X-Git-Url: https://wimlib.net/git/?p=wimlib;a=blobdiff_plain;f=src%2Fencoding.c;h=48bad187ff8e21c7df4333e35854aabf75ef6b5d;hp=30a8889edd63189ccd0d45f5cfbca984cfbc0077;hb=de12c346dc64404821d52d545e2e1b3d44230f2a;hpb=7bde3fc590afbdef8f71cd7f8ccbd24172bffc63 diff --git a/src/encoding.c b/src/encoding.c index 30a8889e..48bad187 100644 --- a/src/encoding.c +++ b/src/encoding.c @@ -1,9 +1,11 @@ /* - * encoding.c: Convert UTF-8 to UTF-16LE strings and vice versa + * 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. Also, convert UTF-8 strings to multibyte strings. */ /* - * Copyright (C) 2012 Eric Biggers + * Copyright (C) 2012, 2013 Eric Biggers * * This file is part of wimlib, a library for working with WIM files. * @@ -21,337 +23,252 @@ * along with wimlib; if not, see http://www.gnu.org/licenses/. */ -#include "wimlib.h" -#include "util.h" -#include "endianness.h" +#include "wimlib_internal.h" #include - -#ifdef WITH_NTFS_3G -#include -#include -#else #include -#endif +#include +#include -/* - * NOTE: - * - * utf16_to_utf8_size() and utf8_to_utf16_size() were taken from - * libntfs-3g/unistr.c in the NTFS-3g sources. (Modified slightly to remove - * unneeded functionality.) - */ -#ifndef WITH_NTFS_3G -/* - * Return the amount of 8-bit elements in UTF-8 needed (without the terminating - * null) to store a given UTF-16LE string. - * - * Return -1 with errno set if string has invalid byte sequence or too long. - */ -static int utf16_to_utf8_size(const u16 *ins, const int ins_len) -{ - int i, ret = -1; - int count = 0; - bool surrog; +bool wimlib_mbs_is_utf8 = true; - surrog = false; - for (i = 0; i < ins_len && ins[i]; i++) { - unsigned short c = le16_to_cpu(ins[i]); - if (surrog) { - if ((c >= 0xdc00) && (c < 0xe000)) { - surrog = false; - count += 4; - } else - goto fail; - } else - if (c < 0x80) - count++; - else if (c < 0x800) - count += 2; - else if (c < 0xd800) - count += 3; - else if (c < 0xdc00) - surrog = true; -#if NOREVBOM - else if ((c >= 0xe000) && (c < 0xfffe)) -#else - else if (c >= 0xe000) -#endif - count += 3; - else - goto fail; - } - if (surrog) - goto fail; +/* 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; + struct list_head list; + pthread_mutex_t mutex; +}; - ret = count; -out: - return ret; -fail: - errno = EILSEQ; - goto out; -} +struct iconv_node { + iconv_t cd; + struct list_head list; + struct iconv_list_head *head; +}; -/* - * Return the amount of 16-bit elements in UTF-16LE needed - * (without the terminating null) to store given UTF-8 string. - * - * Return -1 with errno set if it's longer than PATH_MAX or string is invalid. - * - * Note: This does not check whether the input sequence is a valid utf8 string, - * and should be used only in context where such check is made! - */ -static int utf8_to_utf16_size(const char *s) -{ - unsigned int byte; - size_t count = 0; - while ((byte = *((const unsigned char *)s++))) { - count++; - if (byte >= 0xc0) { - if (byte >= 0xF5) { - errno = EILSEQ; - return -1; - } - if (!*s) - break; - if (byte >= 0xC0) - s++; - if (!*s) - break; - if (byte >= 0xE0) - s++; - if (!*s) - break; - if (byte >= 0xF0) { - s++; - count++; - } - } - } - return count; +#define ICONV_LIST(name, from, to) \ +struct iconv_list_head name = { \ + .from_encoding = from, \ + .to_encoding = to, \ + .list = LIST_HEAD_INIT(name.list), \ + .mutex = PTHREAD_MUTEX_INITIALIZER, \ } -#endif /* !WITH_NTFS_3G */ -#ifndef WITH_NTFS_3G -static iconv_t cd_utf8_to_utf16 = (iconv_t)(-1); -static iconv_t cd_utf16_to_utf8 = (iconv_t)(-1); - -int iconv_global_init() +static iconv_t * +get_iconv(struct iconv_list_head *head) { - if (cd_utf16_to_utf8 == (iconv_t)(-1)) { - cd_utf16_to_utf8 = iconv_open("UTF-8", "UTF-16LE"); - if (cd_utf16_to_utf8 == (iconv_t)-1) { - ERROR_WITH_ERRNO("Failed to get conversion descriptor " - "for converting UTF-16LE to UTF-8"); - if (errno == ENOMEM) - return WIMLIB_ERR_NOMEM; - else - return WIMLIB_ERR_ICONV_NOT_AVAILABLE; - } - } + iconv_t cd; + iconv_t *cd_p; + struct iconv_node *i; - if (cd_utf8_to_utf16 == (iconv_t)(-1)) { - cd_utf8_to_utf16 = iconv_open("UTF-16LE", "UTF-8"); - if (cd_utf8_to_utf16 == (iconv_t)-1) { - ERROR_WITH_ERRNO("Failed to get conversion descriptor " - "for converting UTF-8 to UTF-16LE"); - if (errno == ENOMEM) - return WIMLIB_ERR_NOMEM; - else - return WIMLIB_ERR_ICONV_NOT_AVAILABLE; + pthread_mutex_lock(&head->mutex); + if (list_empty(&head->list)) { + cd = iconv_open(head->to_encoding, head->from_encoding); + if (cd == (iconv_t)-1) { + 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) { + i->head = head; + i->cd = cd; + cd_p = &i->cd; + } else { + iconv_close(cd); + cd_p = NULL; + } } + } else { + i = container_of(head->list.next, struct iconv_node, list); + list_del(head->list.next); + cd_p = &i->cd; } - return 0; + pthread_mutex_unlock(&head->mutex); + return cd_p; } -void iconv_global_cleanup() +static void +put_iconv(iconv_t *cd) { - if (cd_utf8_to_utf16 != (iconv_t)(-1)) - iconv_close(cd_utf8_to_utf16); - if (cd_utf16_to_utf8 != (iconv_t)(-1)) - iconv_close(cd_utf16_to_utf8); + 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; } -#endif -/* Converts a string in the UTF-16LE encoding to a newly allocated string in the - * UTF-8 encoding. - * - * If available, do so by calling a similar function from libntfs-3g. - * Otherwise, use iconv() along with the helper function utf16_to_utf8_size(). - */ -int utf16_to_utf8(const char *utf16_str, size_t utf16_nbytes, - char **utf8_str_ret, size_t *utf8_nbytes_ret) -{ - int 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; - if (utf16_nbytes == 0) { - *utf8_str_ret = NULL; - *utf8_nbytes_ret = 0; - return 0; - } - - if (utf16_nbytes & 1) { - ERROR("UTF-16LE string is invalid (odd number of bytes)!"); - return WIMLIB_ERR_INVALID_UTF16_STRING; - } -#ifdef WITH_NTFS_3G - char *outs = NULL; - int outs_len = ntfs_ucstombs((const ntfschar*)utf16_str, - utf16_nbytes / 2, &outs, 0); - if (outs_len >= 0) { - *utf8_str_ret = outs; - *utf8_nbytes_ret = outs_len; - ret = 0; - } else { - if (errno == ENOMEM) - ret = WIMLIB_ERR_NOMEM; - else - ret = WIMLIB_ERR_INVALID_UTF16_STRING; - } -#else /* !WITH_NTFS_3G */ +#define DEFINE_CHAR_CONVERSION_FUNCTIONS(varname1, longname1, chartype1,\ + varname2, longname2, chartype2,\ + worst_case_len_expr, \ + err_return, \ + err_msg) \ +static ICONV_LIST(iconv_##varname1##_to_##varname2, \ + longname1, longname2); \ + \ +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; \ +} \ + \ +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; \ +} \ + \ +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; \ + \ + 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) { \ + int errno_save = errno; \ + FREE(out); \ + errno = errno_save; \ + } else { \ + *out_ret = out; \ + *out_nbytes_ret = out_nbytes; \ + } \ + return ret; \ +} - ret = iconv_global_init(); - if (ret != 0) - return ret; +DEFINE_CHAR_CONVERSION_FUNCTIONS(utf16le, "UTF-16LE", utf16lechar, + mbs, "", mbchar, + in_nbytes / 2 * MB_CUR_MAX, + WIMLIB_ERR_UNICODE_STRING_NOT_REPRESENTABLE, + ERROR("Failed to convert UTF-16LE string " + "to multibyte string!"); + ERROR("This may be because the UTF-16LE data " + "could not be represented in your " + "locale's character encoding.")) - ret = utf16_to_utf8_size((const u16*)utf16_str, utf16_nbytes / 2); - if (ret >= 0) { - size_t utf8_expected_nbytes; - char *utf8_str; - size_t utf8_bytes_left; - size_t utf16_bytes_left; - size_t num_chars_converted; - char *utf8_str_save; - const char *utf16_str_save; +DEFINE_CHAR_CONVERSION_FUNCTIONS(mbs, "", mbchar, + utf16le, "UTF-16LE", utf16lechar, + in_nbytes * 2, + WIMLIB_ERR_INVALID_MULTIBYTE_STRING, + ERROR_WITH_ERRNO("Failed to convert multibyte " + "string \"%s\" to UTF-16LE string!", in); + ERROR("If the data you provided was UTF-8, please make sure " + "the character encoding of your current locale is UTF-8.")) - utf8_expected_nbytes = ret; - utf8_str = MALLOC(utf8_expected_nbytes + 1); - if (utf8_str) { - utf8_bytes_left = utf8_expected_nbytes; - utf16_bytes_left = utf16_nbytes; - utf8_str_save = utf8_str; - utf16_str_save = utf16_str; - num_chars_converted = iconv(cd_utf16_to_utf8, - (char**)&utf16_str, - &utf16_bytes_left, - &utf8_str, - &utf8_bytes_left); - utf8_str = utf8_str_save; - utf16_str = utf16_str_save; - if (utf16_bytes_left == 0 && - utf8_bytes_left == 0 && - num_chars_converted != (size_t)(-1)) - { - utf8_str[utf8_expected_nbytes] = '\0'; - *utf8_str_ret = utf8_str; - *utf8_nbytes_ret = utf8_expected_nbytes; - ret = 0; - } else { - FREE(utf8_str); - ret = WIMLIB_ERR_INVALID_UTF16_STRING; - } - } else - ret = WIMLIB_ERR_NOMEM; - } else - ret = WIMLIB_ERR_INVALID_UTF16_STRING; -#endif /* WITH_NTFS_3G */ +DEFINE_CHAR_CONVERSION_FUNCTIONS(utf8, "UTF-8", utf8char, + mbs, "", mbchar, + in_nbytes, + 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 " + "in your locale's character encoding.")) -#ifdef ENABLE_ERROR_MESSAGES - if (ret != 0) { - ERROR_WITH_ERRNO("Error converting UTF-16LE string to UTF-8"); - ERROR("The failing string was:"); - print_string(utf16_str, utf16_nbytes); - putchar('\n'); +static void +iconv_cleanup(struct iconv_list_head *head) +{ + pthread_mutex_destroy(&head->mutex); + while (!list_empty(&head->list)) { + struct iconv_node *i; + + i = container_of(head->list.next, struct iconv_node, list); + list_del(&i->list); + iconv_close(i->cd); + FREE(i); } -#endif /* ENABLE_ERROR_MESSAGES */ - return ret; } - -/* Converts a string in the UTF-8 encoding to a newly allocated string in the - * UTF-16 encoding. - * - * If available, do so by calling a similar function from libntfs-3g. - * Otherwise, use iconv() along with the helper function utf8_to_utf16_size(). - */ -int utf8_to_utf16(const char *utf8_str, size_t utf8_nbytes, - char **utf16_str_ret, size_t *utf16_nbytes_ret) +void +iconv_global_cleanup() { - int ret; - if (utf8_nbytes == 0) { - *utf16_str_ret = NULL; - *utf16_nbytes_ret = 0; - return 0; - } -#ifdef WITH_NTFS_3G - char *outs = NULL; - int outs_nchars = ntfs_mbstoucs(utf8_str, (ntfschar**)&outs); - if (outs_nchars >= 0) { - *utf16_str_ret = outs; - *utf16_nbytes_ret = (size_t)outs_nchars * 2; - ret = 0; - } else { - if (errno == ENOMEM) - ret = WIMLIB_ERR_NOMEM; - else - ret = WIMLIB_ERR_INVALID_UTF8_STRING; - } -#else /* !WITH_NTFS_3G */ - - ret = iconv_global_init(); - if (ret != 0) - return ret; - ret = utf8_to_utf16_size(utf8_str); - if (ret >= 0) { - size_t utf16_expected_nbytes; - char *utf16_str; - size_t utf16_bytes_left; - size_t utf8_bytes_left; - size_t num_chars_converted; - const char *utf8_str_save; - char *utf16_str_save; - - utf16_expected_nbytes = (size_t)ret * 2; - utf16_str = MALLOC(utf16_expected_nbytes + 2); - if (utf16_str) { - utf16_bytes_left = utf16_expected_nbytes; - utf8_bytes_left = utf8_nbytes; - utf8_str_save = utf8_str; - utf16_str_save = utf16_str; - num_chars_converted = iconv(cd_utf8_to_utf16, - (char**)&utf8_str, - &utf8_bytes_left, - &utf16_str, - &utf16_bytes_left); - utf8_str = utf8_str_save; - utf16_str = utf16_str_save; - if (utf16_bytes_left == 0 && - utf8_bytes_left == 0 && - num_chars_converted != (size_t)(-1)) - { - utf16_str[utf16_expected_nbytes] = '\0'; - utf16_str[utf16_expected_nbytes + 1] = '\0'; - *utf16_str_ret = utf16_str; - *utf16_nbytes_ret = utf16_expected_nbytes; - ret = 0; - } else { - FREE(utf16_str); - ret = WIMLIB_ERR_INVALID_UTF8_STRING; - } - } else - ret = WIMLIB_ERR_NOMEM; - } else - ret = WIMLIB_ERR_INVALID_UTF8_STRING; -#endif /* WITH_NTFS_3G */ + iconv_cleanup(&iconv_utf16le_to_mbs); + iconv_cleanup(&iconv_mbs_to_utf16le); + iconv_cleanup(&iconv_utf8_to_mbs); +} -#ifdef ENABLE_ERROR_MESSAGES - if (ret != 0) { - ERROR_WITH_ERRNO("Error converting UTF-8 string to UTF-16LE"); - ERROR("The failing string was:"); - print_string(utf8_str, utf8_nbytes); - putchar('\n'); - ERROR("Length: %zu bytes", utf8_nbytes); - } -#endif /* ENABLE_ERROR_MESSAGES */ - return ret; +bool +utf8_str_contains_nonascii_chars(const utf8char *utf8_str) +{ + do { + if ((unsigned char)*utf8_str > 127) + return true; + } while (*++utf8_str); + return false; }