X-Git-Url: http://nsz.repo.hu/git/?a=blobdiff_plain;f=parser.c;h=8639a7a377015963aa0edd265e1e9d6cc89fd704;hb=516b4ed5ac675f2d94becf9fa0815ddc65f637ac;hp=b4bf38b121fb549c2e1c8997ebdf3184d1f889c5;hpb=f1328c63a778a4b0227ced90b16ed71426d93f6d;p=cparser diff --git a/parser.c b/parser.c index b4bf38b..8639a7a 100644 --- a/parser.c +++ b/parser.c @@ -20,6 +20,7 @@ #include #include +#include #include #include @@ -138,9 +139,6 @@ static elf_visibility_tag_t default_visibility = ELF_VISIBILITY_DEFAULT; #define POP_EXTENSION() \ ((void)(in_gcc_extension = old_gcc_extension)) -/** special symbol used for anonymous entities. */ -static symbol_t *sym_anonymous = NULL; - /** The token anchor set */ static unsigned short token_anchor_set[T_LAST_TOKEN]; @@ -247,8 +245,7 @@ static void semantic_comparison(binary_expression_t *expression); case '~': \ case T_ANDAND: \ case T_CHARACTER_CONSTANT: \ - case T_FLOATINGPOINT: \ - case T_INTEGER: \ + case T_NUMBER: \ case T_MINUSMINUS: \ case T_PLUSPLUS: \ case T_STRING_LITERAL: \ @@ -427,8 +424,7 @@ static size_t get_initializer_size(initializer_kind_t kind) { static const size_t sizes[] = { [INITIALIZER_VALUE] = sizeof(initializer_value_t), - [INITIALIZER_STRING] = sizeof(initializer_string_t), - [INITIALIZER_WIDE_STRING] = sizeof(initializer_wide_string_t), + [INITIALIZER_STRING] = sizeof(initializer_value_t), [INITIALIZER_LIST] = sizeof(initializer_list_t), [INITIALIZER_DESIGNATOR] = sizeof(initializer_designator_t) }; @@ -482,12 +478,17 @@ static inline void next_token(void) #endif } -#define eat(token_kind) (assert(token.kind == (token_kind)), next_token()) +static inline void eat(token_kind_t const kind) +{ + assert(token.kind == kind); + (void)kind; + next_token(); +} -static inline bool next_if(token_kind_t const type) +static inline bool next_if(token_kind_t const kind) { - if (token.kind == type) { - eat(type); + if (token.kind == kind) { + eat(kind); return true; } else { return false; @@ -1039,51 +1040,49 @@ static void append_string(string_t const *const s) obstack_grow(&ast_obstack, s->begin, s->size); } -static string_t finish_string(void) +static string_t finish_string(string_encoding_t const enc) { obstack_1grow(&ast_obstack, '\0'); size_t const size = obstack_object_size(&ast_obstack) - 1; char const *const string = obstack_finish(&ast_obstack); - return (string_t){ string, size }; + return (string_t){ string, size, enc }; } -static string_t concat_string_literals(string_encoding_t *const out_enc) +static string_t concat_string_literals(void) { assert(token.kind == T_STRING_LITERAL); - string_t result; - string_encoding_t enc = token.string.encoding; + string_t result; if (look_ahead(1)->kind == T_STRING_LITERAL) { - append_string(&token.string.string); + append_string(&token.literal.string); eat(T_STRING_LITERAL); warningf(WARN_TRADITIONAL, HERE, "traditional C rejects string constant concatenation"); + string_encoding_t enc = token.literal.string.encoding; do { - if (token.string.encoding != STRING_ENCODING_CHAR) { - enc = token.string.encoding; + if (token.literal.string.encoding != STRING_ENCODING_CHAR) { + enc = token.literal.string.encoding; } - append_string(&token.string.string); + append_string(&token.literal.string); eat(T_STRING_LITERAL); } while (token.kind == T_STRING_LITERAL); - result = finish_string(); + result = finish_string(enc); } else { - result = token.string.string; + result = token.literal.string; eat(T_STRING_LITERAL); } - *out_enc = enc; return result; } static string_t parse_string_literals(char const *const context) { if (!skip_till(T_STRING_LITERAL, context)) - return (string_t){ "", 0 }; + return (string_t){ "", 0, STRING_ENCODING_CHAR }; - string_encoding_t enc; source_position_t const pos = *HERE; - string_t const res = concat_string_literals(&enc); + string_t const res = concat_string_literals(); - if (enc != STRING_ENCODING_CHAR) { + if (res.encoding != STRING_ENCODING_CHAR) { errorf(&pos, "expected plain string literal, got wide string literal"); } @@ -1544,31 +1543,6 @@ static designator_t *parse_designation(void) } } -static initializer_t *initializer_from_string(array_type_t *const type, - const string_t *const string) -{ - /* TODO: check len vs. size of array type */ - (void) type; - - initializer_t *initializer = allocate_initializer_zero(INITIALIZER_STRING); - initializer->string.string = *string; - - return initializer; -} - -static initializer_t *initializer_from_wide_string(array_type_t *const type, - const string_t *const string) -{ - /* TODO: check len vs. size of array type */ - (void) type; - - initializer_t *const initializer = - allocate_initializer_zero(INITIALIZER_WIDE_STRING); - initializer->wide_string.string = *string; - - return initializer; -} - /** * Build an initializer from a given expression. */ @@ -1577,37 +1551,32 @@ static initializer_t *initializer_from_expression(type_t *orig_type, { /* TODO check that expression is a constant expression */ - /* §6.7.8.14/15 char array may be initialized by string literals */ - type_t *type = skip_typeref(orig_type); - type_t *expr_type_orig = expression->base.type; - type_t *expr_type = skip_typeref(expr_type_orig); + type_t *const type = skip_typeref(orig_type); - if (is_type_array(type) && expr_type->kind == TYPE_POINTER) { + /* §6.7.8.14/15 char array may be initialized by string literals */ + if (expression->kind == EXPR_STRING_LITERAL && is_type_array(type)) { array_type_t *const array_type = &type->array; type_t *const element_type = skip_typeref(array_type->element_type); - - if (element_type->kind == TYPE_ATOMIC && expression->kind == EXPR_STRING_LITERAL) { - switch (expression->string_literal.encoding) { - case STRING_ENCODING_CHAR: { - atomic_type_kind_t const akind = element_type->atomic.akind; - if (akind == ATOMIC_TYPE_CHAR - || akind == ATOMIC_TYPE_SCHAR - || akind == ATOMIC_TYPE_UCHAR) { - return initializer_from_string(array_type, - &expression->string_literal.value); - } - break; + switch (expression->string_literal.value.encoding) { + case STRING_ENCODING_CHAR: { + if (is_type_atomic(element_type, ATOMIC_TYPE_CHAR) || + is_type_atomic(element_type, ATOMIC_TYPE_SCHAR) || + is_type_atomic(element_type, ATOMIC_TYPE_UCHAR)) { + goto make_string_init; } + break; + } - case STRING_ENCODING_WIDE: { - type_t *bare_wchar_type = skip_typeref(type_wchar_t); - if (get_unqualified_type(element_type) == bare_wchar_type) { - return initializer_from_wide_string(array_type, - &expression->string_literal.value); - } - break; - } + case STRING_ENCODING_WIDE: { + type_t *bare_wchar_type = skip_typeref(type_wchar_t); + if (get_unqualified_type(element_type) == bare_wchar_type) { +make_string_init:; + initializer_t *const init = allocate_initializer_zero(INITIALIZER_STRING); + init->value.value = expression; + return init; } + break; + } } } @@ -2236,13 +2205,10 @@ static initializer_t *parse_initializer(parse_initializer_env_t *env) size = max_index + 1; break; - case INITIALIZER_STRING: - size = result->string.string.size + 1; - break; - - case INITIALIZER_WIDE_STRING: - size = result->wide_string.string.size; + case INITIALIZER_STRING: { + size = get_string_len(&get_init_string(result)->value) + 1; break; + } case INITIALIZER_DESIGNATOR: case INITIALIZER_VALUE: @@ -2496,6 +2462,7 @@ static type_t *parse_typeof(void) } typedef enum specifiers_t { + SPECIFIER_NONE = 0, SPECIFIER_SIGNED = 1 << 0, SPECIFIER_UNSIGNED = 1 << 1, SPECIFIER_LONG = 1 << 2, @@ -3034,7 +3001,8 @@ warn_about_long_long: } else { errorf(pos, "multiple datatypes in declaration"); } - goto end_error; + specifiers->type = type_error_type; + return; } } @@ -3063,10 +3031,6 @@ warn_about_long_long: if (specifiers->attributes != NULL) type = handle_type_attributes(specifiers->attributes, type); specifiers->type = type; - return; - -end_error: - specifiers->type = type_error_type; } static type_qualifiers_t parse_type_qualifiers(void) @@ -4736,7 +4700,6 @@ static bool initializer_returns(initializer_t const *const init) } case INITIALIZER_STRING: - case INITIALIZER_WIDE_STRING: case INITIALIZER_DESIGNATOR: // designators have no payload return true; } @@ -5501,8 +5464,7 @@ static expression_t *find_create_select(const source_position_t *pos, symbol_t *iter_symbol = iter->base.symbol; if (iter_symbol == NULL) { type_t *type = iter->declaration.type; - if (type->kind != TYPE_COMPOUND_STRUCT - && type->kind != TYPE_COMPOUND_UNION) + if (!is_type_compound(type)) continue; compound_t *sub_compound = type->compound.compound; @@ -5725,8 +5687,8 @@ static type_t *get_string_type(string_encoding_t const enc) static expression_t *parse_string_literal(void) { expression_t *const expr = allocate_expression_zero(EXPR_STRING_LITERAL); - expr->string_literal.value = concat_string_literals(&expr->string_literal.encoding); - expr->base.type = get_string_type(expr->string_literal.encoding); + expr->string_literal.value = concat_string_literals(); + expr->base.type = get_string_type(expr->string_literal.value.encoding); return expr; } @@ -5744,110 +5706,204 @@ static expression_t *parse_boolean_literal(bool value) return literal; } -static void warn_traditional_suffix(void) +static void warn_traditional_suffix(char const *const suffix) { - warningf(WARN_TRADITIONAL, HERE, "traditional C rejects the '%S' suffix", - &token.number.suffix); + warningf(WARN_TRADITIONAL, HERE, "traditional C rejects the '%s' suffix", suffix); } -static void check_integer_suffix(void) +static void check_integer_suffix(expression_t *const expr, char const *const suffix) { - const string_t *suffix = &token.number.suffix; - if (suffix->size == 0) - return; - - bool not_traditional = false; - const char *c = suffix->begin; - if (*c == 'l' || *c == 'L') { - ++c; - if (*c == *(c-1)) { - not_traditional = true; - ++c; - if (*c == 'u' || *c == 'U') { + unsigned spec = SPECIFIER_NONE; + char const *c = suffix; + for (;;) { + specifiers_t add; + if (*c == 'L' || *c == 'l') { + add = SPECIFIER_LONG; + if (*c == c[1]) { + add |= SPECIFIER_LONG_LONG; ++c; } - } else if (*c == 'u' || *c == 'U') { - not_traditional = true; - ++c; + } else if (*c == 'U' || *c == 'u') { + add = SPECIFIER_UNSIGNED; + } else { + break; } - } else if (*c == 'u' || *c == 'U') { - not_traditional = true; ++c; - if (*c == 'l' || *c == 'L') { - ++c; - if (*c == *(c-1)) { - ++c; - } - } - } - if (*c != '\0') { - errorf(HERE, "invalid suffix '%S' on integer constant", suffix); - } else if (not_traditional) { - warn_traditional_suffix(); + if (spec & add) + goto error; + spec |= add; + } + + if (*c == '\0') { + type_t *type; + switch (spec) { + case SPECIFIER_NONE: type = type_int; break; + case SPECIFIER_LONG: type = type_long; break; + case SPECIFIER_LONG | SPECIFIER_LONG_LONG: type = type_long_long; break; + case SPECIFIER_UNSIGNED: type = type_unsigned_int; break; + case SPECIFIER_UNSIGNED | SPECIFIER_LONG: type = type_unsigned_long; break; + case SPECIFIER_UNSIGNED | SPECIFIER_LONG | SPECIFIER_LONG_LONG: type = type_unsigned_long_long; break; + default: panic("inconsistent suffix"); + } + if (spec != SPECIFIER_NONE && spec != SPECIFIER_LONG) { + warn_traditional_suffix(suffix); + } + expr->base.type = type; + /* Integer type depends on the size of the number and the size + * representable by the types. The backend/codegeneration has to + * determine that. */ + determine_literal_type(&expr->literal); + } else { +error: + errorf(HERE, "invalid suffix '%s' on integer constant", suffix); } } -static type_t *check_floatingpoint_suffix(void) +static void check_floatingpoint_suffix(expression_t *const expr, char const *const suffix) { - const string_t *suffix = &token.number.suffix; - type_t *type = type_double; - if (suffix->size == 0) - return type; - - bool not_traditional = false; - const char *c = suffix->begin; - if (*c == 'f' || *c == 'F') { - ++c; - type = type_float; - } else if (*c == 'l' || *c == 'L') { - ++c; - type = type_long_double; - } - if (*c != '\0') { - errorf(HERE, "invalid suffix '%S' on floatingpoint constant", suffix); - } else if (not_traditional) { - warn_traditional_suffix(); + type_t *type; + char const *c = suffix; + switch (*c) { + case 'F': + case 'f': type = type_float; ++c; break; + case 'L': + case 'l': type = type_long_double; ++c; break; + default: type = type_double; break; } - return type; + if (*c == '\0') { + expr->base.type = type; + if (suffix[0] != '\0') { + warn_traditional_suffix(suffix); + } + } else { + errorf(HERE, "invalid suffix '%s' on floatingpoint constant", suffix); + } } -/** - * Parse an integer constant. - */ static expression_t *parse_number_literal(void) { - expression_kind_t kind; - type_t *type; + string_t const *const str = &token.literal.string; + char const * i = str->begin; + unsigned digits = 0; + bool is_float = false; - switch (token.kind) { - case T_INTEGER: - kind = EXPR_LITERAL_INTEGER; - check_integer_suffix(); - type = type_int; + /* Parse base prefix. */ + unsigned base; + if (*i == '0') { + switch (*++i) { + case 'B': case 'b': base = 2; ++i; break; + case 'X': case 'x': base = 16; ++i; break; + default: base = 8; digits |= 1U << 0; break; + } + } else { + base = 10; + } + + /* Parse mantissa. */ + for (;; ++i) { + unsigned digit; + switch (*i) { + case '.': + if (is_float) { + errorf(HERE, "multiple decimal points in %K", &token); + i = 0; + goto done; + } + is_float = true; + if (base == 8) + base = 10; + continue; + + case '0': digit = 0; break; + case '1': digit = 1; break; + case '2': digit = 2; break; + case '3': digit = 3; break; + case '4': digit = 4; break; + case '5': digit = 5; break; + case '6': digit = 6; break; + case '7': digit = 7; break; + case '8': digit = 8; break; + case '9': digit = 9; break; + case 'A': case 'a': digit = 10; break; + case 'B': case 'b': digit = 11; break; + case 'C': case 'c': digit = 12; break; + case 'D': case 'd': digit = 13; break; + case 'E': case 'e': digit = 14; break; + case 'F': case 'f': digit = 15; break; + + default: goto done_mantissa; + } + + if (digit >= 10 && base != 16) + goto done_mantissa; + + digits |= 1U << digit; + } +done_mantissa: + + /* Parse exponent. */ + switch (base) { + case 2: + if (is_float) + errorf(HERE, "binary floating %K not allowed", &token); break; - case T_FLOATINGPOINT: - kind = EXPR_LITERAL_FLOATINGPOINT; - type = check_floatingpoint_suffix(); + case 8: + case 10: + if (*i == 'E' || *i == 'e') { + base = 10; + goto parse_exponent; + } + break; + + case 16: + if (*i == 'P' || *i == 'p') { +parse_exponent: + ++i; + is_float = true; + + if (*i == '-' || *i == '+') + ++i; + + if (isdigit(*i)) { + do { + ++i; + } while (isdigit(*i)); + } else { + errorf(HERE, "exponent of %K has no digits", &token); + } + } else if (is_float) { + errorf(HERE, "hexadecimal floating %K requires an exponent", &token); + i = 0; + } break; default: - panic("unexpected token type in parse_number_literal"); + panic("invalid base"); } - expression_t *literal = allocate_expression_zero(kind); - literal->base.type = type; - literal->literal.value = token.number.number; - literal->literal.suffix = token.number.suffix; - next_token(); +done:; + expression_t *const expr = allocate_expression_zero(is_float ? EXPR_LITERAL_FLOATINGPOINT : EXPR_LITERAL_INTEGER); + expr->literal.value = *str; - /* integer type depends on the size of the number and the size - * representable by the types. The backend/codegeneration has to determine - * that - */ - determine_literal_type(&literal->literal); - return literal; + if (i) { + if (digits == 0) { + errorf(HERE, "%K has no digits", &token); + } else if (digits & ~((1U << base) - 1)) { + errorf(HERE, "invalid digit in %K", &token); + } else { + expr->literal.suffix = i; + if (is_float) { + check_floatingpoint_suffix(expr, i); + } else { + check_integer_suffix(expr, i); + } + } + } + + eat(T_NUMBER); + return expr; } /** @@ -5856,13 +5912,13 @@ static expression_t *parse_number_literal(void) static expression_t *parse_character_constant(void) { expression_t *const literal = allocate_expression_zero(EXPR_LITERAL_CHARACTER); - literal->string_literal.encoding = token.string.encoding; - literal->string_literal.value = token.string.string; + literal->string_literal.value = token.literal.string; - switch (token.string.encoding) { + size_t const size = get_string_len(&token.literal.string); + switch (token.literal.string.encoding) { case STRING_ENCODING_CHAR: literal->base.type = c_mode & _CXX ? type_char : type_int; - if (literal->string_literal.value.size > 1) { + if (size > 1) { if (!GNU_MODE && !(c_mode & _C99)) { errorf(HERE, "more than 1 character in character constant"); } else { @@ -5874,7 +5930,7 @@ static expression_t *parse_character_constant(void) case STRING_ENCODING_WIDE: literal->base.type = type_int; - if (wstrlen(&literal->string_literal.value) > 1) { + if (size > 1) { warningf(WARN_MULTICHAR, HERE, "multi-character character constant"); } break; @@ -5969,7 +6025,7 @@ type_t *revert_automatic_type_conversion(const expression_t *expression) } case EXPR_STRING_LITERAL: { - size_t const size = expression->string_literal.value.size + 1; + size_t const size = get_string_len(&expression->string_literal.value) + 1; type_t *const elem = get_unqualified_type(expression->base.type->pointer.points_to); return make_array_type(elem, size, TYPE_QUALIFIER_NONE); } @@ -6663,8 +6719,7 @@ static expression_t *parse_primary_expression(void) switch (token.kind) { case T_false: return parse_boolean_literal(false); case T_true: return parse_boolean_literal(true); - case T_INTEGER: - case T_FLOATINGPOINT: return parse_number_literal(); + case T_NUMBER: return parse_number_literal(); case T_CHARACTER_CONSTANT: return parse_character_constant(); case T_STRING_LITERAL: return parse_string_literal(); case T___func__: return parse_function_keyword(FUNCNAME_FUNCTION); @@ -6887,9 +6942,7 @@ static expression_t *parse_select_expression(expression_t *addr) type_left = type; } - if (type_left->kind != TYPE_COMPOUND_STRUCT && - type_left->kind != TYPE_COMPOUND_UNION) { - + if (!is_type_compound(type_left)) { if (is_type_valid(type_left) && !saw_error) { errorf(&pos, "request for member '%Y' in something not a struct or union, but '%T'", @@ -8521,10 +8574,7 @@ static void register_expression_parser(parse_expression_function parser, { expression_parser_function_t *entry = &expression_parsers[token_kind]; - if (entry->parser != NULL) { - diagnosticf("for token '%k'\n", (token_kind_t)token_kind); - panic("trying to register multiple expression parsers for a token"); - } + assert(!entry->parser); entry->parser = parser; } @@ -8540,11 +8590,7 @@ static void register_infix_parser(parse_expression_infix_function parser, { expression_parser_function_t *entry = &expression_parsers[token_kind]; - if (entry->infix_parser != NULL) { - diagnosticf("for token '%k'\n", (token_kind_t)token_kind); - panic("trying to register multiple infix expression parsers for a " - "token"); - } + assert(!entry->infix_parser); entry->infix_parser = parser; entry->infix_precedence = precedence; } @@ -9765,10 +9811,9 @@ static statement_t *parse_compound_statement(bool inside_expression_statement) add_anchor_token('~'); add_anchor_token(T_CHARACTER_CONSTANT); add_anchor_token(T_COLONCOLON); - add_anchor_token(T_FLOATINGPOINT); add_anchor_token(T_IDENTIFIER); - add_anchor_token(T_INTEGER); add_anchor_token(T_MINUSMINUS); + add_anchor_token(T_NUMBER); add_anchor_token(T_PLUSPLUS); add_anchor_token(T_STRING_LITERAL); add_anchor_token(T__Bool); @@ -9946,10 +9991,9 @@ static statement_t *parse_compound_statement(bool inside_expression_statement) rem_anchor_token(T__Bool); rem_anchor_token(T_STRING_LITERAL); rem_anchor_token(T_PLUSPLUS); + rem_anchor_token(T_NUMBER); rem_anchor_token(T_MINUSMINUS); - rem_anchor_token(T_INTEGER); rem_anchor_token(T_IDENTIFIER); - rem_anchor_token(T_FLOATINGPOINT); rem_anchor_token(T_COLONCOLON); rem_anchor_token(T_CHARACTER_CONSTANT); rem_anchor_token('~'); @@ -10285,8 +10329,6 @@ void parse(void) */ void init_parser(void) { - sym_anonymous = symbol_table_insert(""); - memset(token_anchor_set, 0, sizeof(token_anchor_set)); init_expression_parsers();