X-Git-Url: https://wimlib.net/git/?p=wimlib;a=blobdiff_plain;f=src%2Fencoding.c;h=9337c9a16e6a277b2f4880058103b4cdd1377eee;hp=8c5bd2af6acf4ef9e2d9b2a678084d217d92c883;hb=4a20aae0dd8469a352517a0b107416ffa99ccc55;hpb=dec3bbfa2c3e9fc8e4b1adbaa0323337a75aaadb diff --git a/src/encoding.c b/src/encoding.c index 8c5bd2af..9337c9a1 100644 --- a/src/encoding.c +++ b/src/encoding.c @@ -1,517 +1,377 @@ /* - * encoding.c - */ - -/* - * Copyright (C) 2012, 2013 Eric Biggers + * encoding.c - UTF-8 and UTF-16LE codecs and utility functions * - * This file is part of wimlib, a library for working with WIM files. + * Copyright (C) 2012-2016 Eric Biggers * - * wimlib is free software; you can redistribute it and/or modify it under the - * terms of the GNU General Public License as published by the Free - * Software Foundation; either version 3 of the License, or (at your option) - * any later version. + * 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 + * Software Foundation; either version 3 of the License, or (at your option) any + * later version. * - * wimlib 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 General Public License for more + * This file 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 General Public License - * along with wimlib; if not, see http://www.gnu.org/licenses/. + * You should have received a copy of the GNU Lesser General Public License + * along with this file; if not, see http://www.gnu.org/licenses/. */ #ifdef HAVE_CONFIG_H # include "config.h" #endif -#include "wimlib.h" -#include "wimlib/assert.h" +#include +#include + #include "wimlib/encoding.h" #include "wimlib/endianness.h" #include "wimlib/error.h" -#include "wimlib/list.h" +#include "wimlib/unaligned.h" #include "wimlib/util.h" -#include -#include -#include -#include -#include +/* + * 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, \ - .list = LIST_HEAD_INIT(name.list), \ - .mutex = PTHREAD_MUTEX_INITIALIZER, \ -} +#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) + +/* + * 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 iconv_t * -get_iconv(struct iconv_list_head *head) +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; + } + + 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; } - pthread_mutex_unlock(&head->mutex); - return cd_p; + + 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; + } + + /* 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; } -static void -put_iconv(iconv_t *cd) +static forceinline unsigned +utf8_encode_codepoint(u32 c, u8 *out) { - 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 (likely(c < 0x80)) { + out[0] = c; + return 1; + } -/* 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 (c < 0x800) { + out[0] = 0xC0 | (c >> 6); + out[1] = 0x80 | (c & 0x3F); + return 2; + } + + 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; } -#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."), - ) +static forceinline unsigned +utf16le_decode_codepoint(const u8 *in, size_t remaining, bool validate, + u32 *c_ret) +{ + 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; -/* 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 +invalid: + *c_ret = INVALID_CODEPOINT; + return min(remaining, 2); +} -int -tstr_to_utf8_simple(const tchar *tstr, char **out) +static forceinline unsigned +utf16le_encode_codepoint(u32 c, u8 *out) { - size_t out_nbytes; - return tstr_to_utf8(tstr, tstrlen(tstr) * sizeof(tchar), - out, &out_nbytes); + 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; } -int -utf8_to_tstr_simple(const char *utf8str, tchar **out) +/* + * 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) { - size_t out_nbytes; - return utf8_to_tstr(utf8str, strlen(utf8str), out, &out_nbytes); + 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; } -static void -iconv_cleanup(struct iconv_list_head *head) +int +utf8_to_utf16le(const char *in, size_t in_nbytes, + utf16lechar **out_ret, size_t *out_nbytes_ret) { - 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); - } + 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. */ -static u16 upcase[65536]; +u16 upcase[65536]; -/* - * Initialize the 'upcase' table. - * - * Code modified from NTFS-3g, original copyright notices below: - * - * Copyright (c) 2000-2004 Anton Altaparmakov - * Copyright (c) 2002-2009 Szabolcs Szakacsits - * Copyright (c) 2008-2011 Jean-Pierre Andre - * Copyright (c) 2008 Bernhard Kaindl - * - * License GPLv2 or later. - * - * The expanded table itself is the same as that used by Windows Vista for - * NTFS case insensitivity. - */ void init_upcase(void) { - /* - * This is the table as defined by Vista - */ - /* - * "Start" is inclusive and "End" is exclusive, every value has the - * value of "Add" added to it. - */ - static const u16 uc_run_table[][3] = { /* Start, End, Add */ - {0x0061, 0x007b, -32}, {0x00e0, 0x00f7, -32}, {0x00f8, 0x00ff, -32}, - {0x0256, 0x0258, -205}, {0x028a, 0x028c, -217}, {0x037b, 0x037e, 130}, - {0x03ac, 0x03ad, -38}, {0x03ad, 0x03b0, -37}, {0x03b1, 0x03c2, -32}, - {0x03c2, 0x03c3, -31}, {0x03c3, 0x03cc, -32}, {0x03cc, 0x03cd, -64}, - {0x03cd, 0x03cf, -63}, {0x0430, 0x0450, -32}, {0x0450, 0x0460, -80}, - {0x0561, 0x0587, -48}, {0x1f00, 0x1f08, 8}, {0x1f10, 0x1f16, 8}, - {0x1f20, 0x1f28, 8}, {0x1f30, 0x1f38, 8}, {0x1f40, 0x1f46, 8}, - {0x1f51, 0x1f52, 8}, {0x1f53, 0x1f54, 8}, {0x1f55, 0x1f56, 8}, - {0x1f57, 0x1f58, 8}, {0x1f60, 0x1f68, 8}, {0x1f70, 0x1f72, 74}, - {0x1f72, 0x1f76, 86}, {0x1f76, 0x1f78, 100}, {0x1f78, 0x1f7a, 128}, - {0x1f7a, 0x1f7c, 112}, {0x1f7c, 0x1f7e, 126}, {0x1f80, 0x1f88, 8}, - {0x1f90, 0x1f98, 8}, {0x1fa0, 0x1fa8, 8}, {0x1fb0, 0x1fb2, 8}, - {0x1fb3, 0x1fb4, 9}, {0x1fcc, 0x1fcd, -9}, {0x1fd0, 0x1fd2, 8}, - {0x1fe0, 0x1fe2, 8}, {0x1fe5, 0x1fe6, 7}, {0x1ffc, 0x1ffd, -9}, - {0x2170, 0x2180, -16}, {0x24d0, 0x24ea, -26}, {0x2c30, 0x2c5f, -48}, - {0x2d00, 0x2d26, -7264}, {0xff41, 0xff5b, -32}, - }; - /* - * "Start" is exclusive and "End" is inclusive, every second value is - * decremented by one. - */ - static const u16 uc_dup_table[][2] = { /* Start, End */ - {0x0100, 0x012f}, {0x0132, 0x0137}, {0x0139, 0x0149}, {0x014a, 0x0178}, - {0x0179, 0x017e}, {0x01a0, 0x01a6}, {0x01b3, 0x01b7}, {0x01cd, 0x01dd}, - {0x01de, 0x01ef}, {0x01f4, 0x01f5}, {0x01f8, 0x01f9}, {0x01fa, 0x0220}, - {0x0222, 0x0234}, {0x023b, 0x023c}, {0x0241, 0x0242}, {0x0246, 0x024f}, - {0x03d8, 0x03ef}, {0x03f7, 0x03f8}, {0x03fa, 0x03fb}, {0x0460, 0x0481}, - {0x048a, 0x04bf}, {0x04c1, 0x04c4}, {0x04c5, 0x04c8}, {0x04c9, 0x04ce}, - {0x04ec, 0x04ed}, {0x04d0, 0x04eb}, {0x04ee, 0x04f5}, {0x04f6, 0x0513}, - {0x1e00, 0x1e95}, {0x1ea0, 0x1ef9}, {0x2183, 0x2184}, {0x2c60, 0x2c61}, - {0x2c67, 0x2c6c}, {0x2c75, 0x2c76}, {0x2c80, 0x2ce3}, - }; - /* - * Set the Unicode character at offset "Offset" to "Value". Note, - * "Value" is host endian. - */ - static const u16 uc_byte_table[][2] = { /* Offset, Value */ - {0x00ff, 0x0178}, {0x0180, 0x0243}, {0x0183, 0x0182}, {0x0185, 0x0184}, - {0x0188, 0x0187}, {0x018c, 0x018b}, {0x0192, 0x0191}, {0x0195, 0x01f6}, - {0x0199, 0x0198}, {0x019a, 0x023d}, {0x019e, 0x0220}, {0x01a8, 0x01a7}, - {0x01ad, 0x01ac}, {0x01b0, 0x01af}, {0x01b9, 0x01b8}, {0x01bd, 0x01bc}, - {0x01bf, 0x01f7}, {0x01c6, 0x01c4}, {0x01c9, 0x01c7}, {0x01cc, 0x01ca}, - {0x01dd, 0x018e}, {0x01f3, 0x01f1}, {0x023a, 0x2c65}, {0x023e, 0x2c66}, - {0x0253, 0x0181}, {0x0254, 0x0186}, {0x0259, 0x018f}, {0x025b, 0x0190}, - {0x0260, 0x0193}, {0x0263, 0x0194}, {0x0268, 0x0197}, {0x0269, 0x0196}, - {0x026b, 0x2c62}, {0x026f, 0x019c}, {0x0272, 0x019d}, {0x0275, 0x019f}, - {0x027d, 0x2c64}, {0x0280, 0x01a6}, {0x0283, 0x01a9}, {0x0288, 0x01ae}, - {0x0289, 0x0244}, {0x028c, 0x0245}, {0x0292, 0x01b7}, {0x03f2, 0x03f9}, - {0x04cf, 0x04c0}, {0x1d7d, 0x2c63}, {0x214e, 0x2132}, + /* This is the table used in NTFS volumes formatted by Windows 10. + * It was compressed by tools/compress_upcase_table.c. */ + static const u16 upcase_compressed[] = { + 0x0000, 0x0000, 0x0060, 0x0000, 0x0000, 0xffe0, 0x0019, 0x0061, + 0x0061, 0x0000, 0x001b, 0x005d, 0x0008, 0x0060, 0x0000, 0x0079, + 0x0000, 0x0000, 0x0000, 0xffff, 0x002f, 0x0100, 0x0002, 0x0000, + 0x0007, 0x012b, 0x0011, 0x0121, 0x002f, 0x0103, 0x0006, 0x0101, + 0x0000, 0x00c3, 0x0006, 0x0131, 0x0007, 0x012e, 0x0004, 0x0000, + 0x0003, 0x012f, 0x0000, 0x0061, 0x0004, 0x0130, 0x0000, 0x00a3, + 0x0003, 0x0000, 0x0000, 0x0082, 0x000b, 0x0131, 0x0006, 0x0189, + 0x0008, 0x012f, 0x0007, 0x012e, 0x0000, 0x0038, 0x0006, 0x0000, + 0x0000, 0xfffe, 0x0007, 0x01c4, 0x000f, 0x0101, 0x0000, 0xffb1, + 0x0015, 0x011e, 0x0004, 0x01cc, 0x002a, 0x0149, 0x0014, 0x0149, + 0x0007, 0x0000, 0x0009, 0x018c, 0x000b, 0x0138, 0x0000, 0x2a1f, + 0x0000, 0x2a1c, 0x0000, 0x0000, 0x0000, 0xff2e, 0x0000, 0xff32, + 0x0000, 0x0000, 0x0000, 0xff33, 0x0000, 0xff33, 0x0000, 0x0000, + 0x0000, 0xff36, 0x0000, 0x0000, 0x0000, 0xff35, 0x0004, 0x0000, + 0x0002, 0x0257, 0x0000, 0x0000, 0x0000, 0xff31, 0x0004, 0x0000, + 0x0000, 0xff2f, 0x0000, 0xff2d, 0x0000, 0x0000, 0x0000, 0x29f7, + 0x0003, 0x0000, 0x0002, 0x0269, 0x0000, 0x29fd, 0x0000, 0xff2b, + 0x0002, 0x0000, 0x0000, 0xff2a, 0x0007, 0x0000, 0x0000, 0x29e7, + 0x0002, 0x0000, 0x0000, 0xff26, 0x0005, 0x027e, 0x0003, 0x027e, + 0x0000, 0xffbb, 0x0000, 0xff27, 0x0000, 0xff27, 0x0000, 0xffb9, + 0x0005, 0x0000, 0x0000, 0xff25, 0x0065, 0x007b, 0x0079, 0x0293, + 0x0008, 0x012d, 0x0003, 0x019c, 0x0002, 0x037b, 0x002e, 0x0000, + 0x0000, 0xffda, 0x0000, 0xffdb, 0x0002, 0x03ad, 0x0012, 0x0060, + 0x000a, 0x0060, 0x0000, 0xffc0, 0x0000, 0xffc1, 0x0000, 0xffc1, + 0x0008, 0x0000, 0x0000, 0xfff8, 0x001a, 0x0118, 0x0000, 0x0007, + 0x0008, 0x018d, 0x0009, 0x0233, 0x0046, 0x0035, 0x0006, 0x0061, + 0x0000, 0xffb0, 0x000f, 0x0450, 0x0025, 0x010e, 0x000a, 0x036b, + 0x0032, 0x048b, 0x000e, 0x0100, 0x0000, 0xfff1, 0x0037, 0x048a, + 0x0026, 0x0465, 0x0034, 0x0000, 0x0000, 0xffd0, 0x0025, 0x0561, + 0x00de, 0x0293, 0x1714, 0x0587, 0x0000, 0x8a04, 0x0003, 0x0000, + 0x0000, 0x0ee6, 0x0087, 0x02ee, 0x0092, 0x1e01, 0x0069, 0x1df7, + 0x0000, 0x0008, 0x0007, 0x1f00, 0x0008, 0x0000, 0x000e, 0x1f02, + 0x0008, 0x1f0e, 0x0010, 0x1f06, 0x001a, 0x1f06, 0x0002, 0x1f0f, + 0x0007, 0x1f50, 0x0017, 0x1f19, 0x0000, 0x004a, 0x0000, 0x004a, + 0x0000, 0x0056, 0x0003, 0x1f72, 0x0000, 0x0064, 0x0000, 0x0064, + 0x0000, 0x0080, 0x0000, 0x0080, 0x0000, 0x0070, 0x0000, 0x0070, + 0x0000, 0x007e, 0x0000, 0x007e, 0x0028, 0x1f1e, 0x000c, 0x1f06, + 0x0000, 0x0000, 0x0000, 0x0009, 0x000f, 0x0000, 0x000d, 0x1fb3, + 0x000d, 0x1f44, 0x0008, 0x1fcd, 0x0006, 0x03f2, 0x0015, 0x1fbb, + 0x014e, 0x0587, 0x0000, 0xffe4, 0x0021, 0x0000, 0x0000, 0xfff0, + 0x000f, 0x2170, 0x000a, 0x0238, 0x0346, 0x0587, 0x0000, 0xffe6, + 0x0019, 0x24d0, 0x0746, 0x0587, 0x0026, 0x0561, 0x000b, 0x057e, + 0x0004, 0x012f, 0x0000, 0xd5d5, 0x0000, 0xd5d8, 0x000c, 0x022e, + 0x000e, 0x03f8, 0x006e, 0x1e33, 0x0011, 0x0000, 0x0000, 0xe3a0, + 0x0025, 0x2d00, 0x17f2, 0x0587, 0x6129, 0x2d26, 0x002e, 0x0201, + 0x002a, 0x1def, 0x0098, 0xa5b7, 0x0040, 0x1dff, 0x000e, 0x0368, + 0x000d, 0x022b, 0x034c, 0x2184, 0x5469, 0x2d26, 0x007f, 0x0061, + 0x0040, 0x0000, }; - for (u32 i = 0; i < 65536; i++) - upcase[i] = i; - - for (u32 r = 0; r < ARRAY_LEN(uc_run_table); r++) - for (u32 i = uc_run_table[r][0]; i < uc_run_table[r][1]; i++) - upcase[i] = i + uc_run_table[r][2]; - - for (u32 r = 0; r < ARRAY_LEN(uc_dup_table); r++) - for (u32 i = uc_dup_table[r][0]; i < uc_dup_table[r][1]; i += 2) - upcase[i + 1] = i; - - for (u32 r = 0; r < ARRAY_LEN(uc_byte_table); r++) - upcase[uc_byte_table[r][0]] = uc_byte_table[r][1]; - -#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 + /* Simple LZ decoder */ + const u16 *in_next = upcase_compressed; + for (u32 i = 0; i < ARRAY_LEN(upcase); ) { + u16 length = *in_next++; + u16 src_pos = *in_next++; + if (length == 0) { + /* Literal */ + upcase[i++] = src_pos; + } else { + /* Match */ + do { + upcase[i++] = upcase[src_pos++]; + } while (--length); + } + } + + /* Delta filter */ + for (u32 i = 0; i < ARRAY_LEN(upcase); i++) + upcase[i] += i; } -/* 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, @@ -539,40 +399,66 @@ cmp_utf16le_strings(const utf16lechar *s1, size_t n1, return (n1 < n2) ? -1 : 1; } -/* Duplicates a string of system-dependent encoding into a UTF-16LE string and - * returns the string and its length, in bytes, in the pointer arguments. Frees - * any existing string at the return location before overwriting it. */ +/* Like cmp_utf16le_strings(), but assumes the strings are null terminated. */ int -get_utf16le_string(const tchar *name, utf16lechar **name_utf16le_ret, - u16 *name_utf16le_nbytes_ret) +cmp_utf16le_strings_z(const utf16lechar *s1, const utf16lechar *s2, + bool ignore_case) { - utf16lechar *name_utf16le; - size_t name_utf16le_nbytes; - int ret; -#if TCHAR_IS_UTF16LE - name_utf16le_nbytes = tstrlen(name) * sizeof(utf16lechar); - name_utf16le = MALLOC(name_utf16le_nbytes + sizeof(utf16lechar)); - if (name_utf16le == NULL) - return WIMLIB_ERR_NOMEM; - memcpy(name_utf16le, name, name_utf16le_nbytes + sizeof(utf16lechar)); - ret = 0; -#else - - ret = tstr_to_utf16le(name, tstrlen(name), &name_utf16le, - &name_utf16le_nbytes); - if (ret == 0) { - if (name_utf16le_nbytes > 0xffff) { - FREE(name_utf16le); - ERROR("Multibyte string \"%"TS"\" is too long!", name); - ret = WIMLIB_ERR_INVALID_UTF8_STRING; + if (ignore_case) { + for (;;) { + u16 c1 = upcase[le16_to_cpu(*s1)]; + u16 c2 = upcase[le16_to_cpu(*s2)]; + if (c1 != c2) + return (c1 < c2) ? -1 : 1; + if (c1 == 0) + return 0; + s1++, s2++; } + } else { + while (*s1 && *s1 == *s2) + s1++, s2++; + if (*s1 == *s2) + return 0; + return (le16_to_cpu(*s1) < le16_to_cpu(*s2)) ? -1 : 1; } -#endif - if (ret == 0) { - FREE(*name_utf16le_ret); - *name_utf16le_ret = name_utf16le; - *name_utf16le_nbytes_ret = name_utf16le_nbytes; +} + +/* 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 *s, size_t size) +{ + utf16lechar *dup = MALLOC(size + sizeof(utf16lechar)); + if (dup) { + memcpy(dup, s, size); + dup[size / sizeof(utf16lechar)] = 0; } - return ret; + return dup; +} + +/* Duplicate a null-terminated UTF-16 string. */ +utf16lechar * +utf16le_dup(const utf16lechar *s) +{ + return memdup(s, utf16le_len_bytes(s) + sizeof(utf16lechar)); } +/* Return the length, in bytes, of a null terminated UTF-16 string, excluding + * the null terminator. */ +size_t +utf16le_len_bytes(const utf16lechar *s) +{ + const utf16lechar *p = s; + while (*p) + p++; + return (p - s) * sizeof(utf16lechar); +} + +/* 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) +{ + return utf16le_len_bytes(s) / sizeof(utf16lechar); +}