X-Git-Url: http://nsz.repo.hu/git/?a=blobdiff_plain;f=parser.c;h=560fa4dac0181b997c1ea0e1bd3e77f582a18cfa;hb=02e2aeb532aeea33a0f8af198b3c6e568d54b007;hp=520da3a1e5bda2fdffc45748b4317076a9a90ab0;hpb=3bd432f7fe5fa4d3f84ddeaf5806398c141e4f28;p=cparser diff --git a/parser.c b/parser.c index 520da3a..560fa4d 100644 --- a/parser.c +++ b/parser.c @@ -25,21 +25,28 @@ typedef struct { unsigned short namespace; } 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 context_t *global_context = NULL; -static context_t *context = NULL; -static declaration_t *last_declaration = NULL; -static struct obstack temp_obst; - -static type_t *type_int = NULL; -static type_t *type_double = NULL; -static type_t *type_const_char = NULL; -static type_t *type_string = NULL; -static type_t *type_void = NULL; -static type_t *type_size_t = NULL; +static token_t token; +static token_t lookahead_buffer[MAX_LOOKAHEAD]; +static int lookahead_bufpos; +static stack_entry_t *environment_stack = NULL; +static stack_entry_t *label_stack = NULL; +static context_t *global_context = NULL; +static context_t *context = NULL; +static declaration_t *last_declaration = NULL; +static declaration_t *current_function = NULL; +static struct obstack temp_obst; +static bool found_error; + +static type_t *type_int = NULL; +static type_t *type_uint = NULL; +static type_t *type_long_double = NULL; +static type_t *type_double = NULL; +static type_t *type_float = NULL; +static type_t *type_const_char = NULL; +static type_t *type_string = NULL; +static type_t *type_void = NULL; +static type_t *type_size_t = NULL; +static type_t *type_ptrdiff_t = NULL; static statement_t *parse_compound_statement(void); static statement_t *parse_statement(void); @@ -129,6 +136,11 @@ static inline size_t environment_top(void) return ARR_LEN(environment_stack); } +static inline size_t label_top(void) +{ + return ARR_LEN(label_stack); +} + static inline void next_token(void) @@ -154,14 +166,15 @@ static inline const token_t *look_ahead(int num) #define eat(token_type) do { assert(token.type == token_type); next_token(); } while(0) -void error(void) +static void error(void) { + found_error = true; #ifdef ABORT_ON_ERROR abort(); #endif } -void parser_print_prefix_pos(const source_position_t source_position) +static void parser_print_prefix_pos(const source_position_t source_position) { fputs(source_position.input_name, stderr); fputc(':', stderr); @@ -169,17 +182,17 @@ void parser_print_prefix_pos(const source_position_t source_position) fputs(": ", stderr); } -void parser_print_error_prefix_pos(const source_position_t source_position) +static void parser_print_error_prefix_pos( + const source_position_t source_position) { parser_print_prefix_pos(source_position); fputs("error: ", stderr); error(); } -void parser_print_error_prefix(void) +static void parser_print_error_prefix(void) { - parser_print_prefix_pos(token.source_position); - error(); + parser_print_error_prefix_pos(token.source_position); } static void parse_error(const char *message) @@ -188,13 +201,25 @@ static void parse_error(const char *message) fprintf(stderr, "parse error: %s\n", message); } -__attribute__((unused)) -static void parse_warning(const char *message) +static void parser_print_warning_prefix_pos( + const source_position_t source_position) { - parser_print_prefix_pos(token.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); +} + static void parse_error_expected(const char *message, ...) { va_list args; @@ -224,6 +249,33 @@ static void parse_error_expected(const char *message, ...) fprintf(stderr, "\n"); } +static void print_type_quoted(type_t *type) +{ + fputc('\'', stderr); + print_type(type); + fputc('\'', stderr); +} + +static void type_error(const char *msg, const source_position_t source_position, + type_t *type) +{ + parser_print_error_prefix_pos(source_position); + fprintf(stderr, "%s, but found type ", msg); + print_type_quoted(type); + fputc('\n', stderr); +} + +static void type_error_incompatible(const char *msg, + const source_position_t source_position, type_t *type1, type_t *type2) +{ + parser_print_error_prefix_pos(source_position); + fprintf(stderr, "%s, incompatible types: ", msg); + print_type_quoted(type1); + fprintf(stderr, " - "); + print_type_quoted(type2); + fprintf(stderr, ")\n"); +} + static void eat_block(void) { if(token.type == '{') @@ -342,6 +394,8 @@ static const char *get_namespace_prefix(namespace_t namespace) return "struct "; case NAMESPACE_ENUM: return "enum "; + case NAMESPACE_LABEL: + return "label "; } panic("invalid namespace found"); } @@ -350,15 +404,16 @@ static const char *get_namespace_prefix(namespace_t namespace) * pushs an environment_entry on the environment stack and links the * corresponding symbol to the new entry */ -static declaration_t *environment_push(declaration_t *declaration) +static declaration_t *stack_push(stack_entry_t **stack_ptr, + declaration_t *declaration, + context_t *parent_context) { symbol_t *symbol = declaration->symbol; - namespace_t namespace = declaration->namespace; - assert(declaration->source_position.input_name != NULL); + namespace_t namespace = (namespace_t)declaration->namespace; /* a declaration should be only pushed once */ assert(declaration->parent_context == NULL); - declaration->parent_context = context; + declaration->parent_context = parent_context; declaration_t *previous_declaration = get_declaration(symbol, namespace); assert(declaration != previous_declaration); @@ -368,15 +423,62 @@ static declaration_t *environment_push(declaration_t *declaration) parser_print_error_prefix_pos(declaration->source_position); fprintf(stderr, "definition of symbol %s%s with type ", get_namespace_prefix(namespace), symbol->string); - error(); - print_type(declaration->type); + 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(previous_declaration->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; } @@ -386,49 +488,65 @@ static declaration_t *environment_push(declaration_t *declaration) entry.symbol = symbol; entry.old_declaration = symbol->declaration; entry.namespace = namespace; - ARR_APP1(environment_stack, entry); + ARR_APP1(stack_entry_t, *stack_ptr, entry); /* replace/add declaration into declaration list of the symbol */ if(symbol->declaration == NULL) { symbol->declaration = declaration; } else { - declaration_t *iter = symbol->declaration; - for( ; iter != NULL; iter = iter->symbol_next) { - declaration_t *symbol_next = iter->symbol_next; - if(symbol_next == NULL) { - iter->symbol_next = declaration; - assert(declaration->symbol_next == NULL); - break; - } - if(symbol_next->namespace == namespace) { - iter->symbol_next = declaration; - declaration->symbol_next = symbol_next->symbol_next; + declaration_t *iter_last = NULL; + declaration_t *iter = symbol->declaration; + for( ; iter != NULL; iter_last = iter, iter = iter->symbol_next) { + /* replace an entry? */ + if(iter->namespace == namespace) { + if(iter_last == NULL) { + symbol->declaration = declaration; + } else { + iter_last->symbol_next = declaration; + } + declaration->symbol_next = iter->symbol_next; break; } } + if(iter == NULL) { + assert(iter_last->symbol_next == NULL); + iter_last->symbol_next = declaration; + } } return declaration; } +static declaration_t *environment_push(declaration_t *declaration) +{ + assert(declaration->source_position.input_name != NULL); + return stack_push(&environment_stack, declaration, context); +} + +static declaration_t *label_push(declaration_t *declaration) +{ + return stack_push(&label_stack, declaration, ¤t_function->context); +} + /** * pops symbols from the environment stack until @p new_top is the top element */ -static void environment_pop_to(size_t new_top) +static void stack_pop_to(stack_entry_t **stack_ptr, size_t new_top) { - size_t top = ARR_LEN(environment_stack); - size_t i; + stack_entry_t *stack = *stack_ptr; + size_t top = ARR_LEN(stack); + size_t i; assert(new_top <= top); if(new_top == top) return; for(i = top; i > new_top; --i) { - stack_entry_t *entry = & environment_stack[i - 1]; + stack_entry_t *entry = & stack[i - 1]; declaration_t *old_declaration = entry->old_declaration; symbol_t *symbol = entry->symbol; - namespace_t namespace = entry->namespace; + namespace_t namespace = (namespace_t)entry->namespace; /* replace/remove declaration */ declaration_t *declaration = symbol->declaration; @@ -438,27 +556,179 @@ static void environment_pop_to(size_t new_top) symbol->declaration = declaration->symbol_next; } else { symbol->declaration = old_declaration; - assert(old_declaration->symbol_next == - declaration->symbol_next); } } else { - for(; declaration != NULL; declaration = declaration->symbol_next) { - declaration_t *symbol_next = declaration->symbol_next; - if(symbol_next->namespace == namespace) { - declaration->symbol_next = old_declaration; - assert(old_declaration->symbol_next - == symbol_next->symbol_next); + declaration_t *iter_last = declaration; + declaration_t *iter = declaration->symbol_next; + for( ; iter != NULL; iter_last = iter, iter = iter->symbol_next) { + /* replace an entry? */ + if(iter->namespace == namespace) { + assert(iter_last != NULL); + iter_last->symbol_next = old_declaration; + old_declaration->symbol_next = iter->symbol_next; break; } } - assert(declaration != NULL); + assert(iter != NULL); } } - ARR_SHRINKLEN(environment_stack, (int) new_top); + ARR_SHRINKLEN(*stack_ptr, (int) new_top); +} + +static void environment_pop_to(size_t new_top) +{ + stack_pop_to(&environment_stack, new_top); +} + +static void label_pop_to(size_t new_top) +{ + stack_pop_to(&label_stack, new_top); +} + + +static int get_rank(const type_t *type) +{ + /* The C-standard allows promoting to int or unsigned int (see § 7.2.2 + * and esp. footnote 108). However we can't fold constants (yet), so we + * can't decide wether unsigned int is possible, while int always works. + * (unsigned int would be preferable when possible... for stuff like + * struct { enum { ... } bla : 4; } ) */ + if(type->type == TYPE_ENUM) + return ATOMIC_TYPE_INT; + + assert(type->type == TYPE_ATOMIC); + atomic_type_t *atomic_type = (atomic_type_t*) type; + atomic_type_type_t atype = atomic_type->atype; + return atype; } +static type_t *promote_integer(type_t *type) +{ + if(get_rank(type) < ATOMIC_TYPE_INT) + type = type_int; + return type; +} + +static expression_t *create_cast_expression(expression_t *expression, + type_t *dest_type) +{ + unary_expression_t *cast = allocate_ast_zero(sizeof(cast[0])); + + cast->expression.type = EXPR_UNARY; + cast->type = UNEXPR_CAST; + cast->value = expression; + cast->expression.datatype = dest_type; + + return (expression_t*) cast; +} + +static bool is_null_expression(const expression_t *const expr) +{ + if (expr->type != EXPR_CONST) return false; + + type_t *const type = skip_typeref(expr->datatype); + if (!is_type_integer(type)) return false; + + const const_t *const const_expr = (const const_t*)expr; + return const_expr->v.int_value == 0; +} + +static expression_t *create_implicit_cast(expression_t *expression, + type_t *dest_type) +{ + type_t *source_type = expression->datatype; + + if(source_type == NULL) + return expression; + + source_type = skip_typeref(source_type); + dest_type = skip_typeref(dest_type); + + if(source_type == dest_type) + return expression; + + if(dest_type->type == TYPE_ATOMIC) { + if(source_type->type != TYPE_ATOMIC) + 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); + return expression; + } + + return create_cast_expression(expression, dest_type); + } + if(dest_type->type == TYPE_POINTER) { + pointer_type_t *pointer_type + = (pointer_type_t*) dest_type; + switch (source_type->type) { + case TYPE_ATOMIC: + if (is_null_expression(expression)) { + return create_cast_expression(expression, dest_type); + } + break; + + case TYPE_POINTER: + if (pointers_compatible(source_type, dest_type)) { + return create_cast_expression(expression, dest_type); + } + break; + + case TYPE_ARRAY: { + array_type_t *const array_type = (array_type_t*) source_type; + if (types_compatible(array_type->element_type, + pointer_type->points_to)) { + return create_cast_expression(expression, dest_type); + } + break; + } + + default: + panic("casting of non-atomic types not implemented yet"); + } + + type_error_incompatible("can't implicitely cast types", + expression->source_position, + source_type, dest_type); + return expression; + } + + panic("casting of non-atomic types not implemented yet"); +} + +static void semantic_assign(type_t *orig_type_left, expression_t **right, + const char *context) +{ + type_t *orig_type_right = (*right)->datatype; + + if(orig_type_right == NULL) + return; + + type_t *const type_left = skip_typeref(orig_type_left); + type_t *const type_right = skip_typeref(orig_type_right); + + if(type_left == type_right) { + /* fine */ + } else if ((is_type_arithmetic(type_left) && is_type_arithmetic(type_right)) || + (type_left->type == TYPE_POINTER && is_null_expression(*right)) || + (type_left->type == TYPE_POINTER && type_right->type == TYPE_POINTER)) { + *right = create_implicit_cast(*right, type_left); + } else { + /* TODO: improve error message */ + parser_print_error_prefix(); + fprintf(stderr, "incompatible types in %s\n", context); + parser_print_error_prefix(); + print_type_quoted(type_left); + fputs(" <- ", stderr); + print_type_quoted(type_right); + fputs("\n", stderr); + } + +} static expression_t *parse_constant_expression(void) { @@ -472,17 +742,19 @@ static expression_t *parse_assignment_expression(void) return parse_sub_expression(2); } -static void parse_compound_type_entries(void); -static declaration_t *parse_declarator(storage_class_t storage_class, - type_t *type, int may_be_abstract); -static declaration_t *record_declaration(declaration_t *declaration); - typedef struct declaration_specifiers_t declaration_specifiers_t; struct declaration_specifiers_t { storage_class_t storage_class; + bool is_inline; type_t *type; }; +static void parse_compound_type_entries(void); +static declaration_t *parse_declarator( + const declaration_specifiers_t *specifiers, type_t *type, + bool may_be_abstract); +static declaration_t *record_declaration(declaration_t *declaration); + static const char *parse_string_literals(void) { assert(token.type == T_STRING_LITERAL); @@ -568,7 +840,7 @@ static designator_t *parse_designation(void) designator = allocate_ast_zero(sizeof(designator[0])); next_token(); if(token.type != T_IDENTIFIER) { - parse_error_expected("problem while parsing designator", + parse_error_expected("while parsing designator", T_IDENTIFIER, 0); return NULL; } @@ -590,35 +862,42 @@ static designator_t *parse_designation(void) } } -static initializer_t *parse_initializer_list(void); +static initializer_t *parse_initializer_list(type_t *type); -static initializer_t *parse_initializer(void) +static initializer_t *parse_initializer(type_t *type) { designator_t *designator = parse_designation(); initializer_t *result; if(token.type == '{') { - result = parse_initializer_list(); + result = parse_initializer_list(type); } else { result = allocate_ast_zero(sizeof(result[0])); result->type = INITIALIZER_VALUE; result->v.value = parse_assignment_expression(); + + if(type != NULL) { + semantic_assign(type, &result->v.value, "initializer"); + } } result->designator = designator; return result; } -static initializer_t *parse_initializer_list(void) +static initializer_t *parse_initializer_list(type_t *type) { eat('{'); + /* TODO: semantic */ + (void) type; + initializer_t *result = allocate_ast_zero(sizeof(result[0])); result->type = INITIALIZER_LIST; initializer_t *last = NULL; while(1) { - initializer_t *initializer = parse_initializer(); + initializer_t *initializer = parse_initializer(NULL); if(last != NULL) { last->next = initializer; } else { @@ -630,8 +909,7 @@ static initializer_t *parse_initializer_list(void) break; if(token.type != ',') { - parse_error_expected("problem while parsing initializer list", - ',', '}', 0); + parse_error_expected("while parsing initializer list", ',', '}', 0); eat_block(); return result; } @@ -668,10 +946,10 @@ static declaration_t *parse_compound_type_specifier(bool is_struct) } } else if(token.type != '{') { if(is_struct) { - parse_error_expected("problem while parsing struct type specifier", + parse_error_expected("while parsing struct type specifier", T_IDENTIFIER, '{', 0); } else { - parse_error_expected("problem while parsing union type specifier", + parse_error_expected("while parsing union type specifier", T_IDENTIFIER, '{', 0); } @@ -688,6 +966,7 @@ static declaration_t *parse_compound_type_specifier(bool is_struct) } declaration->source_position = token.source_position; declaration->symbol = symbol; + record_declaration(declaration); } if(token.type == '{') { @@ -698,7 +977,6 @@ static declaration_t *parse_compound_type_specifier(bool is_struct) is_struct ? "struct" : "union", symbol->string); declaration->context.declarations = NULL; } - record_declaration(declaration); declaration->init.is_defined = true; int top = environment_top(); @@ -730,8 +1008,7 @@ static void parse_enum_entries(void) declaration_t *entry = allocate_ast_zero(sizeof(entry[0])); if(token.type != T_IDENTIFIER) { - parse_error_expected("problem while parsing enum entry", - T_IDENTIFIER, 0); + parse_error_expected("while parsing enum entry", T_IDENTIFIER, 0); eat_block(); return; } @@ -742,7 +1019,7 @@ static void parse_enum_entries(void) if(token.type == '=') { next_token(); - entry->init.initializer = parse_initializer(); + entry->init.initializer = parse_initializer(type_int); } record_declaration(entry); @@ -768,7 +1045,7 @@ static declaration_t *parse_enum_specifier(void) declaration = get_declaration(symbol, NAMESPACE_ENUM); } else if(token.type != '{') { - parse_error_expected("problem while parsing enum type specifier", + parse_error_expected("while parsing enum type specifier", T_IDENTIFIER, '{', 0); return NULL; } else { @@ -887,6 +1164,8 @@ static type_t *create_builtin_type(symbol_t *symbol) builtin_type_t *type = allocate_type_zero(sizeof(type[0])); type->type.type = TYPE_BUILTIN; type->symbol = symbol; + /* TODO... */ + type->real_type = type_int; return (type_t*) type; } @@ -942,7 +1221,6 @@ static void parse_declaration_specifiers(declaration_specifiers_t *specifiers) MATCH_TYPE_QUALIFIER(T_const, TYPE_QUALIFIER_CONST); MATCH_TYPE_QUALIFIER(T_restrict, TYPE_QUALIFIER_RESTRICT); MATCH_TYPE_QUALIFIER(T_volatile, TYPE_QUALIFIER_VOLATILE); - MATCH_TYPE_QUALIFIER(T_inline, TYPE_QUALIFIER_INLINE); case T___extension__: /* TODO */ @@ -975,6 +1253,11 @@ static void parse_declaration_specifiers(declaration_specifiers_t *specifiers) #ifdef PROVIDE_IMAGINARY MATCH_SPECIFIER(T__Imaginary, SPECIFIER_IMAGINARY, "_Imaginary") #endif + case T_inline: + next_token(); + specifiers->is_inline = true; + break; + case T_long: next_token(); if(type_specifiers & SPECIFIER_LONG_LONG) { @@ -1171,7 +1454,7 @@ finish_specifiers: } } - type->qualifiers = type_qualifiers; + type->qualifiers = (type_qualifier_t)type_qualifiers; type_t *result = typehash_insert(type); if(newtype && result != (type_t*) type) { @@ -1181,9 +1464,9 @@ finish_specifiers: specifiers->type = result; } -static type_qualifier_t parse_type_qualifiers(void) +static unsigned parse_type_qualifiers(void) { - type_qualifier_t type_qualifiers = 0; + unsigned type_qualifiers = TYPE_QUALIFIER_NONE; while(true) { switch(token.type) { @@ -1191,7 +1474,6 @@ static type_qualifier_t parse_type_qualifiers(void) MATCH_TYPE_QUALIFIER(T_const, TYPE_QUALIFIER_CONST); MATCH_TYPE_QUALIFIER(T_restrict, TYPE_QUALIFIER_RESTRICT); MATCH_TYPE_QUALIFIER(T_volatile, TYPE_QUALIFIER_VOLATILE); - MATCH_TYPE_QUALIFIER(T_inline, TYPE_QUALIFIER_INLINE); default: return type_qualifiers; @@ -1203,8 +1485,8 @@ static void parse_identifier_list(void) { while(true) { if(token.type != T_IDENTIFIER) { - parse_error_expected("problem while parsing parameter identifier " - "list", T_IDENTIFIER, 0); + parse_error_expected("while parsing parameter identifier list", + T_IDENTIFIER, 0); return; } next_token(); @@ -1221,8 +1503,8 @@ static declaration_t *parse_parameter(void) parse_declaration_specifiers(&specifiers); - declaration_t *declaration = parse_declarator(specifiers.storage_class, - specifiers.type, 1); + declaration_t *declaration + = parse_declarator(&specifiers, specifiers.type, true); /* TODO check declaration constraints for parameters */ if(declaration->storage_class == STORAGE_CLASS_TYPEDEF) { @@ -1294,6 +1576,7 @@ static declaration_t *parse_parameters(function_type_t *type) } typedef enum { + CONSTRUCT_INVALID, CONSTRUCT_POINTER, CONSTRUCT_FUNCTION, CONSTRUCT_ARRAY @@ -1338,7 +1621,8 @@ static construct_type_t *parse_pointer_declarator(void) parsed_pointer_t *pointer = obstack_alloc(&temp_obst, sizeof(pointer[0])); memset(pointer, 0, sizeof(pointer[0])); - pointer->type_qualifiers = parse_type_qualifiers(); + pointer->construct_type.type = CONSTRUCT_POINTER; + pointer->type_qualifiers = parse_type_qualifiers(); return (construct_type_t*) pointer; } @@ -1349,6 +1633,7 @@ static construct_type_t *parse_array_declarator(void) parsed_array_t *array = obstack_alloc(&temp_obst, sizeof(array[0])); memset(array, 0, sizeof(array[0])); + array->construct_type.type = CONSTRUCT_ARRAY; if(token.type == T_static) { array->is_static = true; @@ -1438,8 +1723,7 @@ static construct_type_t *parse_inner_declarator(declaration_t *declaration, default: if(may_be_abstract) break; - parse_error_expected("problem while parsing declarator", T_IDENTIFIER, - '(', 0); + parse_error_expected("while parsing declarator", T_IDENTIFIER, '(', 0); } while(true) { @@ -1491,6 +1775,8 @@ static type_t *construct_declarator_type(construct_type_t *construct_list, 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; @@ -1533,11 +1819,13 @@ static type_t *construct_declarator_type(construct_type_t *construct_list, return type; } -static declaration_t *parse_declarator(storage_class_t storage_class, - type_t *type, int may_be_abstract) +static declaration_t *parse_declarator( + const declaration_specifiers_t *specifiers, + type_t *type, bool may_be_abstract) { declaration_t *declaration = allocate_ast_zero(sizeof(declaration[0])); - declaration->storage_class = storage_class; + declaration->storage_class = specifiers->storage_class; + declaration->is_inline = specifiers->is_inline; construct_type_t *construct_type = parse_inner_declarator(declaration, may_be_abstract); @@ -1594,16 +1882,23 @@ static void parser_error_multiple_definition(declaration_t *previous, parser_print_error_prefix_pos(previous->source_position); fprintf(stderr, "this is the location of the previous " "definition.\n"); - error(); } static void parse_init_declarators(const declaration_specifiers_t *specifiers) { while(true) { declaration_t *ndeclaration - = parse_declarator(specifiers->storage_class, specifiers->type, 0); + = parse_declarator(specifiers, specifiers->type, false); declaration_t *declaration = record_declaration(ndeclaration); + + type_t *type = declaration->type; + if(type->type != TYPE_FUNCTION && declaration->is_inline) { + parser_print_warning_prefix_pos(declaration->source_position); + fprintf(stderr, "variable ‘%s’ declared ‘inline’\n", + declaration->symbol->string); + } + if(token.type == '=') { next_token(); @@ -1613,7 +1908,7 @@ static void parse_init_declarators(const declaration_specifiers_t *specifiers) parser_error_multiple_definition(declaration, ndeclaration); } - ndeclaration->init.initializer = parse_initializer(); + ndeclaration->init.initializer = parse_initializer(declaration->type); } else if(token.type == '{') { if(declaration->type->type != TYPE_FUNCTION) { parser_print_error_prefix(); @@ -1642,8 +1937,16 @@ static void parse_init_declarators(const declaration_specifiers_t *specifiers) environment_push(parameter); } + int label_stack_top = label_top(); + declaration_t *old_current_function = current_function; + current_function = declaration; + statement_t *statement = parse_compound_statement(); + assert(current_function == declaration); + current_function = old_current_function; + label_pop_to(label_stack_top); + assert(context == &declaration->context); set_context(last_context); environment_pop_to(top); @@ -1668,8 +1971,7 @@ static void parse_struct_declarators(const declaration_specifiers_t *specifiers) /* TODO (bitfields) */ } else { declaration_t *declaration - = parse_declarator(specifiers->storage_class, - specifiers->type, 1); + = parse_declarator(specifiers, specifiers->type, true); /* TODO: check constraints for struct declarations */ /* TODO: check for doubled fields */ @@ -1715,6 +2017,29 @@ static void parse_declaration(void) parse_declaration_specifiers(&specifiers); if(token.type == ';') { + if (specifiers.storage_class != STORAGE_CLASS_NONE) { + parse_warning_pos(source_position, + "useless keyword in empty declaration"); + } + switch (specifiers.type->type) { + case TYPE_COMPOUND_STRUCT: + case TYPE_COMPOUND_UNION: { + const compound_type_t *const comp_type = + (const compound_type_t*)specifiers.type; + if (comp_type->declaration->symbol == NULL) { + parse_warning_pos(source_position, + "unnamed struct/union that defines no instances"); + } + break; + } + + case TYPE_ENUM: break; + + default: + parse_warning_pos(source_position, "empty declaration"); + break; + } + next_token(); declaration_t *declaration = allocate_ast_zero(sizeof(declaration[0])); @@ -1762,6 +2087,14 @@ struct expression_parser_function_t { expression_parser_function_t expression_parsers[T_LAST_TOKEN]; +static expression_t *make_invalid_expression(void) +{ + expression_t *expression = allocate_ast_zero(sizeof(expression[0])); + expression->type = EXPR_INVALID; + expression->source_position = token.source_position; + return expression; +} + static expression_t *expected_expression_error(void) { parser_print_error_prefix(); @@ -1769,11 +2102,9 @@ static expression_t *expected_expression_error(void) print_token(stderr, & token); fprintf(stderr, "\n"); - expression_t *expression = allocate_ast_zero(sizeof(expression[0])); - expression->type = EXPR_INVALID; next_token(); - return expression; + return make_invalid_expression(); } static expression_t *parse_string_const(void) @@ -1816,7 +2147,8 @@ static expression_t *parse_float_const(void) static declaration_t *create_implicit_function(symbol_t *symbol, const source_position_t source_position) { - function_type_t *function_type = allocate_type_zero(sizeof(function_type)); + function_type_t *function_type + = allocate_type_zero(sizeof(function_type[0])); function_type->type.type = TYPE_FUNCTION; function_type->result_type = type_int; @@ -1892,7 +2224,7 @@ static void check_cast_allowed(expression_t *expression, type_t *dest_type) { (void) expression; (void) dest_type; - /* TODO check if cast is allowed and issue warnings/errors */ + /* TODO check if explicit cast is allowed and issue warnings/errors */ } static expression_t *parse_cast(void) @@ -1921,13 +2253,23 @@ static expression_t *parse_statement_expression(void) statement_expression_t *expression = allocate_ast_zero(sizeof(expression[0])); expression->expression.type = EXPR_STATEMENT; - expression->statement = parse_compound_statement(); + + statement_t *statement = parse_compound_statement(); + expression->statement = statement; + if(statement == NULL) { + expect(')'); + return NULL; + } + + assert(statement->type == STATEMENT_COMPOUND); + compound_statement_t *compound_statement + = (compound_statement_t*) statement; /* find last statement and use it's type */ const statement_t *last_statement = NULL; - const statement_t *statement = expression->statement; - for( ; statement != NULL; statement = statement->next) { - last_statement = statement; + const statement_t *iter = compound_statement->statements; + for( ; iter != NULL; iter = iter->next) { + last_statement = iter; } if(last_statement->type == STATEMENT_EXPRESSION) { @@ -1999,7 +2341,7 @@ static designator_t *parse_designator(void) designator_t *result = allocate_ast_zero(sizeof(result[0])); if(token.type != T_IDENTIFIER) { - parse_error_expected("problem while parsing member designator", + parse_error_expected("while parsing member designator", T_IDENTIFIER, 0); eat_brace(); return NULL; @@ -2012,8 +2354,8 @@ static designator_t *parse_designator(void) if(token.type == '.') { next_token(); if(token.type != T_IDENTIFIER) { - parse_error_expected("problem while parsing member designator", - T_IDENTIFIER, 0); + parse_error_expected("while parsing member designator", + T_IDENTIFIER, 0); eat_brace(); return NULL; } @@ -2128,11 +2470,7 @@ static expression_t *parse_primary_expression(void) fprintf(stderr, "\n"); eat_statement(); - expression_t *expression = allocate_ast_zero(sizeof(expression[0])); - expression->type = EXPR_INVALID; - expression->datatype = type_void; - - return expression; + return make_invalid_expression(); } static expression_t *parse_array_expression(unsigned precedence, @@ -2149,15 +2487,18 @@ static expression_t *parse_array_expression(unsigned precedence, array_access->array_ref = array_ref; array_access->index = parse_expression(); - type_t *array_type = array_ref->datatype; - if(array_type != NULL) { - if(array_type->type == TYPE_POINTER) { - pointer_type_t *pointer = (pointer_type_t*) array_type; + type_t *type = array_ref->datatype; + if(type != NULL) { + if(type->type == TYPE_POINTER) { + pointer_type_t *pointer = (pointer_type_t*) type; array_access->expression.datatype = pointer->points_to; + } else if(type->type == TYPE_ARRAY) { + array_type_t *array_type = (array_type_t*) type; + array_access->expression.datatype = array_type->element_type; } else { parser_print_error_prefix(); fprintf(stderr, "array access on object with non-pointer type "); - print_type(array_type); + print_type_quoted(type); fprintf(stderr, "\n"); } } @@ -2216,8 +2557,9 @@ static expression_t *parse_select_expression(unsigned precedence, expression_t *compound) { (void) precedence; - assert(token.type == '.' || token.type == T_MINUSGREATER); + + bool is_pointer = (token.type == T_MINUSGREATER); next_token(); select_expression_t *select = allocate_ast_zero(sizeof(select[0])); @@ -2225,15 +2567,69 @@ static expression_t *parse_select_expression(unsigned precedence, select->expression.type = EXPR_SELECT; select->compound = compound; - /* TODO: datatype */ - if(token.type != T_IDENTIFIER) { - parse_error_expected("Problem while parsing select", T_IDENTIFIER, 0); + parse_error_expected("while parsing select", T_IDENTIFIER, 0); return (expression_t*) select; } - select->symbol = token.v.symbol; + symbol_t *symbol = token.v.symbol; + select->symbol = symbol; next_token(); + type_t *type = compound->datatype; + if(type == NULL) + return make_invalid_expression(); + + 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(type); + fputc('\n', stderr); + return make_invalid_expression(); + } + pointer_type_t *pointer_type = (pointer_type_t*) type; + type_left = pointer_type->points_to; + } + type_left = skip_typeref(type_left); + + if(type_left->type != TYPE_COMPOUND_STRUCT + && type_left->type != TYPE_COMPOUND_UNION) { + parser_print_error_prefix(); + fprintf(stderr, "request for member '%s' in something not a struct or " + "union, but ", symbol->string); + print_type_quoted(type_left); + fputc('\n', stderr); + return make_invalid_expression(); + } + + compound_type_t *compound_type = (compound_type_t*) type_left; + declaration_t *declaration = compound_type->declaration; + + if(!declaration->init.is_defined) { + parser_print_error_prefix(); + fprintf(stderr, "request for member '%s' of incomplete type ", + symbol->string); + print_type_quoted(type_left); + fputc('\n', stderr); + return make_invalid_expression(); + } + + declaration_t *iter = declaration->context.declarations; + for( ; iter != NULL; iter = iter->next) { + if(iter->symbol == symbol) { + break; + } + } + if(iter == NULL) { + parser_print_error_prefix(); + print_type_quoted(type_left); + fprintf(stderr, " has no memeber named '%s'\n", symbol->string); + return make_invalid_expression(); + } + + select->compound_entry = iter; + select->expression.datatype = iter->type; return (expression_t*) select; } @@ -2242,9 +2638,26 @@ static expression_t *parse_call_expression(unsigned precedence, { (void) precedence; call_expression_t *call = allocate_ast_zero(sizeof(call[0])); + call->expression.type = EXPR_CALL; + call->function = expression; - call->expression.type = EXPR_CALL; - call->function = expression; + function_type_t *function_type; + type_t *type = expression->datatype; + if(type->type != TYPE_FUNCTION) { + /* TODO calling pointers to functions is ok */ + parser_print_error_prefix(); + fputs("called object '", stderr); + print_expression(expression); + fputs("' (type ", stderr); + print_type_quoted(type); + fputs("is not a function\n", stderr); + + function_type = NULL; + call->expression.datatype = NULL; + } else { + function_type = (function_type_t*) type; + call->expression.datatype = function_type->result_type; + } /* parse arguments */ eat('('); @@ -2270,51 +2683,53 @@ static expression_t *parse_call_expression(unsigned precedence, } expect(')'); - type_t *type = expression->datatype; - if(type != NULL) { - /* we can call pointer to function */ - if(type->type == TYPE_POINTER) { - pointer_type_t *pointer = (pointer_type_t*) type; - type = pointer->points_to; + if(function_type != NULL) { + function_parameter_t *parameter = function_type->parameters; + call_argument_t *argument = call->arguments; + for( ; parameter != NULL && argument != NULL; + parameter = parameter->next, argument = argument->next) { + type_t *expected_type = parameter->type; + /* TODO report context in error messages */ + argument->expression = create_implicit_cast(argument->expression, + expected_type); } - - if(type == NULL || type->type != TYPE_FUNCTION) { + /* too few parameters */ + if(parameter != NULL) { parser_print_error_prefix(); - fprintf(stderr, "expected a function type for call but found " - "type "); - print_type(expression->datatype); - fprintf(stderr, "\n"); - } else { - function_type_t *function_type = (function_type_t*) type; - call->expression.datatype = function_type->result_type; + fprintf(stderr, "too few arguments to function '"); + print_expression(expression); + fprintf(stderr, "'\n"); + } else if(argument != NULL) { + /* too many parameters */ + if(!function_type->variadic + && !function_type->unspecified_parameters) { + parser_print_error_prefix(); + fprintf(stderr, "too many arguments to function '"); + print_expression(expression); + fprintf(stderr, "'\n"); + } else { + /* do default promotion */ + for( ; argument != NULL; argument = argument->next) { + type_t *type = argument->expression->datatype; + + if(type == NULL) + continue; + + if(is_type_integer(type)) { + type = promote_integer(type); + } else if(type == type_float) { + type = type_double; + } + argument->expression + = create_implicit_cast(argument->expression, type); + } + } } } return (expression_t*) call; } -static 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(type); - fputc('\n', stderr); - error(); -} - -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(type1); - fprintf(stderr, " - "); - print_type(type2); - fprintf(stderr, ")\n"); - error(); -} - static type_t *get_type_after_conversion(const type_t *type1, const type_t *type2) { @@ -2336,7 +2751,7 @@ static expression_t *parse_conditional_expression(unsigned precedence, /* 6.5.15.2 */ type_t *condition_type = conditional->condition->datatype; if(condition_type != NULL) { - if(!is_type_scalar(condition_type)) { + if(!is_type_scalar(skip_typeref(condition_type))) { type_error("expected a scalar type", expression->source_position, condition_type); } @@ -2353,27 +2768,33 @@ static expression_t *parse_conditional_expression(unsigned precedence, if(false_type == NULL) return (expression_t*) conditional; + type_t *const skipped_true_type = skip_typeref(true_type); + type_t *const skipped_false_type = skip_typeref(false_type); + /* 6.5.15.3 */ - if(true_type == false_type) { - conditional->expression.datatype = true_type; - } else if(is_type_arithmetic(true_type) && is_type_arithmetic(false_type)) { - type_t *result = get_type_after_conversion(true_type, false_type); + 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 = get_type_after_conversion(skipped_true_type, + skipped_false_type); /* TODO: create implicit convs if necessary */ conditional->expression.datatype = result; - } else if(true_type->type == TYPE_POINTER && - false_type->type == TYPE_POINTER && + } 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(true_type) && false_type->type == TYPE_POINTER) - || (is_null_ptr_const(false_type) && - true_type->type == TYPE_POINTER) TODO*/ false) { + } else if(/* (is_null_ptr_const(skipped_true_type) && + skipped_false_type->type == TYPE_POINTER) + || (is_null_ptr_const(skipped_false_type) && + skipped_true_type->type == TYPE_POINTER) TODO*/ false) { /* TODO */ } else if(/* 1 is pointer to object type, other is void* */ false) { /* TODO */ } else { - type_error_incompatible("problem while parsing conditional", + type_error_incompatible("while parsing conditional", expression->source_position, true_type, - false_type); + skipped_false_type); } return (expression_t*) conditional; @@ -2388,26 +2809,77 @@ static expression_t *parse_extension(unsigned precedence) return parse_sub_expression(precedence); } -static type_t *get_unexpr_arithmetic_type(const expression_t *expression) +static void semantic_incdec(unary_expression_t *expression) { - /* TODO */ - return expression->datatype; + type_t *orig_type = expression->value->datatype; + if(orig_type == NULL) + return; + + type_t *type = skip_typeref(orig_type); + if(!is_type_arithmetic(type) && type->type != TYPE_POINTER) { + /* TODO: improve error message */ + parser_print_error_prefix(); + fprintf(stderr, "operation needs an arithmetic or pointer type\n"); + return; + } + + expression->expression.datatype = orig_type; } -static type_t *get_unexpr_dereference_type(const expression_t *expression) +static void semantic_unexpr_arithmetic(unary_expression_t *expression) { - (void) expression; - /* TODO... */ - return NULL; + type_t *orig_type = expression->value->datatype; + if(orig_type == NULL) + return; + + type_t *type = skip_typeref(orig_type); + if(!is_type_arithmetic(type)) { + /* TODO: improve error message */ + parser_print_error_prefix(); + fprintf(stderr, "operation needs an arithmetic type\n"); + return; + } + + expression->expression.datatype = orig_type; } -static type_t *get_unexpr_take_addr_type(const expression_t *expression) +static void semantic_dereference(unary_expression_t *expression) { - type_t *type = expression->datatype; - return make_pointer_type(type, 0); + type_t *orig_type = expression->value->datatype; + if(orig_type == NULL) + return; + + type_t *type = skip_typeref(orig_type); + if(type->type != TYPE_POINTER) { + /* TODO: improve error message */ + parser_print_error_prefix(); + fprintf(stderr, "operation needs a pointer type\n"); + return; + } + + pointer_type_t *pointer_type = (pointer_type_t*) type; + expression->expression.datatype = pointer_type->points_to; +} + +static void semantic_take_addr(unary_expression_t *expression) +{ + type_t *orig_type = expression->value->datatype; + if(orig_type == NULL) + return; + + expression_t *value = expression->value; + if(value->type == EXPR_REFERENCE) { + reference_expression_t *reference = (reference_expression_t*) value; + declaration_t *declaration = reference->declaration; + if(declaration != NULL) { + declaration->address_taken = 1; + } + } + + expression->expression.datatype = make_pointer_type(orig_type, 0); } -#define CREATE_UNARY_EXPRESSION_PARSER(token_type, unexpression_type, tfunc) \ +#define CREATE_UNARY_EXPRESSION_PARSER(token_type, unexpression_type, sfunc) \ static expression_t *parse_##unexpression_type(unsigned precedence) \ { \ eat(token_type); \ @@ -2417,27 +2889,26 @@ static expression_t *parse_##unexpression_type(unsigned precedence) \ unary_expression->expression.type = EXPR_UNARY; \ unary_expression->type = unexpression_type; \ unary_expression->value = parse_sub_expression(precedence); \ - unary_expression->expression.datatype = tfunc(unary_expression->value); \ + \ + sfunc(unary_expression); \ \ return (expression_t*) unary_expression; \ } -CREATE_UNARY_EXPRESSION_PARSER('-', UNEXPR_NEGATE, get_unexpr_arithmetic_type) -CREATE_UNARY_EXPRESSION_PARSER('+', UNEXPR_PLUS, get_unexpr_arithmetic_type) -CREATE_UNARY_EXPRESSION_PARSER('!', UNEXPR_NOT, get_unexpr_arithmetic_type) -CREATE_UNARY_EXPRESSION_PARSER('*', UNEXPR_DEREFERENCE, - get_unexpr_dereference_type) -CREATE_UNARY_EXPRESSION_PARSER('&', UNEXPR_TAKE_ADDRESS, - get_unexpr_take_addr_type) +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_arithmetic) +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, - get_unexpr_arithmetic_type) + semantic_unexpr_arithmetic) CREATE_UNARY_EXPRESSION_PARSER(T_PLUSPLUS, UNEXPR_PREFIX_INCREMENT, - get_unexpr_arithmetic_type) + semantic_incdec) CREATE_UNARY_EXPRESSION_PARSER(T_MINUSMINUS, UNEXPR_PREFIX_DECREMENT, - get_unexpr_arithmetic_type) + semantic_incdec) #define CREATE_UNARY_POSTFIX_EXPRESSION_PARSER(token_type, unexpression_type, \ - tfunc) \ + sfunc) \ static expression_t *parse_##unexpression_type(unsigned precedence, \ expression_t *left) \ { \ @@ -2449,65 +2920,352 @@ static expression_t *parse_##unexpression_type(unsigned precedence, \ unary_expression->expression.type = EXPR_UNARY; \ unary_expression->type = unexpression_type; \ unary_expression->value = left; \ - unary_expression->expression.datatype = tfunc(left); \ + \ + sfunc(unary_expression); \ \ return (expression_t*) unary_expression; \ } CREATE_UNARY_POSTFIX_EXPRESSION_PARSER(T_PLUSPLUS, UNEXPR_POSTFIX_INCREMENT, - get_unexpr_arithmetic_type) + semantic_incdec) CREATE_UNARY_POSTFIX_EXPRESSION_PARSER(T_MINUSMINUS, UNEXPR_POSTFIX_DECREMENT, - get_unexpr_arithmetic_type) - -#define CREATE_BINEXPR_PARSER(token_type, binexpression_type) \ -static \ -expression_t *parse_##binexpression_type(unsigned precedence, \ - expression_t *left) \ -{ \ - eat(token_type); \ - \ - expression_t *right = parse_sub_expression(precedence); \ - \ - binary_expression_t *binexpr \ - = allocate_ast_zero(sizeof(binexpr[0])); \ - binexpr->expression.type = EXPR_BINARY; \ - binexpr->type = binexpression_type; \ - binexpr->left = left; \ - binexpr->right = right; \ - \ - return (expression_t*) binexpr; \ -} - -CREATE_BINEXPR_PARSER(',', BINEXPR_COMMA) -CREATE_BINEXPR_PARSER('*', BINEXPR_MUL) -CREATE_BINEXPR_PARSER('/', BINEXPR_DIV) -CREATE_BINEXPR_PARSER('%', BINEXPR_MOD) -CREATE_BINEXPR_PARSER('+', BINEXPR_ADD) -CREATE_BINEXPR_PARSER('-', BINEXPR_SUB) -CREATE_BINEXPR_PARSER('<', BINEXPR_LESS) -CREATE_BINEXPR_PARSER('>', BINEXPR_GREATER) -CREATE_BINEXPR_PARSER('=', BINEXPR_ASSIGN) -CREATE_BINEXPR_PARSER(T_EQUALEQUAL, BINEXPR_EQUAL) -CREATE_BINEXPR_PARSER(T_EXCLAMATIONMARKEQUAL, BINEXPR_NOTEQUAL) -CREATE_BINEXPR_PARSER(T_LESSEQUAL, BINEXPR_LESSEQUAL) -CREATE_BINEXPR_PARSER(T_GREATEREQUAL, BINEXPR_GREATEREQUAL) -CREATE_BINEXPR_PARSER('&', BINEXPR_BITWISE_AND) -CREATE_BINEXPR_PARSER('|', BINEXPR_BITWISE_OR) -CREATE_BINEXPR_PARSER('^', BINEXPR_BITWISE_XOR) -CREATE_BINEXPR_PARSER(T_ANDAND, BINEXPR_LOGICAL_AND) -CREATE_BINEXPR_PARSER(T_PIPEPIPE, BINEXPR_LOGICAL_OR) -CREATE_BINEXPR_PARSER(T_LESSLESS, BINEXPR_SHIFTLEFT) -CREATE_BINEXPR_PARSER(T_GREATERGREATER, BINEXPR_SHIFTRIGHT) -CREATE_BINEXPR_PARSER(T_PLUSEQUAL, BINEXPR_ADD_ASSIGN) -CREATE_BINEXPR_PARSER(T_MINUSEQUAL, BINEXPR_SUB_ASSIGN) -CREATE_BINEXPR_PARSER(T_ASTERISKEQUAL, BINEXPR_MUL_ASSIGN) -CREATE_BINEXPR_PARSER(T_SLASHEQUAL, BINEXPR_DIV_ASSIGN) -CREATE_BINEXPR_PARSER(T_PERCENTEQUAL, BINEXPR_MOD_ASSIGN) -CREATE_BINEXPR_PARSER(T_LESSLESSEQUAL, BINEXPR_SHIFTLEFT_ASSIGN) -CREATE_BINEXPR_PARSER(T_GREATERGREATEREQUAL, BINEXPR_SHIFTRIGHT_ASSIGN) -CREATE_BINEXPR_PARSER(T_ANDEQUAL, BINEXPR_BITWISE_AND_ASSIGN) -CREATE_BINEXPR_PARSER(T_PIPEEQUAL, BINEXPR_BITWISE_OR_ASSIGN) -CREATE_BINEXPR_PARSER(T_CARETEQUAL, BINEXPR_BITWISE_XOR_ASSIGN) + semantic_incdec) + +static type_t *semantic_arithmetic(type_t *type_left, type_t *type_right) +{ + /* TODO: handle complex + imaginary types */ + + /* § 6.3.1.8 Usual arithmetic conversions */ + if(type_left == type_long_double || type_right == type_long_double) { + return type_long_double; + } else if(type_left == type_double || type_right == type_double) { + return type_double; + } else if(type_left == type_float || type_right == type_float) { + return type_float; + } + + type_right = promote_integer(type_right); + type_left = promote_integer(type_left); + + if(type_left == type_right) + return type_left; + + bool signed_left = is_type_signed(type_left); + bool signed_right = is_type_signed(type_right); + if(get_rank(type_left) < get_rank(type_right)) { + if(signed_left == signed_right || !signed_right) { + return type_right; + } else { + return type_left; + } + } else { + if(signed_left == signed_right || !signed_left) { + return type_left; + } else { + return type_right; + } + } +} + +static 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; + + if(orig_type_left == NULL || orig_type_right == NULL) + return; + + type_t *type_left = skip_typeref(orig_type_left); + type_t *type_right = skip_typeref(orig_type_right); + + if(!is_type_arithmetic(type_left) || !is_type_arithmetic(type_right)) { + /* TODO: improve error message */ + parser_print_error_prefix(); + fprintf(stderr, "operation needs arithmetic types\n"); + return; + } + + type_t *arithmetic_type = semantic_arithmetic(type_left, type_right); + expression->left = create_implicit_cast(left, arithmetic_type); + expression->right = create_implicit_cast(right, arithmetic_type); + expression->expression.datatype = arithmetic_type; +} + +static 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; + + if(orig_type_left == NULL || orig_type_right == NULL) + return; + + type_t *type_left = skip_typeref(orig_type_left); + type_t *type_right = skip_typeref(orig_type_right); + + if(!is_type_integer(type_left) || !is_type_integer(type_right)) { + /* TODO: improve error message */ + parser_print_error_prefix(); + fprintf(stderr, "operation needs integer types\n"); + return; + } + + type_left = promote_integer(type_left); + type_right = promote_integer(type_right); + + expression->left = create_implicit_cast(left, type_left); + expression->right = create_implicit_cast(right, type_right); + expression->expression.datatype = type_left; +} + +static 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; + + if(orig_type_left == NULL || orig_type_right == NULL) + return; + + type_t *type_left = skip_typeref(orig_type_left); + type_t *type_right = skip_typeref(orig_type_right); + + /* § 5.6.5 */ + if(is_type_arithmetic(type_left) && is_type_arithmetic(type_right)) { + type_t *arithmetic_type = semantic_arithmetic(type_left, type_right); + expression->left = create_implicit_cast(left, arithmetic_type); + expression->right = create_implicit_cast(right, arithmetic_type); + expression->expression.datatype = arithmetic_type; + return; + } else if(type_left->type == TYPE_POINTER && is_type_integer(type_right)) { + expression->expression.datatype = type_left; + } else if(type_right->type == TYPE_POINTER && is_type_integer(type_left)) { + expression->expression.datatype = type_right; + } else { + parser_print_error_prefix(); + fprintf(stderr, "invalid operands to binary + ("); + print_type_quoted(orig_type_left); + fprintf(stderr, ", "); + print_type_quoted(orig_type_right); + fprintf(stderr, ")\n"); + } +} + +static void semantic_sub(binary_expression_t *expression) +{ + expression_t *left = expression->left; + expression_t *right = expression->right; + type_t *orig_type_left = left->datatype; + type_t *orig_type_right = right->datatype; + + if(orig_type_left == NULL || orig_type_right == NULL) + return; + + type_t *type_left = skip_typeref(orig_type_left); + type_t *type_right = skip_typeref(orig_type_right); + + /* § 5.6.5 */ + if(is_type_arithmetic(type_left) && is_type_arithmetic(type_right)) { + type_t *arithmetic_type = semantic_arithmetic(type_left, type_right); + expression->left = create_implicit_cast(left, arithmetic_type); + expression->right = create_implicit_cast(right, arithmetic_type); + expression->expression.datatype = arithmetic_type; + return; + } else if(type_left->type == TYPE_POINTER && is_type_integer(type_right)) { + expression->expression.datatype = type_left; + } else if(type_left->type == TYPE_POINTER && + type_right->type == TYPE_POINTER) { + if(!pointers_compatible(type_left, type_right)) { + parser_print_error_prefix(); + fprintf(stderr, "pointers to incompatible objects to binary - ("); + print_type_quoted(orig_type_left); + fprintf(stderr, ", "); + print_type_quoted(orig_type_right); + fprintf(stderr, ")\n"); + } else { + expression->expression.datatype = type_ptrdiff_t; + } + } else { + parser_print_error_prefix(); + fprintf(stderr, "invalid operands to binary - ("); + print_type_quoted(orig_type_left); + fprintf(stderr, ", "); + print_type_quoted(orig_type_right); + fprintf(stderr, ")\n"); + } +} + +static 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; + + if(orig_type_left == NULL || orig_type_right == NULL) + return; + + type_t *type_left = skip_typeref(orig_type_left); + type_t *type_right = skip_typeref(orig_type_right); + + /* TODO non-arithmetic types */ + if(is_type_arithmetic(type_left) && is_type_arithmetic(type_right)) { + type_t *arithmetic_type = semantic_arithmetic(type_left, type_right); + expression->left = create_implicit_cast(left, arithmetic_type); + expression->right = create_implicit_cast(right, arithmetic_type); + expression->expression.datatype = arithmetic_type; + } else if (type_left->type == TYPE_POINTER && + type_right->type == TYPE_POINTER) { + /* TODO check compatibility */ + } else if (type_left->type == TYPE_POINTER) { + expression->right = create_implicit_cast(right, type_left); + } else if (type_right->type == TYPE_POINTER) { + expression->left = create_implicit_cast(left, type_right); + } else { + type_error_incompatible("invalid operands in comparison", + expression->expression.source_position, + type_left, type_right); + } + expression->expression.datatype = type_int; +} + +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; + + if(orig_type_left == NULL || orig_type_right == NULL) + return; + + type_t *type_left = skip_typeref(orig_type_left); + type_t *type_right = skip_typeref(orig_type_right); + + if(!is_type_arithmetic(type_left) || !is_type_arithmetic(type_right)) { + /* TODO: improve error message */ + parser_print_error_prefix(); + fprintf(stderr, "operation needs arithmetic types\n"); + return; + } + + /* combined instructions are tricky. We can't create an implicit cast on + * the left side, because we need the uncasted form for the store. + * The ast2firm pass has to know that left_type must be right_type + * for the arithmeitc operation and create a cast by itself */ + type_t *arithmetic_type = semantic_arithmetic(type_left, type_right); + expression->right = create_implicit_cast(right, arithmetic_type); + expression->expression.datatype = type_left; +} + +static 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; + + if(orig_type_left == NULL || orig_type_right == NULL) + return; + + type_t *type_left = skip_typeref(orig_type_left); + type_t *type_right = skip_typeref(orig_type_right); + + if (!is_type_scalar(type_left) || !is_type_scalar(type_right)) { + /* TODO: improve error message */ + parser_print_error_prefix(); + fprintf(stderr, "operation needs scalar types\n"); + return; + } + + expression->expression.datatype = type_int; +} + +static void semantic_binexpr_assign(binary_expression_t *expression) +{ + expression_t *left = expression->left; + type_t *type_left = left->datatype; + + if(type_left != NULL) { + semantic_assign(type_left, &expression->right, "assignment"); + } + + expression->expression.datatype = type_left; +} + +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, + semantic_comparison, 1) +CREATE_BINEXPR_PARSER(T_LESSEQUAL, BINEXPR_LESSEQUAL, semantic_comparison, 1) +CREATE_BINEXPR_PARSER(T_GREATEREQUAL, BINEXPR_GREATEREQUAL, + semantic_comparison, 1) +CREATE_BINEXPR_PARSER('&', BINEXPR_BITWISE_AND, semantic_binexpr_arithmetic, 1) +CREATE_BINEXPR_PARSER('|', BINEXPR_BITWISE_OR, semantic_binexpr_arithmetic, 1) +CREATE_BINEXPR_PARSER('^', BINEXPR_BITWISE_XOR, semantic_binexpr_arithmetic, 1) +CREATE_BINEXPR_PARSER(T_ANDAND, BINEXPR_LOGICAL_AND, semantic_logical_op, 1) +CREATE_BINEXPR_PARSER(T_PIPEPIPE, BINEXPR_LOGICAL_OR, semantic_logical_op, 1) +/* TODO shift has a bit special semantic */ +CREATE_BINEXPR_PARSER(T_LESSLESS, BINEXPR_SHIFTLEFT, + semantic_shift_op, 1) +CREATE_BINEXPR_PARSER(T_GREATERGREATER, BINEXPR_SHIFTRIGHT, + semantic_shift_op, 1) +CREATE_BINEXPR_PARSER(T_PLUSEQUAL, BINEXPR_ADD_ASSIGN, + semantic_arithmetic_assign, 0) +CREATE_BINEXPR_PARSER(T_MINUSEQUAL, BINEXPR_SUB_ASSIGN, + semantic_arithmetic_assign, 0) +CREATE_BINEXPR_PARSER(T_ASTERISKEQUAL, BINEXPR_MUL_ASSIGN, + semantic_arithmetic_assign, 0) +CREATE_BINEXPR_PARSER(T_SLASHEQUAL, BINEXPR_DIV_ASSIGN, + semantic_arithmetic_assign, 0) +CREATE_BINEXPR_PARSER(T_PERCENTEQUAL, BINEXPR_MOD_ASSIGN, + semantic_arithmetic_assign, 0) +CREATE_BINEXPR_PARSER(T_LESSLESSEQUAL, BINEXPR_SHIFTLEFT_ASSIGN, + semantic_arithmetic_assign, 0) +CREATE_BINEXPR_PARSER(T_GREATERGREATEREQUAL, BINEXPR_SHIFTRIGHT_ASSIGN, + semantic_arithmetic_assign, 0) +CREATE_BINEXPR_PARSER(T_ANDEQUAL, BINEXPR_BITWISE_AND_ASSIGN, + semantic_arithmetic_assign, 0) +CREATE_BINEXPR_PARSER(T_PIPEEQUAL, BINEXPR_BITWISE_OR_ASSIGN, + semantic_arithmetic_assign, 0) +CREATE_BINEXPR_PARSER(T_CARETEQUAL, BINEXPR_BITWISE_XOR_ASSIGN, + semantic_arithmetic_assign, 0) static expression_t *parse_sub_expression(unsigned precedence) { @@ -2542,7 +3300,7 @@ static expression_t *parse_sub_expression(unsigned precedence) left = parser->infix_parser(parser->infix_precedence, left); assert(left != NULL); - assert(left->type != EXPR_INVALID); + assert(left->type != EXPR_UNKNOWN); left->source_position = source_position; } @@ -2556,8 +3314,8 @@ static expression_t *parse_expression(void) -void register_expression_parser(parse_expression_function parser, - int token_type, unsigned precedence) +static void register_expression_parser(parse_expression_function parser, + int token_type, unsigned precedence) { expression_parser_function_t *entry = &expression_parsers[token_type]; @@ -2571,8 +3329,9 @@ void register_expression_parser(parse_expression_function parser, entry->precedence = precedence; } -void register_expression_infix_parser(parse_expression_infix_function parser, - int token_type, unsigned precedence) +static void register_expression_infix_parser( + parse_expression_infix_function parser, int token_type, + unsigned precedence) { expression_parser_function_t *entry = &expression_parsers[token_type]; @@ -2686,13 +3445,63 @@ static statement_t *parse_default_statement(void) return (statement_t*) label; } +static declaration_t *get_label(symbol_t *symbol) +{ + declaration_t *candidate = get_declaration(symbol, NAMESPACE_LABEL); + assert(current_function != NULL); + /* if we found a label in the same function, then we already created the + * declaration */ + if(candidate != NULL + && candidate->parent_context == ¤t_function->context) { + return candidate; + } + + /* otherwise we need to create a new one */ + declaration_t *declaration = allocate_ast_zero(sizeof(declaration[0])); + declaration->namespace = NAMESPACE_LABEL; + declaration->symbol = symbol; + + label_push(declaration); + + return declaration; +} + static statement_t *parse_label_statement(void) { - eat(T_IDENTIFIER); + assert(token.type == T_IDENTIFIER); + symbol_t *symbol = token.v.symbol; + next_token(); + + declaration_t *label = get_label(symbol); + + /* if source position is already set then the label is defined twice, + * otherwise it was just mentioned in a goto so far */ + if(label->source_position.input_name != NULL) { + parser_print_error_prefix(); + fprintf(stderr, "duplicate label '%s'\n", symbol->string); + parser_print_error_prefix_pos(label->source_position); + fprintf(stderr, "previous definition of '%s' was here\n", + symbol->string); + } else { + label->source_position = token.source_position; + } + + label_statement_t *label_statement = allocate_ast_zero(sizeof(label[0])); + + label_statement->statement.type = STATEMENT_LABEL; + label_statement->statement.source_position = token.source_position; + label_statement->label = label; + expect(':'); - parse_statement(); - return NULL; + if(token.type == '}') { + parse_error("label at end of compound statement"); + return (statement_t*) label_statement; + } else { + label_statement->label_statement = parse_statement(); + } + + return (statement_t*) label_statement; } static statement_t *parse_if(void) @@ -2811,10 +3620,27 @@ static statement_t *parse_for(void) static statement_t *parse_goto(void) { eat(T_goto); - expect(T_IDENTIFIER); + + if(token.type != T_IDENTIFIER) { + parse_error_expected("while parsing goto", T_IDENTIFIER, 0); + eat_statement(); + return NULL; + } + symbol_t *symbol = token.v.symbol; + next_token(); + + declaration_t *label = get_label(symbol); + + goto_statement_t *statement = allocate_ast_zero(sizeof(statement[0])); + + statement->statement.type = STATEMENT_GOTO; + statement->statement.source_position = token.source_position; + + statement->label = label; + expect(';'); - return NULL; + return (statement_t*) statement; } static statement_t *parse_continue(void) @@ -2823,8 +3649,8 @@ static statement_t *parse_continue(void) expect(';'); statement_t *statement = allocate_ast_zero(sizeof(statement[0])); - statement->source_position = token.source_position; statement->type = STATEMENT_CONTINUE; + statement->source_position = token.source_position; return statement; } @@ -2835,8 +3661,8 @@ static statement_t *parse_break(void) expect(';'); statement_t *statement = allocate_ast_zero(sizeof(statement[0])); - statement->source_position = token.source_position; statement->type = STATEMENT_BREAK; + statement->source_position = token.source_position; return statement; } @@ -2849,9 +3675,32 @@ static statement_t *parse_return(void) statement->statement.type = STATEMENT_RETURN; statement->statement.source_position = token.source_position; + + assert(current_function->type->type == TYPE_FUNCTION); + function_type_t *function_type = (function_type_t*) current_function->type; + type_t *return_type = function_type->result_type; + + expression_t *return_value; if(token.type != ';') { - statement->return_value = parse_expression(); + return_value = parse_expression(); + + if(return_type == type_void && return_value->datatype != type_void) { + parse_warning("'return' with a value, in function returning void"); + return_value = NULL; + } else { + if(return_type != NULL) { + semantic_assign(return_type, &return_value, "'return'"); + } + } + } else { + return_value = NULL; + if(return_type != type_void) { + parse_warning("'return' without value, in function returning " + "non-void"); + } } + statement->return_value = return_value; + expect(';'); return (statement_t*) statement; @@ -2997,13 +3846,13 @@ static statement_t *parse_statement(void) static statement_t *parse_compound_statement(void) { - eat('{'); - compound_statement_t *compound_statement = allocate_ast_zero(sizeof(compound_statement[0])); compound_statement->statement.type = STATEMENT_COMPOUND; compound_statement->statement.source_position = token.source_position; + eat('{'); + int top = environment_top(); context_t *last_context = context; set_context(&compound_statement->context); @@ -3027,12 +3876,17 @@ static statement_t *parse_compound_statement(void) last_statement = statement; } + if(token.type != '}') { + parser_print_error_prefix_pos( + compound_statement->statement.source_position); + fprintf(stderr, "end of file while looking for closing '}'\n"); + } + next_token(); + assert(context == &compound_statement->context); set_context(last_context); environment_pop_to(top); - next_token(); - return (statement_t*) compound_statement; } @@ -3063,8 +3917,11 @@ static translation_unit_t *parse_translation_unit(void) 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; type_set_output(stderr); + ast_set_output(stderr); lookahead_bufpos = 0; for(int i = 0; i < MAX_LOOKAHEAD + 2; ++i) { @@ -3073,6 +3930,10 @@ translation_unit_t *parse(void) translation_unit_t *unit = parse_translation_unit(); DEL_ARR_F(environment_stack); + DEL_ARR_F(label_stack); + + if(found_error) + return NULL; return unit; } @@ -3082,12 +3943,16 @@ void init_parser(void) init_expression_parsers(); obstack_init(&temp_obst); - type_int = make_atomic_type(ATOMIC_TYPE_INT, 0); - type_double = make_atomic_type(ATOMIC_TYPE_DOUBLE, 0); - type_size_t = make_atomic_type(ATOMIC_TYPE_UINT, 0); - type_const_char = make_atomic_type(ATOMIC_TYPE_CHAR, TYPE_QUALIFIER_CONST); - type_void = make_atomic_type(ATOMIC_TYPE_VOID, 0); - type_string = make_pointer_type(type_const_char, 0); + type_int = make_atomic_type(ATOMIC_TYPE_INT, 0); + type_uint = make_atomic_type(ATOMIC_TYPE_UINT, 0); + type_long_double = make_atomic_type(ATOMIC_TYPE_LONG_DOUBLE, 0); + type_double = make_atomic_type(ATOMIC_TYPE_DOUBLE, 0); + type_float = make_atomic_type(ATOMIC_TYPE_FLOAT, 0); + type_size_t = make_atomic_type(ATOMIC_TYPE_ULONG, 0); + type_ptrdiff_t = make_atomic_type(ATOMIC_TYPE_LONG, 0); + type_const_char = make_atomic_type(ATOMIC_TYPE_CHAR, TYPE_QUALIFIER_CONST); + type_void = make_atomic_type(ATOMIC_TYPE_VOID, 0); + type_string = make_pointer_type(type_const_char, 0); } void exit_parser(void)