X-Git-Url: https://wimlib.net/git/?p=wimlib;a=blobdiff_plain;f=src%2Fencoding.c;h=9337c9a16e6a277b2f4880058103b4cdd1377eee;hp=4fd8712b62a94725f41117ede8a6b8b3961b3518;hb=4a20aae0dd8469a352517a0b107416ffa99ccc55;hpb=542225070ab6583b6f5915172425cac6e0326d77 diff --git a/src/encoding.c b/src/encoding.c index 4fd8712b..9337c9a1 100644 --- a/src/encoding.c +++ b/src/encoding.c @@ -1,9 +1,7 @@ /* - * encoding.c - */ - -/* - * Copyright (C) 2012, 2013 Eric Biggers + * encoding.c - UTF-8 and UTF-16LE codecs and utility functions + * + * Copyright (C) 2012-2016 Eric Biggers * * This file 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 @@ -24,386 +22,265 @@ #endif #include -#include -#include #include -#include "wimlib.h" -#include "wimlib/alloca.h" -#include "wimlib/assert.h" #include "wimlib/encoding.h" #include "wimlib/endianness.h" #include "wimlib/error.h" -#include "wimlib/list.h" +#include "wimlib/unaligned.h" #include "wimlib/util.h" +/* + * Allow unpaired surrogates, such as might exist in Windows-style filenames --- + * which are normally valid UTF-16LE, but are actually treated as opaque + * sequences of 16-bit WCHARs by Windows. When decoding "UTF-16LE", unpaired + * surrogates will be decoded as their surrogate codepoints; and when encoding + * to and from "UTF-8", the encoding will actually be WTF-8 ("Wobbly + * Transformation Format - 8-bit"), a superset of UTF-8 which permits the + * surrogate codepoints. + * + * In combination with also allowing the "non-character" codepoints U+FFFE and + * U+FFFF, the result is that every Windows-style filename can be translated to + * a UNIX-style filename. + * + * Unfortunately, the converse is not true: not every UNIX filename can be + * translated to a Windows filename. Only UNIX filenames that are valid "WTF-8" + * can be translated. I considered ways to define a bijective mapping, but + * there did not seem to be a straightforward way. The "UTF-8b" scheme, for + * example, would map each invalid byte 'b' to a surrogate "escape code" 'U+DC00 + * + b'. The problem with this was that surrogate escape codes can be combined + * to create a valid UTF-8 sequence, thus breaking the bijection by mapping + * multiple Windows filenames to a single UNIX filename. + */ +#define ALLOW_UNPAIRED_SURROGATES 1 -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; - struct list_head list; - pthread_mutex_t mutex; -}; - -struct iconv_node { - iconv_t cd; - struct list_head list; - struct iconv_list_head *head; -}; - -#define ICONV_LIST(name, from, to) \ -struct iconv_list_head name = { \ - .from_encoding = from, \ - .to_encoding = to, \ -} +#define INVALID_CODEPOINT 0xFFFFFFFF +#define VALIDATE(expr) if (validate && unlikely(!(expr))) goto invalid +#define IS_SURROGATE(c) ((c) >= 0xD800 && (c) < 0xE000) +#define IS_HIGH_SURROGATE(c) ((c) >= 0xD800 && (c) < 0xDC00) +#define IS_LOW_SURROGATE(c) ((c) >= 0xDC00 && (c) < 0xE000) +#define IS_UTF8_TAIL(c) (((c) & 0xC0) == 0x80) -static iconv_t * -get_iconv(struct iconv_list_head *head) +/* + * Decode the next Unicode codepoint from the string at @in, which has + * @remaining >= 1 bytes remaining. Return the number of bytes consumed and + * write the decoded codepoint to *c_ret. + * + * If the input might not be a valid string in the source encoding, then + * @validate must be specified as %true, and then on invalid input the function + * consumes at least one byte and sets *c_ret to INVALID_CODEPOINT. If the + * input is guaranteed to be valid, then @validate may be specified as %false. + */ +typedef unsigned (*decode_codepoint_fn)(const u8 *in, size_t remaining, + bool validate, u32 *c_ret); + +/* Encode the Unicode codepoint @c and return the number of bytes used. */ +typedef unsigned (*encode_codepoint_fn)(u32 c, u8 *out); + +static forceinline unsigned +utf8_decode_codepoint(const u8 *in, size_t remaining, bool validate, u32 *c_ret) { - 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) { - 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; + if (likely(in[0] < 0x80)) { /* U+0...U+7F */ + *c_ret = in[0]; + return 1; } - pthread_mutex_unlock(&head->mutex); - 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; -} + if (in[0] < 0xE0) { /* U+80...U+7FF */ + VALIDATE(in[0] >= 0xC2 && remaining >= 2 && + IS_UTF8_TAIL(in[1])); + *c_ret = ((u32)(in[0] & 0x1F) << 6) | + ((u32)(in[1] & 0x3F) << 0); + return 2; + } + + if (in[0] < 0xF0) { /* U+800...U+FFFF, possibly excluding surrogates */ + VALIDATE(remaining >= 3 && + IS_UTF8_TAIL(in[1]) && + IS_UTF8_TAIL(in[2])); + *c_ret = ((u32)(in[0] & 0x0F) << 12) | + ((u32)(in[1] & 0x3F) << 6) | + ((u32)(in[2] & 0x3F) << 0); + VALIDATE(*c_ret >= 0x800); + #if !ALLOW_UNPAIRED_SURROGATES + VALIDATE(!IS_SURROGATE(*c_ret)); + #endif + return 3; + } -#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; \ - \ - chartype2 *buf; \ - size_t bufsize; \ - bool buf_onheap; \ - bufsize = (worst_case_len_expr) * sizeof(chartype2); \ - /* Worst case length */ \ - if (bufsize <= STACK_MAX) { \ - buf = alloca(bufsize); \ - buf_onheap = false; \ - } else { \ - buf = MALLOC(bufsize); \ - if (!buf) \ - return WIMLIB_ERR_NOMEM; \ - buf_onheap = true; \ - } \ - \ - char *inbuf = (char*)in; \ - size_t inbytesleft = in_nbytes; \ - char *outbuf = (char*)buf; \ - size_t outbytesleft = bufsize; \ - size_t len; \ - int ret; \ - \ - len = iconv(*cd, &inbuf, &inbytesleft, &outbuf, &outbytesleft); \ - if (len == (size_t)-1) { \ - err_msg; \ - ret = err_return; \ - } else { \ - *out_nbytes_ret = bufsize - outbytesleft; \ - ret = 0; \ - } \ - put_iconv(cd); \ - if (buf_onheap) \ - FREE(buf); \ - 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) { \ - err_msg; \ - 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; \ + /* U+10000...U+10FFFF */ + VALIDATE(in[0] < 0xF8 && remaining >= 4 && + IS_UTF8_TAIL(in[1]) && + IS_UTF8_TAIL(in[2]) && + IS_UTF8_TAIL(in[3])); + *c_ret = ((u32)(in[0] & 0x07) << 18) | + ((u32)(in[1] & 0x3F) << 12) | + ((u32)(in[2] & 0x3F) << 6) | + ((u32)(in[3] & 0x3F) << 0); + VALIDATE(*c_ret >= 0x10000 && *c_ret <= 0x10FFFF); + return 4; + +invalid: + *c_ret = INVALID_CODEPOINT; + return 1; } -#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 +static forceinline unsigned +utf8_encode_codepoint(u32 c, u8 *out) +{ + if (likely(c < 0x80)) { + out[0] = c; + return 1; + } -/* 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), - ) - -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!"), - ) -#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."), - ) - -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."), - ) -#endif + if (c < 0x800) { + out[0] = 0xC0 | (c >> 6); + out[1] = 0x80 | (c & 0x3F); + return 2; + } -int -tstr_to_utf8_simple(const tchar *tstr, char **out) -{ - size_t out_nbytes; - return tstr_to_utf8(tstr, tstrlen(tstr) * sizeof(tchar), - out, &out_nbytes); + if (c < 0x10000) { + out[0] = 0xE0 | (c >> 12); + out[1] = 0x80 | ((c >> 6) & 0x3F); + out[2] = 0x80 | (c & 0x3F); + return 3; + } + + out[0] = 0xF0 | (c >> 18); + out[1] = 0x80 | ((c >> 12) & 0x3F); + out[2] = 0x80 | ((c >> 6) & 0x3F); + out[3] = 0x80 | (c & 0x3F); + return 4; } -int -utf8_to_tstr_simple(const char *utf8str, tchar **out) +static forceinline unsigned +utf16le_decode_codepoint(const u8 *in, size_t remaining, bool validate, + u32 *c_ret) { - size_t out_nbytes; - return utf8_to_tstr(utf8str, strlen(utf8str), out, &out_nbytes); + u32 h, l; + + VALIDATE(remaining >= 2); + h = get_unaligned_le16(in); + if (unlikely(IS_SURROGATE(h))) { + /* Surrogate pairs are U+10000...U+10FFFF. + * Unpaired surrogates are U+D800...U+DFFF. */ + #if ALLOW_UNPAIRED_SURROGATES + if (unlikely(!IS_HIGH_SURROGATE(h) || remaining < 4)) + goto unpaired; + l = get_unaligned_le16(in + 2); + if (unlikely(!IS_LOW_SURROGATE(l))) + goto unpaired; + #else + VALIDATE(IS_HIGH_SURROGATE(h) && remaining >= 4); + l = get_unaligned_le16(in + 2); + VALIDATE(IS_LOW_SURROGATE(l)); + #endif + *c_ret = 0x10000 + ((h - 0xD800) << 10) + (l - 0xDC00); + return 4; + } +#if ALLOW_UNPAIRED_SURROGATES +unpaired: +#endif + *c_ret = h; + return 2; + +invalid: + *c_ret = INVALID_CODEPOINT; + return min(remaining, 2); } -static void -iconv_init(struct iconv_list_head *head) +static forceinline unsigned +utf16le_encode_codepoint(u32 c, u8 *out) { - pthread_mutex_init(&head->mutex, NULL); - INIT_LIST_HEAD(&head->list); + if (likely(c < 0x10000)) { + put_unaligned_le16(c, out); + return 2; + } + c -= 0x10000; + put_unaligned_le16(0xD800 + (c >> 10), out); + put_unaligned_le16(0xDC00 + (c & 0x3FF), out + 2); + return 4; } -static void -iconv_cleanup(struct iconv_list_head *head) +/* + * Convert the string @in of size @in_nbytes from the encoding given by the + * @decode_codepoint function to the encoding given by the @encode_codepoint + * function. @in does not need to be null-terminated, but a null terminator + * will be added to the output string. + * + * On success, write the allocated output string to @out_ret (must not be NULL) + * and its size excluding the null terminator to @out_nbytes_ret (may be NULL). + * + * If the input string is malformed, return @ilseq_err with errno set to EILSEQ. + * If out of memory, return WIMLIB_ERR_NOMEM with errno set to ENOMEM. + */ +static forceinline int +convert_string(const u8 * const in, const size_t in_nbytes, + u8 **out_ret, size_t *out_nbytes_ret, + int ilseq_err, + decode_codepoint_fn decode_codepoint, + encode_codepoint_fn encode_codepoint) { - 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); + const u8 * const in_end = in + in_nbytes; + const u8 *p_in; + u8 *p_out; + size_t out_nbytes = 0; + u8 *out; + u8 tmp[8]; /* assuming no codepoint requires > 8 bytes to encode */ + u32 c; + + /* Validate the input string and compute the output size. */ + for (p_in = in; p_in != in_end; ) { + p_in += (*decode_codepoint)(p_in, in_end - p_in, true, &c); + if (unlikely(c == INVALID_CODEPOINT)) { + errno = EILSEQ; + return ilseq_err; + } + out_nbytes += (*encode_codepoint)(c, tmp); } + + /* Allocate the output string, including space for a null terminator. */ + out = MALLOC(out_nbytes + (*encode_codepoint)(0, tmp)); + if (unlikely(!out)) + return WIMLIB_ERR_NOMEM; + + /* Do the conversion. */ + for (p_in = in, p_out = out; p_in != in_end; ) { + p_in += (*decode_codepoint)(p_in, in_end - p_in, false, &c); + p_out += (*encode_codepoint)(c, p_out); + } + + /* Add a null terminator. */ + (*encode_codepoint)(0, p_out); + + /* Return the output string and its size (by reference). */ + *out_ret = out; + if (out_nbytes_ret) + *out_nbytes_ret = out_nbytes; + return 0; } -void -iconv_global_init(void) +int +utf8_to_utf16le(const char *in, size_t in_nbytes, + utf16lechar **out_ret, size_t *out_nbytes_ret) { - iconv_init(&iconv_utf8_to_tstr); - iconv_init(&iconv_tstr_to_utf8); -#if !TCHAR_IS_UTF16LE - iconv_init(&iconv_utf16le_to_tstr); - iconv_init(&iconv_tstr_to_utf16le); - iconv_init(&iconv_utf16le_to_utf8); - iconv_init(&iconv_utf8_to_utf16le); -#endif + return convert_string((const u8 *)in, in_nbytes, + (u8 **)out_ret, out_nbytes_ret, + WIMLIB_ERR_INVALID_UTF8_STRING, + utf8_decode_codepoint, utf16le_encode_codepoint); } -void -iconv_global_cleanup(void) +int +utf16le_to_utf8(const utf16lechar *in, size_t in_nbytes, + char **out_ret, size_t *out_nbytes_ret) { - 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 + return convert_string((const u8 *)in, in_nbytes, + (u8 **)out_ret, out_nbytes_ret, + WIMLIB_ERR_INVALID_UTF16_STRING, + utf16le_decode_codepoint, utf8_encode_codepoint); } -/* A table that maps from UCS-2 characters to their upper case equivalents. +/* + * A table that maps from UCS-2 characters to their upper case equivalents. * Index and array values are both CPU endian. * Note: this is only an *approximation* of real UTF-16 case folding. */ @@ -484,38 +361,17 @@ init_upcase(void) /* Delta filter */ for (u32 i = 0; i < ARRAY_LEN(upcase); i++) upcase[i] += i; - -#if 0 - /* Sanity checks */ - wimlib_assert(upcase['a'] == 'A'); - wimlib_assert(upcase['A'] == 'A'); - wimlib_assert(upcase['z'] == 'Z'); - wimlib_assert(upcase['Z'] == 'Z'); - wimlib_assert(upcase['1'] == '1'); - wimlib_assert(upcase[0x00e9] == 0x00c9); /* Latin letter e, with acute accent */ - wimlib_assert(upcase[0x00c9] == 0x00c9); - wimlib_assert(upcase[0x03c1] == 0x03a1); /* Greek letter rho */ - wimlib_assert(upcase[0x03a1] == 0x03a1); - wimlib_assert(upcase[0x0436] == 0x0416); /* Cyrillic letter zhe */ - wimlib_assert(upcase[0x0416] == 0x0416); - wimlib_assert(upcase[0x0567] == 0x0537); /* Armenian letter eh */ - wimlib_assert(upcase[0x0537] == 0x0537); - wimlib_assert(upcase[0x24d0] == 0x24b6); /* Circled Latin letter A - (is that a real character???) */ - wimlib_assert(upcase[0x24b6] == 0x24b6); - wimlib_assert(upcase[0x2603] == 0x2603); /* Note to self: Upper case - snowman symbol does not - exist. */ -#endif } -/* Compare UTF-16LE strings case-sensitively (%ignore_case == false) or +/* + * Compare UTF-16LE strings case-sensitively (%ignore_case == false) or * case-insensitively (%ignore_case == true). * * This is implemented using the default upper-case table used by NTFS. It does * not handle all possible cases allowed by UTF-16LE. For example, different * normalizations of the same sequence of "characters" are not considered equal. - * It hopefully does the right thing most of the time though. */ + * It hopefully does the right thing most of the time though. + */ int cmp_utf16le_strings(const utf16lechar *s1, size_t n1, const utf16lechar *s2, size_t n2, @@ -567,32 +423,29 @@ cmp_utf16le_strings_z(const utf16lechar *s1, const utf16lechar *s2, } } -/* Duplicate a UTF-16LE string. The input string might not be null terminated - * and might be misaligned, but the returned string is guaranteed to be null +/* Duplicate a UTF-16 string. The input string might not be null terminated and + * might be misaligned, but the returned string is guaranteed to be null * terminated and properly aligned. */ utf16lechar * -utf16le_dupz(const void *ustr, size_t usize) +utf16le_dupz(const void *s, size_t size) { - utf16lechar *dup = MALLOC(usize + sizeof(utf16lechar)); + utf16lechar *dup = MALLOC(size + sizeof(utf16lechar)); if (dup) { - memcpy(dup, ustr, usize); - dup[usize / sizeof(utf16lechar)] = 0; + memcpy(dup, s, size); + dup[size / sizeof(utf16lechar)] = 0; } return dup; } -/* Duplicate a null-terminated UTF-16LE string. */ +/* Duplicate a null-terminated UTF-16 string. */ utf16lechar * -utf16le_dup(const utf16lechar *ustr) +utf16le_dup(const utf16lechar *s) { - const utf16lechar *p = ustr; - while (*p++) - ; - return memdup(ustr, (const u8 *)p - (const u8 *)ustr); + return memdup(s, utf16le_len_bytes(s) + sizeof(utf16lechar)); } -/* Return the length, in bytes, of a UTF-null terminated UTF-16 string, - * excluding the null terminator. */ +/* Return the length, in bytes, of a null terminated UTF-16 string, excluding + * the null terminator. */ size_t utf16le_len_bytes(const utf16lechar *s) { @@ -602,7 +455,7 @@ utf16le_len_bytes(const utf16lechar *s) return (p - s) * sizeof(utf16lechar); } -/* Return the length, in UTF-16 coding units, of a UTF-null terminated UTF-16 +/* Return the length, in UTF-16 coding units, of a null terminated UTF-16 * string, excluding the null terminator. */ size_t utf16le_len_chars(const utf16lechar *s)