X-Git-Url: http://nsz.repo.hu/git/?a=blobdiff_plain;f=parser.c;h=fdb85bd97252e3856e0f3d8c25a090a88af77dc4;hb=353985d8d06fcac2f36682820b8ae635c5e6162e;hp=58045f3f65789552a9b5c771e09adf1e66911962;hpb=bc2c469866488880921a9809d8dee3bd1333a25c;p=cparser diff --git a/parser.c b/parser.c index 58045f3..fdb85bd 100644 --- a/parser.c +++ b/parser.c @@ -2,6 +2,7 @@ #include #include +#include #include "parser.h" #include "lexer.h" @@ -16,61 +17,121 @@ //#define PRINT_TOKENS //#define ABORT_ON_ERROR #define MAX_LOOKAHEAD 2 +//#define STRICT_C99 -struct environment_entry_t { - symbol_t *symbol; +typedef struct { declaration_t *old_declaration; - const void *old_context; -}; + symbol_t *symbol; + unsigned short namespace; +} stack_entry_t; static token_t token; static token_t lookahead_buffer[MAX_LOOKAHEAD]; static int lookahead_bufpos; -static struct obstack environment_obstack; -static environment_entry_t **environment_stack = NULL; +static 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 -statement_t *parse_compound_statement(void); -static -statement_t *parse_statement(void); +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 statement_t *parse_compound_statement(void); +static statement_t *parse_statement(void); + +static expression_t *parse_sub_expression(unsigned precedence); +static expression_t *parse_expression(void); +static type_t *parse_typename(void); + +#define STORAGE_CLASSES \ + case T_typedef: \ + case T_extern: \ + case T_static: \ + case T_auto: \ + case T_register: + +#define TYPE_QUALIFIERS \ + case T_const: \ + case T_restrict: \ + case T_volatile: \ + case T_inline: + +#ifdef PROVIDE_COMPLEX +#define COMPLEX_SPECIFIERS \ + case T__Complex: +#else +#define COMPLEX_SPECIFIERS +#endif + +#ifdef PROVIDE_IMAGINARY +#define IMAGINARY_SPECIFIERS \ + case T__Imaginary: +#else +#define IMAGINARY_SPECIFIERS +#endif + +#define TYPE_SPECIFIERS \ + case T_void: \ + case T_char: \ + case T_short: \ + case T_int: \ + case T_long: \ + case T_float: \ + case T_double: \ + case T_signed: \ + case T_unsigned: \ + case T__Bool: \ + case T_struct: \ + case T_union: \ + case T_enum: \ + case T___typeof__: \ + COMPLEX_SPECIFIERS \ + IMAGINARY_SPECIFIERS -static -expression_t *parse_sub_expression(unsigned precedence); -static -expression_t *parse_expression(void); +#define DECLARATION_START \ + STORAGE_CLASSES \ + TYPE_QUALIFIERS \ + TYPE_SPECIFIERS -static inline -void *allocate_ast_zero(size_t size) +#define TYPENAME_START \ + TYPE_QUALIFIERS \ + TYPE_SPECIFIERS + +static inline void *allocate_ast_zero(size_t size) { void *res = allocate_ast(size); memset(res, 0, size); return res; } -static inline -void *allocate_type_zero(size_t size) +static inline void *allocate_type_zero(size_t size) { void *res = obstack_alloc(type_obst, size); memset(res, 0, size); return res; } +static inline void free_type(void *type) +{ + obstack_free(type_obst, type); +} + /** * returns the top element of the environment stack */ -static inline -size_t environment_top(void) +static inline size_t environment_top(void) { return ARR_LEN(environment_stack); } -static inline -void next_token(void) +static inline void next_token(void) { token = lookahead_buffer[lookahead_bufpos]; lookahead_buffer[lookahead_bufpos] = lexer_token; @@ -84,46 +145,58 @@ void next_token(void) #endif } -static inline -const token_t *la(int num) +static inline const token_t *look_ahead(int num) { assert(num > 0 && num <= MAX_LOOKAHEAD); int pos = (lookahead_bufpos+num-1) % MAX_LOOKAHEAD; return & lookahead_buffer[pos]; } -static inline -void eat(token_type_t type) +#define eat(token_type) do { assert(token.type == token_type); next_token(); } while(0) + +static void error(void) { - assert(token.type == type); - next_token(); +#ifdef ABORT_ON_ERROR + abort(); +#endif } -void parser_print_error_prefix_pos(const source_position_t source_position) +static void parser_print_prefix_pos(const source_position_t source_position) { fputs(source_position.input_name, stderr); fputc(':', stderr); fprintf(stderr, "%d", source_position.linenr); - fputs(": error: ", stderr); -#ifdef ABORT_ON_ERROR - abort(); -#endif + fputs(": ", stderr); +} + +static void parser_print_error_prefix_pos( + const source_position_t source_position) +{ + parser_print_prefix_pos(source_position); + fputs("error: ", stderr); + error(); } -void parser_print_error_prefix(void) +static void parser_print_error_prefix(void) { - parser_print_error_prefix_pos(token.source_position); + parser_print_prefix_pos(token.source_position); + error(); } -static -void parse_error(const char *message) +static void parse_error(const char *message) { parser_print_error_prefix(); fprintf(stderr, "parse error: %s\n", message); } -static -void parse_error_expected(const char *message, ...) +__attribute__((unused)) +static void parse_warning(const char *message) +{ + parser_print_prefix_pos(token.source_position); + fprintf(stderr, "warning: %s\n", message); +} + +static void parse_error_expected(const char *message, ...) { va_list args; int first = 1; @@ -152,21 +225,67 @@ void parse_error_expected(const char *message, ...) fprintf(stderr, "\n"); } -static -void eat_until(int token_type) +static void eat_block(void) { - while(token.type != token_type) { + if(token.type == '{') + next_token(); + + while(token.type != '}') { if(token.type == T_EOF) return; + if(token.type == '{') { + eat_block(); + continue; + } next_token(); } - next_token(); + eat('}'); +} + +static void eat_statement(void) +{ + while(token.type != ';') { + if(token.type == T_EOF) + return; + if(token.type == '}') + return; + if(token.type == '{') { + eat_block(); + continue; + } + next_token(); + } + eat(';'); +} + +static void eat_brace(void) +{ + if(token.type == '(') + next_token(); + + while(token.type != ')') { + if(token.type == T_EOF) + return; + if(token.type == ')' || token.type == ';' || token.type == '}') { + return; + } + if(token.type == '(') { + eat_brace(); + continue; + } + if(token.type == '{') { + eat_block(); + continue; + } + next_token(); + } + eat(')'); } #define expect(expected) \ if(UNLIKELY(token.type != (expected))) { \ parse_error_expected(NULL, (expected), 0); \ - eat_until(';'); \ + eat_statement(); \ return NULL; \ } \ next_token(); @@ -174,7 +293,7 @@ void eat_until(int token_type) #define expect_void(expected) \ if(UNLIKELY(token.type != (expected))) { \ parse_error_expected(NULL, (expected), 0); \ - eat_until(';'); \ + eat_statement(); \ return; \ } \ next_token(); @@ -183,82 +302,159 @@ static void set_context(context_t *new_context) { context = new_context; - declaration_t *declaration = new_context->declarations; - if(declaration != NULL) { - while(1) { - if(declaration->next == NULL) - break; - declaration = declaration->next; + last_declaration = new_context->declarations; + if(last_declaration != NULL) { + while(last_declaration->next != NULL) { + last_declaration = last_declaration->next; } } +} - last_declaration = declaration; +/** + * called when we find a 2nd declarator for an identifier we already have a + * declarator for + */ +static bool is_compatible_declaration (declaration_t *declaration, + declaration_t *previous) +{ + /* TODO: not correct yet */ + return declaration->type == previous->type; +} + +static declaration_t *get_declaration(symbol_t *symbol, namespace_t namespace) +{ + declaration_t *declaration = symbol->declaration; + for( ; declaration != NULL; declaration = declaration->symbol_next) { + if(declaration->namespace == namespace) + return declaration; + } + + return NULL; +} + +static const char *get_namespace_prefix(namespace_t namespace) +{ + switch(namespace) { + case NAMESPACE_NORMAL: + return ""; + case NAMESPACE_UNION: + return "union "; + case NAMESPACE_STRUCT: + return "struct "; + case NAMESPACE_ENUM: + return "enum "; + } + panic("invalid namespace found"); } /** * pushs an environment_entry on the environment stack and links the * corresponding symbol to the new entry */ -static inline -void environment_push(declaration_t *declaration, const void *context) +static declaration_t *environment_push(declaration_t *declaration) { - environment_entry_t *entry - = obstack_alloc(&environment_obstack, sizeof(entry[0])); - memset(entry, 0, sizeof(entry[0])); - - int top = ARR_LEN(environment_stack); - ARR_RESIZE(environment_stack, top + 1); - environment_stack[top] = entry; - + symbol_t *symbol = declaration->symbol; + namespace_t namespace = declaration->namespace; assert(declaration->source_position.input_name != NULL); - symbol_t *symbol = declaration->symbol; - assert(declaration != symbol->declaration); + /* a declaration should be only pushed once */ + assert(declaration->parent_context == NULL); + declaration->parent_context = context; - if(symbol->context == context) { - if(symbol->declaration != NULL) { - assert(symbol->declaration != NULL); + declaration_t *previous_declaration = get_declaration(symbol, namespace); + assert(declaration != previous_declaration); + if(previous_declaration != NULL + && previous_declaration->parent_context == context) { + if(!is_compatible_declaration(declaration, previous_declaration)) { parser_print_error_prefix_pos(declaration->source_position); - fprintf(stderr, "multiple definitions for symbol '%s'.\n", - symbol->string); - parser_print_error_prefix_pos(symbol->declaration->source_position); - fprintf(stderr, "this is the location of the previous declaration.\n"); + fprintf(stderr, "definition of symbol %s%s with type ", + get_namespace_prefix(namespace), symbol->string); + error(); + print_type(declaration->type); + fputc('\n', stderr); + parser_print_error_prefix_pos( + previous_declaration->source_position); + fprintf(stderr, "is incompatible with previous declaration " + "of type "); + print_type(previous_declaration->type); + fputc('\n', stderr); + } + return previous_declaration; + } + + /* remember old declaration */ + stack_entry_t entry; + entry.symbol = symbol; + entry.old_declaration = symbol->declaration; + entry.namespace = namespace; + ARR_APP1(environment_stack, 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; + break; + } } } - entry->old_declaration = symbol->declaration; - entry->old_context = symbol->context; - entry->symbol = symbol; - symbol->declaration = declaration; - symbol->context = context; + return declaration; } /** * pops symbols from the environment stack until @p new_top is the top element */ -static inline -void environment_pop_to(size_t new_top) +static void environment_pop_to(size_t new_top) { - environment_entry_t *entry = NULL; size_t top = ARR_LEN(environment_stack); size_t i; + assert(new_top <= top); if(new_top == top) return; - assert(new_top < top); - i = top; - do { - entry = environment_stack[i - 1]; - - symbol_t *symbol = entry->symbol; + for(i = top; i > new_top; --i) { + stack_entry_t *entry = & environment_stack[i - 1]; - symbol->declaration = entry->old_declaration; - symbol->context = entry->old_context; + declaration_t *old_declaration = entry->old_declaration; + symbol_t *symbol = entry->symbol; + namespace_t namespace = entry->namespace; - --i; - } while(i != new_top); - obstack_free(&environment_obstack, entry); + /* replace/remove declaration */ + declaration_t *declaration = symbol->declaration; + assert(declaration != NULL); + if(declaration->namespace == namespace) { + if(old_declaration == NULL) { + 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); + break; + } + } + assert(declaration != NULL); + } + } ARR_SHRINKLEN(environment_stack, (int) new_top); } @@ -267,22 +463,20 @@ void environment_pop_to(size_t new_top) static expression_t *parse_constant_expression(void) { - /* TODO: not correct yet */ - return parse_expression(); + /* start parsing at precedence 7 (conditional expression) */ + return parse_sub_expression(7); } static expression_t *parse_assignment_expression(void) { - /* TODO: not correct yet */ - return parse_expression(); + /* start parsing at precedence 2 (assignment expression) */ + return parse_sub_expression(2); } static void parse_compound_type_entries(void); -static void parse_declarator(declaration_t *declaration, - storage_class_t storage_class, type_t *type, - int may_be_abstract); -static void maybe_push_declaration(declaration_t *declaration); -static void record_declaration(declaration_t *declaration); +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 { @@ -290,95 +484,271 @@ struct declaration_specifiers_t { type_t *type; }; -static type_t *parse_struct_specifier(void) +static const char *parse_string_literals(void) { - eat(T_struct); - - compound_type_t *struct_type = allocate_type_zero(sizeof(struct_type[0])); - struct_type->type.type = TYPE_COMPOUND_STRUCT; - struct_type->source_position = token.source_position; + assert(token.type == T_STRING_LITERAL); + const char *result = token.v.string; - int top = environment_top(); - context_t *last_context = context; - set_context(&struct_type->context); + next_token(); - if(token.type == T_IDENTIFIER) { + while(token.type == T_STRING_LITERAL) { + result = concat_strings(result, token.v.string); next_token(); - if(token.type == '{') { - parse_compound_type_entries(); - } - } else if(token.type == '{') { - parse_compound_type_entries(); - } else { - parse_error_expected("problem while parsing struct type specifiers", - T_IDENTIFIER, '{', 0); - struct_type = NULL; } - assert(context == &struct_type->context); - set_context(last_context); - environment_pop_to(top); + return result; +} + +static void parse_attributes(void) +{ + while(true) { + switch(token.type) { + case T___attribute__: + next_token(); + + expect_void('('); + int depth = 1; + while(depth > 0) { + switch(token.type) { + case T_EOF: + parse_error("EOF while parsing attribute"); + break; + case '(': + next_token(); + depth++; + break; + case ')': + next_token(); + depth--; + break; + default: + next_token(); + } + } + break; + case T_asm: + next_token(); + expect_void('('); + if(token.type != T_STRING_LITERAL) { + parse_error_expected("while parsing assembler attribute", + T_STRING_LITERAL); + eat_brace(); + break; + } else { + parse_string_literals(); + } + expect_void(')'); + break; + default: + goto attributes_finished; + } + } - return (type_t*) struct_type; +attributes_finished: + ; } -static type_t *parse_union_specifier(void) +static designator_t *parse_designation(void) { - eat(T_union); + if(token.type != '[' && token.type != '.') + return NULL; - compound_type_t *union_type = allocate_type_zero(sizeof(union_type[0])); - union_type->type.type = TYPE_COMPOUND_UNION; - union_type->source_position = token.source_position; + designator_t *result = NULL; + designator_t *last = NULL; - int top = environment_top(); - context_t *last_context = context; - set_context(&union_type->context); + while(1) { + designator_t *designator; + switch(token.type) { + case '[': + designator = allocate_ast_zero(sizeof(designator[0])); + next_token(); + designator->array_access = parse_constant_expression(); + expect(']'); + break; + case '.': + designator = allocate_ast_zero(sizeof(designator[0])); + next_token(); + if(token.type != T_IDENTIFIER) { + parse_error_expected("problem while parsing designator", + T_IDENTIFIER, 0); + return NULL; + } + designator->symbol = token.v.symbol; + next_token(); + break; + default: + expect('='); + return result; + } - if(token.type == T_IDENTIFIER) { - union_type->symbol = token.v.symbol; - next_token(); - if(token.type == '{') { - parse_compound_type_entries(); + assert(designator != NULL); + if(last != NULL) { + last->next = designator; + } else { + result = designator; } - } else if(token.type == '{') { - parse_compound_type_entries(); - } else { - parse_error_expected("problem while parsing union type specifiers", - T_IDENTIFIER, '{'); - union_type = NULL; + last = designator; } +} - assert(context == &union_type->context); - set_context(last_context); - environment_pop_to(top); +static initializer_t *parse_initializer_list(void); - return (type_t*) union_type; +static initializer_t *parse_initializer(void) +{ + designator_t *designator = parse_designation(); + + initializer_t *result; + if(token.type == '{') { + result = parse_initializer_list(); + } else { + result = allocate_ast_zero(sizeof(result[0])); + result->type = INITIALIZER_VALUE; + result->v.value = parse_assignment_expression(); + } + result->designator = designator; + + return result; } -static void parse_enum_type_entries(void) +static initializer_t *parse_initializer_list(void) { eat('{'); - if(token.type == '}') { - next_token(); - parse_error("empty enum not allowed"); - return; - } + initializer_t *result = allocate_ast_zero(sizeof(result[0])); + result->type = INITIALIZER_LIST; - do { - if(token.type != T_IDENTIFIER) { - parse_error_expected("problem while parsing enum entry", - T_IDENTIFIER, 0); - eat_until('}'); - return; + initializer_t *last = NULL; + while(1) { + initializer_t *initializer = parse_initializer(); + if(last != NULL) { + last->next = initializer; + } else { + result->v.list = initializer; } - next_token(); + last = initializer; - if(token.type == '=') { - parse_constant_expression(); + if(token.type == '}') + break; + + if(token.type != ',') { + parse_error_expected("problem while parsing initializer list", + ',', '}', 0); + eat_block(); + return result; } + eat(','); - if(token.type != ',') + if(token.type == '}') + break; + } + + expect('}'); + + return result; +} + +static declaration_t *parse_compound_type_specifier(bool is_struct) +{ + if(is_struct) { + eat(T_struct); + } else { + eat(T_union); + } + + symbol_t *symbol = NULL; + declaration_t *declaration = NULL; + + if(token.type == T_IDENTIFIER) { + symbol = token.v.symbol; + next_token(); + + if(is_struct) { + declaration = get_declaration(symbol, NAMESPACE_STRUCT); + } else { + declaration = get_declaration(symbol, NAMESPACE_UNION); + } + } else if(token.type != '{') { + if(is_struct) { + parse_error_expected("problem while parsing struct type specifier", + T_IDENTIFIER, '{', 0); + } else { + parse_error_expected("problem while parsing union type specifier", + T_IDENTIFIER, '{', 0); + } + + return NULL; + } + + if(declaration == NULL) { + declaration = allocate_type_zero(sizeof(declaration[0])); + + if(is_struct) { + declaration->namespace = NAMESPACE_STRUCT; + } else { + declaration->namespace = NAMESPACE_UNION; + } + declaration->source_position = token.source_position; + declaration->symbol = symbol; + } + + if(token.type == '{') { + if(declaration->init.is_defined) { + assert(symbol != NULL); + parser_print_error_prefix(); + fprintf(stderr, "multiple definition of %s %s\n", + is_struct ? "struct" : "union", symbol->string); + declaration->context.declarations = NULL; + } + record_declaration(declaration); + declaration->init.is_defined = true; + + int top = environment_top(); + context_t *last_context = context; + set_context(& declaration->context); + + parse_compound_type_entries(); + parse_attributes(); + + assert(context == & declaration->context); + set_context(last_context); + environment_pop_to(top); + } + + return declaration; +} + +static void parse_enum_entries(void) +{ + eat('{'); + + if(token.type == '}') { + next_token(); + parse_error("empty enum not allowed"); + return; + } + + do { + 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); + eat_block(); + return; + } + entry->storage_class = STORAGE_CLASS_ENUM_ENTRY; + entry->symbol = token.v.symbol; + entry->source_position = token.source_position; + next_token(); + + if(token.type == '=') { + next_token(); + entry->init.initializer = parse_initializer(); + } + + record_declaration(entry); + + if(token.type != ',') break; next_token(); } while(token.type != '}'); @@ -386,28 +756,111 @@ static void parse_enum_type_entries(void) expect_void('}'); } -static type_t *parse_enum_specifier(void) +static declaration_t *parse_enum_specifier(void) { eat(T_enum); - enum_type_t *enum_type = allocate_type_zero(sizeof(enum_type[0])); - enum_type->type.type = TYPE_ENUM; - enum_type->source_position = token.source_position; + declaration_t *declaration; + symbol_t *symbol; if(token.type == T_IDENTIFIER) { - enum_type->symbol = token.v.symbol; + symbol = token.v.symbol; next_token(); - if(token.type == '{') { - parse_enum_type_entries(); - } - } else if(token.type == '{') { - parse_enum_type_entries(); + + declaration = get_declaration(symbol, NAMESPACE_ENUM); + } else if(token.type != '{') { + parse_error_expected("problem while parsing enum type specifier", + T_IDENTIFIER, '{', 0); + return NULL; } else { - parse_error_expected("problem while parsing enum type specifiers", - T_IDENTIFIER, '{'); + declaration = NULL; + symbol = NULL; + } + + if(declaration == NULL) { + declaration = allocate_type_zero(sizeof(declaration[0])); + + declaration->namespace = NAMESPACE_ENUM; + declaration->source_position = token.source_position; + declaration->symbol = symbol; + } + + if(token.type == '{') { + if(declaration->init.is_defined) { + parser_print_error_prefix(); + fprintf(stderr, "multiple definitions of enum %s\n", + symbol->string); + } + record_declaration(declaration); + declaration->init.is_defined = 1; + + parse_enum_entries(); + parse_attributes(); + } + + return declaration; +} + +/** + * if a symbol is a typedef to another type, return true + */ +static bool is_typedef_symbol(symbol_t *symbol) +{ + declaration_t *declaration = get_declaration(symbol, NAMESPACE_NORMAL); + if(declaration == NULL + || declaration->storage_class != STORAGE_CLASS_TYPEDEF) + return false; + + return true; +} + +static type_t *parse_typeof(void) +{ + eat(T___typeof__); + + type_t *type; + + expect('('); + + 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; + + case T_IDENTIFIER: + if(is_typedef_symbol(token.v.symbol)) { + type = parse_typename(); + } else { + expression = parse_expression(); + type = expression->datatype; + } + break; + + TYPENAME_START + type = parse_typename(); + break; + + default: + expression = parse_expression(); + type = expression->datatype; + break; } - return (type_t*) enum_type; + expect(')'); + + typeof_type_t *typeof = allocate_type_zero(sizeof(typeof[0])); + typeof->type.type = TYPE_TYPEOF; + typeof->expression = expression; + typeof->typeof_type = type; + + return (type_t*) typeof; } typedef enum { @@ -430,81 +883,37 @@ typedef enum { #endif } specifiers_t; -#define STORAGE_CLASSES \ - case T_typedef: \ - case T_extern: \ - case T_static: \ - case T_auto: \ - case T_register: - -#define TYPE_QUALIFIERS \ - case T_const: \ - case T_restrict: \ - case T_volatile: \ - case T_inline: \ - case T___extension__: - -#ifdef PROVIDE_COMPLEX -#define COMPLEX_SPECIFIERS \ - case T__Complex: -#else -#define COMPLEX_SPECIFIERS -#endif - -#ifdef PROVIDE_IMAGINARY -#define IMAGINARY_SPECIFIERS \ - case T__Imaginary: -#else -#define IMAGINARY_SPECIFIERS -#endif - -#define TYPE_SPECIFIERS \ - case T_void: \ - case T_char: \ - case T_short: \ - case T_int: \ - case T_long: \ - case T_float: \ - case T_double: \ - case T_signed: \ - case T_unsigned: \ - case T__Bool: \ - case T_struct: \ - case T_union: \ - case T_enum: \ - COMPLEX_SPECIFIERS \ - IMAGINARY_SPECIFIERS - -#define DECLARATION_START \ - STORAGE_CLASSES \ - TYPE_QUALIFIERS \ - TYPE_SPECIFIERS - -static -type_t *create_builtin_type(symbol_t *symbol) +static type_t *create_builtin_type(symbol_t *symbol) { builtin_type_t *type = allocate_type_zero(sizeof(type[0])); type->type.type = TYPE_BUILTIN; type->symbol = symbol; - type_t *result = typehash_insert((type_t*) type); - if(result != (type_t*) type) { - obstack_free(type_obst, type); - } + return (type_t*) type; +} - return result; +static type_t *get_typedef_type(symbol_t *symbol) +{ + declaration_t *declaration = get_declaration(symbol, NAMESPACE_NORMAL); + if(declaration == NULL + || declaration->storage_class != STORAGE_CLASS_TYPEDEF) + return NULL; + + typedef_type_t *typedef_type = allocate_type_zero(sizeof(typedef_type[0])); + typedef_type->type.type = TYPE_TYPEDEF; + typedef_type->declaration = declaration; + + return (type_t*) typedef_type; } -static -void parse_declaration_specifiers(declaration_specifiers_t *specifiers) +static void parse_declaration_specifiers(declaration_specifiers_t *specifiers) { - declaration_t *declaration; type_t *type = NULL; unsigned type_qualifiers = 0; unsigned type_specifiers = 0; int newtype = 0; - while(1) { + while(true) { switch(token.type) { /* storage class */ @@ -578,32 +987,57 @@ void parse_declaration_specifiers(declaration_specifiers_t *specifiers) } break; - /* TODO: if type != NULL for the following rules issue an error */ - case T_struct: - type = parse_struct_specifier(); + /* TODO: if type != NULL for the following rules should issue + * an error */ + case T_struct: { + compound_type_t *compound_type + = allocate_type_zero(sizeof(compound_type[0])); + compound_type->type.type = TYPE_COMPOUND_STRUCT; + compound_type->declaration = parse_compound_type_specifier(true); + + type = (type_t*) compound_type; + break; + } + case T_union: { + compound_type_t *compound_type + = allocate_type_zero(sizeof(compound_type[0])); + compound_type->type.type = TYPE_COMPOUND_UNION; + compound_type->declaration = parse_compound_type_specifier(false); + + type = (type_t*) compound_type; break; - case T_union: - type = parse_union_specifier(); + } + case T_enum: { + enum_type_t *enum_type = allocate_type_zero(sizeof(enum_type[0])); + enum_type->type.type = TYPE_ENUM; + enum_type->declaration = parse_enum_specifier(); + + type = (type_t*) enum_type; break; - case T_enum: - type = parse_enum_specifier(); + } + case T___typeof__: + type = parse_typeof(); break; case T___builtin_va_list: type = create_builtin_type(token.v.symbol); next_token(); break; - case T_IDENTIFIER: - declaration = token.v.symbol->declaration; - if(declaration == NULL || - declaration->storage_class != STORAGE_CLASS_TYPEDEF) { + case T___attribute__: + /* TODO */ + parse_attributes(); + break; + + case T_IDENTIFIER: { + type_t *typedef_type = get_typedef_type(token.v.symbol); + + if(typedef_type == NULL) goto finish_specifiers; - } - type = declaration->type; - assert(type != NULL); next_token(); + type = typedef_type; break; + } /* function specifier */ default: @@ -708,7 +1142,13 @@ finish_specifiers: default: /* invalid specifier combination, give an error message */ if(type_specifiers == 0) { +#ifndef STRICT_C99 + parse_warning("no type specifiers in declaration (using int)"); + atomic_type = ATOMIC_TYPE_INT; + break; +#else parse_error("no type specifiers given in declaration"); +#endif } else if((type_specifiers & SPECIFIER_SIGNED) && (type_specifiers & SPECIFIER_UNSIGNED)) { parse_error("signed and unsigned specifiers gives"); @@ -736,18 +1176,17 @@ finish_specifiers: type_t *result = typehash_insert(type); if(newtype && result != (type_t*) type) { - obstack_free(type_obst, type); + free_type(type); } specifiers->type = result; } -static -unsigned parse_type_qualifiers(void) +static type_qualifier_t parse_type_qualifiers(void) { - unsigned type_qualifiers = 0; + type_qualifier_t type_qualifiers = 0; - while(1) { + while(true) { switch(token.type) { /* type qualifiers */ MATCH_TYPE_QUALIFIER(T_const, TYPE_QUALIFIER_CONST); @@ -761,60 +1200,9 @@ unsigned parse_type_qualifiers(void) } } -typedef struct parsed_pointer_t parsed_pointer_t; -struct parsed_pointer_t { - unsigned type_qualifiers; - parsed_pointer_t *next; -}; - -static -parsed_pointer_t *parse_pointers(void) -{ - parsed_pointer_t *result = NULL; - parsed_pointer_t *last_pointer = NULL; - - while(token.type == '*') { - next_token(); - parsed_pointer_t *pointer - = obstack_alloc(&temp_obst, sizeof(pointer[0])); - pointer->type_qualifiers = parse_type_qualifiers(); - - if(last_pointer != NULL) { - last_pointer->next = pointer; - } else { - result = pointer; - } - last_pointer = pointer; - } - - return result; -} - -static -type_t *make_pointers(type_t *type, parsed_pointer_t *pointer) -{ - for( ; pointer != NULL; pointer = pointer->next) { - pointer_type_t *pointer_type - = allocate_type_zero(sizeof(pointer_type[0])); - pointer_type->type.type = TYPE_POINTER; - pointer_type->points_to = type; - pointer_type->type.qualifiers = pointer->type_qualifiers; - - type_t *result = typehash_insert((type_t*) pointer_type); - if(result != (type_t*) pointer_type) { - obstack_free(type_obst, pointer_type); - } - - type = result; - } - - return type; -} - -static -void parse_identifier_list(void) +static void parse_identifier_list(void) { - while(1) { + while(true) { if(token.type != T_IDENTIFIER) { parse_error_expected("problem while parsing parameter identifier " "list", T_IDENTIFIER, 0); @@ -827,165 +1215,212 @@ void parse_identifier_list(void) } } -static -declaration_t *parse_parameter(void) +static declaration_t *parse_parameter(void) { declaration_specifiers_t specifiers; memset(&specifiers, 0, sizeof(specifiers)); parse_declaration_specifiers(&specifiers); - declaration_t *declaration = allocate_ast_zero(sizeof(declaration[0])); - parse_declarator(declaration, specifiers.storage_class, - specifiers.type, 1); + declaration_t *declaration = parse_declarator(specifiers.storage_class, + specifiers.type, 1); + + /* TODO check declaration constraints for parameters */ + if(declaration->storage_class == STORAGE_CLASS_TYPEDEF) { + parse_error("typedef not allowed in parameter list"); + } return declaration; } -static -void parse_parameters(method_type_t *type) +static declaration_t *parse_parameters(function_type_t *type) { if(token.type == T_IDENTIFIER) { - symbol_t *symbol = token.v.symbol; - declaration_t *declaration = symbol->declaration; - if(declaration == NULL - || declaration->storage_class != STORAGE_CLASS_TYPEDEF) { + symbol_t *symbol = token.v.symbol; + if(!is_typedef_symbol(symbol)) { + /* TODO */ parse_identifier_list(); - return; + return NULL; } } if(token.type == ')') { type->unspecified_parameters = 1; - return; + return NULL; } - if(token.type == T_void && la(1)->type == ')') { + if(token.type == T_void && look_ahead(1)->type == ')') { next_token(); - return; + return NULL; } - declaration_t *declaration; - method_parameter_type_t *parameter_type; - method_parameter_type_t *last_parameter_type = NULL; + declaration_t *declarations = NULL; + declaration_t *declaration; + declaration_t *last_declaration = NULL; + function_parameter_t *parameter; + function_parameter_t *last_parameter = NULL; - while(1) { + while(true) { switch(token.type) { case T_DOTDOTDOT: next_token(); type->variadic = 1; - return; + return declarations; case T_IDENTIFIER: + case T___extension__: DECLARATION_START declaration = parse_parameter(); - parameter_type = allocate_type_zero(sizeof(parameter_type[0])); - parameter_type->type = declaration->type; - parameter_type->symbol = declaration->symbol; + parameter = allocate_type_zero(sizeof(parameter[0])); + parameter->type = declaration->type; - if(last_parameter_type != NULL) { - last_parameter_type->next = parameter_type; + if(last_parameter != NULL) { + last_declaration->next = declaration; + last_parameter->next = parameter; } else { - type->parameter_types = parameter_type; + type->parameters = parameter; + declarations = declaration; } - last_parameter_type = parameter_type; + last_parameter = parameter; + last_declaration = declaration; break; default: - return; + return declarations; } if(token.type != ',') - return; + return declarations; next_token(); } } -static -const char *parse_string_literals(void) -{ - assert(token.type == T_STRING_LITERAL); - const char *result = token.v.string; +typedef enum { + CONSTRUCT_POINTER, + CONSTRUCT_FUNCTION, + CONSTRUCT_ARRAY +} construct_type_type_t; + +typedef struct construct_type_t construct_type_t; +struct construct_type_t { + construct_type_type_t type; + construct_type_t *next; +}; - next_token(); +typedef struct parsed_pointer_t parsed_pointer_t; +struct parsed_pointer_t { + construct_type_t construct_type; + type_qualifier_t type_qualifiers; +}; - while(token.type == T_STRING_LITERAL) { - result = concat_strings(result, token.v.string); - next_token(); - } +typedef struct construct_function_type_t construct_function_type_t; +struct construct_function_type_t { + construct_type_t construct_type; + function_type_t *function_type; +}; - return result; +typedef struct parsed_array_t parsed_array_t; +struct parsed_array_t { + construct_type_t construct_type; + type_qualifier_t type_qualifiers; + bool is_static; + bool is_variable; + expression_t *size; +}; + +typedef struct construct_base_type_t construct_base_type_t; +struct construct_base_type_t { + construct_type_t construct_type; + type_t *type; +}; + +static construct_type_t *parse_pointer_declarator(void) +{ + eat('*'); + + parsed_pointer_t *pointer = obstack_alloc(&temp_obst, sizeof(pointer[0])); + memset(pointer, 0, sizeof(pointer[0])); + pointer->type_qualifiers = parse_type_qualifiers(); + + return (construct_type_t*) pointer; } -static -void parse_attributes(void) +static construct_type_t *parse_array_declarator(void) { - while(1) { - switch(token.type) { - case T___attribute__: - next_token(); + eat('['); - expect_void('('); - int depth = 1; - while(depth > 0) { - switch(token.type) { - case T_EOF: - parse_error("EOF while parsing attribute"); - break; - case '(': - next_token(); - depth++; - break; - case ')': - next_token(); - depth--; - break; - default: - next_token(); - } - } - break; - case T_asm: + parsed_array_t *array = obstack_alloc(&temp_obst, sizeof(array[0])); + memset(array, 0, sizeof(array[0])); + + if(token.type == T_static) { + array->is_static = true; + next_token(); + } + + type_qualifier_t type_qualifiers = parse_type_qualifiers(); + if(type_qualifiers != 0) { + if(token.type == T_static) { + array->is_static = true; next_token(); - expect_void('('); - if(token.type != T_STRING_LITERAL) { - parse_error_expected("while parsing assembler attribute", - T_STRING_LITERAL); - eat_until(')'); - break; - } else { - parse_string_literals(); - } - expect_void(')'); - break; - default: - goto attributes_finished; } } + array->type_qualifiers = type_qualifiers; -attributes_finished: - ; + if(token.type == '*' && look_ahead(1)->type == ']') { + array->is_variable = true; + next_token(); + } else if(token.type != ']') { + array->size = parse_assignment_expression(); + } + + expect(']'); + + return (construct_type_t*) array; } -typedef struct declarator_part declarator_part; -struct declarator_part { - parsed_pointer_t *pointers; - method_type_t *method_type; - declarator_part *inner; -}; +static construct_type_t *parse_function_declarator(declaration_t *declaration) +{ + eat('('); + + function_type_t *type = allocate_type_zero(sizeof(type[0])); + type->type.type = TYPE_FUNCTION; + + declaration_t *parameters = parse_parameters(type); + if(declaration != NULL) { + declaration->context.declarations = parameters; + } + + construct_function_type_t *construct_function_type = + obstack_alloc(&temp_obst, sizeof(construct_function_type[0])); + memset(construct_function_type, 0, sizeof(construct_function_type[0])); + construct_function_type->construct_type.type = CONSTRUCT_FUNCTION; + construct_function_type->function_type = type; + + expect(')'); + return (construct_type_t*) construct_function_type; +} -static -declarator_part *parse_inner_declarator(declaration_t *declaration, - int may_be_abstract) +static construct_type_t *parse_inner_declarator(declaration_t *declaration, + int may_be_abstract) { - declarator_part *part = obstack_alloc(&temp_obst, sizeof(part[0])); - memset(part, 0, sizeof(part[0])); + construct_type_t *result = NULL; + construct_type_t *last = NULL; - part->pointers = parse_pointers(); + while(token.type == '*') { + construct_type_t *type = parse_pointer_declarator(); + if(last != NULL) { + last->next = type; + } else { + result = type; + } + last = type; + } /* TODO: find out if this is correct */ parse_attributes(); + construct_type_t *inner_types = NULL; + switch(token.type) { case T_IDENTIFIER: if(declaration == NULL) { @@ -998,7 +1433,7 @@ declarator_part *parse_inner_declarator(declaration_t *declaration, break; case '(': next_token(); - part->inner = parse_inner_declarator(declaration, may_be_abstract); + inner_types = parse_inner_declarator(declaration, may_be_abstract); expect(')'); break; default: @@ -1008,142 +1443,213 @@ declarator_part *parse_inner_declarator(declaration_t *declaration, '(', 0); } - while(1) { + while(true) { + construct_type_t *type; switch(token.type) { case '(': - next_token(); - - method_type_t *method_type - = allocate_type_zero(sizeof(method_type[0])); - method_type->type.type = TYPE_METHOD; - - parse_parameters(method_type); - - part->method_type = method_type; - - expect(')'); + type = parse_function_declarator(declaration); break; case '[': - next_token(); - - if(token.type == T_static) { - next_token(); - } - - unsigned type_qualifiers = parse_type_qualifiers(); - if(type_qualifiers != 0) { - if(token.type == T_static) { - next_token(); - } - } - - /* TODO */ - - if(token.type == '*' && la(1)->type == ']') { - next_token(); - } else if(token.type != ']') { - parse_assignment_expression(); - } - - expect(']'); + type = parse_array_declarator(); break; default: goto declarator_finished; } + + if(last != NULL) { + last->next = type; + } else { + result = type; + } + last = type; } declarator_finished: parse_attributes(); - return part; + if(inner_types != NULL) { + if(last != NULL) { + last->next = inner_types; + } else { + result = inner_types; + } + last = inner_types; + } + + return result; } -static -type_t *construct_declarator_type(declarator_part *part, type_t *type) +static type_t *construct_declarator_type(construct_type_t *construct_list, + type_t *type) { - do { - type = make_pointers(type, part->pointers); + construct_type_t *iter = construct_list; + for( ; iter != NULL; iter = iter->next) { + parsed_pointer_t *parsed_pointer; + parsed_array_t *parsed_array; + construct_function_type_t *construct_function_type; + function_type_t *function_type; + pointer_type_t *pointer_type; + array_type_t *array_type; - method_type_t *method_type = part->method_type; - if(method_type != NULL) { - method_type->result_type = type; + switch(iter->type) { + case CONSTRUCT_FUNCTION: + construct_function_type = (construct_function_type_t*) iter; + function_type = construct_function_type->function_type; + + function_type->result_type = type; + type = (type_t*) function_type; + break; - type = (type_t*) method_type; + case CONSTRUCT_POINTER: + parsed_pointer = (parsed_pointer_t*) iter; + pointer_type = allocate_type_zero(sizeof(pointer_type[0])); + + pointer_type->type.type = TYPE_POINTER; + pointer_type->points_to = type; + pointer_type->type.qualifiers = parsed_pointer->type_qualifiers; + type = (type_t*) pointer_type; + break; + + case CONSTRUCT_ARRAY: + parsed_array = (parsed_array_t*) iter; + array_type = allocate_type_zero(sizeof(array_type[0])); + + array_type->type.type = TYPE_ARRAY; + array_type->element_type = type; + array_type->type.qualifiers = parsed_array->type_qualifiers; + array_type->is_static = parsed_array->is_static; + array_type->is_variable = parsed_array->is_variable; + array_type->size = parsed_array->size; + type = (type_t*) array_type; + break; } - part = part->inner; - } while(part != NULL); + type_t *hashed_type = typehash_insert((type_t*) type); + if(hashed_type != type) { + free_type(type); + type = hashed_type; + } + } return type; } -static -void parse_declarator(declaration_t *declaration, storage_class_t storage_class, - type_t *type, int may_be_abstract) +static declaration_t *parse_declarator(storage_class_t storage_class, + type_t *type, int may_be_abstract) { - declarator_part *part + declaration_t *declaration = allocate_ast_zero(sizeof(declaration[0])); + declaration->storage_class = storage_class; + + construct_type_t *construct_type = parse_inner_declarator(declaration, may_be_abstract); + declaration->type = construct_declarator_type(construct_type, type); - if(part != NULL) { - declaration->type = construct_declarator_type(part, type); - declaration->storage_class = storage_class; - obstack_free(&temp_obst, part); + if(construct_type != NULL) { + obstack_free(&temp_obst, construct_type); } + + return declaration; } -static -type_t *parse_abstract_declarator(type_t *base_type) +static type_t *parse_abstract_declarator(type_t *base_type) { - declarator_part *part = parse_inner_declarator(NULL, 1); + construct_type_t *construct_type = parse_inner_declarator(NULL, 1); - type_t *result = construct_declarator_type(part, base_type); - obstack_free(&temp_obst, part); + type_t *result = construct_declarator_type(construct_type, base_type); + if(construct_type != NULL) { + obstack_free(&temp_obst, construct_type); + } return result; } -static void record_declaration(declaration_t *declaration) +static declaration_t *record_declaration(declaration_t *declaration) { + assert(context != NULL); + + symbol_t *symbol = declaration->symbol; + if(symbol != NULL) { + declaration_t *alias = environment_push(declaration); + if(alias != declaration) + return alias; + } else { + declaration->parent_context = context; + } + if(last_declaration != NULL) { last_declaration->next = declaration; } else { - if(context != NULL) - context->declarations = declaration; + context->declarations = declaration; } last_declaration = declaration; + + return declaration; } -static -void maybe_push_declaration(declaration_t *declaration) +static void parser_error_multiple_definition(declaration_t *previous, + declaration_t *declaration) { - symbol_t *symbol = declaration->symbol; - - if(symbol != NULL) { - environment_push(declaration, context); - } + parser_print_error_prefix_pos(declaration->source_position); + fprintf(stderr, "multiple definition of symbol '%s'\n", + declaration->symbol->string); + parser_print_error_prefix_pos(previous->source_position); + fprintf(stderr, "this is the location of the previous " + "definition.\n"); + error(); } -static -void parse_init_declarators(const declaration_specifiers_t *specifiers) +static void parse_init_declarators(const declaration_specifiers_t *specifiers) { - while(1) { - declaration_t *declaration = allocate_ast_zero(sizeof(declaration[0])); + while(true) { + declaration_t *ndeclaration + = parse_declarator(specifiers->storage_class, specifiers->type, 0); - parse_declarator(declaration, specifiers->storage_class, - specifiers->type, 0); - maybe_push_declaration(declaration); - record_declaration(declaration); + declaration_t *declaration = record_declaration(ndeclaration); if(token.type == '=') { next_token(); - if(token.type == '{') { - // TODO - expect_void('}'); - } else { - parse_assignment_expression(); + + /* TODO: check that this is an allowed type (no function type) */ + + if(declaration->init.initializer != NULL) { + parser_error_multiple_definition(declaration, ndeclaration); } + + ndeclaration->init.initializer = parse_initializer(); } else if(token.type == '{') { + if(declaration->type->type != TYPE_FUNCTION) { + parser_print_error_prefix(); + fprintf(stderr, "Declarator "); + print_type_ext(declaration->type, declaration->symbol, NULL); + fprintf(stderr, " has a body but is not a function type.\n"); + eat_block(); + continue; + } + + if(declaration->init.statement != NULL) { + parser_error_multiple_definition(declaration, ndeclaration); + } + if(ndeclaration != declaration) { + memcpy(&declaration->context, &ndeclaration->context, + sizeof(declaration->context)); + } + + int top = environment_top(); + context_t *last_context = context; + set_context(&declaration->context); + + /* push function parameters */ + declaration_t *parameter = declaration->context.declarations; + for( ; parameter != NULL; parameter = parameter->next) { + environment_push(parameter); + } + statement_t *statement = parse_compound_statement(); - declaration->statement = statement; + + assert(context == &declaration->context); + set_context(last_context); + environment_pop_to(top); + + declaration->init.statement = statement; return; } @@ -1154,8 +1660,7 @@ void parse_init_declarators(const declaration_specifiers_t *specifiers) expect_void(';'); } -static -void parse_struct_declarators(const declaration_specifiers_t *specifiers) +static void parse_struct_declarators(const declaration_specifiers_t *specifiers) { while(1) { if(token.type == ':') { @@ -1164,10 +1669,11 @@ void parse_struct_declarators(const declaration_specifiers_t *specifiers) /* TODO (bitfields) */ } else { declaration_t *declaration - = allocate_ast_zero(sizeof(declaration[0])); - parse_declarator(declaration, specifiers->storage_class, - specifiers->type, 0); - maybe_push_declaration(declaration); + = parse_declarator(specifiers->storage_class, + specifiers->type, 1); + + /* TODO: check constraints for struct declarations */ + /* TODO: check for doubled fields */ record_declaration(declaration); if(token.type == ':') { @@ -1191,8 +1697,6 @@ static void parse_compound_type_entries(void) while(token.type != '}' && token.type != T_EOF) { declaration_specifiers_t specifiers; memset(&specifiers, 0, sizeof(specifiers)); - /* TODO not correct as this allows storage class stuff... but only - * specifiers and qualifiers sould be allowed here */ parse_declaration_specifiers(&specifiers); parse_struct_declarators(&specifiers); @@ -1200,140 +1704,440 @@ static void parse_compound_type_entries(void) if(token.type == T_EOF) { parse_error("unexpected error while parsing struct"); } - next_token(); -} + next_token(); +} + +static void parse_declaration(void) +{ + source_position_t source_position = token.source_position; + + declaration_specifiers_t specifiers; + memset(&specifiers, 0, sizeof(specifiers)); + parse_declaration_specifiers(&specifiers); + + if(token.type == ';') { + next_token(); + + declaration_t *declaration = allocate_ast_zero(sizeof(declaration[0])); + + declaration->type = specifiers.type; + declaration->storage_class = specifiers.storage_class; + declaration->source_position = source_position; + record_declaration(declaration); + return; + } + parse_init_declarators(&specifiers); +} + +static type_t *parse_typename(void) +{ + declaration_specifiers_t specifiers; + memset(&specifiers, 0, sizeof(specifiers)); + parse_declaration_specifiers(&specifiers); + if(specifiers.storage_class != STORAGE_CLASS_NONE) { + /* TODO: improve error message, user does probably not know what a + * storage class is... + */ + parse_error("typename may not have a storage class"); + } + + type_t *result = parse_abstract_declarator(specifiers.type); + + return result; +} + + + + +typedef expression_t* (*parse_expression_function) (unsigned precedence); +typedef expression_t* (*parse_expression_infix_function) (unsigned precedence, + expression_t *left); + +typedef struct expression_parser_function_t expression_parser_function_t; +struct expression_parser_function_t { + unsigned precedence; + parse_expression_function parser; + unsigned infix_precedence; + parse_expression_infix_function infix_parser; +}; + +expression_parser_function_t expression_parsers[T_LAST_TOKEN]; + +static expression_t *expected_expression_error(void) +{ + parser_print_error_prefix(); + fprintf(stderr, "expected expression, got token "); + print_token(stderr, & token); + fprintf(stderr, "\n"); + + expression_t *expression = allocate_ast_zero(sizeof(expression[0])); + expression->type = EXPR_INVALID; + next_token(); + + return expression; +} + +static expression_t *parse_string_const(void) +{ + string_literal_t *cnst = allocate_ast_zero(sizeof(cnst[0])); + + cnst->expression.type = EXPR_STRING_LITERAL; + cnst->expression.datatype = type_string; + cnst->value = parse_string_literals(); + + return (expression_t*) cnst; +} + +static expression_t *parse_int_const(void) +{ + const_t *cnst = allocate_ast_zero(sizeof(cnst[0])); + + cnst->expression.type = EXPR_CONST; + cnst->expression.datatype = type_int; + cnst->v.int_value = token.v.intvalue; + + next_token(); + + return (expression_t*) cnst; +} + +static expression_t *parse_float_const(void) +{ + const_t *cnst = allocate_ast_zero(sizeof(cnst[0])); + + cnst->expression.type = EXPR_CONST; + cnst->expression.datatype = type_double; + cnst->v.float_value = token.v.floatvalue; + + next_token(); + + return (expression_t*) cnst; +} + +static declaration_t *create_implicit_function(symbol_t *symbol, + const source_position_t source_position) +{ + function_type_t *function_type = allocate_type_zero(sizeof(function_type)); + + function_type->type.type = TYPE_FUNCTION; + function_type->result_type = type_int; + function_type->unspecified_parameters = true; + + type_t *type = typehash_insert((type_t*) function_type); + if(type != (type_t*) function_type) { + free_type(function_type); + } + + declaration_t *declaration = allocate_ast_zero(sizeof(declaration[0])); + + declaration->storage_class = STORAGE_CLASS_EXTERN; + declaration->type = type; + declaration->symbol = symbol; + declaration->source_position = source_position; + + /* prepend the implicit definition to the global context + * this is safe since the symbol wasn't declared as anything else yet + */ + assert(symbol->declaration == NULL); + + context_t *last_context = context; + context = global_context; + + environment_push(declaration); + declaration->next = context->declarations; + context->declarations = declaration; + + context = last_context; + + return declaration; +} + +static expression_t *parse_reference(void) +{ + reference_expression_t *ref = allocate_ast_zero(sizeof(ref[0])); + + ref->expression.type = EXPR_REFERENCE; + ref->symbol = token.v.symbol; + + declaration_t *declaration = get_declaration(ref->symbol, NAMESPACE_NORMAL); + + source_position_t source_position = token.source_position; + next_token(); + + if(declaration == NULL) { +#ifndef STRICT_C99 + /* an implicitely defined function */ + if(token.type == '(') { + parser_print_prefix_pos(token.source_position); + fprintf(stderr, "warning: implicit declaration of function '%s'\n", + ref->symbol->string); + + declaration = create_implicit_function(ref->symbol, + source_position); + } else +#endif + { + parser_print_error_prefix(); + fprintf(stderr, "unknown symbol '%s' found.\n", ref->symbol->string); + return (expression_t*) ref; + } + } + + ref->declaration = declaration; + ref->expression.datatype = declaration->type; + + return (expression_t*) ref; +} + +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 */ +} + +static expression_t *parse_cast(void) +{ + unary_expression_t *cast = allocate_ast_zero(sizeof(cast[0])); + + cast->expression.type = EXPR_UNARY; + cast->type = UNEXPR_CAST; + cast->expression.source_position = token.source_position; + + type_t *type = parse_typename(); + + expect(')'); + expression_t *value = parse_sub_expression(20); + + check_cast_allowed(value, type); + + cast->expression.datatype = type; + cast->value = value; + + return (expression_t*) cast; +} + +static expression_t *parse_statement_expression(void) +{ + statement_expression_t *expression + = allocate_ast_zero(sizeof(expression[0])); + expression->expression.type = EXPR_STATEMENT; + expression->statement = parse_compound_statement(); + + /* 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; + } + + if(last_statement->type == STATEMENT_EXPRESSION) { + const expression_statement_t *expression_statement = + (const expression_statement_t*) last_statement; + expression->expression.datatype + = expression_statement->expression->datatype; + } else { + expression->expression.datatype = type_void; + } -void parse_declaration(void) -{ - declaration_specifiers_t specifiers; - memset(&specifiers, 0, sizeof(specifiers)); - parse_declaration_specifiers(&specifiers); + expect(')'); - if(token.type == ';') { - next_token(); - return; - } - parse_init_declarators(&specifiers); + return (expression_t*) expression; } -type_t *parse_typename(void) +static expression_t *parse_brace_expression(void) { - declaration_specifiers_t specifiers; - memset(&specifiers, 0, sizeof(specifiers)); - parse_declaration_specifiers(&specifiers); - if(specifiers.storage_class != STORAGE_CLASS_NONE) { - /* TODO: improve error message, user does probably not know what a - * storage class is... - */ - parse_error("typename may not have a storage class"); + eat('('); + + switch(token.type) { + case '{': + /* gcc extension: a stement expression */ + return parse_statement_expression(); + + TYPE_QUALIFIERS + TYPE_SPECIFIERS + return parse_cast(); + case T_IDENTIFIER: + if(is_typedef_symbol(token.v.symbol)) { + return parse_cast(); + } } - type_t *result = parse_abstract_declarator(specifiers.type); + expression_t *result = parse_expression(); + expect(')'); return result; } +static expression_t *parse_function_keyword(void) +{ + eat(T___FUNCTION__); + /* TODO */ + string_literal_t *expression = allocate_ast_zero(sizeof(expression[0])); + expression->expression.type = EXPR_FUNCTION; + expression->expression.datatype = type_string; + expression->value = "TODO: FUNCTION"; + return (expression_t*) expression; +} -typedef expression_t* (*parse_expression_function) (unsigned precedence); -typedef expression_t* (*parse_expression_infix_function) (unsigned precedence, - expression_t *left); +static expression_t *parse_pretty_function_keyword(void) +{ + eat(T___PRETTY_FUNCTION__); + /* TODO */ -typedef struct expression_parser_function_t expression_parser_function_t; -struct expression_parser_function_t { - unsigned precedence; - parse_expression_function parser; - unsigned infix_precedence; - parse_expression_infix_function infix_parser; -}; + string_literal_t *expression = allocate_ast_zero(sizeof(expression[0])); + expression->expression.type = EXPR_PRETTY_FUNCTION; + expression->expression.datatype = type_string; + expression->value = "TODO: PRETTY FUNCTION"; -expression_parser_function_t expression_parsers[T_LAST_TOKEN]; + return (expression_t*) expression; +} -static -expression_t *expected_expression_error(void) +static designator_t *parse_designator(void) { - parser_print_error_prefix(); - fprintf(stderr, "expected expression, got token "); - print_token(stderr, & token); - fprintf(stderr, "\n"); + designator_t *result = allocate_ast_zero(sizeof(result[0])); - expression_t *expression = allocate_ast_zero(sizeof(expression[0])); - expression->type = EXPR_INVALID; + if(token.type != T_IDENTIFIER) { + parse_error_expected("problem while parsing member designator", + T_IDENTIFIER, 0); + eat_brace(); + return NULL; + } + result->symbol = token.v.symbol; next_token(); - return expression; -} + designator_t *last_designator = result; + while(true) { + if(token.type == '.') { + next_token(); + if(token.type != T_IDENTIFIER) { + parse_error_expected("problem while parsing member designator", + T_IDENTIFIER, 0); + eat_brace(); + return NULL; + } + designator_t *designator = allocate_ast_zero(sizeof(result[0])); + designator->symbol = token.v.symbol; + next_token(); -static -expression_t *parse_string_const(void) -{ - string_literal_t *cnst = allocate_ast_zero(sizeof(cnst[0])); + last_designator->next = designator; + last_designator = designator; + continue; + } + if(token.type == '[') { + next_token(); + designator_t *designator = allocate_ast_zero(sizeof(result[0])); + designator->array_access = parse_expression(); + if(designator->array_access == NULL) { + eat_brace(); + return NULL; + } + expect(']'); - cnst->expression.type = EXPR_STRING_LITERAL; - cnst->value = parse_string_literals(); + last_designator->next = designator; + last_designator = designator; + continue; + } + break; + } - return (expression_t*) cnst; + return result; } -static -expression_t *parse_int_const(void) +static expression_t *parse_offsetof(void) { - const_t *cnst = allocate_ast_zero(sizeof(cnst[0])); + eat(T___builtin_offsetof); - cnst->expression.type = EXPR_CONST; - cnst->value = token.v.intvalue; + offsetof_expression_t *expression + = allocate_ast_zero(sizeof(expression[0])); + expression->expression.type = EXPR_OFFSETOF; + expression->expression.datatype = type_size_t; - next_token(); + expect('('); + expression->type = parse_typename(); + expect(','); + expression->designator = parse_designator(); + expect(')'); - return (expression_t*) cnst; + return (expression_t*) expression; } -static -expression_t *parse_reference(void) +static expression_t *parse_va_arg(void) { - reference_expression_t *ref = allocate_ast_zero(sizeof(ref[0])); + eat(T___builtin_va_arg); - ref->expression.type = EXPR_REFERENCE; - ref->symbol = token.v.symbol; + va_arg_expression_t *expression = allocate_ast_zero(sizeof(expression[0])); + expression->expression.type = EXPR_VA_ARG; - next_token(); + expect('('); + expression->arg = parse_assignment_expression(); + expect(','); + expression->expression.datatype = parse_typename(); + expect(')'); - return (expression_t*) ref; + return (expression_t*) expression; } -static -expression_t *parse_brace_expression(void) +static expression_t *parse_builtin_symbol(void) { - eat('('); + builtin_symbol_expression_t *expression + = allocate_ast_zero(sizeof(expression[0])); + expression->expression.type = EXPR_BUILTIN_SYMBOL; - expression_t *result = parse_expression(); - expect(')'); + /* TODO: set datatype */ - return result; + expression->symbol = token.v.symbol; + + next_token(); + + return (expression_t*) expression; } -static -expression_t *parse_primary_expression(void) +static expression_t *parse_primary_expression(void) { switch(token.type) { case T_INTEGER: return parse_int_const(); + case T_FLOATINGPOINT: + return parse_float_const(); case T_STRING_LITERAL: return parse_string_const(); case T_IDENTIFIER: return parse_reference(); + case T___FUNCTION__: + return parse_function_keyword(); + case T___PRETTY_FUNCTION__: + return parse_pretty_function_keyword(); + case T___builtin_offsetof: + return parse_offsetof(); + case T___builtin_va_arg: + return parse_va_arg(); + case T___builtin_expect: + case T___builtin_va_start: + case T___builtin_va_end: + return parse_builtin_symbol(); + case '(': return parse_brace_expression(); } - /* TODO: error message */ - return NULL; + parser_print_error_prefix(); + fprintf(stderr, "unexpected token "); + print_token(stderr, &token); + fprintf(stderr, "\n"); + eat_statement(); + + expression_t *expression = allocate_ast_zero(sizeof(expression[0])); + expression->type = EXPR_INVALID; + expression->datatype = type_void; + + return expression; } -static -expression_t *parse_array_expression(unsigned precedence, - expression_t *array_ref) +static expression_t *parse_array_expression(unsigned precedence, + expression_t *array_ref) { (void) precedence; @@ -1342,55 +2146,79 @@ expression_t *parse_array_expression(unsigned precedence, array_access_expression_t *array_access = allocate_ast_zero(sizeof(array_access[0])); - array_access->expression.type = EXPR_ARRAY_ACCESS; - array_access->array_ref = array_ref; - array_access->index = parse_expression(); + array_access->expression.type = EXPR_ARRAY_ACCESS; + 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; + array_access->expression.datatype = pointer->points_to; + } else { + parser_print_error_prefix(); + fprintf(stderr, "array access on object with non-pointer type "); + print_type(array_type); + fprintf(stderr, "\n"); + } + } if(token.type != ']') { parse_error_expected("Problem while parsing array access", ']', 0); - return NULL; + return (expression_t*) array_access; } next_token(); return (expression_t*) array_access; } -static -type_t *get_expression_type(const expression_t *expression) +static bool is_declaration_specifier(const token_t *token, + bool only_type_specifiers) { - (void) expression; - /* TODO */ - return NULL; + switch(token->type) { + TYPE_SPECIFIERS + return 1; + case T_IDENTIFIER: + return is_typedef_symbol(token->v.symbol); + STORAGE_CLASSES + TYPE_QUALIFIERS + if(only_type_specifiers) + return 0; + return 1; + + default: + return 0; + } } -static -expression_t *parse_sizeof(unsigned precedence) +static expression_t *parse_sizeof(unsigned precedence) { eat(T_sizeof); sizeof_expression_t *sizeof_expression = allocate_ast_zero(sizeof(sizeof_expression[0])); - sizeof_expression->expression.type = EXPR_SIZEOF; + sizeof_expression->expression.type = EXPR_SIZEOF; + sizeof_expression->expression.datatype = type_size_t; - if(token.type == '(' /* && LA1 is type_specifier */) { + if(token.type == '(' && is_declaration_specifier(look_ahead(1), true)) { next_token(); sizeof_expression->type = parse_typename(); expect(')'); } else { - expression_t *expression = parse_sub_expression(precedence); - sizeof_expression->type = get_expression_type(expression); + expression_t *expression = parse_sub_expression(precedence); + sizeof_expression->type = expression->datatype; + sizeof_expression->size_expression = expression; } return (expression_t*) sizeof_expression; } -static -expression_t *parse_select_expression(unsigned precedence, - expression_t *compound) +static expression_t *parse_select_expression(unsigned precedence, + expression_t *compound) { (void) precedence; - assert(token.type == '.' || token.type == T_SELECT); + assert(token.type == '.' || token.type == T_MINUSGREATER); next_token(); select_expression_t *select = allocate_ast_zero(sizeof(select[0])); @@ -1398,10 +2226,11 @@ 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 compound select", - T_IDENTIFIER, 0); - return NULL; + parse_error_expected("Problem while parsing select", T_IDENTIFIER, 0); + return (expression_t*) select; } select->symbol = token.v.symbol; next_token(); @@ -1409,15 +2238,14 @@ expression_t *parse_select_expression(unsigned precedence, return (expression_t*) select; } -static -expression_t *parse_call_expression(unsigned precedence, - expression_t *expression) +static expression_t *parse_call_expression(unsigned precedence, + expression_t *expression) { (void) precedence; call_expression_t *call = allocate_ast_zero(sizeof(call[0])); - call->expression.type = EXPR_CALL; - call->method = expression; + call->expression.type = EXPR_CALL; + call->function = expression; /* parse arguments */ eat('('); @@ -1425,10 +2253,10 @@ expression_t *parse_call_expression(unsigned precedence, if(token.type != ')') { call_argument_t *last_argument = NULL; - while(1) { + while(true) { call_argument_t *argument = allocate_ast_zero(sizeof(argument[0])); - argument->expression = parse_expression(); + argument->expression = parse_assignment_expression(); if(last_argument == NULL) { call->arguments = argument; } else { @@ -1443,110 +2271,294 @@ 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(type == NULL || type->type != TYPE_FUNCTION) { + 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; + } + } + return (expression_t*) call; } -static -expression_t *parse_conditional_expression(unsigned precedence, - expression_t *expression) +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) +{ + /* TODO... */ + (void) type2; + return (type_t*) type1; +} + +static expression_t *parse_conditional_expression(unsigned precedence, + expression_t *expression) { eat('?'); conditional_expression_t *conditional = allocate_ast_zero(sizeof(conditional[0])); + conditional->expression.type = EXPR_CONDITIONAL; conditional->condition = expression; + /* 6.5.15.2 */ + type_t *condition_type = conditional->condition->datatype; + if(condition_type != NULL) { + if(!is_type_scalar(condition_type)) { + type_error("expected a scalar type", expression->source_position, + condition_type); + } + } + conditional->true_expression = parse_expression(); expect(':'); conditional->false_expression = parse_sub_expression(precedence); + type_t *true_type = conditional->true_expression->datatype; + if(true_type == NULL) + return (expression_t*) conditional; + type_t *false_type = conditional->false_expression->datatype; + if(false_type == NULL) + return (expression_t*) conditional; + + /* 6.5.15.3 */ + 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); + /* TODO: create implicit convs if necessary */ + conditional->expression.datatype = result; + } else if(true_type->type == TYPE_POINTER && + false_type->type == TYPE_POINTER && + true /* TODO compatible points_to types */) { + /* TODO */ + } else if(/* (is_null_ptr_const(true_type) && false_type->type == TYPE_POINTER) + || (is_null_ptr_const(false_type) && + true_type->type == TYPE_POINTER) TODO*/ false) { + /* TODO */ + } else if(/* 1 is pointer to object type, other is void* */ false) { + /* TODO */ + } else { + type_error_incompatible("problem while parsing conditional", + expression->source_position, true_type, + false_type); + } + return (expression_t*) conditional; } -#define CREATE_UNARY_EXPRESSION_PARSER(token_type, unexpression_type) \ -static \ -expression_t *parse_##unexpression_type(unsigned precedence) \ -{ \ - eat(token_type); \ - \ - unary_expression_t *unary_expression \ - = allocate_ast_zero(sizeof(unary_expression[0])); \ - unary_expression->expression.type = EXPR_UNARY; \ - unary_expression->type = unexpression_type; \ - unary_expression->value = parse_sub_expression(precedence); \ - \ - return (expression_t*) unary_expression; \ -} - -CREATE_UNARY_EXPRESSION_PARSER('-', UNEXPR_NEGATE) -CREATE_UNARY_EXPRESSION_PARSER('+', UNEXPR_PLUS) -CREATE_UNARY_EXPRESSION_PARSER('!', UNEXPR_NOT) -CREATE_UNARY_EXPRESSION_PARSER('*', UNEXPR_DEREFERENCE) -CREATE_UNARY_EXPRESSION_PARSER('&', UNEXPR_TAKE_ADDRESS) -CREATE_UNARY_EXPRESSION_PARSER('~', UNEXPR_BITWISE_NEGATE) -CREATE_UNARY_EXPRESSION_PARSER(T_PLUSPLUS, UNEXPR_PREFIX_INCREMENT) -CREATE_UNARY_EXPRESSION_PARSER(T_MINUSMINUS, UNEXPR_PREFIX_DECREMENT) - -#define CREATE_UNARY_POSTFIX_EXPRESSION_PARSER(token_type, unexpression_type) \ -static \ -expression_t *parse_##unexpression_type(unsigned precedence, \ - expression_t *left) \ +static expression_t *parse_extension(unsigned precedence) +{ + eat(T___extension__); + + /* TODO enable extensions */ + + return parse_sub_expression(precedence); +} + +static type_t *get_unexpr_arithmetic_type(const expression_t *expression) +{ + /* TODO */ + return expression->datatype; +} + +static type_t *get_unexpr_dereference_type(const expression_t *expression) +{ + (void) expression; + /* TODO... */ + return NULL; +} + +static type_t *get_unexpr_take_addr_type(const expression_t *expression) +{ + type_t *type = expression->datatype; + return make_pointer_type(type, 0); +} + +#define CREATE_UNARY_EXPRESSION_PARSER(token_type, unexpression_type, tfunc) \ +static expression_t *parse_##unexpression_type(unsigned precedence) \ +{ \ + eat(token_type); \ + \ + unary_expression_t *unary_expression \ + = allocate_ast_zero(sizeof(unary_expression[0])); \ + unary_expression->expression.type = EXPR_UNARY; \ + unary_expression->type = unexpression_type; \ + unary_expression->value = parse_sub_expression(precedence); \ + unary_expression->expression.datatype = tfunc(unary_expression->value); \ + \ + 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_BITWISE_NEGATE, + get_unexpr_arithmetic_type) +CREATE_UNARY_EXPRESSION_PARSER(T_PLUSPLUS, UNEXPR_PREFIX_INCREMENT, + get_unexpr_arithmetic_type) +CREATE_UNARY_EXPRESSION_PARSER(T_MINUSMINUS, UNEXPR_PREFIX_DECREMENT, + get_unexpr_arithmetic_type) + +#define CREATE_UNARY_POSTFIX_EXPRESSION_PARSER(token_type, unexpression_type, \ + tfunc) \ +static expression_t *parse_##unexpression_type(unsigned precedence, \ + expression_t *left) \ { \ (void) precedence; \ eat(token_type); \ \ unary_expression_t *unary_expression \ = allocate_ast_zero(sizeof(unary_expression[0])); \ - unary_expression->expression.type = EXPR_UNARY; \ - unary_expression->type = unexpression_type; \ - unary_expression->value = left; \ + unary_expression->expression.type = EXPR_UNARY; \ + unary_expression->type = unexpression_type; \ + unary_expression->value = left; \ + unary_expression->expression.datatype = tfunc(left); \ \ return (expression_t*) unary_expression; \ } -CREATE_UNARY_POSTFIX_EXPRESSION_PARSER(T_PLUSPLUS, UNEXPR_POSTFIX_INCREMENT) -CREATE_UNARY_POSTFIX_EXPRESSION_PARSER(T_MINUSMINUS, UNEXPR_POSTFIX_DECREMENT) - -#define CREATE_BINEXPR_PARSER(token_type, binexpression_type) \ -static \ -expression_t *parse_##binexpression_type(unsigned precedence, \ - expression_t *left) \ -{ \ - eat(token_type); \ - \ - expression_t *right = parse_sub_expression(precedence); \ - \ - binary_expression_t *binexpr \ - = allocate_ast_zero(sizeof(binexpr[0])); \ - binexpr->expression.type = EXPR_BINARY; \ - binexpr->type = binexpression_type; \ - binexpr->left = left; \ - binexpr->right = right; \ - \ - return (expression_t*) binexpr; \ -} - -CREATE_BINEXPR_PARSER('*', BINEXPR_MUL) -CREATE_BINEXPR_PARSER('/', BINEXPR_DIV) -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_SLASHEQUAL, 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) - -static -expression_t *parse_sub_expression(unsigned precedence) +CREATE_UNARY_POSTFIX_EXPRESSION_PARSER(T_PLUSPLUS, UNEXPR_POSTFIX_INCREMENT, + get_unexpr_arithmetic_type) +CREATE_UNARY_POSTFIX_EXPRESSION_PARSER(T_MINUSMINUS, UNEXPR_POSTFIX_DECREMENT, + get_unexpr_arithmetic_type) + +static type_t *get_binexpr_int_type(const expression_t *left, + const expression_t *right) +{ + (void) left; + (void) right; + return type_int; +} + +static type_t *get_binexpr_arithmetic_type(const expression_t *left, + const expression_t *right) +{ + (void) right; + return left->datatype; +} + +static type_t *get_binexpr_arithmetic_assign_type(const expression_t *left, + const expression_t *right) +{ + (void) right; + /* TODO */ + return left->datatype; +} + +static type_t *get_binexpr_right_type(const expression_t *left, + const expression_t *right) +{ + (void) left; + return right->datatype; +} + +#define CREATE_BINEXPR_PARSER(token_type, binexpression_type, tfunc) \ +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; \ + binexpr->expression.datatype = tfunc(left, right); \ + \ + return (expression_t*) binexpr; \ +} + +CREATE_BINEXPR_PARSER(',', BINEXPR_COMMA, get_binexpr_right_type) +CREATE_BINEXPR_PARSER('*', BINEXPR_MUL, get_binexpr_arithmetic_type) +CREATE_BINEXPR_PARSER('/', BINEXPR_DIV, get_binexpr_arithmetic_type) +CREATE_BINEXPR_PARSER('%', BINEXPR_MOD, get_binexpr_arithmetic_type) +CREATE_BINEXPR_PARSER('+', BINEXPR_ADD, get_binexpr_arithmetic_type) +CREATE_BINEXPR_PARSER('-', BINEXPR_SUB, get_binexpr_arithmetic_type) +CREATE_BINEXPR_PARSER('<', BINEXPR_LESS, get_binexpr_int_type) +CREATE_BINEXPR_PARSER('>', BINEXPR_GREATER, get_binexpr_int_type) +CREATE_BINEXPR_PARSER('=', BINEXPR_ASSIGN, get_binexpr_right_type) +CREATE_BINEXPR_PARSER(T_EQUALEQUAL, BINEXPR_EQUAL, + get_binexpr_int_type) +CREATE_BINEXPR_PARSER(T_EXCLAMATIONMARKEQUAL, BINEXPR_NOTEQUAL, + get_binexpr_int_type) +CREATE_BINEXPR_PARSER(T_LESSEQUAL, BINEXPR_LESSEQUAL, + get_binexpr_int_type) +CREATE_BINEXPR_PARSER(T_GREATEREQUAL, BINEXPR_GREATEREQUAL, + get_binexpr_int_type) +CREATE_BINEXPR_PARSER('&', BINEXPR_BITWISE_AND, get_binexpr_arithmetic_type) +CREATE_BINEXPR_PARSER('|', BINEXPR_BITWISE_OR, get_binexpr_arithmetic_type) +CREATE_BINEXPR_PARSER('^', BINEXPR_BITWISE_XOR, get_binexpr_arithmetic_type) +CREATE_BINEXPR_PARSER(T_ANDAND, BINEXPR_LOGICAL_AND, + get_binexpr_int_type) +CREATE_BINEXPR_PARSER(T_PIPEPIPE, BINEXPR_LOGICAL_OR, + get_binexpr_int_type) +CREATE_BINEXPR_PARSER(T_LESSLESS, BINEXPR_SHIFTLEFT, + get_binexpr_arithmetic_type) +CREATE_BINEXPR_PARSER(T_GREATERGREATER, BINEXPR_SHIFTRIGHT, + get_binexpr_arithmetic_type) +CREATE_BINEXPR_PARSER(T_PLUSEQUAL, BINEXPR_ADD_ASSIGN, + get_binexpr_arithmetic_assign_type) +CREATE_BINEXPR_PARSER(T_MINUSEQUAL, BINEXPR_SUB_ASSIGN, + get_binexpr_arithmetic_assign_type) +CREATE_BINEXPR_PARSER(T_ASTERISKEQUAL, BINEXPR_MUL_ASSIGN, + get_binexpr_arithmetic_assign_type) +CREATE_BINEXPR_PARSER(T_SLASHEQUAL, BINEXPR_DIV_ASSIGN, + get_binexpr_arithmetic_assign_type) +CREATE_BINEXPR_PARSER(T_PERCENTEQUAL, BINEXPR_MOD_ASSIGN, + get_binexpr_arithmetic_assign_type) +CREATE_BINEXPR_PARSER(T_LESSLESSEQUAL, BINEXPR_SHIFTLEFT_ASSIGN, + get_binexpr_arithmetic_assign_type) +CREATE_BINEXPR_PARSER(T_GREATERGREATEREQUAL, BINEXPR_SHIFTRIGHT_ASSIGN, + get_binexpr_arithmetic_assign_type) +CREATE_BINEXPR_PARSER(T_ANDEQUAL, BINEXPR_BITWISE_AND_ASSIGN, + get_binexpr_arithmetic_assign_type) +CREATE_BINEXPR_PARSER(T_PIPEEQUAL, BINEXPR_BITWISE_OR_ASSIGN, + get_binexpr_arithmetic_assign_type) +CREATE_BINEXPR_PARSER(T_CARETEQUAL, BINEXPR_BITWISE_XOR_ASSIGN, + get_binexpr_arithmetic_assign_type) + +static expression_t *parse_sub_expression(unsigned precedence) { if(token.type < 0) { return expected_expression_error(); @@ -1562,10 +2574,10 @@ expression_t *parse_sub_expression(unsigned precedence) } else { left = parse_primary_expression(); } - if(left != NULL) - left->source_position = source_position; + assert(left != NULL); + left->source_position = source_position; - while(1) { + while(true) { if(token.type < 0) { return expected_expression_error(); } @@ -1577,23 +2589,24 @@ expression_t *parse_sub_expression(unsigned precedence) break; left = parser->infix_parser(parser->infix_precedence, left); - if(left != NULL) - left->source_position = source_position; + + assert(left != NULL); + assert(left->type != EXPR_INVALID); + left->source_position = source_position; } return left; } -static -expression_t *parse_expression(void) +static expression_t *parse_expression(void) { return parse_sub_expression(1); } -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]; @@ -1607,8 +2620,9 @@ 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]; @@ -1623,39 +2637,58 @@ void register_expression_infix_parser(parse_expression_infix_function parser, entry->infix_precedence = precedence; } -static -void init_expression_parsers(void) +static void init_expression_parsers(void) { memset(&expression_parsers, 0, sizeof(expression_parsers)); - register_expression_infix_parser(parse_BINEXPR_MUL, '*', 16); - register_expression_infix_parser(parse_BINEXPR_DIV, '/', 16); - register_expression_infix_parser(parse_BINEXPR_SHIFTLEFT, - T_LESSLESS, 16); + register_expression_infix_parser(parse_BINEXPR_MUL, '*', 16); + register_expression_infix_parser(parse_BINEXPR_DIV, '/', 16); + register_expression_infix_parser(parse_BINEXPR_MOD, '%', 16); + register_expression_infix_parser(parse_BINEXPR_SHIFTLEFT, T_LESSLESS, 16); register_expression_infix_parser(parse_BINEXPR_SHIFTRIGHT, - T_GREATERGREATER, 16); - register_expression_infix_parser(parse_BINEXPR_ADD, '+', 15); - register_expression_infix_parser(parse_BINEXPR_SUB, '-', 15); - register_expression_infix_parser(parse_BINEXPR_LESS, '<', 14); - register_expression_infix_parser(parse_BINEXPR_GREATER, '>', 14); - register_expression_infix_parser(parse_BINEXPR_LESSEQUAL, T_LESSEQUAL, 14); + T_GREATERGREATER, 16); + register_expression_infix_parser(parse_BINEXPR_ADD, '+', 15); + register_expression_infix_parser(parse_BINEXPR_SUB, '-', 15); + register_expression_infix_parser(parse_BINEXPR_LESS, '<', 14); + register_expression_infix_parser(parse_BINEXPR_GREATER, '>', 14); + register_expression_infix_parser(parse_BINEXPR_LESSEQUAL, T_LESSEQUAL, 14); register_expression_infix_parser(parse_BINEXPR_GREATEREQUAL, - T_GREATEREQUAL, 14); + T_GREATEREQUAL, 14); register_expression_infix_parser(parse_BINEXPR_EQUAL, T_EQUALEQUAL, 13); register_expression_infix_parser(parse_BINEXPR_NOTEQUAL, - T_EXCLAMATIONMARKEQUAL, 13); + T_EXCLAMATIONMARKEQUAL, 13); register_expression_infix_parser(parse_BINEXPR_BITWISE_AND, '&', 12); register_expression_infix_parser(parse_BINEXPR_BITWISE_XOR, '^', 11); register_expression_infix_parser(parse_BINEXPR_BITWISE_OR, '|', 10); register_expression_infix_parser(parse_BINEXPR_LOGICAL_AND, T_ANDAND, 9); register_expression_infix_parser(parse_BINEXPR_LOGICAL_OR, T_PIPEPIPE, 8); register_expression_infix_parser(parse_conditional_expression, '?', 7); - register_expression_infix_parser(parse_BINEXPR_ASSIGN, T_EQUAL, 2); + register_expression_infix_parser(parse_BINEXPR_ASSIGN, '=', 2); + register_expression_infix_parser(parse_BINEXPR_ADD_ASSIGN, T_PLUSEQUAL, 2); + register_expression_infix_parser(parse_BINEXPR_SUB_ASSIGN, T_MINUSEQUAL, 2); + register_expression_infix_parser(parse_BINEXPR_MUL_ASSIGN, + T_ASTERISKEQUAL, 2); + register_expression_infix_parser(parse_BINEXPR_DIV_ASSIGN, T_SLASHEQUAL, 2); + register_expression_infix_parser(parse_BINEXPR_MOD_ASSIGN, + T_PERCENTEQUAL, 2); + register_expression_infix_parser(parse_BINEXPR_SHIFTLEFT_ASSIGN, + T_LESSLESSEQUAL, 2); + register_expression_infix_parser(parse_BINEXPR_SHIFTRIGHT_ASSIGN, + T_GREATERGREATEREQUAL, 2); + register_expression_infix_parser(parse_BINEXPR_BITWISE_AND_ASSIGN, + T_ANDEQUAL, 2); + register_expression_infix_parser(parse_BINEXPR_BITWISE_OR_ASSIGN, + T_PIPEEQUAL, 2); + register_expression_infix_parser(parse_BINEXPR_BITWISE_XOR_ASSIGN, + T_CARETEQUAL, 2); + + register_expression_infix_parser(parse_BINEXPR_COMMA, ',', 1); register_expression_infix_parser(parse_array_expression, '[', 30); register_expression_infix_parser(parse_call_expression, '(', 30); register_expression_infix_parser(parse_select_expression, '.', 30); - register_expression_infix_parser(parse_select_expression, T_SELECT, 30); + register_expression_infix_parser(parse_select_expression, + T_MINUSGREATER, 30); register_expression_infix_parser(parse_UNEXPR_POSTFIX_INCREMENT, T_PLUSPLUS, 30); register_expression_infix_parser(parse_UNEXPR_POSTFIX_DECREMENT, @@ -1670,32 +2703,40 @@ void init_expression_parsers(void) register_expression_parser(parse_UNEXPR_PREFIX_INCREMENT, T_PLUSPLUS, 25); register_expression_parser(parse_UNEXPR_PREFIX_DECREMENT, T_MINUSMINUS, 25); register_expression_parser(parse_sizeof, T_sizeof, 25); + register_expression_parser(parse_extension, T___extension__, 25); } -static -statement_t *parse_case_statement(void) +static statement_t *parse_case_statement(void) { eat(T_case); - parse_expression(); + case_label_statement_t *label = allocate_ast_zero(sizeof(label[0])); + label->statement.type = STATEMENT_CASE_LABEL; + label->statement.source_position = token.source_position; + + label->expression = parse_expression(); + expect(':'); - parse_statement(); + label->statement.next = parse_statement(); - return NULL; + return (statement_t*) label; } -static -statement_t *parse_default_statement(void) +static statement_t *parse_default_statement(void) { eat(T_default); + + case_label_statement_t *label = allocate_ast_zero(sizeof(label[0])); + label->statement.type = STATEMENT_CASE_LABEL; + label->statement.source_position = token.source_position; + expect(':'); - parse_statement(); + label->statement.next = parse_statement(); - return NULL; + return (statement_t*) label; } -static -statement_t *parse_label_statement(void) +static statement_t *parse_label_statement(void) { eat(T_IDENTIFIER); expect(':'); @@ -1704,85 +2745,120 @@ statement_t *parse_label_statement(void) return NULL; } -static -statement_t *parse_if(void) +static statement_t *parse_if(void) { eat(T_if); + + if_statement_t *statement = allocate_ast_zero(sizeof(statement[0])); + statement->statement.type = STATEMENT_IF; + statement->statement.source_position = token.source_position; + expect('('); - parse_expression(); + statement->condition = parse_expression(); expect(')'); - parse_statement(); + statement->true_statement = parse_statement(); if(token.type == T_else) { next_token(); - parse_statement(); + statement->false_statement = parse_statement(); } - return NULL; + return (statement_t*) statement; } -static -statement_t *parse_switch(void) +static statement_t *parse_switch(void) { eat(T_switch); + + switch_statement_t *statement = allocate_ast_zero(sizeof(statement[0])); + statement->statement.type = STATEMENT_SWITCH; + statement->statement.source_position = token.source_position; + expect('('); - parse_expression(); + statement->expression = parse_expression(); expect(')'); - parse_statement(); + statement->body = parse_statement(); - return NULL; + return (statement_t*) statement; } -static -statement_t *parse_while(void) +static statement_t *parse_while(void) { eat(T_while); + + while_statement_t *statement = allocate_ast_zero(sizeof(statement[0])); + statement->statement.type = STATEMENT_WHILE; + statement->statement.source_position = token.source_position; + expect('('); - parse_expression(); + statement->condition = parse_expression(); expect(')'); - parse_statement(); + statement->body = parse_statement(); - return NULL; + return (statement_t*) statement; } -static -statement_t *parse_do(void) +static statement_t *parse_do(void) { eat(T_do); - parse_statement(); + + do_while_statement_t *statement = allocate_ast_zero(sizeof(statement[0])); + statement->statement.type = STATEMENT_DO_WHILE; + statement->statement.source_position = token.source_position; + + statement->body = parse_statement(); expect(T_while); expect('('); - parse_expression(); + statement->condition = parse_expression(); expect(')'); + expect(';'); - return NULL; + return (statement_t*) statement; } -static -statement_t *parse_for(void) +static statement_t *parse_for(void) { eat(T_for); + + for_statement_t *statement = allocate_ast_zero(sizeof(statement[0])); + statement->statement.type = STATEMENT_FOR; + statement->statement.source_position = token.source_position; + expect('('); + + int top = environment_top(); + context_t *last_context = context; + set_context(&statement->context); + if(token.type != ';') { - /* TODO not correct... this could also be a declaration */ - parse_expression(); + if(is_declaration_specifier(&token, false)) { + parse_declaration(); + } else { + statement->initialisation = parse_expression(); + expect(';'); + } + } else { + expect(';'); } - expect(';'); + if(token.type != ';') { - parse_expression(); + statement->condition = parse_expression(); } expect(';'); if(token.type != ')') { - parse_expression(); + statement->step = parse_expression(); } expect(')'); - parse_statement(); + statement->body = parse_statement(); - return NULL; + assert(context == &statement->context); + set_context(last_context); + environment_pop_to(top); + + return (statement_t*) statement; } -static -statement_t *parse_goto(void) +static statement_t *parse_goto(void) { eat(T_goto); expect(T_IDENTIFIER); @@ -1791,8 +2867,7 @@ statement_t *parse_goto(void) return NULL; } -static -statement_t *parse_continue(void) +static statement_t *parse_continue(void) { eat(T_continue); expect(';'); @@ -1804,22 +2879,26 @@ statement_t *parse_continue(void) return statement; } -static -statement_t *parse_break(void) +static statement_t *parse_break(void) { eat(T_break); expect(';'); - return NULL; + statement_t *statement = allocate_ast_zero(sizeof(statement[0])); + statement->source_position = token.source_position; + statement->type = STATEMENT_BREAK; + + return statement; } -static -statement_t *parse_return(void) +static statement_t *parse_return(void) { eat(T_return); return_statement_t *statement = allocate_ast_zero(sizeof(statement[0])); - statement->statement.type = STATEMENT_RETURN; + + statement->statement.type = STATEMENT_RETURN; + statement->statement.source_position = token.source_position; if(token.type != ';') { statement->return_value = parse_expression(); } @@ -1828,24 +2907,50 @@ statement_t *parse_return(void) return (statement_t*) statement; } -static -statement_t *parse_declaration_statement(void) +static statement_t *parse_declaration_statement(void) { - parse_declaration(); - return NULL; + declaration_t *before = last_declaration; + + declaration_statement_t *statement + = allocate_ast_zero(sizeof(statement[0])); + statement->statement.type = STATEMENT_DECLARATION; + statement->statement.source_position = token.source_position; + + declaration_specifiers_t specifiers; + memset(&specifiers, 0, sizeof(specifiers)); + parse_declaration_specifiers(&specifiers); + + if(token.type == ';') { + eat(';'); + } else { + parse_init_declarators(&specifiers); + } + + if(before == NULL) { + statement->declarations_begin = context->declarations; + } else { + statement->declarations_begin = before->next; + } + statement->declarations_end = last_declaration; + + return (statement_t*) statement; } -static -statement_t *parse_expression_statement(void) +static statement_t *parse_expression_statement(void) { - parse_expression(); - return NULL; + expression_statement_t *statement = allocate_ast_zero(sizeof(statement[0])); + statement->statement.type = STATEMENT_EXPRESSION; + statement->statement.source_position = token.source_position; + + statement->expression = parse_expression(); + + expect(';'); + + return (statement_t*) statement; } -static -statement_t *parse_statement(void) +static statement_t *parse_statement(void) { - declaration_t *declaration; statement_t *statement = NULL; /* declaration or statement */ @@ -1899,18 +3004,17 @@ statement_t *parse_statement(void) break; case ';': + next_token(); statement = NULL; break; case T_IDENTIFIER: - if(la(1)->type == ':') { + if(look_ahead(1)->type == ':') { statement = parse_label_statement(); break; } - declaration = token.v.symbol->declaration; - if(declaration != NULL && - declaration->storage_class == STORAGE_CLASS_TYPEDEF) { + if(is_typedef_symbol(token.v.symbol)) { statement = parse_declaration_statement(); break; } @@ -1918,22 +3022,37 @@ statement_t *parse_statement(void) statement = parse_expression_statement(); break; + case T___extension__: + /* this can be a prefix to a declaration or an expression statement */ + /* we simply eat it now and parse the rest with tail recursion */ + do { + next_token(); + } while(token.type == T___extension__); + statement = parse_statement(); + break; + DECLARATION_START statement = parse_declaration_statement(); break; + + default: + statement = parse_expression_statement(); + break; } + assert(statement == NULL || statement->source_position.input_name != NULL); + return statement; } -static -statement_t *parse_compound_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.type = STATEMENT_COMPOUND; + compound_statement->statement.source_position = token.source_position; int top = environment_top(); context_t *last_context = context; @@ -1943,12 +3062,18 @@ statement_t *parse_compound_statement(void) while(token.type != '}' && token.type != T_EOF) { statement_t *statement = parse_statement(); + if(statement == NULL) + continue; if(last_statement != NULL) { last_statement->next = statement; } else { compound_statement->statements = statement; } + + while(statement->next != NULL) + statement = statement->next; + last_statement = statement; } @@ -1961,11 +3086,13 @@ statement_t *parse_compound_statement(void) return (statement_t*) compound_statement; } -static -translation_unit_t *parse_translation_unit(void) +static translation_unit_t *parse_translation_unit(void) { translation_unit_t *unit = allocate_ast_zero(sizeof(unit[0])); + assert(global_context == NULL); + global_context = &unit->context; + assert(context == NULL); set_context(&unit->context); @@ -1977,13 +3104,17 @@ translation_unit_t *parse_translation_unit(void) context = NULL; last_declaration = NULL; + assert(global_context == &unit->context); + global_context = NULL; + return unit; } translation_unit_t *parse(void) { - obstack_init(&environment_obstack); - environment_stack = NEW_ARR_F(environment_entry_t*, 0); + environment_stack = NEW_ARR_F(stack_entry_t, 0); + + type_set_output(stderr); lookahead_bufpos = 0; for(int i = 0; i < MAX_LOOKAHEAD + 2; ++i) { @@ -1992,7 +3123,6 @@ translation_unit_t *parse(void) translation_unit_t *unit = parse_translation_unit(); DEL_ARR_F(environment_stack); - obstack_free(&environment_obstack, NULL); return unit; } @@ -2001,6 +3131,13 @@ 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); } void exit_parser(void)