (type_t*) method_type; + case CONSTRUCT_INVALID: + panic("invalid type construction found"); + case CONSTRUCT_FUNCTION: + construct_function_type = (construct_function_type_t*) iter; + function_type = construct_function_type->function_type; + + function_type->result_type = type; + type = (type_t*) function_type; break; case CONSTRUCT_POINTER: @@ -1357,7 +2043,7 @@ static type_t *construct_declarator_type(construct_type_t *construct_list, type_t *hashed_type = typehash_insert((type_t*) type); if(hashed_type != type) { - obstack_free(type_obst, type); + free_type(type); type = hashed_type; } } @@ -1365,50 +2051,54 @@ static type_t *construct_declarator_type(construct_type_t *construct_list, return type; } -static void parse_declarator(declaration_t *declaration, - storage_class_t storage_class, type_t *type, - int may_be_abstract) +static declaration_t *parse_declarator( + const declaration_specifiers_t *specifiers, + type_t *type, bool may_be_abstract) { + 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); - declaration->type = construct_declarator_type(construct_type, type); - declaration->storage_class = storage_class; if(construct_type != NULL) { obstack_free(&temp_obst, construct_type); } + + return declaration; } static type_t *parse_abstract_declarator(type_t *base_type) { - construct_type_t *construct_type - = parse_inner_declarator(NULL, 1); - - if(construct_type == NULL) - return NULL; + construct_type_t *construct_type = parse_inner_declarator(NULL, 1); type_t *result = construct_declarator_type(construct_type, base_type); - obstack_free(&temp_obst, construct_type); + if(construct_type != NULL) { + obstack_free(&temp_obst, construct_type); + } return result; } static declaration_t *record_declaration(declaration_t *declaration) { - if(context == NULL) - return declaration; + assert(context != NULL); symbol_t *symbol = declaration->symbol; if(symbol != NULL) { - declaration_t *alias = environment_push(declaration, context); + declaration_t *alias = environment_push(declaration); if(alias != declaration) return alias; + } else { + declaration->parent_context = context; } if(last_declaration != NULL) { last_declaration->next = declaration; } else { - context->declarations = declaration; + context->declarations = declaration; } last_declaration = declaration; @@ -1422,44 +2112,66 @@ static void parser_error_multiple_definition(declaration_t *previous, 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(); + fprintf(stderr, "this is the location of the previous definition.\n"); } static void parse_init_declarators(const declaration_specifiers_t *specifiers) { while(true) { declaration_t *ndeclaration - = allocate_ast_zero(sizeof(ndeclaration[0])); + = parse_declarator(specifiers, specifiers->type, false); - parse_declarator(ndeclaration, specifiers->storage_class, - specifiers->type, 0); declaration_t *declaration = record_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 == '=') { next_token(); - /* TODO: check that this is an allowed type (esp. no method type) */ + /* TODO: check that this is an allowed type (no function type) */ - if(declaration->initializer != NULL) { + if(declaration->init.initializer != NULL) { parser_error_multiple_definition(declaration, ndeclaration); } - if(token.type == '{') { - // TODO - expect_void('}'); - } else { - declaration->initializer = parse_assignment_expression(); + initializer_t *initializer = parse_initializer(type); + + if(type->type == TYPE_ARRAY && initializer != NULL) { + assert(initializer->type == INITIALIZER_LIST); + + initializer_list_t *list = (initializer_list_t*) initializer; + array_type_t *array_type = (array_type_t*) type; + + if(array_type->size == NULL) { + const_t *cnst = allocate_ast_zero(sizeof(cnst[0])); + + cnst->expression.type = EXPR_CONST; + cnst->expression.datatype = type_size_t; + cnst->v.int_value = list->len; + + array_type->size = (expression_t*) cnst; + } } + + + ndeclaration->init.initializer = initializer; } else if(token.type == '{') { - if(declaration->type->type != TYPE_METHOD) { + if(type->type != TYPE_FUNCTION) { parser_print_error_prefix(); - fprintf(stderr, "Declarator "); - print_type_ext(declaration->type, declaration->symbol, NULL); - fprintf(stderr, " is not a method type.\n"); + fprintf(stderr, "declarator '"); + print_type_ext(orig_type, declaration->symbol, NULL); + fprintf(stderr, "' has a body but is not a function type.\n"); + eat_block(); + continue; } - if(declaration->initializer != NULL) { + if(declaration->init.statement != NULL) { parser_error_multiple_definition(declaration, ndeclaration); } if(ndeclaration != declaration) { @@ -1474,16 +2186,24 @@ static void parse_init_declarators(const declaration_specifiers_t *specifiers) /* push function parameters */ declaration_t *parameter = declaration->context.declarations; for( ; parameter != NULL; parameter = parameter->next) { - environment_push(parameter, context); + environment_push(parameter); } + int label_stack_top = label_top(); + declaration_t *old_current_function = current_function; + current_function = declaration; + statement_t *statement = parse_compound_statement(); + assert(current_function == declaration); + current_function = old_current_function; + label_pop_to(label_stack_top); + assert(context == &declaration->context); set_context(last_context); environment_pop_to(top); - declaration->statement = statement; + declaration->init.statement = statement; return; } @@ -1503,10 +2223,9 @@ static void parse_struct_declarators(const declaration_specifiers_t *specifiers) /* TODO (bitfields) */ } else { declaration_t *declaration - = allocate_ast_zero(sizeof(declaration[0])); - parse_declarator(declaration, specifiers->storage_class, - specifiers->type, 1); + = parse_declarator(specifiers, specifiers->type, true); + /* TODO: check constraints for struct declarations */ /* TODO: check for doubled fields */ record_declaration(declaration); @@ -1531,8 +2250,6 @@ static void parse_compound_type_entries(void) while(token.type != '}' && token.type != T_EOF) { declaration_specifiers_t specifiers; memset(&specifiers, 0, sizeof(specifiers)); - /* TODO not correct as this allows storage class stuff... but only - * specifiers and qualifiers sould be allowed here */ parse_declaration_specifiers(&specifiers); parse_struct_declarators(&specifiers); @@ -1545,12 +2262,44 @@ static void parse_compound_type_entries(void) static void parse_declaration(void) { + source_position_t source_position = token.source_position; + declaration_specifiers_t specifiers; memset(&specifiers, 0, sizeof(specifiers)); parse_declaration_specifiers(&specifiers); if(token.type == ';') { + if (specifiers.storage_class != STORAGE_CLASS_NONE) { + parse_warning_pos(source_position, + "useless keyword in empty declaration"); + } + switch (specifiers.type->type) { + case TYPE_COMPOUND_STRUCT: + case TYPE_COMPOUND_UNION: { + const compound_type_t *const comp_type = + (const compound_type_t*)specifiers.type; + if (comp_type->declaration->symbol == NULL) { + parse_warning_pos(source_position, + "unnamed struct/union that defines no instances"); + } + break; + } + + case TYPE_ENUM: break; + + default: + parse_warning_pos(source_position, "empty declaration"); + break; + } + next_token(); + + declaration_t *declaration = allocate_ast_zero(sizeof(declaration[0])); + + declaration->type = specifiers.type; + declaration->storage_class = specifiers.storage_class; + declaration->source_position = source_position; + record_declaration(declaration); return; } parse_init_declarators(&specifiers); @@ -1590,6 +2339,14 @@ struct expression_parser_function_t { expression_parser_function_t expression_parsers[T_LAST_TOKEN]; +static expression_t *make_invalid_expression(void) +{ + expression_t *expression = allocate_ast_zero(sizeof(expression[0])); + expression->type = EXPR_INVALID; + expression->source_position = token.source_position; + return expression; +} + static expression_t *expected_expression_error(void) { parser_print_error_prefix(); @@ -1597,11 +2354,9 @@ static expression_t *expected_expression_error(void) print_token(stderr, & token); fprintf(stderr, "\n"); - expression_t *expression = allocate_ast_zero(sizeof(expression[0])); - expression->type = EXPR_INVALID; next_token(); - return expression; + return make_invalid_expression(); } static expression_t *parse_string_const(void) @@ -1620,7 +2375,7 @@ static expression_t *parse_int_const(void) const_t *cnst = allocate_ast_zero(sizeof(cnst[0])); cnst->expression.type = EXPR_CONST; - cnst->expression.datatype = type_int; + cnst->expression.datatype = token.datatype; cnst->v.int_value = token.v.intvalue; next_token(); @@ -1633,7 +2388,7 @@ static expression_t *parse_float_const(void) const_t *cnst = allocate_ast_zero(sizeof(cnst[0])); cnst->expression.type = EXPR_CONST; - cnst->expression.datatype = type_int; + cnst->expression.datatype = token.datatype; cnst->v.float_value = token.v.floatvalue; next_token(); @@ -1641,6 +2396,45 @@ static expression_t *parse_float_const(void) return (expression_t*) cnst; } +static declaration_t *create_implicit_function(symbol_t *symbol, + const source_position_t source_position) +{ + function_type_t *function_type + = allocate_type_zero(sizeof(function_type[0])); + + function_type->type.type = TYPE_FUNCTION; + function_type->result_type = type_int; + function_type->unspecified_parameters = true; + + type_t *type = typehash_insert((type_t*) function_type); + if(type != (type_t*) function_type) { + free_type(function_type); + } + + declaration_t *declaration = allocate_ast_zero(sizeof(declaration[0])); + + declaration->storage_class = STORAGE_CLASS_EXTERN; + declaration->type = type; + declaration->symbol = symbol; + declaration->source_position = source_position; + + /* prepend the implicit definition to the global context + * this is safe since the symbol wasn't declared as anything else yet + */ + assert(symbol->declaration == NULL); + + context_t *last_context = context; + context = global_context; + + environment_push(declaration); + declaration->next = context->declarations; + context->declarations = declaration; + + context = last_context; + + return declaration; +} + static expression_t *parse_reference(void) { reference_expression_t *ref = allocate_ast_zero(sizeof(ref[0])); @@ -1648,28 +2442,32 @@ static expression_t *parse_reference(void) ref->expression.type = EXPR_REFERENCE; ref->symbol = token.v.symbol; - declaration_t *declaration = ref->symbol->declaration; + declaration_t *declaration = get_declaration(ref->symbol, NAMESPACE_NORMAL); + + source_position_t source_position = token.source_position; next_token(); if(declaration == NULL) { #ifndef STRICT_C99 - /* is it an implicitely defined function */ + /* 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); - /* TODO: do this correctly */ + + 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; } -#endif - - parser_print_error_prefix(); - fprintf(stderr, "unknown symbol '%s' found.\n", ref->symbol->string); - } else { - ref->declaration = declaration; - ref->expression.datatype = declaration->type; } + ref->declaration = declaration; + ref->expression.datatype = declaration->type; return (expression_t*) ref; } @@ -1678,7 +2476,7 @@ static void check_cast_allowed(expression_t *expression, type_t *dest_type) { (void) expression; (void) dest_type; - /* TODO check if cast is allowed and issue warnings/errors */ + /* TODO check if explicit cast is allowed and issue warnings/errors */ } static expression_t *parse_cast(void) @@ -1707,13 +2505,23 @@ static expression_t *parse_statement_expression(void) statement_expression_t *expression = allocate_ast_zero(sizeof(expression[0])); expression->expression.type = EXPR_STATEMENT; - expression->statement = parse_compound_statement(); + + statement_t *statement = parse_compound_statement(); + expression->statement = statement; + if(statement == NULL) { + expect(')'); + return NULL; + } + + assert(statement->type == STATEMENT_COMPOUND); + compound_statement_t *compound_statement + = (compound_statement_t*) 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; + const statement_t *iter = compound_statement->statements; + for( ; iter != NULL; iter = iter->next) { + last_statement = iter; } if(last_statement->type == STATEMENT_EXPRESSION) { @@ -1734,7 +2542,6 @@ static expression_t *parse_brace_expression(void) { eat('('); - declaration_t *declaration; switch(token.type) { case '{': /* gcc extension: a stement expression */ @@ -1744,9 +2551,7 @@ static expression_t *parse_brace_expression(void) TYPE_SPECIFIERS return parse_cast(); case T_IDENTIFIER: - declaration = token.v.symbol->declaration; - if(declaration != NULL && - (declaration->storage_class == STORAGE_CLASS_TYPEDEF)) { + if(is_typedef_symbol(token.v.symbol)) { return parse_cast(); } } @@ -1759,9 +2564,13 @@ static expression_t *parse_brace_expression(void) static expression_t *parse_function_keyword(void) { - eat(T___FUNCTION__); + next_token(); /* TODO */ + if (current_function == NULL) { + parse_error("'__func__' used outside of a function"); + } + string_literal_t *expression = allocate_ast_zero(sizeof(expression[0])); expression->expression.type = EXPR_FUNCTION; expression->expression.datatype = type_string; @@ -1783,12 +2592,12 @@ static expression_t *parse_pretty_function_keyword(void) return (expression_t*) expression; } -static member_designator_t *parse_member_designators(void) +static designator_t *parse_designator(void) { - member_designator_t *result = allocate_ast_zero(sizeof(result[0])); + designator_t *result = allocate_ast_zero(sizeof(result[0])); if(token.type != T_IDENTIFIER) { - parse_error_expected("problem while parsing member designator", + parse_error_expected("while parsing member designator", T_IDENTIFIER, 0); eat_brace(); return NULL; @@ -1796,19 +2605,18 @@ static member_designator_t *parse_member_designators(void) result->symbol = token.v.symbol; next_token(); - member_designator_t *last_designator = result; + designator_t *last_designator = result; while(true) { if(token.type == '.') { next_token(); if(token.type != T_IDENTIFIER) { - parse_error_expected("problem while parsing member designator", - T_IDENTIFIER, 0); + parse_error_expected("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; + designator_t *designator = allocate_ast_zero(sizeof(result[0])); + designator->symbol = token.v.symbol; next_token(); last_designator->next = designator; @@ -1817,8 +2625,7 @@ static member_designator_t *parse_member_designators(void) } if(token.type == '[') { next_token(); - member_designator_t *designator - = allocate_ast_zero(sizeof(result[0])); + designator_t *designator = allocate_ast_zero(sizeof(result[0])); designator->array_access = parse_expression(); if(designator->array_access == NULL) { eat_brace(); @@ -1848,10 +2655,44 @@ static expression_t *parse_offsetof(void) expect('('); expression->type = parse_typename(); expect(','); - expression->member_designators = parse_member_designators(); + expression->designator = parse_designator(); + expect(')'); + + return (expression_t*) expression; +} + +static expression_t *parse_va_arg(void) +{ + eat(T___builtin_va_arg); + + va_arg_expression_t *expression = allocate_ast_zero(sizeof(expression[0])); + expression->expression.type = EXPR_VA_ARG; + + expect('('); + expression->arg = parse_assignment_expression(); + expect(','); + expression->expression.datatype = parse_typename(); expect(')'); - return (expression_t*) expression; + return (expression_t*) expression; +} + +static type_t *make_function_1_type(type_t *result_type, type_t *argument_type) +{ + function_parameter_t *parameter = allocate_type_zero(sizeof(parameter[0])); + parameter->type = argument_type; + + function_type_t *type = allocate_type_zero(sizeof(type[0])); + type->type.type = TYPE_FUNCTION; + type->result_type = result_type; + type->parameters = parameter; + + type_t *result = typehash_insert((type_t*) type); + if(result != (type_t*) type) { + free_type(type); + } + + return result; } static expression_t *parse_builtin_symbol(void) @@ -1860,12 +2701,18 @@ static expression_t *parse_builtin_symbol(void) = allocate_ast_zero(sizeof(expression[0])); expression->expression.type = EXPR_BUILTIN_SYMBOL; - /* TODO: set datatype */ - expression->symbol = token.v.symbol; + type_t *type; + switch(token.type) { + case T___builtin_alloca: + type = make_function_1_type(type_void_ptr, type_size_t); + break; + } + next_token(); + expression->expression.datatype = type; return (expression_t*) expression; } @@ -1881,14 +2728,17 @@ static expression_t *parse_primary_expression(void) 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_arg: + return parse_va_arg(); + case T___builtin_alloca: case T___builtin_expect: case T___builtin_va_start: - case T___builtin_va_arg: case T___builtin_va_end: return parse_builtin_symbol(); @@ -1902,11 +2752,7 @@ static expression_t *parse_primary_expression(void) fprintf(stderr, "\n"); eat_statement(); - expression_t *expression = allocate_ast_zero(sizeof(expression[0])); - expression->type = EXPR_INVALID; - expression->datatype = type_void; - - return expression; + return make_invalid_expression(); } static expression_t *parse_array_expression(unsigned precedence, @@ -1916,22 +2762,37 @@ static expression_t *parse_array_expression(unsigned precedence, eat('['); + expression_t *index = parse_expression(); + array_access_expression_t *array_access = allocate_ast_zero(sizeof(array_access[0])); - array_access->expression.type = EXPR_ARRAY_ACCESS; - array_access->array_ref = array_ref; - array_access->index = parse_expression(); + array_access->expression.type = EXPR_ARRAY_ACCESS; + array_access->array_ref = array_ref; + array_access->index = index; + + type_t *type_left = skip_typeref(array_ref->datatype); + type_t *type_right = skip_typeref(index->datatype); - 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; + if(type_left != NULL && type_right != NULL) { + if(type_left->type == TYPE_POINTER) { + pointer_type_t *pointer = (pointer_type_t*) type_left; array_access->expression.datatype = pointer->points_to; + } else if(type_left->type == TYPE_ARRAY) { + array_type_t *array_type = (array_type_t*) type_left; + array_access->expression.datatype = array_type->element_type; + } else if(type_right->type == TYPE_POINTER) { + pointer_type_t *pointer = (pointer_type_t*) type_right; + array_access->expression.datatype = pointer->points_to; + } else if(type_right->type == TYPE_ARRAY) { + array_type_t *array_type = (array_type_t*) type_right; + array_access->expression.datatype = array_type->element_type; } else { parser_print_error_prefix(); - fprintf(stderr, "array access on object with non-pointer type "); - print_type(array_type); + fprintf(stderr, "array access on object with non-pointer types "); + print_type_quoted(type_left); + fprintf(stderr, ", "); + print_type_quoted(type_right); fprintf(stderr, "\n"); } } @@ -1942,326 +2803,871 @@ static expression_t *parse_array_expression(unsigned precedence, } next_token(); - return (expression_t*) array_access; + return (expression_t*) array_access; +} + +static bool is_declaration_specifier(const token_t *token, + bool only_type_specifiers) +{ + switch(token->type) { + TYPE_SPECIFIERS + return 1; + case T_IDENTIFIER: + return is_typedef_symbol(token->v.symbol); + STORAGE_CLASSES + TYPE_QUALIFIERS + if(only_type_specifiers) + return 0; + return 1; + + default: + return 0; + } +} + +static expression_t *parse_sizeof(unsigned precedence) +{ + eat(T_sizeof); + + 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(); + sizeof_expression->type = parse_typename(); + expect(')'); + } else { + expression_t *expression = parse_sub_expression(precedence); + sizeof_expression->type = expression->datatype; + sizeof_expression->size_expression = expression; + } + + 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(); + + select_expression_t *select = allocate_ast_zero(sizeof(select[0])); + + select->expression.type = EXPR_SELECT; + select->compound = compound; + + if(token.type != T_IDENTIFIER) { + parse_error_expected("while parsing select", T_IDENTIFIER, 0); + return (expression_t*) select; + } + symbol_t *symbol = token.v.symbol; + select->symbol = symbol; + next_token(); + + type_t *orig_type = compound->datatype; + if(orig_type == NULL) + return make_invalid_expression(); + + type_t *type = skip_typeref(orig_type); + + 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 make_invalid_expression(); + } + pointer_type_t *pointer_type = (pointer_type_t*) type; + type_left = pointer_type->points_to; + } + type_left = skip_typeref(type_left); + + 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 make_invalid_expression(); + } + + compound_type_t *compound_type = (compound_type_t*) type_left; + declaration_t *declaration = compound_type->declaration; + + 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 make_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 make_invalid_expression(); + } + + select->compound_entry = iter; + select->expression.datatype = iter->type; + return (expression_t*) select; +} + +static expression_t *parse_call_expression(unsigned precedence, + expression_t *expression) +{ + (void) precedence; + call_expression_t *call = allocate_ast_zero(sizeof(call[0])); + call->expression.type = EXPR_CALL; + call->function = expression; + + function_type_t *function_type; + type_t *type = expression->datatype; + if (type->type == TYPE_FUNCTION) { + function_type = (function_type_t*) type; + call->expression.datatype = function_type->result_type; + } else if (type->type == TYPE_POINTER && + ((pointer_type_t*)type)->points_to->type == TYPE_FUNCTION) { + pointer_type_t *const ptr_type = (pointer_type_t*)type; + function_type = (function_type_t*)ptr_type->points_to; + call->expression.datatype = function_type->result_type; + } else { + parser_print_error_prefix(); + fputs("called object '", stderr); + print_expression(expression); + fputs("' (type ", stderr); + print_type_quoted(type); + fputs(") is not a function\n", stderr); + + function_type = NULL; + call->expression.datatype = NULL; + } + + /* parse arguments */ + eat('('); + + if(token.type != ')') { + call_argument_t *last_argument = NULL; + + while(true) { + call_argument_t *argument = allocate_ast_zero(sizeof(argument[0])); + + argument->expression = parse_assignment_expression(); + if(last_argument == NULL) { + call->arguments = argument; + } else { + last_argument->next = argument; + } + last_argument = argument; + + if(token.type != ',') + break; + next_token(); + } + } + expect(')'); + + 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->datatype; + + if(type == NULL) + continue; + + 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); + } + } + } + } + + return (expression_t*) call; +} + +static type_t *semantic_arithmetic(type_t *type_left, type_t *type_right); + +static expression_t *parse_conditional_expression(unsigned precedence, + expression_t *expression) +{ + eat('?'); + + conditional_expression_t *conditional + = allocate_ast_zero(sizeof(conditional[0])); + conditional->expression.type = EXPR_CONDITIONAL; + conditional->condition = expression; + + /* 6.5.15.2 */ + type_t *condition_type_orig = conditional->condition->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", expression->source_position, + condition_type_orig); + } + } + + expression_t *const t_expr = parse_expression(); + conditional->true_expression = t_expr; + expect(':'); + expression_t *const f_expr = parse_sub_expression(precedence); + conditional->false_expression = f_expr; + + type_t *const true_type = t_expr->datatype; + if(true_type == NULL) + return (expression_t*) conditional; + type_t *const false_type = f_expr->datatype; + if(false_type == NULL) + return (expression_t*) conditional; + + type_t *const skipped_true_type = skip_typeref(true_type); + type_t *const skipped_false_type = skip_typeref(false_type); + + /* 6.5.15.3 */ + if (skipped_true_type == skipped_false_type) { + conditional->expression.datatype = skipped_true_type; + } else if (is_type_arithmetic(skipped_true_type) && + is_type_arithmetic(skipped_false_type)) { + type_t *const result = semantic_arithmetic(skipped_true_type, + skipped_false_type); + conditional->true_expression = create_implicit_cast(t_expr, result); + conditional->false_expression = create_implicit_cast(f_expr, result); + conditional->expression.datatype = result; + } else if (skipped_true_type->type == TYPE_POINTER && + skipped_false_type->type == TYPE_POINTER && + true /* TODO compatible points_to types */) { + /* TODO */ + } else if(/* (is_null_ptr_const(skipped_true_type) && + skipped_false_type->type == TYPE_POINTER) + || (is_null_ptr_const(skipped_false_type) && + skipped_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("while parsing conditional", + expression->source_position, true_type, + skipped_false_type); + } + + return (expression_t*) conditional; +} + +static expression_t *parse_extension(unsigned precedence) +{ + eat(T___extension__); + + /* TODO enable extensions */ + + return parse_sub_expression(precedence); +} + +static expression_t *parse_builtin_classify_type(const unsigned precedence) +{ + eat(T___builtin_classify_type); + + classify_type_expression_t *const classify_type_expr = + allocate_ast_zero(sizeof(classify_type_expr[0])); + classify_type_expr->expression.type = EXPR_CLASSIFY_TYPE; + classify_type_expr->expression.datatype = type_int; + + expect('('); + expression_t *const expression = parse_sub_expression(precedence); + expect(')'); + classify_type_expr->type_expression = expression; + + return (expression_t*)classify_type_expr; +} + +static void semantic_incdec(unary_expression_t *expression) +{ + type_t *orig_type = expression->value->datatype; + if(orig_type == NULL) + return; + + 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; + } + + expression->expression.datatype = orig_type; +} + +static void semantic_unexpr_arithmetic(unary_expression_t *expression) +{ + type_t *orig_type = expression->value->datatype; + if(orig_type == NULL) + return; + + 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->expression.datatype = orig_type; +} + +static void semantic_unexpr_scalar(unary_expression_t *expression) +{ + type_t *orig_type = expression->value->datatype; + if(orig_type == NULL) + return; + + type_t *type = skip_typeref(orig_type); + if (!is_type_scalar(type)) { + parse_error("operand of ! must be of scalar type\n"); + return; + } + + expression->expression.datatype = orig_type; +} + +static void semantic_unexpr_integer(unary_expression_t *expression) +{ + type_t *orig_type = expression->value->datatype; + if(orig_type == NULL) + return; + + type_t *type = skip_typeref(orig_type); + if (!is_type_integer(type)) { + parse_error("operand of ~ must be of integer type\n"); + return; + } + + expression->expression.datatype = orig_type; +} + +static void semantic_dereference(unary_expression_t *expression) +{ + type_t *orig_type = expression->value->datatype; + if(orig_type == NULL) + return; + + type_t *type = skip_typeref(orig_type); + switch (type->type) { + case TYPE_ARRAY: { + array_type_t *const array_type = (array_type_t*)type; + expression->expression.datatype = array_type->element_type; + break; + } + + case TYPE_POINTER: { + pointer_type_t *pointer_type = (pointer_type_t*)type; + expression->expression.datatype = pointer_type->points_to; + break; + } + + default: + parser_print_error_prefix(); + fputs("'Unary *' needs pointer or arrray type, but type ", stderr); + print_type_quoted(orig_type); + fputs(" given.\n", stderr); + return; + } +} + +static void semantic_take_addr(unary_expression_t *expression) +{ + type_t *orig_type = expression->value->datatype; + if(orig_type == NULL) + return; + + expression_t *value = expression->value; + 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; + } + } + + expression->expression.datatype = make_pointer_type(orig_type, 0); } -static bool is_declaration_specifier(const token_t *token, - bool only_type_specifiers) +#define CREATE_UNARY_EXPRESSION_PARSER(token_type, unexpression_type, sfunc) \ +static expression_t *parse_##unexpression_type(unsigned 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 = parse_sub_expression(precedence); \ + \ + sfunc(unary_expression); \ + \ + return (expression_t*) unary_expression; \ +} + +CREATE_UNARY_EXPRESSION_PARSER('-', UNEXPR_NEGATE, semantic_unexpr_arithmetic) +CREATE_UNARY_EXPRESSION_PARSER('+', UNEXPR_PLUS, semantic_unexpr_arithmetic) +CREATE_UNARY_EXPRESSION_PARSER('!', UNEXPR_NOT, semantic_unexpr_scalar) +CREATE_UNARY_EXPRESSION_PARSER('*', UNEXPR_DEREFERENCE, semantic_dereference) +CREATE_UNARY_EXPRESSION_PARSER('&', UNEXPR_TAKE_ADDRESS, semantic_take_addr) +CREATE_UNARY_EXPRESSION_PARSER('~', UNEXPR_BITWISE_NEGATE, + semantic_unexpr_integer) +CREATE_UNARY_EXPRESSION_PARSER(T_PLUSPLUS, UNEXPR_PREFIX_INCREMENT, + semantic_incdec) +CREATE_UNARY_EXPRESSION_PARSER(T_MINUSMINUS, UNEXPR_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); \ + \ + 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; \ + \ + sfunc(unary_expression); \ + \ + return (expression_t*) unary_expression; \ +} + +CREATE_UNARY_POSTFIX_EXPRESSION_PARSER(T_PLUSPLUS, UNEXPR_POSTFIX_INCREMENT, + semantic_incdec) +CREATE_UNARY_POSTFIX_EXPRESSION_PARSER(T_MINUSMINUS, UNEXPR_POSTFIX_DECREMENT, + semantic_incdec) + +static type_t *semantic_arithmetic(type_t *type_left, type_t *type_right) { - declaration_t *declaration; + /* 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; + } - 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; + type_right = promote_integer(type_right); + type_left = promote_integer(type_left); - default: - return 0; + if(type_left == type_right) + return type_left; + + bool signed_left = is_type_signed(type_left); + bool signed_right = is_type_signed(type_right); + if(get_rank(type_left) < get_rank(type_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_sizeof(unsigned precedence) +static void semantic_binexpr_arithmetic(binary_expression_t *expression) { - eat(T_sizeof); + expression_t *left = expression->left; + expression_t *right = expression->right; + type_t *orig_type_left = left->datatype; + type_t *orig_type_right = right->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(')'); - } else { - expression_t *expression = parse_sub_expression(precedence); - sizeof_expression->type = expression->datatype; - sizeof_expression->size_expression = expression; + 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)) { + /* TODO: improve error message */ + parser_print_error_prefix(); + fprintf(stderr, "operation needs arithmetic types\n"); + return; } - return (expression_t*) sizeof_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_select_expression(unsigned precedence, - expression_t *compound) +static void semantic_shift_op(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->datatype; + type_t *orig_type_right = right->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_integer(type_left) || !is_type_integer(type_right)) { + /* TODO: improve error message */ + parser_print_error_prefix(); + fprintf(stderr, "operation needs integer types\n"); + return; } - select->symbol = token.v.symbol; - next_token(); - return (expression_t*) select; + 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_call_expression(unsigned precedence, - expression_t *expression) +static void semantic_add(binary_expression_t *expression) { - (void) precedence; - call_expression_t *call = allocate_ast_zero(sizeof(call[0])); - - call->expression.type = EXPR_CALL; - call->method = expression; + expression_t *left = expression->left; + expression_t *right = expression->right; + type_t *orig_type_left = left->datatype; + type_t *orig_type_right = right->datatype; - /* parse arguments */ - eat('('); + if(orig_type_left == NULL || orig_type_right == NULL) + return; - if(token.type != ')') { - call_argument_t *last_argument = NULL; + type_t *type_left = skip_typeref(orig_type_left); + type_t *type_right = skip_typeref(orig_type_right); - while(true) { - call_argument_t *argument = allocate_ast_zero(sizeof(argument[0])); + /* § 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(type_left->type == TYPE_POINTER && is_type_integer(type_right)) { + expression->expression.datatype = type_left; + } else if(type_right->type == TYPE_POINTER && is_type_integer(type_left)) { + expression->expression.datatype = type_right; + } else if (type_left->type == TYPE_ARRAY && is_type_integer(type_right)) { + const array_type_t *const arr_type = (const array_type_t*)type_left; + expression->expression.datatype = + make_pointer_type(arr_type->element_type, TYPE_QUALIFIER_NONE); + } else if (type_right->type == TYPE_ARRAY && is_type_integer(type_left)) { + const array_type_t *const arr_type = (const array_type_t*)type_right; + expression->expression.datatype = + make_pointer_type(arr_type->element_type, TYPE_QUALIFIER_NONE); + } 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"); + } +} - argument->expression = parse_assignment_expression(); - if(last_argument == NULL) { - call->arguments = argument; - } else { - last_argument->next = argument; - } - last_argument = argument; +static void semantic_sub(binary_expression_t *expression) +{ + expression_t *left = expression->left; + expression_t *right = expression->right; + type_t *orig_type_left = left->datatype; + type_t *orig_type_right = right->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) { + /* § 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(type_left->type == TYPE_POINTER && is_type_integer(type_right)) { + expression->expression.datatype = type_left; + } else if(type_left->type == TYPE_POINTER && + type_right->type == TYPE_POINTER) { + if(!pointers_compatible(type_left, type_right)) { parser_print_error_prefix(); - fprintf(stderr, "expected a method type for call but found type "); - print_type(expression->datatype); - fprintf(stderr, "\n"); + 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 { - method_type_t *method_type = (method_type_t*) type; - call->expression.datatype = method_type->result_type; + 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"); } - - return (expression_t*) call; } -static void type_error(const char *msg, const source_position_t source_position, - const type_t *type) +static void semantic_comparison(binary_expression_t *expression) { - parser_print_error_prefix_pos(source_position); - fprintf(stderr, "%s, but found type ", msg); - print_type(type); - fputc('\n', stderr); - error(); -} + expression_t *left = expression->left; + expression_t *right = expression->right; + type_t *orig_type_left = left->datatype; + type_t *orig_type_right = right->datatype; -static void type_error_incompatible(const char *msg, - const source_position_t source_position, const type_t *type1, - const type_t *type2) -{ - 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(); + if(orig_type_left == NULL || orig_type_right == NULL) + return; + + 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 (type_left->type == TYPE_POINTER && + type_right->type == TYPE_POINTER) { + /* TODO check compatibility */ + } else if (type_left->type == TYPE_POINTER) { + expression->right = create_implicit_cast(right, type_left); + } else if (type_right->type == TYPE_POINTER) { + expression->left = create_implicit_cast(left, type_right); + } else { + type_error_incompatible("invalid operands in comparison", + token.source_position, type_left, type_right); + } + expression->expression.datatype = type_int; } -static type_t *get_type_after_conversion(const type_t *type1, - const type_t *type2) +static void semantic_arithmetic_assign(binary_expression_t *expression) { - /* TODO... */ - (void) type2; - return (type_t*) type1; + expression_t *left = expression->left; + expression_t *right = expression->right; + type_t *orig_type_left = left->datatype; + type_t *orig_type_right = right->datatype; + + if(orig_type_left == NULL || orig_type_right == NULL) + return; + + 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)) { + /* TODO: improve error message */ + parser_print_error_prefix(); + fprintf(stderr, "operation needs arithmetic types\n"); + return; + } + + /* 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_conditional_expression(unsigned precedence, - expression_t *expression) +static void semantic_arithmetic_addsubb_assign(binary_expression_t *expression) { - eat('?'); + expression_t *left = expression->left; + expression_t *right = expression->right; + type_t *orig_type_left = left->datatype; + type_t *orig_type_right = right->datatype; - conditional_expression_t *conditional - = allocate_ast_zero(sizeof(conditional[0])); - conditional->expression.type = EXPR_CONDITIONAL; - conditional->condition = expression; + if(orig_type_left == NULL || orig_type_right == NULL) + return; - /* 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); - } + 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 (type_left->type == TYPE_POINTER && 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; } +} - conditional->true_expression = parse_expression(); - expect(':'); - conditional->false_expression = parse_sub_expression(precedence); +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->datatype; + type_t *orig_type_right = right->datatype; - 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; + if(orig_type_left == NULL || orig_type_right == NULL) + return; - /* 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); + type_t *type_left = skip_typeref(orig_type_left); + type_t *type_right = skip_typeref(orig_type_right); + + 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*) conditional; + expression->expression.datatype = type_int; } -static expression_t *parse_extension(unsigned precedence) +static void semantic_binexpr_assign(binary_expression_t *expression) { - eat(T___extension__); + expression_t *left = expression->left; + type_t *type_left = left->datatype; - /* TODO enable extensions */ + if(type_left == NULL) + return; - return parse_sub_expression(precedence); + if (type_left->type == TYPE_ARRAY) { + parse_error("Cannot assign to arrays."); + } else if (type_left != NULL) { + semantic_assign(type_left, &expression->right, "assignment"); + } + + expression->expression.datatype = type_left; } -#define CREATE_UNARY_EXPRESSION_PARSER(token_type, unexpression_type) \ -static \ -expression_t *parse_##unexpression_type(unsigned 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 = parse_sub_expression(precedence); \ - \ - 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; \ +static void semantic_comma(binary_expression_t *expression) +{ + expression->expression.datatype = expression->right->datatype; +} + +#define CREATE_BINEXPR_PARSER(token_type, binexpression_type, sfunc, lr) \ +static expression_t *parse_##binexpression_type(unsigned precedence, \ + expression_t *left) \ +{ \ + eat(token_type); \ + \ + expression_t *right = parse_sub_expression(precedence + lr); \ + \ + 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; \ + sfunc(binexpr); \ + \ + return (expression_t*) binexpr; \ } -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) +CREATE_BINEXPR_PARSER(',', BINEXPR_COMMA, semantic_comma, 1) +CREATE_BINEXPR_PARSER('*', BINEXPR_MUL, semantic_binexpr_arithmetic, 1) +CREATE_BINEXPR_PARSER('/', BINEXPR_DIV, semantic_binexpr_arithmetic, 1) +CREATE_BINEXPR_PARSER('%', BINEXPR_MOD, semantic_binexpr_arithmetic, 1) +CREATE_BINEXPR_PARSER('+', BINEXPR_ADD, semantic_add, 1) +CREATE_BINEXPR_PARSER('-', BINEXPR_SUB, semantic_sub, 1) +CREATE_BINEXPR_PARSER('<', BINEXPR_LESS, semantic_comparison, 1) +CREATE_BINEXPR_PARSER('>', BINEXPR_GREATER, semantic_comparison, 1) +CREATE_BINEXPR_PARSER('=', BINEXPR_ASSIGN, semantic_binexpr_assign, 0) +CREATE_BINEXPR_PARSER(T_EQUALEQUAL, BINEXPR_EQUAL, semantic_comparison, 1) +CREATE_BINEXPR_PARSER(T_EXCLAMATIONMARKEQUAL, BINEXPR_NOTEQUAL, + semantic_comparison, 1) +CREATE_BINEXPR_PARSER(T_LESSEQUAL, BINEXPR_LESSEQUAL, semantic_comparison, 1) +CREATE_BINEXPR_PARSER(T_GREATEREQUAL, BINEXPR_GREATEREQUAL, + semantic_comparison, 1) +CREATE_BINEXPR_PARSER('&', BINEXPR_BITWISE_AND, semantic_binexpr_arithmetic, 1) +CREATE_BINEXPR_PARSER('|', BINEXPR_BITWISE_OR, semantic_binexpr_arithmetic, 1) +CREATE_BINEXPR_PARSER('^', BINEXPR_BITWISE_XOR, semantic_binexpr_arithmetic, 1) +CREATE_BINEXPR_PARSER(T_ANDAND, BINEXPR_LOGICAL_AND, semantic_logical_op, 1) +CREATE_BINEXPR_PARSER(T_PIPEPIPE, BINEXPR_LOGICAL_OR, semantic_logical_op, 1) +/* TODO shift has a bit special semantic */ +CREATE_BINEXPR_PARSER(T_LESSLESS, BINEXPR_SHIFTLEFT, + semantic_shift_op, 1) +CREATE_BINEXPR_PARSER(T_GREATERGREATER, BINEXPR_SHIFTRIGHT, + semantic_shift_op, 1) +CREATE_BINEXPR_PARSER(T_PLUSEQUAL, BINEXPR_ADD_ASSIGN, + semantic_arithmetic_addsubb_assign, 0) +CREATE_BINEXPR_PARSER(T_MINUSEQUAL, BINEXPR_SUB_ASSIGN, + semantic_arithmetic_addsubb_assign, 0) +CREATE_BINEXPR_PARSER(T_ASTERISKEQUAL, BINEXPR_MUL_ASSIGN, + semantic_arithmetic_assign, 0) +CREATE_BINEXPR_PARSER(T_SLASHEQUAL, BINEXPR_DIV_ASSIGN, + semantic_arithmetic_assign, 0) +CREATE_BINEXPR_PARSER(T_PERCENTEQUAL, BINEXPR_MOD_ASSIGN, + semantic_arithmetic_assign, 0) +CREATE_BINEXPR_PARSER(T_LESSLESSEQUAL, BINEXPR_SHIFTLEFT_ASSIGN, + semantic_arithmetic_assign, 0) +CREATE_BINEXPR_PARSER(T_GREATERGREATEREQUAL, BINEXPR_SHIFTRIGHT_ASSIGN, + semantic_arithmetic_assign, 0) +CREATE_BINEXPR_PARSER(T_ANDEQUAL, BINEXPR_BITWISE_AND_ASSIGN, + semantic_arithmetic_assign, 0) +CREATE_BINEXPR_PARSER(T_PIPEEQUAL, BINEXPR_BITWISE_OR_ASSIGN, + semantic_arithmetic_assign, 0) +CREATE_BINEXPR_PARSER(T_CARETEQUAL, BINEXPR_BITWISE_XOR_ASSIGN, + semantic_arithmetic_assign, 0) static expression_t *parse_sub_expression(unsigned precedence) { @@ -2296,7 +3702,7 @@ static expression_t *parse_sub_expression(unsigned precedence) left = parser->infix_parser(parser->infix_precedence, left); assert(left != NULL); - assert(left->type != EXPR_INVALID); + assert(left->type != EXPR_UNKNOWN); left->source_position = source_position; } @@ -2310,8 +3716,8 @@ 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]; @@ -2325,8 +3731,9 @@ 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_expression_infix_parser( + parse_expression_infix_function parser, int token_type, + unsigned precedence) { expression_parser_function_t *entry = &expression_parsers[token_type]; @@ -2408,6 +3815,8 @@ static void init_expression_parsers(void) register_expression_parser(parse_UNEXPR_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); } @@ -2421,7 +3830,7 @@ static statement_t *parse_case_statement(void) label->expression = parse_expression(); expect(':'); - label->statement.next = parse_statement(); + label->label_statement = parse_statement(); return (statement_t*) label; } @@ -2435,18 +3844,68 @@ static statement_t *parse_default_statement(void) label->statement.source_position = token.source_position; expect(':'); - label->statement.next = parse_statement(); + label->label_statement = parse_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) @@ -2565,10 +4024,27 @@ 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) @@ -2577,8 +4053,8 @@ static statement_t *parse_continue(void) expect(';'); statement_t *statement = allocate_ast_zero(sizeof(statement[0])); - statement->source_position = token.source_position; statement->type = STATEMENT_CONTINUE; + statement->source_position = token.source_position; return statement; } @@ -2589,8 +4065,8 @@ static statement_t *parse_break(void) expect(';'); statement_t *statement = allocate_ast_zero(sizeof(statement[0])); - statement->source_position = token.source_position; statement->type = STATEMENT_BREAK; + statement->source_position = token.source_position; return statement; } @@ -2603,9 +4079,32 @@ static statement_t *parse_return(void) statement->statement.type = STATEMENT_RETURN; statement->statement.source_position = token.source_position; + + assert(current_function->type->type == TYPE_FUNCTION); + function_type_t *function_type = (function_type_t*) current_function->type; + type_t *return_type = function_type->result_type; + + expression_t *return_value; if(token.type != ';') { - statement->return_value = parse_expression(); + return_value = parse_expression(); + + if(return_type == type_void && return_value->datatype != 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 { + return_value = NULL; + if(return_type != type_void) { + parse_warning("'return' without value, in function returning " + "non-void"); + } } + statement->return_value = return_value; + expect(';'); return (statement_t*) statement; @@ -2655,7 +4154,6 @@ static statement_t *parse_expression_statement(void) static statement_t *parse_statement(void) { - declaration_t *declaration; statement_t *statement = NULL; /* declaration or statement */ @@ -2719,9 +4217,7 @@ 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; } @@ -2754,13 +4250,13 @@ static statement_t *parse_statement(void) 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); @@ -2784,12 +4280,17 @@ static statement_t *parse_compound_statement(void) 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; } @@ -2797,6 +4298,9 @@ static translation_unit_t *parse_translation_unit(void) { translation_unit_t *unit = allocate_ast_zero(sizeof(unit[0])); + assert(global_context == NULL); + global_context = &unit->context; + assert(context == NULL); set_context(&unit->context); @@ -2808,15 +4312,20 @@ static translation_unit_t *parse_translation_unit(void) context = NULL; last_declaration = NULL; + assert(global_context == &unit->context); + global_context = NULL; + return unit; } translation_unit_t *parse(void) { - obstack_init(&environment_obstack); - environment_stack = NEW_ARR_F(environment_entry_t*, 0); + environment_stack = NEW_ARR_F(stack_entry_t, 0); + 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) { @@ -2825,7 +4334,10 @@ 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; } @@ -2835,11 +4347,17 @@ void init_parser(void) init_expression_parsers(); obstack_init(&temp_obst); - type_int = make_atomic_type(ATOMIC_TYPE_INT, 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, 0); + type_uint = make_atomic_type(ATOMIC_TYPE_UINT, 0); + type_long_double = make_atomic_type(ATOMIC_TYPE_LONG_DOUBLE, 0); + type_double = make_atomic_type(ATOMIC_TYPE_DOUBLE, 0); + type_float = make_atomic_type(ATOMIC_TYPE_FLOAT, 0); + type_size_t = make_atomic_type(ATOMIC_TYPE_ULONG, 0); + type_ptrdiff_t = make_atomic_type(ATOMIC_TYPE_LONG, 0); + type_const_char = make_atomic_type(ATOMIC_TYPE_CHAR, TYPE_QUALIFIER_CONST); + type_void = make_atomic_type(ATOMIC_TYPE_VOID, 0); + type_void_ptr = make_pointer_type(type_void, 0); + type_string = make_pointer_type(type_const_char, 0); } void exit_parser(void)