Permalink
Switch branches/tags
Find file
Fetching contributors…
Cannot retrieve contributors at this time
567 lines (496 sloc) 15 KB
/*
* Binary encode X templates from text format.
*
* Copyright 2011 Dylan Smith
*
* This library is free software; you can redistribute it and/or
* modify it under the terms of the GNU Lesser General Public
* License as published by the Free Software Foundation; either
* version 2.1 of the License, or (at your option) any later version.
*
* This library is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
* Lesser General Public License for more details.
*
* You should have received a copy of the GNU Lesser General Public
* License along with this library; if not, write to the Free Software
* Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301, USA
*/
#include "config.h"
#include "wine/port.h"
#include <signal.h>
#include <stdio.h>
#include <stdlib.h>
#ifdef HAVE_GETOPT_H
# include <getopt.h>
#endif
#ifdef HAVE_UNISTD_H
# include <unistd.h>
#endif
#include "windef.h"
#include "guiddef.h"
#define ARRAY_SIZE(array) (sizeof(array)/sizeof(array[0]))
#define TOKEN_NAME 1
#define TOKEN_STRING 2
#define TOKEN_INTEGER 3
#define TOKEN_GUID 5
#define TOKEN_INTEGER_LIST 6
#define TOKEN_FLOAT_LIST 7
#define TOKEN_OBRACE 10
#define TOKEN_CBRACE 11
#define TOKEN_OPAREN 12
#define TOKEN_CPAREN 13
#define TOKEN_OBRACKET 14
#define TOKEN_CBRACKET 15
#define TOKEN_OANGLE 16
#define TOKEN_CANGLE 17
#define TOKEN_DOT 18
#define TOKEN_COMMA 19
#define TOKEN_SEMICOLON 20
#define TOKEN_TEMPLATE 31
#define TOKEN_WORD 40
#define TOKEN_DWORD 41
#define TOKEN_FLOAT 42
#define TOKEN_DOUBLE 43
#define TOKEN_CHAR 44
#define TOKEN_UCHAR 45
#define TOKEN_SWORD 46
#define TOKEN_SDWORD 47
#define TOKEN_VOID 48
#define TOKEN_LPSTR 49
#define TOKEN_UNICODE 50
#define TOKEN_CSTRING 51
#define TOKEN_ARRAY 52
struct parser
{
FILE *infile;
FILE *outfile;
int line_no;
UINT bytes_output;
BOOL(*write_bytes)(struct parser *parser, const BYTE *data, DWORD size);
BOOL error;
};
struct keyword
{
const char *word;
WORD token;
};
static const struct keyword reserved_words[] = {
{"ARRAY", TOKEN_ARRAY},
{"CHAR", TOKEN_CHAR},
{"CSTRING", TOKEN_CSTRING},
{"DOUBLE", TOKEN_DOUBLE},
{"DWORD", TOKEN_DWORD},
{"FLOAT", TOKEN_FLOAT},
{"SDWORD", TOKEN_SDWORD},
{"STRING", TOKEN_LPSTR},
{"SWORD", TOKEN_SWORD},
{"TEMPLATE", TOKEN_TEMPLATE},
{"UCHAR", TOKEN_UCHAR},
{"UNICODE", TOKEN_UNICODE},
{"VOID", TOKEN_VOID},
{"WORD", TOKEN_WORD}
};
static char *program_name;
static inline BOOL read_byte(struct parser *parser, char *byte)
{
int c = fgetc(parser->infile);
*byte = c;
return c != EOF;
}
static inline BOOL unread_byte(struct parser *parser, char last_byte)
{
return ungetc(last_byte, parser->infile) != EOF;
}
static inline BOOL read_bytes(struct parser *parser, void *data, DWORD size)
{
return fread(data, size, 1, parser->infile) > 0;
}
static BOOL write_c_hex_bytes(struct parser *parser, const BYTE *data, DWORD size)
{
while (size--)
{
if (parser->bytes_output % 12 == 0)
fprintf(parser->outfile, "\n ");
fprintf(parser->outfile, " 0x%02x,", *data++);
parser->bytes_output++;
}
return TRUE;
}
static BOOL write_raw_bytes(struct parser *parser, const BYTE *data, DWORD size)
{
return fwrite(data, size, 1, parser->outfile) > 0;
}
static inline BOOL write_bytes(struct parser *parser, const void *data, DWORD size)
{
return parser->write_bytes(parser, data, size);
}
static inline BOOL write_byte(struct parser *parser, BYTE value)
{
return write_bytes(parser, &value, sizeof(value));
}
static inline BOOL write_word(struct parser *parser, WORD value)
{
return write_bytes(parser, &value, sizeof(value));
}
static inline BOOL write_dword(struct parser *parser, DWORD value)
{
return write_bytes(parser, &value, sizeof(value));
}
static int compare_names(const void *a, const void *b)
{
return strcasecmp(*(const char **)a, *(const char **)b);
}
static BOOL parse_keyword(struct parser *parser, const char *name)
{
const struct keyword *keyword;
keyword = bsearch(&name, reserved_words, ARRAY_SIZE(reserved_words),
sizeof(reserved_words[0]), compare_names);
if (!keyword)
return FALSE;
return write_word(parser, keyword->token);
}
static BOOL parse_guid(struct parser *parser)
{
char buf[39];
GUID guid;
DWORD tab[10];
BOOL ret;
static const char *guidfmt = "<%08X-%04X-%04X-%02X%02X-%02X%02X%02X%02X%02X%02X>";
buf[0] = '<';
if (!read_bytes(parser, buf + 1, 37)) {
fprintf(stderr, "%s: Truncated GUID (line %d)\n",
program_name, parser->line_no);
parser->error = TRUE;
return FALSE;
}
buf[38] = 0;
ret = sscanf(buf, guidfmt, &guid.Data1, tab, tab+1, tab+2, tab+3, tab+4, tab+5, tab+6, tab+7, tab+8, tab+9);
if (ret != 11) {
fprintf(stderr, "%s: Invalid GUID '%s' (line %d)\n",
program_name, buf, parser->line_no);
parser->error = TRUE;
return FALSE;
}
guid.Data2 = tab[0];
guid.Data3 = tab[1];
guid.Data4[0] = tab[2];
guid.Data4[1] = tab[3];
guid.Data4[2] = tab[4];
guid.Data4[3] = tab[5];
guid.Data4[4] = tab[6];
guid.Data4[5] = tab[7];
guid.Data4[6] = tab[8];
guid.Data4[7] = tab[9];
return write_word(parser, TOKEN_GUID) &&
write_bytes(parser, &guid, sizeof(guid));
}
static BOOL parse_name(struct parser *parser)
{
char c;
int len = 0;
char name[512];
while (read_byte(parser, &c) && len < sizeof(name) &&
(isalnum(c) || c == '_' || c == '-'))
{
if (len + 1 < sizeof(name))
name[len++] = c;
}
unread_byte(parser, c);
name[len] = 0;
if (parse_keyword(parser, name)) {
return TRUE;
} else {
return write_word(parser, TOKEN_NAME) &&
write_dword(parser, len) &&
write_bytes(parser, name, len);
}
}
static BOOL parse_number(struct parser *parser)
{
int len = 0;
char c;
char buffer[512];
BOOL dot = FALSE;
BOOL ret;
while (read_byte(parser, &c) &&
((!len && c == '-') || (!dot && c == '.') || isdigit(c)))
{
if (len + 1 < sizeof(buffer))
buffer[len++] = c;
if (c == '.')
dot = TRUE;
}
unread_byte(parser, c);
buffer[len] = 0;
if (dot) {
float value;
ret = sscanf(buffer, "%f", &value);
if (!ret) {
fprintf(stderr, "%s: Invalid float token (line %d).\n",
program_name, parser->line_no);
parser->error = TRUE;
} else {
ret = write_word(parser, TOKEN_FLOAT) &&
write_bytes(parser, &value, sizeof(value));
}
} else {
int value;
ret = sscanf(buffer, "%d", &value);
if (!ret) {
fprintf(stderr, "%s: Invalid integer token (line %d).\n",
program_name, parser->line_no);
parser->error = TRUE;
} else {
ret = write_word(parser, TOKEN_INTEGER) &&
write_dword(parser, value);
}
}
return ret;
}
static BOOL parse_token(struct parser *parser)
{
char c;
if (!read_byte(parser, &c))
return FALSE;
switch (c)
{
case '\n':
parser->line_no++;
/* fall through */
case '\r':
case ' ':
case '\t':
return TRUE;
case '{': return write_word(parser, TOKEN_OBRACE);
case '}': return write_word(parser, TOKEN_CBRACE);
case '[': return write_word(parser, TOKEN_OBRACKET);
case ']': return write_word(parser, TOKEN_CBRACKET);
case '(': return write_word(parser, TOKEN_OPAREN);
case ')': return write_word(parser, TOKEN_CPAREN);
case ',': return write_word(parser, TOKEN_COMMA);
case ';': return write_word(parser, TOKEN_SEMICOLON);
case '.': return write_word(parser, TOKEN_DOT);
case '/':
if (!read_byte(parser, &c) || c != '/') {
fprintf(stderr, "%s: Invalid single '/' comment token (line %d).\n",
program_name, parser->line_no);
parser->error = TRUE;
return FALSE;
}
/* fall through */
case '#':
while (read_byte(parser, &c) && c != '\n');
return c == '\n';
case '<':
return parse_guid(parser);
case '"':
{
int len = 0;
char buffer[512];
/* FIXME: Handle '\' (e.g. "valid\"string") */
while (read_byte(parser, &c) && c != '"') {
if (len + 1 < sizeof(buffer))
buffer[len++] = c;
}
if (c != '"') {
fprintf(stderr, "%s: Unterminated string (line %d).\n",
program_name, parser->line_no);
parser->error = TRUE;
return FALSE;
}
return write_word(parser, TOKEN_STRING) &&
write_dword(parser, len) &&
write_bytes(parser, buffer, len);
}
default:
unread_byte(parser, c);
if (isdigit(c) || c == '-')
return parse_number(parser);
if (isalpha(c) || c == '_')
return parse_name(parser);
fprintf(stderr, "%s: Invalid character (%d) to start token (line %d).\n",
program_name, c, parser->line_no);
parser->error = TRUE;
return FALSE;
}
return TRUE;
}
static const char *output_file;
static void cleanup_files(void)
{
if (output_file) unlink(output_file);
}
static void exit_on_signal( int sig )
{
exit(1); /* this will call the atexit functions */
}
static void usage(void)
{
fprintf(stderr, "Usage: %s [OPTIONS] INFILE\n"
"Options:\n"
" -i NAME Output to a c header file, data in variable NAME\n"
" -s NAME In a c header file, define NAME to be the data size\n"
" -o FILE Write output to FILE\n",
program_name);
}
static char *option_inc_var_name = NULL;
static char *option_inc_size_name = NULL;
static const char *option_outfile_name = "-";
static char **parse_options(int argc, char **argv)
{
int optc;
while ((optc = getopt(argc, argv, "hi:o:s:")) != -1)
{
switch (optc)
{
case 'h':
usage();
exit(0);
case 'i':
option_inc_var_name = strdup(optarg);
break;
case 'o':
option_outfile_name = strdup(optarg);
break;
case 's':
option_inc_size_name = strdup(optarg);
break;
}
}
return &argv[optind];
}
int main(int argc, char **argv)
{
const char *infile_name;
char header[16];
struct parser parser;
char **args;
char *header_name = NULL;
program_name = argv[0];
args = parse_options(argc, argv);
infile_name = *args++;
if (!infile_name || *args)
{
usage();
return 1;
}
parser.infile = stdin;
parser.outfile = NULL;
parser.error = FALSE;
if (!strcmp(infile_name, "-")) {
infile_name = "stdin";
} else if (!(parser.infile = fopen(infile_name, "rb"))) {
perror(infile_name);
goto error;
}
if (!read_bytes(&parser, header, sizeof(header))) {
fprintf(stderr, "%s: Failed to read file header\n", program_name);
goto error;
}
if (strncmp(header, "xof ", 4))
{
fprintf(stderr, "%s: Invalid magic value '%.4s'\n", program_name, header);
goto error;
}
if (strncmp(header + 4, "0302", 4) && strncmp(header + 4, "0303", 4))
{
fprintf(stderr, "%s: Unsupported version '%.4s'\n", program_name, header + 4);
goto error;
}
if (strncmp(header + 8, "txt ", 4))
{
fprintf(stderr, "%s: Only support conversion from text encoded X files.",
program_name);
goto error;
}
if (strncmp(header + 12, "0032", 4) && strncmp(header + 12, "0064", 4))
{
fprintf(stderr, "%s: Only 32-bit or 64-bit float format supported, not '%.4s'.\n",
program_name, header + 12);
goto error;
}
if (!strcmp(option_outfile_name, "-")) {
option_outfile_name = "stdout";
parser.outfile = stdout;
} else {
output_file = option_outfile_name;
atexit(cleanup_files);
signal(SIGTERM, exit_on_signal);
signal(SIGINT, exit_on_signal);
#ifdef SIGHUP
signal(SIGHUP, exit_on_signal);
#endif
if (!(parser.outfile = fopen(output_file, "wb"))) {
perror(option_outfile_name);
goto error;
}
}
if (option_inc_var_name)
{
char *str_ptr;
header_name = strrchr(option_outfile_name, '/');
if (header_name)
header_name = strdup(header_name + 1);
else
header_name = strdup(option_outfile_name);
if (!header_name) {
fprintf(stderr, "Out of memory\n");
goto error;
}
str_ptr = header_name;
while (*str_ptr) {
if (*str_ptr == '.')
*str_ptr = '_';
else
*str_ptr = toupper(*str_ptr);
str_ptr++;
}
fprintf(parser.outfile,
"/* File generated automatically from %s; do not edit */\n"
"\n"
"#ifndef __WINE_%s\n"
"#define __WINE_%s\n"
"\n"
"unsigned char %s[] = {",
infile_name, header_name, header_name, option_inc_var_name);
if (ferror(parser.outfile))
goto error;
parser.write_bytes = &write_c_hex_bytes;
} else {
parser.write_bytes = &write_raw_bytes;
}
parser.bytes_output = 0;
if (!write_bytes(&parser, "xof 0302bin 0064", 16))
goto error;
parser.line_no = 1;
while (parse_token(&parser));
if (parser.error || ferror(parser.outfile) || ferror(parser.infile))
goto error;
if (option_inc_var_name)
{
fprintf(parser.outfile, "\n};\n\n");
if (option_inc_size_name)
fprintf(parser.outfile, "#define %s %u\n\n", option_inc_size_name, parser.bytes_output);
fprintf(parser.outfile, "#endif /* __WINE_%s */\n", header_name);
if (ferror(parser.outfile))
goto error;
}
fclose(parser.infile);
fclose(parser.outfile);
output_file = NULL;
return 0;
error:
if (parser.infile) {
if (ferror(parser.infile))
perror(infile_name);
fclose(parser.infile);
}
if (parser.outfile) {
if (ferror(parser.outfile))
perror(option_outfile_name);
fclose(parser.outfile);
}
return 1;
}