X-Git-Url: http://nsz.repo.hu/git/?a=blobdiff_plain;f=parser.c;h=fdc2839f956266d4e7ec5f34b74ac98b24c57784;hb=f21a830ddb9b69adb0d55dcd04e7e9d8375a83f0;hp=d35396f2cebd706bbe9bd1d83c7c78f3c496b629;hpb=b1933de107b720d0efda752ef102a6e9ead64e20;p=cparser diff --git a/parser.c b/parser.c index d35396f..fdc2839 100644 --- a/parser.c +++ b/parser.c @@ -4,12 +4,16 @@ #include #include +#include "diagnostic.h" +#include "format_check.h" #include "parser.h" #include "lexer.h" #include "token_t.h" +#include "types.h" #include "type_t.h" #include "type_hash.h" #include "ast_t.h" +#include "lang_features.h" #include "adt/bitfiddle.h" #include "adt/error.h" #include "adt/array.h" @@ -17,7 +21,6 @@ //#define PRINT_TOKENS //#define ABORT_ON_ERROR #define MAX_LOOKAHEAD 2 -//#define STRICT_C99 typedef struct { declaration_t *old_declaration; @@ -25,29 +28,36 @@ typedef struct { 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; +typedef struct declaration_specifiers_t declaration_specifiers_t; +struct declaration_specifiers_t { + source_position_t source_position; + unsigned char storage_class; + bool is_inline; + decl_modifiers_t decl_modifiers; + type_t *type; +}; + +typedef declaration_t* (*parsed_declaration_func) (declaration_t *declaration); + +static token_t token; +static token_t lookahead_buffer[MAX_LOOKAHEAD]; +static int lookahead_bufpos; +static stack_entry_t *environment_stack = NULL; +static stack_entry_t *label_stack = NULL; +static context_t *global_context = NULL; +static context_t *context = NULL; +static declaration_t *last_declaration = NULL; +static declaration_t *current_function = NULL; +static switch_statement_t *current_switch = NULL; +static statement_t *current_loop = NULL; +static goto_statement_t *goto_first = NULL; +static goto_statement_t *goto_last = NULL; static struct obstack temp_obst; -static bool found_error; - -static type_t *type_int = NULL; -static type_t *type_uint = NULL; -static type_t *type_long_double = NULL; -static type_t *type_double = NULL; -static type_t *type_float = NULL; -static type_t *type_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; + +/** The current source position. */ +#define HERE token.source_position + +static type_t *type_valist; static statement_t *parse_compound_statement(void); static statement_t *parse_statement(void); @@ -56,6 +66,13 @@ static expression_t *parse_sub_expression(unsigned precedence); static expression_t *parse_expression(void); static type_t *parse_typename(void); +static void parse_compound_type_entries(void); +static declaration_t *parse_declarator( + const declaration_specifiers_t *specifiers, bool may_be_abstract); +static declaration_t *record_declaration(declaration_t *declaration); + +static void semantic_comparison(binary_expression_t *expression); + #define STORAGE_CLASSES \ case T_typedef: \ case T_extern: \ @@ -67,7 +84,8 @@ static type_t *parse_typename(void); case T_const: \ case T_restrict: \ case T_volatile: \ - case T_inline: + case T_inline: \ + case T_forceinline: #ifdef PROVIDE_COMPLEX #define COMPLEX_SPECIFIERS \ @@ -79,22 +97,23 @@ static type_t *parse_typename(void); #define IMAGINARY_SPECIFIERS #endif -#define TYPE_SPECIFIERS \ - case T_void: \ - case T_char: \ - case T_short: \ - case T_int: \ - case T_long: \ - case T_float: \ - case T_double: \ - case T_signed: \ - case T_unsigned: \ - case T__Bool: \ - case T_struct: \ - case T_union: \ - case T_enum: \ - case T___typeof__: \ - COMPLEX_SPECIFIERS \ +#define TYPE_SPECIFIERS \ + case T_void: \ + case T_char: \ + case T_short: \ + case T_int: \ + case T_long: \ + case T_float: \ + case T_double: \ + case T_signed: \ + case T_unsigned: \ + case T__Bool: \ + case T_struct: \ + case T_union: \ + case T_enum: \ + case T___typeof__: \ + case T___builtin_va_list: \ + COMPLEX_SPECIFIERS \ IMAGINARY_SPECIFIERS #define DECLARATION_START \ @@ -106,60 +125,217 @@ static type_t *parse_typename(void); TYPE_QUALIFIERS \ TYPE_SPECIFIERS -static inline void *allocate_ast_zero(size_t size) +/** + * Allocate an AST node with given size and + * initialize all fields with zero. + */ +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 declaration_t *allocate_declaration_zero(void) +{ + return allocate_ast_zero(sizeof(*allocate_declaration_zero())); +} + +/** + * Returns the size of a statement node. + * + * @param kind the statement kind + */ +static size_t get_statement_struct_size(statement_kind_t kind) +{ + static const size_t sizes[] = { + [STATEMENT_COMPOUND] = sizeof(compound_statement_t), + [STATEMENT_RETURN] = sizeof(return_statement_t), + [STATEMENT_DECLARATION] = sizeof(declaration_statement_t), + [STATEMENT_IF] = sizeof(if_statement_t), + [STATEMENT_SWITCH] = sizeof(switch_statement_t), + [STATEMENT_EXPRESSION] = sizeof(expression_statement_t), + [STATEMENT_CONTINUE] = sizeof(statement_base_t), + [STATEMENT_BREAK] = sizeof(statement_base_t), + [STATEMENT_GOTO] = sizeof(goto_statement_t), + [STATEMENT_LABEL] = sizeof(label_statement_t), + [STATEMENT_CASE_LABEL] = sizeof(case_label_statement_t), + [STATEMENT_WHILE] = sizeof(while_statement_t), + [STATEMENT_DO_WHILE] = sizeof(do_while_statement_t), + [STATEMENT_FOR] = sizeof(for_statement_t), + [STATEMENT_ASM] = sizeof(asm_statement_t) + }; + assert(kind <= sizeof(sizes) / sizeof(sizes[0])); + assert(sizes[kind] != 0); + return sizes[kind]; +} + +/** + * Allocate a statement node of given kind and initialize all + * fields with zero. + */ +static statement_t *allocate_statement_zero(statement_kind_t kind) +{ + size_t size = get_statement_struct_size(kind); + statement_t *res = allocate_ast_zero(size); + + res->base.kind = kind; + return res; +} + +/** + * Returns the size of an expression node. + * + * @param kind the expression kind + */ +static size_t get_expression_struct_size(expression_kind_t kind) +{ + static const size_t sizes[] = { + [EXPR_INVALID] = sizeof(expression_base_t), + [EXPR_REFERENCE] = sizeof(reference_expression_t), + [EXPR_CONST] = sizeof(const_expression_t), + [EXPR_STRING_LITERAL] = sizeof(string_literal_expression_t), + [EXPR_WIDE_STRING_LITERAL] = sizeof(wide_string_literal_expression_t), + [EXPR_CALL] = sizeof(call_expression_t), + [EXPR_UNARY_FIRST] = sizeof(unary_expression_t), + [EXPR_BINARY_FIRST] = sizeof(binary_expression_t), + [EXPR_CONDITIONAL] = sizeof(conditional_expression_t), + [EXPR_SELECT] = sizeof(select_expression_t), + [EXPR_ARRAY_ACCESS] = sizeof(array_access_expression_t), + [EXPR_SIZEOF] = sizeof(sizeof_expression_t), + [EXPR_CLASSIFY_TYPE] = sizeof(classify_type_expression_t), + [EXPR_FUNCTION] = sizeof(string_literal_expression_t), + [EXPR_PRETTY_FUNCTION] = sizeof(string_literal_expression_t), + [EXPR_BUILTIN_SYMBOL] = sizeof(builtin_symbol_expression_t), + [EXPR_BUILTIN_CONSTANT_P] = sizeof(builtin_constant_expression_t), + [EXPR_BUILTIN_PREFETCH] = sizeof(builtin_prefetch_expression_t), + [EXPR_OFFSETOF] = sizeof(offsetof_expression_t), + [EXPR_VA_START] = sizeof(va_start_expression_t), + [EXPR_VA_ARG] = sizeof(va_arg_expression_t), + [EXPR_STATEMENT] = sizeof(statement_expression_t), + }; + if(kind >= EXPR_UNARY_FIRST && kind <= EXPR_UNARY_LAST) { + return sizes[EXPR_UNARY_FIRST]; + } + if(kind >= EXPR_BINARY_FIRST && kind <= EXPR_BINARY_LAST) { + return sizes[EXPR_BINARY_FIRST]; + } + assert(kind <= sizeof(sizes) / sizeof(sizes[0])); + assert(sizes[kind] != 0); + return sizes[kind]; +} + +/** + * Allocate an expression node of given kind and initialize all + * fields with zero. + */ +static expression_t *allocate_expression_zero(expression_kind_t kind) +{ + size_t size = get_expression_struct_size(kind); + expression_t *res = allocate_ast_zero(size); + + res->base.kind = kind; + return res; +} + +/** + * Returns the size of a type node. + * + * @param kind the type kind + */ +static size_t get_type_struct_size(type_kind_t kind) +{ + static const size_t sizes[] = { + [TYPE_ATOMIC] = sizeof(atomic_type_t), + [TYPE_BITFIELD] = sizeof(bitfield_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(kind <= TYPE_TYPEOF); + assert(sizes[kind] != 0); + return sizes[kind]; +} + +/** + * Allocate a type node of given kind and initialize all + * fields with zero. + */ +static type_t *allocate_type_zero(type_kind_t kind) { - void *res = obstack_alloc(type_obst, size); + size_t size = get_type_struct_size(kind); + type_t *res = obstack_alloc(type_obst, size); memset(res, 0, size); + + res->base.kind = kind; return res; } -static inline size_t get_initializer_size(initializer_type_t type) +/** + * Returns the size of an initializer node. + * + * @param kind the initializer kind + */ +static size_t get_initializer_size(initializer_kind_t kind) { - static const size_t size[] = { - [INITIALIZER_VALUE] = sizeof(initializer_value_t), - [INITIALIZER_STRING] = sizeof(initializer_string_t), - [INITIALIZER_LIST] = sizeof(initializer_list_t) + static const size_t sizes[] = { + [INITIALIZER_VALUE] = sizeof(initializer_value_t), + [INITIALIZER_STRING] = sizeof(initializer_string_t), + [INITIALIZER_WIDE_STRING] = sizeof(initializer_wide_string_t), + [INITIALIZER_LIST] = sizeof(initializer_list_t) }; - assert(type < INITIALIZER_COUNT); - assert(size[type] != 0); - return size[type]; + assert(kind < sizeof(sizes) / sizeof(*sizes)); + assert(sizes[kind] != 0); + return sizes[kind]; } -static inline initializer_t *allocate_initializer(initializer_type_t type) +/** + * Allocate an initializer node of given kind and initialize all + * fields with zero. + */ +static initializer_t *allocate_initializer_zero(initializer_kind_t kind) { - initializer_t *result = allocate_ast_zero(get_initializer_size(type)); - result->type = type; + initializer_t *result = allocate_ast_zero(get_initializer_size(kind)); + result->kind = kind; return result; } -static inline void free_type(void *type) +/** + * Free a type from the type obstack. + */ +static void free_type(void *type) { obstack_free(type_obst, type); } /** - * returns the top element of the environment stack + * Returns the index of 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 inline size_t label_top(void) +/** + * Returns the index of the top element of the label stack. + */ +static size_t label_top(void) { return ARR_LEN(label_stack); } - +/** + * Return the next token. + */ static inline void next_token(void) { token = lookahead_buffer[lookahead_bufpos]; @@ -174,125 +350,53 @@ static inline void next_token(void) #endif } +/** + * Return the next token with a given lookahead. + */ static inline const token_t *look_ahead(int num) { assert(num > 0 && num <= MAX_LOOKAHEAD); int pos = (lookahead_bufpos+num-1) % MAX_LOOKAHEAD; - return & lookahead_buffer[pos]; + return &lookahead_buffer[pos]; } #define eat(token_type) do { assert(token.type == token_type); next_token(); } while(0) -static void error(void) -{ - found_error = true; -#ifdef ABORT_ON_ERROR - abort(); -#endif -} - -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); - fputs(": ", stderr); -} - -static void parser_print_error_prefix_pos( - const source_position_t source_position) -{ - parser_print_prefix_pos(source_position); - fputs("error: ", stderr); - error(); -} - -static void parser_print_error_prefix(void) -{ - parser_print_error_prefix_pos(token.source_position); -} - -static void parse_error(const char *message) -{ - parser_print_error_prefix(); - fprintf(stderr, "parse error: %s\n", 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(source_position); - fprintf(stderr, "warning: %s\n", message); -} - -static void parse_warning(const char *message) -{ - parse_warning_pos(token.source_position, message); -} - +/** + * Report a parse error because an expected token was not found. + */ static void parse_error_expected(const char *message, ...) { - va_list args; - int first = 1; - if(message != NULL) { - parser_print_error_prefix(); - fprintf(stderr, "%s\n", message); - } - parser_print_error_prefix(); - fputs("Parse error: got ", stderr); - print_token(stderr, &token); - fputs(", expected ", stderr); - - va_start(args, message); - token_type_t token_type = va_arg(args, token_type_t); - while(token_type != 0) { - if(first == 1) { - first = 0; - } else { - fprintf(stderr, ", "); - } - print_token_type(stderr, token_type); - token_type = va_arg(args, token_type_t); + errorf(HERE, "%s", message); } - va_end(args); - fprintf(stderr, "\n"); -} - -static void print_type_quoted(type_t *type) -{ - fputc('\'', stderr); - print_type(type); - fputc('\'', stderr); + va_list ap; + va_start(ap, message); + errorf(HERE, "got '%K', expected %#k", &token, &ap, ", "); + va_end(ap); } +/** + * Report a type error. + */ 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); + errorf(source_position, "%s, but found type '%T'", msg, type); } +/** + * Report an incompatible type. + */ 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"); + errorf(source_position, "%s, incompatible types: '%T' - '%T'", msg, type1, type2); } +/** + * Eat an complete block, ie. '{ ... }'. + */ static void eat_block(void) { if(token.type == '{') @@ -310,6 +414,9 @@ static void eat_block(void) eat('}'); } +/** + * Eat a statement until an ';' token. + */ static void eat_statement(void) { while(token.type != ';') { @@ -326,7 +433,10 @@ static void eat_statement(void) eat(';'); } -static void eat_brace(void) +/** + * Eat a parenthesed term, ie. '( ... )'. + */ +static void eat_paren(void) { if(token.type == '(') next_token(); @@ -338,7 +448,7 @@ static void eat_brace(void) return; } if(token.type == '(') { - eat_brace(); + eat_paren(); continue; } if(token.type == '{') { @@ -387,17 +497,10 @@ static void set_context(context_t *new_context) } /** - * called when we find a 2nd declarator for an identifier we already have a - * declarator for + * Search a symbol in a given namespace and returns its declaration or + * NULL if this symbol was not found. */ -static bool is_compatible_declaration (declaration_t *declaration, - declaration_t *previous) -{ - /* TODO: not correct yet */ - return declaration->type == previous->type; -} - -static declaration_t *get_declaration(symbol_t *symbol, namespace_t namespc) +static declaration_t *get_declaration(const symbol_t *const symbol, const namespace_t namespc) { declaration_t *declaration = symbol->declaration; for( ; declaration != NULL; declaration = declaration->symbol_next) { @@ -408,111 +511,20 @@ static declaration_t *get_declaration(symbol_t *symbol, namespace_t namespc) 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 declaration_t *stack_push(stack_entry_t **stack_ptr, - declaration_t *declaration, - context_t *parent_context) +static void stack_push(stack_entry_t **stack_ptr, declaration_t *declaration) { symbol_t *symbol = declaration->symbol; namespace_t namespc = (namespace_t)declaration->namespc; - /* a declaration should be only pushed once */ - assert(declaration->parent_context == NULL); - declaration->parent_context = parent_context; - - declaration_t *previous_declaration = get_declaration(symbol, namespc); - assert(declaration != previous_declaration); - if(previous_declaration != NULL - && previous_declaration->parent_context == context) { - if(!is_compatible_declaration(declaration, previous_declaration)) { - parser_print_error_prefix_pos(declaration->source_position); - fprintf(stderr, "definition of symbol %s%s with type ", - get_namespace_prefix(namespc), symbol->string); - print_type_quoted(declaration->type); - fputc('\n', stderr); - parser_print_error_prefix_pos( - previous_declaration->source_position); - fprintf(stderr, "is incompatible with previous declaration " - "of type "); - print_type_quoted(previous_declaration->type); - fputc('\n', stderr); - } else { - const storage_class_t old_storage = previous_declaration->storage_class; - const storage_class_t new_storage = declaration->storage_class; - if (current_function == NULL) { - if (old_storage != STORAGE_CLASS_STATIC && - new_storage == STORAGE_CLASS_STATIC) { - parser_print_error_prefix_pos(declaration->source_position); - fprintf(stderr, - "static declaration of '%s' follows non-static declaration\n", - symbol->string); - parser_print_error_prefix_pos(previous_declaration->source_position); - fprintf(stderr, "previous declaration of '%s' was here\n", - symbol->string); - } else { - if (old_storage == STORAGE_CLASS_EXTERN) { - if (new_storage == STORAGE_CLASS_NONE) { - previous_declaration->storage_class = STORAGE_CLASS_NONE; - } - } else { - parser_print_warning_prefix_pos(declaration->source_position); - fprintf(stderr, "redundant declaration for '%s'\n", - symbol->string); - parser_print_warning_prefix_pos(previous_declaration->source_position); - fprintf(stderr, "previous declaration of '%s' was here\n", - symbol->string); - } - } - } else { - if (old_storage == STORAGE_CLASS_EXTERN && - new_storage == STORAGE_CLASS_EXTERN) { - parser_print_warning_prefix_pos(declaration->source_position); - fprintf(stderr, "redundant extern declaration for '%s'\n", - symbol->string); - parser_print_warning_prefix_pos(previous_declaration->source_position); - fprintf(stderr, "previous declaration of '%s' was here\n", - symbol->string); - } else { - parser_print_error_prefix_pos(declaration->source_position); - if (old_storage == new_storage) { - fprintf(stderr, "redeclaration of '%s'\n", symbol->string); - } else { - fprintf(stderr, "redeclaration of '%s' with different linkage\n", symbol->string); - } - parser_print_error_prefix_pos(previous_declaration->source_position); - fprintf(stderr, "previous declaration of '%s' was here\n", - symbol->string); - } - } - } - return previous_declaration; - } - /* remember old declaration */ stack_entry_t entry; entry.symbol = symbol; entry.old_declaration = symbol->declaration; - entry.namespc = namespc; + entry.namespc = (unsigned short) namespc; ARR_APP1(stack_entry_t, *stack_ptr, entry); /* replace/add declaration into declaration list of the symbol */ @@ -538,19 +550,19 @@ static declaration_t *stack_push(stack_entry_t **stack_ptr, iter_last->symbol_next = declaration; } } - - return declaration; } -static declaration_t *environment_push(declaration_t *declaration) +static void environment_push(declaration_t *declaration) { assert(declaration->source_position.input_name != NULL); - return stack_push(&environment_stack, declaration, context); + assert(declaration->parent_context != NULL); + stack_push(&environment_stack, declaration); } -static declaration_t *label_push(declaration_t *declaration) +static void label_push(declaration_t *declaration) { - return stack_push(&label_stack, declaration, ¤t_function->context); + declaration->parent_context = ¤t_function->context; + stack_push(&label_stack, declaration); } /** @@ -567,7 +579,7 @@ static void stack_pop_to(stack_entry_t **stack_ptr, size_t new_top) return; for(i = top; i > new_top; --i) { - stack_entry_t *entry = & stack[i - 1]; + stack_entry_t *entry = &stack[i - 1]; declaration_t *old_declaration = entry->old_declaration; symbol_t *symbol = entry->symbol; @@ -614,56 +626,82 @@ static void label_pop_to(size_t new_top) static int get_rank(const type_t *type) { + assert(!is_typeref(type)); /* The C-standard allows promoting to int or unsigned int (see § 7.2.2 * and esp. footnote 108). However we can't fold constants (yet), so we - * can't decide wether unsigned int is possible, while int always works. + * can't decide whether 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) + if(type->kind == TYPE_ENUM) return ATOMIC_TYPE_INT; - assert(type->type == TYPE_ATOMIC); - atomic_type_t *atomic_type = (atomic_type_t*) type; - atomic_type_type_t atype = atomic_type->atype; + assert(type->kind == TYPE_ATOMIC); + const atomic_type_t *atomic_type = &type->atomic; + atomic_type_kind_t atype = atomic_type->akind; return atype; } static type_t *promote_integer(type_t *type) { + if(type->kind == TYPE_BITFIELD) + return promote_integer(type->bitfield.base); + if(get_rank(type) < ATOMIC_TYPE_INT) type = type_int; return type; } +/** + * Create a cast expression. + * + * @param expression the expression to cast + * @param dest_type the destination type + */ static expression_t *create_cast_expression(expression_t *expression, type_t *dest_type) { - unary_expression_t *cast = allocate_ast_zero(sizeof(cast[0])); + expression_t *cast = allocate_expression_zero(EXPR_UNARY_CAST_IMPLICIT); - cast->expression.type = EXPR_UNARY; - cast->type = UNEXPR_CAST; - cast->value = expression; - cast->expression.datatype = dest_type; + cast->unary.value = expression; + cast->base.datatype = dest_type; - return (expression_t*) cast; + return cast; } -static bool is_null_expression(const expression_t *const expr) +/** + * Check if a given expression represents the 0 pointer constant. + */ +static bool is_null_pointer_constant(const expression_t *expression) { - if (expr->type != EXPR_CONST) return false; + /* skip void* cast */ + if(expression->kind == EXPR_UNARY_CAST + || expression->kind == EXPR_UNARY_CAST_IMPLICIT) { + expression = expression->unary.value; + } - type_t *const type = skip_typeref(expr->datatype); - if (!is_type_integer(type)) return false; + /* TODO: not correct yet, should be any constant integer expression + * which evaluates to 0 */ + if (expression->kind != EXPR_CONST) + return false; - const const_t *const const_expr = (const const_t*)expr; - return const_expr->v.int_value == 0; + type_t *const type = skip_typeref(expression->base.datatype); + if (!is_type_integer(type)) + return false; + + return expression->conste.v.int_value == 0; } +/** + * Create an implicit cast expression. + * + * @param expression the expression to cast + * @param dest_type the destination type + */ static expression_t *create_implicit_cast(expression_t *expression, type_t *dest_type) { - type_t *source_type = expression->datatype; + type_t *source_type = expression->base.datatype; if(source_type == NULL) return expression; @@ -674,27 +712,30 @@ static expression_t *create_implicit_cast(expression_t *expression, if(source_type == dest_type) return expression; - switch (dest_type->type) { + switch (dest_type->kind) { case TYPE_ENUM: /* TODO warning for implicitly converting to enum */ + case TYPE_BITFIELD: case TYPE_ATOMIC: - if (source_type->type != TYPE_ATOMIC && - source_type->type != TYPE_ENUM) { + if (source_type->kind != TYPE_ATOMIC && + source_type->kind != TYPE_ENUM && + source_type->kind != TYPE_BITFIELD) { 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->source_position, source_type, dest_type); + expression->base.source_position, source_type, + dest_type); return expression; } return create_cast_expression(expression, dest_type); case TYPE_POINTER: - switch (source_type->type) { + switch (source_type->kind) { case TYPE_ATOMIC: - if (is_null_expression(expression)) { + if (is_null_pointer_constant(expression)) { return create_cast_expression(expression, dest_type); } break; @@ -706,9 +747,8 @@ static expression_t *create_implicit_cast(expression_t *expression, break; case TYPE_ARRAY: { - array_type_t *array_type = (array_type_t*) source_type; - pointer_type_t *pointer_type - = (pointer_type_t*) dest_type; + 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); @@ -721,7 +761,7 @@ static expression_t *create_implicit_cast(expression_t *expression, } type_error_incompatible("can't implicitly cast types", - expression->source_position, source_type, dest_type); + expression->base.source_position, source_type, dest_type); return expression; default: @@ -733,7 +773,7 @@ static expression_t *create_implicit_cast(expression_t *expression, static void semantic_assign(type_t *orig_type_left, expression_t **right, const char *context) { - type_t *orig_type_right = (*right)->datatype; + type_t *orig_type_right = (*right)->base.datatype; if(orig_type_right == NULL) return; @@ -742,7 +782,7 @@ static void semantic_assign(type_t *orig_type_left, expression_t **right, 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_pointer(type_left) && is_null_pointer_constant(*right)) || (is_type_atomic(type_left, ATOMIC_TYPE_BOOL) && is_type_pointer(type_right))) { *right = create_implicit_cast(*right, type_left); @@ -750,35 +790,31 @@ static void semantic_assign(type_t *orig_type_left, expression_t **right, } if (is_type_pointer(type_left) && is_type_pointer(type_right)) { - pointer_type_t *pointer_type_left = (pointer_type_t*) type_left; - pointer_type_t *pointer_type_right = (pointer_type_t*) 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->qualifiers & ~points_to_left->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"); + errorf(HERE, "destination type '%T' in %s from type '%T' lacks qualifiers '%Q' in pointed-to type", type_left, context, type_right, missing_qualifiers); return; } + points_to_left = get_unqualified_type(points_to_left); + points_to_right = get_unqualified_type(points_to_right); + + if(!is_type_atomic(points_to_left, ATOMIC_TYPE_VOID) + && !is_type_atomic(points_to_right, ATOMIC_TYPE_VOID) + && !types_compatible(points_to_left, points_to_right)) { + goto incompatible_assign_types; + } + *right = create_implicit_cast(*right, type_left); return; } @@ -791,19 +827,20 @@ static void semantic_assign(type_t *orig_type_left, expression_t **right, 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); + errorf(HERE, "incompatible types in %s: '%T' <- '%T'", + context, orig_type_left, orig_type_right); } static expression_t *parse_constant_expression(void) { /* start parsing at precedence 7 (conditional expression) */ - return parse_sub_expression(7); + expression_t *result = parse_sub_expression(7); + + if(!is_constant_expression(result)) { + errorf(result->base.source_position, "expression '%E' is not constant\n", result); + } + + return result; } static expression_t *parse_assignment_expression(void) @@ -812,28 +849,34 @@ static expression_t *parse_assignment_expression(void) return parse_sub_expression(2); } -typedef struct declaration_specifiers_t declaration_specifiers_t; -struct declaration_specifiers_t { - storage_class_t storage_class; - bool is_inline; - type_t *type; -}; +static type_t *make_global_typedef(const char *name, type_t *type) +{ + symbol_t *const symbol = symbol_table_insert(name); -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); + declaration_t *const declaration = allocate_declaration_zero(); + declaration->namespc = NAMESPACE_NORMAL; + declaration->storage_class = STORAGE_CLASS_TYPEDEF; + declaration->type = type; + declaration->symbol = symbol; + declaration->source_position = builtin_source_position; + + record_declaration(declaration); -static const char *parse_string_literals(void) + type_t *typedef_type = allocate_type_zero(TYPE_TYPEDEF); + typedef_type->typedeft.declaration = declaration; + + return typedef_type; +} + +static string_t parse_string_literals(void) { assert(token.type == T_STRING_LITERAL); - const char *result = token.v.string; + string_t result = token.v.string; next_token(); - while(token.type == T_STRING_LITERAL) { - result = concat_strings(result, token.v.string); + while (token.type == T_STRING_LITERAL) { + result = concat_strings(&result, &token.v.string); next_token(); } @@ -844,7 +887,7 @@ static void parse_attributes(void) { while(true) { switch(token.type) { - case T___attribute__: + case T___attribute__: { next_token(); expect_void('('); @@ -852,7 +895,7 @@ static void parse_attributes(void) while(depth > 0) { switch(token.type) { case T_EOF: - parse_error("EOF while parsing attribute"); + errorf(HERE, "EOF while parsing attribute"); break; case '(': next_token(); @@ -867,13 +910,14 @@ static void parse_attributes(void) } } break; + } case T_asm: next_token(); expect_void('('); if(token.type != T_STRING_LITERAL) { parse_error_expected("while parsing assembler attribute", T_STRING_LITERAL); - eat_brace(); + eat_paren(); break; } else { parse_string_literals(); @@ -935,13 +979,26 @@ static designator_t *parse_designation(void) #endif static initializer_t *initializer_from_string(array_type_t *type, - const char *string) + const string_t *const string) +{ + /* TODO: check len vs. size of array type */ + (void) type; + + initializer_t *initializer = allocate_initializer_zero(INITIALIZER_STRING); + initializer->string.string = *string; + + return initializer; +} + +static initializer_t *initializer_from_wide_string(array_type_t *const type, + wide_string_t *const string) { /* TODO: check len vs. size of array type */ (void) type; - initializer_t *initializer = allocate_initializer(INITIALIZER_STRING); - initializer->string.string = string; + initializer_t *const initializer = + allocate_initializer_zero(INITIALIZER_WIDE_STRING); + initializer->wide_string.string = *string; return initializer; } @@ -952,31 +1009,44 @@ static initializer_t *initializer_from_expression(type_t *type, /* TODO check that expression is a constant expression */ /* § 6.7.8.14/15 char array may be initialized by string literals */ - if(type->type == TYPE_ARRAY && expression->type == EXPR_STRING_LITERAL) { - array_type_t *array_type = (array_type_t*) type; - type_t *element_type = array_type->element_type; - - if(element_type->type == TYPE_ATOMIC) { - atomic_type_t *atomic_type = (atomic_type_t*) element_type; - atomic_type_type_t atype = atomic_type->atype; + type_t *const expr_type = expression->base.datatype; + if (is_type_array(type) && expr_type->kind == TYPE_POINTER) { + array_type_t *const array_type = &type->array; + type_t *const element_type = skip_typeref(array_type->element_type); + + if (element_type->kind == TYPE_ATOMIC) { + switch (expression->kind) { + case EXPR_STRING_LITERAL: + if (element_type->atomic.akind == ATOMIC_TYPE_CHAR) { + return initializer_from_string(array_type, + &expression->string.value); + } - /* TODO handle wide strings */ - if(atype == ATOMIC_TYPE_CHAR - || atype == ATOMIC_TYPE_SCHAR - || atype == ATOMIC_TYPE_UCHAR) { + case EXPR_WIDE_STRING_LITERAL: { + type_t *bare_wchar_type = skip_typeref(type_wchar_t); + if (get_unqualified_type(element_type) == bare_wchar_type) { + return initializer_from_wide_string(array_type, + &expression->wide_string.value); + } + } - string_literal_t *literal = (string_literal_t*) expression; - return initializer_from_string(array_type, literal->value); + default: + break; } } } - semantic_assign(type, &expression, "initializer"); + type_t *expression_type = skip_typeref(expression->base.datatype); + if(is_type_scalar(type) || types_compatible(type, expression_type)) { + semantic_assign(type, &expression, "initializer"); - initializer_t *result = allocate_initializer(INITIALIZER_VALUE); - result->value.value = expression; + initializer_t *result = allocate_initializer_zero(INITIALIZER_VALUE); + result->value.value = expression; - return result; + return result; + } + + return NULL; } static initializer_t *parse_sub_initializer(type_t *type, @@ -990,13 +1060,31 @@ static initializer_t *parse_sub_initializer_elem(type_t *type) } expression_t *expression = parse_assignment_expression(); - type_t *expression_type = skip_typeref(expression->datatype); + type_t *expression_type = skip_typeref(expression->base.datatype); return parse_sub_initializer(type, expression, expression_type); } static bool had_initializer_brace_warning; +static void skip_designator(void) +{ + while(1) { + if(token.type == '.') { + next_token(); + if(token.type == T_IDENTIFIER) + next_token(); + } else if(token.type == '[') { + next_token(); + parse_constant_expression(); + if(token.type == ']') + next_token(); + } else { + break; + } + } +} + static initializer_t *parse_sub_initializer(type_t *type, expression_t *expression, type_t *expression_type) @@ -1006,7 +1094,7 @@ static initializer_t *parse_sub_initializer(type_t *type, if(token.type == '{') { next_token(); if(!had_initializer_brace_warning) { - parse_warning("braces around scalar initializer"); + warningf(HERE, "braces around scalar initializer"); had_initializer_brace_warning = true; } initializer_t *result = parse_sub_initializer(type, NULL, NULL); @@ -1024,17 +1112,11 @@ static initializer_t *parse_sub_initializer(type_t *type, return initializer_from_expression(type, expression); } - /* TODO: ignore qualifiers, comparing pointers is probably - * not correct */ - if(expression != NULL && expression_type == type) { - initializer_t *result = allocate_initializer(INITIALIZER_VALUE); - - if(type != NULL) { - semantic_assign(type, &expression, "initializer"); - } - result->value.value = expression; - - return result; + /* does the expression match the currently looked at object to initialize */ + if(expression != NULL) { + initializer_t *result = initializer_from_expression(type, expression); + if(result != NULL) + return result; } bool read_paren = false; @@ -1046,11 +1128,18 @@ static initializer_t *parse_sub_initializer(type_t *type, /* descend into subtype */ initializer_t *result = NULL; initializer_t **elems; - if(type->type == TYPE_ARRAY) { - array_type_t *array_type = (array_type_t*) type; + if(is_type_array(type)) { + array_type_t *array_type = &type->array; type_t *element_type = array_type->element_type; element_type = skip_typeref(element_type); + if(token.type == '.') { + errorf(HERE, + "compound designator in initializer for array type '%T'", + type); + skip_designator(); + } + initializer_t *sub; had_initializer_brace_warning = false; if(expression == NULL) { @@ -1076,21 +1165,26 @@ static initializer_t *parse_sub_initializer(type_t *type, if(token.type == '}') break; - initializer_t *sub - = parse_sub_initializer(element_type, NULL, NULL); + sub = parse_sub_initializer_elem(element_type); if(sub == NULL) { /* TODO error, do nicer cleanup */ - parse_error("member initializer didn't match"); + errorf(HERE, "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 = (compound_type_t*) type; - context_t *context = & compound_type->declaration->context; + assert(is_type_compound(type)); + compound_type_t *compound_type = &type->compound; + context_t *context = &compound_type->declaration->context; + + if(token.type == '[') { + errorf(HERE, + "array designator in initializer for compound type '%T'", + type); + skip_designator(); + } declaration_t *first = context->declarations; if(first == NULL) @@ -1131,10 +1225,10 @@ static initializer_t *parse_sub_initializer(type_t *type, type_t *iter_type = iter->type; iter_type = skip_typeref(iter_type); - initializer_t *sub = parse_sub_initializer(iter_type, NULL, NULL); + sub = parse_sub_initializer_elem(iter_type); if(sub == NULL) { - /* TODO error, do nicer cleanup*/ - parse_error("member initializer didn't match"); + /* TODO error, do nicer cleanup */ + errorf(HERE, "member initializer didn't match"); DEL_ARR_F(elems); return NULL; } @@ -1147,7 +1241,7 @@ static initializer_t *parse_sub_initializer(type_t *type, initializer_list_t *init = allocate_ast_zero(sizeof(init[0]) + elems_size); - init->initializer.type = INITIALIZER_LIST; + init->initializer.kind = INITIALIZER_LIST; init->len = len; memcpy(init->initializers, elems, elems_size); DEL_ARR_F(elems); @@ -1169,8 +1263,14 @@ static initializer_t *parse_initializer(type_t *type) type = skip_typeref(type); if(token.type != '{') { - expression_t *expression = parse_assignment_expression(); - return initializer_from_expression(type, expression); + expression_t *expression = parse_assignment_expression(); + initializer_t *initializer = initializer_from_expression(type, expression); + if(initializer == NULL) { + errorf(HERE, + "initializer expression '%E' of type '%T' is incompatible with type '%T'", + expression, expression->base.datatype, type); + } + return initializer; } if(is_type_scalar(type)) { @@ -1192,7 +1292,7 @@ static initializer_t *parse_initializer(type_t *type) return result; } - +static declaration_t *append_declaration(declaration_t *declaration); static declaration_t *parse_compound_type_specifier(bool is_struct) { @@ -1232,36 +1332,35 @@ static declaration_t *parse_compound_type_specifier(bool is_struct) } if(declaration == NULL) { - declaration = allocate_type_zero(sizeof(declaration[0])); - - if(is_struct) { - declaration->namespc = NAMESPACE_STRUCT; - } else { - declaration->namespc = NAMESPACE_UNION; - } + declaration = allocate_declaration_zero(); + declaration->namespc = + (is_struct ? NAMESPACE_STRUCT : NAMESPACE_UNION); declaration->source_position = token.source_position; declaration->symbol = symbol; - record_declaration(declaration); + declaration->parent_context = context; + if (symbol != NULL) { + environment_push(declaration); + } + append_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); + errorf(HERE, "multiple definition of '%s %Y'", + is_struct ? "struct" : "union", symbol); declaration->context.declarations = NULL; } declaration->init.is_defined = true; int top = environment_top(); context_t *last_context = context; - set_context(& declaration->context); + set_context(&declaration->context); parse_compound_type_entries(); parse_attributes(); - assert(context == & declaration->context); + assert(context == &declaration->context); set_context(last_context); environment_pop_to(top); } @@ -1269,26 +1368,26 @@ static declaration_t *parse_compound_type_specifier(bool is_struct) return declaration; } -static void parse_enum_entries(enum_type_t *const enum_type) +static void parse_enum_entries(type_t *const enum_type) { eat('{'); if(token.type == '}') { next_token(); - parse_error("empty enum not allowed"); + errorf(HERE, "empty enum not allowed"); return; } do { - declaration_t *entry = allocate_ast_zero(sizeof(entry[0])); - if(token.type != T_IDENTIFIER) { parse_error_expected("while parsing enum entry", T_IDENTIFIER, 0); eat_block(); return; } + + declaration_t *const entry = allocate_declaration_zero(); entry->storage_class = STORAGE_CLASS_ENUM_ENTRY; - entry->type = (type_t*) enum_type; + entry->type = enum_type; entry->symbol = token.v.symbol; entry->source_position = token.source_position; next_token(); @@ -1332,31 +1431,31 @@ static type_t *parse_enum_specifier(void) } if(declaration == NULL) { - declaration = allocate_type_zero(sizeof(declaration[0])); - - declaration->namespc = NAMESPACE_ENUM; + declaration = allocate_declaration_zero(); + declaration->namespc = NAMESPACE_ENUM; declaration->source_position = token.source_position; declaration->symbol = symbol; + declaration->parent_context = context; } - enum_type_t *const enum_type = allocate_type_zero(sizeof(enum_type[0])); - enum_type->type.type = TYPE_ENUM; - enum_type->declaration = declaration; + type_t *const type = allocate_type_zero(TYPE_ENUM); + type->enumt.declaration = declaration; if(token.type == '{') { if(declaration->init.is_defined) { - parser_print_error_prefix(); - fprintf(stderr, "multiple definitions of enum %s\n", - symbol->string); + errorf(HERE, "multiple definitions of enum %Y", symbol); } - record_declaration(declaration); + if (symbol != NULL) { + environment_push(declaration); + } + append_declaration(declaration); declaration->init.is_defined = 1; - parse_enum_entries(enum_type); + parse_enum_entries(type); parse_attributes(); } - return (type_t*) enum_type; + return type; } /** @@ -1396,7 +1495,7 @@ restart: type = parse_typename(); } else { expression = parse_expression(); - type = expression->datatype; + type = expression->base.datatype; } break; @@ -1406,18 +1505,17 @@ restart: default: expression = parse_expression(); - type = expression->datatype; + type = expression->base.datatype; break; } expect(')'); - typeof_type_t *typeof = allocate_type_zero(sizeof(typeof[0])); - typeof->type.type = TYPE_TYPEOF; - typeof->expression = expression; - typeof->typeof_type = type; + type_t *typeof_type = allocate_type_zero(TYPE_TYPEOF); + typeof_type->typeoft.expression = expression; + typeof_type->typeoft.typeof_type = type; - return (type_t*) typeof; + return typeof_type; } typedef enum { @@ -1438,15 +1536,19 @@ typedef enum { #endif } specifiers_t; -static type_t *create_builtin_type(symbol_t *symbol) +static type_t *create_builtin_type(symbol_t *const symbol, + type_t *const real_type) { - builtin_type_t *type = allocate_type_zero(sizeof(type[0])); - type->type.type = TYPE_BUILTIN; - type->symbol = symbol; - /* TODO... */ - type->real_type = type_int; + type_t *type = allocate_type_zero(TYPE_BUILTIN); + type->builtin.symbol = symbol; + type->builtin.real_type = real_type; + + type_t *result = typehash_insert(type); + if (type != result) { + free_type(type); + } - return (type_t*) type; + return result; } static type_t *get_typedef_type(symbol_t *symbol) @@ -1456,19 +1558,20 @@ static type_t *get_typedef_type(symbol_t *symbol) || declaration->storage_class != STORAGE_CLASS_TYPEDEF) return NULL; - typedef_type_t *typedef_type = allocate_type_zero(sizeof(typedef_type[0])); - typedef_type->type.type = TYPE_TYPEDEF; - typedef_type->declaration = declaration; + type_t *type = allocate_type_zero(TYPE_TYPEDEF); + type->typedeft.declaration = declaration; - return (type_t*) typedef_type; + return type; } static void parse_declaration_specifiers(declaration_specifiers_t *specifiers) { - type_t *type = NULL; - unsigned type_qualifiers = 0; - unsigned type_specifiers = 0; - int newtype = 0; + type_t *type = NULL; + unsigned type_qualifiers = 0; + unsigned type_specifiers = 0; + int newtype = 0; + + specifiers->source_position = token.source_position; while(true) { switch(token.type) { @@ -1477,8 +1580,7 @@ static void parse_declaration_specifiers(declaration_specifiers_t *specifiers) #define MATCH_STORAGE_CLASS(token, class) \ case token: \ if(specifiers->storage_class != STORAGE_CLASS_NONE) { \ - parse_error("multiple storage classes in declaration " \ - "specifiers"); \ + errorf(HERE, "multiple storage classes in declaration specifiers"); \ } \ specifiers->storage_class = class; \ next_token(); \ @@ -1490,10 +1592,31 @@ 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) - /* type qualifiers */ -#define MATCH_TYPE_QUALIFIER(token, qualifier) \ - case token: \ - type_qualifiers |= qualifier; \ + 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: + errorf(HERE, "multiple storage classes in declaration specifiers"); + break; + } + next_token(); + break; + + /* type qualifiers */ +#define MATCH_TYPE_QUALIFIER(token, qualifier) \ + case token: \ + type_qualifiers |= qualifier; \ next_token(); \ break; @@ -1511,7 +1634,7 @@ static void parse_declaration_specifiers(declaration_specifiers_t *specifiers) case token: \ next_token(); \ if(type_specifiers & specifier) { \ - parse_error("multiple " name " type specifiers given"); \ + errorf(HERE, "multiple " name " type specifiers given"); \ } else { \ type_specifiers |= specifier; \ } \ @@ -1530,6 +1653,10 @@ static void parse_declaration_specifiers(declaration_specifiers_t *specifiers) MATCH_SPECIFIER(T__Complex, SPECIFIER_COMPLEX, "_Complex") MATCH_SPECIFIER(T__Imaginary, SPECIFIER_IMAGINARY, "_Imaginary") #endif + case T_forceinline: + /* only in microsoft mode */ + specifiers->decl_modifiers |= DM_FORCEINLINE; + case T_inline: next_token(); specifiers->is_inline = true; @@ -1538,7 +1665,7 @@ static void parse_declaration_specifiers(declaration_specifiers_t *specifiers) case T_long: next_token(); if(type_specifiers & SPECIFIER_LONG_LONG) { - parse_error("multiple type specifiers given"); + errorf(HERE, "multiple type specifiers given"); } else if(type_specifiers & SPECIFIER_LONG) { type_specifiers |= SPECIFIER_LONG_LONG; } else { @@ -1549,19 +1676,15 @@ static void parse_declaration_specifiers(declaration_specifiers_t *specifiers) /* TODO: if type != NULL for the following rules should issue * an error */ case T_struct: { - type = allocate_type_zero(sizeof(struct compound_type_t)); + type = allocate_type_zero(TYPE_COMPOUND_STRUCT); - compound_type_t *compound_type = (compound_type_t*) type; - compound_type->type.type = TYPE_COMPOUND_STRUCT; - compound_type->declaration = parse_compound_type_specifier(true); + type->compound.declaration = parse_compound_type_specifier(true); break; } case T_union: { - type = allocate_type_zero(sizeof(compound_type_t)); + type = allocate_type_zero(TYPE_COMPOUND_STRUCT); - compound_type_t *compound_type = (compound_type_t*) type; - compound_type->type.type = TYPE_COMPOUND_UNION; - compound_type->declaration = parse_compound_type_specifier(false); + type->compound.declaration = parse_compound_type_specifier(false); break; } case T_enum: @@ -1571,7 +1694,7 @@ static void parse_declaration_specifiers(declaration_specifiers_t *specifiers) type = parse_typeof(); break; case T___builtin_va_list: - type = create_builtin_type(token.v.symbol); + type = duplicate_type(type_valist); next_token(); break; @@ -1600,7 +1723,7 @@ static void parse_declaration_specifiers(declaration_specifiers_t *specifiers) finish_specifiers: if(type == NULL) { - atomic_type_type_t atomic_type; + atomic_type_kind_t atomic_type; /* match valid basic types */ switch(type_specifiers) { @@ -1692,40 +1815,37 @@ finish_specifiers: default: /* invalid specifier combination, give an error message */ if(type_specifiers == 0) { -#ifndef STRICT_C99 - parse_warning("no type specifiers in declaration (using int)"); - atomic_type = ATOMIC_TYPE_INT; - break; -#else - parse_error("no type specifiers given in declaration"); -#endif + if (! strict_mode) { + warningf(HERE, "no type specifiers in declaration, using int"); + atomic_type = ATOMIC_TYPE_INT; + break; + } else { + errorf(HERE, "no type specifiers given in declaration"); + } } else if((type_specifiers & SPECIFIER_SIGNED) && (type_specifiers & SPECIFIER_UNSIGNED)) { - parse_error("signed and unsigned specifiers gives"); + errorf(HERE, "signed and unsigned specifiers gives"); } else if(type_specifiers & (SPECIFIER_SIGNED | SPECIFIER_UNSIGNED)) { - parse_error("only integer types can be signed or unsigned"); + errorf(HERE, "only integer types can be signed or unsigned"); } else { - parse_error("multiple datatypes in declaration"); + errorf(HERE, "multiple datatypes in declaration"); } 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.akind = atomic_type; + newtype = 1; } else { if(type_specifiers != 0) { - parse_error("multiple datatypes in declaration"); + errorf(HERE, "multiple datatypes in declaration"); } } - type->qualifiers = (type_qualifier_t)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); } @@ -1749,18 +1869,61 @@ static type_qualifiers_t parse_type_qualifiers(void) } } -static void parse_identifier_list(void) +static declaration_t *parse_identifier_list(void) { - while(true) { - if(token.type != T_IDENTIFIER) { - parse_error_expected("while parsing parameter identifier list", - T_IDENTIFIER, 0); - return; - } + declaration_t *declarations = NULL; + declaration_t *last_declaration = NULL; + do { + declaration_t *const declaration = allocate_declaration_zero(); + declaration->source_position = token.source_position; + declaration->symbol = token.v.symbol; next_token(); + + if(last_declaration != NULL) { + last_declaration->next = declaration; + } else { + declarations = declaration; + } + last_declaration = declaration; + if(token.type != ',') break; next_token(); + } while(token.type == T_IDENTIFIER); + + return declarations; +} + +static void semantic_parameter(declaration_t *declaration) +{ + /* TODO: improve error messages */ + + if(declaration->storage_class == STORAGE_CLASS_TYPEDEF) { + errorf(HERE, "typedef not allowed in parameter list"); + } else if(declaration->storage_class != STORAGE_CLASS_NONE + && declaration->storage_class != STORAGE_CLASS_REGISTER) { + errorf(HERE, "parameter may only have none or register storage class"); + } + + type_t *orig_type = declaration->type; + if(orig_type == NULL) + return; + type_t *type = skip_typeref(orig_type); + + /* Array as last part of a parameter type is just syntactic sugar. Turn it + * into a pointer. § 6.7.5.3 (7) */ + if (is_type_array(type)) { + const array_type_t *arr_type = &type->array; + type_t *element_type = arr_type->element_type; + + type = make_pointer_type(element_type, type->base.qualifiers); + + declaration->type = type; + } + + if(is_type_incomplete(type)) { + errorf(HERE, "incomplete type ('%T') not allowed for parameter '%Y'", + orig_type, declaration->symbol); } } @@ -1771,22 +1934,9 @@ static declaration_t *parse_parameter(void) parse_declaration_specifiers(&specifiers); - 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"); - } + declaration_t *declaration = parse_declarator(&specifiers, /*may_be_abstract=*/true); - /* Array as last part of a paramter type is just syntactic sugar. Turn it - * into a pointer */ - if (declaration->type->type == TYPE_ARRAY) { - const array_type_t *const arr_type = - (const array_type_t*)declaration->type; - type_t *element_type = arr_type->element_type; - declaration->type = make_pointer_type(element_type, TYPE_QUALIFIER_NONE); - } + semantic_parameter(declaration); return declaration; } @@ -1794,11 +1944,10 @@ static declaration_t *parse_parameter(void) static declaration_t *parse_parameters(function_type_t *type) { if(token.type == T_IDENTIFIER) { - symbol_t *symbol = token.v.symbol; + symbol_t *symbol = token.v.symbol; if(!is_typedef_symbol(symbol)) { - /* TODO: K&R style C parameters */ - parse_identifier_list(); - return NULL; + type->kr_style_parameters = true; + return parse_identifier_list(); } } @@ -1829,7 +1978,8 @@ static declaration_t *parse_parameters(function_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) { @@ -1873,8 +2023,8 @@ struct parsed_pointer_t { typedef struct construct_function_type_t construct_function_type_t; struct construct_function_type_t { - construct_type_t construct_type; - function_type_t *function_type; + construct_type_t construct_type; + type_t *function_type; }; typedef struct parsed_array_t parsed_array_t; @@ -1942,10 +2092,9 @@ static construct_type_t *parse_function_declarator(declaration_t *declaration) { eat('('); - function_type_t *type = allocate_type_zero(sizeof(type[0])); - type->type.type = TYPE_FUNCTION; + type_t *type = allocate_type_zero(TYPE_FUNCTION); - declaration_t *parameters = parse_parameters(type); + declaration_t *parameters = parse_parameters(&type->function); if(declaration != NULL) { declaration->context.declarations = parameters; } @@ -1990,7 +2139,7 @@ static construct_type_t *parse_inner_declarator(declaration_t *declaration, switch(token.type) { case T_IDENTIFIER: if(declaration == NULL) { - parse_error("no identifier expected in typename"); + errorf(HERE, "no identifier expected in typename"); } else { declaration->symbol = token.v.symbol; declaration->source_position = token.source_position; @@ -2064,49 +2213,47 @@ 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_function_type_t *construct_function_type; - function_type_t *function_type; - pointer_type_t *pointer_type; - array_type_t *array_type; - switch(iter->type) { case CONSTRUCT_INVALID: panic("invalid type construction found"); - case CONSTRUCT_FUNCTION: - construct_function_type = (construct_function_type_t*) iter; - function_type = construct_function_type->function_type; + case CONSTRUCT_FUNCTION: { + construct_function_type_t *construct_function_type + = (construct_function_type_t*) iter; + + type_t *function_type = construct_function_type->function_type; - function_type->result_type = type; - type = (type_t*) function_type; + function_type->function.return_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) { /* the function type was constructed earlier freeing it here will * destroy other types... */ @@ -2121,12 +2268,13 @@ static type_t *construct_declarator_type(construct_type_t *construct_list, } static declaration_t *parse_declarator( - const declaration_specifiers_t *specifiers, - type_t *type, bool may_be_abstract) + const declaration_specifiers_t *specifiers, 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; + type_t *type = specifiers->type; + declaration_t *const declaration = allocate_declaration_zero(); + declaration->storage_class = specifiers->storage_class; + declaration->modifiers = specifiers->decl_modifiers; + declaration->is_inline = specifiers->is_inline; construct_type_t *construct_type = parse_inner_declarator(declaration, may_be_abstract); @@ -2151,250 +2299,608 @@ static type_t *parse_abstract_declarator(type_t *base_type) return result; } -static declaration_t *record_declaration(declaration_t *declaration) +static declaration_t *append_declaration(declaration_t* const declaration) { - assert(context != NULL); - - symbol_t *symbol = declaration->symbol; - if(symbol != NULL) { - declaration_t *alias = environment_push(declaration); - if(alias != declaration) - return alias; - } else { - declaration->parent_context = context; - } - - if(last_declaration != NULL) { + if (last_declaration != NULL) { last_declaration->next = declaration; } else { context->declarations = declaration; } last_declaration = declaration; - return declaration; } -static void parser_error_multiple_definition(declaration_t *previous, - declaration_t *declaration) +static declaration_t *internal_record_declaration( + declaration_t *const declaration, + const bool is_function_definition) { - parser_print_error_prefix_pos(declaration->source_position); - fprintf(stderr, "multiple definition of symbol '%s'\n", - declaration->symbol->string); - parser_print_error_prefix_pos(previous->source_position); - fprintf(stderr, "this is the location of the previous definition.\n"); + const symbol_t *const symbol = declaration->symbol; + const namespace_t namespc = (namespace_t)declaration->namespc; + + const type_t *const type = skip_typeref(declaration->type); + if (is_type_function(type) && type->function.unspecified_parameters) { + warningf(declaration->source_position, + "function declaration '%#T' is not a prototype", + type, declaration->symbol); + } + + declaration_t *const previous_declaration = get_declaration(symbol, namespc); + assert(declaration != previous_declaration); + if (previous_declaration != NULL + && previous_declaration->parent_context == context) { + /* can happen for K&R style declarations */ + if(previous_declaration->type == NULL) { + previous_declaration->type = declaration->type; + } + + const type_t *const prev_type = skip_typeref(previous_declaration->type); + if (!types_compatible(type, prev_type)) { + errorf(declaration->source_position, + "declaration '%#T' is incompatible with previous declaration '%#T'", + type, symbol, previous_declaration->type, symbol); + errorf(previous_declaration->source_position, "previous declaration of '%Y' was here", symbol); + } else { + unsigned old_storage_class = previous_declaration->storage_class; + unsigned new_storage_class = declaration->storage_class; + + /* pretend no storage class means extern for function declarations + * (except if the previous declaration is neither none nor extern) */ + if (is_type_function(type)) { + switch (old_storage_class) { + case STORAGE_CLASS_NONE: + old_storage_class = STORAGE_CLASS_EXTERN; + + case STORAGE_CLASS_EXTERN: + if (new_storage_class == STORAGE_CLASS_NONE && !is_function_definition) { + new_storage_class = STORAGE_CLASS_EXTERN; + } + break; + + default: break; + } + } + + if (old_storage_class == STORAGE_CLASS_EXTERN && + new_storage_class == STORAGE_CLASS_EXTERN) { +warn_redundant_declaration: + warningf(declaration->source_position, "redundant declaration for '%Y'", symbol); + warningf(previous_declaration->source_position, "previous declaration of '%Y' was here", symbol); + } else if (current_function == NULL) { + if (old_storage_class != STORAGE_CLASS_STATIC && + new_storage_class == STORAGE_CLASS_STATIC) { + errorf(declaration->source_position, "static declaration of '%Y' follows non-static declaration", symbol); + errorf(previous_declaration->source_position, "previous declaration of '%Y' was here", symbol); + } else { + if (old_storage_class != STORAGE_CLASS_EXTERN) { + goto warn_redundant_declaration; + } + if (new_storage_class == STORAGE_CLASS_NONE) { + previous_declaration->storage_class = STORAGE_CLASS_NONE; + } + } + } else { + if (old_storage_class == new_storage_class) { + errorf(declaration->source_position, "redeclaration of '%Y'", symbol); + } else { + errorf(declaration->source_position, "redeclaration of '%Y' with different linkage", symbol); + } + errorf(previous_declaration->source_position, "previous declaration of '%Y' was here", symbol); + } + } + return previous_declaration; + } + + assert(declaration->parent_context == NULL); + assert(declaration->symbol != NULL); + assert(context != NULL); + + declaration->parent_context = context; + + environment_push(declaration); + return append_declaration(declaration); } -static void parse_init_declarators(const declaration_specifiers_t *specifiers) +static declaration_t *record_declaration(declaration_t *declaration) { - while(true) { - declaration_t *ndeclaration - = parse_declarator(specifiers, specifiers->type, false); + return internal_record_declaration(declaration, false); +} - declaration_t *declaration = record_declaration(ndeclaration); +static declaration_t *record_function_definition(declaration_t *const declaration) +{ + return internal_record_declaration(declaration, true); +} - 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); - } +static void parser_error_multiple_definition(declaration_t *declaration, + const source_position_t source_position) +{ + errorf(source_position, "multiple definition of symbol '%Y'", + declaration->symbol); + errorf(declaration->source_position, + "this is the location of the previous definition."); +} - if(token.type == '=') { - next_token(); +static bool is_declaration_specifier(const token_t *token, + bool only_type_specifiers) +{ + switch(token->type) { + TYPE_SPECIFIERS + return true; + case T_IDENTIFIER: + return is_typedef_symbol(token->v.symbol); - /* TODO: check that this is an allowed type (no function type) */ + case T___extension__: + STORAGE_CLASSES + TYPE_QUALIFIERS + return !only_type_specifiers; - if(declaration->init.initializer != NULL) { - parser_error_multiple_definition(declaration, ndeclaration); - } + default: + return false; + } +} - initializer_t *initializer = parse_initializer(type); +static void parse_init_declarator_rest(declaration_t *declaration) +{ + eat('='); - if(type->type == TYPE_ARRAY && initializer != NULL) { - array_type_t *array_type = (array_type_t*) type; + type_t *orig_type = declaration->type; + type_t *type = NULL; + if(orig_type != NULL) + type = skip_typeref(orig_type); - if(array_type->size == NULL) { - const_t *cnst = allocate_ast_zero(sizeof(cnst[0])); + if(declaration->init.initializer != NULL) { + parser_error_multiple_definition(declaration, token.source_position); + } - cnst->expression.type = EXPR_CONST; - cnst->expression.datatype = type_size_t; + initializer_t *initializer = parse_initializer(type); - if(initializer->type == INITIALIZER_LIST) { - initializer_list_t *list = &initializer->list; - cnst->v.int_value = list->len; - } else { - assert(initializer->type == INITIALIZER_STRING); - initializer_string_t *string = &initializer->string; - cnst->v.int_value = strlen(string->string) + 1; - } + /* § 6.7.5 (22) array initializers for arrays with unknown size determine + * the array type size */ + if(type != NULL && is_type_array(type) && initializer != NULL) { + array_type_t *array_type = &type->array; - array_type->size = (expression_t*) cnst; - } - } + if(array_type->size == NULL) { + expression_t *cnst = allocate_expression_zero(EXPR_CONST); + cnst->base.datatype = type_size_t; - ndeclaration->init.initializer = initializer; - } else if(token.type == '{') { - if(type->type != TYPE_FUNCTION) { - parser_print_error_prefix(); - 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; - } + switch (initializer->kind) { + case INITIALIZER_LIST: { + initializer_list_t *const list = &initializer->list; + cnst->conste.v.int_value = list->len; + break; + } - if(declaration->init.statement != NULL) { - parser_error_multiple_definition(declaration, ndeclaration); - } - if(ndeclaration != declaration) { - memcpy(&declaration->context, &ndeclaration->context, - sizeof(declaration->context)); - } + case INITIALIZER_STRING: { + initializer_string_t *const string = &initializer->string; + cnst->conste.v.int_value = string->string.size; + break; + } - int top = environment_top(); - context_t *last_context = context; - set_context(&declaration->context); + case INITIALIZER_WIDE_STRING: { + initializer_wide_string_t *const string = &initializer->wide_string; + cnst->conste.v.int_value = string->string.size; + break; + } - /* push function parameters */ - declaration_t *parameter = declaration->context.declarations; - for( ; parameter != NULL; parameter = parameter->next) { - environment_push(parameter); + default: + panic("invalid initializer type"); } - int label_stack_top = label_top(); - declaration_t *old_current_function = current_function; - current_function = declaration; + array_type->size = cnst; + } + } + + if(type != NULL && is_type_function(type)) { + errorf(declaration->source_position, + "initializers not allowed for function types at declator '%Y' (type '%T')", + declaration->symbol, orig_type); + } else { + declaration->init.initializer = initializer; + } +} - statement_t *statement = parse_compound_statement(); +/* parse rest of a declaration without any declarator */ +static void parse_anonymous_declaration_rest( + const declaration_specifiers_t *specifiers, + parsed_declaration_func finished_declaration) +{ + eat(';'); - assert(current_function == declaration); - current_function = old_current_function; - label_pop_to(label_stack_top); + declaration_t *const declaration = allocate_declaration_zero(); + declaration->type = specifiers->type; + declaration->storage_class = specifiers->storage_class; + declaration->source_position = specifiers->source_position; - assert(context == &declaration->context); - set_context(last_context); - environment_pop_to(top); + if (declaration->storage_class != STORAGE_CLASS_NONE) { + warningf(declaration->source_position, "useless storage class in empty declaration"); + } - declaration->init.statement = statement; - return; + type_t *type = declaration->type; + switch (type->kind) { + case TYPE_COMPOUND_STRUCT: + case TYPE_COMPOUND_UNION: { + const compound_type_t *compound_type = &type->compound; + if (compound_type->declaration->symbol == NULL) { + warningf(declaration->source_position, "unnamed struct/union that defines no instances"); + } + break; } - if(token.type != ',') + case TYPE_ENUM: + break; + + default: + warningf(declaration->source_position, "empty declaration"); break; - next_token(); } - expect_void(';'); + + finished_declaration(declaration); } -static void parse_struct_declarators(const declaration_specifiers_t *specifiers) +static void parse_declaration_rest(declaration_t *ndeclaration, + const declaration_specifiers_t *specifiers, + parsed_declaration_func finished_declaration) { - while(1) { - if(token.type == ':') { - next_token(); - parse_constant_expression(); - /* TODO (bitfields) */ - } else { - declaration_t *declaration - = parse_declarator(specifiers, specifiers->type, true); + while(true) { + declaration_t *declaration = finished_declaration(ndeclaration); - /* TODO: check constraints for struct declarations */ - /* TODO: check for doubled fields */ - record_declaration(declaration); + type_t *orig_type = declaration->type; + type_t *type = skip_typeref(orig_type); - if(token.type == ':') { - next_token(); - parse_constant_expression(); - /* TODO (bitfields) */ - } + if(type->kind != TYPE_FUNCTION && declaration->is_inline) { + warningf(declaration->source_position, + "variable '%Y' declared 'inline'\n", declaration->symbol); + } + + if(token.type == '=') { + parse_init_declarator_rest(declaration); } if(token.type != ',') break; - next_token(); + eat(','); + + ndeclaration = parse_declarator(specifiers, /*may_be_abstract=*/false); } expect_void(';'); } -static void parse_compound_type_entries(void) +static declaration_t *finished_kr_declaration(declaration_t *declaration) { - eat('{'); - - while(token.type != '}' && token.type != T_EOF) { - declaration_specifiers_t specifiers; - memset(&specifiers, 0, sizeof(specifiers)); - parse_declaration_specifiers(&specifiers); + symbol_t *symbol = declaration->symbol; + if(symbol == NULL) { + errorf(HERE, "anonymous declaration not valid as function parameter"); + return declaration; + } + namespace_t namespc = (namespace_t) declaration->namespc; + if(namespc != NAMESPACE_NORMAL) { + return record_declaration(declaration); + } - parse_struct_declarators(&specifiers); + declaration_t *previous_declaration = get_declaration(symbol, namespc); + if(previous_declaration == NULL || + previous_declaration->parent_context != context) { + errorf(HERE, "expected declaration of a function parameter, found '%Y'", + symbol); + return declaration; } - if(token.type == T_EOF) { - parse_error("unexpected error while parsing struct"); + + if(previous_declaration->type == NULL) { + previous_declaration->type = declaration->type; + previous_declaration->storage_class = declaration->storage_class; + previous_declaration->parent_context = context; + return previous_declaration; + } else { + return record_declaration(declaration); } - next_token(); } -static void parse_declaration(void) +static void parse_declaration(parsed_declaration_func finished_declaration) { - source_position_t source_position = token.source_position; - declaration_specifiers_t specifiers; memset(&specifiers, 0, sizeof(specifiers)); parse_declaration_specifiers(&specifiers); if(token.type == ';') { - if (specifiers.storage_class != STORAGE_CLASS_NONE) { - parse_warning_pos(source_position, - "useless keyword in empty declaration"); - } - switch (specifiers.type->type) { - case TYPE_COMPOUND_STRUCT: - case TYPE_COMPOUND_UNION: { - const compound_type_t *const comp_type = - (const compound_type_t*)specifiers.type; - if (comp_type->declaration->symbol == NULL) { - parse_warning_pos(source_position, - "unnamed struct/union that defines no instances"); - } - break; + parse_anonymous_declaration_rest(&specifiers, finished_declaration); + } else { + declaration_t *declaration = parse_declarator(&specifiers, /*may_be_abstract=*/false); + parse_declaration_rest(declaration, &specifiers, finished_declaration); + } +} + +static void parse_kr_declaration_list(declaration_t *declaration) +{ + type_t *type = skip_typeref(declaration->type); + if(!is_type_function(type)) + return; + + if(!type->function.kr_style_parameters) + return; + + /* push function parameters */ + int top = environment_top(); + context_t *last_context = context; + set_context(&declaration->context); + + declaration_t *parameter = declaration->context.declarations; + for( ; parameter != NULL; parameter = parameter->next) { + assert(parameter->parent_context == NULL); + parameter->parent_context = context; + environment_push(parameter); + } + + /* parse declaration list */ + while(is_declaration_specifier(&token, false)) { + parse_declaration(finished_kr_declaration); + } + + /* pop function parameters */ + assert(context == &declaration->context); + set_context(last_context); + environment_pop_to(top); + + /* update function type */ + type_t *new_type = duplicate_type(type); + new_type->function.kr_style_parameters = false; + + function_parameter_t *parameters = NULL; + function_parameter_t *last_parameter = NULL; + + declaration_t *parameter_declaration = declaration->context.declarations; + for( ; parameter_declaration != NULL; + parameter_declaration = parameter_declaration->next) { + type_t *parameter_type = parameter_declaration->type; + if(parameter_type == NULL) { + if (strict_mode) { + errorf(HERE, "no type specified for function parameter '%Y'", + parameter_declaration->symbol); + } else { + warningf(HERE, "no type specified for function parameter '%Y', using int", + parameter_declaration->symbol); + parameter_type = type_int; + parameter_declaration->type = parameter_type; } + } - case TYPE_ENUM: break; + semantic_parameter(parameter_declaration); + parameter_type = parameter_declaration->type; - default: - parse_warning_pos(source_position, "empty declaration"); - break; + function_parameter_t *function_parameter + = obstack_alloc(type_obst, sizeof(function_parameter[0])); + memset(function_parameter, 0, sizeof(function_parameter[0])); + + function_parameter->type = parameter_type; + if(last_parameter != NULL) { + last_parameter->next = function_parameter; + } else { + parameters = function_parameter; } + last_parameter = function_parameter; + } + new_type->function.parameters = parameters; - next_token(); + type = typehash_insert(new_type); + if(type != new_type) { + obstack_free(type_obst, new_type); + } - declaration_t *declaration = allocate_ast_zero(sizeof(declaration[0])); + declaration->type = type; +} - declaration->type = specifiers.type; - declaration->storage_class = specifiers.storage_class; - declaration->source_position = source_position; - record_declaration(declaration); - return; +/** + * Check if all labels are defined in the current function. + */ +static void check_for_missing_labels(void) +{ + bool first_err = true; + for (const goto_statement_t *goto_statement = goto_first; + goto_statement != NULL; + goto_statement = goto_statement->next) { + const declaration_t *label = goto_statement->label; + + if (label->source_position.input_name == NULL) { + if (first_err) { + first_err = false; + diagnosticf("%s: In function '%Y':\n", + current_function->source_position.input_name, + current_function->symbol); + } + errorf(goto_statement->statement.source_position, + "label '%Y' used but not defined", label->symbol); + } } - parse_init_declarators(&specifiers); + goto_first = goto_last = NULL; } -static type_t *parse_typename(void) +static void parse_external_declaration(void) { + /* function-definitions and declarations both start with declaration + * specifiers */ declaration_specifiers_t specifiers; memset(&specifiers, 0, sizeof(specifiers)); parse_declaration_specifiers(&specifiers); - if(specifiers.storage_class != STORAGE_CLASS_NONE) { - /* TODO: improve error message, user does probably not know what a - * storage class is... - */ - parse_error("typename may not have a storage class"); - } - type_t *result = parse_abstract_declarator(specifiers.type); + /* must be a declaration */ + if(token.type == ';') { + parse_anonymous_declaration_rest(&specifiers, append_declaration); + return; + } - return result; -} + /* declarator is common to both function-definitions and declarations */ + declaration_t *ndeclaration = parse_declarator(&specifiers, /*may_be_abstract=*/false); + + /* must be a declaration */ + if(token.type == ',' || token.type == '=' || token.type == ';') { + parse_declaration_rest(ndeclaration, &specifiers, record_declaration); + return; + } + + /* must be a function definition */ + parse_kr_declaration_list(ndeclaration); + + if(token.type != '{') { + parse_error_expected("while parsing function definition", '{', 0); + eat_statement(); + return; + } + + type_t *type = ndeclaration->type; + if(type == NULL) { + eat_block(); + return; + } + + /* note that we don't skip typerefs: the standard doesn't allow them here + * (so we can't use is_type_function here) */ + if(type->kind != TYPE_FUNCTION) { + errorf(HERE, "declarator '%#T' has a body but is not a function type", + type, ndeclaration->symbol); + eat_block(); + return; + } + + /* § 6.7.5.3 (14) a function definition with () means no + * parameters (and not unspecified parameters) */ + if(type->function.unspecified_parameters) { + type_t *duplicate = duplicate_type(type); + duplicate->function.unspecified_parameters = false; + + type = typehash_insert(duplicate); + if(type != duplicate) { + obstack_free(type_obst, duplicate); + } + ndeclaration->type = type; + } + + declaration_t *const declaration = record_function_definition(ndeclaration); + if(ndeclaration != declaration) { + declaration->context = ndeclaration->context; + } + type = skip_typeref(declaration->type); + + /* push function parameters and switch context */ + int top = environment_top(); + context_t *last_context = context; + set_context(&declaration->context); + + declaration_t *parameter = declaration->context.declarations; + for( ; parameter != NULL; parameter = parameter->next) { + if(parameter->parent_context == &ndeclaration->context) { + parameter->parent_context = context; + } + assert(parameter->parent_context == NULL + || parameter->parent_context == context); + parameter->parent_context = context; + environment_push(parameter); + } + + if(declaration->init.statement != NULL) { + parser_error_multiple_definition(declaration, token.source_position); + eat_block(); + goto end_of_parse_external_declaration; + } else { + /* parse function body */ + int label_stack_top = label_top(); + declaration_t *old_current_function = current_function; + current_function = declaration; + + declaration->init.statement = parse_compound_statement(); + check_for_missing_labels(); + + assert(current_function == declaration); + current_function = old_current_function; + label_pop_to(label_stack_top); + } + +end_of_parse_external_declaration: + assert(context == &declaration->context); + set_context(last_context); + environment_pop_to(top); +} + +static type_t *make_bitfield_type(type_t *base, expression_t *size) +{ + type_t *type = allocate_type_zero(TYPE_BITFIELD); + type->bitfield.base = base; + type->bitfield.size = size; + + return type; +} + +static void parse_struct_declarators(const declaration_specifiers_t *specifiers) +{ + /* TODO: check constraints for struct declarations (in specifiers) */ + while(1) { + declaration_t *declaration; + + if(token.type == ':') { + next_token(); + + type_t *base_type = specifiers->type; + expression_t *size = parse_constant_expression(); + + type_t *type = make_bitfield_type(base_type, size); + + declaration = allocate_declaration_zero(); + declaration->namespc = NAMESPACE_NORMAL; + declaration->storage_class = STORAGE_CLASS_NONE; + declaration->source_position = token.source_position; + declaration->modifiers = specifiers->decl_modifiers; + declaration->type = type; + } else { + declaration = parse_declarator(specifiers,/*may_be_abstract=*/true); + + if(token.type == ':') { + next_token(); + expression_t *size = parse_constant_expression(); + + type_t *type = make_bitfield_type(declaration->type, size); + declaration->type = type; + } + } + record_declaration(declaration); + + if(token.type != ',') + break; + next_token(); + } + expect_void(';'); +} + +static void parse_compound_type_entries(void) +{ + eat('{'); + + while(token.type != '}' && token.type != T_EOF) { + declaration_specifiers_t specifiers; + memset(&specifiers, 0, sizeof(specifiers)); + parse_declaration_specifiers(&specifiers); + + parse_struct_declarators(&specifiers); + } + if(token.type == T_EOF) { + errorf(HERE, "EOF while parsing struct"); + } + next_token(); +} + +static type_t *parse_typename(void) +{ + declaration_specifiers_t specifiers; + memset(&specifiers, 0, sizeof(specifiers)); + parse_declaration_specifiers(&specifiers); + if(specifiers.storage_class != STORAGE_CLASS_NONE) { + /* TODO: improve error message, user does probably not know what a + * storage class is... + */ + errorf(HERE, "typename may not have a storage class"); + } + + type_t *result = parse_abstract_declarator(specifiers.type); + + return result; +} @@ -2413,151 +2919,183 @@ struct expression_parser_function_t { expression_parser_function_t expression_parsers[T_LAST_TOKEN]; -static expression_t *make_invalid_expression(void) +/** + * Creates a new invalid expression. + */ +static expression_t *create_invalid_expression(void) { - expression_t *expression = allocate_ast_zero(sizeof(expression[0])); - expression->type = EXPR_INVALID; - expression->source_position = token.source_position; + 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(); - fprintf(stderr, "expected expression, got token "); - print_token(stderr, & token); - fprintf(stderr, "\n"); + errorf(HERE, "expected expression, got token '%K'", &token); next_token(); - return make_invalid_expression(); + return create_invalid_expression(); } +/** + * Parse a string constant. + */ 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 cnst; +} - return (expression_t*) cnst; +/** + * Parse a wide string constant. + */ +static expression_t *parse_wide_string_const(void) +{ + expression_t *const cnst = allocate_expression_zero(EXPR_WIDE_STRING_LITERAL); + cnst->base.datatype = type_wchar_t_ptr; + cnst->wide_string.value = token.v.wide_string; /* TODO concatenate */ + next_token(); + return cnst; } +/** + * Parse an integer constant. + */ static expression_t *parse_int_const(void) { - const_t *cnst = allocate_ast_zero(sizeof(cnst[0])); - - cnst->expression.type = EXPR_CONST; - cnst->expression.datatype = token.datatype; - 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; } +/** + * Parse a float constant. + */ static expression_t *parse_float_const(void) { - const_t *cnst = allocate_ast_zero(sizeof(cnst[0])); - - cnst->expression.type = EXPR_CONST; - cnst->expression.datatype = token.datatype; - 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) { - function_type_t *function_type - = allocate_type_zero(sizeof(function_type[0])); + type_t *ntype = allocate_type_zero(TYPE_FUNCTION); + ntype->function.return_type = type_int; + ntype->function.unspecified_parameters = true; - function_type->type.type = TYPE_FUNCTION; - function_type->result_type = type_int; - function_type->unspecified_parameters = true; - - type_t *type = typehash_insert((type_t*) function_type); - if(type != (type_t*) function_type) { - free_type(function_type); + type_t *type = typehash_insert(ntype); + if(type != ntype) { + free_type(ntype); } - declaration_t *declaration = allocate_ast_zero(sizeof(declaration[0])); - + declaration_t *const declaration = allocate_declaration_zero(); declaration->storage_class = STORAGE_CLASS_EXTERN; declaration->type = type; declaration->symbol = symbol; declaration->source_position = source_position; + declaration->parent_context = 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); - - context_t *last_context = context; - context = global_context; + context_t *old_context = context; + set_context(global_context); environment_push(declaration); + /* prepend the declaration to the global declarations list */ declaration->next = context->declarations; context->declarations = declaration; - context = last_context; + assert(context == global_context); + set_context(old_context); return declaration; } -static type_t *make_function_1_type(type_t *result_type, type_t *argument_type) +/** + * Creates a return_type (func)(argument_type) function type if not + * already exists. + * + * @param return_type the return type + * @param argument_type the argument type + */ +static type_t *make_function_1_type(type_t *return_type, type_t *argument_type) { - function_parameter_t *parameter = allocate_type_zero(sizeof(parameter[0])); + function_parameter_t *parameter + = obstack_alloc(type_obst, sizeof(parameter[0])); + memset(parameter, 0, sizeof(parameter[0])); parameter->type = argument_type; - function_type_t *type = allocate_type_zero(sizeof(type[0])); - type->type.type = TYPE_FUNCTION; - type->result_type = result_type; - type->parameters = parameter; + type_t *type = allocate_type_zero(TYPE_FUNCTION); + type->function.return_type = return_type; + type->function.parameters = parameter; - type_t *result = typehash_insert((type_t*) type); - if(result != (type_t*) type) { + type_t *result = typehash_insert(type); + if(result != type) { free_type(type); } return result; } +/** + * Creates a function type for some function like builtins. + * + * @param symbol the symbol describing the builtin + */ static type_t *get_builtin_symbol_type(symbol_t *symbol) { switch(symbol->ID) { case T___builtin_alloca: return make_function_1_type(type_void_ptr, type_size_t); + case T___builtin_nan: + return make_function_1_type(type_double, type_string); + case T___builtin_nanf: + return make_function_1_type(type_float, type_string); + case T___builtin_nand: + return make_function_1_type(type_long_double, type_string); + case T___builtin_va_end: + return make_function_1_type(type_void, type_valist); default: panic("not implemented builtin symbol found"); } } /** - * performs automatic type cast as described in § 6.3.2.1 + * Performs automatic type cast as described in § 6.3.2.1. + * + * @param orig_type the original type */ -static type_t *automatic_type_conversion(type_t *type) +static type_t *automatic_type_conversion(type_t *orig_type) { - if(type == NULL) + if(orig_type == NULL) return NULL; - if(type->type == TYPE_ARRAY) { - array_type_t *array_type = (array_type_t*) type; + type_t *type = skip_typeref(orig_type); + if(is_type_array(type)) { + array_type_t *array_type = &type->array; type_t *element_type = array_type->element_type; unsigned qualifiers = array_type->type.qualifiers; return make_pointer_type(element_type, qualifiers); } - if(type->type == TYPE_FUNCTION) { - return make_pointer_type(type, TYPE_QUALIFIER_NONE); + if(is_type_function(type)) { + return make_pointer_type(orig_type, TYPE_QUALIFIER_NONE); } - return type; + return orig_type; } /** @@ -2566,43 +3104,37 @@ static type_t *automatic_type_conversion(type_t *type) */ type_t *revert_automatic_type_conversion(const expression_t *expression) { - if(expression->datatype == NULL) + if(expression->base.datatype == NULL) return NULL; - switch(expression->type) { + switch(expression->kind) { case EXPR_REFERENCE: { - const reference_expression_t *ref - = (const reference_expression_t*) expression; + const reference_expression_t *ref = &expression->reference; return ref->declaration->type; } case EXPR_SELECT: { - const select_expression_t *select - = (const select_expression_t*) expression; + const select_expression_t *select = &expression->select; return select->compound_entry->type; } - case EXPR_UNARY: { - const unary_expression_t *unary - = (const unary_expression_t*) expression; - if(unary->type == UNEXPR_DEREFERENCE) { - expression_t *value = unary->value; - type_t *type = skip_typeref(value->datatype); - pointer_type_t *pointer_type = (pointer_type_t*) type; + case EXPR_UNARY_DEREFERENCE: { + expression_t *value = expression->unary.value; + type_t *type = skip_typeref(value->base.datatype); + pointer_type_t *pointer_type = &type->pointer; - return pointer_type->points_to; - } - break; + return pointer_type->points_to; } case EXPR_BUILTIN_SYMBOL: { const builtin_symbol_expression_t *builtin - = (const builtin_symbol_expression_t*) expression; + = &expression->builtin_symbol; return get_builtin_symbol_type(builtin->symbol); } case EXPR_ARRAY_ACCESS: { const array_access_expression_t *array_access - = (const array_access_expression_t*) expression; - type_t *type_left = skip_typeref(array_access->array_ref->datatype); + = &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 = (pointer_type_t*) type_left; + pointer_type_t *pointer_type = &type_left->pointer; return pointer_type->points_to; } @@ -2610,15 +3142,15 @@ type_t *revert_automatic_type_conversion(const expression_t *expression) break; } - return expression->datatype; + 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 = get_declaration(ref->symbol, NAMESPACE_NORMAL); @@ -2626,25 +3158,21 @@ static expression_t *parse_reference(void) next_token(); if(declaration == NULL) { -#ifndef STRICT_C99 - /* an implicitly defined function */ - if(token.type == '(') { - parser_print_prefix_pos(token.source_position); - fprintf(stderr, "warning: implicit declaration of function '%s'\n", - ref->symbol->string); + if (! strict_mode && token.type == '(') { + /* an implicitly defined function */ + warningf(HERE, "implicit declaration of function '%Y'", + ref->symbol); declaration = create_implicit_function(ref->symbol, source_position); - } else -#endif - { - parser_print_error_prefix(); - fprintf(stderr, "unknown symbol '%s' found.\n", ref->symbol->string); - return (expression_t*) ref; + } else { + errorf(HERE, "unknown symbol '%Y' found.", ref->symbol); + return expression; } } - type_t *type = declaration->type; + 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); @@ -2652,7 +3180,7 @@ static expression_t *parse_reference(void) ref->declaration = declaration; ref->expression.datatype = type; - return (expression_t*) ref; + return expression; } static void check_cast_allowed(expression_t *expression, type_t *dest_type) @@ -2664,11 +3192,9 @@ static void check_cast_allowed(expression_t *expression, type_t *dest_type) 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); - cast->expression.type = EXPR_UNARY; - cast->type = UNEXPR_CAST; - cast->expression.source_position = token.source_position; + cast->base.source_position = token.source_position; type_t *type = parse_typename(); @@ -2677,48 +3203,45 @@ 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_t *expression = allocate_expression_zero(EXPR_STATEMENT); - statement_t *statement = parse_compound_statement(); - expression->statement = 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 - = (compound_statement_t*) statement; + assert(statement->kind == 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 *iter = compound_statement->statements; - for( ; iter != NULL; iter = iter->next) { + 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; + if(last_statement->kind == STATEMENT_EXPRESSION) { + 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) @@ -2727,7 +3250,7 @@ static expression_t *parse_brace_expression(void) switch(token.type) { case '{': - /* gcc extension: a stement expression */ + /* gcc extension: a statement expression */ return parse_statement_expression(); TYPE_QUALIFIERS @@ -2751,13 +3274,14 @@ static expression_t *parse_function_keyword(void) /* TODO */ if (current_function == NULL) { - parse_error("'__func__' used outside of a function"); + errorf(HERE, "'__func__' used outside of a function"); } - string_literal_t *expression = allocate_ast_zero(sizeof(expression[0])); - expression->expression.type = EXPR_FUNCTION; + string_literal_expression_t *expression + = allocate_ast_zero(sizeof(expression[0])); + + expression->expression.kind = EXPR_FUNCTION; expression->expression.datatype = type_string; - expression->value = "TODO: FUNCTION"; return (expression_t*) expression; } @@ -2767,10 +3291,15 @@ static expression_t *parse_pretty_function_keyword(void) eat(T___PRETTY_FUNCTION__); /* TODO */ - string_literal_t *expression = allocate_ast_zero(sizeof(expression[0])); - expression->expression.type = EXPR_PRETTY_FUNCTION; + if (current_function == NULL) { + errorf(HERE, "'__PRETTY_FUNCTION__' used outside of a function"); + } + + string_literal_expression_t *expression + = allocate_ast_zero(sizeof(expression[0])); + + expression->expression.kind = EXPR_PRETTY_FUNCTION; expression->expression.datatype = type_string; - expression->value = "TODO: PRETTY FUNCTION"; return (expression_t*) expression; } @@ -2782,7 +3311,7 @@ static designator_t *parse_designator(void) if(token.type != T_IDENTIFIER) { parse_error_expected("while parsing member designator", T_IDENTIFIER, 0); - eat_brace(); + eat_paren(); return NULL; } result->symbol = token.v.symbol; @@ -2795,7 +3324,7 @@ static designator_t *parse_designator(void) if(token.type != T_IDENTIFIER) { parse_error_expected("while parsing member designator", T_IDENTIFIER, 0); - eat_brace(); + eat_paren(); return NULL; } designator_t *designator = allocate_ast_zero(sizeof(result[0])); @@ -2811,7 +3340,7 @@ static designator_t *parse_designator(void) designator_t *designator = allocate_ast_zero(sizeof(result[0])); designator->array_access = parse_expression(); if(designator->array_access == NULL) { - eat_brace(); + eat_paren(); return NULL; } expect(']'); @@ -2830,50 +3359,205 @@ 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_start(void) +{ + eat(T___builtin_va_start); + + expression_t *expression = allocate_expression_zero(EXPR_VA_START); + + expect('('); + expression->va_starte.ap = parse_assignment_expression(); + expect(','); + expression_t *const expr = parse_assignment_expression(); + if (expr->kind == EXPR_REFERENCE) { + declaration_t *const decl = expr->reference.declaration; + if (decl->parent_context == ¤t_function->context && + decl->next == NULL) { + expression->va_starte.parameter = decl; + expect(')'); + return expression; + } + } + errorf(expr->base.source_position, "second argument of 'va_start' must be last parameter of the current function"); + + return create_invalid_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.ap = 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; + expression_t *expression = allocate_expression_zero(EXPR_BUILTIN_SYMBOL); + + symbol_t *symbol = token.v.symbol; - expression->symbol = token.v.symbol; + expression->builtin_symbol.symbol = symbol; next_token(); - type_t *type = get_builtin_symbol_type(expression->symbol); + type_t *type = get_builtin_symbol_type(symbol); type = automatic_type_conversion(type); - expression->expression.datatype = type; - return (expression_t*) expression; + expression->base.datatype = type; + return expression; +} + +static expression_t *parse_builtin_constant(void) +{ + eat(T___builtin_constant_p); + + expression_t *expression = allocate_expression_zero(EXPR_BUILTIN_CONSTANT_P); + + expect('('); + expression->builtin_constant.value = parse_assignment_expression(); + expect(')'); + expression->base.datatype = type_int; + + return expression; +} + +static expression_t *parse_builtin_prefetch(void) +{ + eat(T___builtin_prefetch); + + expression_t *expression = allocate_expression_zero(EXPR_BUILTIN_PREFETCH); + + expect('('); + expression->builtin_prefetch.adr = parse_assignment_expression(); + if (token.type == ',') { + next_token(); + expression->builtin_prefetch.rw = parse_assignment_expression(); + } + if (token.type == ',') { + next_token(); + expression->builtin_prefetch.locality = parse_assignment_expression(); + } + expect(')'); + expression->base.datatype = type_void; + + return expression; +} + +static expression_t *parse_compare_builtin(void) +{ + expression_t *expression; + + switch(token.type) { + case T___builtin_isgreater: + expression = allocate_expression_zero(EXPR_BINARY_ISGREATER); + break; + case T___builtin_isgreaterequal: + expression = allocate_expression_zero(EXPR_BINARY_ISGREATEREQUAL); + break; + case T___builtin_isless: + expression = allocate_expression_zero(EXPR_BINARY_ISLESS); + break; + case T___builtin_islessequal: + expression = allocate_expression_zero(EXPR_BINARY_ISLESSEQUAL); + break; + case T___builtin_islessgreater: + expression = allocate_expression_zero(EXPR_BINARY_ISLESSGREATER); + break; + case T___builtin_isunordered: + expression = allocate_expression_zero(EXPR_BINARY_ISUNORDERED); + break; + default: + panic("invalid compare builtin found"); + break; + } + next_token(); + + expect('('); + expression->binary.left = parse_assignment_expression(); + expect(','); + expression->binary.right = parse_assignment_expression(); + expect(')'); + + type_t *orig_type_left = expression->binary.left->base.datatype; + type_t *orig_type_right = expression->binary.right->base.datatype; + if(orig_type_left == NULL || orig_type_right == NULL) + return expression; + + type_t *type_left = skip_typeref(orig_type_left); + type_t *type_right = skip_typeref(orig_type_right); + if(!is_type_floating(type_left) && !is_type_floating(type_right)) { + type_error_incompatible("invalid operands in comparison", + token.source_position, type_left, type_right); + } else { + semantic_comparison(&expression->binary); + } + + return expression; +} + +static expression_t *parse_builtin_expect(void) +{ + eat(T___builtin_expect); + + expression_t *expression + = allocate_expression_zero(EXPR_BINARY_BUILTIN_EXPECT); + + expect('('); + expression->binary.left = parse_assignment_expression(); + expect(','); + expression->binary.right = parse_constant_expression(); + expect(')'); + + expression->base.datatype = expression->binary.left->base.datatype; + + return expression; +} + +static expression_t *parse_assume(void) { + eat(T_assume); + + expression_t *expression + = allocate_expression_zero(EXPR_UNARY_ASSUME); + + expect('('); + expression->unary.value = parse_assignment_expression(); + expect(')'); + + expression->base.datatype = type_void; + return expression; +} + +static expression_t *parse_alignof(void) { + eat(T___alignof__); + + expression_t *expression + = allocate_expression_zero(EXPR_ALIGNOF); + + expect('('); + expression->alignofe.type = parse_typename(); + expect(')'); + + expression->base.datatype = type_size_t; + return expression; } static expression_t *parse_primary_expression(void) @@ -2885,6 +3569,8 @@ static expression_t *parse_primary_expression(void) return parse_float_const(); case T_STRING_LITERAL: return parse_string_const(); + case T_WIDE_STRING_LITERAL: + return parse_wide_string_const(); case T_IDENTIFIER: return parse_reference(); case T___FUNCTION__: @@ -2894,25 +3580,55 @@ static expression_t *parse_primary_expression(void) return parse_pretty_function_keyword(); case T___builtin_offsetof: return parse_offsetof(); + case T___builtin_va_start: + return parse_va_start(); case T___builtin_va_arg: return parse_va_arg(); - case T___builtin_alloca: case T___builtin_expect: - case T___builtin_va_start: + return parse_builtin_expect(); + case T___builtin_nanf: + case T___builtin_alloca: case T___builtin_va_end: return parse_builtin_symbol(); + case T___builtin_isgreater: + case T___builtin_isgreaterequal: + case T___builtin_isless: + case T___builtin_islessequal: + case T___builtin_islessgreater: + case T___builtin_isunordered: + return parse_compare_builtin(); + case T___builtin_constant_p: + return parse_builtin_constant(); + case T___builtin_prefetch: + return parse_builtin_prefetch(); + case T___alignof__: + return parse_alignof(); + case T_assume: + return parse_assume(); case '(': return parse_brace_expression(); } - parser_print_error_prefix(); - fprintf(stderr, "unexpected token "); - print_token(stderr, &token); - fprintf(stderr, "\n"); + errorf(HERE, "unexpected token '%K'", &token); eat_statement(); - return make_invalid_expression(); + return create_invalid_expression(); +} + +/** + * Check if the expression has the character type and issue a warning then. + */ +static void check_for_char_index_type(const expression_t *expression) { + type_t *type = expression->base.datatype; + type_t *base_type = skip_typeref(type); + + if (base_type->base.kind == TYPE_ATOMIC) { + if (base_type->atomic.akind == ATOMIC_TYPE_CHAR) { + warningf(expression->base.source_position, + "array subscript has type '%T'", type); + } + } } static expression_t *parse_array_expression(unsigned precedence, @@ -2927,31 +3643,31 @@ static expression_t *parse_array_expression(unsigned precedence, array_access_expression_t *array_access = allocate_ast_zero(sizeof(array_access[0])); - array_access->expression.type = EXPR_ARRAY_ACCESS; + array_access->expression.kind = EXPR_ARRAY_ACCESS; - type_t *type_left = skip_typeref(left->datatype); - type_t *type_inside = skip_typeref(inside->datatype); - type_t *result_type; + type_t *type_left = left->base.datatype; + type_t *type_inside = inside->base.datatype; + type_t *return_type = NULL; if(type_left != NULL && type_inside != NULL) { + type_left = skip_typeref(type_left); + type_inside = skip_typeref(type_inside); + if(is_type_pointer(type_left)) { - pointer_type_t *pointer = (pointer_type_t*) type_left; - result_type = pointer->points_to; + pointer_type_t *pointer = &type_left->pointer; + return_type = pointer->points_to; array_access->array_ref = left; array_access->index = inside; + check_for_char_index_type(inside); } else if(is_type_pointer(type_inside)) { - pointer_type_t *pointer = (pointer_type_t*) type_inside; - result_type = pointer->points_to; + pointer_type_t *pointer = &type_inside->pointer; + return_type = pointer->points_to; array_access->array_ref = inside; array_access->index = left; array_access->flipped = true; + check_for_char_index_type(left); } 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"); + errorf(HERE, "array access on object with non-pointer types '%T', '%T'", type_left, type_inside); } } else { array_access->array_ref = left; @@ -2964,38 +3680,19 @@ static expression_t *parse_array_expression(unsigned precedence, } next_token(); - result_type = automatic_type_conversion(result_type); - array_access->expression.datatype = result_type; + return_type = automatic_type_conversion(return_type); + array_access->expression.datatype = return_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.kind = EXPR_SIZEOF; sizeof_expression->expression.datatype = type_size_t; if(token.type == '(' && is_declaration_specifier(look_ahead(1), true)) { @@ -3003,10 +3700,10 @@ static expression_t *parse_sizeof(unsigned precedence) sizeof_expression->type = parse_typename(); expect(')'); } else { - expression_t *expression = parse_sub_expression(precedence); - expression->datatype = revert_automatic_type_conversion(expression); + expression_t *expression = parse_sub_expression(precedence); + expression->base.datatype = revert_automatic_type_conversion(expression); - sizeof_expression->type = expression->datatype; + sizeof_expression->type = expression->base.datatype; sizeof_expression->size_expression = expression; } @@ -3022,59 +3719,48 @@ static expression_t *parse_select_expression(unsigned precedence, bool is_pointer = (token.type == T_MINUSGREATER); next_token(); - select_expression_t *select = allocate_ast_zero(sizeof(select[0])); - - select->expression.type = EXPR_SELECT; - select->compound = compound; + 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 (expression_t*) select; + return select; } - symbol_t *symbol = token.v.symbol; - select->symbol = symbol; + symbol_t *symbol = token.v.symbol; + select->select.symbol = symbol; next_token(); - type_t *orig_type = compound->datatype; + type_t *orig_type = compound->base.datatype; if(orig_type == NULL) - return make_invalid_expression(); + return create_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(); + if(type->kind != TYPE_POINTER) { + errorf(HERE, "left hand side of '->' is not a pointer, but '%T'", orig_type); + return create_invalid_expression(); } - pointer_type_t *pointer_type = (pointer_type_t*) type; + 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(); + if(type_left->kind != TYPE_COMPOUND_STRUCT + && type_left->kind != TYPE_COMPOUND_UNION) { + errorf(HERE, "request for member '%Y' in something not a struct or " + "union, but '%T'", symbol, type_left); + return create_invalid_expression(); } - compound_type_t *compound_type = (compound_type_t*) type_left; + 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(); + errorf(HERE, "request for member '%Y' of incomplete type '%T'", + symbol, type_left); + return create_invalid_expression(); } declaration_t *iter = declaration->context.declarations; @@ -3084,52 +3770,60 @@ static expression_t *parse_select_expression(unsigned precedence, } } 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(); + errorf(HERE, "'%T' has no member named '%Y'", orig_type, symbol); + return create_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->compound_entry = iter; - select->expression.datatype = expression_type; - return (expression_t*) select; + select->select.compound_entry = iter; + select->base.datatype = expression_type; + + if(expression_type->kind == TYPE_BITFIELD) { + expression_t *extract + = allocate_expression_zero(EXPR_UNARY_BITFIELD_EXTRACT); + extract->unary.value = select; + extract->base.datatype = expression_type->bitfield.base; + + return extract; + } + + return select; } +/** + * Parse a call expression, ie. expression '( ... )'. + * + * @param expression the function address + */ static expression_t *parse_call_expression(unsigned precedence, expression_t *expression) { (void) precedence; - call_expression_t *call = allocate_ast_zero(sizeof(call[0])); - call->expression.type = EXPR_CALL; + expression_t *result = allocate_expression_zero(EXPR_CALL); + + call_expression_t *call = &result->call; call->function = expression; - function_type_t *function_type; - type_t *orig_type = expression->datatype; + function_type_t *function_type = NULL; + type_t *orig_type = expression->base.datatype; if(orig_type != NULL) { - function_type = NULL; type_t *type = skip_typeref(orig_type); if(is_type_pointer(type)) { - pointer_type_t *pointer_type = (pointer_type_t*) type; + pointer_type_t *pointer_type = &type->pointer; type = skip_typeref(pointer_type->points_to); - if (type->type == TYPE_FUNCTION) { - function_type = (function_type_t*) type; - call->expression.datatype = function_type->result_type; + if (is_type_function(type)) { + function_type = &type->function; + call->expression.datatype = function_type->return_type; } } if(function_type == NULL) { - parser_print_error_prefix(); - fputs("called object '", stderr); - print_expression(expression); - fputs("' (type ", stderr); - print_type_quoted(orig_type); - fputs(") is not a pointer to a function\n", stderr); + errorf(HERE, "called object '%E' (type '%T') is not a pointer to a function", expression, orig_type); function_type = NULL; call->expression.datatype = NULL; @@ -3172,146 +3866,165 @@ static expression_t *parse_call_expression(unsigned precedence, } /* too few parameters */ if(parameter != NULL) { - parser_print_error_prefix(); - fprintf(stderr, "too few arguments to function '"); - print_expression(expression); - fprintf(stderr, "'\n"); + errorf(HERE, "too few arguments to function '%E'", expression); } 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"); + errorf(HERE, "too many arguments to function '%E'", expression); } else { /* do default promotion */ for( ; argument != NULL; argument = argument->next) { - type_t *type = argument->expression->datatype; + type_t *type = argument->expression->base.datatype; if(type == NULL) continue; + type = skip_typeref(type); if(is_type_integer(type)) { type = promote_integer(type); } else if(type == type_float) { type = type_double; } + argument->expression = create_implicit_cast(argument->expression, type); } + + check_format(&result->call); } + } else { + check_format(&result->call); } } - return (expression_t*) call; + return result; } static type_t *semantic_arithmetic(type_t *type_left, type_t *type_right); +static bool same_compound_type(const type_t *type1, const type_t *type2) +{ + if(!is_type_compound(type1)) + return false; + if(type1->kind != type2->kind) + return false; + + const compound_type_t *compound1 = &type1->compound; + const compound_type_t *compound2 = &type2->compound; + + return compound1->declaration == compound2->declaration; +} + +/** + * Parse a conditional expression, ie. 'expression ? ... : ...'. + * + * @param expression the conditional expression + */ 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; + expression_t *result = allocate_expression_zero(EXPR_CONDITIONAL); + + conditional_expression_t *conditional = &result->conditional; + conditional->condition = expression; /* 6.5.15.2 */ - type_t *condition_type_orig = conditional->condition->datatype; + type_t *condition_type_orig = expression->base.datatype; if(condition_type_orig != NULL) { - type_t *condition_type = skip_typeref(condition_type_orig); + type_t *condition_type = skip_typeref(condition_type_orig); if(condition_type != NULL && !is_type_scalar(condition_type)) { - type_error("expected a scalar type", expression->source_position, - condition_type_orig); + type_error("expected a scalar type in conditional condition", + expression->base.source_position, condition_type_orig); } } - expression_t *const t_expr = parse_expression(); - conditional->true_expression = t_expr; + expression_t *true_expression = parse_expression(); expect(':'); - expression_t *const f_expr = parse_sub_expression(precedence); - conditional->false_expression = f_expr; + expression_t *false_expression = parse_sub_expression(precedence); - type_t *const true_type = t_expr->datatype; - if(true_type == NULL) - return (expression_t*) conditional; - type_t *const false_type = f_expr->datatype; - if(false_type == NULL) - return (expression_t*) conditional; + conditional->true_expression = true_expression; + conditional->false_expression = false_expression; + + type_t *orig_true_type = true_expression->base.datatype; + type_t *orig_false_type = false_expression->base.datatype; + if(orig_true_type == NULL || orig_false_type == NULL) + return result; - type_t *const skipped_true_type = skip_typeref(true_type); - type_t *const skipped_false_type = skip_typeref(false_type); + type_t *true_type = skip_typeref(orig_true_type); + type_t *false_type = skip_typeref(orig_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 */ + type_t *result_type = NULL; + if (is_type_arithmetic(true_type) && is_type_arithmetic(false_type)) { + result_type = semantic_arithmetic(true_type, false_type); + + true_expression = create_implicit_cast(true_expression, result_type); + false_expression = create_implicit_cast(false_expression, result_type); + + conditional->true_expression = true_expression; + conditional->false_expression = false_expression; + conditional->expression.datatype = result_type; + } else if (same_compound_type(true_type, false_type) + || (is_type_atomic(true_type, ATOMIC_TYPE_VOID) && + is_type_atomic(false_type, ATOMIC_TYPE_VOID))) { + /* just take 1 of the 2 types */ + result_type = true_type; + } else if (is_type_pointer(true_type) && is_type_pointer(false_type) + && pointers_compatible(true_type, false_type)) { + /* ok */ + result_type = true_type; } else { + /* TODO */ type_error_incompatible("while parsing conditional", - expression->source_position, true_type, - skipped_false_type); + expression->base.source_position, true_type, + false_type); } - return (expression_t*) conditional; + conditional->expression.datatype = result_type; + return result; } +/** + * Parse an extension expression. + */ static expression_t *parse_extension(unsigned precedence) { eat(T___extension__); /* TODO enable extensions */ - - return parse_sub_expression(precedence); + expression_t *expression = parse_sub_expression(precedence); + /* TODO disable extensions */ + return expression; } 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; + expression_t *result = allocate_expression_zero(EXPR_CLASSIFY_TYPE); + result->base.datatype = type_int; expect('('); - expression_t *const expression = parse_sub_expression(precedence); + expression_t *expression = parse_sub_expression(precedence); expect(')'); - classify_type_expr->type_expression = expression; + result->classify_type.type_expression = expression; - return (expression_t*)classify_type_expr; + return result; } static void semantic_incdec(unary_expression_t *expression) { - type_t *orig_type = expression->value->datatype; + 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) { + if(!is_type_arithmetic(type) && type->kind != TYPE_POINTER) { /* TODO: improve error message */ - parser_print_error_prefix(); - fprintf(stderr, "operation needs an arithmetic or pointer type\n"); + errorf(HERE, "operation needs an arithmetic or pointer type"); return; } @@ -3320,15 +4033,14 @@ static void semantic_incdec(unary_expression_t *expression) static void semantic_unexpr_arithmetic(unary_expression_t *expression) { - type_t *orig_type = expression->value->datatype; + 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"); + errorf(HERE, "operation needs an arithmetic type"); return; } @@ -3337,13 +4049,13 @@ static void semantic_unexpr_arithmetic(unary_expression_t *expression) static void semantic_unexpr_scalar(unary_expression_t *expression) { - type_t *orig_type = expression->value->datatype; + 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"); + errorf(HERE, "operand of ! must be of scalar type"); return; } @@ -3352,13 +4064,13 @@ static void semantic_unexpr_scalar(unary_expression_t *expression) static void semantic_unexpr_integer(unary_expression_t *expression) { - type_t *orig_type = expression->value->datatype; + 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"); + errorf(HERE, "operand of ~ must be of integer type"); return; } @@ -3367,39 +4079,44 @@ static void semantic_unexpr_integer(unary_expression_t *expression) static void semantic_dereference(unary_expression_t *expression) { - type_t *orig_type = expression->value->datatype; + 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); + errorf(HERE, "Unary '*' needs pointer or arrray type, but type '%T' given", orig_type); return; } - pointer_type_t *pointer_type = (pointer_type_t*)type; + 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; } +/** + * Check the semantic of the address taken expression. + */ static void semantic_take_addr(unary_expression_t *expression) { - expression_t *value = expression->value; - value->datatype = revert_automatic_type_conversion(value); + expression_t *value = expression->value; + value->base.datatype = revert_automatic_type_conversion(value); - type_t *orig_type = value->datatype; + type_t *orig_type = value->base.datatype; if(orig_type == NULL) return; - if(value->type == EXPR_REFERENCE) { + if(value->kind == EXPR_REFERENCE) { reference_expression_t *reference = (reference_expression_t*) value; declaration_t *declaration = reference->declaration; if(declaration != NULL) { + if (declaration->storage_class == STORAGE_CLASS_REGISTER) { + errorf(expression->expression.source_position, + "address of register variable '%Y' requested", + declaration->symbol); + } declaration->address_taken = 1; } } @@ -3411,28 +4128,32 @@ static void semantic_take_addr(unary_expression_t *expression) 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, + expression_t *unary_expression \ + = allocate_expression_zero(unexpression_type); \ + unary_expression->base.source_position = HERE; \ + unary_expression->unary.value = parse_sub_expression(precedence); \ + \ + sfunc(&unary_expression->unary); \ + \ + return unary_expression; \ +} + +CREATE_UNARY_EXPRESSION_PARSER('-', EXPR_UNARY_NEGATE, + semantic_unexpr_arithmetic) +CREATE_UNARY_EXPRESSION_PARSER('+', EXPR_UNARY_PLUS, + semantic_unexpr_arithmetic) +CREATE_UNARY_EXPRESSION_PARSER('!', EXPR_UNARY_NOT, + semantic_unexpr_scalar) +CREATE_UNARY_EXPRESSION_PARSER('*', EXPR_UNARY_DEREFERENCE, + semantic_dereference) +CREATE_UNARY_EXPRESSION_PARSER('&', EXPR_UNARY_TAKE_ADDRESS, + semantic_take_addr) +CREATE_UNARY_EXPRESSION_PARSER('~', EXPR_UNARY_BITWISE_NEGATE, semantic_unexpr_integer) -CREATE_UNARY_EXPRESSION_PARSER(T_PLUSPLUS, UNEXPR_PREFIX_INCREMENT, +CREATE_UNARY_EXPRESSION_PARSER(T_PLUSPLUS, EXPR_UNARY_PREFIX_INCREMENT, semantic_incdec) -CREATE_UNARY_EXPRESSION_PARSER(T_MINUSMINUS, UNEXPR_PREFIX_DECREMENT, +CREATE_UNARY_EXPRESSION_PARSER(T_MINUSMINUS, EXPR_UNARY_PREFIX_DECREMENT, semantic_incdec) #define CREATE_UNARY_POSTFIX_EXPRESSION_PARSER(token_type, unexpression_type, \ @@ -3443,20 +4164,20 @@ static expression_t *parse_##unexpression_type(unsigned precedence, \ (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; \ + expression_t *unary_expression \ + = allocate_expression_zero(unexpression_type); \ + unary_expression->unary.value = left; \ \ - sfunc(unary_expression); \ + sfunc(&unary_expression->unary); \ \ - return (expression_t*) unary_expression; \ + return unary_expression; \ } -CREATE_UNARY_POSTFIX_EXPRESSION_PARSER(T_PLUSPLUS, UNEXPR_POSTFIX_INCREMENT, +CREATE_UNARY_POSTFIX_EXPRESSION_PARSER(T_PLUSPLUS, + EXPR_UNARY_POSTFIX_INCREMENT, semantic_incdec) -CREATE_UNARY_POSTFIX_EXPRESSION_PARSER(T_MINUSMINUS, UNEXPR_POSTFIX_DECREMENT, +CREATE_UNARY_POSTFIX_EXPRESSION_PARSER(T_MINUSMINUS, + EXPR_UNARY_POSTFIX_DECREMENT, semantic_incdec) static type_t *semantic_arithmetic(type_t *type_left, type_t *type_right) @@ -3480,7 +4201,9 @@ static type_t *semantic_arithmetic(type_t *type_left, type_t *type_right) bool signed_left = is_type_signed(type_left); bool signed_right = is_type_signed(type_right); - if(get_rank(type_left) < get_rank(type_right)) { + 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 { @@ -3495,12 +4218,15 @@ static type_t *semantic_arithmetic(type_t *type_left, type_t *type_right) } } +/** + * Check the semantic restrictions for a binary expression. + */ 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->datatype; - type_t *orig_type_right = right->datatype; + 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; @@ -3510,8 +4236,7 @@ static void semantic_binexpr_arithmetic(binary_expression_t *expression) 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"); + errorf(HERE, "operation needs arithmetic types"); return; } @@ -3525,8 +4250,8 @@ static void semantic_shift_op(binary_expression_t *expression) { expression_t *left = expression->left; expression_t *right = expression->right; - type_t *orig_type_left = left->datatype; - type_t *orig_type_right = right->datatype; + type_t *orig_type_left = left->base.datatype; + type_t *orig_type_right = right->base.datatype; if(orig_type_left == NULL || orig_type_right == NULL) return; @@ -3536,8 +4261,7 @@ static void semantic_shift_op(binary_expression_t *expression) 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"); + errorf(HERE, "operation needs integer types"); return; } @@ -3553,8 +4277,8 @@ static void semantic_add(binary_expression_t *expression) { expression_t *left = expression->left; expression_t *right = expression->right; - type_t *orig_type_left = left->datatype; - type_t *orig_type_right = right->datatype; + type_t *orig_type_left = left->base.datatype; + type_t *orig_type_right = right->base.datatype; if(orig_type_left == NULL || orig_type_right == NULL) return; @@ -3574,12 +4298,7 @@ static void semantic_add(binary_expression_t *expression) } 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"); + errorf(HERE, "invalid operands to binary + ('%T', '%T')", orig_type_left, orig_type_right); } } @@ -3587,8 +4306,8 @@ static void semantic_sub(binary_expression_t *expression) { expression_t *left = expression->left; expression_t *right = expression->right; - type_t *orig_type_left = left->datatype; - type_t *orig_type_right = right->datatype; + 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; @@ -3603,27 +4322,16 @@ static void semantic_sub(binary_expression_t *expression) 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)) { + } else if(is_type_pointer(type_left) && is_type_integer(type_right)) { expression->expression.datatype = type_left; - } else if(type_left->type == TYPE_POINTER && - type_right->type == TYPE_POINTER) { + } else if(is_type_pointer(type_left) && is_type_pointer(type_right)) { if(!pointers_compatible(type_left, type_right)) { - parser_print_error_prefix(); - fprintf(stderr, "pointers to incompatible objects to binary - ("); - print_type_quoted(orig_type_left); - fprintf(stderr, ", "); - print_type_quoted(orig_type_right); - fprintf(stderr, ")\n"); + errorf(HERE, "pointers to incompatible objects to binary - ('%T', '%T')", orig_type_left, orig_type_right); } else { expression->expression.datatype = type_ptrdiff_t; } } else { - parser_print_error_prefix(); - fprintf(stderr, "invalid operands to binary - ("); - print_type_quoted(orig_type_left); - fprintf(stderr, ", "); - print_type_quoted(orig_type_right); - fprintf(stderr, ")\n"); + errorf(HERE, "invalid operands to binary - ('%T', '%T')", orig_type_left, orig_type_right); } } @@ -3631,8 +4339,8 @@ static void semantic_comparison(binary_expression_t *expression) { expression_t *left = expression->left; expression_t *right = expression->right; - type_t *orig_type_left = left->datatype; - type_t *orig_type_right = right->datatype; + type_t *orig_type_left = left->base.datatype; + type_t *orig_type_right = right->base.datatype; if(orig_type_left == NULL || orig_type_right == NULL) return; @@ -3646,12 +4354,11 @@ static void semantic_comparison(binary_expression_t *expression) 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) { + } else if (is_type_pointer(type_left) && is_type_pointer(type_right)) { /* TODO check compatibility */ - } else if (type_left->type == TYPE_POINTER) { + } else if (is_type_pointer(type_left)) { expression->right = create_implicit_cast(right, type_left); - } else if (type_right->type == TYPE_POINTER) { + } else if (is_type_pointer(type_right)) { expression->left = create_implicit_cast(left, type_right); } else { type_error_incompatible("invalid operands in comparison", @@ -3664,8 +4371,8 @@ 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->datatype; - type_t *orig_type_right = right->datatype; + 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; @@ -3675,15 +4382,14 @@ static void semantic_arithmetic_assign(binary_expression_t *expression) 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"); + errorf(HERE, "operation needs arithmetic types"); return; } /* combined instructions are tricky. We can't create an implicit cast on * the left side, because we need the uncasted form for the store. * The ast2firm pass has to know that left_type must be right_type - * for the arithmeitc operation and create a cast by itself */ + * for the arithmetic 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; @@ -3693,8 +4399,8 @@ static void semantic_arithmetic_addsubb_assign(binary_expression_t *expression) { expression_t *left = expression->left; expression_t *right = expression->right; - type_t *orig_type_left = left->datatype; - type_t *orig_type_right = right->datatype; + type_t *orig_type_left = left->base.datatype; + type_t *orig_type_right = right->base.datatype; if(orig_type_left == NULL || orig_type_right == NULL) return; @@ -3706,29 +4412,27 @@ static void semantic_arithmetic_addsubb_assign(binary_expression_t *expression) /* 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 */ + * for the arithmetic 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)) { + } else if (is_type_pointer(type_left) && is_type_integer(type_right)) { expression->expression.datatype = type_left; } else { - parser_print_error_prefix(); - fputs("Incompatible types ", stderr); - print_type_quoted(orig_type_left); - fputs(" and ", stderr); - print_type_quoted(orig_type_right); - fputs(" in assignment\n", stderr); + errorf(HERE, "incompatible types '%T' and '%T' in assignment", orig_type_left, orig_type_right); return; } } +/** + * Check the semantic restrictions of a logical expression. + */ static void semantic_logical_op(binary_expression_t *expression) { expression_t *left = expression->left; expression_t *right = expression->right; - type_t *orig_type_left = left->datatype; - type_t *orig_type_right = right->datatype; + type_t *orig_type_left = left->base.datatype; + type_t *orig_type_right = right->base.datatype; if(orig_type_left == NULL || orig_type_right == NULL) return; @@ -3738,65 +4442,66 @@ static void semantic_logical_op(binary_expression_t *expression) 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"); + errorf(HERE, "operation needs scalar types"); return; } expression->expression.datatype = type_int; } -static bool has_const_fields(type_t *type) +/** + * Checks if a compound type has constant fields. + */ +static bool has_const_fields(const compound_type_t *type) { - (void) type; + const context_t *context = &type->declaration->context; + const declaration_t *declaration = context->declarations; + + for (; declaration != NULL; declaration = declaration->next) { + if (declaration->namespc != NAMESPACE_NORMAL) + continue; + + const type_t *decl_type = skip_typeref(declaration->type); + if (decl_type->base.qualifiers & TYPE_QUALIFIER_CONST) + return true; + } /* TODO */ return false; } +/** + * Check the semantic restrictions of a binary assign expression. + */ static void semantic_binexpr_assign(binary_expression_t *expression) { expression_t *left = expression->left; - type_t *orig_type_left = left->datatype; + type_t *orig_type_left = left->base.datatype; if(orig_type_left == NULL) return; type_t *type_left = revert_automatic_type_conversion(left); - type_left = skip_typeref(orig_type_left); + type_left = skip_typeref(orig_type_left); /* 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"); + if (is_type_array(type_left)) { + errorf(HERE, "cannot assign to arrays ('%E')", left); return; } - if(type_left->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"); + if(type_left->base.qualifiers & TYPE_QUALIFIER_CONST) { + errorf(HERE, "assignment to readonly location '%E' (type '%T')", left, + orig_type_left); 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"); + errorf(HERE, + "left-hand side of assignment '%E' has incomplete type '%T'", + left, orig_type_left); 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"); + if(is_type_compound(type_left) && has_const_fields(&type_left->compound)) { + errorf(HERE, "cannot assign to '%E' because compound type '%T' has readonly fields", + left, orig_type_left); return; } @@ -3807,71 +4512,77 @@ static void semantic_binexpr_assign(binary_expression_t *expression) static void semantic_comma(binary_expression_t *expression) { - expression->expression.datatype = expression->right->datatype; -} - -#define CREATE_BINEXPR_PARSER(token_type, binexpression_type, sfunc, lr) \ -static expression_t *parse_##binexpression_type(unsigned precedence, \ - expression_t *left) \ -{ \ - eat(token_type); \ - \ - expression_t *right = parse_sub_expression(precedence + lr); \ - \ - binary_expression_t *binexpr \ - = allocate_ast_zero(sizeof(binexpr[0])); \ - binexpr->expression.type = EXPR_BINARY; \ - binexpr->type = binexpression_type; \ - binexpr->left = left; \ - binexpr->right = right; \ - sfunc(binexpr); \ - \ - return (expression_t*) binexpr; \ -} - -CREATE_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, + expression->expression.datatype = expression->right->base.datatype; +} + +#define CREATE_BINEXPR_PARSER(token_type, binexpression_type, sfunc, lr) \ +static expression_t *parse_##binexpression_type(unsigned precedence, \ + expression_t *left) \ +{ \ + eat(token_type); \ + \ + expression_t *right = parse_sub_expression(precedence + lr); \ + \ + expression_t *binexpr = allocate_expression_zero(binexpression_type); \ + binexpr->binary.left = left; \ + binexpr->binary.right = right; \ + sfunc(&binexpr->binary); \ + \ + return binexpr; \ +} + +CREATE_BINEXPR_PARSER(',', EXPR_BINARY_COMMA, semantic_comma, 1) +CREATE_BINEXPR_PARSER('*', EXPR_BINARY_MUL, semantic_binexpr_arithmetic, 1) +CREATE_BINEXPR_PARSER('/', EXPR_BINARY_DIV, semantic_binexpr_arithmetic, 1) +CREATE_BINEXPR_PARSER('%', EXPR_BINARY_MOD, semantic_binexpr_arithmetic, 1) +CREATE_BINEXPR_PARSER('+', EXPR_BINARY_ADD, semantic_add, 1) +CREATE_BINEXPR_PARSER('-', EXPR_BINARY_SUB, semantic_sub, 1) +CREATE_BINEXPR_PARSER('<', EXPR_BINARY_LESS, semantic_comparison, 1) +CREATE_BINEXPR_PARSER('>', EXPR_BINARY_GREATER, semantic_comparison, 1) +CREATE_BINEXPR_PARSER('=', EXPR_BINARY_ASSIGN, semantic_binexpr_assign, 0) + +CREATE_BINEXPR_PARSER(T_EQUALEQUAL, EXPR_BINARY_EQUAL, + semantic_comparison, 1) +CREATE_BINEXPR_PARSER(T_EXCLAMATIONMARKEQUAL, EXPR_BINARY_NOTEQUAL, semantic_comparison, 1) -CREATE_BINEXPR_PARSER(T_LESSEQUAL, BINEXPR_LESSEQUAL, semantic_comparison, 1) -CREATE_BINEXPR_PARSER(T_GREATEREQUAL, BINEXPR_GREATEREQUAL, +CREATE_BINEXPR_PARSER(T_LESSEQUAL, EXPR_BINARY_LESSEQUAL, 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, +CREATE_BINEXPR_PARSER(T_GREATEREQUAL, EXPR_BINARY_GREATEREQUAL, + semantic_comparison, 1) + +CREATE_BINEXPR_PARSER('&', EXPR_BINARY_BITWISE_AND, + semantic_binexpr_arithmetic, 1) +CREATE_BINEXPR_PARSER('|', EXPR_BINARY_BITWISE_OR, + semantic_binexpr_arithmetic, 1) +CREATE_BINEXPR_PARSER('^', EXPR_BINARY_BITWISE_XOR, + semantic_binexpr_arithmetic, 1) +CREATE_BINEXPR_PARSER(T_ANDAND, EXPR_BINARY_LOGICAL_AND, + semantic_logical_op, 1) +CREATE_BINEXPR_PARSER(T_PIPEPIPE, EXPR_BINARY_LOGICAL_OR, + semantic_logical_op, 1) +CREATE_BINEXPR_PARSER(T_LESSLESS, EXPR_BINARY_SHIFTLEFT, semantic_shift_op, 1) -CREATE_BINEXPR_PARSER(T_GREATERGREATER, BINEXPR_SHIFTRIGHT, +CREATE_BINEXPR_PARSER(T_GREATERGREATER, EXPR_BINARY_SHIFTRIGHT, semantic_shift_op, 1) -CREATE_BINEXPR_PARSER(T_PLUSEQUAL, BINEXPR_ADD_ASSIGN, +CREATE_BINEXPR_PARSER(T_PLUSEQUAL, EXPR_BINARY_ADD_ASSIGN, semantic_arithmetic_addsubb_assign, 0) -CREATE_BINEXPR_PARSER(T_MINUSEQUAL, BINEXPR_SUB_ASSIGN, +CREATE_BINEXPR_PARSER(T_MINUSEQUAL, EXPR_BINARY_SUB_ASSIGN, semantic_arithmetic_addsubb_assign, 0) -CREATE_BINEXPR_PARSER(T_ASTERISKEQUAL, BINEXPR_MUL_ASSIGN, +CREATE_BINEXPR_PARSER(T_ASTERISKEQUAL, EXPR_BINARY_MUL_ASSIGN, semantic_arithmetic_assign, 0) -CREATE_BINEXPR_PARSER(T_SLASHEQUAL, BINEXPR_DIV_ASSIGN, +CREATE_BINEXPR_PARSER(T_SLASHEQUAL, EXPR_BINARY_DIV_ASSIGN, semantic_arithmetic_assign, 0) -CREATE_BINEXPR_PARSER(T_PERCENTEQUAL, BINEXPR_MOD_ASSIGN, +CREATE_BINEXPR_PARSER(T_PERCENTEQUAL, EXPR_BINARY_MOD_ASSIGN, semantic_arithmetic_assign, 0) -CREATE_BINEXPR_PARSER(T_LESSLESSEQUAL, BINEXPR_SHIFTLEFT_ASSIGN, +CREATE_BINEXPR_PARSER(T_LESSLESSEQUAL, EXPR_BINARY_SHIFTLEFT_ASSIGN, semantic_arithmetic_assign, 0) -CREATE_BINEXPR_PARSER(T_GREATERGREATEREQUAL, BINEXPR_SHIFTRIGHT_ASSIGN, +CREATE_BINEXPR_PARSER(T_GREATERGREATEREQUAL, EXPR_BINARY_SHIFTRIGHT_ASSIGN, semantic_arithmetic_assign, 0) -CREATE_BINEXPR_PARSER(T_ANDEQUAL, BINEXPR_BITWISE_AND_ASSIGN, +CREATE_BINEXPR_PARSER(T_ANDEQUAL, EXPR_BINARY_BITWISE_AND_ASSIGN, semantic_arithmetic_assign, 0) -CREATE_BINEXPR_PARSER(T_PIPEEQUAL, BINEXPR_BITWISE_OR_ASSIGN, +CREATE_BINEXPR_PARSER(T_PIPEEQUAL, EXPR_BINARY_BITWISE_OR_ASSIGN, semantic_arithmetic_assign, 0) -CREATE_BINEXPR_PARSER(T_CARETEQUAL, BINEXPR_BITWISE_XOR_ASSIGN, +CREATE_BINEXPR_PARSER(T_CARETEQUAL, EXPR_BINARY_BITWISE_XOR_ASSIGN, semantic_arithmetic_assign, 0) static expression_t *parse_sub_expression(unsigned precedence) @@ -3891,7 +4602,7 @@ 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) { @@ -3907,45 +4618,55 @@ static expression_t *parse_sub_expression(unsigned precedence) left = parser->infix_parser(parser->infix_precedence, left); assert(left != NULL); - assert(left->type != EXPR_UNKNOWN); - left->source_position = source_position; + assert(left->kind != EXPR_UNKNOWN); + left->base.source_position = source_position; } return left; } +/** + * Parse an expression. + */ static expression_t *parse_expression(void) { return parse_sub_expression(1); } - - +/** + * Register a parser for a prefix-like operator with given precedence. + * + * @param parser the parser function + * @param token_type the token type of the prefix token + * @param precedence the precedence of the operator + */ 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); - fprintf(stderr, "\n"); + diagnosticf("for token '%k'\n", (token_type_t)token_type); panic("trying to register multiple expression parsers for a token"); } entry->parser = parser; entry->precedence = precedence; } -static void register_expression_infix_parser( - parse_expression_infix_function parser, int token_type, - unsigned precedence) +/** + * Register a parser for an infix operator with given precedence. + * + * @param parser the parser function + * @param token_type the token type of the infix operator + * @param precedence the precedence of the operator + */ +static void register_infix_parser(parse_expression_infix_function parser, + int token_type, unsigned precedence) { expression_parser_function_t *entry = &expression_parsers[token_type]; if(entry->infix_parser != NULL) { - fprintf(stderr, "for token "); - print_token_type(stderr, token_type); - fprintf(stderr, "\n"); + diagnosticf("for token '%k'\n", (token_type_t)token_type); panic("trying to register multiple infix expression parsers for a " "token"); } @@ -3953,107 +4674,281 @@ static void register_expression_infix_parser( entry->infix_precedence = precedence; } +/** + * Initialize the expression parsers. + */ static void init_expression_parsers(void) { memset(&expression_parsers, 0, sizeof(expression_parsers)); - register_expression_infix_parser(parse_BINEXPR_MUL, '*', 16); - register_expression_infix_parser(parse_BINEXPR_DIV, '/', 16); - register_expression_infix_parser(parse_BINEXPR_MOD, '%', 16); - register_expression_infix_parser(parse_BINEXPR_SHIFTLEFT, T_LESSLESS, 16); - register_expression_infix_parser(parse_BINEXPR_SHIFTRIGHT, - T_GREATERGREATER, 16); - register_expression_infix_parser(parse_BINEXPR_ADD, '+', 15); - register_expression_infix_parser(parse_BINEXPR_SUB, '-', 15); - register_expression_infix_parser(parse_BINEXPR_LESS, '<', 14); - register_expression_infix_parser(parse_BINEXPR_GREATER, '>', 14); - register_expression_infix_parser(parse_BINEXPR_LESSEQUAL, T_LESSEQUAL, 14); - register_expression_infix_parser(parse_BINEXPR_GREATEREQUAL, - T_GREATEREQUAL, 14); - register_expression_infix_parser(parse_BINEXPR_EQUAL, T_EQUALEQUAL, 13); - register_expression_infix_parser(parse_BINEXPR_NOTEQUAL, - T_EXCLAMATIONMARKEQUAL, 13); - register_expression_infix_parser(parse_BINEXPR_BITWISE_AND, '&', 12); - register_expression_infix_parser(parse_BINEXPR_BITWISE_XOR, '^', 11); - register_expression_infix_parser(parse_BINEXPR_BITWISE_OR, '|', 10); - register_expression_infix_parser(parse_BINEXPR_LOGICAL_AND, T_ANDAND, 9); - register_expression_infix_parser(parse_BINEXPR_LOGICAL_OR, T_PIPEPIPE, 8); - register_expression_infix_parser(parse_conditional_expression, '?', 7); - register_expression_infix_parser(parse_BINEXPR_ASSIGN, '=', 2); - register_expression_infix_parser(parse_BINEXPR_ADD_ASSIGN, T_PLUSEQUAL, 2); - register_expression_infix_parser(parse_BINEXPR_SUB_ASSIGN, T_MINUSEQUAL, 2); - register_expression_infix_parser(parse_BINEXPR_MUL_ASSIGN, - T_ASTERISKEQUAL, 2); - register_expression_infix_parser(parse_BINEXPR_DIV_ASSIGN, T_SLASHEQUAL, 2); - register_expression_infix_parser(parse_BINEXPR_MOD_ASSIGN, - T_PERCENTEQUAL, 2); - register_expression_infix_parser(parse_BINEXPR_SHIFTLEFT_ASSIGN, + register_infix_parser(parse_array_expression, '[', 30); + register_infix_parser(parse_call_expression, '(', 30); + register_infix_parser(parse_select_expression, '.', 30); + register_infix_parser(parse_select_expression, T_MINUSGREATER, 30); + register_infix_parser(parse_EXPR_UNARY_POSTFIX_INCREMENT, + T_PLUSPLUS, 30); + register_infix_parser(parse_EXPR_UNARY_POSTFIX_DECREMENT, + T_MINUSMINUS, 30); + + register_infix_parser(parse_EXPR_BINARY_MUL, '*', 16); + register_infix_parser(parse_EXPR_BINARY_DIV, '/', 16); + register_infix_parser(parse_EXPR_BINARY_MOD, '%', 16); + register_infix_parser(parse_EXPR_BINARY_SHIFTLEFT, T_LESSLESS, 16); + register_infix_parser(parse_EXPR_BINARY_SHIFTRIGHT, T_GREATERGREATER, 16); + register_infix_parser(parse_EXPR_BINARY_ADD, '+', 15); + register_infix_parser(parse_EXPR_BINARY_SUB, '-', 15); + register_infix_parser(parse_EXPR_BINARY_LESS, '<', 14); + register_infix_parser(parse_EXPR_BINARY_GREATER, '>', 14); + register_infix_parser(parse_EXPR_BINARY_LESSEQUAL, T_LESSEQUAL, 14); + register_infix_parser(parse_EXPR_BINARY_GREATEREQUAL, T_GREATEREQUAL, 14); + register_infix_parser(parse_EXPR_BINARY_EQUAL, T_EQUALEQUAL, 13); + register_infix_parser(parse_EXPR_BINARY_NOTEQUAL, + T_EXCLAMATIONMARKEQUAL, 13); + register_infix_parser(parse_EXPR_BINARY_BITWISE_AND, '&', 12); + register_infix_parser(parse_EXPR_BINARY_BITWISE_XOR, '^', 11); + register_infix_parser(parse_EXPR_BINARY_BITWISE_OR, '|', 10); + register_infix_parser(parse_EXPR_BINARY_LOGICAL_AND, T_ANDAND, 9); + register_infix_parser(parse_EXPR_BINARY_LOGICAL_OR, T_PIPEPIPE, 8); + register_infix_parser(parse_conditional_expression, '?', 7); + register_infix_parser(parse_EXPR_BINARY_ASSIGN, '=', 2); + register_infix_parser(parse_EXPR_BINARY_ADD_ASSIGN, T_PLUSEQUAL, 2); + register_infix_parser(parse_EXPR_BINARY_SUB_ASSIGN, T_MINUSEQUAL, 2); + register_infix_parser(parse_EXPR_BINARY_MUL_ASSIGN, T_ASTERISKEQUAL, 2); + register_infix_parser(parse_EXPR_BINARY_DIV_ASSIGN, T_SLASHEQUAL, 2); + register_infix_parser(parse_EXPR_BINARY_MOD_ASSIGN, T_PERCENTEQUAL, 2); + register_infix_parser(parse_EXPR_BINARY_SHIFTLEFT_ASSIGN, T_LESSLESSEQUAL, 2); - register_expression_infix_parser(parse_BINEXPR_SHIFTRIGHT_ASSIGN, + register_infix_parser(parse_EXPR_BINARY_SHIFTRIGHT_ASSIGN, T_GREATERGREATEREQUAL, 2); - register_expression_infix_parser(parse_BINEXPR_BITWISE_AND_ASSIGN, + register_infix_parser(parse_EXPR_BINARY_BITWISE_AND_ASSIGN, T_ANDEQUAL, 2); - register_expression_infix_parser(parse_BINEXPR_BITWISE_OR_ASSIGN, + register_infix_parser(parse_EXPR_BINARY_BITWISE_OR_ASSIGN, T_PIPEEQUAL, 2); - register_expression_infix_parser(parse_BINEXPR_BITWISE_XOR_ASSIGN, + register_infix_parser(parse_EXPR_BINARY_BITWISE_XOR_ASSIGN, T_CARETEQUAL, 2); - register_expression_infix_parser(parse_BINEXPR_COMMA, ',', 1); - - register_expression_infix_parser(parse_array_expression, '[', 30); - register_expression_infix_parser(parse_call_expression, '(', 30); - register_expression_infix_parser(parse_select_expression, '.', 30); - register_expression_infix_parser(parse_select_expression, - T_MINUSGREATER, 30); - register_expression_infix_parser(parse_UNEXPR_POSTFIX_INCREMENT, - T_PLUSPLUS, 30); - register_expression_infix_parser(parse_UNEXPR_POSTFIX_DECREMENT, - T_MINUSMINUS, 30); - - register_expression_parser(parse_UNEXPR_NEGATE, '-', 25); - register_expression_parser(parse_UNEXPR_PLUS, '+', 25); - register_expression_parser(parse_UNEXPR_NOT, '!', 25); - register_expression_parser(parse_UNEXPR_BITWISE_NEGATE, '~', 25); - register_expression_parser(parse_UNEXPR_DEREFERENCE, '*', 25); - register_expression_parser(parse_UNEXPR_TAKE_ADDRESS, '&', 25); - register_expression_parser(parse_UNEXPR_PREFIX_INCREMENT, T_PLUSPLUS, 25); - register_expression_parser(parse_UNEXPR_PREFIX_DECREMENT, T_MINUSMINUS, 25); + register_infix_parser(parse_EXPR_BINARY_COMMA, ',', 1); + + register_expression_parser(parse_EXPR_UNARY_NEGATE, '-', 25); + register_expression_parser(parse_EXPR_UNARY_PLUS, '+', 25); + register_expression_parser(parse_EXPR_UNARY_NOT, '!', 25); + register_expression_parser(parse_EXPR_UNARY_BITWISE_NEGATE, '~', 25); + register_expression_parser(parse_EXPR_UNARY_DEREFERENCE, '*', 25); + register_expression_parser(parse_EXPR_UNARY_TAKE_ADDRESS, '&', 25); + register_expression_parser(parse_EXPR_UNARY_PREFIX_INCREMENT, + T_PLUSPLUS, 25); + register_expression_parser(parse_EXPR_UNARY_PREFIX_DECREMENT, + T_MINUSMINUS, 25); register_expression_parser(parse_sizeof, T_sizeof, 25); register_expression_parser(parse_extension, T___extension__, 25); register_expression_parser(parse_builtin_classify_type, T___builtin_classify_type, 25); } +/** + * Parse a asm statement constraints specification. + */ +static asm_constraint_t *parse_asm_constraints(void) +{ + asm_constraint_t *result = NULL; + asm_constraint_t *last = NULL; + + while(token.type == T_STRING_LITERAL || token.type == '[') { + asm_constraint_t *constraint = allocate_ast_zero(sizeof(constraint[0])); + memset(constraint, 0, sizeof(constraint[0])); + + if(token.type == '[') { + eat('['); + if(token.type != T_IDENTIFIER) { + parse_error_expected("while parsing asm constraint", + T_IDENTIFIER, 0); + return NULL; + } + constraint->symbol = token.v.symbol; + + expect(']'); + } + + constraint->constraints = parse_string_literals(); + expect('('); + constraint->expression = parse_expression(); + expect(')'); + + if(last != NULL) { + last->next = constraint; + } else { + result = constraint; + } + last = constraint; + + if(token.type != ',') + break; + eat(','); + } + + return result; +} + +/** + * Parse a asm statement clobber specification. + */ +static asm_clobber_t *parse_asm_clobbers(void) +{ + asm_clobber_t *result = NULL; + asm_clobber_t *last = NULL; + + while(token.type == T_STRING_LITERAL) { + asm_clobber_t *clobber = allocate_ast_zero(sizeof(clobber[0])); + clobber->clobber = parse_string_literals(); + + if(last != NULL) { + last->next = clobber; + } else { + result = clobber; + } + last = clobber; + + if(token.type != ',') + break; + eat(','); + } + + return result; +} + +/** + * Parse an asm statement. + */ +static statement_t *parse_asm_statement(void) +{ + eat(T_asm); + + statement_t *statement = allocate_statement_zero(STATEMENT_ASM); + statement->base.source_position = token.source_position; + asm_statement_t *asm_statement = &statement->asms; + + if(token.type == T_volatile) { + next_token(); + asm_statement->is_volatile = true; + } + + expect('('); + asm_statement->asm_text = parse_string_literals(); + + if(token.type != ':') + goto end_of_asm; + eat(':'); + + asm_statement->inputs = parse_asm_constraints(); + if(token.type != ':') + goto end_of_asm; + eat(':'); + + asm_statement->outputs = parse_asm_constraints(); + if(token.type != ':') + goto end_of_asm; + eat(':'); + + asm_statement->clobbers = parse_asm_clobbers(); + +end_of_asm: + expect(')'); + expect(';'); + return statement; +} + +/** + * Parse a case statement. + */ static statement_t *parse_case_statement(void) { eat(T_case); - case_label_statement_t *label = allocate_ast_zero(sizeof(label[0])); - label->statement.type = STATEMENT_CASE_LABEL; - label->statement.source_position = token.source_position; - label->expression = parse_expression(); + statement_t *statement = allocate_statement_zero(STATEMENT_CASE_LABEL); + + statement->base.source_position = token.source_position; + statement->case_label.expression = parse_expression(); expect(':'); - label->label_statement = parse_statement(); - return (statement_t*) label; + if (! is_constant_expression(statement->case_label.expression)) { + errorf(statement->base.source_position, + "case label does not reduce to an integer constant"); + } else { + /* TODO: check if the case label is already known */ + if (current_switch != NULL) { + /* link all cases into the switch statement */ + if (current_switch->last_case == NULL) { + current_switch->first_case = + current_switch->last_case = &statement->case_label; + } else { + current_switch->last_case->next = &statement->case_label; + } + } else { + errorf(statement->base.source_position, + "case label not within a switch statement"); + } + } + statement->case_label.label_statement = parse_statement(); + + return statement; +} + +/** + * Finds an existing default label of a switch statement. + */ +static case_label_statement_t * +find_default_label(const switch_statement_t *statement) +{ + for (case_label_statement_t *label = statement->first_case; + label != NULL; + label = label->next) { + if (label->expression == NULL) + return label; + } + return NULL; } +/** + * Parse a default statement. + */ static statement_t *parse_default_statement(void) { eat(T_default); - case_label_statement_t *label = allocate_ast_zero(sizeof(label[0])); - label->statement.type = STATEMENT_CASE_LABEL; - label->statement.source_position = token.source_position; + statement_t *statement = allocate_statement_zero(STATEMENT_CASE_LABEL); + + statement->base.source_position = token.source_position; expect(':'); - label->label_statement = parse_statement(); + if (current_switch != NULL) { + const case_label_statement_t *def_label = find_default_label(current_switch); + if (def_label != NULL) { + errorf(HERE, "multiple default labels in one switch"); + errorf(def_label->statement.source_position, + "this is the first default label"); + } else { + /* link all cases into the switch statement */ + if (current_switch->last_case == NULL) { + current_switch->first_case = + current_switch->last_case = &statement->case_label; + } else { + current_switch->last_case->next = &statement->case_label; + } + } + } else { + errorf(statement->base.source_position, + "'default' label not within a switch statement"); + } + statement->label.label_statement = parse_statement(); - return (statement_t*) label; + return statement; } +/** + * Return the declaration for a given label symbol or create a new one. + */ static declaration_t *get_label(symbol_t *symbol) { declaration_t *candidate = get_declaration(symbol, NAMESPACE_LABEL); @@ -4066,8 +4961,8 @@ static declaration_t *get_label(symbol_t *symbol) } /* otherwise we need to create a new one */ - declaration_t *declaration = allocate_ast_zero(sizeof(declaration[0])); - declaration->namespc = NAMESPACE_LABEL; + declaration_t *const declaration = allocate_declaration_zero(); + declaration->namespc = NAMESPACE_LABEL; declaration->symbol = symbol; label_push(declaration); @@ -4075,6 +4970,9 @@ static declaration_t *get_label(symbol_t *symbol) return declaration; } +/** + * Parse a label statement. + */ static statement_t *parse_label_statement(void) { assert(token.type == T_IDENTIFIER); @@ -4086,25 +4984,24 @@ static statement_t *parse_label_statement(void) /* 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); + errorf(HERE, "duplicate label '%Y'", symbol); + errorf(label->source_position, "previous definition of '%Y' was here", + symbol); } 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.kind = STATEMENT_LABEL; label_statement->statement.source_position = token.source_position; label_statement->label = label; - expect(':'); + eat(':'); if(token.type == '}') { - parse_error("label at end of compound statement"); + /* TODO only warn? */ + errorf(HERE, "label at end of compound statement"); return (statement_t*) label_statement; } else { label_statement->label_statement = parse_statement(); @@ -4113,12 +5010,15 @@ static statement_t *parse_label_statement(void) return (statement_t*) label_statement; } +/** + * Parse an if statement. + */ static statement_t *parse_if(void) { eat(T_if); if_statement_t *statement = allocate_ast_zero(sizeof(statement[0])); - statement->statement.type = STATEMENT_IF; + statement->statement.kind = STATEMENT_IF; statement->statement.source_position = token.source_position; expect('('); @@ -4134,47 +5034,72 @@ static statement_t *parse_if(void) return (statement_t*) statement; } +/** + * Parse a switch statement. + */ static statement_t *parse_switch(void) { eat(T_switch); switch_statement_t *statement = allocate_ast_zero(sizeof(statement[0])); - statement->statement.type = STATEMENT_SWITCH; + statement->statement.kind = STATEMENT_SWITCH; statement->statement.source_position = token.source_position; expect('('); - statement->expression = parse_expression(); + expression_t *const expr = parse_expression(); + type_t *const type = promote_integer(skip_typeref(expr->base.datatype)); + statement->expression = create_implicit_cast(expr, type); expect(')'); + + switch_statement_t *rem = current_switch; + current_switch = statement; statement->body = parse_statement(); + current_switch = rem; return (statement_t*) statement; } +static statement_t *parse_loop_body(statement_t *const loop) +{ + statement_t *const rem = current_loop; + current_loop = loop; + statement_t *const body = parse_statement(); + current_loop = rem; + return body; +} + +/** + * Parse a while statement. + */ static statement_t *parse_while(void) { eat(T_while); while_statement_t *statement = allocate_ast_zero(sizeof(statement[0])); - statement->statement.type = STATEMENT_WHILE; + statement->statement.kind = STATEMENT_WHILE; statement->statement.source_position = token.source_position; expect('('); statement->condition = parse_expression(); expect(')'); - statement->body = parse_statement(); + + statement->body = parse_loop_body((statement_t*)statement); return (statement_t*) statement; } +/** + * Parse a do statement. + */ static statement_t *parse_do(void) { eat(T_do); do_while_statement_t *statement = allocate_ast_zero(sizeof(statement[0])); - statement->statement.type = STATEMENT_DO_WHILE; + statement->statement.kind = STATEMENT_DO_WHILE; statement->statement.source_position = token.source_position; - statement->body = parse_statement(); + statement->body = parse_loop_body((statement_t*)statement); expect(T_while); expect('('); statement->condition = parse_expression(); @@ -4184,12 +5109,15 @@ static statement_t *parse_do(void) return (statement_t*) statement; } +/** + * Parse a for statement. + */ static statement_t *parse_for(void) { eat(T_for); for_statement_t *statement = allocate_ast_zero(sizeof(statement[0])); - statement->statement.type = STATEMENT_FOR; + statement->statement.kind = STATEMENT_FOR; statement->statement.source_position = token.source_position; expect('('); @@ -4200,7 +5128,7 @@ static statement_t *parse_for(void) if(token.type != ';') { if(is_declaration_specifier(&token, false)) { - parse_declaration(); + parse_declaration(record_declaration); } else { statement->initialisation = parse_expression(); expect(';'); @@ -4217,7 +5145,7 @@ static statement_t *parse_for(void) statement->step = parse_expression(); } expect(')'); - statement->body = parse_statement(); + statement->body = parse_loop_body((statement_t*)statement); assert(context == &statement->context); set_context(last_context); @@ -4226,6 +5154,9 @@ static statement_t *parse_for(void) return (statement_t*) statement; } +/** + * Parse a goto statement. + */ static statement_t *parse_goto(void) { eat(T_goto); @@ -4242,127 +5173,208 @@ static statement_t *parse_goto(void) goto_statement_t *statement = allocate_ast_zero(sizeof(statement[0])); - statement->statement.type = STATEMENT_GOTO; + statement->statement.kind = STATEMENT_GOTO; statement->statement.source_position = token.source_position; statement->label = label; + /* remember the goto's in a list for later checking */ + if (goto_last == NULL) { + goto_first = goto_last = statement; + } else { + goto_last->next = statement; + } + expect(';'); return (statement_t*) statement; } +/** + * Parse a continue statement. + */ static statement_t *parse_continue(void) { + statement_t *statement; + if (current_loop == NULL) { + errorf(HERE, "continue statement not within loop"); + statement = NULL; + } else { + statement = allocate_statement_zero(STATEMENT_CONTINUE); + + statement->base.source_position = token.source_position; + } + eat(T_continue); expect(';'); - statement_t *statement = allocate_ast_zero(sizeof(statement[0])); - statement->type = STATEMENT_CONTINUE; - statement->source_position = token.source_position; - return statement; } +/** + * Parse a break statement. + */ static statement_t *parse_break(void) { + statement_t *statement; + if (current_switch == NULL && current_loop == NULL) { + errorf(HERE, "break statement not within loop or switch"); + statement = NULL; + } else { + statement = allocate_statement_zero(STATEMENT_BREAK); + + statement->base.source_position = token.source_position; + } + eat(T_break); expect(';'); - statement_t *statement = allocate_ast_zero(sizeof(statement[0])); - statement->type = STATEMENT_BREAK; - statement->source_position = token.source_position; - return statement; } +/** + * Check if a given declaration represents a local variable. + */ +static bool is_local_var_declaration(const declaration_t *declaration) { + switch ((storage_class_tag_t) declaration->storage_class) { + case STORAGE_CLASS_NONE: + case STORAGE_CLASS_AUTO: + case STORAGE_CLASS_REGISTER: { + const type_t *type = skip_typeref(declaration->type); + if(is_type_function(type)) { + return false; + } else { + return true; + } + } + default: + return false; + } +} + +/** + * Check if a given expression represents a local variable. + */ +static bool is_local_variable(const expression_t *expression) +{ + if (expression->base.kind != EXPR_REFERENCE) { + return false; + } + const declaration_t *declaration = expression->reference.declaration; + return is_local_var_declaration(declaration); +} + +/** + * Parse a return statement. + */ static statement_t *parse_return(void) { eat(T_return); return_statement_t *statement = allocate_ast_zero(sizeof(statement[0])); - statement->statement.type = STATEMENT_RETURN; + statement->statement.kind = STATEMENT_RETURN; statement->statement.source_position = token.source_position; - assert(current_function->type->type == TYPE_FUNCTION); - function_type_t *function_type = (function_type_t*) current_function->type; - type_t *return_type = function_type->result_type; + assert(is_type_function(current_function->type)); + function_type_t *function_type = ¤t_function->type->function; + type_t *return_type = function_type->return_type; - expression_t *return_value; + expression_t *return_value = NULL; if(token.type != ';') { return_value = parse_expression(); + } + expect(';'); + + if(return_type == NULL) + return (statement_t*) statement; + if(return_value != NULL && return_value->base.datatype == NULL) + return (statement_t*) statement; + + return_type = skip_typeref(return_type); - if(return_type == type_void && return_value->datatype != type_void) { - parse_warning("'return' with a value, in function returning void"); + 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)) { + warningf(statement->statement.source_position, + "'return' with a value, in function returning void"); return_value = NULL; } else { if(return_type != NULL) { semantic_assign(return_type, &return_value, "'return'"); } } + /* check for returning address of a local var */ + if (return_value->base.kind == EXPR_UNARY_TAKE_ADDRESS) { + const expression_t *expression = return_value->unary.value; + if (is_local_variable(expression)) { + warningf(statement->statement.source_position, + "function returns address of local variable"); + } + } } else { - return_value = NULL; - if(return_type != type_void) { - parse_warning("'return' without value, in function returning " - "non-void"); + if(!is_type_atomic(return_type, ATOMIC_TYPE_VOID)) { + warningf(statement->statement.source_position, + "'return' without value, in function returning non-void"); } } statement->return_value = return_value; - expect(';'); - return (statement_t*) statement; } +/** + * Parse a declaration statement. + */ static statement_t *parse_declaration_statement(void) { - declaration_t *before = last_declaration; + statement_t *statement = allocate_statement_zero(STATEMENT_DECLARATION); - declaration_statement_t *statement - = allocate_ast_zero(sizeof(statement[0])); - statement->statement.type = STATEMENT_DECLARATION; - statement->statement.source_position = token.source_position; - - declaration_specifiers_t specifiers; - memset(&specifiers, 0, sizeof(specifiers)); - parse_declaration_specifiers(&specifiers); + statement->base.source_position = token.source_position; - if(token.type == ';') { - eat(';'); - } else { - parse_init_declarators(&specifiers); - } + declaration_t *before = last_declaration; + parse_declaration(record_declaration); if(before == NULL) { - statement->declarations_begin = context->declarations; + statement->declaration.declarations_begin = context->declarations; } else { - statement->declarations_begin = before->next; + statement->declaration.declarations_begin = before->next; } - statement->declarations_end = last_declaration; + statement->declaration.declarations_end = last_declaration; - return (statement_t*) statement; + return statement; } +/** + * Parse an expression statement, ie. expr ';'. + */ static statement_t *parse_expression_statement(void) { - expression_statement_t *statement = allocate_ast_zero(sizeof(statement[0])); - statement->statement.type = STATEMENT_EXPRESSION; - statement->statement.source_position = token.source_position; + statement_t *statement = allocate_statement_zero(STATEMENT_EXPRESSION); - statement->expression = parse_expression(); + statement->base.source_position = token.source_position; + statement->expression.expression = parse_expression(); expect(';'); - return (statement_t*) statement; + return statement; } +/** + * Parse a statement. + */ static statement_t *parse_statement(void) { statement_t *statement = NULL; /* declaration or statement */ switch(token.type) { + case T_asm: + statement = parse_asm_statement(); + break; + case T_case: statement = parse_case_statement(); break; @@ -4448,16 +5460,20 @@ 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; } +/** + * Parse a compound statement. + */ static statement_t *parse_compound_statement(void) { compound_statement_t *compound_statement = allocate_ast_zero(sizeof(compound_statement[0])); - compound_statement->statement.type = STATEMENT_COMPOUND; + compound_statement->statement.kind = STATEMENT_COMPOUND; compound_statement->statement.source_position = token.source_position; eat('{'); @@ -4474,23 +5490,22 @@ 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"); + if(token.type == '}') { + next_token(); + } else { + errorf(compound_statement->statement.source_position, "end of file while looking for closing '}'"); } - next_token(); assert(context == &compound_statement->context); set_context(last_context); @@ -4499,6 +5514,29 @@ static statement_t *parse_compound_statement(void) return (statement_t*) compound_statement; } +/** + * Initialize builtin types. + */ +static void initialize_builtin_types(void) +{ + type_intmax_t = make_global_typedef("__intmax_t__", type_long_long); + type_size_t = make_global_typedef("__SIZE_TYPE__", type_unsigned_long); + type_ssize_t = make_global_typedef("__SSIZE_TYPE__", type_long); + type_ptrdiff_t = make_global_typedef("__PTRDIFF_TYPE__", type_long); + type_uintmax_t = make_global_typedef("__uintmax_t__", type_unsigned_long_long); + type_uptrdiff_t = make_global_typedef("__UPTRDIFF_TYPE__", type_unsigned_long); + type_wchar_t = make_global_typedef("__WCHAR_TYPE__", type_int); + type_wint_t = make_global_typedef("__WINT_TYPE__", type_int); + + type_intmax_t_ptr = make_pointer_type(type_intmax_t, TYPE_QUALIFIER_NONE); + type_ptrdiff_t_ptr = make_pointer_type(type_ptrdiff_t, TYPE_QUALIFIER_NONE); + type_ssize_t_ptr = make_pointer_type(type_ssize_t, TYPE_QUALIFIER_NONE); + type_wchar_t_ptr = make_pointer_type(type_wchar_t, TYPE_QUALIFIER_NONE); +} + +/** + * Parse a translation unit. + */ static translation_unit_t *parse_translation_unit(void) { translation_unit_t *unit = allocate_ast_zero(sizeof(unit[0])); @@ -4509,8 +5547,16 @@ static translation_unit_t *parse_translation_unit(void) assert(context == NULL); set_context(&unit->context); + initialize_builtin_types(); + while(token.type != T_EOF) { - parse_declaration(); + if (token.type == ';') { + /* TODO error in strict mode */ + warningf(HERE, "stray ';' outside of function"); + next_token(); + } else { + parse_external_declaration(); + } } assert(context == &unit->context); @@ -4523,11 +5569,18 @@ static translation_unit_t *parse_translation_unit(void) return unit; } +/** + * Parse the input. + * + * @return the translation unit or NULL if errors occurred. + */ translation_unit_t *parse(void) { environment_stack = NEW_ARR_F(stack_entry_t, 0); label_stack = NEW_ARR_F(stack_entry_t, 0); - found_error = false; + diagnostic_count = 0; + error_count = 0; + warning_count = 0; type_set_output(stderr); ast_set_output(stderr); @@ -4541,30 +5594,27 @@ translation_unit_t *parse(void) DEL_ARR_F(environment_stack); DEL_ARR_F(label_stack); - if(found_error) + if(error_count > 0) return NULL; return unit; } +/** + * Initialize the parser. + */ void init_parser(void) { init_expression_parsers(); obstack_init(&temp_obst); - type_int = make_atomic_type(ATOMIC_TYPE_INT, TYPE_QUALIFIER_NONE); - type_uint = make_atomic_type(ATOMIC_TYPE_UINT, 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); + symbol_t *const va_list_sym = symbol_table_insert("__builtin_va_list"); + type_valist = create_builtin_type(va_list_sym, type_void_ptr); } +/** + * Terminate the parser. + */ void exit_parser(void) { obstack_free(&temp_obst, NULL);