freebsd-skq/usr.bin/gencat/gencat.c

785 lines
18 KiB
C
Raw Normal View History

/* ex:ts=4
*/
/* $NetBSD: gencat.c,v 1.18 2003/10/27 00:12:43 lukem Exp $ */
/*
* Copyright (c) 1996 The NetBSD Foundation, Inc.
* All rights reserved.
*
* This code is derived from software contributed to The NetBSD Foundation
* by J.T. Conklin.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
* 3. All advertising materials mentioning features or use of this software
* must display the following acknowledgement:
* This product includes software developed by the NetBSD
* Foundation, Inc. and its contributors.
* 4. Neither the name of The NetBSD Foundation nor the names of its
* contributors may be used to endorse or promote products derived
* from this software without specific prior written permission.
*
* THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
* ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
* TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
* PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
* BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
* CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
* SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
* INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
* CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
* ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
* POSSIBILITY OF SUCH DAMAGE.
*/
/***********************************************************
Copyright 1990, by Alfalfa Software Incorporated, Cambridge, Massachusetts.
All Rights Reserved
Permission to use, copy, modify, and distribute this software and its
documentation for any purpose and without fee is hereby granted,
provided that the above copyright notice appear in all copies and that
both that copyright notice and this permission notice appear in
supporting documentation, and that Alfalfa's name not be used in
advertising or publicity pertaining to distribution of the software
without specific, written prior permission.
ALPHALPHA DISCLAIMS ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING
ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT SHALL
ALPHALPHA BE LIABLE FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR
ANY DAMAGES WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS,
WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION,
ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS
SOFTWARE.
If you make any modifications, bugfixes or other changes to this software
we'd appreciate it if you could send a copy to us so we can keep things
up-to-date. Many thanks.
Kee Hinckley
Alfalfa Software, Inc.
267 Allston St., #3
Cambridge, MA 02139 USA
nazgul@alfalfa.com
1995-05-30 06:41:30 +00:00
******************************************************************/
#include <sys/cdefs.h>
__FBSDID("$FreeBSD$");
#define _NLS_PRIVATE
#include <sys/types.h>
#include <sys/queue.h>
#include <arpa/inet.h> /* for htonl() */
#include <ctype.h>
#include <err.h>
#include <fcntl.h>
#include <limits.h>
#include <nl_types.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <unistd.h>
struct _msgT {
long msgId;
char *str;
LIST_ENTRY(_msgT) entries;
};
struct _setT {
long setId;
LIST_HEAD(msghead, _msgT) msghead;
LIST_ENTRY(_setT) entries;
};
LIST_HEAD(sethead, _setT) sethead;
static struct _setT *curSet;
static char *curline = NULL;
static long lineno = 0;
static char *cskip(char *);
static void error(const char *);
static char *getline(int);
static char *getmsg(int, char *, char);
static void warning(const char *, const char *);
static char *wskip(char *);
static char *xstrdup(const char *);
static void *xmalloc(size_t);
static void *xrealloc(void *, size_t);
void MCParse(int);
void MCReadCat(int);
void MCWriteCat(int);
void MCDelMsg(int);
void MCAddMsg(int, const char *);
void MCAddSet(int);
void MCDelSet(int);
void usage(void);
int main(int, char **);
void
usage()
1997-07-08 11:05:59 +00:00
{
fprintf(stderr, "usage: %s catfile msgfile ...\n", getprogname());
exit(1);
}
int
main(int argc, char **argv)
{
int ofd, ifd;
char *catfile = NULL;
int c;
#define DEPRECATEDMSG 1
#ifdef DEPRECATEDMSG
while ((c = getopt(argc, argv, "new")) != -1) {
#else
while ((c = getopt(argc, argv, "")) != -1) {
#endif
switch (c) {
#ifdef DEPRECATEDMSG
case 'n':
fprintf(stderr, "WARNING: Usage of \"-new\" argument is deprecated.\n");
case 'e':
case 'w':
break;
#endif
case '?':
default:
usage();
/* NOTREACHED */
}
}
argc -= optind;
argv += optind;
if (argc < 2) {
usage();
/* NOTREACHED */
}
catfile = *argv++;
for (; *argv; argv++) {
if ((ifd = open(*argv, O_RDONLY)) < 0)
err(1, "Unable to read %s", *argv);
MCParse(ifd);
close(ifd);
}
if ((ofd = open(catfile, O_WRONLY | O_TRUNC | O_CREAT, 0666)) < 0)
err(1, "Unable to create a new %s", catfile);
MCWriteCat(ofd);
exit(0);
}
static void
warning(const char *cptr, const char *msg)
{
fprintf(stderr, "%s: %s on line %ld\n", getprogname(), msg, lineno);
fprintf(stderr, "%s\n", curline);
if (cptr) {
char *tptr;
for (tptr = curline; tptr < cptr; ++tptr)
putc(' ', stderr);
fprintf(stderr, "^\n");
}
}
#define CORRUPT() { error("corrupt message catalog"); }
#define NOMEM() { error("out of memory"); }
static void
error(const char *msg)
{
warning(NULL, msg);
exit(1);
}
static void *
xmalloc(size_t len)
{
void *p;
if ((p = malloc(len)) == NULL)
NOMEM();
return (p);
}
static void *
xrealloc(void *ptr, size_t size)
{
if ((ptr = realloc(ptr, size)) == NULL)
NOMEM();
return (ptr);
}
static char *
xstrdup(const char *str)
{
char *nstr;
if ((nstr = strdup(str)) == NULL)
NOMEM();
return (nstr);
}
static char *
getline(int fd)
{
static long curlen = BUFSIZ;
static char buf[BUFSIZ], *bptr = buf, *bend = buf;
char *cptr, *cend;
long buflen;
if (!curline) {
curline = xmalloc(curlen);
}
++lineno;
cptr = curline;
cend = curline + curlen;
for (;;) {
for (; bptr < bend && cptr < cend; ++cptr, ++bptr) {
if (*bptr == '\n') {
*cptr = '\0';
++bptr;
return (curline);
} else
*cptr = *bptr;
}
if (cptr == cend) {
cptr = curline = xrealloc(curline, curlen *= 2);
cend = curline + curlen;
}
if (bptr == bend) {
buflen = read(fd, buf, BUFSIZ);
if (buflen <= 0) {
if (cptr > curline) {
*cptr = '\0';
return (curline);
}
return (NULL);
}
bend = buf + buflen;
bptr = buf;
}
}
}
static char *
wskip(char *cptr)
{
if (!*cptr || !isspace((unsigned char) *cptr)) {
warning(cptr, "expected a space");
return (cptr);
}
while (*cptr && isspace((unsigned char) *cptr))
++cptr;
return (cptr);
}
static char *
cskip(char *cptr)
{
if (!*cptr || isspace((unsigned char) *cptr)) {
warning(cptr, "wasn't expecting a space");
return (cptr);
}
while (*cptr && !isspace((unsigned char) *cptr))
++cptr;
return (cptr);
}
static char *
getmsg(int fd, char *cptr, char quote)
{
static char *msg = NULL;
static long msglen = 0;
long clen, i;
char *tptr;
if (quote && *cptr == quote) {
++cptr;
}
clen = strlen(cptr) + 1;
if (clen > msglen) {
if (msglen)
msg = xrealloc(msg, clen);
else
msg = xmalloc(clen);
msglen = clen;
}
tptr = msg;
while (*cptr) {
if (quote && *cptr == quote) {
char *tmp;
tmp = cptr + 1;
if (*tmp && (!isspace((unsigned char) *tmp) || *wskip(tmp))) {
warning(cptr, "unexpected quote character, ignoring");
*tptr++ = *cptr++;
} else {
*cptr = '\0';
}
} else
if (*cptr == '\\') {
++cptr;
switch (*cptr) {
case '\0':
cptr = getline(fd);
if (!cptr)
error("premature end of file");
msglen += strlen(cptr);
i = tptr - msg;
msg = xrealloc(msg, msglen);
tptr = msg + i;
break;
#define CASEOF(CS, CH) \
case CS: \
*tptr++ = CH; \
++cptr; \
break; \
CASEOF('n', '\n');
CASEOF('t', '\t');
CASEOF('v', '\v');
CASEOF('b', '\b');
CASEOF('r', '\r');
CASEOF('f', '\f');
CASEOF('"', '"');
CASEOF('\\', '\\');
default:
if (quote && *cptr == quote) {
*tptr++ = *cptr++;
} else if (isdigit((unsigned char) *cptr)) {
*tptr = 0;
for (i = 0; i < 3; ++i) {
if (!isdigit((unsigned char) *cptr))
break;
if (*cptr > '7')
warning(cptr, "octal number greater than 7?!");
*tptr *= 8;
*tptr += (*cptr - '0');
++cptr;
}
} else {
warning(cptr, "unrecognized escape sequence");
}
break;
}
} else {
*tptr++ = *cptr++;
}
}
*tptr = '\0';
return (msg);
}
void
MCParse(int fd)
{
char *cptr, *str;
int setid, msgid = 0;
char quote = 0;
/* XXX: init sethead? */
while ((cptr = getline(fd))) {
if (*cptr == '$') {
++cptr;
if (strncmp(cptr, "set", 3) == 0) {
cptr += 3;
cptr = wskip(cptr);
setid = atoi(cptr);
MCAddSet(setid);
msgid = 0;
} else if (strncmp(cptr, "delset", 6) == 0) {
cptr += 6;
cptr = wskip(cptr);
setid = atoi(cptr);
MCDelSet(setid);
} else if (strncmp(cptr, "quote", 5) == 0) {
cptr += 5;
if (!*cptr)
quote = 0;
else {
cptr = wskip(cptr);
if (!*cptr)
quote = 0;
else
quote = *cptr;
1995-05-30 06:41:30 +00:00
}
} else if (isspace((unsigned char) *cptr)) {
;
} else {
if (*cptr) {
cptr = wskip(cptr);
if (*cptr)
warning(cptr, "unrecognized line");
}
}
} else {
/*
* First check for (and eat) empty lines....
*/
if (!*cptr)
continue;
/*
* We have a digit? Start of a message. Else,
* syntax error.
*/
if (isdigit((unsigned char) *cptr)) {
msgid = atoi(cptr);
cptr = cskip(cptr);
cptr = wskip(cptr);
/* if (*cptr) ++cptr; */
} else {
warning(cptr, "neither blank line nor start of a message id");
continue;
}
/*
* If we have a message ID, but no message,
* then this means "delete this message id
* from the catalog".
*/
if (!*cptr) {
MCDelMsg(msgid);
} else {
str = getmsg(fd, cptr, quote);
MCAddMsg(msgid, str);
}
}
}
}
void
MCReadCat(int fd)
{
fd = 0;
#if 0
MCHeaderT mcHead;
MCMsgT mcMsg;
MCSetT mcSet;
msgT *msg;
setT *set;
int i;
char *data;
/* XXX init sethead? */
if (read(fd, &mcHead, sizeof(mcHead)) != sizeof(mcHead))
CORRUPT();
if (strncmp(mcHead.magic, MCMagic, MCMagicLen) != 0)
CORRUPT();
if (mcHead.majorVer != MCMajorVer)
error("unrecognized catalog version");
if ((mcHead.flags & MCGetByteOrder()) == 0)
error("wrong byte order");
if (lseek(fd, mcHead.firstSet, SEEK_SET) == -1)
CORRUPT();
for (;;) {
if (read(fd, &mcSet, sizeof(mcSet)) != sizeof(mcSet))
CORRUPT();
if (mcSet.invalid)
continue;
set = xmalloc(sizeof(setT));
memset(set, '\0', sizeof(*set));
if (cat->first) {
cat->last->next = set;
set->prev = cat->last;
cat->last = set;
} else
cat->first = cat->last = set;
set->setId = mcSet.setId;
/* Get the data */
if (mcSet.dataLen) {
data = xmalloc(mcSet.dataLen);
if (lseek(fd, mcSet.data.off, SEEK_SET) == -1)
CORRUPT();
if (read(fd, data, mcSet.dataLen) != mcSet.dataLen)
CORRUPT();
if (lseek(fd, mcSet.u.firstMsg, SEEK_SET) == -1)
CORRUPT();
for (i = 0; i < mcSet.numMsgs; ++i) {
if (read(fd, &mcMsg, sizeof(mcMsg)) != sizeof(mcMsg))
CORRUPT();
if (mcMsg.invalid) {
--i;
continue;
}
msg = xmalloc(sizeof(msgT));
memset(msg, '\0', sizeof(*msg));
if (set->first) {
set->last->next = msg;
msg->prev = set->last;
set->last = msg;
} else
set->first = set->last = msg;
msg->msgId = mcMsg.msgId;
msg->str = xstrdup((char *) (data + mcMsg.msg.off));
}
free(data);
}
if (!mcSet.nextSet)
break;
if (lseek(fd, mcSet.nextSet, SEEK_SET) == -1)
CORRUPT();
}
#endif
}
/*
* Write message catalog.
*
* The message catalog is first converted from its internal to its
* external representation in a chunk of memory allocated for this
* purpose. Then the completed catalog is written. This approach
* avoids additional housekeeping variables and/or a lot of seeks
* that would otherwise be required.
*/
void
MCWriteCat(int fd)
{
int nsets; /* number of sets */
int nmsgs; /* number of msgs */
int string_size; /* total size of string pool */
int msgcat_size; /* total size of message catalog */
void *msgcat; /* message catalog data */
struct _nls_cat_hdr *cat_hdr;
struct _nls_set_hdr *set_hdr;
struct _nls_msg_hdr *msg_hdr;
char *strings;
struct _setT *set;
struct _msgT *msg;
int msg_index;
int msg_offset;
/* determine number of sets, number of messages, and size of the
* string pool */
nsets = 0;
nmsgs = 0;
string_size = 0;
for (set = sethead.lh_first; set != NULL;
set = set->entries.le_next) {
nsets++;
for (msg = set->msghead.lh_first; msg != NULL;
msg = msg->entries.le_next) {
nmsgs++;
string_size += strlen(msg->str) + 1;
}
}
#ifdef DEBUG
printf("number of sets: %d\n", nsets);
printf("number of msgs: %d\n", nmsgs);
printf("string pool size: %d\n", string_size);
#endif
/* determine size and then allocate buffer for constructing external
* message catalog representation */
msgcat_size = sizeof(struct _nls_cat_hdr)
+ (nsets * sizeof(struct _nls_set_hdr))
+ (nmsgs * sizeof(struct _nls_msg_hdr))
+ string_size;
msgcat = xmalloc(msgcat_size);
memset(msgcat, '\0', msgcat_size);
/* fill in msg catalog header */
cat_hdr = (struct _nls_cat_hdr *) msgcat;
cat_hdr->__magic = htonl(_NLS_MAGIC);
cat_hdr->__nsets = htonl(nsets);
cat_hdr->__mem = htonl(msgcat_size - sizeof(struct _nls_cat_hdr));
cat_hdr->__msg_hdr_offset =
htonl(nsets * sizeof(struct _nls_set_hdr));
cat_hdr->__msg_txt_offset =
htonl(nsets * sizeof(struct _nls_set_hdr) +
nmsgs * sizeof(struct _nls_msg_hdr));
/* compute offsets for set & msg header tables and string pool */
2005-02-27 19:37:51 +00:00
set_hdr = (struct _nls_set_hdr *)(void *)((char *)msgcat +
sizeof(struct _nls_cat_hdr));
2005-02-27 19:37:51 +00:00
msg_hdr = (struct _nls_msg_hdr *)(void *)((char *)msgcat +
sizeof(struct _nls_cat_hdr) +
nsets * sizeof(struct _nls_set_hdr));
strings = (char *) msgcat +
sizeof(struct _nls_cat_hdr) +
nsets * sizeof(struct _nls_set_hdr) +
nmsgs * sizeof(struct _nls_msg_hdr);
msg_index = 0;
msg_offset = 0;
for (set = sethead.lh_first; set != NULL;
set = set->entries.le_next) {
nmsgs = 0;
for (msg = set->msghead.lh_first; msg != NULL;
msg = msg->entries.le_next) {
int msg_len = strlen(msg->str) + 1;
msg_hdr->__msgno = htonl(msg->msgId);
msg_hdr->__msglen = htonl(msg_len);
msg_hdr->__offset = htonl(msg_offset);
memcpy(strings, msg->str, msg_len);
strings += msg_len;
msg_offset += msg_len;
nmsgs++;
msg_hdr++;
}
set_hdr->__setno = htonl(set->setId);
set_hdr->__nmsgs = htonl(nmsgs);
set_hdr->__index = htonl(msg_index);
msg_index += nmsgs;
set_hdr++;
}
/* write out catalog. XXX: should this be done in small chunks? */
write(fd, msgcat, msgcat_size);
}
void
MCAddSet(int setId)
{
struct _setT *p, *q;
if (setId <= 0) {
error("setId's must be greater than zero");
/* NOTREACHED */
}
if (setId > NL_SETMAX) {
error("setId exceeds limit");
/* NOTREACHED */
}
p = sethead.lh_first;
q = NULL;
for (; p != NULL && p->setId < setId; q = p, p = p->entries.le_next);
if (p && p->setId == setId) {
;
} else {
p = xmalloc(sizeof(struct _setT));
memset(p, '\0', sizeof(struct _setT));
LIST_INIT(&p->msghead);
p->setId = setId;
if (q == NULL) {
LIST_INSERT_HEAD(&sethead, p, entries);
} else {
LIST_INSERT_AFTER(q, p, entries);
}
}
curSet = p;
}
void
MCAddMsg(int msgId, const char *str)
{
struct _msgT *p, *q;
if (!curSet)
error("can't specify a message when no set exists");
if (msgId <= 0) {
error("msgId's must be greater than zero");
/* NOTREACHED */
}
if (msgId > NL_MSGMAX) {
error("msgID exceeds limit");
/* NOTREACHED */
}
p = curSet->msghead.lh_first;
q = NULL;
for (; p != NULL && p->msgId < msgId; q = p, p = p->entries.le_next);
if (p && p->msgId == msgId) {
free(p->str);
} else {
p = xmalloc(sizeof(struct _msgT));
memset(p, '\0', sizeof(struct _msgT));
if (q == NULL) {
LIST_INSERT_HEAD(&curSet->msghead, p, entries);
} else {
LIST_INSERT_AFTER(q, p, entries);
}
}
p->msgId = msgId;
p->str = xstrdup(str);
}
void
MCDelSet(int setId)
{
struct _setT *set;
struct _msgT *msg;
set = sethead.lh_first;
for (; set != NULL && set->setId < setId; set = set->entries.le_next);
if (set && set->setId == setId) {
msg = set->msghead.lh_first;
while (msg) {
free(msg->str);
LIST_REMOVE(msg, entries);
}
LIST_REMOVE(set, entries);
return;
}
warning(NULL, "specified set doesn't exist");
}
void
MCDelMsg(int msgId)
{
struct _msgT *msg;
if (!curSet)
error("you can't delete a message before defining the set");
msg = curSet->msghead.lh_first;
for (; msg != NULL && msg->msgId < msgId; msg = msg->entries.le_next);
if (msg && msg->msgId == msgId) {
free(msg->str);
LIST_REMOVE(msg, entries);
return;
}
warning(NULL, "specified msg doesn't exist");
}