X-Git-Url: http://nsz.repo.hu/git/?a=blobdiff_plain;f=parser.c;h=b800fb8c94ec5fa6d5b1b22a02167f2da8787104;hb=582f473864187c81edea9cf95171876fb6b8c68e;hp=7aaa859db06333d0d41c22fbafaf23ccd96404b2;hpb=5668cfa66fad0283559dc1f9deaf2bd16440efe4;p=cparser diff --git a/parser.c b/parser.c index 7aaa859..b800fb8 100644 --- a/parser.c +++ b/parser.c @@ -105,15 +105,21 @@ typedef struct parse_initializer_env_t { typedef declaration_t* (*parsed_declaration_func) (declaration_t *declaration, bool is_definition); +/** The current token. */ static token_t token; +/** The lookahead ring-buffer. */ static token_t lookahead_buffer[MAX_LOOKAHEAD]; +/** Position of the next token in the lookahead buffer. */ static int lookahead_bufpos; static stack_entry_t *environment_stack = NULL; static stack_entry_t *label_stack = NULL; static stack_entry_t *local_label_stack = NULL; +/** The global file scope. */ static scope_t *global_scope = NULL; +/** The current scope. */ static scope_t *scope = NULL; static declaration_t *last_declaration = NULL; +/** Point to the current function declaration if inside a function. */ static declaration_t *current_function = NULL; static declaration_t *current_init_decl = NULL; static switch_statement_t *current_switch = NULL; @@ -124,9 +130,15 @@ static goto_statement_t *goto_first = NULL; static goto_statement_t *goto_last = NULL; static label_statement_t *label_first = NULL; static label_statement_t *label_last = NULL; +/** current translation unit. */ static translation_unit_t *unit = NULL; +/** true if we are in a type property context (evaluation only for type. */ +static bool in_type_prop = false; +/** true in we are in a __extension__ context. */ +static bool in_gcc_extension = false; static struct obstack temp_obst; + #define PUSH_PARENT(stmt) \ statement_t *const prev_parent = current_parent; \ current_parent = (stmt); @@ -134,6 +146,9 @@ static struct obstack temp_obst; static source_position_t null_position = { NULL, 0 }; +/** special symbol used for anonymous entities. */ +static const symbol_t *sym_anonymous = NULL; + /* symbols for Microsoft extended-decl-modifier */ static const symbol_t *sym_align = NULL; static const symbol_t *sym_allocate = NULL; @@ -160,6 +175,9 @@ static unsigned char token_anchor_set[T_LAST_TOKEN]; /** The current source position. */ #define HERE (&token.source_position) +/** true if we are in GCC mode. */ +#define GNU_MODE ((c_mode & _GNUC) || in_gcc_extension) + static type_t *type_valist; static statement_t *parse_compound_statement(bool inside_expression_statement); @@ -552,6 +570,7 @@ static void restore_anchor_state(int token_type, int count) static void rem_anchor_token(int token_type) { assert(0 <= token_type && token_type < T_LAST_TOKEN); + assert(token_anchor_set[token_type] != 0); --token_anchor_set[token_type]; } @@ -639,16 +658,6 @@ static void eat_block(void) next_token(); } -/** - * eat all token until a ';' is reached or a stop token is found. - */ -static void eat_statement(void) -{ - eat_until_matching_token(';'); - if (token.type == ';') - next_token(); -} - #define eat(token_type) do { assert(token.type == token_type); next_token(); } while (0) /** @@ -693,30 +702,39 @@ static void type_error_incompatible(const char *msg, * If not, generate an error, eat the current statement, * and goto the end_error label. */ -#define expect(expected) \ - do { \ - if (UNLIKELY(token.type != (expected))) { \ - parse_error_expected(NULL, (expected), NULL); \ - add_anchor_token(expected); \ - eat_until_anchor(); \ - if (token.type == expected) \ - next_token(); \ - rem_anchor_token(expected); \ - goto end_error; \ - } \ - next_token(); \ +#define expect(expected) \ + do { \ + if (UNLIKELY(token.type != (expected))) { \ + parse_error_expected(NULL, (expected), NULL); \ + add_anchor_token(expected); \ + eat_until_anchor(); \ + if (token.type == expected) \ + next_token(); \ + rem_anchor_token(expected); \ + goto end_error; \ + } \ + next_token(); \ } while (0) -static void set_scope(scope_t *new_scope) +static void scope_push(scope_t *new_scope) { if (scope != NULL) { scope->last_declaration = last_declaration; + new_scope->depth = scope->depth + 1; } - scope = new_scope; + new_scope->parent = scope; + scope = new_scope; last_declaration = new_scope->last_declaration; } +static void scope_pop(void) +{ + scope->last_declaration = last_declaration; + scope = scope->parent; + last_declaration = scope->last_declaration; +} + /** * Search a symbol in a given namespace and returns its declaration or * NULL if this symbol was not found. @@ -743,28 +761,20 @@ static void stack_push(stack_entry_t **stack_ptr, declaration_t *declaration) namespace_t namespc = (namespace_t) declaration->namespc; /* replace/add declaration into declaration list of the symbol */ - declaration_t *iter = symbol->declaration; - if (iter == NULL) { - symbol->declaration = declaration; - } else { - declaration_t *iter_last = NULL; - for( ; iter != NULL; iter_last = iter, iter = iter->symbol_next) { - /* replace an entry? */ - if (iter->namespc == namespc) { - if (iter_last == NULL) { - symbol->declaration = declaration; - } else { - iter_last->symbol_next = declaration; - } - declaration->symbol_next = iter->symbol_next; - break; - } - } - if (iter == NULL) { - assert(iter_last->symbol_next == NULL); - iter_last->symbol_next = declaration; + declaration_t **anchor; + declaration_t *iter; + for (anchor = &symbol->declaration;; anchor = &iter->symbol_next) { + iter = *anchor; + if (iter == NULL) + break; + + /* replace an entry? */ + if (iter->namespc == namespc) { + declaration->symbol_next = iter->symbol_next; + break; } } + *anchor = declaration; /* remember old declaration */ stack_entry_t entry; @@ -829,30 +839,20 @@ static void stack_pop_to(stack_entry_t **stack_ptr, size_t new_top) namespace_t namespc = (namespace_t)entry->namespc; /* replace/remove declaration */ - declaration_t *declaration = symbol->declaration; - assert(declaration != NULL); - if (declaration->namespc == namespc) { - if (old_declaration == NULL) { - symbol->declaration = declaration->symbol_next; - } else { - symbol->declaration = old_declaration; - } - } else { - declaration_t *iter_last = declaration; - declaration_t *iter = declaration->symbol_next; - for( ; iter != NULL; iter_last = iter, iter = iter->symbol_next) { - /* replace an entry? */ - if (iter->namespc == namespc) { - assert(iter_last != NULL); - iter_last->symbol_next = old_declaration; - if (old_declaration != NULL) { - old_declaration->symbol_next = iter->symbol_next; - } - break; - } - } + declaration_t **anchor; + declaration_t *iter; + for (anchor = &symbol->declaration;; anchor = &iter->symbol_next) { + iter = *anchor; assert(iter != NULL); + /* replace an entry? */ + if (iter->namespc == namespc) + break; } + + /* Because of scopes and appending other namespaces to the end of + * the list, this must hold. */ + assert((old_declaration != NULL ? old_declaration->symbol_next : NULL) == iter->symbol_next); + *anchor = old_declaration; } ARR_SHRINKLEN(*stack_ptr, (int) new_top); @@ -1497,12 +1497,12 @@ static void parse_gnu_attribute_format_args(gnu_attribute_t *attribute) add_anchor_token(','); parse_constant_expression(); rem_anchor_token(','); - rem_anchor_token('('); + rem_anchor_token(')'); expect(','); add_anchor_token(')'); parse_constant_expression(); - rem_anchor_token('('); + rem_anchor_token(')'); expect(')'); return; end_error: @@ -2471,6 +2471,10 @@ finish_designator: if (type == NULL) { /* we are already outside, ... */ + if (is_type_compound(outer_type) && + !outer_type->compound.declaration->init.complete) { + goto error_parse_next; + } goto error_excess; } @@ -2855,16 +2859,16 @@ static type_t *parse_typeof(void) expression_t *expression = NULL; -restart: - switch(token.type) { - case T___extension__: - /* This can be a prefix to a typename or an expression. We simply eat - * it now. */ - do { - next_token(); - } while (token.type == T___extension__); - goto restart; + bool old_type_prop = in_type_prop; + bool old_gcc_extension = in_gcc_extension; + in_type_prop = true; + while (token.type == T___extension__) { + /* This can be a prefix to a typename or an expression. */ + next_token(); + in_gcc_extension = true; + } + switch (token.type) { case T_IDENTIFIER: if (is_typedef_symbol(token.v.symbol)) { type = parse_typename(); @@ -2883,6 +2887,8 @@ restart: type = expression->base.type; break; } + in_type_prop = old_type_prop; + in_gcc_extension = old_gcc_extension; rem_anchor_token(')'); expect(')'); @@ -2955,7 +2961,7 @@ static bool check_alignment_value(long long intvalue) return false; } unsigned v = (unsigned)intvalue; - for(unsigned i = 1; i <= 8192; i += i) { + for (unsigned i = 1; i <= 8192; i += i) { if (i == v) return true; } @@ -3026,7 +3032,7 @@ static void parse_microsoft_extended_decl_modifier(declaration_specifiers_t *spe } else if (symbol == sym_property) { next_token(); expect('('); - for(;;) { + for (;;) { bool is_get = false; if (token.type != T_IDENTIFIER) goto end_error; @@ -3216,12 +3222,13 @@ static void finish_union_type(compound_type_t *type) { static void parse_declaration_specifiers(declaration_specifiers_t *specifiers) { - type_t *type = NULL; - type_qualifiers_t qualifiers = TYPE_QUALIFIER_NONE; - type_modifiers_t modifiers = TYPE_MODIFIER_NONE; - unsigned type_specifiers = 0; - bool newtype = false; - bool saw_error = false; + type_t *type = NULL; + type_qualifiers_t qualifiers = TYPE_QUALIFIER_NONE; + type_modifiers_t modifiers = TYPE_MODIFIER_NONE; + unsigned type_specifiers = 0; + bool newtype = false; + bool saw_error = false; + bool old_gcc_extension = in_gcc_extension; specifiers->source_position = token.source_position; @@ -3231,7 +3238,7 @@ static void parse_declaration_specifiers(declaration_specifiers_t *specifiers) if (specifiers->modifiers & DM_TRANSPARENT_UNION) modifiers |= TYPE_MODIFIER_TRANSPARENT_UNION; - switch(token.type) { + switch (token.type) { /* storage class */ #define MATCH_STORAGE_CLASS(token, class) \ @@ -3296,8 +3303,8 @@ static void parse_declaration_specifiers(declaration_specifiers_t *specifiers) MATCH_TYPE_QUALIFIER(T___sptr, TYPE_QUALIFIER_SPTR); case T___extension__: - /* TODO */ next_token(); + in_gcc_extension = true; break; /* type specifiers */ @@ -3442,6 +3449,8 @@ static void parse_declaration_specifiers(declaration_specifiers_t *specifiers) } finish_specifiers: + in_gcc_extension = old_gcc_extension; + if (type == NULL || (saw_error && type_specifiers != 0)) { atomic_type_kind_t atomic_type; @@ -3580,10 +3589,8 @@ warn_about_long_long: default: /* invalid specifier combination, give an error message */ if (type_specifiers == 0) { - if (saw_error) { - specifiers->type = type_error_type; - return; - } + if (saw_error) + goto end_error; if (!strict_mode) { if (warning.implicit_int) { @@ -3602,15 +3609,13 @@ warn_about_long_long: } else { errorf(HERE, "multiple datatypes in declaration"); } - atomic_type = ATOMIC_TYPE_INVALID; + goto end_error; } - if (type_specifiers & SPECIFIER_COMPLEX && - atomic_type != ATOMIC_TYPE_INVALID) { + if (type_specifiers & SPECIFIER_COMPLEX) { type = allocate_type_zero(TYPE_COMPLEX, &builtin_source_position); type->complex.akind = atomic_type; - } else if (type_specifiers & SPECIFIER_IMAGINARY && - atomic_type != ATOMIC_TYPE_INVALID) { + } else if (type_specifiers & SPECIFIER_IMAGINARY) { type = allocate_type_zero(TYPE_IMAGINARY, &builtin_source_position); type->imaginary.akind = atomic_type; } else { @@ -3633,7 +3638,10 @@ warn_about_long_long: } specifiers->type = result; + return; + end_error: + specifiers->type = type_error_type; return; } @@ -3660,7 +3668,13 @@ static type_qualifiers_t parse_type_qualifiers(void) } } -static declaration_t *parse_identifier_list(void) +/** + * Parses an K&R identifier list and return a list of declarations. + * + * @param last points to the last declaration in the list + * @return the list of declarations + */ +static declaration_t *parse_identifier_list(declaration_t **last) { declaration_t *declarations = NULL; declaration_t *last_declaration = NULL; @@ -3684,6 +3698,7 @@ static declaration_t *parse_identifier_list(void) next_token(); } while (token.type == T_IDENTIFIER); + *last = last_declaration; return declarations; } @@ -3736,7 +3751,13 @@ static declaration_t *parse_parameter(void) return declaration; } -static declaration_t *parse_parameters(function_type_t *type) +/** + * Parses a function type parameter list and return a list of declarations. + * + * @param last point to the last element of the list + * @return the parameter list + */ +static declaration_t *parse_parameters(function_type_t *type, declaration_t **last) { declaration_t *declarations = NULL; @@ -3749,7 +3770,7 @@ static declaration_t *parse_parameters(function_type_t *type) token_type_t la1_type = (token_type_t)look_ahead(1)->type; if (la1_type == ',' || la1_type == ')') { type->kr_style_parameters = true; - declarations = parse_identifier_list(); + declarations = parse_identifier_list(last); goto parameters_finished; } } @@ -3815,10 +3836,12 @@ parameters_finished: expect(')'); restore_anchor_state(',', saved_comma_state); + *last = last_declaration; return declarations; end_error: restore_anchor_state(',', saved_comma_state); + *last = NULL; return NULL; } @@ -3907,9 +3930,8 @@ static construct_type_t *parse_array_declarator(void) rem_anchor_token(']'); expect(']'); - return (construct_type_t*) array; end_error: - return NULL; + return (construct_type_t*) array; } static construct_type_t *parse_function_declarator(declaration_t *declaration) @@ -3955,9 +3977,12 @@ static construct_type_t *parse_function_declarator(declaration_t *declaration) type = allocate_type_zero(TYPE_FUNCTION, HERE); } - declaration_t *parameters = parse_parameters(&type->function); + declaration_t *last; + declaration_t *parameters = parse_parameters(&type->function, &last); if (declaration != NULL) { - declaration->scope.declarations = parameters; + declaration->scope.declarations = parameters; + declaration->scope.last_declaration = last; + declaration->scope.is_parameter = true; } construct_function_type_t *construct_function_type = @@ -4048,13 +4073,7 @@ static construct_type_t *parse_inner_declarator(declaration_t *declaration, if (may_be_abstract) break; parse_error_expected("while parsing declarator", T_IDENTIFIER, '(', NULL); - /* avoid a loop in the outermost scope, because eat_statement doesn't - * eat '}' */ - if (token.type == '}' && current_function == NULL) { - next_token(); - } else { - eat_statement(); - } + eat_until_anchor(); return NULL; } @@ -4153,15 +4172,19 @@ static type_t *construct_declarator_type(construct_type_t *construct_list, function_type->function.return_type = type; type_t *skipped_return_type = skip_typeref(type); + /* §6.7.5.3(1) */ if (is_type_function(skipped_return_type)) { errorf(HERE, "function returning function is not allowed"); - type = type_error_type; } else if (is_type_array(skipped_return_type)) { errorf(HERE, "function returning array is not allowed"); - type = type_error_type; } else { - type = function_type; + if (skipped_return_type->base.qualifiers != 0) { + warningf(HERE, + "type qualifiers in return type of function type are meaningless"); + } } + + type = function_type; break; } @@ -4202,12 +4225,13 @@ static type_t *construct_declarator_type(construct_type_t *construct_list, } type_t *skipped_type = skip_typeref(type); - if (is_type_atomic(skipped_type, ATOMIC_TYPE_VOID)) { - errorf(HERE, "array of void is not allowed"); - type = type_error_type; - } else { - type = array_type; + /* §6.7.5.2(1) */ + if (is_type_incomplete(skipped_type)) { + errorf(HERE, "array of incomplete type '%T' is not allowed", type); + } else if (is_type_function(skipped_type)) { + errorf(HERE, "array of functions is not allowed"); } + type = array_type; break; } } @@ -4304,7 +4328,7 @@ static void check_type_of_main(const declaration_t *const decl, const function_t warningf(&decl->source_position, "'main' is normally a non-static function"); } - if (skip_typeref(func_type->return_type) != type_int) { + if (!types_compatible(skip_typeref(func_type->return_type), type_int)) { warningf(&decl->source_position, "return type of 'main' should be 'int', but is '%T'", func_type->return_type); @@ -4367,7 +4391,7 @@ static declaration_t *record_declaration( previous_declaration == NULL) { warningf(&declaration->source_position, "function declaration '%#T' is not a prototype", - orig_type, declaration->symbol); + orig_type, symbol); } if (warning.main && is_type_function(type) && is_sym_main(symbol)) { @@ -4382,8 +4406,17 @@ static declaration_t *record_declaration( } assert(declaration != previous_declaration); - if (previous_declaration != NULL - && previous_declaration->parent_scope == scope) { + if (previous_declaration != NULL && + previous_declaration->parent_scope->is_parameter && + scope->depth == previous_declaration->parent_scope->depth + 1) { + errorf(&declaration->source_position, + "declaration '%#T' redeclares the parameter '%#T' (declared %P)", + orig_type, symbol, previous_declaration->type, symbol, + &previous_declaration->source_position); + goto finish; + } + if (previous_declaration != NULL && + previous_declaration->parent_scope == scope) { /* can happen for K&R style declarations */ if (previous_declaration->type == NULL) { previous_declaration->type = declaration->type; @@ -4515,7 +4548,7 @@ warn_redundant_declaration: "no previous declaration for '%#T'", orig_type, symbol); } } - +finish: assert(declaration->parent_scope == NULL); assert(scope != NULL); @@ -4535,7 +4568,7 @@ static void parser_error_multiple_definition(declaration_t *declaration, static bool is_declaration_specifier(const token_t *token, bool only_specifiers_qualifiers) { - switch(token->type) { + switch (token->type) { TYPE_SPECIFIERS TYPE_QUALIFIERS return true; @@ -4598,24 +4631,17 @@ static void parse_anonymous_declaration_rest( { eat(';'); - declaration_t *const declaration = allocate_declaration_zero(); - declaration->type = specifiers->type; - declaration->declared_storage_class = specifiers->declared_storage_class; - declaration->source_position = specifiers->source_position; - declaration->modifiers = specifiers->modifiers; - - if (declaration->declared_storage_class != STORAGE_CLASS_NONE) { - warningf(&declaration->source_position, + if (specifiers->declared_storage_class != STORAGE_CLASS_NONE) { + warningf(&specifiers->source_position, "useless storage class in empty declaration"); } - declaration->storage_class = STORAGE_CLASS_NONE; - type_t *type = declaration->type; + type_t *type = specifiers->type; switch (type->kind) { case TYPE_COMPOUND_STRUCT: case TYPE_COMPOUND_UNION: { if (type->compound.declaration->symbol == NULL) { - warningf(&declaration->source_position, + warningf(&specifiers->source_position, "unnamed struct/union that defines no instances"); } break; @@ -4625,11 +4651,20 @@ static void parse_anonymous_declaration_rest( break; default: - warningf(&declaration->source_position, "empty declaration"); + warningf(&specifiers->source_position, "empty declaration"); break; } +#ifdef RECORD_EMPTY_DECLARATIONS + declaration_t *const declaration = allocate_declaration_zero(); + declaration->type = specifiers->type; + declaration->declared_storage_class = specifiers->declared_storage_class; + declaration->source_position = specifiers->source_position; + declaration->modifiers = specifiers->modifiers; + declaration->storage_class = STORAGE_CLASS_NONE; + append_declaration(declaration); +#endif } static void parse_declaration_rest(declaration_t *ndeclaration, @@ -4637,7 +4672,6 @@ static void parse_declaration_rest(declaration_t *ndeclaration, parsed_declaration_func finished_declaration) { add_anchor_token(';'); - add_anchor_token('='); add_anchor_token(','); while(true) { declaration_t *declaration = @@ -4661,13 +4695,14 @@ static void parse_declaration_rest(declaration_t *ndeclaration, break; eat(','); + add_anchor_token('='); ndeclaration = parse_declarator(specifiers, /*may_be_abstract=*/false); + rem_anchor_token('='); } expect(';'); end_error: rem_anchor_token(';'); - rem_anchor_token('='); rem_anchor_token(','); } @@ -4710,7 +4745,10 @@ static void parse_declaration(parsed_declaration_func finished_declaration) { declaration_specifiers_t specifiers; memset(&specifiers, 0, sizeof(specifiers)); + + add_anchor_token(';'); parse_declaration_specifiers(&specifiers); + rem_anchor_token(';'); if (token.type == ';') { parse_anonymous_declaration_rest(&specifiers); @@ -4743,10 +4781,11 @@ static void parse_kr_declaration_list(declaration_t *declaration) if (!type->function.kr_style_parameters) return; + add_anchor_token('{'); + /* push function parameters */ - int top = environment_top(); - scope_t *last_scope = scope; - set_scope(&declaration->scope); + size_t const top = environment_top(); + scope_push(&declaration->scope); declaration_t *parameter = declaration->scope.declarations; for ( ; parameter != NULL; parameter = parameter->next) { @@ -4762,7 +4801,7 @@ static void parse_kr_declaration_list(declaration_t *declaration) /* pop function parameters */ assert(scope == &declaration->scope); - set_scope(last_scope); + scope_pop(); environment_pop_to(top); /* update function type */ @@ -4821,6 +4860,8 @@ static void parse_kr_declaration_list(declaration_t *declaration) } declaration->type = type; + + rem_anchor_token('{'); } static bool first_err = true; @@ -5380,14 +5421,16 @@ static void parse_external_declaration(void) add_anchor_token(','); add_anchor_token('='); - rem_anchor_token(';'); + add_anchor_token(';'); + add_anchor_token('{'); /* declarator is common to both function-definitions and declarations */ declaration_t *ndeclaration = parse_declarator(&specifiers, /*may_be_abstract=*/false); - rem_anchor_token(','); - rem_anchor_token('='); + rem_anchor_token('{'); rem_anchor_token(';'); + rem_anchor_token('='); + rem_anchor_token(','); /* must be a declaration */ switch (token.type) { @@ -5456,9 +5499,8 @@ static void parse_external_declaration(void) type = skip_typeref(declaration->type); /* push function parameters and switch scope */ - int top = environment_top(); - scope_t *last_scope = scope; - set_scope(&declaration->scope); + size_t const top = environment_top(); + scope_push(&declaration->scope); declaration_t *parameter = declaration->scope.declarations; for( ; parameter != NULL; parameter = parameter->next) { @@ -5513,20 +5555,44 @@ static void parse_external_declaration(void) } assert(scope == &declaration->scope); - set_scope(last_scope); + scope_pop(); environment_pop_to(top); } static type_t *make_bitfield_type(type_t *base_type, expression_t *size, - source_position_t *source_position) + source_position_t *source_position, + const symbol_t *symbol) { type_t *type = allocate_type_zero(TYPE_BITFIELD, source_position); type->bitfield.base_type = base_type; type->bitfield.size_expression = size; + il_size_t bit_size; + type_t *skipped_type = skip_typeref(base_type); + if (!is_type_integer(skipped_type)) { + errorf(HERE, "bitfield base type '%T' is not an integer type", + base_type); + bit_size = 0; + } else { + bit_size = skipped_type->base.size * 8; + } + if (is_constant_expression(size)) { - type->bitfield.bit_size = fold_constant(size); + long v = fold_constant(size); + + if (v < 0) { + errorf(source_position, "negative width in bit-field '%Y'", + symbol); + } else if (v == 0) { + errorf(source_position, "zero width for bit-field '%Y'", + symbol); + } else if (bit_size > 0 && (il_size_t)v > bit_size) { + errorf(source_position, "width of '%Y' exceeds its type", + symbol); + } else { + type->bitfield.bit_size = v; + } } return type; @@ -5579,12 +5645,8 @@ static void parse_compound_declarators(declaration_t *struct_declaration, type_t *base_type = specifiers->type; expression_t *size = parse_constant_expression(); - if (!is_type_integer(skip_typeref(base_type))) { - errorf(HERE, "bitfield base type '%T' is not an integer type", - base_type); - } - - type_t *type = make_bitfield_type(base_type, size, &source_position); + type_t *type = make_bitfield_type(base_type, size, + &source_position, sym_anonymous); declaration = allocate_declaration_zero(); declaration->namespc = NAMESPACE_NORMAL; @@ -5604,12 +5666,8 @@ static void parse_compound_declarators(declaration_t *struct_declaration, next_token(); expression_t *size = parse_constant_expression(); - if (!is_type_integer(type)) { - errorf(HERE, "bitfield base type '%T' is not an integer type", - orig_type); - } - - type_t *bitfield_type = make_bitfield_type(orig_type, size, &source_position); + type_t *bitfield_type = make_bitfield_type(orig_type, size, + &source_position, declaration->symbol); declaration->type = bitfield_type; } else { /* TODO we ignore arrays for now... what is missing is a check @@ -5802,8 +5860,7 @@ static expression_t *parse_character_constant(void) cnst->conste.v.character = token.v.string; if (cnst->conste.v.character.size != 1) { - if (warning.multichar && (c_mode & _GNUC)) { - /* TODO */ + if (warning.multichar && GNU_MODE) { warningf(HERE, "multi-character character constant"); } else { errorf(HERE, "more than 1 characters in character constant"); @@ -5826,8 +5883,7 @@ static expression_t *parse_wide_character_constant(void) cnst->conste.v.wide_character = token.v.wide_string; if (cnst->conste.v.wide_character.size != 1) { - if (warning.multichar && (c_mode & _GNUC)) { - /* TODO */ + if (warning.multichar && GNU_MODE) { warningf(HERE, "multi-character character constant"); } else { errorf(HERE, "more than 1 characters in character constant"); @@ -6075,11 +6131,11 @@ static expression_t *parse_reference(void) } } - type_t *type = declaration->type; + type_t *orig_type = declaration->type; /* we always do the auto-type conversions; the & and sizeof parser contains * code to revert this! */ - type = automatic_type_conversion(type); + type_t *type = automatic_type_conversion(orig_type); ref->declaration = declaration; ref->base.type = type; @@ -6087,6 +6143,15 @@ static expression_t *parse_reference(void) /* this declaration is used */ declaration->used = true; + if (declaration->parent_scope != global_scope && + declaration->parent_scope->depth < current_function->scope.depth && + is_type_valid(orig_type) && !is_type_function(orig_type)) { + /* access of a variable from an outer function */ + declaration->address_taken = true; + ref->is_outer_ref = true; + current_function->need_closure = true; + } + /* check for deprecated functions */ if (warning.deprecated_declarations && declaration->modifiers & DM_DEPRECATED) { @@ -6102,7 +6167,7 @@ static expression_t *parse_reference(void) declaration->symbol, &declaration->source_position); } } - if (warning.init_self && declaration == current_init_decl) { + if (warning.init_self && declaration == current_init_decl && !in_type_prop) { current_init_decl = NULL; warningf(HERE, "variable '%#T' is initialized by itself", declaration->type, declaration->symbol); @@ -6183,11 +6248,12 @@ static expression_t *parse_compound_literal(type_t *type) */ static expression_t *parse_cast(void) { + add_anchor_token(')'); + source_position_t source_position = token.source_position; type_t *type = parse_typename(); - /* matching add_anchor_token() is at call site */ rem_anchor_token(')'); expect(')'); @@ -6216,6 +6282,8 @@ end_error: */ static expression_t *parse_statement_expression(void) { + add_anchor_token(')'); + expression_t *expression = allocate_expression_zero(EXPR_STATEMENT); statement_t *statement = parse_compound_statement(true); @@ -6237,11 +6305,11 @@ static expression_t *parse_statement_expression(void) } expression->base.type = type; + rem_anchor_token(')'); expect(')'); - return expression; end_error: - return create_invalid_expression(); + return expression; } /** @@ -6250,7 +6318,6 @@ end_error: static expression_t *parse_parenthesized_expression(void) { eat('('); - add_anchor_token(')'); switch(token.type) { case '{': @@ -6266,13 +6333,13 @@ static expression_t *parse_parenthesized_expression(void) } } + add_anchor_token(')'); expression_t *result = parse_expression(); rem_anchor_token(')'); expect(')'); - return result; end_error: - return create_invalid_expression(); + return result; } static expression_t *parse_function_keyword(void) @@ -6449,17 +6516,15 @@ static expression_t *parse_va_start(void) expression_t *const expr = parse_assignment_expression(); if (expr->kind == EXPR_REFERENCE) { declaration_t *const decl = expr->reference.declaration; - if (decl == NULL) - return create_invalid_expression(); - if (decl->parent_scope == ¤t_function->scope && - decl->next == NULL) { - expression->va_starte.parameter = decl; - expect(')'); - return expression; + if (decl->parent_scope != ¤t_function->scope || decl->next != NULL) { + errorf(&expr->base.source_position, + "second argument of 'va_start' must be last parameter of the current function"); } + 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"); + expect(')'); end_error: return create_invalid_expression(); } @@ -6668,7 +6733,10 @@ static declaration_t *get_label(symbol_t *symbol) candidate = get_declaration(symbol, NAMESPACE_LOCAL_LABEL); /* if we found a local label, we already created the declaration */ if (candidate != NULL) { - assert(candidate->parent_scope == scope); + if (candidate->parent_scope != scope) { + assert(candidate->parent_scope->depth < scope->depth); + current_function->goto_to_outer = true; + } return candidate; } @@ -6798,7 +6866,7 @@ static expression_t *parse_primary_expression(void) case T___builtin_prefetch: return parse_builtin_prefetch(); case T__assume: return parse_assume(); case T_ANDAND: - if (c_mode & _GNUC) + if (GNU_MODE) return parse_label_address(); break; @@ -6883,12 +6951,15 @@ static expression_t *parse_typeprop(expression_kind_t const kind, source_position_t const pos, unsigned const precedence) { - expression_t *tp_expression = allocate_expression_zero(kind); + expression_t *tp_expression = allocate_expression_zero(kind); tp_expression->base.type = type_size_t; tp_expression->base.source_position = pos; char const* const what = kind == EXPR_SIZEOF ? "sizeof" : "alignof"; + /* we only refer to a type property, mark this case */ + bool old = in_type_prop; + in_type_prop = true; if (token.type == '(' && is_declaration_specifier(look_ahead(1), true)) { next_token(); add_anchor_token(')'); @@ -6928,9 +6999,9 @@ static expression_t *parse_typeprop(expression_kind_t const kind, tp_expression->typeprop.tp_expression = expression; } - return tp_expression; end_error: - return create_invalid_expression(); + in_type_prop = old; + return tp_expression; } static expression_t *parse_sizeof(unsigned precedence) @@ -7186,9 +7257,8 @@ static expression_t *parse_call_expression(unsigned precedence, "function call has aggregate value"); } - return result; end_error: - return create_invalid_expression(); + return result; } static type_t *semantic_arithmetic(type_t *type_left, type_t *type_right); @@ -7228,7 +7298,7 @@ static expression_t *parse_conditional_expression(unsigned precedence, expression_t *true_expression = expression; bool gnu_cond = false; - if ((c_mode & _GNUC) && token.type == ':') { + if (GNU_MODE && token.type == ':') { gnu_cond = true; } else true_expression = parse_expression(); @@ -7339,9 +7409,10 @@ static expression_t *parse_extension(unsigned precedence) { eat(T___extension__); - /* TODO enable extensions */ + bool old_gcc_extension = in_gcc_extension; + in_gcc_extension = true; expression_t *expression = parse_sub_expression(precedence); - /* TODO disable extensions */ + in_gcc_extension = old_gcc_extension; return expression; } @@ -7375,7 +7446,7 @@ static bool check_pointer_arithmetic(const source_position_t *source_position, points_to = skip_typeref(points_to); if (is_type_incomplete(points_to)) { - if (!(c_mode & _GNUC) || !is_type_atomic(points_to, ATOMIC_TYPE_VOID)) { + if (!GNU_MODE || !is_type_atomic(points_to, ATOMIC_TYPE_VOID)) { errorf(source_position, "arithmetic with pointer to incomplete type '%T' not allowed", orig_pointer_type); @@ -7386,7 +7457,7 @@ static bool check_pointer_arithmetic(const source_position_t *source_position, orig_pointer_type); } } else if (is_type_function(points_to)) { - if (!(c_mode && _GNUC)) { + if (!GNU_MODE) { errorf(source_position, "arithmetic with pointer to function type '%T' not allowed", orig_pointer_type); @@ -7460,6 +7531,46 @@ static void semantic_unexpr_plus(unary_expression_t *expression) "traditional C rejects the unary plus operator"); } +static expression_t const *get_reference_address(expression_t const *expr) +{ + bool regular_take_address = true; + for (;;) { + if (expr->kind == EXPR_UNARY_TAKE_ADDRESS) { + expr = expr->unary.value; + } else { + regular_take_address = false; + } + + if (expr->kind != EXPR_UNARY_DEREFERENCE) + break; + + expr = expr->unary.value; + } + + if (expr->kind != EXPR_REFERENCE) + return NULL; + + if (!regular_take_address && + !is_type_function(skip_typeref(expr->reference.declaration->type))) { + return NULL; + } + + return expr; +} + +static void warn_function_address_as_bool(expression_t const* expr) +{ + if (!warning.address) + return; + + expr = get_reference_address(expr); + if (expr != NULL) { + warningf(&expr->base.source_position, + "the address of '%Y' will always evaluate as 'true'", + expr->reference.declaration->symbol); + } +} + static void semantic_not(unary_expression_t *expression) { type_t *const orig_type = expression->value->base.type; @@ -7469,6 +7580,8 @@ static void semantic_not(unary_expression_t *expression) "operand of ! must be of scalar type"); } + warn_function_address_as_bool(expression->value); + expression->base.type = type_int; } @@ -7816,6 +7929,22 @@ static void semantic_sub(binary_expression_t *expression) } } +static void warn_string_literal_address(expression_t const* expr) +{ + while (expr->kind == EXPR_UNARY_TAKE_ADDRESS) { + expr = expr->unary.value; + if (expr->kind != EXPR_UNARY_DEREFERENCE) + return; + expr = expr->unary.value; + } + + if (expr->kind == EXPR_STRING_LITERAL || + expr->kind == EXPR_WIDE_STRING_LITERAL) { + warningf(&expr->base.source_position, + "comparison with string literal results in unspecified behaviour"); + } +} + /** * Check the semantics of comparison expressions. * @@ -7823,13 +7952,32 @@ static void semantic_sub(binary_expression_t *expression) */ static void semantic_comparison(binary_expression_t *expression) { - expression_t *left = expression->left; - expression_t *right = expression->right; - type_t *orig_type_left = left->base.type; - type_t *orig_type_right = right->base.type; + expression_t *left = expression->left; + expression_t *right = expression->right; - type_t *type_left = skip_typeref(orig_type_left); - type_t *type_right = skip_typeref(orig_type_right); + if (warning.address) { + warn_string_literal_address(left); + warn_string_literal_address(right); + + expression_t const* const func_left = get_reference_address(left); + if (func_left != NULL && is_null_pointer_constant(right)) { + warningf(&expression->base.source_position, + "the address of '%Y' will never be NULL", + func_left->reference.declaration->symbol); + } + + expression_t const* const func_right = get_reference_address(right); + if (func_right != NULL && is_null_pointer_constant(right)) { + warningf(&expression->base.source_position, + "the address of '%Y' will never be NULL", + func_right->reference.declaration->symbol); + } + } + + type_t *orig_type_left = left->base.type; + type_t *orig_type_right = right->base.type; + 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)) { @@ -8022,6 +8170,9 @@ static void semantic_logical_op(binary_expression_t *expression) type_t *const type_left = skip_typeref(orig_type_left); type_t *const type_right = skip_typeref(orig_type_right); + warn_function_address_as_bool(left); + warn_function_address_as_bool(right); + if (!is_type_scalar(type_left) || !is_type_scalar(type_right)) { /* TODO: improve error message */ if (is_type_valid(type_left) && is_type_valid(type_right)) { @@ -8645,7 +8796,7 @@ static statement_t *parse_case_statement(void) statement->case_label.last_case = val; } - if (c_mode & _GNUC) { + if (GNU_MODE) { if (token.type == T_DOTDOTDOT) { next_token(); expression_t *const end_range = parse_expression(); @@ -8663,7 +8814,7 @@ static statement_t *parse_case_statement(void) statement->case_label.last_case = val; if (val < statement->case_label.first_case) { - statement->case_label.is_empty = true; + statement->case_label.is_empty_range = true; warningf(pos, "empty range specified"); } } @@ -8679,7 +8830,7 @@ static statement_t *parse_case_statement(void) /* Check for duplicate case values */ case_label_statement_t *c = &statement->case_label; for (case_label_statement_t *l = current_switch->first_case; l != NULL; l = l->next) { - if (l->is_bad || l->is_empty || l->expression == NULL) + if (l->is_bad || l->is_empty_range || l->expression == NULL) continue; if (c->last_case < l->first_case || c->first_case > l->last_case) @@ -9033,9 +9184,8 @@ static statement_t *parse_for(void) PUSH_PARENT(statement); - int top = environment_top(); - scope_t *last_scope = scope; - set_scope(&statement->fors.scope); + size_t const top = environment_top(); + scope_push(&statement->fors.scope); expect('('); add_anchor_token(')'); @@ -9077,7 +9227,7 @@ static statement_t *parse_for(void) statement->fors.body = parse_loop_body(statement); assert(scope == &statement->fors.scope); - set_scope(last_scope); + scope_pop(); environment_pop_to(top); POP_PARENT; @@ -9087,7 +9237,7 @@ end_error: POP_PARENT; rem_anchor_token(')'); assert(scope == &statement->fors.scope); - set_scope(last_scope); + scope_pop(); environment_pop_to(top); return create_invalid_statement(); @@ -9102,7 +9252,7 @@ static statement_t *parse_goto(void) eat(T_goto); statement_t *statement; - if (c_mode & _GNUC && token.type == '*') { + if (GNU_MODE && token.type == '*') { next_token(); expression_t *expression = parse_expression(); @@ -9125,11 +9275,11 @@ static statement_t *parse_goto(void) statement->gotos.expression = expression; } else { if (token.type != T_IDENTIFIER) { - if (c_mode & _GNUC) + if (GNU_MODE) parse_error_expected("while parsing goto", T_IDENTIFIER, '*', NULL); else parse_error_expected("while parsing goto", T_IDENTIFIER, NULL); - eat_statement(); + eat_until_anchor(); goto end_error; } symbol_t *symbol = token.v.symbol; @@ -9138,6 +9288,10 @@ static statement_t *parse_goto(void) statement = allocate_statement_zero(STATEMENT_GOTO); statement->base.source_position = source_position; statement->gotos.label = get_label(symbol); + + if (statement->gotos.label->parent_scope->depth < current_function->scope.depth) { + statement->gotos.outer_fkt_jmp = true; + } } /* remember the goto's in a list for later checking */ @@ -9337,7 +9491,10 @@ static statement_t *parse_declaration_statement(void) statement->base.source_position = token.source_position; declaration_t *before = last_declaration; - parse_declaration(record_declaration); + if (GNU_MODE) + parse_external_declaration(); + else + parse_declaration(record_declaration); if (before == NULL) { statement->declaration.declarations_begin = scope->declarations; @@ -9515,7 +9672,10 @@ expression_statment: do { next_token(); } while (token.type == T___extension__); + bool old_gcc_extension = in_gcc_extension; + in_gcc_extension = true; statement = parse_statement(); + in_gcc_extension = old_gcc_extension; break; DECLARATION_START @@ -9528,7 +9688,7 @@ expression_statment: case ';': statement = parse_empty_statement(); break; case '{': statement = parse_compound_statement(false); break; - case T___leave: statement = parse_leave_statement(); break; + case T___leave: statement = parse_leave_statement(); break; case T___try: statement = parse_ms_try_statment(); break; case T_asm: statement = parse_asm_statement(); break; case T_break: statement = parse_break(); break; @@ -9542,7 +9702,58 @@ expression_statment: case T_return: statement = parse_return(); break; case T_switch: statement = parse_switch(); break; case T_while: statement = parse_while(); break; - default: statement = parse_expression_statement(); break; + + case '!': + case '&': + case '(': + case '*': + case '+': + case '-': + case '~': + case T_ANDAND: + case T_CHARACTER_CONSTANT: + case T_FLOATINGPOINT: + case T_INTEGER: + case T_MINUSMINUS: + case T_PLUSPLUS: + case T_STRING_LITERAL: + case T_WIDE_CHARACTER_CONSTANT: + case T_WIDE_STRING_LITERAL: + case T___FUNCDNAME__: + case T___FUNCSIG__: + case T___FUNCTION__: + case T___PRETTY_FUNCTION__: + case T___builtin_alloca: + case T___builtin_classify_type: + case T___builtin_constant_p: + case T___builtin_expect: + case T___builtin_huge_val: + case T___builtin_isgreater: + case T___builtin_isgreaterequal: + case T___builtin_isless: + case T___builtin_islessequal: + case T___builtin_islessgreater: + case T___builtin_isunordered: + case T___builtin_nan: + case T___builtin_nand: + case T___builtin_nanf: + case T___builtin_offsetof: + case T___builtin_prefetch: + case T___builtin_va_arg: + case T___builtin_va_end: + case T___builtin_va_start: + case T___func__: + case T___noop: + case T__assume: + statement = parse_expression_statement(); + break; + + default: + errorf(HERE, "unexpected token %K while parsing statement", &token); + statement = create_invalid_statement(); + if (!at_anchor()) + next_token(); + break; } rem_anchor_token(';'); @@ -9586,10 +9797,9 @@ static statement_t *parse_compound_statement(bool inside_expression_statement) eat('{'); add_anchor_token('}'); - int top = environment_top(); - int top_local = local_label_top(); - scope_t *last_scope = scope; - set_scope(&statement->compound.scope); + size_t const top = environment_top(); + size_t const top_local = local_label_top(); + scope_push(&statement->compound.scope); statement_t **anchor = &statement->compound.statements; bool only_decls_so_far = true; @@ -9648,7 +9858,7 @@ static statement_t *parse_compound_statement(bool inside_expression_statement) end_error: rem_anchor_token('}'); assert(scope == &statement->compound.scope); - set_scope(last_scope); + scope_pop(); environment_pop_to(top); local_label_pop_to(top_local); @@ -9739,32 +9949,52 @@ end_error:; */ static void parse_translation_unit(void) { - for (;;) switch (token.type) { - DECLARATION_START - case T_IDENTIFIER: - case T___extension__: - parse_external_declaration(); - break; + for (;;) { +#ifndef NDEBUG + bool anchor_leak = false; + for (int i = 0; i != T_LAST_TOKEN; ++i) { + unsigned char count = token_anchor_set[i]; + if (count != 0) { + errorf(HERE, "Leaked anchor token %k %d times", i, count); + anchor_leak = true; + } + } + if (in_gcc_extension) { + errorf(HERE, "Leaked __extension__"); + anchor_leak = true; + } - case T_asm: - parse_global_asm(); - break; + if (anchor_leak) + abort(); +#endif - case T_EOF: - return; + switch (token.type) { + DECLARATION_START + case T_IDENTIFIER: + case T___extension__: + parse_external_declaration(); + break; - case ';': - /* TODO error in strict mode */ - warningf(HERE, "stray ';' outside of function"); - next_token(); - break; + case T_asm: + parse_global_asm(); + break; - default: - errorf(HERE, "stray %K outside of function", &token); - if (token.type == '(' || token.type == '{' || token.type == '[') - eat_until_matching_token(token.type); - next_token(); - break; + case T_EOF: + return; + + case ';': + /* TODO error in strict mode */ + warningf(HERE, "stray ';' outside of function"); + next_token(); + break; + + default: + errorf(HERE, "stray %K outside of function", &token); + if (token.type == '(' || token.type == '{' || token.type == '[') + eat_until_matching_token(token.type); + next_token(); + break; + } } } @@ -9792,15 +10022,16 @@ void start_parsing(void) global_scope = &unit->scope; assert(scope == NULL); - set_scope(&unit->scope); + scope_push(&unit->scope); initialize_builtin_types(); } translation_unit_t *finish_parsing(void) { + /* do NOT use scope_pop() here, this will crash, will it by hand */ assert(scope == &unit->scope); - scope = NULL; + scope = NULL; last_declaration = NULL; assert(global_scope == &unit->scope); @@ -9830,6 +10061,8 @@ void parse(void) */ void init_parser(void) { + sym_anonymous = symbol_table_insert(""); + if (c_mode & _MS) { /* add predefined symbols for extended-decl-modifier */ sym_align = symbol_table_insert("align");