glcpp-parse.y

%{
/*
 * Copyright © 2010 Intel Corporation
 *
 * Permission is hereby granted, free of charge, to any person obtaining a
 * copy of this software and associated documentation files (the "Software"),
 * to deal in the Software without restriction, including without limitation
 * the rights to use, copy, modify, merge, publish, distribute, sublicense,
 * and/or sell copies of the Software, and to permit persons to whom the
 * Software is furnished to do so, subject to the following conditions:
 *
 * The above copyright notice and this permission notice (including the next
 * paragraph) shall be included in all copies or substantial portions of the
 * Software.
 *
 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
 * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
 * DEALINGS IN THE SOFTWARE.
 */

#include <stdio.h>
#include <stdlib.h>
#include <assert.h>
#include <inttypes.h>

#include "glcpp.h"

static void
yyerror (void *scanner, const char *error);

static void
_define_object_macro (glcpp_parser_t *parser,
		      const char *macro,
		      token_list_t *replacements);

static void
_define_function_macro (glcpp_parser_t *parser,
			const char *macro,
			string_list_t *parameters,
			token_list_t *replacements);

static string_list_t *
_string_list_create (void *ctx);

static void
_string_list_append_item (string_list_t *list, const char *str);

static void
_string_list_append_list (string_list_t *list, string_list_t *tail);

static int
_string_list_contains (string_list_t *list, const char *member, int *index);

static int
_string_list_length (string_list_t *list);

static argument_list_t *
_argument_list_create (void *ctx);

static void
_argument_list_append (argument_list_t *list, token_list_t *argument);

static int
_argument_list_length (argument_list_t *list);

static token_list_t *
_argument_list_member_at (argument_list_t *list, int index);

/* Note: This function talloc_steal()s the str pointer. */
static token_t *
_token_create_str (void *ctx, int type, char *str);

static token_t *
_token_create_ival (void *ctx, int type, int ival);

static token_list_t *
_token_list_create (void *ctx);

/* Note: This function adds a talloc_reference() to token.
 *
 * You may want to talloc_unlink any current reference if you no
 * longer need it. */
static void
_token_list_append (token_list_t *list, token_t *token);

static void
_token_list_append_list (token_list_t *list, token_list_t *tail);

static int
_token_list_length (token_list_t *list);

static active_list_t *
_active_list_push (active_list_t *list,
		   const char *identifier,
		   token_node_t *marker);

static active_list_t *
_active_list_pop (active_list_t *list);

int
_active_list_contains (active_list_t *list, const char *identifier);

static void
_glcpp_parser_evaluate_defined (glcpp_parser_t *parser,
				token_list_t *list);

static void
_glcpp_parser_expand_token_list (glcpp_parser_t *parser,
				 token_list_t *list);

static void
_glcpp_parser_print_expanded_token_list (glcpp_parser_t *parser,
					 token_list_t *list);

static void
_glcpp_parser_skip_stack_push_if (glcpp_parser_t *parser, int condition);

static void
_glcpp_parser_skip_stack_change_if (glcpp_parser_t *parser, const char *type,
				    int condition);

static void
_glcpp_parser_skip_stack_pop (glcpp_parser_t *parser);

#define yylex glcpp_parser_lex

static int
glcpp_parser_lex (glcpp_parser_t *parser);

static void
glcpp_parser_lex_from (glcpp_parser_t *parser, token_list_t *list);

%}

%parse-param {glcpp_parser_t *parser}
%lex-param {glcpp_parser_t *parser}

%token COMMA_FINAL DEFINED ELIF_EXPANDED HASH HASH_DEFINE_FUNC HASH_DEFINE_OBJ HASH_ELIF HASH_ELSE HASH_ENDIF HASH_IF HASH_IFDEF HASH_IFNDEF HASH_UNDEF IDENTIFIER IF_EXPANDED INTEGER INTEGER_STRING NEWLINE OTHER PLACEHOLDER SPACE
%token PASTE
%type <ival> expression INTEGER operator SPACE
%type <str> IDENTIFIER INTEGER_STRING OTHER
%type <string_list> identifier_list
%type <token> preprocessing_token
%type <token_list> pp_tokens replacement_list text_line
%left OR
%left AND
%left '|'
%left '^'
%left '&'
%left EQUAL NOT_EQUAL
%left '<' '>' LESS_OR_EQUAL GREATER_OR_EQUAL
%left LEFT_SHIFT RIGHT_SHIFT
%left '+' '-'
%left '*' '/' '%'
%right UNARY

%%

input:
	/* empty */
|	input line
;

line:
	control_line {
		printf ("\n");
	}
|	text_line {
		_glcpp_parser_print_expanded_token_list (parser, $1);
		printf ("\n");
		talloc_free ($1);
	}
|	expanded_line
|	HASH non_directive
;

expanded_line:
	IF_EXPANDED expression NEWLINE {
		_glcpp_parser_skip_stack_push_if (parser, $2);
	}
|	ELIF_EXPANDED expression NEWLINE {
		_glcpp_parser_skip_stack_change_if (parser, "elif", $2);
	}
;

control_line:
	HASH_DEFINE_OBJ	IDENTIFIER replacement_list NEWLINE {
		_define_object_macro (parser, $2, $3);
	}
|	HASH_DEFINE_FUNC IDENTIFIER '(' ')' replacement_list NEWLINE {
		_define_function_macro (parser, $2, NULL, $5);
	}
|	HASH_DEFINE_FUNC IDENTIFIER '(' identifier_list ')' replacement_list NEWLINE {
		_define_function_macro (parser, $2, $4, $6);
	}
|	HASH_UNDEF IDENTIFIER NEWLINE {
		macro_t *macro = hash_table_find (parser->defines, $2);
		if (macro) {
			/* XXX: Need hash table to support a real way
			 * to remove an element rather than prefixing
			 * a new node with data of NULL like this. */
			hash_table_insert (parser->defines, NULL, $2);
			talloc_free (macro);
		}
		talloc_free ($2);
	}
|	HASH_IF pp_tokens NEWLINE {
		token_list_t *expanded;
		token_t *token;

		expanded = _token_list_create (parser);
		token = _token_create_ival (parser, IF_EXPANDED, IF_EXPANDED);
		_token_list_append (expanded, token);
		talloc_unlink (parser, token);
		_glcpp_parser_evaluate_defined (parser, $2);
		_glcpp_parser_expand_token_list (parser, $2);
		_token_list_append_list (expanded, $2);
		glcpp_parser_lex_from (parser, expanded);
	}
|	HASH_IFDEF IDENTIFIER NEWLINE {
		macro_t *macro = hash_table_find (parser->defines, $2);
		talloc_free ($2);
		_glcpp_parser_skip_stack_push_if (parser, macro != NULL);
	}
|	HASH_IFNDEF IDENTIFIER NEWLINE {
		macro_t *macro = hash_table_find (parser->defines, $2);
		talloc_free ($2);
		_glcpp_parser_skip_stack_push_if (parser, macro == NULL);
	}
|	HASH_ELIF pp_tokens NEWLINE {
		token_list_t *expanded;
		token_t *token;

		expanded = _token_list_create (parser);
		token = _token_create_ival (parser, ELIF_EXPANDED, ELIF_EXPANDED);
		_token_list_append (expanded, token);
		talloc_unlink (parser, token);
		_glcpp_parser_evaluate_defined (parser, $2);
		_glcpp_parser_expand_token_list (parser, $2);
		_token_list_append_list (expanded, $2);
		glcpp_parser_lex_from (parser, expanded);
	}
|	HASH_ELSE NEWLINE {
		_glcpp_parser_skip_stack_change_if (parser, "else", 1);
	}
|	HASH_ENDIF NEWLINE {
		_glcpp_parser_skip_stack_pop (parser);
	}
|	HASH NEWLINE
;

expression:
	INTEGER_STRING {
		if (strlen ($1) >= 3 && strncmp ($1, "0x", 2) == 0) {
			$$ = strtoll ($1 + 2, NULL, 16);
		} else if ($1[0] == '0') {
			$$ = strtoll ($1, NULL, 8);
		} else {
			$$ = strtoll ($1, NULL, 10);
		}
	}
|	INTEGER {
		$$ = $1;
	}
|	expression OR expression {
		$$ = $1 || $3;
	}
|	expression AND expression {
		$$ = $1 && $3;
	}
|	expression '|' expression {
		$$ = $1 | $3;
	}
|	expression '^' expression {
		$$ = $1 ^ $3;
	}
|	expression '&' expression {
		$$ = $1 & $3;
	}
|	expression NOT_EQUAL expression {
		$$ = $1 != $3;
	}
|	expression EQUAL expression {
		$$ = $1 == $3;
	}
|	expression GREATER_OR_EQUAL expression {
		$$ = $1 >= $3;
	}
|	expression LESS_OR_EQUAL expression {
		$$ = $1 <= $3;
	}
|	expression '>' expression {
		$$ = $1 > $3;
	}
|	expression '<' expression {
		$$ = $1 < $3;
	}
|	expression RIGHT_SHIFT expression {
		$$ = $1 >> $3;
	}
|	expression LEFT_SHIFT expression {
		$$ = $1 << $3;
	}
|	expression '-' expression {
		$$ = $1 - $3;
	}
|	expression '+' expression {
		$$ = $1 + $3;
	}
|	expression '%' expression {
		$$ = $1 % $3;
	}
|	expression '/' expression {
		$$ = $1 / $3;
	}
|	expression '*' expression {
		$$ = $1 * $3;
	}
|	'!' expression %prec UNARY {
		$$ = ! $2;
	}
|	'~' expression %prec UNARY {
		$$ = ~ $2;
	}
|	'-' expression %prec UNARY {
		$$ = - $2;
	}
|	'+' expression %prec UNARY {
		$$ = + $2;
	}
|	'(' expression ')' {
		$$ = $2;
	}
;

identifier_list:
	IDENTIFIER {
		$$ = _string_list_create (parser);
		_string_list_append_item ($$, $1);
		talloc_steal ($$, $1);
	}
|	identifier_list ',' IDENTIFIER {
		$$ = $1;	
		_string_list_append_item ($$, $3);
		talloc_steal ($$, $3);
	}
;

text_line:
	NEWLINE { $$ = NULL; }
|	pp_tokens NEWLINE
;

non_directive:
	pp_tokens NEWLINE
;

replacement_list:
	/* empty */ { $$ = NULL; }
|	pp_tokens
;

pp_tokens:
	preprocessing_token {
		parser->space_tokens = 1;
		$$ = _token_list_create (parser);
		_token_list_append ($$, $1);
		talloc_unlink (parser, $1);
	}
|	pp_tokens preprocessing_token {
		$$ = $1;
		_token_list_append ($$, $2);
		talloc_unlink (parser, $2);
	}
;

preprocessing_token:
	IDENTIFIER {
		$$ = _token_create_str (parser, IDENTIFIER, $1);
	}
|	INTEGER_STRING {
		$$ = _token_create_str (parser, INTEGER_STRING, $1);
	}
|	operator {
		$$ = _token_create_ival (parser, $1, $1);
	}
|	OTHER {
		$$ = _token_create_str (parser, OTHER, $1);
	}
|	SPACE {
		$$ = _token_create_ival (parser, SPACE, SPACE);
	}
;

operator:
	'['			{ $$ = '['; }
|	']'			{ $$ = ']'; }
|	'('			{ $$ = '('; }
|	')'			{ $$ = ')'; }
|	'{'			{ $$ = '{'; }
|	'}'			{ $$ = '}'; }
|	'.'			{ $$ = '.'; }
|	'&'			{ $$ = '&'; }
|	'*'			{ $$ = '*'; }
|	'+'			{ $$ = '+'; }
|	'-'			{ $$ = '-'; }
|	'~'			{ $$ = '~'; }
|	'!'			{ $$ = '!'; }
|	'/'			{ $$ = '/'; }
|	'%'			{ $$ = '%'; }
|	LEFT_SHIFT		{ $$ = LEFT_SHIFT; }
|	RIGHT_SHIFT		{ $$ = RIGHT_SHIFT; }
|	'<'			{ $$ = '<'; }
|	'>'			{ $$ = '>'; }
|	LESS_OR_EQUAL		{ $$ = LESS_OR_EQUAL; }
|	GREATER_OR_EQUAL	{ $$ = GREATER_OR_EQUAL; }
|	EQUAL			{ $$ = EQUAL; }
|	NOT_EQUAL		{ $$ = NOT_EQUAL; }
|	'^'			{ $$ = '^'; }
|	'|'			{ $$ = '|'; }
|	AND			{ $$ = AND; }
|	OR			{ $$ = OR; }
|	';'			{ $$ = ';'; }
|	','			{ $$ = ','; }
|	'='			{ $$ = '='; }
|	PASTE			{ $$ = PASTE; }
|	DEFINED			{ $$ = DEFINED; }
;

%%

string_list_t *
_string_list_create (void *ctx)
{
	string_list_t *list;

	list = xtalloc (ctx, string_list_t);
	list->head = NULL;
	list->tail = NULL;

	return list;
}

void
_string_list_append_list (string_list_t *list, string_list_t *tail)
{
	if (list->head == NULL) {
		list->head = tail->head;
	} else {
		list->tail->next = tail->head;
	}

	list->tail = tail->tail;
}

void
_string_list_append_item (string_list_t *list, const char *str)
{
	string_node_t *node;

	node = xtalloc (list, string_node_t);
	node->str = xtalloc_strdup (node, str);

	node->next = NULL;

	if (list->head == NULL) {
		list->head = node;
	} else {
		list->tail->next = node;
	}

	list->tail = node;
}

int
_string_list_contains (string_list_t *list, const char *member, int *index)
{
	string_node_t *node;
	int i;

	if (list == NULL)
		return 0;

	for (i = 0, node = list->head; node; i++, node = node->next) {
		if (strcmp (node->str, member) == 0) {
			if (index)
				*index = i;
			return 1;
		}
	}

	return 0;
}

int
_string_list_length (string_list_t *list)
{
	int length = 0;
	string_node_t *node;

	if (list == NULL)
		return 0;

	for (node = list->head; node; node = node->next)
		length++;

	return length;
}

argument_list_t *
_argument_list_create (void *ctx)
{
	argument_list_t *list;

	list = xtalloc (ctx, argument_list_t);
	list->head = NULL;
	list->tail = NULL;

	return list;
}

void
_argument_list_append (argument_list_t *list, token_list_t *argument)
{
	argument_node_t *node;

	node = xtalloc (list, argument_node_t);
	node->argument = argument;

	node->next = NULL;

	if (list->head == NULL) {
		list->head = node;
	} else {
		list->tail->next = node;
	}

	list->tail = node;
}

int
_argument_list_length (argument_list_t *list)
{
	int length = 0;
	argument_node_t *node;

	if (list == NULL)
		return 0;

	for (node = list->head; node; node = node->next)
		length++;

	return length;
}

token_list_t *
_argument_list_member_at (argument_list_t *list, int index)
{
	argument_node_t *node;
	int i;

	if (list == NULL)
		return NULL;

	node = list->head;
	for (i = 0; i < index; i++) {
		node = node->next;
		if (node == NULL)
			break;
	}

	if (node)
		return node->argument;

	return NULL;
}

/* Note: This function talloc_steal()s the str pointer. */
token_t *
_token_create_str (void *ctx, int type, char *str)
{
	token_t *token;

	token = xtalloc (ctx, token_t);
	token->type = type;
	token->value.str = talloc_steal (token, str);

	return token;
}

token_t *
_token_create_ival (void *ctx, int type, int ival)
{
	token_t *token;

	token = xtalloc (ctx, token_t);
	token->type = type;
	token->value.ival = ival;

	return token;
}

token_list_t *
_token_list_create (void *ctx)
{
	token_list_t *list;

	list = xtalloc (ctx, token_list_t);
	list->head = NULL;
	list->tail = NULL;
	list->non_space_tail = NULL;

	return list;
}

void
_token_list_append (token_list_t *list, token_t *token)
{
	token_node_t *node;

	node = xtalloc (list, token_node_t);
	node->token = xtalloc_reference (list, token);

	node->next = NULL;

	if (list->head == NULL) {
		list->head = node;
	} else {
		list->tail->next = node;
	}

	list->tail = node;
	if (token->type != SPACE)
		list->non_space_tail = node;
}

void
_token_list_append_list (token_list_t *list, token_list_t *tail)
{
	if (tail == NULL || tail->head == NULL)
		return;

	if (list->head == NULL) {
		list->head = tail->head;
	} else {
		list->tail->next = tail->head;
	}

	list->tail = tail->tail;
	list->non_space_tail = tail->non_space_tail;
}

token_list_t *
_token_list_copy (void *ctx, token_list_t *other)
{
	token_list_t *copy;
	token_node_t *node;

	if (other == NULL)
		return NULL;

	copy = _token_list_create (ctx);
	for (node = other->head; node; node = node->next)
		_token_list_append (copy, node->token);

	return copy;
}

void
_token_list_trim_trailing_space (token_list_t *list)
{
	token_node_t *tail, *next;

	if (list->non_space_tail) {
		tail = list->non_space_tail->next;
		list->non_space_tail->next = NULL;
		list->tail = list->non_space_tail;

		while (tail) {
			next = tail->next;
			talloc_free (tail);
			tail = next;
		}
	}
}

static int
_token_list_length (token_list_t *list)
{
	int length = 0;
	token_node_t *node;

	if (list == NULL)
		return 0;

	for (node = list->head; node; node = node->next)
		length++;

	return length;
}

static void
_token_print (token_t *token)
{
	if (token->type < 256) {
		printf ("%c", token->type);
		return;
	}

	switch (token->type) {
	case INTEGER:
		printf ("%" PRIxMAX, token->value.ival);
		break;
	case IDENTIFIER:
	case INTEGER_STRING:
	case OTHER:
		printf ("%s", token->value.str);
		break;
	case SPACE:
		printf (" ");
		break;
	case LEFT_SHIFT:
		printf ("<<");
		break;
	case RIGHT_SHIFT:
		printf (">>");
		break;
	case LESS_OR_EQUAL:
		printf ("<=");
		break;
	case GREATER_OR_EQUAL:
		printf (">=");
		break;
	case EQUAL:
		printf ("==");
		break;
	case NOT_EQUAL:
		printf ("!=");
		break;
	case AND:
		printf ("&&");
		break;
	case OR:
		printf ("||");
		break;
	case PASTE:
		printf ("##");
		break;
	case COMMA_FINAL:
		printf (",");
		break;
	case PLACEHOLDER:
		/* Nothing to print. */
		break;
	default:
		fprintf (stderr, "Error: Don't know how to print token type %d\n", token->type);
		break;
	}
}

/* Return a new token (talloc()ed off of 'token') formed by pasting
 * 'token' and 'other'. Note that this function may return 'token' or
 * 'other' directly rather than allocating anything new.
 *
 * Caution: Only very cursory error-checking is performed to see if
 * the final result is a valid single token. */
static token_t *
_token_paste (token_t *token, token_t *other)
{
	/* Pasting a placeholder onto anything makes no change. */
	if (other->type == PLACEHOLDER)
		return token;

	/* When 'token' is a placeholder, just return 'other'. */
	if (token->type == PLACEHOLDER)
		return other;

	/* A very few single-character punctuators can be combined
	 * with another to form a multi-character punctuator. */
	switch (token->type) {
	case '<':
		if (other->type == '<')
			return _token_create_ival (token, LEFT_SHIFT, LEFT_SHIFT);
		else if (other->type == '=')
			return _token_create_ival (token, LESS_OR_EQUAL, LESS_OR_EQUAL);
		break;
	case '>':
		if (other->type == '>')
			return _token_create_ival (token, RIGHT_SHIFT, RIGHT_SHIFT);
		else if (other->type == '=')
			return _token_create_ival (token, GREATER_OR_EQUAL, GREATER_OR_EQUAL);
		break;
	case '=':
		if (other->type == '=')
			return _token_create_ival (token, EQUAL, EQUAL);
		break;
	case '!':
		if (other->type == '=')
			return _token_create_ival (token, NOT_EQUAL, NOT_EQUAL);
		break;
	case '&':
		if (other->type == '&')
			return _token_create_ival (token, AND, AND);
		break;
	case '|':
		if (other->type == '|')
			return _token_create_ival (token, OR, OR);
		break;
	}

	/* Two string-valued tokens can usually just be mashed
	 * together.
	 *
	 * XXX: This isn't actually legitimate. Several things here
	 * should result in a diagnostic since the result cannot be a
	 * valid, single pre-processing token. For example, pasting
	 * "123" and "abc" is not legal, but we don't catch that
	 * here. */
	if ((token->type == IDENTIFIER || token->type == OTHER || token->type == INTEGER_STRING) &&
	    (other->type == IDENTIFIER || other->type == OTHER || other->type == INTEGER_STRING))
	{
		char *str;

		str = xtalloc_asprintf (token, "%s%s",
					token->value.str, other->value.str);
		return _token_create_str (token, token->type, str);
	}

	printf ("Error: Pasting \"");
	_token_print (token);
	printf ("\" and \"");
	_token_print (other);
	printf ("\" does not give a valid preprocessing token.\n");

	return token;
}

static void
_token_list_print (token_list_t *list)
{
	token_node_t *node;

	if (list == NULL)
		return;

	for (node = list->head; node; node = node->next)
		_token_print (node->token);
}

void
yyerror (void *scanner, const char *error)
{
	fprintf (stderr, "Parse error: %s\n", error);
}

glcpp_parser_t *
glcpp_parser_create (void)
{
	glcpp_parser_t *parser;

	parser = xtalloc (NULL, glcpp_parser_t);

	glcpp_lex_init_extra (parser, &parser->scanner);
	parser->defines = hash_table_ctor (32, hash_table_string_hash,
					   hash_table_string_compare);
	parser->active = NULL;
	parser->lexing_if = 0;
	parser->space_tokens = 1;
	parser->newline_as_space = 0;
	parser->in_control_line = 0;
	parser->paren_count = 0;

	parser->skip_stack = NULL;

	parser->lex_from_list = NULL;
	parser->lex_from_node = NULL;

	return parser;
}

int
glcpp_parser_parse (glcpp_parser_t *parser)
{
	return yyparse (parser);
}

void
glcpp_parser_destroy (glcpp_parser_t *parser)
{
	if (parser->skip_stack)
		fprintf (stderr, "Error: Unterminated #if\n");
	glcpp_lex_destroy (parser->scanner);
	hash_table_dtor (parser->defines);
	talloc_free (parser);
}

/* Replace any occurences of DEFINED tokens in 'list' with either a
 * '0' or '1' INTEGER token depending on whether the next token in the
 * list is defined or not. */
static void
_glcpp_parser_evaluate_defined (glcpp_parser_t *parser,
				token_list_t *list)
{
	token_node_t *node, *next;
	macro_t *macro;

	if (list == NULL)
		return;

	for (node = list->head; node; node = node->next) {
		if (node->token->type != DEFINED)
			continue;
		next = node->next;
		while (next && next->token->type == SPACE)
			next = next->next;
		if (next == NULL || next->token->type != IDENTIFIER) {
			fprintf (stderr, "Error: operator \"defined\" requires an identifier\n");
			exit (1);
		}
		macro = hash_table_find (parser->defines,
					 next->token->value.str);

		node->token->type = INTEGER;
		node->token->value.ival = (macro != NULL);
		node->next = next->next;
	}
}
	
typedef enum function_status
{
	FUNCTION_STATUS_SUCCESS,
	FUNCTION_NOT_A_FUNCTION,
	FUNCTION_UNBALANCED_PARENTHESES
} function_status_t;

/* Find a set of function-like macro arguments by looking for a
 * balanced set of parentheses.
 *
 * When called, 'node' should be the opening-parenthesis token, (or
 * perhaps preceeding SPACE tokens). Upon successful return *last will
 * be the last consumed node, (corresponding to the closing right
 * parenthesis).
 *
 * Return values:
 *
 *   FUNCTION_STATUS_SUCCESS:
 *
 *	Successfully parsed a set of function arguments.	
 *
 *   FUNCTION_NOT_A_FUNCTION:
 *
 *	Macro name not followed by a '('. This is not an error, but
 *	simply that the macro name should be treated as a non-macro.
 *
 *   FUNCTION_UNBLANCED_PARENTHESES
 *
 *	Macro name is not followed by a balanced set of parentheses.
 */
static function_status_t
_arguments_parse (argument_list_t *arguments,
		  token_node_t *node,
		  token_node_t **last)
{
	token_list_t *argument;
	int paren_count;

	node = node->next;

	/* Ignore whitespace before first parenthesis. */
	while (node && node->token->type == SPACE)
		node = node->next;

	if (node == NULL || node->token->type != '(')
		return FUNCTION_NOT_A_FUNCTION;

	node = node->next;

	argument = _token_list_create (arguments);
	_argument_list_append (arguments, argument);

	for (paren_count = 1; node; node = node->next) {
		if (node->token->type == '(')
		{
			paren_count++;
		}
		else if (node->token->type == ')')
		{
			paren_count--;
			if (paren_count == 0)
				break;
		}

		if (node->token->type == ',' &&
			 paren_count == 1)
		{
			_token_list_trim_trailing_space (argument);
			argument = _token_list_create (arguments);
			_argument_list_append (arguments, argument);
		}
		else {
			if (argument->head == NULL) {
				/* Don't treat initial whitespace as
				 * part of the arguement. */
				if (node->token->type == SPACE)
					continue;
			}
			_token_list_append (argument, node->token);
		}
	}

	if (paren_count)
		return FUNCTION_UNBALANCED_PARENTHESES;

	*last = node;

	return FUNCTION_STATUS_SUCCESS;
}

/* This is a helper function that's essentially part of the
 * implementation of _glcpp_parser_expand_node. It shouldn't be called
 * except for by that function.
 *
 * Returns NULL if node is a simple token with no expansion, (that is,
 * although 'node' corresponds to an identifier defined as a
 * function-like macro, it is not followed with a parenthesized
 * argument list).
 *
 * Compute the complete expansion of node (which is a function-like
 * macro) and subsequent nodes which are arguments.
 *
 * Returns the token list that results from the expansion and sets
 * *last to the last node in the list that was consumed by the
 * expansion. Specificallty, *last will be set as follows: as the
 * token of the closing right parenthesis.
 */
static token_list_t *
_glcpp_parser_expand_function (glcpp_parser_t *parser,
			       token_node_t *node,
			       token_node_t **last)
			       
{
	macro_t *macro;
	const char *identifier;
	argument_list_t *arguments;
	function_status_t status;
	token_list_t *substituted;
	int parameter_index;

	identifier = node->token->value.str;

	macro = hash_table_find (parser->defines, identifier);

	assert (macro->is_function);

	arguments = _argument_list_create (parser);
	status = _arguments_parse (arguments, node, last);

	switch (status) {
	case FUNCTION_STATUS_SUCCESS:
		break;
	case FUNCTION_NOT_A_FUNCTION:
		return NULL;
	case FUNCTION_UNBALANCED_PARENTHESES:
		return NULL;
	}

	if (macro->replacements == NULL) {
		talloc_free (arguments);
		return _token_list_create (parser);
	}

	if (! ((_argument_list_length (arguments) == 
		_string_list_length (macro->parameters)) ||
	       (_string_list_length (macro->parameters) == 0 &&
		_argument_list_length (arguments) == 1 &&
		arguments->head->argument->head == NULL)))
	{
		fprintf (stderr,
			 "Error: macro %s invoked with %d arguments (expected %d)\n",
			 identifier,
			 _argument_list_length (arguments),
			 _string_list_length (macro->parameters));
		return NULL;
	}

	/* Perform argument substitution on the replacement list. */
	substituted = _token_list_create (arguments);

	for (node = macro->replacements->head; node; node = node->next)
	{
		if (node->token->type == IDENTIFIER &&
		    _string_list_contains (macro->parameters,
					   node->token->value.str,
					   &parameter_index))
		{
			token_list_t *argument;
			argument = _argument_list_member_at (arguments,
							     parameter_index);
			/* Before substituting, we expand the argument
			 * tokens, or append a placeholder token for
			 * an empty argument. */
			if (argument->head) {
				_glcpp_parser_expand_token_list (parser,
								 argument);
				_token_list_append_list (substituted, argument);
			} else {
				token_t *new_token;

				new_token = _token_create_ival (substituted,
								PLACEHOLDER,
								PLACEHOLDER);
				_token_list_append (substituted, new_token);
			}
		} else {
			_token_list_append (substituted, node->token);
		}
	}

	/* After argument substitution, and before further expansion
	 * below, implement token pasting. */

	_token_list_trim_trailing_space (substituted);

	node = substituted->head;
	while (node)
	{
		token_node_t *next_non_space;

		/* Look ahead for a PASTE token, skipping space. */
		next_non_space = node->next;
		while (next_non_space && next_non_space->token->type == SPACE)
			next_non_space = next_non_space->next;

		if (next_non_space == NULL)
			break;

		if (next_non_space->token->type != PASTE) {
			node = next_non_space;
			continue;
		}

		/* Now find the next non-space token after the PASTE. */
		next_non_space = next_non_space->next;
		while (next_non_space && next_non_space->token->type == SPACE)
			next_non_space = next_non_space->next;

		if (next_non_space == NULL) {
			fprintf (stderr, "Error: '##' cannot appear at either end of a macro expansion\n");
			return NULL;
		}

		node->token = _token_paste (node->token, next_non_space->token);
		node->next = next_non_space->next;
		if (next_non_space == substituted->tail)
			substituted->tail = node;

		node = node->next;
	}

	substituted->non_space_tail = substituted->tail;

	return substituted;
}

/* Compute the complete expansion of node, (and subsequent nodes after
 * 'node' in the case that 'node' is a function-like macro and
 * subsequent nodes are arguments).
 *
 * Returns NULL if node is a simple token with no expansion.
 *
 * Otherwise, returns the token list that results from the expansion
 * and sets *last to the last node in the list that was consumed by
 * the expansion. Specificallty, *last will be set as follows:
 *
 *	As 'node' in the case of object-like macro expansion.
 *
 *	As the token of the closing right parenthesis in the case of
 *	function-like macro expansion.
 */
static token_list_t *
_glcpp_parser_expand_node (glcpp_parser_t *parser,
			   token_node_t *node,
			   token_node_t **last)
{
	token_t *token = node->token;
	const char *identifier;
	macro_t *macro;

	/* We only expand identifiers */
	if (token->type != IDENTIFIER) {
		/* We change any COMMA into a COMMA_FINAL to prevent
		 * it being mistaken for an argument separator
		 * later. */
		if (token->type == ',') {
			token->type = COMMA_FINAL;
			token->value.ival = COMMA_FINAL;
		}

		return NULL;
	}

	/* Look up this identifier in the hash table. */
	identifier = token->value.str;
	macro = hash_table_find (parser->defines, identifier);

	/* Not a macro, so no expansion needed. */
	if (macro == NULL)
		return NULL;

	/* Finally, don't expand this macro if we're already actively
	 * expanding it, (to avoid infinite recursion). */
	if (_active_list_contains (parser->active, identifier)) {
		/* We change the token type here from IDENTIFIER to
		 * OTHER to prevent any future expansion of this
		 * unexpanded token. */
		char *str;
		token_list_t *expansion;
		token_t *final;

		str = xtalloc_strdup (parser, token->value.str);
		final = _token_create_str (parser, OTHER, str);
		expansion = _token_list_create (parser);
		_token_list_append (expansion, final);
		*last = node;
		return expansion;
	}

	if (! macro->is_function)
	{
		*last = node;

		if (macro->replacements == NULL)
			return _token_list_create (parser);

		return _token_list_copy (parser, macro->replacements);
	}

	return _glcpp_parser_expand_function (parser, node, last);
}

/* Push a new identifier onto the active list, returning the new list.
 *
 * Here, 'marker' is the token node that appears in the list after the
 * expansion of 'identifier'. That is, when the list iterator begins
 * examinging 'marker', then it is time to pop this node from the
 * active stack.
 */
active_list_t *
_active_list_push (active_list_t *list,
		   const char *identifier,
		   token_node_t *marker)
{
	active_list_t *node;

	node = xtalloc (list, active_list_t);
	node->identifier = xtalloc_strdup (node, identifier);
	node->marker = marker;
	node->next = list;

	return node;
}

active_list_t *
_active_list_pop (active_list_t *list)
{
	active_list_t *node = list;

	if (node == NULL)
		return NULL;

	node = list->next;
	talloc_free (list);

	return node;
}

int
_active_list_contains (active_list_t *list, const char *identifier)
{
	active_list_t *node;

	if (list == NULL)
		return 0;

	for (node = list; node; node = node->next)
		if (strcmp (node->identifier, identifier) == 0)
			return 1;

	return 0;
}

/* Walk over the token list replacing nodes with their expansion.
 * Whenever nodes are expanded the walking will walk over the new
 * nodes, continuing to expand as necessary. The results are placed in
 * 'list' itself;
 */
static void
_glcpp_parser_expand_token_list (glcpp_parser_t *parser,
				 token_list_t *list)
{
	token_node_t *node_prev;
	token_node_t *node, *last;
	token_list_t *expansion;

	if (list == NULL)
		return;

	_token_list_trim_trailing_space (list);

	node_prev = NULL;
	node = list->head;

	while (node) {

		while (parser->active && parser->active->marker == node)
			parser->active = _active_list_pop (parser->active);

		/* Find the expansion for node, which will replace all
		 * nodes from node to last, inclusive. */
		expansion = _glcpp_parser_expand_node (parser, node, &last);
		if (expansion) {
			token_node_t *n;

			for (n = node; n != last->next; n = n->next)
				while (parser->active &&
				       parser->active->marker == n)
				{
					parser->active = _active_list_pop (parser->active);
				}

			parser->active = _active_list_push (parser->active,
							    node->token->value.str,
							    last->next);
			
			/* Splice expansion into list, supporting a
			 * simple deletion if the expansion is
			 * empty. */
			if (expansion->head) {
				if (node_prev)
					node_prev->next = expansion->head;
				else
					list->head = expansion->head;
				expansion->tail->next = last->next;
				if (last == list->tail)
					list->tail = expansion->tail;
			} else {
				if (node_prev)
					node_prev->next = last->next;
				else
					list->head = last->next;
				if (last == list->tail)
					list->tail == NULL;
			}
		} else {
			node_prev = node;
		}
		node = node_prev ? node_prev->next : list->head;
	}

	while (parser->active)
		parser->active = _active_list_pop (parser->active);

	list->non_space_tail = list->tail;
}

void
_glcpp_parser_print_expanded_token_list (glcpp_parser_t *parser,
					 token_list_t *list)
{
	if (list == NULL)
		return;

	_glcpp_parser_expand_token_list (parser, list);

	_token_list_trim_trailing_space (list);

	_token_list_print (list);
}

void
_define_object_macro (glcpp_parser_t *parser,
		      const char *identifier,
		      token_list_t *replacements)
{
	macro_t *macro;

	macro = xtalloc (parser, macro_t);

	macro->is_function = 0;
	macro->parameters = NULL;
	macro->identifier = talloc_strdup (macro, identifier);
	macro->replacements = talloc_steal (macro, replacements);

	hash_table_insert (parser->defines, macro, identifier);
}

void
_define_function_macro (glcpp_parser_t *parser,
			const char *identifier,
			string_list_t *parameters,
			token_list_t *replacements)
{
	macro_t *macro;

	macro = xtalloc (parser, macro_t);

	macro->is_function = 1;
	macro->parameters = talloc_steal (macro, parameters);
	macro->identifier = talloc_strdup (macro, identifier);
	macro->replacements = talloc_steal (macro, replacements);

	hash_table_insert (parser->defines, macro, identifier);
}

static int
glcpp_parser_lex (glcpp_parser_t *parser)
{
	token_node_t *node;
	int ret;

	if (parser->lex_from_list == NULL) {
		ret = glcpp_lex (parser->scanner);

		/* XXX: This ugly block of code exists for the sole
		 * purpose of converting a NEWLINE token into a SPACE
		 * token, but only in the case where we have seen a
		 * function-like macro name, but have not yet seen its
		 * closing parenthesis.
		 *
		 * There's perhaps a more compact way to do this with
		 * mid-rule actions in the grammar.
		 *
		 * I'm definitely not pleased with the complexity of
		 * this code here.
		 */
		if (parser->newline_as_space)
		{
			if (ret == '(') {
				parser->paren_count++;
			} else if (ret == ')') {
				parser->paren_count--;
				if (parser->paren_count == 0)
					parser->newline_as_space = 0;
			} else if (ret == NEWLINE) {
				ret = SPACE;
			} else if (ret != SPACE) {
				if (parser->paren_count == 0)
					parser->newline_as_space = 0;
			}
		}
		else if (parser->in_control_line)
		{
			if (ret == NEWLINE)
				parser->in_control_line = 0;
		}
		else if (ret == HASH_DEFINE_OBJ || ret == HASH_DEFINE_FUNC ||
			   ret == HASH_UNDEF || ret == HASH_IF ||
			   ret == HASH_IFDEF || ret == HASH_IFNDEF ||
			   ret == HASH_ELIF || ret == HASH_ELSE ||
			   ret == HASH_ENDIF || ret == HASH)
		{
			parser->in_control_line = 1;
		}
		else if (ret == IDENTIFIER)
		{
			macro_t *macro;
			macro = hash_table_find (parser->defines,
						 yylval.str);
			if (macro && macro->is_function) {
				parser->newline_as_space = 1;
				parser->paren_count = 0;
			}
		}

		return ret;
	}

	node = parser->lex_from_node;

	if (node == NULL) {
		talloc_free (parser->lex_from_list);
		parser->lex_from_list = NULL;
		return NEWLINE;
	}

	yylval = node->token->value;
	ret = node->token->type;

	parser->lex_from_node = node->next;

	return ret;
}

static void
glcpp_parser_lex_from (glcpp_parser_t *parser, token_list_t *list)
{
	token_node_t *node;

	assert (parser->lex_from_list == NULL);

	/* Copy list, eliminating any space tokens. */
	parser->lex_from_list = _token_list_create (parser);

	for (node = list->head; node; node = node->next) {
		if (node->token->type == SPACE)
			continue;
		_token_list_append (parser->lex_from_list, node->token);
	}

	talloc_free (list);

	parser->lex_from_node = parser->lex_from_list->head;

	/* It's possible the list consisted of nothing but whitespace. */
	if (parser->lex_from_node == NULL) {
		talloc_free (parser->lex_from_list);
		parser->lex_from_list = NULL;
	}
}

static void
_glcpp_parser_skip_stack_push_if (glcpp_parser_t *parser, int condition)
{
	skip_type_t current = SKIP_NO_SKIP;
	skip_node_t *node;

	if (parser->skip_stack)
		current = parser->skip_stack->type;

	node = xtalloc (parser, skip_node_t);

	if (current == SKIP_NO_SKIP) {
		if (condition)
			node->type = SKIP_NO_SKIP;
		else
			node->type = SKIP_TO_ELSE;
	} else {
		node->type = SKIP_TO_ENDIF;
	}

	node->next = parser->skip_stack;
	parser->skip_stack = node;
}

static void
_glcpp_parser_skip_stack_change_if (glcpp_parser_t *parser, const char *type,
				    int condition)
{
	if (parser->skip_stack == NULL) {
		fprintf (stderr, "Error: %s without #if\n", type);
		exit (1);
	}

	if (parser->skip_stack->type == SKIP_TO_ELSE) {
		if (condition)
			parser->skip_stack->type = SKIP_NO_SKIP;
	} else {
		parser->skip_stack->type = SKIP_TO_ENDIF;
	}
}

static void
_glcpp_parser_skip_stack_pop (glcpp_parser_t *parser)
{
	skip_node_t *node;

	if (parser->skip_stack == NULL) {
		fprintf (stderr, "Error: #endif without #if\n");
		exit (1);
	}

	node = parser->skip_stack;
	parser->skip_stack = node->next;
	talloc_free (node);
}