wimlib.net Git - wimlib/blob - src/encoding.c

   1 /*
   2  * encoding.c
   3  */
   4
   5 /*
   6  * Copyright (C) 2012, 2013 Eric Biggers
   7  *
   8  * This file is part of wimlib, a library for working with WIM files.
   9  *
  10  * wimlib is free software; you can redistribute it and/or modify it under the
  11  * terms of the GNU General Public License as published by the Free
  12  * Software Foundation; either version 3 of the License, or (at your option)
  13  * any later version.
  14  *
  15  * wimlib is distributed in the hope that it will be useful, but WITHOUT ANY
  16  * WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR
  17  * A PARTICULAR PURPOSE. See the GNU General Public License for more
  18  * details.
  19  *
  20  * You should have received a copy of the GNU General Public License
  21  * along with wimlib; if not, see http://www.gnu.org/licenses/.
  22  */
  23
  24 #ifdef HAVE_CONFIG_H
  25 #  include "config.h"
  26 #endif
  27
  28 #include "wimlib.h"
  29 #include "wimlib/assert.h"
  30 #include "wimlib/encoding.h"
  31 #include "wimlib/endianness.h"
  32 #include "wimlib/error.h"
  33 #include "wimlib/list.h"
  34 #include "wimlib/util.h"
  35
  36 #include <errno.h>
  37 #include <iconv.h>
  38 #include <pthread.h>
  39 #include <stdlib.h>
  40 #include <string.h>
  41
  42 #ifdef HAVE_ALLOCA_H
  43 #  include <alloca.h>
  44 #endif
  45
  46 bool wimlib_mbs_is_utf8 = !TCHAR_IS_UTF16LE;
  47
  48 /* List of iconv_t conversion descriptors for a specific character conversion.
  49  * The idea is that it is not thread-safe to have just one conversion
  50  * descriptor, but it also is inefficient to open a new conversion descriptor to
  51  * convert every string.  Both these problems can be solved by maintaining a
  52  * list of conversion descriptors; then, a thread can use an existing conversion
  53  * descriptor if available. */
  54 struct iconv_list_head {
  55         const char *from_encoding;
  56         const char *to_encoding;
  57         struct list_head list;
  58         pthread_mutex_t mutex;
  59 };
  60
  61 struct iconv_node {
  62         iconv_t cd;
  63         struct list_head list;
  64         struct iconv_list_head *head;
  65 };
  66
  67 #define ICONV_LIST(name, from, to)                      \
  68 struct iconv_list_head name = {                         \
  69         .from_encoding = from,                          \
  70         .to_encoding = to,                              \
  71         .list = LIST_HEAD_INIT(name.list),              \
  72         .mutex = PTHREAD_MUTEX_INITIALIZER,             \
  73 }
  74
  75 static iconv_t *
  76 get_iconv(struct iconv_list_head *head)
  77 {
  78         iconv_t cd;
  79         iconv_t *cd_p;
  80         struct iconv_node *i;
  81
  82         pthread_mutex_lock(&head->mutex);
  83         if (list_empty(&head->list)) {
  84                 cd = iconv_open(head->to_encoding, head->from_encoding);
  85                 if (cd == (iconv_t)-1) {
  86                         ERROR_WITH_ERRNO("Failed to open iconv from %s to %s",
  87                                          head->from_encoding, head->to_encoding);
  88                         cd_p = NULL;
  89                 } else {
  90                         i = MALLOC(sizeof(struct iconv_node));
  91                         if (i) {
  92                                 i->head = head;
  93                                 i->cd = cd;
  94                                 cd_p = &i->cd;
  95                         } else {
  96                                 iconv_close(cd);
  97                                 cd_p = NULL;
  98                         }
  99                 }
 100         } else {
 101                 i = container_of(head->list.next, struct iconv_node, list);
 102                 list_del(head->list.next);
 103                 cd_p = &i->cd;
 104         }
 105         pthread_mutex_unlock(&head->mutex);
 106         return cd_p;
 107 }
 108
 109 static void
 110 put_iconv(iconv_t *cd)
 111 {
 112         int errno_save = errno;
 113         struct iconv_node *i = container_of(cd, struct iconv_node, cd);
 114         struct iconv_list_head *head = i->head;
 115
 116         pthread_mutex_lock(&head->mutex);
 117         list_add(&i->list, &head->list);
 118         pthread_mutex_unlock(&head->mutex);
 119         errno = errno_save;
 120 }
 121
 122 /* Prevent printing an error message if a character conversion error occurs
 123  * while printing an error message.  (This variable is not per-thread but it
 124  * doesn't matter too much since it's just the error messages.) */
 125 static bool error_message_being_printed = false;
 126
 127 #define DEFINE_CHAR_CONVERSION_FUNCTIONS(varname1, longname1, chartype1,\
 128                                          varname2, longname2, chartype2,\
 129                                          earlyreturn_on_utf8_locale,    \
 130                                          earlyreturn_expr,              \
 131                                          worst_case_len_expr,           \
 132                                          err_return,                    \
 133                                          err_msg,                       \
 134                                          modifier)                      \
 135 static ICONV_LIST(iconv_##varname1##_to_##varname2,                     \
 136                   longname1, longname2);                                \
 137                                                                         \
 138 modifier int                                                            \
 139 varname1##_to_##varname2##_nbytes(const chartype1 *in, size_t in_nbytes,\
 140                                   size_t *out_nbytes_ret)               \
 141 {                                                                       \
 142         iconv_t *cd = get_iconv(&iconv_##varname1##_to_##varname2);     \
 143         if (cd == NULL)                                                 \
 144                 return WIMLIB_ERR_ICONV_NOT_AVAILABLE;                  \
 145                                                                         \
 146         chartype2 *buf;                                                 \
 147         size_t bufsize;                                                 \
 148         bool buf_onheap;                                                \
 149         bufsize = (worst_case_len_expr) * sizeof(chartype2);            \
 150         /* Worst case length */                                         \
 151         if (bufsize <= STACK_MAX) {                                     \
 152                 buf = alloca(bufsize);                                  \
 153                 buf_onheap = false;                                     \
 154         } else {                                                        \
 155                 buf = MALLOC(bufsize);                                  \
 156                 if (!buf)                                               \
 157                         return WIMLIB_ERR_NOMEM;                        \
 158                 buf_onheap = true;                                      \
 159         }                                                               \
 160                                                                         \
 161         char *inbuf = (char*)in;                                        \
 162         size_t inbytesleft = in_nbytes;                                 \
 163         char *outbuf = (char*)buf;                                      \
 164         size_t outbytesleft = bufsize;                                  \
 165         size_t len;                                                     \
 166         int ret;                                                        \
 167                                                                         \
 168         len = iconv(*cd, &inbuf, &inbytesleft, &outbuf, &outbytesleft); \
 169         if (len == (size_t)-1) {                                        \
 170                 if (!error_message_being_printed) {                     \
 171                         error_message_being_printed = true;             \
 172                         err_msg;                                        \
 173                         error_message_being_printed = false;            \
 174                 }                                                       \
 175                 ret = err_return;                                       \
 176         } else {                                                        \
 177                 *out_nbytes_ret = bufsize - outbytesleft;               \
 178                 ret = 0;                                                \
 179         }                                                               \
 180         put_iconv(cd);                                                  \
 181         if (buf_onheap)                                                 \
 182                 FREE(buf);                                              \
 183         return ret;                                                     \
 184 }                                                                       \
 185                                                                         \
 186 modifier int                                                            \
 187 varname1##_to_##varname2##_buf(const chartype1 *in, size_t in_nbytes,   \
 188                                chartype2 *out)                          \
 189 {                                                                       \
 190         iconv_t *cd = get_iconv(&iconv_##varname1##_to_##varname2);     \
 191         if (cd == NULL)                                                 \
 192                 return WIMLIB_ERR_ICONV_NOT_AVAILABLE;                  \
 193                                                                         \
 194         char *inbuf = (char*)in;                                        \
 195         size_t inbytesleft = in_nbytes;                                 \
 196         char *outbuf = (char*)out;                                      \
 197         const size_t LARGE_NUMBER = 1000000000;                         \
 198         size_t outbytesleft = LARGE_NUMBER;                             \
 199         size_t len;                                                     \
 200         int ret;                                                        \
 201                                                                         \
 202         len = iconv(*cd, &inbuf, &inbytesleft, &outbuf, &outbytesleft); \
 203         if (len == (size_t)-1) {                                        \
 204                 if (!error_message_being_printed) {                     \
 205                         error_message_being_printed = true;             \
 206                         err_msg;                                        \
 207                         error_message_being_printed = false;            \
 208                 }                                                       \
 209                 ret = err_return;                                       \
 210         } else {                                                        \
 211                 out[(LARGE_NUMBER-outbytesleft)/sizeof(chartype2)] = 0; \
 212                 ret = 0;                                                \
 213         }                                                               \
 214         put_iconv(cd);                                                  \
 215         return ret;                                                     \
 216 }                                                                       \
 217                                                                         \
 218 modifier int                                                            \
 219 varname1##_to_##varname2(const chartype1 *in, size_t in_nbytes,         \
 220                          chartype2 **out_ret,                           \
 221                          size_t *out_nbytes_ret)                        \
 222 {                                                                       \
 223         int ret;                                                        \
 224         chartype2 *out;                                                 \
 225         size_t out_nbytes;                                              \
 226                                                                         \
 227         if (earlyreturn_on_utf8_locale && wimlib_mbs_is_utf8) {         \
 228                 earlyreturn_expr;                                       \
 229                 /* Out same as in */                                    \
 230                 out = MALLOC(in_nbytes + sizeof(chartype2));            \
 231                 if (!out)                                               \
 232                         return WIMLIB_ERR_NOMEM;                        \
 233                 memcpy(out, in, in_nbytes);                             \
 234                 out[in_nbytes / sizeof(chartype2)] = 0;                 \
 235                 *out_ret = out;                                         \
 236                 *out_nbytes_ret = in_nbytes;                            \
 237                 return 0;                                               \
 238         }                                                               \
 239                                                                         \
 240         ret = varname1##_to_##varname2##_nbytes(in, in_nbytes,          \
 241                                                 &out_nbytes);           \
 242         if (ret)                                                        \
 243                 return ret;                                             \
 244                                                                         \
 245         out = MALLOC(out_nbytes + sizeof(chartype2));                   \
 246         if (!out)                                                       \
 247                 return WIMLIB_ERR_NOMEM;                                \
 248                                                                         \
 249         ret = varname1##_to_##varname2##_buf(in, in_nbytes, out);       \
 250         if (ret) {                                                      \
 251                 FREE(out);                                              \
 252         } else {                                                        \
 253                 *out_ret = out;                                         \
 254                 *out_nbytes_ret = out_nbytes;                           \
 255         }                                                               \
 256         return ret;                                                     \
 257 }
 258
 259 #if !TCHAR_IS_UTF16LE
 260
 261 /* UNIX */
 262
 263 DEFINE_CHAR_CONVERSION_FUNCTIONS(utf8, "UTF-8", tchar,
 264                                  utf16le, "UTF-16LE", utf16lechar,
 265                                  false,
 266                                  ,
 267                                  in_nbytes * 2,
 268                                  WIMLIB_ERR_INVALID_UTF8_STRING,
 269                                  ERROR_WITH_ERRNO("Failed to convert UTF-8 string "
 270                                                   "to UTF-16LE string!"),
 271                                  static)
 272
 273 DEFINE_CHAR_CONVERSION_FUNCTIONS(utf16le, "UTF-16LE", utf16lechar,
 274                                  utf8, "UTF-8", tchar,
 275                                  false,
 276                                  ,
 277                                  in_nbytes * 2,
 278                                  WIMLIB_ERR_INVALID_UTF16_STRING,
 279                                  ERROR_WITH_ERRNO("Failed to convert UTF-16LE string "
 280                                                   "to UTF-8 string!"),
 281                                  static)
 282
 283 DEFINE_CHAR_CONVERSION_FUNCTIONS(tstr, "", tchar,
 284                                  utf16le, "UTF-16LE", utf16lechar,
 285                                  true,
 286                                  return utf8_to_utf16le(in, in_nbytes, out_ret, out_nbytes_ret),
 287                                  in_nbytes * 2,
 288                                  WIMLIB_ERR_INVALID_MULTIBYTE_STRING,
 289                                  ERROR_WITH_ERRNO("Failed to convert multibyte "
 290                                                   "string \"%"TS"\" to UTF-16LE string!", in);
 291                                  ERROR("If the data you provided was UTF-8, please make sure "
 292                                        "the character encoding\n"
 293                                        "        of your current locale is UTF-8."),
 294                                  )
 295
 296 DEFINE_CHAR_CONVERSION_FUNCTIONS(utf16le, "UTF-16LE", utf16lechar,
 297                                  tstr, "", tchar,
 298                                  true,
 299                                  return utf16le_to_utf8(in, in_nbytes, out_ret, out_nbytes_ret),
 300                                  in_nbytes * 2,
 301                                  WIMLIB_ERR_UNICODE_STRING_NOT_REPRESENTABLE,
 302                                  ERROR("Failed to convert UTF-16LE string to "
 303                                        "multibyte string!");
 304                                  ERROR("This may be because the UTF-16LE string "
 305                                        "could not be represented\n"
 306                                        "        in your locale's character encoding."),
 307                                  )
 308 #endif
 309
 310 /* tchar to UTF-8 and back */
 311 #if TCHAR_IS_UTF16LE
 312
 313 /* Windows */
 314 DEFINE_CHAR_CONVERSION_FUNCTIONS(tstr, "UTF-16LE", tchar,
 315                                  utf8, "UTF-8", char,
 316                                  false,
 317                                  ,
 318                                  in_nbytes * 2,
 319                                  WIMLIB_ERR_INVALID_UTF16_STRING,
 320                                  ERROR_WITH_ERRNO("Failed to convert UTF-16LE "
 321                                                   "string \"%"TS"\" to UTF-8 string!", in),
 322                                  )
 323
 324 DEFINE_CHAR_CONVERSION_FUNCTIONS(utf8, "UTF-8", char,
 325                                  tstr, "UTF-16LE", tchar,
 326                                  false,
 327                                  ,
 328                                  in_nbytes * 2,
 329                                  WIMLIB_ERR_INVALID_UTF8_STRING,
 330                                  ERROR_WITH_ERRNO("Failed to convert UTF-8 string "
 331                                                   "to UTF-16LE string!"),
 332                                  )
 333 #else
 334
 335 /* UNIX */
 336
 337 DEFINE_CHAR_CONVERSION_FUNCTIONS(tstr, "", tchar,
 338                                  utf8, "UTF-8", char,
 339                                  true,
 340                                  ,
 341                                  in_nbytes * 4,
 342                                  WIMLIB_ERR_INVALID_MULTIBYTE_STRING,
 343                                  ERROR_WITH_ERRNO("Failed to convert multibyte "
 344                                                   "string \"%"TS"\" to UTF-8 string!", in);
 345                                  ERROR("If the data you provided was UTF-8, please make sure "
 346                                        "the character\n"
 347                                        "        encoding of your current locale is UTF-8."),
 348                                  )
 349
 350 DEFINE_CHAR_CONVERSION_FUNCTIONS(utf8, "UTF-8", char,
 351                                  tstr, "", tchar,
 352                                  true,
 353                                  ,
 354                                  in_nbytes * 4,
 355                                  WIMLIB_ERR_UNICODE_STRING_NOT_REPRESENTABLE,
 356                                  ERROR("Failed to convert UTF-8 string to "
 357                                        "multibyte string!");
 358                                  ERROR("This may be because the UTF-8 data "
 359                                        "could not be represented\n"
 360                                        "        in your locale's character encoding."),
 361                                  )
 362 #endif
 363
 364 int
 365 tstr_to_utf8_simple(const tchar *tstr, char **out)
 366 {
 367         size_t out_nbytes;
 368         return tstr_to_utf8(tstr, tstrlen(tstr) * sizeof(tchar),
 369                             out, &out_nbytes);
 370 }
 371
 372 int
 373 utf8_to_tstr_simple(const char *utf8str, tchar **out)
 374 {
 375         size_t out_nbytes;
 376         return utf8_to_tstr(utf8str, strlen(utf8str), out, &out_nbytes);
 377 }
 378
 379 static void
 380 iconv_cleanup(struct iconv_list_head *head)
 381 {
 382         pthread_mutex_destroy(&head->mutex);
 383         while (!list_empty(&head->list)) {
 384                 struct iconv_node *i;
 385
 386                 i = container_of(head->list.next, struct iconv_node, list);
 387                 list_del(&i->list);
 388                 iconv_close(i->cd);
 389                 FREE(i);
 390         }
 391 }
 392
 393 void
 394 iconv_global_cleanup(void)
 395 {
 396         iconv_cleanup(&iconv_utf8_to_tstr);
 397         iconv_cleanup(&iconv_tstr_to_utf8);
 398 #if !TCHAR_IS_UTF16LE
 399         iconv_cleanup(&iconv_utf16le_to_tstr);
 400         iconv_cleanup(&iconv_tstr_to_utf16le);
 401         iconv_cleanup(&iconv_utf16le_to_utf8);
 402         iconv_cleanup(&iconv_utf8_to_utf16le);
 403 #endif
 404 }
 405
 406 /* A table that maps from UCS-2 characters to their upper case equivalents.
 407  * Index and array values are both CPU endian.
 408  * Note: this is only an *approximation* of real UTF-16 case folding.
 409  */
 410 static u16 upcase[65536];
 411
 412 /*
 413  * Initialize the 'upcase' table.
 414  *
 415  * Code modified from NTFS-3g, original copyright notices below:
 416  *
 417  * Copyright (c) 2000-2004 Anton Altaparmakov
 418  * Copyright (c) 2002-2009 Szabolcs Szakacsits
 419  * Copyright (c) 2008-2011 Jean-Pierre Andre
 420  * Copyright (c) 2008      Bernhard Kaindl
 421  *
 422  * License GPLv2 or later.
 423  *
 424  * The expanded table itself is the same as that used by Windows Vista for
 425  * NTFS case insensitivity.
 426  */
 427 void
 428 init_upcase(void)
 429 {
 430         /*
 431          *      This is the table as defined by Vista
 432          */
 433         /*
 434          * "Start" is inclusive and "End" is exclusive, every value has the
 435          * value of "Add" added to it.
 436          */
 437         static const u16 uc_run_table[][3] = { /* Start, End, Add */
 438         {0x0061, 0x007b,   -32}, {0x00e0, 0x00f7,  -32}, {0x00f8, 0x00ff, -32},
 439         {0x0256, 0x0258,  -205}, {0x028a, 0x028c, -217}, {0x037b, 0x037e, 130},
 440         {0x03ac, 0x03ad,   -38}, {0x03ad, 0x03b0,  -37}, {0x03b1, 0x03c2, -32},
 441         {0x03c2, 0x03c3,   -31}, {0x03c3, 0x03cc,  -32}, {0x03cc, 0x03cd, -64},
 442         {0x03cd, 0x03cf,   -63}, {0x0430, 0x0450,  -32}, {0x0450, 0x0460, -80},
 443         {0x0561, 0x0587,   -48}, {0x1f00, 0x1f08,    8}, {0x1f10, 0x1f16,   8},
 444         {0x1f20, 0x1f28,     8}, {0x1f30, 0x1f38,    8}, {0x1f40, 0x1f46,   8},
 445         {0x1f51, 0x1f52,     8}, {0x1f53, 0x1f54,    8}, {0x1f55, 0x1f56,   8},
 446         {0x1f57, 0x1f58,     8}, {0x1f60, 0x1f68,    8}, {0x1f70, 0x1f72,  74},
 447         {0x1f72, 0x1f76,    86}, {0x1f76, 0x1f78,  100}, {0x1f78, 0x1f7a, 128},
 448         {0x1f7a, 0x1f7c,   112}, {0x1f7c, 0x1f7e,  126}, {0x1f80, 0x1f88,   8},
 449         {0x1f90, 0x1f98,     8}, {0x1fa0, 0x1fa8,    8}, {0x1fb0, 0x1fb2,   8},
 450         {0x1fb3, 0x1fb4,     9}, {0x1fcc, 0x1fcd,   -9}, {0x1fd0, 0x1fd2,   8},
 451         {0x1fe0, 0x1fe2,     8}, {0x1fe5, 0x1fe6,    7}, {0x1ffc, 0x1ffd,  -9},
 452         {0x2170, 0x2180,   -16}, {0x24d0, 0x24ea,  -26}, {0x2c30, 0x2c5f, -48},
 453         {0x2d00, 0x2d26, -7264}, {0xff41, 0xff5b,  -32},
 454         };
 455         /*
 456          * "Start" is exclusive and "End" is inclusive, every second value is
 457          * decremented by one.
 458          */
 459         static const u16 uc_dup_table[][2] = { /* Start, End */
 460         {0x0100, 0x012f}, {0x0132, 0x0137}, {0x0139, 0x0149}, {0x014a, 0x0178},
 461         {0x0179, 0x017e}, {0x01a0, 0x01a6}, {0x01b3, 0x01b7}, {0x01cd, 0x01dd},
 462         {0x01de, 0x01ef}, {0x01f4, 0x01f5}, {0x01f8, 0x01f9}, {0x01fa, 0x0220},
 463         {0x0222, 0x0234}, {0x023b, 0x023c}, {0x0241, 0x0242}, {0x0246, 0x024f},
 464         {0x03d8, 0x03ef}, {0x03f7, 0x03f8}, {0x03fa, 0x03fb}, {0x0460, 0x0481},
 465         {0x048a, 0x04bf}, {0x04c1, 0x04c4}, {0x04c5, 0x04c8}, {0x04c9, 0x04ce},
 466         {0x04ec, 0x04ed}, {0x04d0, 0x04eb}, {0x04ee, 0x04f5}, {0x04f6, 0x0513},
 467         {0x1e00, 0x1e95}, {0x1ea0, 0x1ef9}, {0x2183, 0x2184}, {0x2c60, 0x2c61},
 468         {0x2c67, 0x2c6c}, {0x2c75, 0x2c76}, {0x2c80, 0x2ce3},
 469         };
 470         /*
 471          * Set the Unicode character at offset "Offset" to "Value".  Note,
 472          * "Value" is host endian.
 473          */
 474         static const u16 uc_byte_table[][2] = { /* Offset, Value */
 475         {0x00ff, 0x0178}, {0x0180, 0x0243}, {0x0183, 0x0182}, {0x0185, 0x0184},
 476         {0x0188, 0x0187}, {0x018c, 0x018b}, {0x0192, 0x0191}, {0x0195, 0x01f6},
 477         {0x0199, 0x0198}, {0x019a, 0x023d}, {0x019e, 0x0220}, {0x01a8, 0x01a7},
 478         {0x01ad, 0x01ac}, {0x01b0, 0x01af}, {0x01b9, 0x01b8}, {0x01bd, 0x01bc},
 479         {0x01bf, 0x01f7}, {0x01c6, 0x01c4}, {0x01c9, 0x01c7}, {0x01cc, 0x01ca},
 480         {0x01dd, 0x018e}, {0x01f3, 0x01f1}, {0x023a, 0x2c65}, {0x023e, 0x2c66},
 481         {0x0253, 0x0181}, {0x0254, 0x0186}, {0x0259, 0x018f}, {0x025b, 0x0190},
 482         {0x0260, 0x0193}, {0x0263, 0x0194}, {0x0268, 0x0197}, {0x0269, 0x0196},
 483         {0x026b, 0x2c62}, {0x026f, 0x019c}, {0x0272, 0x019d}, {0x0275, 0x019f},
 484         {0x027d, 0x2c64}, {0x0280, 0x01a6}, {0x0283, 0x01a9}, {0x0288, 0x01ae},
 485         {0x0289, 0x0244}, {0x028c, 0x0245}, {0x0292, 0x01b7}, {0x03f2, 0x03f9},
 486         {0x04cf, 0x04c0}, {0x1d7d, 0x2c63}, {0x214e, 0x2132},
 487         };
 488
 489         for (u32 i = 0; i < 65536; i++)
 490                 upcase[i] = i;
 491
 492         for (u32 r = 0; r < ARRAY_LEN(uc_run_table); r++)
 493                 for (u32 i = uc_run_table[r][0]; i < uc_run_table[r][1]; i++)
 494                         upcase[i] = i + uc_run_table[r][2];
 495
 496         for (u32 r = 0; r < ARRAY_LEN(uc_dup_table); r++)
 497                 for (u32 i = uc_dup_table[r][0]; i < uc_dup_table[r][1]; i += 2)
 498                         upcase[i + 1] = i;
 499
 500         for (u32 r = 0; r < ARRAY_LEN(uc_byte_table); r++)
 501                 upcase[uc_byte_table[r][0]] = uc_byte_table[r][1];
 502
 503 #if 0
 504         /* Sanity checks  */
 505         wimlib_assert(upcase['a'] == 'A');
 506         wimlib_assert(upcase['A'] == 'A');
 507         wimlib_assert(upcase['z'] == 'Z');
 508         wimlib_assert(upcase['Z'] == 'Z');
 509         wimlib_assert(upcase['1'] == '1');
 510         wimlib_assert(upcase[0x00e9] == 0x00c9); /* Latin letter e, with acute accent  */
 511         wimlib_assert(upcase[0x00c9] == 0x00c9);
 512         wimlib_assert(upcase[0x03c1] == 0x03a1); /* Greek letter rho  */
 513         wimlib_assert(upcase[0x03a1] == 0x03a1);
 514         wimlib_assert(upcase[0x0436] == 0x0416); /* Cyrillic letter zhe  */
 515         wimlib_assert(upcase[0x0416] == 0x0416);
 516         wimlib_assert(upcase[0x0567] == 0x0537); /* Armenian letter eh  */
 517         wimlib_assert(upcase[0x0537] == 0x0537);
 518         wimlib_assert(upcase[0x24d0] == 0x24b6); /* Circled Latin letter A
 519                                                     (is that a real character???)  */
 520         wimlib_assert(upcase[0x24b6] == 0x24b6);
 521         wimlib_assert(upcase[0x2603] == 0x2603); /* Note to self: Upper case
 522                                                     snowman symbol does not
 523                                                     exist.  */
 524 #endif
 525 }
 526
 527 /* Compare UTF-16LE strings case-sensitively (%ignore_case == false) or
 528  * case-insensitively (%ignore_case == true).
 529  *
 530  * This is implemented using the default upper-case table used by NTFS.  It does
 531  * not handle all possible cases allowed by UTF-16LE.  For example, different
 532  * normalizations of the same sequence of "characters" are not considered equal.
 533  * It hopefully does the right thing most of the time though.  */
 534 int
 535 cmp_utf16le_strings(const utf16lechar *s1, size_t n1,
 536                     const utf16lechar *s2, size_t n2,
 537                     bool ignore_case)
 538 {
 539         size_t n = min(n1, n2);
 540
 541         if (ignore_case) {
 542                 for (size_t i = 0; i < n; i++) {
 543                         u16 c1 = upcase[le16_to_cpu(s1[i])];
 544                         u16 c2 = upcase[le16_to_cpu(s2[i])];
 545                         if (c1 != c2)
 546                                 return (c1 < c2) ? -1 : 1;
 547                 }
 548         } else {
 549                 for (size_t i = 0; i < n; i++) {
 550                         u16 c1 = le16_to_cpu(s1[i]);
 551                         u16 c2 = le16_to_cpu(s2[i]);
 552                         if (c1 != c2)
 553                                 return (c1 < c2) ? -1 : 1;
 554                 }
 555         }
 556         if (n1 == n2)
 557                 return 0;
 558         return (n1 < n2) ? -1 : 1;
 559 }
 560
 561 /* Duplicates a string of system-dependent encoding into a UTF-16LE string and
 562  * returns the string and its length, in bytes, in the pointer arguments.  Frees
 563  * any existing string at the return location before overwriting it. */
 564 int
 565 get_utf16le_string(const tchar *name, utf16lechar **name_utf16le_ret,
 566                    u16 *name_utf16le_nbytes_ret)
 567 {
 568         utf16lechar *name_utf16le;
 569         size_t name_utf16le_nbytes;
 570         int ret;
 571 #if TCHAR_IS_UTF16LE
 572         name_utf16le_nbytes = tstrlen(name) * sizeof(utf16lechar);
 573         name_utf16le = MALLOC(name_utf16le_nbytes + sizeof(utf16lechar));
 574         if (name_utf16le == NULL)
 575                 return WIMLIB_ERR_NOMEM;
 576         memcpy(name_utf16le, name, name_utf16le_nbytes + sizeof(utf16lechar));
 577         ret = 0;
 578 #else
 579
 580         ret = tstr_to_utf16le(name, tstrlen(name), &name_utf16le,
 581                               &name_utf16le_nbytes);
 582         if (ret == 0) {
 583                 if (name_utf16le_nbytes > 0xffff) {
 584                         FREE(name_utf16le);
 585                         ERROR("Multibyte string \"%"TS"\" is too long!", name);
 586                         ret = WIMLIB_ERR_INVALID_UTF8_STRING;
 587                 }
 588         }
 589 #endif
 590         if (ret == 0) {
 591                 FREE(*name_utf16le_ret);
 592                 *name_utf16le_ret = name_utf16le;
 593                 *name_utf16le_nbytes_ret = name_utf16le_nbytes;
 594         }
 595         return ret;
 596 }
 597