#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 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
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
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 *la(int num)
+{
+ assert(num > 0 && num <= MAX_LOOKAHEAD);
+ int pos = (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);
fputs(": error: ", stderr);
}
+void parser_print_error_prefix(void)
+{
+ parser_print_error_prefix_pos(token.source_position);
+}
+
static
void parse_error(const char *message)
{
} \
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;
+}
+
+/**
+ * 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)
+{
+ 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) {
+ if(symbol->declaration != NULL) {
+ assert(symbol->declaration != NULL);
+ 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");
+ }
+ }
+
+ entry->old_declaration = symbol->declaration;
+ entry->old_context = symbol->context;
+ entry->symbol = symbol;
+ symbol->declaration = declaration;
+ symbol->context = context;
+}
+
+/**
+ * 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 void parse_compound_type_entries(void);
static void parse_declarator(declaration_t *declaration,
storage_class_t storage_class, type_t *type);
+static void maybe_push_declaration(declaration_t *declaration);
+static void record_declaration(declaration_t *declaration);
typedef struct declaration_specifiers_t declaration_specifiers_t;
struct declaration_specifiers_t {
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;
+ struct_type->source_position = token.source_position;
+
+ int top = environment_top();
+ context_t *last_context = context;
+ set_context(&struct_type->context);
- if(token.type == T_IDENTIFIER || token.type == T_TYPENAME) {
- /* TODO */
+ if(token.type == T_IDENTIFIER) {
next_token();
if(token.type == '{') {
parse_compound_type_entries();
} else {
parse_error_expected("problem while parsing struct type specifiers",
T_IDENTIFIER, '{', 0);
- return NULL;
+ struct_type = NULL;
}
+ assert(context == &struct_type->context);
+ set_context(last_context);
+ environment_pop_to(top);
+
return (type_t*) struct_type;
}
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;
+ union_type->source_position = token.source_position;
+
+ int top = environment_top();
+ context_t *last_context = context;
+ set_context(&union_type->context);
if(token.type == T_IDENTIFIER) {
union_type->symbol = token.v.symbol;
} else {
parse_error_expected("problem while parsing union type specifiers",
T_IDENTIFIER, '{');
+ union_type = NULL;
}
+ assert(context == &union_type->context);
+ set_context(last_context);
+ environment_pop_to(top);
+
return (type_t*) union_type;
}
enum_type_t *enum_type = allocate_type_zero(sizeof(enum_type[0]));
enum_type->type.type = TYPE_ENUM;
- enum_type->source_position = source_position;
+ enum_type->source_position = token.source_position;
if(token.type == T_IDENTIFIER) {
enum_type->symbol = token.v.symbol;
#endif
#define TYPE_SPECIFIERS \
- case T_TYPENAME: \
case T_void: \
case T_char: \
case T_short: \
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) {
}
break;
+ /* TODO: if type != NULL for the following rules issue an error */
case T_struct:
type = parse_struct_specifier();
break;
next_token();
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;
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_t *result = typehash_insert(type);
- if(result != (type_t*) type) {
+ if(newtype && result != (type_t*) type) {
obstack_free(type_obst, type);
}
specifiers->type = result;
-
- fprintf(stderr, "Specifiers type: ");
- print_type(stderr, result);
- fprintf(stderr, "\n");
}
static
parse_declaration_specifiers(&specifiers);
specifiers.type = parse_pointer(specifiers.type);
- if(token.type == '(' || token.type == T_IDENTIFIER
- || token.type == T_TYPENAME) {
- declaration_t declaration;
- memset(&declaration, 0, sizeof(declaration));
- parse_declarator(&declaration, specifiers.storage_class,
+ if(token.type == '(' || token.type == T_IDENTIFIER) {
+ declaration_t *declaration = allocate_ast_zero(sizeof(declaration[0]));
+ parse_declarator(declaration, specifiers.storage_class,
specifiers.type);
+ maybe_push_declaration(declaration);
+ record_declaration(declaration);
}
}
void parse_parameters()
{
if(token.type == T_IDENTIFIER) {
- parse_identifier_list();
- return;
+ symbol_t *symbol = token.v.symbol;
+ declaration_t *declaration = symbol->declaration;
+ if(declaration == NULL
+ || declaration->storage_class != STORAGE_CLASS_TYPEDEF) {
+ parse_identifier_list();
+ return;
+ }
}
while(1) {
switch(token.type) {
case T_DOTDOTDOT:
+ case T_IDENTIFIER:
DECLARATION_START
parse_parameter();
break;
{
while(token.type == T___attribute__) {
next_token();
- fprintf(stderr, "TODO: __attribute__ not handled yet\n");
expect_void('(');
int depth = 1;
declaration->type = type;
switch(token.type) {
- case T_TYPENAME:
case T_IDENTIFIER:
- declaration->symbol = token.v.symbol;
+ declaration->symbol = token.v.symbol;
+ declaration->source_position = token.source_position;
next_token();
break;
case '(':
expect_void(')');
break;
default:
- parse_error_expected("problem while parsing declarator", T_TYPENAME,
- T_IDENTIFIER, '(', 0);
+ parse_error_expected("problem while parsing declarator", T_IDENTIFIER,
+ '(', 0);
}
while(1) {
case '(':
next_token();
+ int top = environment_top();
+ context_t *last_context = context;
+ set_context(&declaration->context);
+
parse_parameters();
+ assert(context == &declaration->context);
+ set_context(last_context);
+ environment_pop_to(top);
+
expect_void(')');
break;
case '[':
}
}
- if(token.type == '*' /* TODO: && lookahead == ']' */) {
+ if(token.type == '*' && la(1)->type == ']') {
next_token();
} else if(token.type != ']') {
parse_assignment_expression();
declarator_finished:
parse_attributes();
+}
- fprintf(stderr, "Declarator type: ");
- print_type(stderr, type);
- fprintf(stderr, "\n");
+static void record_declaration(declaration_t *declaration)
+{
+ if(last_declaration != NULL) {
+ last_declaration->next = declaration;
+ } else {
+ if(context != NULL)
+ context->declarations = declaration;
+ }
+ last_declaration = declaration;
+}
+static
+void maybe_push_declaration(declaration_t *declaration)
+{
symbol_t *symbol = declaration->symbol;
if(symbol != NULL) {
- environment_entry_t *entry = environment_push(symbol);
- entry->declaration = declaration;
- entry->old_symbol_ID = symbol->ID;
-
- if(storage_class == STORAGE_CLASS_TYPEDEF) {
- symbol->ID = T_TYPENAME;
- fprintf(stderr, "typedef '%s'\n", symbol->string);
- } else {
- symbol->ID = T_IDENTIFIER;
- }
+ environment_push(declaration, context);
}
}
parse_declarator(declaration, specifiers->storage_class,
specifiers->type);
+ maybe_push_declaration(declaration);
+ record_declaration(declaration);
if(token.type == '=') {
next_token();
if(token.type == '{') {
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,
+ declaration_t *declaration
+ = allocate_ast_zero(sizeof(declaration[0]));
+ parse_declarator(declaration, specifiers->storage_class,
specifiers->type);
+ maybe_push_declaration(declaration);
+ 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)
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) {
statement_t *parse_label_statement(void)
{
eat(T_IDENTIFIER);
- expect(';');
+ expect(':');
parse_statement();
return NULL;
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;
statement = NULL;
break;
+ case T_IDENTIFIER:
+ if(la(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;
{
eat('{');
- int top = environment_top();
+ compound_statement_t *compound_statement
+ = allocate_ast_zero(sizeof(compound_statement[0]));
+ compound_statement->statement.type = STATEMENT_COMPOUND;
+
+ int top = environment_top();
+ context_t *last_context = context;
+ set_context(&compound_statement->context);
while(token.type != '}') {
parse_statement();
}
+ assert(context == &compound_statement->context);
+ set_context(last_context);
environment_pop_to(top);
next_token();
{
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();
+ 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);