/*
- * encoding.c: Convert "multibyte" strings (the locale-default encoding---
- * generally, UTF-8 or something like ISO-8859-1) to UTF-16LE strings, and vice
- * versa.
+ * encoding.c
*/
/*
* along with wimlib; if not, see http://www.gnu.org/licenses/.
*/
-#include "config.h"
-#include "wimlib_internal.h"
-#include <pthread.h>
-#include "list.h"
+#ifdef HAVE_CONFIG_H
+# include "config.h"
+#endif
+
+#include "wimlib.h"
+#include "wimlib/encoding.h"
+#include "wimlib/error.h"
+#include "wimlib/list.h"
+#include "wimlib/util.h"
+#include <errno.h>
#include <iconv.h>
+#include <pthread.h>
#include <stdlib.h>
+#include <string.h>
-bool wimlib_mbs_is_utf8 = false;
+bool wimlib_mbs_is_utf8 = !TCHAR_IS_UTF16LE;
+/* List of iconv_t conversion descriptors for a specific character conversion.
+ * The idea is that it is not thread-safe to have just one conversion
+ * descriptor, but it also is inefficient to open a new conversion descriptor to
+ * convert every string. Both these problems can be solved by maintaining a
+ * list of conversion descriptors; then, a thread can use an existing conversion
+ * descriptor if available. */
struct iconv_list_head {
const char *from_encoding;
const char *to_encoding;
.mutex = PTHREAD_MUTEX_INITIALIZER, \
}
-static ICONV_LIST(iconv_mbs_to_utf16le, "", "UTF-16LE");
-static ICONV_LIST(iconv_utf16le_to_mbs, "UTF-16LE", "");
-
static iconv_t *
get_iconv(struct iconv_list_head *head)
{
iconv_t cd;
+ iconv_t *cd_p;
struct iconv_node *i;
pthread_mutex_lock(&head->mutex);
if (list_empty(&head->list)) {
cd = iconv_open(head->to_encoding, head->from_encoding);
if (cd == (iconv_t)-1) {
- goto out_unlock;
+ ERROR_WITH_ERRNO("Failed to open iconv from %s to %s",
+ head->from_encoding, head->to_encoding);
+ cd_p = NULL;
} else {
i = MALLOC(sizeof(struct iconv_node));
- if (!i) {
+ if (i) {
+ i->head = head;
+ i->cd = cd;
+ cd_p = &i->cd;
+ } else {
iconv_close(cd);
- cd = (iconv_t)-1;
- goto out_unlock;
+ cd_p = NULL;
}
- i->head = head;
}
} else {
i = container_of(head->list.next, struct iconv_node, list);
list_del(head->list.next);
+ cd_p = &i->cd;
}
- cd = i->cd;
-out_unlock:
pthread_mutex_unlock(&head->mutex);
- return cd;
+ return cd_p;
}
static void
put_iconv(iconv_t *cd)
{
+ int errno_save = errno;
struct iconv_node *i = container_of(cd, struct iconv_node, cd);
struct iconv_list_head *head = i->head;
-
+
pthread_mutex_lock(&head->mutex);
list_add(&i->list, &head->list);
pthread_mutex_unlock(&head->mutex);
+ errno = errno_save;
}
-int
-mbs_to_utf16le_nbytes(const mbchar *mbs, size_t mbs_nbytes,
- size_t *utf16le_nbytes_ret)
-{
- iconv_t *cd = get_iconv(&iconv_mbs_to_utf16le);
- if (*cd == (iconv_t)-1)
- return WIMLIB_ERR_ICONV_NOT_AVAILABLE;
-
- /* Worst case length */
- utf16lechar buf[mbs_nbytes * 2];
- char *inbuf = (char*)mbs;
- char *outbuf = (char*)buf;
- size_t outbytesleft = sizeof(buf);
- size_t inbytesleft = mbs_nbytes;
- size_t len;
- int ret;
-
- len = iconv(*cd, &inbuf, &inbytesleft, &outbuf, &outbytesleft);
- if (len == (size_t)-1) {
- ret = WIMLIB_ERR_INVALID_MULTIBYTE_STRING;
- } else {
- *utf16le_nbytes_ret = sizeof(buf) - outbytesleft;
- ret = 0;
- }
- put_iconv(cd);
- return ret;
+/* Prevent printing an error message if a character conversion error occurs
+ * while printing an error message. (This variable is not per-thread but it
+ * doesn't matter too much since it's just the error messages.) */
+static bool error_message_being_printed = false;
+
+#define DEFINE_CHAR_CONVERSION_FUNCTIONS(varname1, longname1, chartype1,\
+ varname2, longname2, chartype2,\
+ earlyreturn_on_utf8_locale, \
+ earlyreturn_expr, \
+ worst_case_len_expr, \
+ err_return, \
+ err_msg, \
+ modifier) \
+static ICONV_LIST(iconv_##varname1##_to_##varname2, \
+ longname1, longname2); \
+ \
+modifier int \
+varname1##_to_##varname2##_nbytes(const chartype1 *in, size_t in_nbytes,\
+ size_t *out_nbytes_ret) \
+{ \
+ iconv_t *cd = get_iconv(&iconv_##varname1##_to_##varname2); \
+ if (cd == NULL) \
+ return WIMLIB_ERR_ICONV_NOT_AVAILABLE; \
+ \
+ /* Worst case length */ \
+ chartype2 buf[worst_case_len_expr]; \
+ char *inbuf = (char*)in; \
+ size_t inbytesleft = in_nbytes; \
+ char *outbuf = (char*)buf; \
+ size_t outbytesleft = sizeof(buf); \
+ size_t len; \
+ int ret; \
+ \
+ len = iconv(*cd, &inbuf, &inbytesleft, &outbuf, &outbytesleft); \
+ if (len == (size_t)-1) { \
+ if (!error_message_being_printed) { \
+ error_message_being_printed = true; \
+ err_msg; \
+ error_message_being_printed = false; \
+ } \
+ ret = err_return; \
+ } else { \
+ *out_nbytes_ret = sizeof(buf) - outbytesleft; \
+ ret = 0; \
+ } \
+ put_iconv(cd); \
+ return ret; \
+} \
+ \
+modifier int \
+varname1##_to_##varname2##_buf(const chartype1 *in, size_t in_nbytes, \
+ chartype2 *out) \
+{ \
+ iconv_t *cd = get_iconv(&iconv_##varname1##_to_##varname2); \
+ if (cd == NULL) \
+ return WIMLIB_ERR_ICONV_NOT_AVAILABLE; \
+ \
+ char *inbuf = (char*)in; \
+ size_t inbytesleft = in_nbytes; \
+ char *outbuf = (char*)out; \
+ const size_t LARGE_NUMBER = 1000000000; \
+ size_t outbytesleft = LARGE_NUMBER; \
+ size_t len; \
+ int ret; \
+ \
+ len = iconv(*cd, &inbuf, &inbytesleft, &outbuf, &outbytesleft); \
+ if (len == (size_t)-1) { \
+ if (!error_message_being_printed) { \
+ error_message_being_printed = true; \
+ err_msg; \
+ error_message_being_printed = false; \
+ } \
+ ret = err_return; \
+ } else { \
+ out[(LARGE_NUMBER-outbytesleft)/sizeof(chartype2)] = 0; \
+ ret = 0; \
+ } \
+ put_iconv(cd); \
+ return ret; \
+} \
+ \
+modifier int \
+varname1##_to_##varname2(const chartype1 *in, size_t in_nbytes, \
+ chartype2 **out_ret, \
+ size_t *out_nbytes_ret) \
+{ \
+ int ret; \
+ chartype2 *out; \
+ size_t out_nbytes; \
+ \
+ if (earlyreturn_on_utf8_locale && wimlib_mbs_is_utf8) { \
+ earlyreturn_expr; \
+ /* Out same as in */ \
+ out = MALLOC(in_nbytes + sizeof(chartype2)); \
+ if (!out) \
+ return WIMLIB_ERR_NOMEM; \
+ memcpy(out, in, in_nbytes); \
+ out[in_nbytes / sizeof(chartype2)] = 0; \
+ *out_ret = out; \
+ *out_nbytes_ret = in_nbytes; \
+ return 0; \
+ } \
+ \
+ ret = varname1##_to_##varname2##_nbytes(in, in_nbytes, \
+ &out_nbytes); \
+ if (ret) \
+ return ret; \
+ \
+ out = MALLOC(out_nbytes + sizeof(chartype2)); \
+ if (!out) \
+ return WIMLIB_ERR_NOMEM; \
+ \
+ ret = varname1##_to_##varname2##_buf(in, in_nbytes, out); \
+ if (ret) { \
+ FREE(out); \
+ } else { \
+ *out_ret = out; \
+ *out_nbytes_ret = out_nbytes; \
+ } \
+ return ret; \
}
+#if !TCHAR_IS_UTF16LE
+
+/* UNIX */
+
+DEFINE_CHAR_CONVERSION_FUNCTIONS(utf8, "UTF-8", tchar,
+ utf16le, "UTF-16LE", utf16lechar,
+ false,
+ ,
+ in_nbytes * 2,
+ WIMLIB_ERR_INVALID_UTF8_STRING,
+ ERROR_WITH_ERRNO("Failed to convert UTF-8 string "
+ "to UTF-16LE string!"),
+ static)
+
+DEFINE_CHAR_CONVERSION_FUNCTIONS(utf16le, "UTF-16LE", utf16lechar,
+ utf8, "UTF-8", tchar,
+ false,
+ ,
+ in_nbytes * 2,
+ WIMLIB_ERR_INVALID_UTF16_STRING,
+ ERROR_WITH_ERRNO("Failed to convert UTF-16LE string "
+ "to UTF-8 string!"),
+ static)
+
+DEFINE_CHAR_CONVERSION_FUNCTIONS(tstr, "", tchar,
+ utf16le, "UTF-16LE", utf16lechar,
+ true,
+ return utf8_to_utf16le(in, in_nbytes, out_ret, out_nbytes_ret),
+ in_nbytes * 2,
+ WIMLIB_ERR_INVALID_MULTIBYTE_STRING,
+ ERROR_WITH_ERRNO("Failed to convert multibyte "
+ "string \"%"TS"\" to UTF-16LE string!", in);
+ ERROR("If the data you provided was UTF-8, please make sure "
+ "the character encoding\n"
+ " of your current locale is UTF-8."),
+ )
+
+DEFINE_CHAR_CONVERSION_FUNCTIONS(utf16le, "UTF-16LE", utf16lechar,
+ tstr, "", tchar,
+ true,
+ return utf16le_to_utf8(in, in_nbytes, out_ret, out_nbytes_ret),
+ in_nbytes * 2,
+ WIMLIB_ERR_UNICODE_STRING_NOT_REPRESENTABLE,
+ ERROR("Failed to convert UTF-16LE string to "
+ "multibyte string!");
+ ERROR("This may be because the UTF-16LE string "
+ "could not be represented\n"
+ " in your locale's character encoding."),
+ )
+#endif
+
+/* tchar to UTF-8 and back */
+#if TCHAR_IS_UTF16LE
+
+/* Windows */
+DEFINE_CHAR_CONVERSION_FUNCTIONS(tstr, "UTF-16LE", tchar,
+ utf8, "UTF-8", char,
+ false,
+ ,
+ in_nbytes * 2,
+ WIMLIB_ERR_INVALID_UTF16_STRING,
+ ERROR_WITH_ERRNO("Failed to convert UTF-16LE "
+ "string \"%"TS"\" to UTF-8 string!", in),
+ static)
+
+DEFINE_CHAR_CONVERSION_FUNCTIONS(utf8, "UTF-8", char,
+ tstr, "UTF-16LE", tchar,
+ false,
+ ,
+ in_nbytes * 2,
+ WIMLIB_ERR_INVALID_UTF8_STRING,
+ ERROR_WITH_ERRNO("Failed to convert UTF-8 string "
+ "to UTF-16LE string!"),
+ static)
+#else
+
+/* UNIX */
+
+DEFINE_CHAR_CONVERSION_FUNCTIONS(tstr, "", tchar,
+ utf8, "UTF-8", char,
+ true,
+ ,
+ in_nbytes * 4,
+ WIMLIB_ERR_INVALID_MULTIBYTE_STRING,
+ ERROR_WITH_ERRNO("Failed to convert multibyte "
+ "string \"%"TS"\" to UTF-8 string!", in);
+ ERROR("If the data you provided was UTF-8, please make sure "
+ "the character\n"
+ " encoding of your current locale is UTF-8."),
+ static)
+
+DEFINE_CHAR_CONVERSION_FUNCTIONS(utf8, "UTF-8", char,
+ tstr, "", tchar,
+ true,
+ ,
+ in_nbytes * 4,
+ WIMLIB_ERR_UNICODE_STRING_NOT_REPRESENTABLE,
+ ERROR("Failed to convert UTF-8 string to "
+ "multibyte string!");
+ ERROR("This may be because the UTF-8 data "
+ "could not be represented\n"
+ " in your locale's character encoding."),
+ static)
+#endif
int
-utf16le_to_mbs_nbytes(const utf16lechar *utf16le_str, size_t utf16le_nbytes,
- size_t *mbs_nbytes_ret)
+tstr_to_utf8_simple(const tchar *tstr, char **out)
{
- iconv_t *cd = get_iconv(&iconv_utf16le_to_mbs);
- if (*cd == (iconv_t)-1)
- return WIMLIB_ERR_ICONV_NOT_AVAILABLE;
-
- /* Worst case length */
- mbchar buf[utf16le_nbytes / 2 * MB_CUR_MAX];
- char *inbuf = (char*)utf16le_str;
- char *outbuf = (char*)buf;
- size_t outbytesleft = sizeof(buf);
- size_t inbytesleft = utf16le_nbytes;
- size_t len;
- int ret;
-
- len = iconv(*cd, &inbuf, &inbytesleft, &outbuf, &outbytesleft);
- if (len == (size_t)-1) {
- ERROR("Could not convert \"%W\" to encoding of current locale",
- utf16le_str);
- /* EILSEQ is supposed to mean that the *input* is invalid, but
- * it's also returned if any input characters are not
- * representable in the output encoding. (The actual behavior
- * in this case is undefined for some reason...). Assume it's
- * the latter error case. */
- ret = WIMLIB_ERR_UNICODE_STRING_NOT_REPRESENTABLE;
- } else {
- *mbs_nbytes_ret = sizeof(buf) - outbytesleft;
- ret = 0;
- }
- put_iconv(cd);
- return ret;
+ size_t out_nbytes;
+ return tstr_to_utf8(tstr, tstrlen(tstr) * sizeof(tchar),
+ out, &out_nbytes);
}
int
-mbs_to_utf16le_buf(const mbchar *mbs, size_t mbs_nbytes,
- utf16lechar *utf16le_str)
+utf8_to_tstr_simple(const char *utf8str, tchar **out)
{
- iconv_t *cd = get_iconv(&iconv_mbs_to_utf16le);
- if (*cd == (iconv_t)-1)
- return WIMLIB_ERR_ICONV_NOT_AVAILABLE;
-
- char *inbuf = (char*)mbs;
- size_t inbytesleft = mbs_nbytes;
- char *outbuf = (char*)utf16le_str;
- size_t outbytesleft = SIZE_MAX;
- size_t len;
- int ret;
-
- len = iconv(*cd, &inbuf, &inbytesleft, &outbuf, &outbytesleft);
- if (len == (size_t)-1) {
- ret = WIMLIB_ERR_INVALID_MULTIBYTE_STRING;
- } else {
- ret = 0;
- }
- put_iconv(cd);
- return ret;
+ size_t out_nbytes;
+ return utf8_to_tstr(utf8str, strlen(utf8str), out, &out_nbytes);
}
-int
-utf16le_to_mbs_buf(const utf16lechar *utf16le_str, size_t utf16le_nbytes,
- mbchar *mbs)
-{
- int ret;
- iconv_t *cd = get_iconv(&iconv_utf16le_to_mbs);
- if (*cd == (iconv_t)-1)
- return WIMLIB_ERR_ICONV_NOT_AVAILABLE;
-
- char *inbuf = (char*)utf16le_str;
- size_t inbytesleft;
- char *outbuf = (char*)mbs;
- size_t outbytesleft = SIZE_MAX;
- size_t len;
-
- len = iconv(*cd, &inbuf, &inbytesleft, &outbuf, &outbytesleft);
- if (len == (size_t)-1) {
- ret = WIMLIB_ERR_INVALID_UTF16_STRING;
- } else {
- ret = 0;
- }
- mbs[SIZE_MAX - inbytesleft] = '\0';
- put_iconv(cd);
- return ret;
-}
-
-int
-mbs_to_utf16le(const mbchar *mbs, size_t mbs_nbytes,
- utf16lechar **utf16le_ret, size_t *utf16le_nbytes_ret)
-{
- int ret;
- utf16lechar *utf16le_str;
- size_t utf16le_nbytes;
-
- ret = mbs_to_utf16le_nbytes(mbs, mbs_nbytes,
- &utf16le_nbytes);
- if (ret)
- return ret;
-
- utf16le_str = MALLOC(utf16le_nbytes + 1);
- if (!utf16le_str)
- return WIMLIB_ERR_NOMEM;
-
- ret = mbs_to_utf16le_buf(mbs, mbs_nbytes, utf16le_str);
- if (ret) {
- FREE(utf16le_str);
- } else {
- *utf16le_ret = utf16le_str;
- *utf16le_nbytes_ret = utf16le_nbytes;
- }
- return ret;
-}
-
-
-int
-utf16le_to_mbs(const utf16lechar *utf16le_str, size_t utf16le_nbytes,
- mbchar **mbs_ret, size_t *mbs_nbytes_ret)
+static void
+iconv_cleanup(struct iconv_list_head *head)
{
- int ret;
- mbchar *mbs;
- size_t mbs_nbytes;
-
- ret = utf16le_to_mbs_nbytes(utf16le_str, utf16le_nbytes,
- &mbs_nbytes);
- if (ret)
- return ret;
-
- mbs = MALLOC(mbs_nbytes + 1);
- if (!mbs)
- return WIMLIB_ERR_NOMEM;
+ pthread_mutex_destroy(&head->mutex);
+ while (!list_empty(&head->list)) {
+ struct iconv_node *i;
- ret = utf16le_to_mbs_buf(utf16le_str, utf16le_nbytes, mbs);
- if (ret) {
- FREE(mbs);
- } else {
- *mbs_ret = mbs;
- *mbs_nbytes_ret = mbs_nbytes;
+ i = container_of(head->list.next, struct iconv_node, list);
+ list_del(&i->list);
+ iconv_close(i->cd);
+ FREE(i);
}
- return ret;
}
-bool
-utf8_str_contains_nonascii_chars(const utf8char *utf8_str)
+void
+iconv_global_cleanup(void)
{
- do {
- if ((unsigned char)*utf8_str > 127)
- return false;
- } while (*++utf8_str);
- return true;
+ iconv_cleanup(&iconv_utf8_to_tstr);
+ iconv_cleanup(&iconv_tstr_to_utf8);
+#if !TCHAR_IS_UTF16LE
+ iconv_cleanup(&iconv_utf16le_to_tstr);
+ iconv_cleanup(&iconv_tstr_to_utf16le);
+ iconv_cleanup(&iconv_utf16le_to_utf8);
+ iconv_cleanup(&iconv_utf8_to_utf16le);
+#endif
}