Sweden-Number/tools/wmc/utils.c

599 lines
14 KiB
C

/*
* Utility routines
*
* Copyright 1998,2000 Bertho A. Stultiens
*
* 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 <stdio.h>
#include <stdlib.h>
#include <stdarg.h>
#include <string.h>
#include <assert.h>
#include <ctype.h>
#include "wmctypes.h"
#include "winnls.h"
#include "utils.h"
#include "wmc.h"
#define SUPPRESS_YACC_ERROR_MESSAGE
static void generic_msg(const char *s, const char *t, va_list ap)
{
fprintf(stderr, "%s:%d:%d: %s: ", input_name ? input_name : "stdin", line_number, char_number, t);
vfprintf(stderr, s, ap);
}
/*
* The yyerror routine should not exit because we use the error-token
* to determine the syntactic error in the source. However, YACC
* uses the same routine to print an error just before the error
* token is reduced.
* The extra routine 'xyyerror' is used to exit after giving a real
* message.
*/
int mcy_error(const char *s, ...)
{
#ifndef SUPPRESS_YACC_ERROR_MESSAGE
va_list ap;
va_start(ap, s);
generic_msg(s, "Yacc error", ap);
va_end(ap);
#endif
return 1;
}
int xyyerror(const char *s, ...)
{
va_list ap;
va_start(ap, s);
generic_msg(s, "Error", ap);
va_end(ap);
exit(1);
return 1;
}
int mcy_warning(const char *s, ...)
{
va_list ap;
va_start(ap, s);
generic_msg(s, "Warning", ap);
va_end(ap);
return 0;
}
void internal_error(const char *file, int line, const char *s, ...)
{
va_list ap;
va_start(ap, s);
fprintf(stderr, "Internal error (please report) %s %d: ", file, line);
vfprintf(stderr, s, ap);
va_end(ap);
exit(3);
}
void fatal_perror( const char *msg, ... )
{
va_list valist;
va_start( valist, msg );
fprintf(stderr, "Error: ");
vfprintf( stderr, msg, valist );
perror( " " );
va_end( valist );
exit(2);
}
void error(const char *s, ...)
{
va_list ap;
va_start(ap, s);
fprintf(stderr, "Error: ");
vfprintf(stderr, s, ap);
va_end(ap);
exit(2);
}
void warning(const char *s, ...)
{
va_list ap;
va_start(ap, s);
fprintf(stderr, "Warning: ");
vfprintf(stderr, s, ap);
va_end(ap);
}
char *dup_basename(const char *name, const char *ext)
{
int namelen;
int extlen = strlen(ext);
char *base;
char *slash;
if(!name)
name = "wmc.tab";
slash = strrchr(name, '/');
if (slash)
name = slash + 1;
namelen = strlen(name);
/* +4 for later extension and +1 for '\0' */
base = xmalloc(namelen +4 +1);
strcpy(base, name);
if(!strcasecmp(name + namelen-extlen, ext))
{
base[namelen - extlen] = '\0';
}
return base;
}
void *xmalloc(size_t size)
{
void *res;
assert(size > 0);
res = malloc(size);
if(res == NULL)
{
error("Virtual memory exhausted.\n");
}
memset(res, 0x55, size);
return res;
}
void *xrealloc(void *p, size_t size)
{
void *res;
assert(size > 0);
res = realloc(p, size);
if(res == NULL)
{
error("Virtual memory exhausted.\n");
}
return res;
}
char *xstrdup(const char *str)
{
char *s;
assert(str != NULL);
s = xmalloc(strlen(str)+1);
return strcpy(s, str);
}
char *strmake( const char* fmt, ... )
{
int n;
size_t size = 100;
va_list ap;
for (;;)
{
char *p = xmalloc( size );
va_start( ap, fmt );
n = vsnprintf( p, size, fmt, ap );
va_end( ap );
if (n == -1) size *= 2;
else if ((size_t)n >= size) size = n + 1;
else return p;
free( p );
}
}
int strendswith( const char *str, const char *end )
{
int l = strlen(str);
int m = strlen(end);
return l >= m && !strcmp( str + l - m, end );
}
int unistrlen(const WCHAR *s)
{
int n;
for(n = 0; *s; n++, s++)
;
return n;
}
WCHAR *unistrcpy(WCHAR *dst, const WCHAR *src)
{
WCHAR *t = dst;
while(*src)
*t++ = *src++;
*t = 0;
return dst;
}
WCHAR *xunistrdup(const WCHAR * str)
{
WCHAR *s;
assert(str != NULL);
s = xmalloc((unistrlen(str)+1) * sizeof(WCHAR));
return unistrcpy(s, str);
}
int unistricmp(const WCHAR *s1, const WCHAR *s2)
{
int i;
int once = 0;
static const char warn[] = "Don't know the uppercase equivalent of non ascii characters;"
"comparison might yield wrong results";
while(*s1 && *s2)
{
if((*s1 & 0xffff) > 0x7f || (*s2 & 0xffff) > 0x7f)
{
if(!once)
{
once++;
mcy_warning(warn);
}
i = *s1++ - *s2++;
}
else
i = toupper(*s1++) - toupper(*s2++);
if(i)
return i;
}
if((*s1 & 0xffff) > 0x7f || (*s2 & 0xffff) > 0x7f)
{
if(!once)
mcy_warning(warn);
return *s1 - *s2;
}
else
return toupper(*s1) - toupper(*s2);
}
int unistrcmp(const WCHAR *s1, const WCHAR *s2)
{
int i;
while(*s1 && *s2)
{
i = *s1++ - *s2++;
if(i)
return i;
}
return *s1 - *s2;
}
WCHAR *utf8_to_unicode( const char *src, int srclen, int *dstlen )
{
static const char utf8_length[128] =
{
0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, /* 0x80-0x8f */
0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, /* 0x90-0x9f */
0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, /* 0xa0-0xaf */
0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, /* 0xb0-0xbf */
0,0,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* 0xc0-0xcf */
1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, /* 0xd0-0xdf */
2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2, /* 0xe0-0xef */
3,3,3,3,3,0,0,0,0,0,0,0,0,0,0,0 /* 0xf0-0xff */
};
static const unsigned char utf8_mask[4] = { 0x7f, 0x1f, 0x0f, 0x07 };
const char *srcend = src + srclen;
int len, res;
WCHAR *ret, *dst;
dst = ret = xmalloc( (srclen + 1) * sizeof(WCHAR) );
while (src < srcend)
{
unsigned char ch = *src++;
if (ch < 0x80) /* special fast case for 7-bit ASCII */
{
*dst++ = ch;
continue;
}
len = utf8_length[ch - 0x80];
if (len && src + len <= srcend)
{
res = ch & utf8_mask[len];
switch (len)
{
case 3:
if ((ch = *src ^ 0x80) >= 0x40) break;
res = (res << 6) | ch;
src++;
if (res < 0x10) break;
case 2:
if ((ch = *src ^ 0x80) >= 0x40) break;
res = (res << 6) | ch;
if (res >= 0x110000 >> 6) break;
src++;
if (res < 0x20) break;
if (res >= 0xd800 >> 6 && res <= 0xdfff >> 6) break;
case 1:
if ((ch = *src ^ 0x80) >= 0x40) break;
res = (res << 6) | ch;
src++;
if (res < 0x80) break;
if (res <= 0xffff) *dst++ = res;
else
{
res -= 0x10000;
*dst++ = 0xd800 | (res >> 10);
*dst++ = 0xdc00 | (res & 0x3ff);
}
continue;
}
}
*dst++ = 0xfffd;
}
*dst = 0;
*dstlen = dst - ret;
return ret;
}
char *unicode_to_utf8( const WCHAR *src, int srclen, int *dstlen )
{
char *ret, *dst;
dst = ret = xmalloc( srclen * 3 + 1 );
for ( ; srclen; srclen--, src++)
{
unsigned int ch = *src;
if (ch < 0x80) /* 0x00-0x7f: 1 byte */
{
*dst++ = ch;
continue;
}
if (ch < 0x800) /* 0x80-0x7ff: 2 bytes */
{
dst[1] = 0x80 | (ch & 0x3f);
ch >>= 6;
dst[0] = 0xc0 | ch;
dst += 2;
continue;
}
if (ch >= 0xd800 && ch <= 0xdbff && srclen > 1 && src[1] >= 0xdc00 && src[1] <= 0xdfff)
{
/* 0x10000-0x10ffff: 4 bytes */
ch = 0x10000 + ((ch & 0x3ff) << 10) + (src[1] & 0x3ff);
dst[3] = 0x80 | (ch & 0x3f);
ch >>= 6;
dst[2] = 0x80 | (ch & 0x3f);
ch >>= 6;
dst[1] = 0x80 | (ch & 0x3f);
ch >>= 6;
dst[0] = 0xf0 | ch;
dst += 4;
src++;
srclen--;
continue;
}
if (ch >= 0xd800 && ch <= 0xdfff) ch = 0xfffd; /* invalid surrogate pair */
/* 0x800-0xffff: 3 bytes */
dst[2] = 0x80 | (ch & 0x3f);
ch >>= 6;
dst[1] = 0x80 | (ch & 0x3f);
ch >>= 6;
dst[0] = 0xe0 | ch;
dst += 3;
}
*dst = 0;
*dstlen = dst - ret;
return ret;
}
#ifdef _WIN32
int is_valid_codepage(int id)
{
return IsValidCodePage( id );
}
WCHAR *codepage_to_unicode( int codepage, const char *src, int srclen, int *dstlen )
{
WCHAR *dst = xmalloc( (srclen + 1) * sizeof(WCHAR) );
DWORD ret = MultiByteToWideChar( codepage, MB_ERR_INVALID_CHARS, src, srclen, dst, srclen );
if (!ret) return NULL;
dst[ret] = 0;
*dstlen = ret;
return dst;
}
#else /* _WIN32 */
struct nls_info
{
unsigned short codepage;
unsigned short unidef;
unsigned short trans_unidef;
unsigned short *cp2uni;
unsigned short *dbcs_offsets;
};
static struct nls_info nlsinfo[128];
static void init_nls_info( struct nls_info *info, unsigned short *ptr )
{
unsigned short hdr_size = ptr[0];
info->codepage = ptr[1];
info->unidef = ptr[4];
info->trans_unidef = ptr[6];
ptr += hdr_size;
info->cp2uni = ++ptr;
ptr += 256;
if (*ptr++) ptr += 256; /* glyph table */
info->dbcs_offsets = *ptr ? ptr + 1 : NULL;
}
static const struct nls_info *get_nls_info( unsigned int codepage )
{
struct stat st;
unsigned short *data;
char *path;
unsigned int i;
int fd;
for (i = 0; i < ARRAY_SIZE(nlsinfo) && nlsinfo[i].codepage; i++)
if (nlsinfo[i].codepage == codepage) return &nlsinfo[i];
assert( i < ARRAY_SIZE(nlsinfo) );
for (i = 0; nlsdirs[i]; i++)
{
path = strmake( "%s/c_%03u.nls", nlsdirs[i], codepage );
if ((fd = open( path, O_RDONLY )) != -1) break;
free( path );
}
if (!nlsdirs[i]) return NULL;
fstat( fd, &st );
data = xmalloc( st.st_size );
if (read( fd, data, st.st_size ) != st.st_size) error( "failed to load %s\n", path );
close( fd );
free( path );
init_nls_info( &nlsinfo[i], data );
return &nlsinfo[i];
}
int is_valid_codepage(int cp)
{
return cp == CP_UTF8 || get_nls_info( cp );
}
WCHAR *codepage_to_unicode( int codepage, const char *src, int srclen, int *dstlen )
{
const struct nls_info *info = get_nls_info( codepage );
unsigned int i;
WCHAR dbch, *dst = xmalloc( (srclen + 1) * sizeof(WCHAR) );
if (!info) error( "codepage %u not supported\n", codepage );
if (info->dbcs_offsets)
{
for (i = 0; srclen; i++, srclen--, src++)
{
unsigned short off = info->dbcs_offsets[(unsigned char)*src];
if (off)
{
if (srclen == 1) return NULL;
dbch = (src[0] << 8) | (unsigned char)src[1];
src++;
srclen--;
dst[i] = info->dbcs_offsets[off + (unsigned char)*src];
if (dst[i] == info->unidef && dbch != info->trans_unidef) return NULL;
}
else
{
dst[i] = info->cp2uni[(unsigned char)*src];
if (dst[i] == info->unidef && *src != info->trans_unidef) return NULL;
}
}
}
else
{
for (i = 0; i < srclen; i++)
{
dst[i] = info->cp2uni[(unsigned char)src[i]];
if (dst[i] == info->unidef && src[i] != info->trans_unidef) return NULL;
}
}
dst[i] = 0;
*dstlen = i;
return dst;
}
#endif /* _WIN32 */
/*******************************************************************
* buffer management
*
* Function for writing to a memory buffer.
*/
int byte_swapped = 0;
unsigned char *output_buffer;
size_t output_buffer_pos;
size_t output_buffer_size;
static void check_output_buffer_space( size_t size )
{
if (output_buffer_pos + size >= output_buffer_size)
{
output_buffer_size = max( output_buffer_size * 2, output_buffer_pos + size );
output_buffer = xrealloc( output_buffer, output_buffer_size );
}
}
void init_output_buffer(void)
{
output_buffer_size = 1024;
output_buffer_pos = 0;
output_buffer = xmalloc( output_buffer_size );
}
void flush_output_buffer( const char *name )
{
int fd = open( name, O_WRONLY | O_CREAT | O_TRUNC | O_BINARY, 0666 );
if (fd == -1) error( "Error creating %s\n", name );
if (write( fd, output_buffer, output_buffer_pos ) != output_buffer_pos)
error( "Error writing to %s\n", name );
close( fd );
free( output_buffer );
}
void put_data( const void *data, size_t size )
{
check_output_buffer_space( size );
memcpy( output_buffer + output_buffer_pos, data, size );
output_buffer_pos += size;
}
void put_byte( unsigned char val )
{
check_output_buffer_space( 1 );
output_buffer[output_buffer_pos++] = val;
}
void put_word( unsigned short val )
{
if (byte_swapped) val = (val << 8) | (val >> 8);
put_data( &val, sizeof(val) );
}
void put_dword( unsigned int val )
{
if (byte_swapped)
val = ((val << 24) | ((val << 8) & 0x00ff0000) | ((val >> 8) & 0x0000ff00) | (val >> 24));
put_data( &val, sizeof(val) );
}
void align_output( unsigned int align )
{
size_t size = align - (output_buffer_pos % align);
if (size == align) return;
check_output_buffer_space( size );
memset( output_buffer + output_buffer_pos, 0, size );
output_buffer_pos += size;
}