X-Git-Url: http://nsz.repo.hu/git/?a=blobdiff_plain;f=parser.c;h=3b2b6fd97a6912a7c316bb0840f27863a92ea82b;hb=fc5cb406afbc60111167f594dd1fade033716e1f;hp=1c627a320172b04939a374ce38b275c98e74c63c;hpb=659fb8e34ebdf612319f58c3ffe67f91d24efde7;p=cparser diff --git a/parser.c b/parser.c index 1c627a3..3b2b6fd 100644 --- a/parser.c +++ b/parser.c @@ -19,28 +19,34 @@ #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 *global_context = NULL; -static context_t *context = NULL; -static declaration_t *last_declaration = NULL; -static struct obstack temp_obst; - -static type_t *type_int = NULL; -static type_t *type_double = NULL; -static type_t *type_const_char = NULL; -static type_t *type_string = NULL; -static type_t *type_void = NULL; -static type_t *type_size_t = NULL; + 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_long_double = NULL; +static type_t *type_double = NULL; +static type_t *type_float = NULL; +static type_t *type_char = NULL; +static type_t *type_string = NULL; +static type_t *type_void = NULL; +static type_t *type_void_ptr = NULL; +static type_t *type_size_t = NULL; +static type_t *type_ptrdiff_t = NULL; static statement_t *parse_compound_statement(void); static statement_t *parse_statement(void); @@ -65,14 +71,10 @@ 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 @@ -103,21 +105,101 @@ static type_t *parse_typename(void); TYPE_QUALIFIERS \ TYPE_SPECIFIERS -static inline void *allocate_ast_zero(size_t size) +static void *allocate_ast_zero(size_t size) { void *res = allocate_ast(size); memset(res, 0, size); return res; } -static inline void *allocate_type_zero(size_t size) +static size_t get_expression_struct_size(expression_type_t type) +{ + static const size_t sizes[] = { + [EXPR_INVALID] = sizeof(expression_base_t), + [EXPR_REFERENCE] = sizeof(reference_expression_t), + [EXPR_CONST] = sizeof(const_expression_t), + [EXPR_STRING_LITERAL] = sizeof(string_literal_expression_t), + [EXPR_CALL] = sizeof(call_expression_t), + [EXPR_UNARY] = sizeof(unary_expression_t), + [EXPR_BINARY] = sizeof(binary_expression_t), + [EXPR_CONDITIONAL] = sizeof(conditional_expression_t), + [EXPR_SELECT] = sizeof(select_expression_t), + [EXPR_ARRAY_ACCESS] = sizeof(array_access_expression_t), + [EXPR_SIZEOF] = sizeof(sizeof_expression_t), + [EXPR_CLASSIFY_TYPE] = sizeof(classify_type_expression_t), + [EXPR_FUNCTION] = sizeof(string_literal_expression_t), + [EXPR_PRETTY_FUNCTION] = sizeof(string_literal_expression_t), + [EXPR_BUILTIN_SYMBOL] = sizeof(builtin_symbol_expression_t), + [EXPR_OFFSETOF] = sizeof(offsetof_expression_t), + [EXPR_VA_ARG] = sizeof(va_arg_expression_t), + [EXPR_STATEMENT] = sizeof(statement_expression_t) + }; + assert(sizeof(sizes) / sizeof(sizes[0]) == EXPR_STATEMENT + 1); + assert(type <= EXPR_STATEMENT); + assert(sizes[type] != 0); + return sizes[type]; +} + +static expression_t *allocate_expression_zero(expression_type_t type) +{ + size_t size = get_expression_struct_size(type); + expression_t *res = allocate_ast_zero(size); + + res->base.type = type; + return res; +} + +static size_t get_type_struct_size(type_type_t type) +{ + static const size_t sizes[] = { + [TYPE_ATOMIC] = sizeof(atomic_type_t), + [TYPE_COMPOUND_STRUCT] = sizeof(compound_type_t), + [TYPE_COMPOUND_UNION] = sizeof(compound_type_t), + [TYPE_ENUM] = sizeof(enum_type_t), + [TYPE_FUNCTION] = sizeof(function_type_t), + [TYPE_POINTER] = sizeof(pointer_type_t), + [TYPE_ARRAY] = sizeof(array_type_t), + [TYPE_BUILTIN] = sizeof(builtin_type_t), + [TYPE_TYPEDEF] = sizeof(typedef_type_t), + [TYPE_TYPEOF] = sizeof(typeof_type_t), + }; + assert(sizeof(sizes) / sizeof(sizes[0]) == (int) TYPE_TYPEOF + 1); + assert(type <= TYPE_TYPEOF); + assert(sizes[type] != 0); + return sizes[type]; +} + +static type_t *allocate_type_zero(type_type_t type) { - void *res = obstack_alloc(type_obst, size); + size_t size = get_type_struct_size(type); + type_t *res = obstack_alloc(type_obst, size); memset(res, 0, size); + + res->base.type = type; return res; } -static inline void free_type(void *type) +static size_t get_initializer_size(initializer_type_t type) +{ + static const size_t sizes[] = { + [INITIALIZER_VALUE] = sizeof(initializer_value_t), + [INITIALIZER_STRING] = sizeof(initializer_string_t), + [INITIALIZER_LIST] = sizeof(initializer_list_t) + }; + assert(type < INITIALIZER_COUNT); + assert(sizes[type] != 0); + return sizes[type]; +} + +static initializer_t *allocate_initializer(initializer_type_t type) +{ + initializer_t *result = allocate_ast_zero(get_initializer_size(type)); + result->type = type; + + return result; +} + +static void free_type(void *type) { obstack_free(type_obst, type); } @@ -125,11 +207,16 @@ static inline void free_type(void *type) /** * returns the top element of the environment stack */ -static inline size_t environment_top(void) +static size_t environment_top(void) { return ARR_LEN(environment_stack); } +static size_t label_top(void) +{ + return ARR_LEN(label_stack); +} + static inline void next_token(void) @@ -155,32 +242,33 @@ static inline const token_t *look_ahead(int num) #define eat(token_type) do { assert(token.type == token_type); next_token(); } while(0) -void error(void) +static void error(void) { + found_error = true; #ifdef ABORT_ON_ERROR abort(); #endif } -void parser_print_prefix_pos(const source_position_t source_position) +static void parser_print_prefix_pos(const source_position_t source_position) { fputs(source_position.input_name, stderr); fputc(':', stderr); - fprintf(stderr, "%d", source_position.linenr); + fprintf(stderr, "%u", source_position.linenr); fputs(": ", stderr); } -void parser_print_error_prefix_pos(const source_position_t source_position) +static void parser_print_error_prefix_pos( + const source_position_t source_position) { parser_print_prefix_pos(source_position); fputs("error: ", stderr); error(); } -void parser_print_error_prefix(void) +static void parser_print_error_prefix(void) { - parser_print_prefix_pos(token.source_position); - error(); + parser_print_error_prefix_pos(token.source_position); } static void parse_error(const char *message) @@ -189,13 +277,25 @@ 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; @@ -225,6 +325,33 @@ 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 == '{') @@ -290,6 +417,14 @@ 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); \ @@ -302,16 +437,12 @@ 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; } /** @@ -321,89 +452,423 @@ static void set_context(context_t *new_context) static bool is_compatible_declaration (declaration_t *declaration, declaration_t *previous) { + if (declaration->type->type == TYPE_FUNCTION && + previous->type->type == TYPE_FUNCTION && + previous->type->function.unspecified_parameters) { + function_type_t* const prev_func = &previous->type->function; + function_type_t* const decl_func = &declaration->type->function; + if (prev_func->unspecified_parameters && + prev_func->result_type == decl_func->result_type) { + declaration->type = previous->type; + return true; + } + } /* TODO: not correct yet */ 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 { + unsigned old_storage_class = previous_declaration->storage_class; + unsigned new_storage_class = declaration->storage_class; + if (current_function == NULL) { + if (old_storage_class != STORAGE_CLASS_STATIC && + new_storage_class == STORAGE_CLASS_STATIC) { + parser_print_error_prefix_pos(declaration->source_position); + fprintf(stderr, + "static declaration of '%s' follows non-static declaration\n", + symbol->string); + parser_print_error_prefix_pos(previous_declaration->source_position); + fprintf(stderr, "previous declaration of '%s' was here\n", + symbol->string); + } else { + if (old_storage_class == STORAGE_CLASS_EXTERN) { + if (new_storage_class == STORAGE_CLASS_NONE) { + previous_declaration->storage_class = STORAGE_CLASS_NONE; + } + } else { + parser_print_warning_prefix_pos(declaration->source_position); + fprintf(stderr, "redundant declaration for '%s'\n", + symbol->string); + parser_print_warning_prefix_pos(previous_declaration->source_position); + fprintf(stderr, "previous declaration of '%s' was here\n", + symbol->string); + } + } + } else { + if (old_storage_class == STORAGE_CLASS_EXTERN && + new_storage_class == STORAGE_CLASS_EXTERN) { + parser_print_warning_prefix_pos(declaration->source_position); + fprintf(stderr, "redundant extern declaration for '%s'\n", + symbol->string); + parser_print_warning_prefix_pos(previous_declaration->source_position); + fprintf(stderr, "previous declaration of '%s' was here\n", + symbol->string); + } else { + parser_print_error_prefix_pos(declaration->source_position); + if (old_storage_class == new_storage_class) { + fprintf(stderr, "redeclaration of '%s'\n", symbol->string); + } else { + fprintf(stderr, "redeclaration of '%s' with different linkage\n", symbol->string); + } + parser_print_error_prefix_pos(previous_declaration->source_position); + fprintf(stderr, "previous declaration of '%s' was here\n", + symbol->string); + } } - return previous_declaration; } + return previous_declaration; } - environment_entry_t *entry - = obstack_alloc(&environment_obstack, sizeof(entry[0])); - memset(entry, 0, sizeof(entry[0])); - - int top = ARR_LEN(environment_stack); - ARR_RESIZE(environment_stack, top + 1); - environment_stack[top] = entry; + /* remember old declaration */ + stack_entry_t entry; + entry.symbol = symbol; + entry.old_declaration = symbol->declaration; + entry.namespc = (unsigned short) namespc; + ARR_APP1(stack_entry_t, *stack_ptr, entry); - entry->old_declaration = symbol->declaration; - entry->old_context = symbol->context; - entry->symbol = symbol; - symbol->declaration = declaration; - symbol->context = context; + /* replace/add declaration into declaration list of the symbol */ + if(symbol->declaration == NULL) { + symbol->declaration = declaration; + } else { + declaration_t *iter_last = NULL; + declaration_t *iter = symbol->declaration; + for( ; iter != NULL; iter_last = iter, iter = iter->symbol_next) { + /* replace an entry? */ + if(iter->namespc == namespc) { + if(iter_last == NULL) { + symbol->declaration = declaration; + } else { + iter_last->symbol_next = declaration; + } + declaration->symbol_next = iter->symbol_next; + break; + } + } + if(iter == NULL) { + assert(iter_last->symbol_next == NULL); + iter_last->symbol_next = declaration; + } + } return declaration; } +static declaration_t *environment_push(declaration_t *declaration) +{ + assert(declaration->source_position.input_name != NULL); + return stack_push(&environment_stack, declaration, context); +} + +static declaration_t *label_push(declaration_t *declaration) +{ + return stack_push(&label_stack, declaration, ¤t_function->context); +} + /** * pops symbols from the environment stack until @p new_top is the top element */ -static inline void environment_pop_to(size_t new_top) +static void stack_pop_to(stack_entry_t **stack_ptr, size_t new_top) { - environment_entry_t *entry = NULL; - size_t top = ARR_LEN(environment_stack); - size_t i; + stack_entry_t *stack = *stack_ptr; + size_t top = ARR_LEN(stack); + size_t i; + assert(new_top <= top); if(new_top == top) return; - assert(new_top < top); - i = top; - do { - entry = environment_stack[i - 1]; + for(i = top; i > new_top; --i) { + stack_entry_t *entry = & stack[i - 1]; + + declaration_t *old_declaration = entry->old_declaration; + symbol_t *symbol = entry->symbol; + namespace_t namespc = (namespace_t)entry->namespc; + + /* 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(*stack_ptr, (int) new_top); +} + +static void environment_pop_to(size_t new_top) +{ + stack_pop_to(&environment_stack, new_top); +} + +static void label_pop_to(size_t new_top) +{ + stack_pop_to(&label_stack, new_top); +} + + +static 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); + const atomic_type_t *atomic_type = &type->atomic; + atomic_type_type_t atype = atomic_type->atype; + return atype; +} + +static type_t *promote_integer(type_t *type) +{ + if(get_rank(type) < ATOMIC_TYPE_INT) + type = type_int; + + return type; +} + +static expression_t *create_cast_expression(expression_t *expression, + type_t *dest_type) +{ + expression_t *cast = allocate_expression_zero(EXPR_UNARY); + + cast->unary.type = UNEXPR_CAST; + cast->unary.value = expression; + cast->base.datatype = dest_type; + + return cast; +} + +static bool is_null_expression(const expression_t *const expression) +{ + if (expression->type != EXPR_CONST) + return false; + + type_t *const type = skip_typeref(expression->base.datatype); + if (!is_type_integer(type)) + return false; + + return expression->conste.v.int_value == 0; +} + +static expression_t *create_implicit_cast(expression_t *expression, + type_t *dest_type) +{ + type_t *source_type = expression->base.datatype; + + if(source_type == NULL) + return expression; + + source_type = skip_typeref(source_type); + dest_type = skip_typeref(dest_type); + + if(source_type == dest_type) + return expression; + + switch (dest_type->type) { + case TYPE_ENUM: + /* TODO warning for implicitly converting to enum */ + case TYPE_ATOMIC: + if (source_type->type != TYPE_ATOMIC && + source_type->type != TYPE_ENUM) { + panic("casting of non-atomic types not implemented yet"); + } + + if(is_type_floating(dest_type) && !is_type_scalar(source_type)) { + type_error_incompatible("can't cast types", + expression->base.source_position, source_type, + dest_type); + return expression; + } + + return create_cast_expression(expression, dest_type); + + case TYPE_POINTER: + switch (source_type->type) { + case TYPE_ATOMIC: + if (is_null_expression(expression)) { + return create_cast_expression(expression, dest_type); + } + break; + + case TYPE_POINTER: + if (pointers_compatible(source_type, dest_type)) { + return create_cast_expression(expression, dest_type); + } + break; - symbol_t *symbol = entry->symbol; + case TYPE_ARRAY: { + array_type_t *array_type = &source_type->array; + pointer_type_t *pointer_type = &dest_type->pointer; + if (types_compatible(array_type->element_type, + pointer_type->points_to)) { + return create_cast_expression(expression, dest_type); + } + break; + } - symbol->declaration = entry->old_declaration; - symbol->context = entry->old_context; + default: + panic("casting of non-atomic types not implemented yet"); + } - --i; - } while(i != new_top); - obstack_free(&environment_obstack, entry); + type_error_incompatible("can't implicitly cast types", + expression->base.source_position, source_type, dest_type); + return expression; - ARR_SHRINKLEN(environment_stack, (int) new_top); + default: + panic("casting of non-atomic types not implemented yet"); + } } +/** Implements the rules from § 6.5.16.1 */ +static void semantic_assign(type_t *orig_type_left, expression_t **right, + const char *context) +{ + type_t *orig_type_right = (*right)->base.datatype; + + if(orig_type_right == NULL) + return; + + type_t *const type_left = skip_typeref(orig_type_left); + type_t *const type_right = skip_typeref(orig_type_right); + + if ((is_type_arithmetic(type_left) && is_type_arithmetic(type_right)) || + (is_type_pointer(type_left) && is_null_expression(*right)) || + (is_type_atomic(type_left, ATOMIC_TYPE_BOOL) + && is_type_pointer(type_right))) { + *right = create_implicit_cast(*right, type_left); + return; + } + + if (is_type_pointer(type_left) && is_type_pointer(type_right)) { + pointer_type_t *pointer_type_left = &type_left->pointer; + pointer_type_t *pointer_type_right = &type_right->pointer; + type_t *points_to_left = pointer_type_left->points_to; + type_t *points_to_right = pointer_type_right->points_to; + + points_to_left = skip_typeref(points_to_left); + points_to_right = skip_typeref(points_to_right); + + if(!is_type_atomic(points_to_left, ATOMIC_TYPE_VOID) + && !is_type_atomic(points_to_right, ATOMIC_TYPE_VOID) + && !types_compatible(points_to_left, points_to_right)) { + goto incompatible_assign_types; + } + + /* the left type has all qualifiers from the right type */ + unsigned missing_qualifiers + = points_to_right->base.qualifiers & ~points_to_left->base.qualifiers; + if(missing_qualifiers != 0) { + parser_print_error_prefix(); + fprintf(stderr, "destination type "); + print_type_quoted(type_left); + fprintf(stderr, " in %s from type ", context); + print_type_quoted(type_right); + fprintf(stderr, " lacks qualifiers '"); + print_type_qualifiers(missing_qualifiers); + fprintf(stderr, "' in pointed-to type\n"); + return; + } + + *right = create_implicit_cast(*right, type_left); + return; + } + + if (is_type_compound(type_left) + && types_compatible(type_left, type_right)) { + *right = create_implicit_cast(*right, type_left); + return; + } +incompatible_assign_types: + /* TODO: improve error message */ + parser_print_error_prefix(); + fprintf(stderr, "incompatible types in %s\n", context); + parser_print_error_prefix(); + print_type_quoted(type_left); + fputs(" <- ", stderr); + print_type_quoted(type_right); + fputs("\n", stderr); +} static expression_t *parse_constant_expression(void) { @@ -417,18 +882,19 @@ static expression_t *parse_assignment_expression(void) return parse_sub_expression(2); } -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; + unsigned char storage_class; + bool is_inline; + type_t *type; }; +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) { assert(token.type == T_STRING_LITERAL); @@ -448,7 +914,7 @@ static void parse_attributes(void) { while(true) { switch(token.type) { - case T___attribute__: + case T___attribute__: { next_token(); expect_void('('); @@ -471,6 +937,7 @@ static void parse_attributes(void) } } break; + } case T_asm: next_token(); expect_void('('); @@ -493,6 +960,7 @@ attributes_finished: ; } +#if 0 static designator_t *parse_designation(void) { if(token.type != '[' && token.type != '.') @@ -514,7 +982,7 @@ static designator_t *parse_designation(void) designator = allocate_ast_zero(sizeof(designator[0])); next_token(); if(token.type != T_IDENTIFIER) { - parse_error_expected("problem while parsing designator", + parse_error_expected("while parsing designator", T_IDENTIFIER, 0); return NULL; } @@ -535,146 +1003,343 @@ static designator_t *parse_designation(void) last = designator; } } +#endif -static initializer_t *parse_initializer_list(void); - -static initializer_t *parse_initializer(void) +static initializer_t *initializer_from_string(array_type_t *type, + const char *string) { - designator_t *designator = parse_designation(); + /* TODO: check len vs. size of array type */ + (void) type; - initializer_t *result; - if(token.type == '{') { - result = parse_initializer_list(); - } else { - result = allocate_ast_zero(sizeof(result[0])); - result->type = INITIALIZER_VALUE; - result->v.value = parse_assignment_expression(); - } - result->designator = designator; + initializer_t *initializer = allocate_initializer(INITIALIZER_STRING); + initializer->string.string = string; - return result; + return initializer; } -static initializer_t *parse_initializer_list(void) +static initializer_t *initializer_from_expression(type_t *type, + expression_t *expression) { - eat('{'); + /* TODO check that expression is a constant expression */ - initializer_t *result = allocate_ast_zero(sizeof(result[0])); - result->type = INITIALIZER_LIST; + /* § 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 = &type->array; + type_t *element_type = array_type->element_type; - initializer_t *last = NULL; - while(1) { - initializer_t *initializer = parse_initializer(); - if(last != NULL) { - last->next = initializer; - } else { - result->v.list = initializer; - } - last = initializer; + if(element_type->type == TYPE_ATOMIC) { + atomic_type_t *atomic_type = &element_type->atomic; + atomic_type_type_t atype = atomic_type->atype; - if(token.type == '}') - break; + /* TODO handle wide strings */ + if(atype == ATOMIC_TYPE_CHAR + || atype == ATOMIC_TYPE_SCHAR + || atype == ATOMIC_TYPE_UCHAR) { - if(token.type != ',') { - parse_error_expected("problem while parsing initializer list", - ',', '}', 0); - eat_block(); - return result; + string_literal_expression_t *literal = &expression->string; + return initializer_from_string(array_type, literal->value); + } } - eat(','); - - if(token.type == '}') - break; } - expect('}'); + semantic_assign(type, &expression, "initializer"); + + initializer_t *result = allocate_initializer(INITIALIZER_VALUE); + result->value.value = expression; return result; } -static compound_type_t *find_compound_type(compound_type_t *types, - const symbol_t *symbol) -{ - compound_type_t *type = types; - for( ; type != NULL; type = type->next) { - if(type->symbol == symbol) - return type; - } - - return NULL; -} +static initializer_t *parse_sub_initializer(type_t *type, + expression_t *expression, + type_t *expression_type); -static type_t *parse_compound_type_specifier(bool is_struct) +static initializer_t *parse_sub_initializer_elem(type_t *type) { - if(is_struct) { - eat(T_struct); - } else { - eat(T_union); + if(token.type == '{') { + return parse_sub_initializer(type, NULL, NULL); } - symbol_t *symbol = NULL; - compound_type_t *compound_type = NULL; + expression_t *expression = parse_assignment_expression(); + type_t *expression_type = skip_typeref(expression->base.datatype); - if(token.type == T_IDENTIFIER) { - symbol = token.v.symbol; - next_token(); + return parse_sub_initializer(type, expression, expression_type); +} - if(context != NULL) { - if(is_struct) { - compound_type = find_compound_type(context->structs, symbol); - } else { - compound_type = find_compound_type(context->unions, symbol); +static bool had_initializer_brace_warning; + +static initializer_t *parse_sub_initializer(type_t *type, + expression_t *expression, + type_t *expression_type) +{ + if(is_type_scalar(type)) { + /* there might be extra {} hierarchies */ + if(token.type == '{') { + next_token(); + if(!had_initializer_brace_warning) { + parse_warning("braces around scalar initializer"); + had_initializer_brace_warning = true; } - } - } 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); + initializer_t *result = parse_sub_initializer(type, NULL, NULL); + if(token.type == ',') { + next_token(); + /* TODO: warn about excessive elements */ + } + expect_block('}'); + return result; } - return NULL; + if(expression == NULL) { + expression = parse_assignment_expression(); + } + return initializer_from_expression(type, expression); } - if(compound_type == NULL) { - compound_type = allocate_type_zero(sizeof(compound_type[0])); + /* TODO: ignore qualifiers, comparing pointers is probably + * not correct */ + if(expression != NULL && expression_type == type) { + initializer_t *result = allocate_initializer(INITIALIZER_VALUE); - if(is_struct) { - compound_type->type.type = TYPE_COMPOUND_STRUCT; - } else { - compound_type->type.type = TYPE_COMPOUND_UNION; + if(type != NULL) { + semantic_assign(type, &expression, "initializer"); } - compound_type->source_position = token.source_position; - compound_type->symbol = symbol; + result->value.value = expression; + + return result; } + bool read_paren = false; if(token.type == '{') { - if(compound_type->defined) { + 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 = &type->array; + type_t *element_type = array_type->element_type; + element_type = skip_typeref(element_type); + + initializer_t *sub; + had_initializer_brace_warning = false; + if(expression == NULL) { + sub = parse_sub_initializer_elem(element_type); + } else { + sub = parse_sub_initializer(element_type, expression, + expression_type); + } + + /* didn't match the subtypes -> try the parent type */ + if(sub == NULL) { + assert(!read_paren); + return NULL; + } + + elems = NEW_ARR_F(initializer_t*, 0); + ARR_APP1(initializer_t*, elems, sub); + + while(true) { + if(token.type == '}') + break; + expect_block(','); + if(token.type == '}') + break; + + sub = parse_sub_initializer(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; + } + ARR_APP1(initializer_t*, elems, sub); + } + } else { + assert(type->type == TYPE_COMPOUND_STRUCT + || type->type == TYPE_COMPOUND_UNION); + compound_type_t *compound_type = &type->compound; + context_t *context = & compound_type->declaration->context; + + declaration_t *first = context->declarations; + if(first == NULL) + return NULL; + type_t *first_type = first->type; + first_type = skip_typeref(first_type); + + 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); + } + + /* 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(','); + if(token.type == '}') + break; + + type_t *iter_type = iter->type; + iter_type = skip_typeref(iter_type); + + 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; + + initializer_list_t *init = allocate_ast_zero(sizeof(init[0]) + elems_size); + + init->initializer.type = INITIALIZER_LIST; + init->len = len; + memcpy(init->initializers, elems, elems_size); + DEL_ARR_F(elems); + + result = (initializer_t*) init; + + if(read_paren) { + if(token.type == ',') + next_token(); + expect('}'); + } + return result; +} + +static initializer_t *parse_initializer(type_t *type) +{ + initializer_t *result; + + type = skip_typeref(type); + + if(token.type != '{') { + expression_t *expression = parse_assignment_expression(); + return initializer_from_expression(type, expression); + } + + if(is_type_scalar(type)) { + /* § 6.7.8.11 */ + eat('{'); + + expression_t *expression = parse_assignment_expression(); + result = initializer_from_expression(type, expression); + + if(token.type == ',') + next_token(); + + expect('}'); + return result; + } else { + result = parse_sub_initializer(type, NULL, NULL); + } + + return result; +} + + + +static declaration_t *parse_compound_type_specifier(bool is_struct) +{ + if(is_struct) { + eat(T_struct); + } else { + eat(T_union); + } + + symbol_t *symbol = NULL; + declaration_t *declaration = NULL; + + if (token.type == T___attribute__) { + /* TODO */ + parse_attributes(); + } + + if(token.type == T_IDENTIFIER) { + symbol = token.v.symbol; + next_token(); + + if(is_struct) { + declaration = get_declaration(symbol, NAMESPACE_STRUCT); + } else { + declaration = get_declaration(symbol, NAMESPACE_UNION); + } + } else if(token.type != '{') { + if(is_struct) { + parse_error_expected("while parsing struct type specifier", + T_IDENTIFIER, '{', 0); + } else { + parse_error_expected("while parsing union type specifier", + T_IDENTIFIER, '{', 0); + } + + return NULL; + } + + if(declaration == NULL) { + declaration = allocate_ast_zero(sizeof(declaration[0])); + + if(is_struct) { + declaration->namespc = NAMESPACE_STRUCT; + } else { + declaration->namespc = NAMESPACE_UNION; + } + declaration->source_position = token.source_position; + declaration->symbol = symbol; + record_declaration(declaration); + } + + if(token.type == '{') { + if(declaration->init.is_defined) { + assert(symbol != NULL); parser_print_error_prefix(); fprintf(stderr, "multiple definition of %s %s\n", is_struct ? "struct" : "union", symbol->string); - compound_type->context.declarations = NULL; + declaration->context.declarations = NULL; } - compound_type->defined = 1; + declaration->init.is_defined = true; int top = environment_top(); context_t *last_context = context; - set_context(&compound_type->context); + set_context(& declaration->context); parse_compound_type_entries(); parse_attributes(); - assert(context == &compound_type->context); + assert(context == & declaration->context); set_context(last_context); environment_pop_to(top); } - return (type_t*) compound_type; + return declaration; } -static void parse_enum_entries(void) +static void parse_enum_entries(enum_type_t *const enum_type) { eat('{'); @@ -688,19 +1353,21 @@ static void parse_enum_entries(void) declaration_t *entry = allocate_ast_zero(sizeof(entry[0])); if(token.type != T_IDENTIFIER) { - parse_error_expected("problem while parsing enum entry", - T_IDENTIFIER, 0); + parse_error_expected("while parsing enum entry", T_IDENTIFIER, 0); eat_block(); return; } entry->storage_class = STORAGE_CLASS_ENUM_ENTRY; + entry->type = (type_t*) enum_type; entry->symbol = token.v.symbol; entry->source_position = token.source_position; next_token(); if(token.type == '=') { next_token(); - entry->initializer = parse_initializer(); + entry->init.enum_value = parse_constant_expression(); + + /* TODO semantic */ } record_declaration(entry); @@ -713,80 +1380,75 @@ static void parse_enum_entries(void) expect_void('}'); } -static enum_type_t *find_enum_type(enum_type_t *types, const symbol_t *symbol) -{ - enum_type_t *type = types; - for( ; type != NULL; type = type->next) { - if(type->symbol == symbol) - return type; - } - - return NULL; -} - static type_t *parse_enum_specifier(void) { eat(T_enum); - symbol_t *symbol = NULL; - enum_type_t *enum_type = NULL; + declaration_t *declaration; + symbol_t *symbol; if(token.type == T_IDENTIFIER) { symbol = token.v.symbol; next_token(); - if(context != NULL) { - enum_type = find_enum_type(context->enums, symbol); - } + declaration = get_declaration(symbol, NAMESPACE_ENUM); } else if(token.type != '{') { - parse_error_expected("problem while parsing enum type specifier", + parse_error_expected("while parsing enum type specifier", T_IDENTIFIER, '{', 0); return NULL; + } else { + declaration = NULL; + symbol = NULL; } - if(enum_type == NULL) { - enum_type = allocate_type_zero(sizeof(enum_type[0])); - enum_type->type.type = TYPE_ENUM; - enum_type->source_position = token.source_position; - enum_type->symbol = symbol; + if(declaration == NULL) { + declaration = allocate_ast_zero(sizeof(declaration[0])); + + declaration->namespc = NAMESPACE_ENUM; + declaration->source_position = token.source_position; + declaration->symbol = symbol; } + type_t *const type = allocate_type_zero(TYPE_ENUM); + type->enumt.declaration = declaration; + if(token.type == '{') { - if(enum_type->defined) { + if(declaration->init.is_defined) { parser_print_error_prefix(); fprintf(stderr, "multiple definitions of enum %s\n", symbol->string); - enum_type->entries_begin = NULL; - enum_type->entries_end = NULL; } - enum_type->defined = 1; + record_declaration(declaration); + declaration->init.is_defined = 1; - declaration_t *before = last_declaration; - - parse_enum_entries(); + parse_enum_entries(&type->enumt); parse_attributes(); - - if(before == NULL) { - enum_type->entries_begin = context->declarations; - } else { - enum_type->entries_begin = before->next; - } - enum_type->entries_end = last_declaration; } - return (type_t*) enum_type; + return type; +} + +/** + * if a symbol is a typedef to another type, return true + */ +static bool is_typedef_symbol(symbol_t *symbol) +{ + const declaration_t *const declaration = + get_declaration(symbol, NAMESPACE_NORMAL); + return + declaration != NULL && + declaration->storage_class == STORAGE_CLASS_TYPEDEF; } static type_t *parse_typeof(void) { eat(T___typeof__); - type_t *result; + type_t *type; expect('('); - declaration_t *declaration; - expression_t *expression; + expression_t *expression = NULL; restart: switch(token.type) { @@ -799,29 +1461,31 @@ restart: goto restart; case T_IDENTIFIER: - declaration = token.v.symbol->declaration; - if(declaration != NULL - && declaration->storage_class == STORAGE_CLASS_TYPEDEF) { - result = parse_typename(); - break; + if(is_typedef_symbol(token.v.symbol)) { + type = parse_typename(); + } else { + expression = parse_expression(); + type = expression->base.datatype; } - expression = parse_expression(); - result = expression->datatype; break; TYPENAME_START - result = parse_typename(); + type = parse_typename(); break; default: expression = parse_expression(); - result = expression->datatype; + type = expression->base.datatype; break; } expect(')'); - return result; + type_t *typeof_type = allocate_type_zero(TYPE_TYPEOF); + typeof_type->typeoft.expression = expression; + typeof_type->typeoft.typeof_type = type; + + return typeof_type; } typedef enum { @@ -838,33 +1502,39 @@ typedef enum { SPECIFIER_VOID = 1 << 10, #ifdef PROVIDE_COMPLEX SPECIFIER_COMPLEX = 1 << 11, -#endif -#ifdef PROVIDE_IMAGINARY 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; + type_t *type = allocate_type_zero(TYPE_BUILTIN); + type->builtin.symbol = symbol; + /* TODO... */ + type->builtin.real_type = type_int; - type_t *result = typehash_insert((type_t*) type); - if(result != (type_t*) type) { - free_type(type); - } + return type; +} - return result; +static type_t *get_typedef_type(symbol_t *symbol) +{ + declaration_t *declaration = get_declaration(symbol, NAMESPACE_NORMAL); + if(declaration == NULL + || declaration->storage_class != STORAGE_CLASS_TYPEDEF) + return NULL; + + type_t *type = allocate_type_zero(TYPE_TYPEDEF); + type->typedeft.declaration = declaration; + + return type; } static void parse_declaration_specifiers(declaration_specifiers_t *specifiers) { - declaration_t *declaration; - type_t *type = NULL; - unsigned type_qualifiers = 0; - unsigned type_specifiers = 0; - int newtype = 0; + type_t *type = NULL; + unsigned type_qualifiers = 0; + unsigned type_specifiers = 0; + int newtype = 0; while(true) { switch(token.type) { @@ -886,6 +1556,27 @@ static void parse_declaration_specifiers(declaration_specifiers_t *specifiers) MATCH_STORAGE_CLASS(T_auto, STORAGE_CLASS_AUTO) MATCH_STORAGE_CLASS(T_register, STORAGE_CLASS_REGISTER) + case T___thread: + switch (specifiers->storage_class) { + case STORAGE_CLASS_NONE: + specifiers->storage_class = STORAGE_CLASS_THREAD; + break; + + case STORAGE_CLASS_EXTERN: + specifiers->storage_class = STORAGE_CLASS_THREAD_EXTERN; + break; + + case STORAGE_CLASS_STATIC: + specifiers->storage_class = STORAGE_CLASS_THREAD_STATIC; + break; + + default: + parse_error("multiple storage classes in declaration specifiers"); + break; + } + next_token(); + break; + /* type qualifiers */ #define MATCH_TYPE_QUALIFIER(token, qualifier) \ case token: \ @@ -896,7 +1587,6 @@ static void parse_declaration_specifiers(declaration_specifiers_t *specifiers) 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); case T___extension__: /* TODO */ @@ -925,10 +1615,13 @@ static void parse_declaration_specifiers(declaration_specifiers_t *specifiers) 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_inline: + next_token(); + specifiers->is_inline = true; + break; + case T_long: next_token(); if(type_specifiers & SPECIFIER_LONG_LONG) { @@ -940,13 +1633,20 @@ static void parse_declaration_specifiers(declaration_specifiers_t *specifiers) } break; - /* TODO: if type != NULL for the following rules issue an error */ - case T_struct: - type = parse_compound_type_specifier(true); + /* TODO: if type != NULL for the following rules should issue + * an error */ + case T_struct: { + type = allocate_type_zero(TYPE_COMPOUND_STRUCT); + + type->compound.declaration = parse_compound_type_specifier(true); break; - case T_union: - type = parse_compound_type_specifier(false); + } + case T_union: { + type = allocate_type_zero(TYPE_COMPOUND_STRUCT); + + type->compound.declaration = parse_compound_type_specifier(false); break; + } case T_enum: type = parse_enum_specifier(); break; @@ -963,17 +1663,16 @@ static void parse_declaration_specifiers(declaration_specifiers_t *specifiers) parse_attributes(); break; - case T_IDENTIFIER: - declaration = token.v.symbol->declaration; - if(declaration == NULL || - declaration->storage_class != STORAGE_CLASS_TYPEDEF) { + case T_IDENTIFIER: { + type_t *typedef_type = get_typedef_type(token.v.symbol); + + if(typedef_type == NULL) goto finish_specifiers; - } - type = declaration->type; - assert(type != NULL); next_token(); + type = typedef_type; break; + } /* function specifier */ default: @@ -1063,8 +1762,6 @@ 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; @@ -1096,31 +1793,28 @@ finish_specifiers: atomic_type = ATOMIC_TYPE_INVALID; } - atomic_type_t *atype = allocate_type_zero(sizeof(atype[0])); - atype->type.type = TYPE_ATOMIC; - atype->atype = atomic_type; - newtype = 1; - - type = (type_t*) atype; + type = allocate_type_zero(TYPE_ATOMIC); + type->atomic.atype = atomic_type; + newtype = 1; } else { if(type_specifiers != 0) { parse_error("multiple datatypes in declaration"); } } - type->qualifiers = type_qualifiers; + type->base.qualifiers = type_qualifiers; type_t *result = typehash_insert(type); - if(newtype && result != (type_t*) type) { + if(newtype && result != type) { free_type(type); } specifiers->type = result; } -static type_qualifier_t parse_type_qualifiers(void) +static type_qualifiers_t parse_type_qualifiers(void) { - type_qualifier_t type_qualifiers = 0; + type_qualifiers_t type_qualifiers = TYPE_QUALIFIER_NONE; while(true) { switch(token.type) { @@ -1128,7 +1822,6 @@ 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; @@ -1140,11 +1833,15 @@ 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; } + declaration_t *declaration = allocate_ast_zero(sizeof(declaration[0])); + declaration->symbol = token.v.symbol; + next_token(); + if(token.type != ',') break; next_token(); @@ -1158,21 +1855,31 @@ 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 = &declaration->type->array; + type_t *element_type = arr_type->element_type; + declaration->type = make_pointer_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; } @@ -1187,11 +1894,11 @@ 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) { @@ -1205,7 +1912,8 @@ static declaration_t *parse_parameters(method_type_t *type) DECLARATION_START declaration = parse_parameter(); - parameter = allocate_type_zero(sizeof(parameter[0])); + parameter = obstack_alloc(type_obst, sizeof(parameter[0])); + memset(parameter, 0, sizeof(parameter[0])); parameter->type = declaration->type; if(last_parameter != NULL) { @@ -1229,8 +1937,9 @@ 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; @@ -1243,19 +1952,19 @@ struct construct_type_t { typedef struct parsed_pointer_t parsed_pointer_t; struct parsed_pointer_t { construct_type_t construct_type; - type_qualifier_t type_qualifiers; + type_qualifiers_t type_qualifiers; }; -typedef struct construct_method_type_t construct_method_type_t; -struct construct_method_type_t { +typedef struct construct_function_type_t construct_function_type_t; +struct construct_function_type_t { construct_type_t construct_type; - method_type_t *method_type; + type_t *function_type; }; typedef struct parsed_array_t parsed_array_t; struct parsed_array_t { construct_type_t construct_type; - type_qualifier_t type_qualifiers; + type_qualifiers_t type_qualifiers; bool is_static; bool is_variable; expression_t *size; @@ -1273,7 +1982,8 @@ 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; } @@ -1284,13 +1994,14 @@ 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; next_token(); } - type_qualifier_t type_qualifiers = parse_type_qualifiers(); + type_qualifiers_t type_qualifiers = parse_type_qualifiers(); if(type_qualifiers != 0) { if(token.type == T_static) { array->is_static = true; @@ -1311,44 +2022,47 @@ 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; + type_t *type = allocate_type_zero(TYPE_FUNCTION); - declaration_t *parameters = parse_parameters(method_type); + declaration_t *parameters = parse_parameters(&type->function); 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, - int may_be_abstract) + bool may_be_abstract) { - construct_type_t *result = NULL; - construct_type_t *last = NULL; + /* construct a single linked list of construct_type_t's which describe + * how to construct the final declarator type */ + construct_type_t *first = NULL; + construct_type_t *last = NULL; + /* pointers */ while(token.type == '*') { construct_type_t *type = parse_pointer_declarator(); - if(last != NULL) { - last->next = type; + + if(last == NULL) { + first = type; + last = type; } else { - result = type; + last->next = type; + last = type; } - last = type; } /* TODO: find out if this is correct */ @@ -1374,15 +2088,24 @@ 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; } + construct_type_t *p = last; + 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(); @@ -1391,27 +2114,32 @@ static construct_type_t *parse_inner_declarator(declaration_t *declaration, goto declarator_finished; } - if(last != NULL) { - last->next = type; + /* insert in the middle of the list (behind p) */ + if(p != NULL) { + type->next = p->next; + p->next = type; } else { - result = type; + type->next = first; + first = type; + } + if(last == p) { + last = type; } - last = type; } declarator_finished: parse_attributes(); - if(inner_types != NULL) { - if(last != NULL) { - last->next = inner_types; - } else { - result = inner_types; - } - last = inner_types; + /* append inner_types at the end of the list, we don't to set last anymore + * as it's not needed anymore */ + if(last == NULL) { + assert(first == NULL); + first = inner_types; + } else { + last->next = inner_types; } - return result; + return first; } static type_t *construct_declarator_type(construct_type_t *construct_list, @@ -1419,49 +2147,53 @@ 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; - switch(iter->type) { - case CONSTRUCT_METHOD: - construct_method_type = (construct_method_type_t*) iter; - method_type = construct_method_type->method_type; + case CONSTRUCT_INVALID: + panic("invalid type construction found"); + case CONSTRUCT_FUNCTION: { + construct_function_type_t *construct_function_type + = (construct_function_type_t*) iter; + + type_t *function_type = construct_function_type->function_type; - method_type->result_type = type; - type = (type_t*) method_type; + function_type->function.result_type = type; + + type = function_type; break; + } - case CONSTRUCT_POINTER: - parsed_pointer = (parsed_pointer_t*) iter; - pointer_type = allocate_type_zero(sizeof(pointer_type[0])); + case CONSTRUCT_POINTER: { + parsed_pointer_t *parsed_pointer = (parsed_pointer_t*) iter; + type_t *pointer_type = allocate_type_zero(TYPE_POINTER); + pointer_type->pointer.points_to = type; + pointer_type->base.qualifiers = parsed_pointer->type_qualifiers; - pointer_type->type.type = TYPE_POINTER; - pointer_type->points_to = type; - pointer_type->type.qualifiers = parsed_pointer->type_qualifiers; - type = (type_t*) pointer_type; + type = pointer_type; break; + } + + case CONSTRUCT_ARRAY: { + parsed_array_t *parsed_array = (parsed_array_t*) iter; + type_t *array_type = allocate_type_zero(TYPE_ARRAY); - case CONSTRUCT_ARRAY: - parsed_array = (parsed_array_t*) iter; - array_type = allocate_type_zero(sizeof(array_type[0])); - - array_type->type.type = TYPE_ARRAY; - array_type->element_type = type; - array_type->type.qualifiers = parsed_array->type_qualifiers; - array_type->is_static = parsed_array->is_static; - array_type->is_variable = parsed_array->is_variable; - array_type->size = parsed_array->size; - type = (type_t*) array_type; + array_type->base.qualifiers = parsed_array->type_qualifiers; + array_type->array.element_type = type; + array_type->array.is_static = parsed_array->is_static; + array_type->array.is_variable = parsed_array->is_variable; + array_type->array.size = parsed_array->size; + + type = array_type; break; } + } - type_t *hashed_type = typehash_insert((type_t*) type); + type_t *hashed_type = typehash_insert(type); if(hashed_type != type) { - free_type(type); + /* the function type was constructed earlier freeing it here will + * destroy other types... */ + if(iter->type != CONSTRUCT_FUNCTION) { + free_type(type); + } type = hashed_type; } } @@ -1469,50 +2201,54 @@ 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; @@ -1526,39 +2262,82 @@ 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); } - ndeclaration->initializer = parse_initializer(); + initializer_t *initializer = parse_initializer(type); + + if(type->type == TYPE_ARRAY && initializer != NULL) { + array_type_t *array_type = &type->array; + + if(array_type->size == NULL) { + expression_t *cnst = allocate_expression_zero(EXPR_CONST); + + cnst->base.datatype = type_size_t; + + if(initializer->type == INITIALIZER_LIST) { + initializer_list_t *list = &initializer->list; + cnst->conste.v.int_value = list->len; + } else { + assert(initializer->type == INITIALIZER_STRING); + initializer_string_t *string = &initializer->string; + cnst->conste.v.int_value = strlen(string->string) + 1; + } + + array_type->size = 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; + } + function_type_t *function_type = &type->function; + /* § 6.7.5.3 (14) a function definition with () */ + if(function_type->unspecified_parameters) { + type_t *duplicate = duplicate_type(type); + duplicate->function.unspecified_parameters = false; + + type = typehash_insert(duplicate); + if(type != duplicate) { + obstack_free(type_obst, duplicate); + } + function_type = &type->function; } - if(declaration->initializer != NULL) { + if(declaration->init.statement != NULL) { parser_error_multiple_definition(declaration, ndeclaration); } if(ndeclaration != declaration) { @@ -1573,16 +2352,24 @@ 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; } @@ -1602,10 +2389,9 @@ 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); @@ -1630,8 +2416,6 @@ 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); @@ -1644,12 +2428,44 @@ 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 + = &specifiers.type->compound; + 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); @@ -1689,6 +2505,13 @@ 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_expression_zero(EXPR_INVALID); + expression->base.source_position = token.source_position; + return expression; +} + static expression_t *expected_expression_error(void) { parser_print_error_prefix(); @@ -1696,62 +2519,52 @@ 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) { - string_literal_t *cnst = allocate_ast_zero(sizeof(cnst[0])); + expression_t *cnst = allocate_expression_zero(EXPR_STRING_LITERAL); + cnst->base.datatype = type_string; + cnst->string.value = parse_string_literals(); - cnst->expression.type = EXPR_STRING_LITERAL; - cnst->expression.datatype = type_string; - cnst->value = parse_string_literals(); - - return (expression_t*) cnst; + return cnst; } 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->v.int_value = token.v.intvalue; + expression_t *cnst = allocate_expression_zero(EXPR_CONST); + cnst->base.datatype = token.datatype; + cnst->conste.v.int_value = token.v.intvalue; next_token(); - return (expression_t*) cnst; + return cnst; } static expression_t *parse_float_const(void) { - const_t *cnst = allocate_ast_zero(sizeof(cnst[0])); - - cnst->expression.type = EXPR_CONST; - cnst->expression.datatype = type_double; - cnst->v.float_value = token.v.floatvalue; + expression_t *cnst = allocate_expression_zero(EXPR_CONST); + cnst->base.datatype = token.datatype; + cnst->conste.v.float_value = token.v.floatvalue; next_token(); - return (expression_t*) cnst; + return cnst; } static declaration_t *create_implicit_function(symbol_t *symbol, const source_position_t source_position) { - method_type_t *method_type = allocate_type_zero(sizeof(method_type)); - - method_type->type.type = TYPE_METHOD; - method_type->result_type = type_int; - method_type->unspecified_parameters = true; + type_t *ntype = allocate_type_zero(TYPE_FUNCTION); + ntype->function.result_type = type_int; + ntype->function.unspecified_parameters = true; - type_t *type = typehash_insert((type_t*) method_type); - if(type != (type_t*) method_type) { - free_type(method_type); + type_t *type = typehash_insert(ntype); + if(type != ntype) { + free_type(ntype); } declaration_t *declaration = allocate_ast_zero(sizeof(declaration[0])); @@ -1761,31 +2574,141 @@ static declaration_t *create_implicit_function(symbol_t *symbol, declaration->symbol = symbol; declaration->source_position = source_position; - /* we have to violate the environment push/pop rules here and assign - * the declaration directly to its symbol and append it to the - * global context */ + /* 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); - symbol->declaration = declaration; - declaration->next = global_context->declarations; - global_context->declarations = declaration; + + context_t *last_context = context; + context = global_context; + + environment_push(declaration); + declaration->next = context->declarations; + context->declarations = declaration; + + context = last_context; return declaration; } +static type_t *make_function_1_type(type_t *result_type, type_t *argument_type) +{ + function_parameter_t *parameter + = obstack_alloc(type_obst, sizeof(parameter[0])); + memset(parameter, 0, sizeof(parameter[0])); + parameter->type = argument_type; + + type_t *type = allocate_type_zero(TYPE_FUNCTION); + type->function.result_type = result_type; + type->function.parameters = parameter; + + type_t *result = typehash_insert(type); + if(result != type) { + free_type(type); + } + + return result; +} + +static type_t *get_builtin_symbol_type(symbol_t *symbol) +{ + switch(symbol->ID) { + case T___builtin_alloca: + return make_function_1_type(type_void_ptr, type_size_t); + default: + panic("not implemented builtin symbol found"); + } +} + +/** + * performs automatic type cast as described in § 6.3.2.1 + */ +static type_t *automatic_type_conversion(type_t *type) +{ + if(type == NULL) + return NULL; + + if(type->type == TYPE_ARRAY) { + array_type_t *array_type = &type->array; + type_t *element_type = array_type->element_type; + unsigned qualifiers = array_type->type.qualifiers; + + return make_pointer_type(element_type, qualifiers); + } + + if(type->type == TYPE_FUNCTION) { + return make_pointer_type(type, TYPE_QUALIFIER_NONE); + } + + return type; +} + +/** + * reverts the automatic casts of array to pointer types and function + * to function-pointer types as defined § 6.3.2.1 + */ +type_t *revert_automatic_type_conversion(const expression_t *expression) +{ + if(expression->base.datatype == NULL) + return NULL; + + switch(expression->type) { + case EXPR_REFERENCE: { + const reference_expression_t *ref = &expression->reference; + return ref->declaration->type; + } + case EXPR_SELECT: { + const select_expression_t *select = &expression->select; + return select->compound_entry->type; + } + case EXPR_UNARY: { + const unary_expression_t *unary = &expression->unary; + if(unary->type == UNEXPR_DEREFERENCE) { + expression_t *value = unary->value; + type_t *type = skip_typeref(value->base.datatype); + pointer_type_t *pointer_type = &type->pointer; + + return pointer_type->points_to; + } + break; + } + case EXPR_BUILTIN_SYMBOL: { + const builtin_symbol_expression_t *builtin + = &expression->builtin_symbol; + return get_builtin_symbol_type(builtin->symbol); + } + case EXPR_ARRAY_ACCESS: { + const array_access_expression_t *array_access + = &expression->array_access; + const expression_t *array_ref = array_access->array_ref; + type_t *type_left = skip_typeref(array_ref->base.datatype); + assert(is_type_pointer(type_left)); + pointer_type_t *pointer_type = &type_left->pointer; + return pointer_type->points_to; + } + + default: + break; + } + + return expression->base.datatype; +} + static expression_t *parse_reference(void) { - reference_expression_t *ref = allocate_ast_zero(sizeof(ref[0])); + expression_t *expression = allocate_expression_zero(EXPR_REFERENCE); - ref->expression.type = EXPR_REFERENCE; - ref->symbol = token.v.symbol; + reference_expression_t *ref = &expression->reference; + ref->symbol = token.v.symbol; - declaration_t *declaration = ref->symbol->declaration; - source_position_t source_position = token.source_position; + declaration_t *declaration = get_declaration(ref->symbol, NAMESPACE_NORMAL); + + source_position_t source_position = token.source_position; next_token(); if(declaration == NULL) { #ifndef STRICT_C99 - /* an implicitely defined function */ + /* an implicitly defined function */ if(token.type == '(') { parser_print_prefix_pos(token.source_position); fprintf(stderr, "warning: implicit declaration of function '%s'\n", @@ -1798,30 +2721,34 @@ static expression_t *parse_reference(void) { parser_print_error_prefix(); fprintf(stderr, "unknown symbol '%s' found.\n", ref->symbol->string); - return (expression_t*) ref; + return expression; } } + type_t *type = declaration->type; + /* we always do the auto-type conversions; the & and sizeof parser contains + * code to revert this! */ + type = automatic_type_conversion(type); + ref->declaration = declaration; - ref->expression.datatype = declaration->type; + ref->expression.datatype = type; - return (expression_t*) ref; + return expression; } 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) { - unary_expression_t *cast = allocate_ast_zero(sizeof(cast[0])); + expression_t *cast = allocate_expression_zero(EXPR_UNARY); - cast->expression.type = EXPR_UNARY; - cast->type = UNEXPR_CAST; - cast->expression.source_position = token.source_position; + cast->unary.type = UNEXPR_CAST; + cast->base.source_position = token.source_position; type_t *type = parse_typename(); @@ -1830,45 +2757,51 @@ static expression_t *parse_cast(void) check_cast_allowed(value, type); - cast->expression.datatype = type; - cast->value = value; + cast->base.datatype = type; + cast->unary.value = value; - return (expression_t*) cast; + return cast; } 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(); + expression_t *expression = allocate_expression_zero(EXPR_STATEMENT); + + statement_t *statement = parse_compound_statement(); + expression->statement.statement = statement; + if(statement == NULL) { + expect(')'); + return NULL; + } + + assert(statement->type == STATEMENT_COMPOUND); + compound_statement_t *compound_statement = &statement->compound; /* 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->base.next) { + last_statement = iter; } if(last_statement->type == STATEMENT_EXPRESSION) { - const expression_statement_t *expression_statement = - (const expression_statement_t*) last_statement; - expression->expression.datatype - = expression_statement->expression->datatype; + const expression_statement_t *expression_statement + = &last_statement->expression; + expression->base.datatype + = expression_statement->expression->base.datatype; } else { - expression->expression.datatype = type_void; + expression->base.datatype = type_void; } expect(')'); - return (expression_t*) expression; + return expression; } static expression_t *parse_brace_expression(void) { eat('('); - declaration_t *declaration; switch(token.type) { case '{': /* gcc extension: a stement expression */ @@ -1878,9 +2811,7 @@ 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(); } } @@ -1893,10 +2824,16 @@ static expression_t *parse_brace_expression(void) static expression_t *parse_function_keyword(void) { - eat(T___FUNCTION__); + next_token(); /* TODO */ - string_literal_t *expression = allocate_ast_zero(sizeof(expression[0])); + if (current_function == NULL) { + parse_error("'__func__' used outside of a function"); + } + + string_literal_expression_t *expression + = allocate_ast_zero(sizeof(expression[0])); + expression->expression.type = EXPR_FUNCTION; expression->expression.datatype = type_string; expression->value = "TODO: FUNCTION"; @@ -1909,7 +2846,9 @@ static expression_t *parse_pretty_function_keyword(void) eat(T___PRETTY_FUNCTION__); /* TODO */ - string_literal_t *expression = allocate_ast_zero(sizeof(expression[0])); + string_literal_expression_t *expression + = allocate_ast_zero(sizeof(expression[0])); + expression->expression.type = EXPR_PRETTY_FUNCTION; expression->expression.datatype = type_string; expression->value = "TODO: PRETTY FUNCTION"; @@ -1922,7 +2861,7 @@ static designator_t *parse_designator(void) 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; @@ -1935,8 +2874,8 @@ static designator_t *parse_designator(void) 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; } @@ -1972,49 +2911,47 @@ static expression_t *parse_offsetof(void) { eat(T___builtin_offsetof); - offsetof_expression_t *expression - = allocate_ast_zero(sizeof(expression[0])); - expression->expression.type = EXPR_OFFSETOF; - expression->expression.datatype = type_size_t; + expression_t *expression = allocate_expression_zero(EXPR_OFFSETOF); + expression->base.datatype = type_size_t; expect('('); - expression->type = parse_typename(); + expression->offsetofe.type = parse_typename(); expect(','); - expression->designator = parse_designator(); + expression->offsetofe.designator = parse_designator(); expect(')'); - return (expression_t*) expression; + return 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; + expression_t *expression = allocate_expression_zero(EXPR_VA_ARG); expect('('); - expression->arg = parse_assignment_expression(); + expression->va_arge.arg = parse_assignment_expression(); expect(','); - expression->expression.datatype = parse_typename(); + expression->base.datatype = parse_typename(); expect(')'); - return (expression_t*) expression; + return expression; } static expression_t *parse_builtin_symbol(void) { - builtin_symbol_expression_t *expression - = allocate_ast_zero(sizeof(expression[0])); - expression->expression.type = EXPR_BUILTIN_SYMBOL; - - /* TODO: set datatype */ + expression_t *expression = allocate_expression_zero(EXPR_BUILTIN_SYMBOL); - expression->symbol = token.v.symbol; + symbol_t *symbol = token.v.symbol; + expression->builtin_symbol.symbol = symbol; next_token(); - return (expression_t*) expression; + type_t *type = get_builtin_symbol_type(symbol); + type = automatic_type_conversion(type); + + expression->base.datatype = type; + return expression; } static expression_t *parse_primary_expression(void) @@ -2029,6 +2966,7 @@ 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(); @@ -2036,6 +2974,7 @@ static expression_t *parse_primary_expression(void) 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_end: @@ -2051,366 +2990,972 @@ 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 make_invalid_expression(); +} + +static expression_t *parse_array_expression(unsigned precedence, + expression_t *left) +{ + (void) precedence; + + eat('['); + + expression_t *inside = parse_expression(); + + array_access_expression_t *array_access + = allocate_ast_zero(sizeof(array_access[0])); + + array_access->expression.type = EXPR_ARRAY_ACCESS; + + type_t *type_left = left->base.datatype; + type_t *type_inside = inside->base.datatype; + type_t *result_type = NULL; + + if(type_left != NULL && type_inside != NULL) { + type_left = skip_typeref(type_left); + type_inside = skip_typeref(type_inside); + + if(is_type_pointer(type_left)) { + pointer_type_t *pointer = &type_left->pointer; + result_type = pointer->points_to; + array_access->array_ref = left; + array_access->index = inside; + } else if(is_type_pointer(type_inside)) { + pointer_type_t *pointer = &type_inside->pointer; + result_type = pointer->points_to; + array_access->array_ref = inside; + array_access->index = left; + array_access->flipped = true; + } else { + parser_print_error_prefix(); + fprintf(stderr, "array access on object with non-pointer types "); + print_type_quoted(type_left); + fprintf(stderr, ", "); + print_type_quoted(type_inside); + fprintf(stderr, "\n"); + } + } else { + array_access->array_ref = left; + array_access->index = inside; + } + + if(token.type != ']') { + parse_error_expected("Problem while parsing array access", ']', 0); + return (expression_t*) array_access; + } + next_token(); + + result_type = automatic_type_conversion(result_type); + array_access->expression.datatype = result_type; + + 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); + expression->base.datatype = revert_automatic_type_conversion(expression); + + sizeof_expression->type = expression->base.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(); + + expression_t *select = allocate_expression_zero(EXPR_SELECT); + select->select.compound = compound; + + if(token.type != T_IDENTIFIER) { + parse_error_expected("while parsing select", T_IDENTIFIER, 0); + return select; + } + symbol_t *symbol = token.v.symbol; + select->select.symbol = symbol; + next_token(); + + type_t *orig_type = compound->base.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 = &type->pointer; + 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 = &type_left->compound; + 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(); + } + + /* we always do the auto-type conversions; the & and sizeof parser contains + * code to revert this! */ + type_t *expression_type = automatic_type_conversion(iter->type); + + select->select.compound_entry = iter; + select->base.datatype = expression_type; + return select; +} + +static expression_t *parse_call_expression(unsigned precedence, + expression_t *expression) +{ + (void) precedence; + expression_t *result = allocate_expression_zero(EXPR_CALL); + + call_expression_t *call = &result->call; + call->function = expression; + + function_type_t *function_type = NULL; + type_t *orig_type = expression->base.datatype; + if(orig_type != NULL) { + type_t *type = skip_typeref(orig_type); + + if(is_type_pointer(type)) { + pointer_type_t *pointer_type = &type->pointer; + + type = skip_typeref(pointer_type->points_to); + + if (type->type == TYPE_FUNCTION) { + function_type = &type->function; + call->expression.datatype = function_type->result_type; + } + } + if(function_type == NULL) { + parser_print_error_prefix(); + fputs("called object '", stderr); + print_expression(expression); + fputs("' (type ", stderr); + print_type_quoted(orig_type); + fputs(") is not a pointer to a function\n", stderr); + + function_type = NULL; + call->expression.datatype = NULL; + } + } + + /* 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->base.datatype; + type = skip_typeref(type); + + 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 result; +} + +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->base.datatype; + if(condition_type_orig != NULL) { + type_t *condition_type = skip_typeref(condition_type_orig); + if(condition_type != NULL && !is_type_scalar(condition_type)) { + type_error("expected a scalar type", + expression->base.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->base.datatype; + if(true_type == NULL) + return (expression_t*) conditional; + type_t *const false_type = f_expr->base.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->base.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->base.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->base.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->base.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->base.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->base.datatype; + if(orig_type == NULL) + return; + + type_t *type = skip_typeref(orig_type); + if(!is_type_pointer(type)) { + parser_print_error_prefix(); + fputs("Unary '*' needs pointer or arrray type, but type ", stderr); + print_type_quoted(orig_type); + fputs(" given.\n", stderr); + return; + } + + pointer_type_t *pointer_type = &type->pointer; + type_t *result_type = pointer_type->points_to; + + result_type = automatic_type_conversion(result_type); + expression->expression.datatype = result_type; +} + +static void semantic_take_addr(unary_expression_t *expression) +{ + expression_t *value = expression->value; + value->base.datatype = revert_automatic_type_conversion(value); + + type_t *orig_type = value->base.datatype; + if(orig_type == NULL) + return; + + if(value->type == EXPR_REFERENCE) { + reference_expression_t *reference = (reference_expression_t*) value; + declaration_t *declaration = reference->declaration; + if(declaration != NULL) { + declaration->address_taken = 1; + } + } + + expression->expression.datatype = make_pointer_type(orig_type, TYPE_QUALIFIER_NONE); +} + +#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) +{ + /* 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; + } + + type_right = promote_integer(type_right); + type_left = promote_integer(type_left); + + if(type_left == type_right) + return type_left; + + bool signed_left = is_type_signed(type_left); + bool signed_right = is_type_signed(type_right); + int rank_left = get_rank(type_left); + int rank_right = get_rank(type_right); + if(rank_left < rank_right) { + if(signed_left == signed_right || !signed_right) { + return type_right; + } else { + return type_left; + } + } else { + if(signed_left == signed_right || !signed_left) { + return type_left; + } else { + return type_right; + } + } +} + +static void semantic_binexpr_arithmetic(binary_expression_t *expression) +{ + expression_t *left = expression->left; + expression_t *right = expression->right; + type_t *orig_type_left = left->base.datatype; + type_t *orig_type_right = right->base.datatype; + + if(orig_type_left == NULL || orig_type_right == NULL) + return; - return 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; + } + + 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_array_expression(unsigned precedence, - expression_t *array_ref) +static void semantic_shift_op(binary_expression_t *expression) { - (void) precedence; - - eat('['); + expression_t *left = expression->left; + expression_t *right = expression->right; + type_t *orig_type_left = left->base.datatype; + type_t *orig_type_right = right->base.datatype; - array_access_expression_t *array_access - = allocate_ast_zero(sizeof(array_access[0])); + if(orig_type_left == NULL || orig_type_right == NULL) + return; - array_access->expression.type = EXPR_ARRAY_ACCESS; - array_access->array_ref = array_ref; - array_access->index = parse_expression(); + type_t *type_left = skip_typeref(orig_type_left); + type_t *type_right = skip_typeref(orig_type_right); - type_t *array_type = array_ref->datatype; - if(array_type != NULL) { - if(array_type->type == TYPE_POINTER) { - pointer_type_t *pointer = (pointer_type_t*) array_type; - array_access->expression.datatype = pointer->points_to; - } else { - parser_print_error_prefix(); - fprintf(stderr, "array access on object with non-pointer type "); - print_type(array_type); - fprintf(stderr, "\n"); - } + 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; } - if(token.type != ']') { - parse_error_expected("Problem while parsing array access", ']', 0); - return (expression_t*) array_access; - } - next_token(); + type_left = promote_integer(type_left); + type_right = promote_integer(type_right); - return (expression_t*) array_access; + expression->left = create_implicit_cast(left, type_left); + expression->right = create_implicit_cast(right, type_right); + expression->expression.datatype = type_left; } -static bool is_declaration_specifier(const token_t *token, - bool only_type_specifiers) +static void semantic_add(binary_expression_t *expression) { - declaration_t *declaration; + expression_t *left = expression->left; + expression_t *right = expression->right; + type_t *orig_type_left = left->base.datatype; + type_t *orig_type_right = right->base.datatype; - switch(token->type) { - TYPE_SPECIFIERS - return 1; - case T_IDENTIFIER: - declaration = token->v.symbol->declaration; - if(declaration == NULL) - return 0; - if(declaration->storage_class != STORAGE_CLASS_TYPEDEF) - return 0; - return 1; - STORAGE_CLASSES - TYPE_QUALIFIERS - if(only_type_specifiers) - return 0; - return 1; + if(orig_type_left == NULL || orig_type_right == NULL) + return; - default: - return 0; + type_t *type_left = skip_typeref(orig_type_left); + type_t *type_right = skip_typeref(orig_type_right); + + /* § 5.6.5 */ + if(is_type_arithmetic(type_left) && is_type_arithmetic(type_right)) { + type_t *arithmetic_type = semantic_arithmetic(type_left, type_right); + expression->left = create_implicit_cast(left, arithmetic_type); + expression->right = create_implicit_cast(right, arithmetic_type); + expression->expression.datatype = arithmetic_type; + return; + } else if(is_type_pointer(type_left) && is_type_integer(type_right)) { + expression->expression.datatype = type_left; + } else if(is_type_pointer(type_right) && is_type_integer(type_left)) { + expression->expression.datatype = type_right; + } else { + parser_print_error_prefix(); + fprintf(stderr, "invalid operands to binary + ("); + print_type_quoted(orig_type_left); + fprintf(stderr, ", "); + print_type_quoted(orig_type_right); + fprintf(stderr, ")\n"); } } -static expression_t *parse_sizeof(unsigned precedence) +static void semantic_sub(binary_expression_t *expression) { - eat(T_sizeof); + expression_t *left = expression->left; + expression_t *right = expression->right; + type_t *orig_type_left = left->base.datatype; + type_t *orig_type_right = right->base.datatype; - sizeof_expression_t *sizeof_expression - = allocate_ast_zero(sizeof(sizeof_expression[0])); - sizeof_expression->expression.type = EXPR_SIZEOF; - sizeof_expression->expression.datatype = type_size_t; + if(orig_type_left == NULL || orig_type_right == NULL) + return; - if(token.type == '(' && is_declaration_specifier(look_ahead(1), true)) { - next_token(); - sizeof_expression->type = parse_typename(); - expect(')'); + type_t *type_left = skip_typeref(orig_type_left); + type_t *type_right = skip_typeref(orig_type_right); + + /* § 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, "pointers to incompatible objects to binary - ("); + print_type_quoted(orig_type_left); + fprintf(stderr, ", "); + print_type_quoted(orig_type_right); + fprintf(stderr, ")\n"); + } else { + expression->expression.datatype = type_ptrdiff_t; + } } else { - expression_t *expression = parse_sub_expression(precedence); - sizeof_expression->type = expression->datatype; - sizeof_expression->size_expression = expression; + 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*) sizeof_expression; } -static expression_t *parse_select_expression(unsigned precedence, - expression_t *compound) +static void semantic_comparison(binary_expression_t *expression) { - (void) precedence; + expression_t *left = expression->left; + expression_t *right = expression->right; + type_t *orig_type_left = left->base.datatype; + type_t *orig_type_right = right->base.datatype; - assert(token.type == '.' || token.type == T_MINUSGREATER); - next_token(); + 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; +} - select_expression_t *select = allocate_ast_zero(sizeof(select[0])); +static void semantic_arithmetic_assign(binary_expression_t *expression) +{ + expression_t *left = expression->left; + expression_t *right = expression->right; + type_t *orig_type_left = left->base.datatype; + type_t *orig_type_right = right->base.datatype; - select->expression.type = EXPR_SELECT; - select->compound = compound; + if(orig_type_left == NULL || orig_type_right == NULL) + return; - /* TODO: datatype */ + type_t *type_left = skip_typeref(orig_type_left); + type_t *type_right = skip_typeref(orig_type_right); - if(token.type != T_IDENTIFIER) { - parse_error_expected("Problem while parsing select", T_IDENTIFIER, 0); - return (expression_t*) select; + if(!is_type_arithmetic(type_left) || !is_type_arithmetic(type_right)) { + /* TODO: improve error message */ + parser_print_error_prefix(); + fprintf(stderr, "operation needs arithmetic types\n"); + return; } - select->symbol = token.v.symbol; - next_token(); - return (expression_t*) select; + /* combined instructions are tricky. We can't create an implicit cast on + * the left side, because we need the uncasted form for the store. + * The ast2firm pass has to know that left_type must be right_type + * for the arithmeitc operation and create a cast by itself */ + type_t *arithmetic_type = semantic_arithmetic(type_left, type_right); + expression->right = create_implicit_cast(right, arithmetic_type); + expression->expression.datatype = type_left; } -static expression_t *parse_call_expression(unsigned precedence, - expression_t *expression) +static void semantic_arithmetic_addsubb_assign(binary_expression_t *expression) { - (void) precedence; - call_expression_t *call = allocate_ast_zero(sizeof(call[0])); + expression_t *left = expression->left; + expression_t *right = expression->right; + type_t *orig_type_left = left->base.datatype; + type_t *orig_type_right = right->base.datatype; - call->expression.type = EXPR_CALL; - call->method = expression; - - /* parse arguments */ - eat('('); - - if(token.type != ')') { - call_argument_t *last_argument = NULL; + if(orig_type_left == NULL || orig_type_right == NULL) + return; - while(true) { - call_argument_t *argument = allocate_ast_zero(sizeof(argument[0])); + type_t *type_left = skip_typeref(orig_type_left); + type_t *type_right = skip_typeref(orig_type_right); + + if (is_type_arithmetic(type_left) && is_type_arithmetic(type_right)) { + /* combined instructions are tricky. We can't create an implicit cast on + * the left side, because we need the uncasted form for the store. + * The ast2firm pass has to know that left_type must be right_type + * for the arithmeitc operation and create a cast by itself */ + type_t *const arithmetic_type = semantic_arithmetic(type_left, type_right); + expression->right = create_implicit_cast(right, arithmetic_type); + expression->expression.datatype = type_left; + } else if (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; + } +} - argument->expression = parse_assignment_expression(); - if(last_argument == NULL) { - call->arguments = argument; - } else { - last_argument->next = argument; - } - last_argument = argument; +static void semantic_logical_op(binary_expression_t *expression) +{ + expression_t *left = expression->left; + expression_t *right = expression->right; + type_t *orig_type_left = left->base.datatype; + type_t *orig_type_right = right->base.datatype; - if(token.type != ',') - break; - next_token(); - } - } - expect(')'); + if(orig_type_left == NULL || orig_type_right == NULL) + return; - type_t *type = expression->datatype; - if(type != NULL) { - /* we can call pointer to function */ - if(type->type == TYPE_POINTER) { - pointer_type_t *pointer = (pointer_type_t*) type; - type = pointer->points_to; - } + type_t *type_left = skip_typeref(orig_type_left); + type_t *type_right = skip_typeref(orig_type_right); - if(type == NULL || type->type != TYPE_METHOD) { - parser_print_error_prefix(); - fprintf(stderr, "expected a method type for call but found type "); - print_type(expression->datatype); - fprintf(stderr, "\n"); - } else { - method_type_t *method_type = (method_type_t*) type; - call->expression.datatype = method_type->result_type; - } + if (!is_type_scalar(type_left) || !is_type_scalar(type_right)) { + /* TODO: improve error message */ + parser_print_error_prefix(); + fprintf(stderr, "operation needs scalar types\n"); + return; } - return (expression_t*) call; -} - -static void type_error(const char *msg, const source_position_t source_position, - const type_t *type) -{ - parser_print_error_prefix_pos(source_position); - fprintf(stderr, "%s, but found type ", msg); - print_type(type); - fputc('\n', stderr); - error(); + expression->expression.datatype = type_int; } -static void type_error_incompatible(const char *msg, - const source_position_t source_position, const type_t *type1, - const type_t *type2) +static bool has_const_fields(type_t *type) { - parser_print_error_prefix_pos(source_position); - fprintf(stderr, "%s, incompatible types: ", msg); - print_type(type1); - fprintf(stderr, " - "); - print_type(type2); - fprintf(stderr, ")\n"); - error(); + (void) type; + /* TODO */ + return false; } -static type_t *get_type_after_conversion(const type_t *type1, - const type_t *type2) +static void semantic_binexpr_assign(binary_expression_t *expression) { - /* TODO... */ - (void) type2; - return (type_t*) type1; -} + expression_t *left = expression->left; + type_t *orig_type_left = left->base.datatype; -static expression_t *parse_conditional_expression(unsigned precedence, - expression_t *expression) -{ - eat('?'); + if(orig_type_left == NULL) + return; - conditional_expression_t *conditional - = allocate_ast_zero(sizeof(conditional[0])); - conditional->expression.type = EXPR_CONDITIONAL; - conditional->condition = expression; + type_t *type_left = revert_automatic_type_conversion(left); + type_left = skip_typeref(orig_type_left); - /* 6.5.15.2 */ - type_t *condition_type = conditional->condition->datatype; - if(condition_type != NULL) { - if(!is_type_scalar(condition_type)) { - type_error("expected a scalar type", expression->source_position, - condition_type); - } + /* must be a modifiable lvalue */ + if (type_left->type == TYPE_ARRAY) { + parser_print_error_prefix(); + fprintf(stderr, "Cannot assign to arrays ('"); + print_expression(left); + fprintf(stderr, "')\n"); + return; } - - conditional->true_expression = parse_expression(); - expect(':'); - conditional->false_expression = parse_sub_expression(precedence); - - type_t *true_type = conditional->true_expression->datatype; - if(true_type == NULL) - return (expression_t*) conditional; - type_t *false_type = conditional->false_expression->datatype; - if(false_type == NULL) - return (expression_t*) conditional; - - /* 6.4.15.3 */ - if(true_type == false_type) { - conditional->expression.datatype = true_type; - } else if(is_type_arithmetic(true_type) && is_type_arithmetic(false_type)) { - type_t *result = get_type_after_conversion(true_type, false_type); - /* TODO: create implicit convs if necessary */ - conditional->expression.datatype = result; - } else if(true_type->type == TYPE_POINTER && - false_type->type == TYPE_POINTER && - true /* TODO compatible points_to types */) { - /* TODO */ - } else if(/* (is_null_ptr_const(true_type) && false_type->type == TYPE_POINTER) - || (is_null_ptr_const(false_type) && - true_type->type == TYPE_POINTER) TODO*/ false) { - /* TODO */ - } else if(/* 1 is pointer to object type, other is void* */ false) { - /* TODO */ - } else { - type_error_incompatible("problem while parsing conditional", - expression->source_position, true_type, - false_type); + if(type_left->base.qualifiers & TYPE_QUALIFIER_CONST) { + parser_print_error_prefix(); + fprintf(stderr, "assignment to readonly location '"); + print_expression(left); + fprintf(stderr, "' (type "); + print_type_quoted(orig_type_left); + fprintf(stderr, ")\n"); + return; + } + if(is_type_incomplete(type_left)) { + parser_print_error_prefix(); + fprintf(stderr, "left-hand side of assignment '"); + print_expression(left); + fprintf(stderr, "' has incomplete type "); + print_type_quoted(orig_type_left); + fprintf(stderr, "\n"); + return; + } + if(is_type_compound(type_left) && has_const_fields(type_left)) { + parser_print_error_prefix(); + fprintf(stderr, "can't assign to '"); + print_expression(left); + fprintf(stderr, "' because compound type "); + print_type_quoted(orig_type_left); + fprintf(stderr, " has readonly fields\n"); + return; } - return (expression_t*) conditional; + semantic_assign(orig_type_left, &expression->right, "assignment"); + + expression->expression.datatype = orig_type_left; } -static expression_t *parse_extension(unsigned precedence) +static void semantic_comma(binary_expression_t *expression) { - eat(T___extension__); - - /* TODO enable extensions */ - - return parse_sub_expression(precedence); + expression->expression.datatype = expression->right->base.datatype; } -#define CREATE_UNARY_EXPRESSION_PARSER(token_type, unexpression_type) \ -static \ -expression_t *parse_##unexpression_type(unsigned precedence) \ -{ \ - 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; \ +#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) +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) { @@ -2429,7 +3974,7 @@ static expression_t *parse_sub_expression(unsigned precedence) left = parse_primary_expression(); } assert(left != NULL); - left->source_position = source_position; + left->base.source_position = source_position; while(true) { if(token.type < 0) { @@ -2445,8 +3990,8 @@ static expression_t *parse_sub_expression(unsigned precedence) left = parser->infix_parser(parser->infix_precedence, left); assert(left != NULL); - assert(left->type != EXPR_INVALID); - left->source_position = source_position; + assert(left->type != EXPR_UNKNOWN); + left->base.source_position = source_position; } return left; @@ -2459,14 +4004,14 @@ static expression_t *parse_expression(void) -void register_expression_parser(parse_expression_function parser, - int token_type, unsigned precedence) +static void register_expression_parser(parse_expression_function parser, + int token_type, unsigned precedence) { expression_parser_function_t *entry = &expression_parsers[token_type]; if(entry->parser != NULL) { fprintf(stderr, "for token "); - print_token_type(stderr, token_type); + print_token_type(stderr, (token_type_t) token_type); fprintf(stderr, "\n"); panic("trying to register multiple expression parsers for a token"); } @@ -2474,14 +4019,15 @@ 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]; if(entry->infix_parser != NULL) { fprintf(stderr, "for token "); - print_token_type(stderr, token_type); + print_token_type(stderr, (token_type_t) token_type); fprintf(stderr, "\n"); panic("trying to register multiple infix expression parsers for a " "token"); @@ -2557,6 +4103,8 @@ 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); } @@ -2570,7 +4118,7 @@ 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; } @@ -2584,18 +4132,68 @@ 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) @@ -2714,10 +4312,27 @@ 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) @@ -2725,9 +4340,9 @@ static statement_t *parse_continue(void) eat(T_continue); expect(';'); - statement_t *statement = allocate_ast_zero(sizeof(statement[0])); - statement->source_position = token.source_position; - statement->type = STATEMENT_CONTINUE; + statement_t *statement = allocate_ast_zero(sizeof(statement[0])); + statement->type = STATEMENT_CONTINUE; + statement->base.source_position = token.source_position; return statement; } @@ -2737,9 +4352,9 @@ static statement_t *parse_break(void) eat(T_break); expect(';'); - statement_t *statement = allocate_ast_zero(sizeof(statement[0])); - statement->source_position = token.source_position; - statement->type = STATEMENT_BREAK; + statement_t *statement = allocate_ast_zero(sizeof(statement[0])); + statement->type = STATEMENT_BREAK; + statement->base.source_position = token.source_position; return statement; } @@ -2752,11 +4367,42 @@ 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 = ¤t_function->type->function; + type_t *return_type = function_type->result_type; + + expression_t *return_value = NULL; if(token.type != ';') { - statement->return_value = parse_expression(); + return_value = parse_expression(); } expect(';'); + if(return_type == NULL) + return (statement_t*) statement; + + return_type = skip_typeref(return_type); + + if(return_value != NULL) { + type_t *return_value_type = skip_typeref(return_value->base.datatype); + + if(is_type_atomic(return_type, ATOMIC_TYPE_VOID) + && !is_type_atomic(return_value_type, ATOMIC_TYPE_VOID)) { + parse_warning("'return' with a value, in function returning void"); + return_value = NULL; + } else { + if(return_type != NULL) { + semantic_assign(return_type, &return_value, "'return'"); + } + } + } else { + if(!is_type_atomic(return_type, ATOMIC_TYPE_VOID)) { + parse_warning("'return' without value, in function returning " + "non-void"); + } + } + statement->return_value = return_value; + return (statement_t*) statement; } @@ -2804,7 +4450,6 @@ static statement_t *parse_expression_statement(void) static statement_t *parse_statement(void) { - declaration_t *declaration; statement_t *statement = NULL; /* declaration or statement */ @@ -2868,9 +4513,7 @@ 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; } @@ -2896,20 +4539,21 @@ static statement_t *parse_statement(void) break; } - assert(statement == NULL || statement->source_position.input_name != NULL); + assert(statement == NULL + || statement->base.source_position.input_name != NULL); return statement; } static statement_t *parse_compound_statement(void) { - eat('{'); - compound_statement_t *compound_statement = allocate_ast_zero(sizeof(compound_statement[0])); compound_statement->statement.type = STATEMENT_COMPOUND; compound_statement->statement.source_position = token.source_position; + eat('{'); + int top = environment_top(); context_t *last_context = context; set_context(&compound_statement->context); @@ -2922,23 +4566,28 @@ static statement_t *parse_compound_statement(void) continue; if(last_statement != NULL) { - last_statement->next = statement; + last_statement->base.next = statement; } else { compound_statement->statements = statement; } - while(statement->next != NULL) - statement = statement->next; + while(statement->base.next != NULL) + statement = statement->base.next; last_statement = statement; } + if(token.type != '}') { + parser_print_error_prefix_pos( + compound_statement->statement.source_position); + fprintf(stderr, "end of file while looking for closing '}'\n"); + } + next_token(); + assert(context == &compound_statement->context); set_context(last_context); environment_pop_to(top); - next_token(); - return (statement_t*) compound_statement; } @@ -2968,10 +4617,12 @@ static translation_unit_t *parse_translation_unit(void) translation_unit_t *parse(void) { - obstack_init(&environment_obstack); - environment_stack = NEW_ARR_F(environment_entry_t*, 0); + environment_stack = NEW_ARR_F(stack_entry_t, 0); + label_stack = NEW_ARR_F(stack_entry_t, 0); + found_error = false; type_set_output(stderr); + ast_set_output(stderr); lookahead_bufpos = 0; for(int i = 0; i < MAX_LOOKAHEAD + 2; ++i) { @@ -2980,7 +4631,10 @@ 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; } @@ -2990,12 +4644,16 @@ void init_parser(void) init_expression_parsers(); obstack_init(&temp_obst); - type_int = make_atomic_type(ATOMIC_TYPE_INT, 0); - type_double = make_atomic_type(ATOMIC_TYPE_DOUBLE, 0); - type_size_t = make_atomic_type(ATOMIC_TYPE_UINT, 0); - type_const_char = make_atomic_type(ATOMIC_TYPE_CHAR, TYPE_QUALIFIER_CONST); - type_void = make_atomic_type(ATOMIC_TYPE_VOID, 0); - type_string = make_pointer_type(type_const_char, 0); + type_int = make_atomic_type(ATOMIC_TYPE_INT, TYPE_QUALIFIER_NONE); + type_long_double = make_atomic_type(ATOMIC_TYPE_LONG_DOUBLE, TYPE_QUALIFIER_NONE); + type_double = make_atomic_type(ATOMIC_TYPE_DOUBLE, TYPE_QUALIFIER_NONE); + type_float = make_atomic_type(ATOMIC_TYPE_FLOAT, TYPE_QUALIFIER_NONE); + type_size_t = make_atomic_type(ATOMIC_TYPE_ULONG, TYPE_QUALIFIER_NONE); + type_ptrdiff_t = make_atomic_type(ATOMIC_TYPE_LONG, TYPE_QUALIFIER_NONE); + type_char = make_atomic_type(ATOMIC_TYPE_CHAR, TYPE_QUALIFIER_NONE); + type_void = make_atomic_type(ATOMIC_TYPE_VOID, TYPE_QUALIFIER_NONE); + type_void_ptr = make_pointer_type(type_void, TYPE_QUALIFIER_NONE); + type_string = make_pointer_type(type_char, TYPE_QUALIFIER_NONE); } void exit_parser(void)