#include <stdarg.h>
#include "parser.h"
-#include "lexer_t.h"
+#include "lexer.h"
#include "token_t.h"
#include "type_t.h"
#include "type_hash.h"
#include "adt/error.h"
#include "adt/array.h"
-#define PRINT_TOKENS
+//#define PRINT_TOKENS
+//#define ABORT_ON_ERROR
+#define MAX_LOOKAHEAD 2
struct environment_entry_t {
- symbol_t *symbol;
- environment_entry_t *old_entry;
- declaration_t *declaration;
- unsigned short old_symbol_ID;
+ symbol_t *symbol;
+ declaration_t *old_declaration;
+ const void *old_context;
};
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 translation_unit_t *translation_unit = NULL;
-static block_statement_t *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);
return res;
}
-/**
- * pushs an environment_entry on the environment stack and links the
- * corresponding symbol to the new entry
- */
-static inline
-environment_entry_t *environment_push(symbol_t *symbol)
-{
- 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;
-
- entry->old_entry = symbol->thing;
- entry->symbol = symbol;
- symbol->thing = entry;
-
- return entry;
-}
-
-/**
- * pops symbols from the environment stack until @p new_top is the top element
- */
-static inline
-void environment_pop_to(size_t new_top)
-{
- environment_entry_t *entry = NULL;
- size_t top = ARR_LEN(environment_stack);
- size_t i;
-
- if(new_top == top)
- return;
-
- assert(new_top < top);
- i = top;
- do {
- entry = environment_stack[i - 1];
-
- symbol_t *symbol = entry->symbol;
-
-#if 0
- if(entry->type == ENTRY_LOCAL_VARIABLE
- && entry->e.variable->refs == 0) {
- variable_declaration_statement_t *variable = entry->e.variable;
- print_warning_prefix(env, variable->statement.source_position);
- fprintf(stderr, "variable '%s' was declared but never read\n",
- symbol->string);
- }
-#endif
-
- if(entry->declaration->storage_class == STORAGE_CLASS_TYPEDEF) {
- fprintf(stderr, "pop typename '%s'\n", entry->symbol->string);
- symbol->ID = entry->old_symbol_ID;
- }
-
- assert(symbol->thing == entry);
- symbol->thing = entry->old_entry;
-
- --i;
- } while(i != new_top);
- obstack_free(&environment_obstack, entry);
-
- ARR_SHRINKLEN(environment_stack, (int) new_top);
-}
-
/**
* returns the top element of the environment stack
*/
static inline
-size_t environment_top()
+size_t environment_top(void)
{
return ARR_LEN(environment_stack);
}
static inline
void next_token(void)
{
- lexer_next_token(&token);
+ token = lookahead_buffer[lookahead_bufpos];
+ lookahead_buffer[lookahead_bufpos] = lexer_token;
+ lexer_next_token();
+
+ lookahead_bufpos = (lookahead_bufpos+1) % MAX_LOOKAHEAD;
#ifdef PRINT_TOKENS
print_token(stderr, &token);
#endif
}
+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)
{
next_token();
}
-void parser_print_error_prefix(void)
+void parser_print_error_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
+}
+
+void parser_print_error_prefix(void)
+{
+ parser_print_error_prefix_pos(token.source_position);
}
static
fprintf(stderr, "\n");
}
-static
-void eat_until_semi(void)
+static void eat_block(void)
+{
+ if(token.type == '{')
+ next_token();
+
+ while(token.type != '}') {
+ if(token.type == T_EOF)
+ return;
+ if(token.type == '{') {
+ eat_block();
+ continue;
+ }
+ 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 == '{') {
+ eat_block();
+ continue;
+ }
+ next_token();
}
- next_token();
+ eat(')');
}
#define expect(expected) \
if(UNLIKELY(token.type != (expected))) { \
parse_error_expected(NULL, (expected), 0); \
- eat_until_semi(); \
+ eat_statement(); \
return NULL; \
} \
next_token();
#define expect_void(expected) \
if(UNLIKELY(token.type != (expected))) { \
parse_error_expected(NULL, (expected), 0); \
- eat_until_semi(); \
+ eat_statement(); \
return; \
} \
next_token();
+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 = declaration;
+}
+
+/**
+ * called when we find a 2nd declarator for an identifier we already have a
+ * declarator for
+ */
+static int is_compatible_declaration (declaration_t *declaration,
+ declaration_t *previous)
+{
+ /* TODO: not correct yet */
+ return declaration->type == previous->type;
+}
+
+/**
+ * pushs an environment_entry on the environment stack and links the
+ * corresponding symbol to the new entry
+ */
+static inline
+declaration_t *environment_push(declaration_t *declaration, const void *context)
+{
+ 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;
+
+ assert(declaration->source_position.input_name != NULL);
+
+ symbol_t *symbol = declaration->symbol;
+ assert(declaration != symbol->declaration);
+
+ if(symbol->context == context) {
+ declaration_t *previous_declaration = symbol->declaration;
+ if(symbol->declaration != NULL) {
+ if(!is_compatible_declaration(declaration, previous_declaration)) {
+ parser_print_error_prefix_pos(declaration->source_position);
+ fprintf(stderr, "definition of symbol '%s' with type ",
+ declaration->symbol->string);
+ print_type(declaration->type, NULL);
+ 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, NULL);
+ fputc('\n', stderr);
+ }
+ return previous_declaration;
+ }
+ }
+
+ 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)
+{
+ environment_entry_t *entry = NULL;
+ size_t top = ARR_LEN(environment_stack);
+ size_t i;
+
+ if(new_top == top)
+ return;
+
+ assert(new_top < top);
+ i = top;
+ do {
+ entry = environment_stack[i - 1];
+
+ symbol_t *symbol = entry->symbol;
+
+ symbol->declaration = entry->old_declaration;
+ symbol->context = entry->old_context;
+
+ --i;
+ } while(i != new_top);
+ obstack_free(&environment_obstack, entry);
+
+ ARR_SHRINKLEN(environment_stack, (int) new_top);
+}
+
+
+
static expression_t *parse_constant_expression(void)
{
/* TODO: not correct yet */
return parse_expression();
}
-static compound_entry_t *parse_compound_type_entries(void);
+static expression_t *parse_assignment_expression(void)
+{
+ /* TODO: not correct yet */
+ return parse_expression();
+}
+
+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 declaration_t *record_declaration(declaration_t *declaration);
typedef struct declaration_specifiers_t declaration_specifiers_t;
struct declaration_specifiers_t {
type_t *type;
};
-static type_t *parse_struct_specifier(void)
+static compound_type_t *find_compound_type(compound_type_t *types,
+ const symbol_t *symbol)
{
- eat(T_struct);
+ compound_type_t *type = types;
+ for( ; type != NULL; type = type->next) {
+ if(type->symbol == symbol)
+ return type;
+ }
+
+ return NULL;
+}
+
+static type_t *parse_compound_type_specifier(int is_struct)
+{
+ if(is_struct) {
+ eat(T_struct);
+ } else {
+ eat(T_union);
+ }
- compound_type_t *struct_type = allocate_type_zero(sizeof(struct_type[0]));
- struct_type->type.type = TYPE_COMPOUND_STRUCT;
- struct_type->source_position = source_position;
+ symbol_t *symbol = NULL;
+ compound_type_t *compound_type = NULL;
if(token.type == T_IDENTIFIER) {
- /* TODO */
+ symbol = token.v.symbol;
next_token();
- if(token.type == '{') {
- parse_compound_type_entries();
+
+ if(context != NULL) {
+ if(is_struct) {
+ compound_type = find_compound_type(context->structs, symbol);
+ } else {
+ compound_type = find_compound_type(context->unions, symbol);
+ }
}
- } else if(token.type == '{') {
+ } else if(token.type != '{') {
+ if(is_struct) {
+ parse_error_expected("problem while parsing struct type specifiers",
+ T_IDENTIFIER, '{', 0);
+ } else {
+ parse_error_expected("problem while parsing union type specifiers",
+ T_IDENTIFIER, '{', 0);
+ }
+
+ return NULL;
+ }
+
+ if(compound_type == NULL) {
+ compound_type = allocate_type_zero(sizeof(compound_type[0]));
+
+ if(is_struct) {
+ compound_type->type.type = TYPE_COMPOUND_STRUCT;
+ } else {
+ compound_type->type.type = TYPE_COMPOUND_UNION;
+ }
+ compound_type->source_position = token.source_position;
+ compound_type->symbol = symbol;
+ }
+
+ if(token.type == '{') {
+ if(compound_type->defined) {
+ parser_print_error_prefix();
+ fprintf(stderr, "multiple definition of %s %s\n",
+ is_struct ? "struct" : "union", symbol->string);
+ compound_type->context.declarations = NULL;
+ }
+ compound_type->defined = 1;
+
+ int top = environment_top();
+ context_t *last_context = context;
+ set_context(&compound_type->context);
+
parse_compound_type_entries();
- } else {
- parse_error_expected("problem while parsing struct type specifiers",
- T_IDENTIFIER, '{');
+
+ assert(context == &compound_type->context);
+ set_context(last_context);
+ environment_pop_to(top);
+ }
+
+ return (type_t*) compound_type;
+}
+
+static enum_entry_t *parse_enum_type_entries(void)
+{
+ eat('{');
+
+ if(token.type == '}') {
+ next_token();
+ parse_error("empty enum not allowed");
return NULL;
}
- return (type_t*) struct_type;
+ enum_entry_t *result = NULL;
+ enum_entry_t *last_entry = NULL;
+ do {
+ enum_entry_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 result;
+ }
+ entry->symbol = token.v.symbol;
+ next_token();
+
+ if(token.type == '=') {
+ next_token();
+ entry->value = parse_constant_expression();
+ }
+
+ if(last_entry != NULL) {
+ last_entry->next = entry;
+ } else {
+ result = entry;
+ }
+ last_entry = entry;
+
+ if(token.type != ',')
+ break;
+ next_token();
+ } while(token.type != '}');
+
+ expect('}');
+ return result;
}
-static type_t *parse_union_specifier(void)
+static type_t *parse_enum_specifier(void)
{
- eat(T_union);
+ 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;
- compound_type_t *union_type = allocate_type_zero(sizeof(union_type[0]));
- union_type->type.type = TYPE_COMPOUND_UNION;
- union_type->source_position = source_position;
+ /* TODO: rewrite to the same style as struct/union above to handle
+ * type identities correctly
+ */
if(token.type == T_IDENTIFIER) {
- /* TODO */
+ enum_type->symbol = token.v.symbol;
next_token();
if(token.type == '{') {
- parse_compound_type_entries();
+ enum_type->entries = parse_enum_type_entries();
}
} else if(token.type == '{') {
- parse_compound_type_entries();
+ enum_type->entries = parse_enum_type_entries();
} else {
- parse_error_expected("problem while parsing union type specifiers",
+ parse_error_expected("problem while parsing enum type specifiers",
T_IDENTIFIER, '{');
}
- return (type_t*) union_type;
+ return (type_t*) enum_type;
+}
+
+static
+const char *parse_string_literals(void)
+{
+ assert(token.type == T_STRING_LITERAL);
+ const char *result = token.v.string;
+
+ next_token();
+
+ while(token.type == T_STRING_LITERAL) {
+ result = concat_strings(result, token.v.string);
+ next_token();
+ }
+
+ return result;
+}
+
+static
+void parse_attributes(void)
+{
+ while(1) {
+ 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;
+ }
+ }
+
+attributes_finished:
+ ;
}
typedef enum {
case T_restrict: \
case T_volatile: \
case T_inline: \
- case T___extension__: \
- case T___attribute__:
+ case T___extension__:
#ifdef PROVIDE_COMPLEX
#define COMPLEX_SPECIFIERS \
#endif
#define TYPE_SPECIFIERS \
- case T_TYPENAME: \
case T_void: \
case T_char: \
case T_short: \
case T_struct: \
case T_union: \
case T_enum: \
- case T___quad_t: \
- case T___u_quad_t: \
COMPLEX_SPECIFIERS \
IMAGINARY_SPECIFIERS
+#define DECLARATION_START \
+ STORAGE_CLASSES \
+ TYPE_QUALIFIERS \
+ TYPE_SPECIFIERS
+
+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 result;
+}
+
static
void parse_declaration_specifiers(declaration_specifiers_t *specifiers)
{
- type_t *type = NULL;
- unsigned type_qualifiers = 0;
- unsigned type_specifiers = 0;
+ declaration_t *declaration;
+ type_t *type = NULL;
+ unsigned type_qualifiers = 0;
+ unsigned type_specifiers = 0;
+ int newtype = 0;
while(1) {
switch(token.type) {
next_token();
break;
- case T___attribute__:
- fprintf(stderr, "TODO: __attribute__ not handled yet\n");
- next_token();
- break;
-
/* type specifiers */
#define MATCH_SPECIFIER(token, specifier, name) \
case token: \
}
break;
- case T___quad_t:
- next_token();
- if(type_specifiers & SPECIFIER_LONG_LONG ||
- type_specifiers & SPECIFIER_LONG) {
- parse_error("multiple type specifiers given");
- } else {
- type_specifiers |= specifier;
- }
- break;
-
- type_specifiers
-
+ /* TODO: if type != NULL for the following rules issue an error */
case T_struct:
- type = parse_struct_specifier();
+ type = parse_compound_type_specifier(1);
break;
case T_union:
- type = parse_union_specifier();
+ type = parse_compound_type_specifier(0);
break;
case T_enum:
+ type = parse_enum_specifier();
+ break;
+ case T___builtin_va_list:
+ type = create_builtin_type(token.v.symbol);
+ next_token();
+ break;
+
+ case T___attribute__:
/* TODO */
- assert(0);
+ parse_attributes();
break;
- case T_TYPENAME:
- if(type != NULL || type_specifiers != 0) {
+ case T_IDENTIFIER:
+ declaration = token.v.symbol->declaration;
+ if(declaration == NULL ||
+ declaration->storage_class != STORAGE_CLASS_TYPEDEF) {
goto finish_specifiers;
}
- type = token.v.symbol->thing->declaration->type;
+ type = declaration->type;
assert(type != NULL);
next_token();
break;
case SPECIFIER_UNSIGNED | SPECIFIER_LONG | SPECIFIER_INT:
atomic_type = ATOMIC_TYPE_ULONG;
break;
- case SPECIFIER_LONG_LONG:
- case SPECIFIER_SIGNED | SPECIFIER_LONG_LONG:
- case SPECIFIER_LONG_LONG | SPECIFIER_INT:
- case SPECIFIER_SIGNED | SPECIFIER_LONG_LONG | SPECIFIER_INT:
+ case SPECIFIER_LONG | SPECIFIER_LONG_LONG:
+ case SPECIFIER_SIGNED | SPECIFIER_LONG | SPECIFIER_LONG_LONG:
+ case SPECIFIER_LONG | SPECIFIER_LONG_LONG | SPECIFIER_INT:
+ case SPECIFIER_SIGNED | SPECIFIER_LONG | SPECIFIER_LONG_LONG
+ | SPECIFIER_INT:
atomic_type = ATOMIC_TYPE_LONGLONG;
break;
- case SPECIFIER_UNSIGNED | SPECIFIER_LONG_LONG:
- case SPECIFIER_UNSIGNED | SPECIFIER_LONG_LONG | SPECIFIER_INT:
+ case SPECIFIER_UNSIGNED | SPECIFIER_LONG | SPECIFIER_LONG_LONG:
+ case SPECIFIER_UNSIGNED | SPECIFIER_LONG | SPECIFIER_LONG_LONG
+ | SPECIFIER_INT:
atomic_type = ATOMIC_TYPE_ULONGLONG;
break;
case SPECIFIER_FLOAT:
atomic_type_t *atype = allocate_type_zero(sizeof(atype[0]));
atype->type.type = TYPE_ATOMIC;
atype->atype = atomic_type;
+ newtype = 1;
type = (type_t*) atype;
} else {
}
}
- type->qualifiers = type_qualifiers;
+ type->qualifiers = type_qualifiers;
+
+ type_t *result = typehash_insert(type);
+ if(newtype && result != (type_t*) type) {
+ obstack_free(type_obst, type);
+ }
+
+ specifiers->type = result;
+}
+
+static
+unsigned parse_type_qualifiers(void)
+{
+ unsigned type_qualifiers = 0;
+
+ while(1) {
+ switch(token.type) {
+ /* type qualifiers */
+ MATCH_TYPE_QUALIFIER(T_const, TYPE_QUALIFIER_CONST);
+ MATCH_TYPE_QUALIFIER(T_restrict, TYPE_QUALIFIER_RESTRICT);
+ MATCH_TYPE_QUALIFIER(T_volatile, TYPE_QUALIFIER_VOLATILE);
+ MATCH_TYPE_QUALIFIER(T_inline, TYPE_QUALIFIER_INLINE);
+
+ default:
+ return type_qualifiers;
+ }
+ }
+}
+
+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)
+{
+ while(1) {
+ if(token.type != T_IDENTIFIER) {
+ parse_error_expected("problem while parsing parameter identifier "
+ "list", T_IDENTIFIER, 0);
+ return;
+ }
+ next_token();
+ if(token.type != ',')
+ break;
+ next_token();
+ }
+}
+
+static
+declaration_t *parse_parameter(void)
+{
+ declaration_specifiers_t specifiers;
+ memset(&specifiers, 0, sizeof(specifiers));
- type_t *result = typehash_insert(type);
- if(result != (type_t*) type) {
- obstack_free(type_obst, type);
- }
+ parse_declaration_specifiers(&specifiers);
- specifiers->type = result;
+ declaration_t *declaration = allocate_ast_zero(sizeof(declaration[0]));
+ parse_declarator(declaration, specifiers.storage_class,
+ specifiers.type, 1);
- fprintf(stderr, "Specifiers type: ");
- print_type(stderr, result);
- fprintf(stderr, "\n");
+ return declaration;
}
static
-unsigned parse_type_qualifiers()
+declaration_t *parse_parameters(method_type_t *type)
{
- unsigned type_qualifiers = 0;
+ 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) {
+ /* TODO */
+ parse_identifier_list();
+ return NULL;
+ }
+ }
+
+ if(token.type == ')') {
+ type->unspecified_parameters = 1;
+ return NULL;
+ }
+ if(token.type == T_void && look_ahead(1)->type == ')') {
+ next_token();
+ return NULL;
+ }
+
+ declaration_t *declarations = NULL;
+ declaration_t *declaration;
+ declaration_t *last_declaration = NULL;
+ method_parameter_t *parameter;
+ method_parameter_t *last_parameter = NULL;
while(1) {
switch(token.type) {
- /* type qualifiers */
- MATCH_TYPE_QUALIFIER(T_const, TYPE_QUALIFIER_CONST);
- MATCH_TYPE_QUALIFIER(T_restrict, TYPE_QUALIFIER_RESTRICT);
- MATCH_TYPE_QUALIFIER(T_volatile, TYPE_QUALIFIER_VOLATILE);
- MATCH_TYPE_QUALIFIER(T_inline, TYPE_QUALIFIER_INLINE);
+ case T_DOTDOTDOT:
+ next_token();
+ type->variadic = 1;
+ return declarations;
+
+ case T_IDENTIFIER:
+ DECLARATION_START
+ declaration = parse_parameter();
+
+ parameter = allocate_type_zero(sizeof(parameter[0]));
+ parameter->type = declaration->type;
+
+ if(last_parameter != NULL) {
+ last_declaration->next = declaration;
+ last_parameter->next = parameter;
+ } else {
+ type->parameters = parameter;
+ declarations = declaration;
+ }
+ last_parameter = parameter;
+ last_declaration = declaration;
+ break;
default:
- return type_qualifiers;
+ return declarations;
}
+ if(token.type != ',')
+ return declarations;
+ next_token();
}
}
-static
-void parse_declarator(declaration_t *declaration, storage_class_t storage_class,
- type_t *type)
-{
- while(token.type == '*') {
- /* pointer */
- next_token();
+typedef struct declarator_part declarator_part;
+struct declarator_part {
+ parsed_pointer_t *pointers;
+ method_type_t *method_type;
+ declarator_part *inner;
+};
- 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 = parse_type_qualifiers();
+static
+declarator_part *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]));
- type_t *result = typehash_insert((type_t*) pointer_type);
- if(result != (type_t*) pointer_type) {
- obstack_free(type_obst, pointer_type);
- }
+ part->pointers = parse_pointers();
- type = result;
- }
- declaration->storage_class = storage_class;
- declaration->type = type;
+ /* TODO: find out if this is correct */
+ parse_attributes();
switch(token.type) {
- case T_TYPENAME:
case T_IDENTIFIER:
- declaration->symbol = token.v.symbol;
+ if(declaration == NULL) {
+ parse_error("no identifier expected in typename");
+ } else {
+ declaration->symbol = token.v.symbol;
+ declaration->source_position = token.source_position;
+ }
next_token();
break;
case '(':
next_token();
- parse_declarator(declaration, storage_class, type);
- expect_void(')');
+ part->inner = parse_inner_declarator(declaration, may_be_abstract);
+ expect(')');
break;
default:
- parse_error("problem while parsing declarator");
+ if(may_be_abstract)
+ break;
+ parse_error_expected("problem while parsing declarator", T_IDENTIFIER,
+ '(', 0);
}
- if(token.type == '(') {
- next_token();
+ while(1) {
+ switch(token.type) {
+ case '(':
+ next_token();
- /* parse parameter-type-list or identifier-list */
+ method_type_t *method_type
+ = allocate_type_zero(sizeof(method_type[0]));
+ method_type->type.type = TYPE_METHOD;
- expect_void(')');
- } else if(token.type == '[') {
- next_token();
+ declaration_t *parameters = parse_parameters(method_type);
+ if(declaration != NULL) {
+ declaration->context.declarations = parameters;
+ }
+
+ part->method_type = method_type;
+
+ expect(')');
+ 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();
+ }
+ }
- /* multiple type qualifiers, and static */
+ /* TODO */
- /* assignment_expression or '*' or nothing */
+ if(token.type == '*' && look_ahead(1)->type == ']') {
+ next_token();
+ } else if(token.type != ']') {
+ parse_assignment_expression();
+ }
- expect_void(']');
+ expect(']');
+ break;
+ default:
+ goto declarator_finished;
+ }
}
- fprintf(stderr, "Declarator type: ");
- print_type(stderr, type);
- fprintf(stderr, "\n");
+declarator_finished:
+ parse_attributes();
- symbol_t *symbol = declaration->symbol;
+ return part;
+}
+
+static
+type_t *construct_declarator_type(declarator_part *part, type_t *type)
+{
+ do {
+ type = make_pointers(type, part->pointers);
+
+ method_type_t *method_type = part->method_type;
+ if(method_type != NULL) {
+ method_type->result_type = type;
+
+ type_t *result = typehash_insert((type_t*) method_type);
+ if(result != (type_t*) method_type) {
+ obstack_free(type_obst, method_type);
+ }
+ type = result;
+ }
+
+ part = part->inner;
+ } while(part != NULL);
+
+ return type;
+}
- environment_entry_t *entry = environment_push(symbol);
- entry->declaration = declaration;
- entry->old_symbol_ID = symbol->ID;
+static
+void parse_declarator(declaration_t *declaration, storage_class_t storage_class,
+ type_t *type, int may_be_abstract)
+{
+ declarator_part *part
+ = parse_inner_declarator(declaration, may_be_abstract);
+
+ if(part != NULL) {
+ declaration->type = construct_declarator_type(part, type);
+ declaration->storage_class = storage_class;
+ obstack_free(&temp_obst, part);
+ }
+}
+
+static
+type_t *parse_abstract_declarator(type_t *base_type)
+{
+ declarator_part *part = parse_inner_declarator(NULL, 1);
+
+ if(part == NULL)
+ return NULL;
+
+ type_t *result = construct_declarator_type(part, base_type);
+ obstack_free(&temp_obst, part);
+
+ return result;
+}
+
+static declaration_t *record_declaration(declaration_t *declaration)
+{
+ if(context == NULL)
+ return declaration;
+
+ symbol_t *symbol = declaration->symbol;
+ if(symbol != NULL) {
+ declaration_t *alias = environment_push(declaration, context);
+ if(alias != declaration)
+ return alias;
+ }
- if(storage_class == STORAGE_CLASS_TYPEDEF) {
- symbol->ID = T_TYPENAME;
+ if(last_declaration != NULL) {
+ last_declaration->next = declaration;
} else {
- symbol->ID = T_IDENTIFIER;
+ context->declarations = declaration;
}
+ last_declaration = declaration;
+
+ return declaration;
+}
+
+static
+void parser_error_multiple_definition(declaration_t *previous,
+ declaration_t *declaration)
+{
+ 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");
}
static
void parse_init_declarators(const declaration_specifiers_t *specifiers)
{
while(1) {
- declaration_t *declaration = allocate_ast_zero(sizeof(declaration[0]));
+ declaration_t *ndeclaration
+ = allocate_ast_zero(sizeof(ndeclaration[0]));
- parse_declarator(declaration, specifiers->storage_class,
- specifiers->type);
+ parse_declarator(ndeclaration, specifiers->storage_class,
+ specifiers->type, 0);
+ declaration_t *declaration = record_declaration(ndeclaration);
if(token.type == '=') {
next_token();
- // parse_initializer TODO
+
+ /* TODO: check that this is an allowed type (esp. no method type) */
+
+ if(declaration->initializer != NULL) {
+ parser_error_multiple_definition(declaration, ndeclaration);
+ }
+
+ if(token.type == '{') {
+ // TODO
+ expect_void('}');
+ } else {
+ declaration->initializer = parse_assignment_expression();
+ }
} else if(token.type == '{') {
- parse_compound_statement();
+ if(declaration->type->type != TYPE_METHOD) {
+ parser_print_error_prefix();
+ fprintf(stderr, "Declarator ");
+ print_type(declaration->type, declaration->symbol);
+ fprintf(stderr, " is not a method type.\n");
+ }
+
+ if(declaration->initializer != NULL) {
+ parser_error_multiple_definition(declaration, ndeclaration);
+ }
+
+ 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, context);
+ }
+
+ statement_t *statement = parse_compound_statement();
+
+ assert(context == &declaration->context);
+ set_context(last_context);
+ environment_pop_to(top);
+
+ declaration->statement = statement;
return;
}
void parse_struct_declarators(const declaration_specifiers_t *specifiers)
{
while(1) {
- declaration_t declaration;
- compound_entry_t *entry = allocate_ast_zero(sizeof(entry[0]));
-
if(token.type == ':') {
next_token();
parse_constant_expression();
- /* TODO */
+ /* TODO (bitfields) */
} else {
- parse_declarator(&declaration, specifiers->storage_class,
- specifiers->type);
+ declaration_t *declaration
+ = allocate_ast_zero(sizeof(declaration[0]));
+ parse_declarator(declaration, specifiers->storage_class,
+ specifiers->type, 1);
+
+ /* TODO: check for doubled fields */
+ record_declaration(declaration);
if(token.type == ':') {
next_token();
parse_constant_expression();
- /* TODO */
+ /* TODO (bitfields) */
}
}
expect_void(';');
}
-static compound_entry_t *parse_compound_type_entries(void)
+static void parse_compound_type_entries(void)
{
eat('{');
- compound_entry_t *entries = NULL;
-
while(token.type != '}' && token.type != T_EOF) {
declaration_specifiers_t specifiers;
memset(&specifiers, 0, sizeof(specifiers));
parse_struct_declarators(&specifiers);
}
+ if(token.type == T_EOF) {
+ parse_error("unexpected error while parsing struct");
+ }
next_token();
-
- return entries;
}
void parse_declaration(void)
parse_init_declarators(&specifiers);
}
+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);
string_literal_t *cnst = allocate_ast_zero(sizeof(cnst[0]));
cnst->expression.type = EXPR_STRING_LITERAL;
- cnst->value = token.v.string;
-
- next_token();
+ cnst->value = parse_string_literals();
return (expression_t*) cnst;
}
{
reference_expression_t *ref = allocate_ast_zero(sizeof(ref[0]));
- ref->expression.type = EXPR_REFERENCE;
- ref->symbol = token.v.symbol;
+ ref->expression.type = EXPR_REFERENCE;
+ ref->symbol = token.v.symbol;
+
+ if(ref->symbol->declaration == NULL) {
+ parser_print_error_prefix();
+ fprintf(stderr, "unknown symbol '%s' found.\n", ref->symbol->string);
+ }
+ ref->declaration = ref->symbol->declaration;
next_token();
return (expression_t*) ref;
}
+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);
+
+ cast->expression.datatype = type;
+ cast->value = value;
+
+ return (expression_t*) cast;
+}
+
static
expression_t *parse_brace_expression(void)
{
eat('(');
+ declaration_t *declaration;
+ switch(token.type) {
+ TYPE_QUALIFIERS
+ TYPE_SPECIFIERS
+ return parse_cast();
+ case T_IDENTIFIER:
+ declaration = token.v.symbol->declaration;
+ if(declaration != NULL &&
+ (declaration->storage_class & STORAGE_CLASS_TYPEDEF)) {
+ return parse_cast();
+ }
+ }
+
expression_t *result = parse_expression();
expect(')');
return parse_brace_expression();
}
- /* TODO: error message */
+ parser_print_error_prefix();
+ fprintf(stderr, "unexpected token ");
+ print_token(stderr, &token);
+ fprintf(stderr, "\n");
+ eat_statement();
return NULL;
}
return (expression_t*) array_access;
}
+static
+type_t *get_expression_type(const expression_t *expression)
+{
+ (void) expression;
+ /* TODO */
+ return NULL;
+}
+
+static
+int is_type_specifier(const token_t *token)
+{
+ declaration_t *declaration;
+
+ switch(token->type) {
+ TYPE_SPECIFIERS
+ return 1;
+ case T_IDENTIFIER:
+ declaration = token->v.symbol->declaration;
+ if(declaration == NULL)
+ return 0;
+ if(declaration->storage_class != STORAGE_CLASS_TYPEDEF)
+ return 0;
+ return 1;
+ default:
+ return 0;
+ }
+}
+
static
expression_t *parse_sizeof(unsigned precedence)
{
- (void) precedence;
eat(T_sizeof);
- /* TODO... */
- return NULL;
+ sizeof_expression_t *sizeof_expression
+ = allocate_ast_zero(sizeof(sizeof_expression[0]));
+ sizeof_expression->expression.type = EXPR_SIZEOF;
+
+ if(token.type == '(' && is_type_specifier(look_ahead(1))) {
+ next_token();
+ sizeof_expression->type = parse_typename();
+ expect(')');
+ } else {
+ expression_t *expression = parse_sub_expression(precedence);
+ sizeof_expression->type = get_expression_type(expression);
+ }
+
+ return (expression_t*) sizeof_expression;
}
static
return (expression_t*) call;
}
+static
+expression_t *parse_conditional_expression(unsigned precedence,
+ expression_t *expression)
+{
+ eat('?');
+
+ conditional_expression_t *conditional
+ = allocate_ast_zero(sizeof(conditional[0]));
+ conditional->condition = expression;
+
+ conditional->true_expression = parse_expression();
+ expect(':');
+ conditional->false_expression = parse_sub_expression(precedence);
+
+ return (expression_t*) conditional;
+}
+
#define CREATE_UNARY_EXPRESSION_PARSER(token_type, unexpression_type) \
static \
expression_t *parse_##unexpression_type(unsigned precedence) \
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_EXCLAMATIONMARKEQUAL, BINEXPR_NOTEQUAL)
CREATE_BINEXPR_PARSER(T_LESSEQUAL, BINEXPR_LESSEQUAL)
CREATE_BINEXPR_PARSER(T_GREATEREQUAL, BINEXPR_GREATEREQUAL)
CREATE_BINEXPR_PARSER('&', BINEXPR_BITWISE_AND)
CREATE_BINEXPR_PARSER('|', BINEXPR_BITWISE_OR)
CREATE_BINEXPR_PARSER('^', BINEXPR_BITWISE_XOR)
+CREATE_BINEXPR_PARSER(T_ANDAND, BINEXPR_LOGICAL_AND)
+CREATE_BINEXPR_PARSER(T_PIPEPIPE, BINEXPR_LOGICAL_OR)
CREATE_BINEXPR_PARSER(T_LESSLESS, BINEXPR_SHIFTLEFT)
CREATE_BINEXPR_PARSER(T_GREATERGREATER, BINEXPR_SHIFTRIGHT)
expression_parser_function_t *parser
= &expression_parsers[token.type];
- source_position_t source_position = source_position;
+ source_position_t source_position = token.source_position;
expression_t *left;
if(parser->parser != NULL) {
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_array_expression, '[', 30);
statement_t *parse_label_statement(void)
{
eat(T_IDENTIFIER);
- expect(';');
+ expect(':');
parse_statement();
return NULL;
statement_t *parse_if(void)
{
eat(T_if);
+
+ if_statement_t *statement = allocate_ast_zero(sizeof(statement[0]));
+ statement->statement.type = STATEMENT_IF;
+
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
eat(T_continue);
expect(';');
- return NULL;
+ statement_t *statement = allocate_ast_zero(sizeof(statement[0]));
+ statement->source_position = token.source_position;
+ statement->type = STATEMENT_CONTINUE;
+
+ return statement;
}
static
statement_t *parse_return(void)
{
eat(T_return);
- parse_expression();
+
+ return_statement_t *statement = allocate_ast_zero(sizeof(statement[0]));
+ statement->statement.type = STATEMENT_RETURN;
+ if(token.type != ';') {
+ statement->return_value = parse_expression();
+ }
expect(';');
- return NULL;
+ return (statement_t*) statement;
}
static
return NULL;
}
+static
+statement_t *parse_expression_statement(void)
+{
+ parse_expression();
+ return NULL;
+}
+
static
statement_t *parse_statement(void)
{
- statement_t *statement = NULL;
+ declaration_t *declaration;
+ statement_t *statement = NULL;
/* declaration or statement */
switch(token.type) {
statement = parse_default_statement();
break;
- case T_IDENTIFIER:
- statement = parse_label_statement();
- break;
-
case '{':
statement = parse_compound_statement();
break;
break;
case ';':
+ next_token();
statement = NULL;
break;
- STORAGE_CLASSES
- TYPE_QUALIFIERS
- TYPE_SPECIFIERS
+ case T_IDENTIFIER:
+ if(look_ahead(1)->type == ':') {
+ statement = parse_label_statement();
+ break;
+ }
+
+ declaration = token.v.symbol->declaration;
+ if(declaration != NULL &&
+ declaration->storage_class == STORAGE_CLASS_TYPEDEF) {
+ statement = parse_declaration_statement();
+ break;
+ }
+
+ statement = parse_expression_statement();
+ break;
+
+ DECLARATION_START
statement = parse_declaration_statement();
break;
+
+ default:
+ statement = parse_expression_statement();
+ break;
}
return statement;
{
eat('{');
- int top = environment_top();
+ compound_statement_t *compound_statement
+ = allocate_ast_zero(sizeof(compound_statement[0]));
+ compound_statement->statement.type = STATEMENT_COMPOUND;
- while(token.type != '}') {
- parse_statement();
+ int top = environment_top();
+ context_t *last_context = context;
+ set_context(&compound_statement->context);
+
+ statement_t *last_statement = NULL;
+
+ while(token.type != '}' && token.type != T_EOF) {
+ statement_t *statement = parse_statement();
+
+ if(last_statement != NULL) {
+ last_statement->next = statement;
+ } else {
+ compound_statement->statements = statement;
+ }
+ last_statement = statement;
}
+ assert(context == &compound_statement->context);
+ set_context(last_context);
environment_pop_to(top);
next_token();
- return NULL;
+ return (statement_t*) compound_statement;
}
static
{
translation_unit_t *unit = allocate_ast_zero(sizeof(unit[0]));
- assert(translation_unit == NULL);
assert(context == NULL);
- translation_unit = unit;
+ set_context(&unit->context);
while(token.type != T_EOF) {
parse_declaration();
}
- translation_unit = NULL;
+ assert(context == &unit->context);
+ context = NULL;
+ last_declaration = NULL;
+
return unit;
}
obstack_init(&environment_obstack);
environment_stack = NEW_ARR_F(environment_entry_t*, 0);
- next_token();
+ type_set_output(stderr);
+
+ lookahead_bufpos = 0;
+ for(int i = 0; i < MAX_LOOKAHEAD + 2; ++i) {
+ next_token();
+ }
translation_unit_t *unit = parse_translation_unit();
DEL_ARR_F(environment_stack);
void init_parser(void)
{
init_expression_parsers();
+ obstack_init(&temp_obst);
}
void exit_parser(void)
{
+ obstack_free(&temp_obst, NULL);
}