src/common/strconv.cpp

   1 /////////////////////////////////////////////////////////////////////////////
   2 // Name:        strconv.cpp
   3 // Purpose:     Unicode conversion classes
   4 // Author:      Ove Kaaven, Robert Roebling, Vadim Zeitlin, Vaclav Slavik,
   5 //              Ryan Norton, Fredrik Roubert (UTF7)
   6 // Modified by:
   7 // Created:     29/01/98
   8 // RCS-ID:      $Id$
   9 // Copyright:   (c) 1999 Ove Kaaven, Robert Roebling, Vaclav Slavik
  10 //              (c) 2000-2003 Vadim Zeitlin
  11 //              (c) 2004 Ryan Norton, Fredrik Roubert
  12 // Licence:     wxWindows licence
  13 /////////////////////////////////////////////////////////////////////////////
  14
  15 // ============================================================================
  16 // declarations
  17 // ============================================================================
  18
  19 // ----------------------------------------------------------------------------
  20 // headers
  21 // ----------------------------------------------------------------------------
  22
  23 #if defined(__GNUG__) && !defined(NO_GCC_PRAGMA)
  24   #pragma implementation "strconv.h"
  25 #endif
  26
  27 // For compilers that support precompilation, includes "wx.h".
  28 #include "wx/wxprec.h"
  29
  30 #ifdef __BORLANDC__
  31   #pragma hdrstop
  32 #endif
  33
  34 #ifndef WX_PRECOMP
  35     #include "wx/intl.h"
  36     #include "wx/log.h"
  37 #endif // WX_PRECOMP
  38
  39 #include "wx/strconv.h"
  40
  41 #if wxUSE_WCHAR_T
  42
  43 #ifdef __WINDOWS__
  44     #include "wx/msw/private.h"
  45     #include "wx/msw/missing.h"
  46 #endif
  47
  48 #ifndef __WXWINCE__
  49 #include <errno.h>
  50 #endif
  51
  52 #include <ctype.h>
  53 #include <string.h>
  54 #include <stdlib.h>
  55
  56 #if defined(__WIN32__) && !defined(__WXMICROWIN__)
  57     #define wxHAVE_WIN32_MB2WC
  58 #endif // __WIN32__ but !__WXMICROWIN__
  59
  60 #ifdef __SALFORDC__
  61     #include <clib.h>
  62 #endif
  63
  64 #ifdef HAVE_ICONV
  65     #include <iconv.h>
  66     #include "wx/thread.h"
  67 #endif
  68
  69 #include "wx/encconv.h"
  70 #include "wx/fontmap.h"
  71 #include "wx/utils.h"
  72
  73 #ifdef __WXMAC__
  74 #ifndef __DARWIN__
  75 #include <ATSUnicode.h>
  76 #include <TextCommon.h>
  77 #include <TextEncodingConverter.h>
  78 #endif
  79
  80 #include  "wx/mac/private.h"  // includes mac headers
  81 #endif
  82
  83 #define TRACE_STRCONV _T("strconv")
  84
  85 // ============================================================================
  86 // implementation
  87 // ============================================================================
  88
  89 // ----------------------------------------------------------------------------
  90 // UTF-16 en/decoding to/from UCS-4
  91 // ----------------------------------------------------------------------------
  92
  93
  94 static size_t encode_utf16(wxUint32 input, wxUint16 *output)
  95 {
  96     if (input<=0xffff)
  97     {
  98         if (output)
  99             *output = (wxUint16) input;
 100         return 1;
 101     }
 102     else if (input>=0x110000)
 103     {
 104         return (size_t)-1;
 105     }
 106     else
 107     {
 108         if (output)
 109         {
 110             *output++ = (wxUint16) ((input >> 10)+0xd7c0);
 111             *output = (wxUint16) ((input&0x3ff)+0xdc00);
 112         }
 113         return 2;
 114     }
 115 }
 116
 117 static size_t decode_utf16(const wxUint16* input, wxUint32& output)
 118 {
 119     if ((*input<0xd800) || (*input>0xdfff))
 120     {
 121         output = *input;
 122         return 1;
 123     }
 124     else if ((input[1]<0xdc00) || (input[1]>0xdfff))
 125     {
 126         output = *input;
 127         return (size_t)-1;
 128     }
 129     else
 130     {
 131         output = ((input[0] - 0xd7c0) << 10) + (input[1] - 0xdc00);
 132         return 2;
 133     }
 134 }
 135
 136
 137 // ----------------------------------------------------------------------------
 138 // wxMBConv
 139 // ----------------------------------------------------------------------------
 140
 141 wxMBConv::~wxMBConv()
 142 {
 143     // nothing to do here (necessary for Darwin linking probably)
 144 }
 145
 146 const wxWCharBuffer wxMBConv::cMB2WC(const char *psz) const
 147 {
 148     if ( psz )
 149     {
 150         // calculate the length of the buffer needed first
 151         size_t nLen = MB2WC(NULL, psz, 0);
 152         if ( nLen != (size_t)-1 )
 153         {
 154             // now do the actual conversion
 155             wxWCharBuffer buf(nLen);
 156             nLen = MB2WC(buf.data(), psz, nLen + 1); // with the trailing NULL
 157             if ( nLen != (size_t)-1 )
 158             {
 159                 return buf;
 160             }
 161         }
 162     }
 163
 164     wxWCharBuffer buf((wchar_t *)NULL);
 165
 166     return buf;
 167 }
 168
 169 const wxCharBuffer wxMBConv::cWC2MB(const wchar_t *pwz) const
 170 {
 171     if ( pwz )
 172     {
 173         size_t nLen = WC2MB(NULL, pwz, 0);
 174         if ( nLen != (size_t)-1 )
 175         {
 176             wxCharBuffer buf(nLen+3);       // space for a wxUint32 trailing zero
 177             nLen = WC2MB(buf.data(), pwz, nLen + 4);
 178             if ( nLen != (size_t)-1 )
 179             {
 180                 return buf;
 181             }
 182         }
 183     }
 184
 185     wxCharBuffer buf((char *)NULL);
 186
 187     return buf;
 188 }
 189
 190 const wxWCharBuffer wxMBConv::cMB2WC(const char *szString, size_t nStringLen, size_t* pOutSize) const
 191 {
 192     wxASSERT(pOutSize != NULL);
 193
 194     const char* szEnd = szString + nStringLen + 1;
 195     const char* szPos = szString;
 196     const char* szStart = szPos;
 197
 198     size_t nActualLength = 0;
 199     size_t nCurrentSize = nStringLen; //try normal size first (should never resize?)
 200
 201     wxWCharBuffer theBuffer(nCurrentSize);
 202
 203     //Convert the string until the length() is reached, continuing the
 204     //loop every time a null character is reached
 205     while(szPos != szEnd)
 206     {
 207         wxASSERT(szPos < szEnd); //something is _really_ screwed up if this rings true
 208
 209         //Get the length of the current (sub)string
 210         size_t nLen = MB2WC(NULL, szPos, 0);
 211
 212         //Invalid conversion?
 213         if( nLen == (size_t)-1 )
 214         {
 215             *pOutSize = 0;
 216             theBuffer.data()[0u] = wxT('\0');
 217             return theBuffer;
 218         }
 219
 220
 221         //Increase the actual length (+1 for current null character)
 222         nActualLength += nLen + 1;
 223
 224         //if buffer too big, realloc the buffer
 225         if (nActualLength > (nCurrentSize+1))
 226         {
 227             wxWCharBuffer theNewBuffer(nCurrentSize << 1);
 228             memcpy(theNewBuffer.data(), theBuffer.data(), nCurrentSize * sizeof(wchar_t));
 229             theBuffer = theNewBuffer;
 230             nCurrentSize <<= 1;
 231         }
 232
 233         //Convert the current (sub)string
 234         if ( MB2WC(&theBuffer.data()[szPos - szStart], szPos, nLen + 1) == (size_t)-1 )
 235         {
 236             *pOutSize = 0;
 237             theBuffer.data()[0u] = wxT('\0');
 238             return theBuffer;
 239         }
 240
 241         //Increment to next (sub)string
 242         //Note that we have to use strlen instead of nLen here
 243         //because XX2XX gives us the size of the output buffer,
 244         //which is not necessarily the length of the string
 245         szPos += strlen(szPos) + 1;
 246     }
 247
 248     //success - return actual length and the buffer
 249     *pOutSize = nActualLength;
 250     return theBuffer;
 251 }
 252
 253 const wxCharBuffer wxMBConv::cWC2MB(const wchar_t *szString, size_t nStringLen, size_t* pOutSize) const
 254 {
 255     wxASSERT(pOutSize != NULL);
 256
 257     const wchar_t* szEnd = szString + nStringLen + 1;
 258     const wchar_t* szPos = szString;
 259     const wchar_t* szStart = szPos;
 260
 261     size_t nActualLength = 0;
 262     size_t nCurrentSize = nStringLen << 2; //try * 4 first
 263
 264     wxCharBuffer theBuffer(nCurrentSize);
 265
 266     //Convert the string until the length() is reached, continuing the
 267     //loop every time a null character is reached
 268     while(szPos != szEnd)
 269     {
 270         wxASSERT(szPos < szEnd); //something is _really_ screwed up if this rings true
 271
 272         //Get the length of the current (sub)string
 273         size_t nLen = WC2MB(NULL, szPos, 0);
 274
 275         //Invalid conversion?
 276         if( nLen == (size_t)-1 )
 277         {
 278             *pOutSize = 0;
 279             theBuffer.data()[0u] = wxT('\0');
 280             return theBuffer;
 281         }
 282
 283         //Increase the actual length (+1 for current null character)
 284         nActualLength += nLen + 1;
 285
 286         //if buffer too big, realloc the buffer
 287         if (nActualLength > (nCurrentSize+1))
 288         {
 289             wxCharBuffer theNewBuffer(nCurrentSize << 1);
 290             memcpy(theNewBuffer.data(), theBuffer.data(), nCurrentSize);
 291             theBuffer = theNewBuffer;
 292             nCurrentSize <<= 1;
 293         }
 294
 295         //Convert the current (sub)string
 296         if(WC2MB(&theBuffer.data()[szPos - szStart], szPos, nLen + 1) == (size_t)-1 )
 297         {
 298             *pOutSize = 0;
 299             theBuffer.data()[0u] = wxT('\0');
 300             return theBuffer;
 301         }
 302
 303         //Increment to next (sub)string
 304         //Note that we have to use wxWcslen instead of nLen here
 305         //because XX2XX gives us the size of the output buffer,
 306         //which is not necessarily the length of the string
 307         szPos += wxWcslen(szPos) + 1;
 308     }
 309
 310     //success - return actual length and the buffer
 311     *pOutSize = nActualLength;
 312     return theBuffer;
 313 }
 314
 315 // ----------------------------------------------------------------------------
 316 // wxMBConvLibc
 317 // ----------------------------------------------------------------------------
 318
 319 size_t wxMBConvLibc::MB2WC(wchar_t *buf, const char *psz, size_t n) const
 320 {
 321     return wxMB2WC(buf, psz, n);
 322 }
 323
 324 size_t wxMBConvLibc::WC2MB(char *buf, const wchar_t *psz, size_t n) const
 325 {
 326     return wxWC2MB(buf, psz, n);
 327 }
 328
 329 #ifdef __UNIX__
 330
 331 // ----------------------------------------------------------------------------
 332 // wxConvBrokenFileNames
 333 // ----------------------------------------------------------------------------
 334
 335 wxConvBrokenFileNames::wxConvBrokenFileNames(const wxChar *charset)
 336 {
 337     if ( !charset || wxStricmp(charset, _T("UTF-8")) == 0
 338                   || wxStricmp(charset, _T("UTF8")) == 0  )
 339         m_conv = new wxMBConvUTF8(wxMBConvUTF8::MAP_INVALID_UTF8_TO_OCTAL);
 340     else
 341         m_conv = new wxCSConv(charset);
 342 }
 343
 344 size_t
 345 wxConvBrokenFileNames::MB2WC(wchar_t *outputBuf,
 346                              const char *psz,
 347                              size_t outputSize) const
 348 {
 349     return m_conv->MB2WC( outputBuf, psz, outputSize );
 350 }
 351
 352 size_t
 353 wxConvBrokenFileNames::WC2MB(char *outputBuf,
 354                              const wchar_t *psz,
 355                              size_t outputSize) const
 356 {
 357     return m_conv->WC2MB( outputBuf, psz, outputSize );
 358 }
 359
 360 #endif
 361
 362 // ----------------------------------------------------------------------------
 363 // UTF-7
 364 // ----------------------------------------------------------------------------
 365
 366 // Implementation (C) 2004 Fredrik Roubert
 367
 368 //
 369 // BASE64 decoding table
 370 //
 371 static const unsigned char utf7unb64[] =
 372 {
 373     0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff,
 374     0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff,
 375     0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff,
 376     0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff,
 377     0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff,
 378     0xff, 0xff, 0xff, 0x3e, 0xff, 0xff, 0xff, 0x3f,
 379     0x34, 0x35, 0x36, 0x37, 0x38, 0x39, 0x3a, 0x3b,
 380     0x3c, 0x3d, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff,
 381     0xff, 0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06,
 382     0x07, 0x08, 0x09, 0x0a, 0x0b, 0x0c, 0x0d, 0x0e,
 383     0x0f, 0x10, 0x11, 0x12, 0x13, 0x14, 0x15, 0x16,
 384     0x17, 0x18, 0x19, 0xff, 0xff, 0xff, 0xff, 0xff,
 385     0xff, 0x1a, 0x1b, 0x1c, 0x1d, 0x1e, 0x1f, 0x20,
 386     0x21, 0x22, 0x23, 0x24, 0x25, 0x26, 0x27, 0x28,
 387     0x29, 0x2a, 0x2b, 0x2c, 0x2d, 0x2e, 0x2f, 0x30,
 388     0x31, 0x32, 0x33, 0xff, 0xff, 0xff, 0xff, 0xff,
 389     0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff,
 390     0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff,
 391     0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff,
 392     0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff,
 393     0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff,
 394     0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff,
 395     0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff,
 396     0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff,
 397     0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff,
 398     0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff,
 399     0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff,
 400     0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff,
 401     0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff,
 402     0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff,
 403     0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff,
 404     0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff
 405 };
 406
 407 size_t wxMBConvUTF7::MB2WC(wchar_t *buf, const char *psz, size_t n) const
 408 {
 409     size_t len = 0;
 410
 411     while (*psz && ((!buf) || (len < n)))
 412     {
 413         unsigned char cc = *psz++;
 414         if (cc != '+')
 415         {
 416             // plain ASCII char
 417             if (buf)
 418                 *buf++ = cc;
 419             len++;
 420         }
 421         else if (*psz == '-')
 422         {
 423             // encoded plus sign
 424             if (buf)
 425                 *buf++ = cc;
 426             len++;
 427             psz++;
 428         }
 429         else
 430         {
 431             // BASE64 encoded string
 432             bool lsb;
 433             unsigned char c;
 434             unsigned int d, l;
 435             for (lsb = false, d = 0, l = 0;
 436                 (cc = utf7unb64[(unsigned char)*psz]) != 0xff; psz++)
 437             {
 438                 d <<= 6;
 439                 d += cc;
 440                 for (l += 6; l >= 8; lsb = !lsb)
 441                 {
 442                     c = (unsigned char)((d >> (l -= 8)) % 256);
 443                     if (lsb)
 444                     {
 445                         if (buf)
 446                             *buf++ |= c;
 447                         len ++;
 448                     }
 449                     else
 450                         if (buf)
 451                             *buf = (wchar_t)(c << 8);
 452                 }
 453             }
 454             if (*psz == '-')
 455                 psz++;
 456         }
 457     }
 458     if (buf && (len < n))
 459         *buf = 0;
 460     return len;
 461 }
 462
 463 //
 464 // BASE64 encoding table
 465 //
 466 static const unsigned char utf7enb64[] =
 467 {
 468     'A', 'B', 'C', 'D', 'E', 'F', 'G', 'H',
 469     'I', 'J', 'K', 'L', 'M', 'N', 'O', 'P',
 470     'Q', 'R', 'S', 'T', 'U', 'V', 'W', 'X',
 471     'Y', 'Z', 'a', 'b', 'c', 'd', 'e', 'f',
 472     'g', 'h', 'i', 'j', 'k', 'l', 'm', 'n',
 473     'o', 'p', 'q', 'r', 's', 't', 'u', 'v',
 474     'w', 'x', 'y', 'z', '0', '1', '2', '3',
 475     '4', '5', '6', '7', '8', '9', '+', '/'
 476 };
 477
 478 //
 479 // UTF-7 encoding table
 480 //
 481 // 0 - Set D (directly encoded characters)
 482 // 1 - Set O (optional direct characters)
 483 // 2 - whitespace characters (optional)
 484 // 3 - special characters
 485 //
 486 static const unsigned char utf7encode[128] =
 487 {
 488     3, 3, 3, 3, 3, 3, 3, 3, 3, 2, 2, 3, 3, 2, 3, 3,
 489     3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3,
 490     2, 1, 1, 1, 1, 1, 1, 0, 0, 0, 1, 3, 0, 0, 0, 3,
 491     0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 0,
 492     1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
 493     0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 3, 1, 1, 1,
 494     1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
 495     0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 3, 3
 496 };
 497
 498 size_t wxMBConvUTF7::WC2MB(char *buf, const wchar_t *psz, size_t n) const
 499 {
 500
 501
 502     size_t len = 0;
 503
 504     while (*psz && ((!buf) || (len < n)))
 505     {
 506         wchar_t cc = *psz++;
 507         if (cc < 0x80 && utf7encode[cc] < 1)
 508         {
 509             // plain ASCII char
 510             if (buf)
 511                 *buf++ = (char)cc;
 512             len++;
 513         }
 514 #ifndef WC_UTF16
 515         else if (((wxUint32)cc) > 0xffff)
 516         {
 517             // no surrogate pair generation (yet?)
 518             return (size_t)-1;
 519         }
 520 #endif
 521         else
 522         {
 523             if (buf)
 524                 *buf++ = '+';
 525             len++;
 526             if (cc != '+')
 527             {
 528                 // BASE64 encode string
 529                 unsigned int lsb, d, l;
 530                 for (d = 0, l = 0;; psz++)
 531                 {
 532                     for (lsb = 0; lsb < 2; lsb ++)
 533                     {
 534                         d <<= 8;
 535                         d += lsb ? cc & 0xff : (cc & 0xff00) >> 8;
 536
 537                         for (l += 8; l >= 6; )
 538                         {
 539                             l -= 6;
 540                             if (buf)
 541                                 *buf++ = utf7enb64[(d >> l) % 64];
 542                             len++;
 543                         }
 544                     }
 545                     cc = *psz;
 546                     if (!(cc) || (cc < 0x80 && utf7encode[cc] < 1))
 547                         break;
 548                 }
 549                 if (l != 0)
 550                 {
 551                     if (buf)
 552                         *buf++ = utf7enb64[((d % 16) << (6 - l)) % 64];
 553                     len++;
 554                 }
 555             }
 556             if (buf)
 557                 *buf++ = '-';
 558             len++;
 559         }
 560     }
 561     if (buf && (len < n))
 562         *buf = 0;
 563     return len;
 564 }
 565
 566 // ----------------------------------------------------------------------------
 567 // UTF-8
 568 // ----------------------------------------------------------------------------
 569
 570 static wxUint32 utf8_max[]=
 571     { 0x7f, 0x7ff, 0xffff, 0x1fffff, 0x3ffffff, 0x7fffffff, 0xffffffff };
 572
 573 // boundaries of the private use area we use to (temporarily) remap invalid
 574 // characters invalid in a UTF-8 encoded string
 575 const wxUint32 wxUnicodePUA = 0x100000;
 576 const wxUint32 wxUnicodePUAEnd = wxUnicodePUA + 256;
 577
 578 size_t wxMBConvUTF8::MB2WC(wchar_t *buf, const char *psz, size_t n) const
 579 {
 580     size_t len = 0;
 581
 582     while (*psz && ((!buf) || (len < n)))
 583     {
 584         const char *opsz = psz;
 585         bool invalid = false;
 586         unsigned char cc = *psz++, fc = cc;
 587         unsigned cnt;
 588         for (cnt = 0; fc & 0x80; cnt++)
 589             fc <<= 1;
 590         if (!cnt)
 591         {
 592             // plain ASCII char
 593             if (buf)
 594                 *buf++ = cc;
 595             len++;
 596
 597             // escape the escape character for octal escapes
 598             if ((m_options & MAP_INVALID_UTF8_TO_OCTAL)
 599                     && cc == '\\' && (!buf || len < n))
 600             {
 601                 if (buf)
 602                     *buf++ = cc;
 603                 len++;
 604             }
 605         }
 606         else
 607         {
 608             cnt--;
 609             if (!cnt)
 610             {
 611                 // invalid UTF-8 sequence
 612                 invalid = true;
 613             }
 614             else
 615             {
 616                 unsigned ocnt = cnt - 1;
 617                 wxUint32 res = cc & (0x3f >> cnt);
 618                 while (cnt--)
 619                 {
 620                     cc = *psz;
 621                     if ((cc & 0xC0) != 0x80)
 622                     {
 623                         // invalid UTF-8 sequence
 624                         invalid = true;
 625                         break;
 626                     }
 627                     psz++;
 628                     res = (res << 6) | (cc & 0x3f);
 629                 }
 630                 if (invalid || res <= utf8_max[ocnt])
 631                 {
 632                     // illegal UTF-8 encoding
 633                     invalid = true;
 634                 }
 635                 else if ((m_options & MAP_INVALID_UTF8_TO_PUA) &&
 636                         res >= wxUnicodePUA && res < wxUnicodePUAEnd)
 637                 {
 638                     // if one of our PUA characters turns up externally
 639                     // it must also be treated as an illegal sequence
 640                     // (a bit like you have to escape an escape character)
 641                     invalid = true;
 642                 }
 643                 else
 644                 {
 645 #ifdef WC_UTF16
 646                     // cast is ok because wchar_t == wxUuint16 if WC_UTF16
 647                     size_t pa = encode_utf16(res, (wxUint16 *)buf);
 648                     if (pa == (size_t)-1)
 649                     {
 650                         invalid = true;
 651                     }
 652                     else
 653                     {
 654                         if (buf)
 655                             buf += pa;
 656                         len += pa;
 657                     }
 658 #else // !WC_UTF16
 659                     if (buf)
 660                         *buf++ = res;
 661                     len++;
 662 #endif // WC_UTF16/!WC_UTF16
 663                 }
 664             }
 665             if (invalid)
 666             {
 667                 if (m_options & MAP_INVALID_UTF8_TO_PUA)
 668                 {
 669                     while (opsz < psz && (!buf || len < n))
 670                     {
 671 #ifdef WC_UTF16
 672                         // cast is ok because wchar_t == wxUuint16 if WC_UTF16
 673                         size_t pa = encode_utf16((unsigned char)*opsz + wxUnicodePUA, (wxUint16 *)buf);
 674                         wxASSERT(pa != (size_t)-1);
 675                         if (buf)
 676                             buf += pa;
 677                         opsz++;
 678                         len += pa;
 679 #else
 680                         if (buf)
 681                             *buf++ = wxUnicodePUA + (unsigned char)*opsz;
 682                         opsz++;
 683                         len++;
 684 #endif
 685                     }
 686                 }
 687                 else if (m_options & MAP_INVALID_UTF8_TO_OCTAL)
 688                 {
 689                     while (opsz < psz && (!buf || len < n))
 690                     {
 691                         if ( buf && len + 3 < n )
 692                         {
 693                             unsigned char n = *opsz;
 694                             *buf++ = L'\\';
 695                             *buf++ = (wchar_t)( L'0' + n / 0100 );
 696                             *buf++ = (wchar_t)( L'0' + (n % 0100) / 010 );
 697                             *buf++ = (wchar_t)( L'0' + n % 010 );
 698                         }
 699                         opsz++;
 700                         len += 4;
 701                     }
 702                 }
 703                 else // MAP_INVALID_UTF8_NOT
 704                 {
 705                     return (size_t)-1;
 706                 }
 707             }
 708         }
 709     }
 710     if (buf && (len < n))
 711         *buf = 0;
 712     return len;
 713 }
 714
 715 static inline bool isoctal(wchar_t wch)
 716 {
 717     return L'0' <= wch && wch <= L'7';
 718 }
 719
 720 size_t wxMBConvUTF8::WC2MB(char *buf, const wchar_t *psz, size_t n) const
 721 {
 722     size_t len = 0;
 723
 724     while (*psz && ((!buf) || (len < n)))
 725     {
 726         wxUint32 cc;
 727 #ifdef WC_UTF16
 728         // cast is ok for WC_UTF16
 729         size_t pa = decode_utf16((const wxUint16 *)psz, cc);
 730         psz += (pa == (size_t)-1) ? 1 : pa;
 731 #else
 732         cc=(*psz++) & 0x7fffffff;
 733 #endif
 734
 735         if ( (m_options & MAP_INVALID_UTF8_TO_PUA)
 736                 && cc >= wxUnicodePUA && cc < wxUnicodePUAEnd )
 737         {
 738             if (buf)
 739                 *buf++ = (char)(cc - wxUnicodePUA);
 740             len++;
 741         }
 742         else if ( (m_options & MAP_INVALID_UTF8_TO_OCTAL)
 743                     && cc == L'\\' && psz[0] == L'\\' )
 744         {
 745             if (buf)
 746                 *buf++ = (char)cc;
 747             psz++;
 748             len++;
 749         }
 750         else if ( (m_options & MAP_INVALID_UTF8_TO_OCTAL) &&
 751                     cc == L'\\' &&
 752                         isoctal(psz[0]) && isoctal(psz[1]) && isoctal(psz[2]) )
 753         {
 754             if (buf)
 755             {
 756                 *buf++ = (char) ((psz[0] - L'0')*0100 +
 757                                  (psz[1] - L'0')*010 +
 758                                  (psz[2] - L'0'));
 759             }
 760
 761             psz += 3;
 762             len++;
 763         }
 764         else
 765         {
 766             unsigned cnt;
 767             for (cnt = 0; cc > utf8_max[cnt]; cnt++) {}
 768             if (!cnt)
 769             {
 770                 // plain ASCII char
 771                 if (buf)
 772                     *buf++ = (char) cc;
 773                 len++;
 774             }
 775
 776             else
 777             {
 778                 len += cnt + 1;
 779                 if (buf)
 780                 {
 781                     *buf++ = (char) ((-128 >> cnt) | ((cc >> (cnt * 6)) & (0x3f >> cnt)));
 782                     while (cnt--)
 783                         *buf++ = (char) (0x80 | ((cc >> (cnt * 6)) & 0x3f));
 784                 }
 785             }
 786         }
 787     }
 788
 789     if (buf && (len<n))
 790         *buf = 0;
 791
 792     return len;
 793 }
 794
 795 // ----------------------------------------------------------------------------
 796 // UTF-16
 797 // ----------------------------------------------------------------------------
 798
 799 #ifdef WORDS_BIGENDIAN
 800     #define wxMBConvUTF16straight wxMBConvUTF16BE
 801     #define wxMBConvUTF16swap     wxMBConvUTF16LE
 802 #else
 803     #define wxMBConvUTF16swap     wxMBConvUTF16BE
 804     #define wxMBConvUTF16straight wxMBConvUTF16LE
 805 #endif
 806
 807
 808 #ifdef WC_UTF16
 809
 810 // copy 16bit MB to 16bit String
 811 size_t wxMBConvUTF16straight::MB2WC(wchar_t *buf, const char *psz, size_t n) const
 812 {
 813     size_t len=0;
 814
 815     while (*(wxUint16*)psz && (!buf || len < n))
 816     {
 817         if (buf)
 818             *buf++ = *(wxUint16*)psz;
 819         len++;
 820
 821         psz += sizeof(wxUint16);
 822     }
 823     if (buf && len<n)   *buf=0;
 824
 825     return len;
 826 }
 827
 828
 829 // copy 16bit String to 16bit MB
 830 size_t wxMBConvUTF16straight::WC2MB(char *buf, const wchar_t *psz, size_t n) const
 831 {
 832     size_t len=0;
 833
 834     while (*psz && (!buf || len < n))
 835     {
 836         if (buf)
 837         {
 838             *(wxUint16*)buf = *psz;
 839             buf += sizeof(wxUint16);
 840         }
 841         len += sizeof(wxUint16);
 842         psz++;
 843     }
 844     if (buf && len<=n-sizeof(wxUint16))   *(wxUint16*)buf=0;
 845
 846     return len;
 847 }
 848
 849
 850 // swap 16bit MB to 16bit String
 851 size_t wxMBConvUTF16swap::MB2WC(wchar_t *buf, const char *psz, size_t n) const
 852 {
 853     size_t len=0;
 854
 855     while (*(wxUint16*)psz && (!buf || len < n))
 856     {
 857         if (buf)
 858         {
 859             ((char *)buf)[0] = psz[1];
 860             ((char *)buf)[1] = psz[0];
 861             buf++;
 862         }
 863         len++;
 864         psz += sizeof(wxUint16);
 865     }
 866     if (buf && len<n)   *buf=0;
 867
 868     return len;
 869 }
 870
 871
 872 // swap 16bit MB to 16bit String
 873 size_t wxMBConvUTF16swap::WC2MB(char *buf, const wchar_t *psz, size_t n) const
 874 {
 875     size_t len=0;
 876
 877     while (*psz && (!buf || len < n))
 878     {
 879         if (buf)
 880         {
 881             *buf++ = ((char*)psz)[1];
 882             *buf++ = ((char*)psz)[0];
 883         }
 884         len += sizeof(wxUint16);
 885         psz++;
 886     }
 887     if (buf && len<=n-sizeof(wxUint16))   *(wxUint16*)buf=0;
 888
 889     return len;
 890 }
 891
 892
 893 #else // WC_UTF16
 894
 895
 896 // copy 16bit MB to 32bit String
 897 size_t wxMBConvUTF16straight::MB2WC(wchar_t *buf, const char *psz, size_t n) const
 898 {
 899     size_t len=0;
 900
 901     while (*(wxUint16*)psz && (!buf || len < n))
 902     {
 903         wxUint32 cc;
 904         size_t pa=decode_utf16((wxUint16*)psz, cc);
 905         if (pa == (size_t)-1)
 906             return pa;
 907
 908         if (buf)
 909             *buf++ = cc;
 910         len++;
 911         psz += pa * sizeof(wxUint16);
 912     }
 913     if (buf && len<n)   *buf=0;
 914
 915     return len;
 916 }
 917
 918
 919 // copy 32bit String to 16bit MB
 920 size_t wxMBConvUTF16straight::WC2MB(char *buf, const wchar_t *psz, size_t n) const
 921 {
 922     size_t len=0;
 923
 924     while (*psz && (!buf || len < n))
 925     {
 926         wxUint16 cc[2];
 927         size_t pa=encode_utf16(*psz, cc);
 928
 929         if (pa == (size_t)-1)
 930             return pa;
 931
 932         if (buf)
 933         {
 934             *(wxUint16*)buf = cc[0];
 935             buf += sizeof(wxUint16);
 936             if (pa > 1)
 937             {
 938                 *(wxUint16*)buf = cc[1];
 939                 buf += sizeof(wxUint16);
 940             }
 941         }
 942
 943         len += pa*sizeof(wxUint16);
 944         psz++;
 945     }
 946     if (buf && len<=n-sizeof(wxUint16))   *(wxUint16*)buf=0;
 947
 948     return len;
 949 }
 950
 951
 952 // swap 16bit MB to 32bit String
 953 size_t wxMBConvUTF16swap::MB2WC(wchar_t *buf, const char *psz, size_t n) const
 954 {
 955     size_t len=0;
 956
 957     while (*(wxUint16*)psz && (!buf || len < n))
 958     {
 959         wxUint32 cc;
 960         char tmp[4];
 961         tmp[0]=psz[1];  tmp[1]=psz[0];
 962         tmp[2]=psz[3];  tmp[3]=psz[2];
 963
 964         size_t pa=decode_utf16((wxUint16*)tmp, cc);
 965         if (pa == (size_t)-1)
 966             return pa;
 967
 968         if (buf)
 969             *buf++ = cc;
 970
 971         len++;
 972         psz += pa * sizeof(wxUint16);
 973     }
 974     if (buf && len<n)   *buf=0;
 975
 976     return len;
 977 }
 978
 979
 980 // swap 32bit String to 16bit MB
 981 size_t wxMBConvUTF16swap::WC2MB(char *buf, const wchar_t *psz, size_t n) const
 982 {
 983     size_t len=0;
 984
 985     while (*psz && (!buf || len < n))
 986     {
 987         wxUint16 cc[2];
 988         size_t pa=encode_utf16(*psz, cc);
 989
 990         if (pa == (size_t)-1)
 991             return pa;
 992
 993         if (buf)
 994         {
 995             *buf++ = ((char*)cc)[1];
 996             *buf++ = ((char*)cc)[0];
 997             if (pa > 1)
 998             {
 999                 *buf++ = ((char*)cc)[3];
1000                 *buf++ = ((char*)cc)[2];
1001             }
1002         }
1003
1004         len += pa*sizeof(wxUint16);
1005         psz++;
1006     }
1007     if (buf && len<=n-sizeof(wxUint16))   *(wxUint16*)buf=0;
1008
1009     return len;
1010 }
1011
1012 #endif // WC_UTF16
1013
1014
1015 // ----------------------------------------------------------------------------
1016 // UTF-32
1017 // ----------------------------------------------------------------------------
1018
1019 #ifdef WORDS_BIGENDIAN
1020 #define wxMBConvUTF32straight  wxMBConvUTF32BE
1021 #define wxMBConvUTF32swap      wxMBConvUTF32LE
1022 #else
1023 #define wxMBConvUTF32swap      wxMBConvUTF32BE
1024 #define wxMBConvUTF32straight  wxMBConvUTF32LE
1025 #endif
1026
1027
1028 WXDLLIMPEXP_DATA_BASE(wxMBConvUTF32LE) wxConvUTF32LE;
1029 WXDLLIMPEXP_DATA_BASE(wxMBConvUTF32BE) wxConvUTF32BE;
1030
1031
1032 #ifdef WC_UTF16
1033
1034 // copy 32bit MB to 16bit String
1035 size_t wxMBConvUTF32straight::MB2WC(wchar_t *buf, const char *psz, size_t n) const
1036 {
1037     size_t len=0;
1038
1039     while (*(wxUint32*)psz && (!buf || len < n))
1040     {
1041         wxUint16 cc[2];
1042
1043         size_t pa=encode_utf16(*(wxUint32*)psz, cc);
1044         if (pa == (size_t)-1)
1045             return pa;
1046
1047         if (buf)
1048         {
1049             *buf++ = cc[0];
1050             if (pa > 1)
1051                 *buf++ = cc[1];
1052         }
1053         len += pa;
1054         psz += sizeof(wxUint32);
1055     }
1056     if (buf && len<n)   *buf=0;
1057
1058     return len;
1059 }
1060
1061
1062 // copy 16bit String to 32bit MB
1063 size_t wxMBConvUTF32straight::WC2MB(char *buf, const wchar_t *psz, size_t n) const
1064 {
1065     size_t len=0;
1066
1067     while (*psz && (!buf || len < n))
1068     {
1069         wxUint32 cc;
1070
1071         // cast is ok for WC_UTF16
1072         size_t pa = decode_utf16((const wxUint16 *)psz, cc);
1073         if (pa == (size_t)-1)
1074             return pa;
1075
1076         if (buf)
1077         {
1078             *(wxUint32*)buf = cc;
1079             buf += sizeof(wxUint32);
1080         }
1081         len += sizeof(wxUint32);
1082         psz += pa;
1083     }
1084
1085     if (buf && len<=n-sizeof(wxUint32))
1086         *(wxUint32*)buf=0;
1087
1088     return len;
1089 }
1090
1091
1092
1093 // swap 32bit MB to 16bit String
1094 size_t wxMBConvUTF32swap::MB2WC(wchar_t *buf, const char *psz, size_t n) const
1095 {
1096     size_t len=0;
1097
1098     while (*(wxUint32*)psz && (!buf || len < n))
1099     {
1100         char tmp[4];
1101         tmp[0] = psz[3];   tmp[1] = psz[2];
1102         tmp[2] = psz[1];   tmp[3] = psz[0];
1103
1104
1105         wxUint16 cc[2];
1106
1107         size_t pa=encode_utf16(*(wxUint32*)tmp, cc);
1108         if (pa == (size_t)-1)
1109             return pa;
1110
1111         if (buf)
1112         {
1113             *buf++ = cc[0];
1114             if (pa > 1)
1115                 *buf++ = cc[1];
1116         }
1117         len += pa;
1118         psz += sizeof(wxUint32);
1119     }
1120
1121     if (buf && len<n)
1122         *buf=0;
1123
1124     return len;
1125 }
1126
1127
1128 // swap 16bit String to 32bit MB
1129 size_t wxMBConvUTF32swap::WC2MB(char *buf, const wchar_t *psz, size_t n) const
1130 {
1131     size_t len=0;
1132
1133     while (*psz && (!buf || len < n))
1134     {
1135         char cc[4];
1136
1137         // cast is ok for WC_UTF16
1138         size_t pa=decode_utf16((const wxUint16 *)psz, *(wxUint32*)cc);
1139         if (pa == (size_t)-1)
1140             return pa;
1141
1142         if (buf)
1143         {
1144             *buf++ = cc[3];
1145             *buf++ = cc[2];
1146             *buf++ = cc[1];
1147             *buf++ = cc[0];
1148         }
1149         len += sizeof(wxUint32);
1150         psz += pa;
1151     }
1152
1153     if (buf && len<=n-sizeof(wxUint32))
1154         *(wxUint32*)buf=0;
1155
1156     return len;
1157 }
1158
1159 #else // WC_UTF16
1160
1161
1162 // copy 32bit MB to 32bit String
1163 size_t wxMBConvUTF32straight::MB2WC(wchar_t *buf, const char *psz, size_t n) const
1164 {
1165     size_t len=0;
1166
1167     while (*(wxUint32*)psz && (!buf || len < n))
1168     {
1169         if (buf)
1170             *buf++ = *(wxUint32*)psz;
1171         len++;
1172         psz += sizeof(wxUint32);
1173     }
1174
1175     if (buf && len<n)
1176         *buf=0;
1177
1178     return len;
1179 }
1180
1181
1182 // copy 32bit String to 32bit MB
1183 size_t wxMBConvUTF32straight::WC2MB(char *buf, const wchar_t *psz, size_t n) const
1184 {
1185     size_t len=0;
1186
1187     while (*psz && (!buf || len < n))
1188     {
1189         if (buf)
1190         {
1191             *(wxUint32*)buf = *psz;
1192             buf += sizeof(wxUint32);
1193         }
1194
1195         len += sizeof(wxUint32);
1196         psz++;
1197     }
1198
1199     if (buf && len<=n-sizeof(wxUint32))
1200         *(wxUint32*)buf=0;
1201
1202     return len;
1203 }
1204
1205
1206 // swap 32bit MB to 32bit String
1207 size_t wxMBConvUTF32swap::MB2WC(wchar_t *buf, const char *psz, size_t n) const
1208 {
1209     size_t len=0;
1210
1211     while (*(wxUint32*)psz && (!buf || len < n))
1212     {
1213         if (buf)
1214         {
1215             ((char *)buf)[0] = psz[3];
1216             ((char *)buf)[1] = psz[2];
1217             ((char *)buf)[2] = psz[1];
1218             ((char *)buf)[3] = psz[0];
1219             buf++;
1220         }
1221         len++;
1222         psz += sizeof(wxUint32);
1223     }
1224
1225     if (buf && len<n)
1226         *buf=0;
1227
1228     return len;
1229 }
1230
1231
1232 // swap 32bit String to 32bit MB
1233 size_t wxMBConvUTF32swap::WC2MB(char *buf, const wchar_t *psz, size_t n) const
1234 {
1235     size_t len=0;
1236
1237     while (*psz && (!buf || len < n))
1238     {
1239         if (buf)
1240         {
1241             *buf++ = ((char *)psz)[3];
1242             *buf++ = ((char *)psz)[2];
1243             *buf++ = ((char *)psz)[1];
1244             *buf++ = ((char *)psz)[0];
1245         }
1246         len += sizeof(wxUint32);
1247         psz++;
1248     }
1249
1250     if (buf && len<=n-sizeof(wxUint32))
1251         *(wxUint32*)buf=0;
1252
1253     return len;
1254 }
1255
1256
1257 #endif // WC_UTF16
1258
1259
1260 // ============================================================================
1261 // The classes doing conversion using the iconv_xxx() functions
1262 // ============================================================================
1263
1264 #ifdef HAVE_ICONV
1265
1266 // VS: glibc 2.1.3 is broken in that iconv() conversion to/from UCS4 fails with
1267 //     E2BIG if output buffer is _exactly_ as big as needed. Such case is
1268 //     (unless there's yet another bug in glibc) the only case when iconv()
1269 //     returns with (size_t)-1 (which means error) and says there are 0 bytes
1270 //     left in the input buffer -- when _real_ error occurs,
1271 //     bytes-left-in-input buffer is non-zero. Hence, this alternative test for
1272 //     iconv() failure.
1273 //     [This bug does not appear in glibc 2.2.]
1274 #if defined(__GLIBC__) && __GLIBC__ == 2 && __GLIBC_MINOR__ <= 1
1275 #define ICONV_FAILED(cres, bufLeft) ((cres == (size_t)-1) && \
1276                                      (errno != E2BIG || bufLeft != 0))
1277 #else
1278 #define ICONV_FAILED(cres, bufLeft)  (cres == (size_t)-1)
1279 #endif
1280
1281 #define ICONV_CHAR_CAST(x)  ((ICONV_CONST char **)(x))
1282
1283 #define ICONV_T_INVALID ((iconv_t)-1)
1284
1285 #if SIZEOF_WCHAR_T == 4
1286     #define WC_BSWAP    wxUINT32_SWAP_ALWAYS
1287     #define WC_ENC      wxFONTENCODING_UTF32
1288 #elif SIZEOF_WCHAR_T == 2
1289     #define WC_BSWAP    wxUINT16_SWAP_ALWAYS
1290     #define WC_ENC      wxFONTENCODING_UTF16
1291 #else // sizeof(wchar_t) != 2 nor 4
1292     // does this ever happen?
1293     #error "Unknown sizeof(wchar_t): please report this to wx-dev@lists.wxwindows.org"
1294 #endif
1295
1296 // ----------------------------------------------------------------------------
1297 // wxMBConv_iconv: encapsulates an iconv character set
1298 // ----------------------------------------------------------------------------
1299
1300 class wxMBConv_iconv : public wxMBConv
1301 {
1302 public:
1303     wxMBConv_iconv(const wxChar *name);
1304     virtual ~wxMBConv_iconv();
1305
1306     virtual size_t MB2WC(wchar_t *buf, const char *psz, size_t n) const;
1307     virtual size_t WC2MB(char *buf, const wchar_t *psz, size_t n) const;
1308
1309     bool IsOk() const
1310         { return (m2w != ICONV_T_INVALID) && (w2m != ICONV_T_INVALID); }
1311
1312 protected:
1313     // the iconv handlers used to translate from multibyte to wide char and in
1314     // the other direction
1315     iconv_t m2w,
1316             w2m;
1317 #if wxUSE_THREADS
1318     // guards access to m2w and w2m objects
1319     wxMutex m_iconvMutex;
1320 #endif
1321
1322 private:
1323     // the name (for iconv_open()) of a wide char charset -- if none is
1324     // available on this machine, it will remain NULL
1325     static wxString ms_wcCharsetName;
1326
1327     // true if the wide char encoding we use (i.e. ms_wcCharsetName) has
1328     // different endian-ness than the native one
1329     static bool ms_wcNeedsSwap;
1330 };
1331
1332 // make the constructor available for unit testing
1333 WXDLLIMPEXP_BASE wxMBConv* new_wxMBConv_iconv( const wxChar* name )
1334 {
1335     wxMBConv_iconv* result = new wxMBConv_iconv( name );
1336     if ( !result->IsOk() )
1337     {
1338         delete result;
1339         return 0;
1340     }
1341     return result;
1342 }
1343
1344 wxString wxMBConv_iconv::ms_wcCharsetName = NULL;
1345 bool wxMBConv_iconv::ms_wcNeedsSwap = false;
1346
1347 wxMBConv_iconv::wxMBConv_iconv(const wxChar *name)
1348 {
1349     // iconv operates with chars, not wxChars, but luckily it uses only ASCII
1350     // names for the charsets
1351     const wxCharBuffer cname(wxString(name).ToAscii());
1352
1353     // check for charset that represents wchar_t:
1354     if ( ms_wcCharsetName.empty() )
1355     {
1356 #if wxUSE_FONTMAP
1357         const wxChar **names = wxFontMapperBase::GetAllEncodingNames(WC_ENC);
1358 #else // !wxUSE_FONTMAP
1359         static const wxChar *names[] =
1360         {
1361 #if SIZEOF_WCHAR_T == 4
1362             _T("UCS-4"),
1363 #elif SIZEOF_WCHAR_T = 2
1364             _T("UCS-2"),
1365 #endif
1366             NULL
1367         };
1368 #endif // wxUSE_FONTMAP/!wxUSE_FONTMAP
1369
1370         for ( ; *names; ++names )
1371         {
1372             const wxString name(*names);
1373
1374             // first try charset with explicit bytesex info (e.g. "UCS-4LE"):
1375             wxString nameXE(name);
1376             #ifdef WORDS_BIGENDIAN
1377                 nameXE += _T("BE");
1378             #else // little endian
1379                 nameXE += _T("LE");
1380             #endif
1381
1382             m2w = iconv_open(nameXE.ToAscii(), cname);
1383             if ( m2w == ICONV_T_INVALID )
1384             {
1385                 // try charset w/o bytesex info (e.g. "UCS4")
1386                 m2w = iconv_open(name.ToAscii(), cname);
1387
1388                 // and check for bytesex ourselves:
1389                 if ( m2w != ICONV_T_INVALID )
1390                 {
1391                     char    buf[2], *bufPtr;
1392                     wchar_t wbuf[2], *wbufPtr;
1393                     size_t  insz, outsz;
1394                     size_t  res;
1395
1396                     buf[0] = 'A';
1397                     buf[1] = 0;
1398                     wbuf[0] = 0;
1399                     insz = 2;
1400                     outsz = SIZEOF_WCHAR_T * 2;
1401                     wbufPtr = wbuf;
1402                     bufPtr = buf;
1403
1404                     res = iconv(m2w, ICONV_CHAR_CAST(&bufPtr), &insz,
1405                                 (char**)&wbufPtr, &outsz);
1406
1407                     if (ICONV_FAILED(res, insz))
1408                     {
1409                         wxLogLastError(wxT("iconv"));
1410                         wxLogError(_("Conversion to charset '%s' doesn't work."), name);
1411                     }
1412                     else // ok, can convert to this encoding, remember it
1413                     {
1414                         ms_wcCharsetName = name;
1415                         ms_wcNeedsSwap = wbuf[0] != (wchar_t)buf[0];
1416                     }
1417                 }
1418             }
1419             else // use charset not requiring byte swapping
1420             {
1421                 ms_wcCharsetName = nameXE;
1422             }
1423         }
1424
1425         wxLogTrace(TRACE_STRCONV,
1426                    wxT("iconv wchar_t charset is \"%s\"%s"),
1427                    ms_wcCharsetName.empty() ? "<none>"
1428                                             : ms_wcCharsetName.c_str(),
1429                    ms_wcNeedsSwap ? _T(" (needs swap)")
1430                                   : _T(""));
1431     }
1432     else // we already have ms_wcCharsetName
1433     {
1434         m2w = iconv_open(ms_wcCharsetName.ToAscii(), cname);
1435     }
1436
1437     if ( ms_wcCharsetName.empty() )
1438     {
1439         w2m = ICONV_T_INVALID;
1440     }
1441     else
1442     {
1443         w2m = iconv_open(cname, ms_wcCharsetName.ToAscii());
1444         if ( w2m == ICONV_T_INVALID )
1445         {
1446             wxLogTrace(TRACE_STRCONV,
1447                        wxT("\"%s\" -> \"%s\" works but not the converse!?"),
1448                        ms_wcCharsetName.c_str(), cname);
1449         }
1450     }
1451 }
1452
1453 wxMBConv_iconv::~wxMBConv_iconv()
1454 {
1455     if ( m2w != ICONV_T_INVALID )
1456         iconv_close(m2w);
1457     if ( w2m != ICONV_T_INVALID )
1458         iconv_close(w2m);
1459 }
1460
1461 size_t wxMBConv_iconv::MB2WC(wchar_t *buf, const char *psz, size_t n) const
1462 {
1463 #if wxUSE_THREADS
1464     // NB: iconv() is MT-safe, but each thread must use it's own iconv_t handle.
1465     //     Unfortunately there is a couple of global wxCSConv objects such as
1466     //     wxConvLocal that are used all over wx code, so we have to make sure
1467     //     the handle is used by at most one thread at the time. Otherwise
1468     //     only a few wx classes would be safe to use from non-main threads
1469     //     as MB<->WC conversion would fail "randomly".
1470     wxMutexLocker lock(wxConstCast(this, wxMBConv_iconv)->m_iconvMutex);
1471 #endif
1472
1473     size_t inbuf = strlen(psz);
1474     size_t outbuf = n * SIZEOF_WCHAR_T;
1475     size_t res, cres;
1476     // VS: Use these instead of psz, buf because iconv() modifies its arguments:
1477     wchar_t *bufPtr = buf;
1478     const char *pszPtr = psz;
1479
1480     if (buf)
1481     {
1482         // have destination buffer, convert there
1483         cres = iconv(m2w,
1484                      ICONV_CHAR_CAST(&pszPtr), &inbuf,
1485                      (char**)&bufPtr, &outbuf);
1486         res = n - (outbuf / SIZEOF_WCHAR_T);
1487
1488         if (ms_wcNeedsSwap)
1489         {
1490             // convert to native endianness
1491             for ( unsigned n = 0; n < res; n++ )
1492                 buf[n] = WC_BSWAP(buf[n]);
1493         }
1494
1495         // NB: iconv was given only strlen(psz) characters on input, and so
1496         //     it couldn't convert the trailing zero. Let's do it ourselves
1497         //     if there's some room left for it in the output buffer.
1498         if (res < n)
1499             buf[res] = 0;
1500     }
1501     else
1502     {
1503         // no destination buffer... convert using temp buffer
1504         // to calculate destination buffer requirement
1505         wchar_t tbuf[8];
1506         res = 0;
1507         do {
1508             bufPtr = tbuf;
1509             outbuf = 8*SIZEOF_WCHAR_T;
1510
1511             cres = iconv(m2w,
1512                          ICONV_CHAR_CAST(&pszPtr), &inbuf,
1513                          (char**)&bufPtr, &outbuf );
1514
1515             res += 8-(outbuf/SIZEOF_WCHAR_T);
1516         } while ((cres==(size_t)-1) && (errno==E2BIG));
1517     }
1518
1519     if (ICONV_FAILED(cres, inbuf))
1520     {
1521         //VS: it is ok if iconv fails, hence trace only
1522         wxLogTrace(TRACE_STRCONV, wxT("iconv failed: %s"), wxSysErrorMsg(wxSysErrorCode()));
1523         return (size_t)-1;
1524     }
1525
1526     return res;
1527 }
1528
1529 size_t wxMBConv_iconv::WC2MB(char *buf, const wchar_t *psz, size_t n) const
1530 {
1531 #if wxUSE_THREADS
1532     // NB: explained in MB2WC
1533     wxMutexLocker lock(wxConstCast(this, wxMBConv_iconv)->m_iconvMutex);
1534 #endif
1535
1536     size_t inbuf = wxWcslen(psz) * SIZEOF_WCHAR_T;
1537     size_t outbuf = n;
1538     size_t res, cres;
1539
1540     wchar_t *tmpbuf = 0;
1541
1542     if (ms_wcNeedsSwap)
1543     {
1544         // need to copy to temp buffer to switch endianness
1545         // (doing WC_BSWAP twice on the original buffer won't help, as it
1546         //  could be in read-only memory, or be accessed in some other thread)
1547         tmpbuf = (wchar_t *)malloc(inbuf + SIZEOF_WCHAR_T);
1548         for ( size_t n = 0; n < inbuf; n++ )
1549             tmpbuf[n] = WC_BSWAP(psz[n]);
1550         tmpbuf[inbuf] = L'\0';
1551         psz = tmpbuf;
1552     }
1553
1554     if (buf)
1555     {
1556         // have destination buffer, convert there
1557         cres = iconv( w2m, ICONV_CHAR_CAST(&psz), &inbuf, &buf, &outbuf );
1558
1559         res = n-outbuf;
1560
1561         // NB: iconv was given only wcslen(psz) characters on input, and so
1562         //     it couldn't convert the trailing zero. Let's do it ourselves
1563         //     if there's some room left for it in the output buffer.
1564         if (res < n)
1565             buf[0] = 0;
1566     }
1567     else
1568     {
1569         // no destination buffer... convert using temp buffer
1570         // to calculate destination buffer requirement
1571         char tbuf[16];
1572         res = 0;
1573         do {
1574             buf = tbuf; outbuf = 16;
1575
1576             cres = iconv( w2m, ICONV_CHAR_CAST(&psz), &inbuf, &buf, &outbuf );
1577
1578             res += 16 - outbuf;
1579         } while ((cres==(size_t)-1) && (errno==E2BIG));
1580     }
1581
1582     if (ms_wcNeedsSwap)
1583     {
1584         free(tmpbuf);
1585     }
1586
1587     if (ICONV_FAILED(cres, inbuf))
1588     {
1589         //VS: it is ok if iconv fails, hence trace only
1590         wxLogTrace(TRACE_STRCONV, wxT("iconv failed: %s"), wxSysErrorMsg(wxSysErrorCode()));
1591         return (size_t)-1;
1592     }
1593
1594     return res;
1595 }
1596
1597 #endif // HAVE_ICONV
1598
1599
1600 // ============================================================================
1601 // Win32 conversion classes
1602 // ============================================================================
1603
1604 #ifdef wxHAVE_WIN32_MB2WC
1605
1606 // from utils.cpp
1607 #if wxUSE_FONTMAP
1608 extern WXDLLIMPEXP_BASE long wxCharsetToCodepage(const wxChar *charset);
1609 extern WXDLLIMPEXP_BASE long wxEncodingToCodepage(wxFontEncoding encoding);
1610 #endif
1611
1612 class wxMBConv_win32 : public wxMBConv
1613 {
1614 public:
1615     wxMBConv_win32()
1616     {
1617         m_CodePage = CP_ACP;
1618     }
1619
1620 #if wxUSE_FONTMAP
1621     wxMBConv_win32(const wxChar* name)
1622     {
1623         m_CodePage = wxCharsetToCodepage(name);
1624     }
1625
1626     wxMBConv_win32(wxFontEncoding encoding)
1627     {
1628         m_CodePage = wxEncodingToCodepage(encoding);
1629     }
1630 #endif
1631
1632     size_t MB2WC(wchar_t *buf, const char *psz, size_t n) const
1633     {
1634         // note that we have to use MB_ERR_INVALID_CHARS flag as it without it
1635         // the behaviour is not compatible with the Unix version (using iconv)
1636         // and break the library itself, e.g. wxTextInputStream::NextChar()
1637         // wouldn't work if reading an incomplete MB char didn't result in an
1638         // error
1639         //
1640         // note however that using MB_ERR_INVALID_CHARS with CP_UTF7 results in
1641         // an error (tested under Windows Server 2003) and apparently it is
1642         // done on purpose, i.e. the function accepts any input in this case
1643         // and although I'd prefer to return error on ill-formed output, our
1644         // own wxMBConvUTF7 doesn't detect errors (e.g. lone "+" which is
1645         // explicitly ill-formed according to RFC 2152) neither so we don't
1646         // even have any fallback here...
1647         int flags = m_CodePage == CP_UTF7 ? 0 : MB_ERR_INVALID_CHARS;
1648
1649         const size_t len = ::MultiByteToWideChar
1650                              (
1651                                 m_CodePage,     // code page
1652                                 flags,          // flags: fall on error
1653                                 psz,            // input string
1654                                 -1,             // its length (NUL-terminated)
1655                                 buf,            // output string
1656                                 buf ? n : 0     // size of output buffer
1657                              );
1658
1659         // note that it returns count of written chars for buf != NULL and size
1660         // of the needed buffer for buf == NULL so in either case the length of
1661         // the string (which never includes the terminating NUL) is one less
1662         return len ? len - 1 : (size_t)-1;
1663     }
1664
1665     size_t WC2MB(char *buf, const wchar_t *pwz, size_t n) const
1666     {
1667         /*
1668             we have a problem here: by default, WideCharToMultiByte() may
1669             replace characters unrepresentable in the target code page with bad
1670             quality approximations such as turning "1/2" symbol (U+00BD) into
1671             "1" for the code pages which don't have it and we, obviously, want
1672             to avoid this at any price
1673
1674             the trouble is that this function does it _silently_, i.e. it won't
1675             even tell us whether it did or not... Win98/2000 and higher provide
1676             WC_NO_BEST_FIT_CHARS but it doesn't work for the older systems and
1677             we have to resort to a round trip, i.e. check that converting back
1678             results in the same string -- this is, of course, expensive but
1679             otherwise we simply can't be sure to not garble the data.
1680          */
1681
1682         // determine if we can rely on WC_NO_BEST_FIT_CHARS: according to MSDN
1683         // it doesn't work with CJK encodings (which we test for rather roughly
1684         // here...) nor with UTF-7/8 nor, of course, with Windows versions not
1685         // supporting it
1686         BOOL usedDef wxDUMMY_INITIALIZE(false);
1687         BOOL *pUsedDef;
1688         int flags;
1689         if ( CanUseNoBestFit() && m_CodePage < 50000 )
1690         {
1691             // it's our lucky day
1692             flags = WC_NO_BEST_FIT_CHARS;
1693             pUsedDef = &usedDef;
1694         }
1695         else // old system or unsupported encoding
1696         {
1697             flags = 0;
1698             pUsedDef = NULL;
1699         }
1700
1701         const size_t len = ::WideCharToMultiByte
1702                              (
1703                                 m_CodePage,     // code page
1704                                 flags,          // either none or no best fit
1705                                 pwz,            // input string
1706                                 -1,             // it is (wide) NUL-terminated
1707                                 buf,            // output buffer
1708                                 buf ? n : 0,    // and its size
1709                                 NULL,           // default "replacement" char
1710                                 pUsedDef        // [out] was it used?
1711                              );
1712
1713         if ( !len )
1714         {
1715             // function totally failed
1716             return (size_t)-1;
1717         }
1718
1719         // if we were really converting, check if we succeeded
1720         if ( buf )
1721         {
1722             if ( flags )
1723             {
1724                 // check if the conversion failed, i.e. if any replacements
1725                 // were done
1726                 if ( usedDef )
1727                     return (size_t)-1;
1728             }
1729             else // we must resort to double tripping...
1730             {
1731                 wxWCharBuffer wcBuf(n);
1732                 if ( MB2WC(wcBuf.data(), buf, n) == (size_t)-1 ||
1733                         wcscmp(wcBuf, pwz) != 0 )
1734                 {
1735                     // we didn't obtain the same thing we started from, hence
1736                     // the conversion was lossy and we consider that it failed
1737                     return (size_t)-1;
1738                 }
1739             }
1740         }
1741
1742         // see the comment above for the reason of "len - 1"
1743         return len - 1;
1744     }
1745
1746     bool IsOk() const { return m_CodePage != -1; }
1747
1748 private:
1749     static bool CanUseNoBestFit()
1750     {
1751         static int s_isWin98Or2k = -1;
1752
1753         if ( s_isWin98Or2k == -1 )
1754         {
1755             int verMaj, verMin;
1756             switch ( wxGetOsVersion(&verMaj, &verMin) )
1757             {
1758                 case wxWIN95:
1759                     s_isWin98Or2k = verMaj >= 4 && verMin >= 10;
1760                     break;
1761
1762                 case wxWINDOWS_NT:
1763                     s_isWin98Or2k = verMaj >= 5;
1764                     break;
1765
1766                 default:
1767                     // unknown, be conseravtive by default
1768                     s_isWin98Or2k = 0;
1769             }
1770
1771             wxASSERT_MSG( s_isWin98Or2k != -1, _T("should be set above") );
1772         }
1773
1774         return s_isWin98Or2k == 1;
1775     }
1776
1777     long m_CodePage;
1778 };
1779
1780 #endif // wxHAVE_WIN32_MB2WC
1781
1782 // ============================================================================
1783 // Cocoa conversion classes
1784 // ============================================================================
1785
1786 #if defined(__WXCOCOA__)
1787
1788 // RN:  There is no UTF-32 support in either Core Foundation or
1789 // Cocoa.  Strangely enough, internally Core Foundation uses
1790 // UTF 32 internally quite a bit - its just not public (yet).
1791
1792 #include <CoreFoundation/CFString.h>
1793 #include <CoreFoundation/CFStringEncodingExt.h>
1794
1795 CFStringEncoding wxCFStringEncFromFontEnc(wxFontEncoding encoding)
1796 {
1797     CFStringEncoding enc = kCFStringEncodingInvalidId ;
1798     if ( encoding == wxFONTENCODING_DEFAULT )
1799     {
1800         enc = CFStringGetSystemEncoding();
1801     }
1802     else switch( encoding)
1803     {
1804         case wxFONTENCODING_ISO8859_1 :
1805             enc = kCFStringEncodingISOLatin1 ;
1806             break ;
1807         case wxFONTENCODING_ISO8859_2 :
1808             enc = kCFStringEncodingISOLatin2;
1809             break ;
1810         case wxFONTENCODING_ISO8859_3 :
1811             enc = kCFStringEncodingISOLatin3 ;
1812             break ;
1813         case wxFONTENCODING_ISO8859_4 :
1814             enc = kCFStringEncodingISOLatin4;
1815             break ;
1816         case wxFONTENCODING_ISO8859_5 :
1817             enc = kCFStringEncodingISOLatinCyrillic;
1818             break ;
1819         case wxFONTENCODING_ISO8859_6 :
1820             enc = kCFStringEncodingISOLatinArabic;
1821             break ;
1822         case wxFONTENCODING_ISO8859_7 :
1823             enc = kCFStringEncodingISOLatinGreek;
1824             break ;
1825         case wxFONTENCODING_ISO8859_8 :
1826             enc = kCFStringEncodingISOLatinHebrew;
1827             break ;
1828         case wxFONTENCODING_ISO8859_9 :
1829             enc = kCFStringEncodingISOLatin5;
1830             break ;
1831         case wxFONTENCODING_ISO8859_10 :
1832             enc = kCFStringEncodingISOLatin6;
1833             break ;
1834         case wxFONTENCODING_ISO8859_11 :
1835             enc = kCFStringEncodingISOLatinThai;
1836             break ;
1837         case wxFONTENCODING_ISO8859_13 :
1838             enc = kCFStringEncodingISOLatin7;
1839             break ;
1840         case wxFONTENCODING_ISO8859_14 :
1841             enc = kCFStringEncodingISOLatin8;
1842             break ;
1843         case wxFONTENCODING_ISO8859_15 :
1844             enc = kCFStringEncodingISOLatin9;
1845             break ;
1846
1847         case wxFONTENCODING_KOI8 :
1848             enc = kCFStringEncodingKOI8_R;
1849             break ;
1850         case wxFONTENCODING_ALTERNATIVE : // MS-DOS CP866
1851             enc = kCFStringEncodingDOSRussian;
1852             break ;
1853
1854 //      case wxFONTENCODING_BULGARIAN :
1855 //          enc = ;
1856 //          break ;
1857
1858         case wxFONTENCODING_CP437 :
1859             enc =kCFStringEncodingDOSLatinUS ;
1860             break ;
1861         case wxFONTENCODING_CP850 :
1862             enc = kCFStringEncodingDOSLatin1;
1863             break ;
1864         case wxFONTENCODING_CP852 :
1865             enc = kCFStringEncodingDOSLatin2;
1866             break ;
1867         case wxFONTENCODING_CP855 :
1868             enc = kCFStringEncodingDOSCyrillic;
1869             break ;
1870         case wxFONTENCODING_CP866 :
1871             enc =kCFStringEncodingDOSRussian ;
1872             break ;
1873         case wxFONTENCODING_CP874 :
1874             enc = kCFStringEncodingDOSThai;
1875             break ;
1876         case wxFONTENCODING_CP932 :
1877             enc = kCFStringEncodingDOSJapanese;
1878             break ;
1879         case wxFONTENCODING_CP936 :
1880             enc =kCFStringEncodingDOSChineseSimplif ;
1881             break ;
1882         case wxFONTENCODING_CP949 :
1883             enc = kCFStringEncodingDOSKorean;
1884             break ;
1885         case wxFONTENCODING_CP950 :
1886             enc = kCFStringEncodingDOSChineseTrad;
1887             break ;
1888         case wxFONTENCODING_CP1250 :
1889             enc = kCFStringEncodingWindowsLatin2;
1890             break ;
1891         case wxFONTENCODING_CP1251 :
1892             enc =kCFStringEncodingWindowsCyrillic ;
1893             break ;
1894         case wxFONTENCODING_CP1252 :
1895             enc =kCFStringEncodingWindowsLatin1 ;
1896             break ;
1897         case wxFONTENCODING_CP1253 :
1898             enc = kCFStringEncodingWindowsGreek;
1899             break ;
1900         case wxFONTENCODING_CP1254 :
1901             enc = kCFStringEncodingWindowsLatin5;
1902             break ;
1903         case wxFONTENCODING_CP1255 :
1904             enc =kCFStringEncodingWindowsHebrew ;
1905             break ;
1906         case wxFONTENCODING_CP1256 :
1907             enc =kCFStringEncodingWindowsArabic ;
1908             break ;
1909         case wxFONTENCODING_CP1257 :
1910             enc = kCFStringEncodingWindowsBalticRim;
1911             break ;
1912 //   This only really encodes to UTF7 (if that) evidently
1913 //        case wxFONTENCODING_UTF7 :
1914 //            enc = kCFStringEncodingNonLossyASCII ;
1915 //            break ;
1916         case wxFONTENCODING_UTF8 :
1917             enc = kCFStringEncodingUTF8 ;
1918             break ;
1919         case wxFONTENCODING_EUC_JP :
1920             enc = kCFStringEncodingEUC_JP;
1921             break ;
1922         case wxFONTENCODING_UTF16 :
1923             enc = kCFStringEncodingUnicode ;
1924             break ;
1925         case wxFONTENCODING_MACROMAN :
1926             enc = kCFStringEncodingMacRoman ;
1927             break ;
1928         case wxFONTENCODING_MACJAPANESE :
1929             enc = kCFStringEncodingMacJapanese ;
1930             break ;
1931         case wxFONTENCODING_MACCHINESETRAD :
1932             enc = kCFStringEncodingMacChineseTrad ;
1933             break ;
1934         case wxFONTENCODING_MACKOREAN :
1935             enc = kCFStringEncodingMacKorean ;
1936             break ;
1937         case wxFONTENCODING_MACARABIC :
1938             enc = kCFStringEncodingMacArabic ;
1939             break ;
1940         case wxFONTENCODING_MACHEBREW :
1941             enc = kCFStringEncodingMacHebrew ;
1942             break ;
1943         case wxFONTENCODING_MACGREEK :
1944             enc = kCFStringEncodingMacGreek ;
1945             break ;
1946         case wxFONTENCODING_MACCYRILLIC :
1947             enc = kCFStringEncodingMacCyrillic ;
1948             break ;
1949         case wxFONTENCODING_MACDEVANAGARI :
1950             enc = kCFStringEncodingMacDevanagari ;
1951             break ;
1952         case wxFONTENCODING_MACGURMUKHI :
1953             enc = kCFStringEncodingMacGurmukhi ;
1954             break ;
1955         case wxFONTENCODING_MACGUJARATI :
1956             enc = kCFStringEncodingMacGujarati ;
1957             break ;
1958         case wxFONTENCODING_MACORIYA :
1959             enc = kCFStringEncodingMacOriya ;
1960             break ;
1961         case wxFONTENCODING_MACBENGALI :
1962             enc = kCFStringEncodingMacBengali ;
1963             break ;
1964         case wxFONTENCODING_MACTAMIL :
1965             enc = kCFStringEncodingMacTamil ;
1966             break ;
1967         case wxFONTENCODING_MACTELUGU :
1968             enc = kCFStringEncodingMacTelugu ;
1969             break ;
1970         case wxFONTENCODING_MACKANNADA :
1971             enc = kCFStringEncodingMacKannada ;
1972             break ;
1973         case wxFONTENCODING_MACMALAJALAM :
1974             enc = kCFStringEncodingMacMalayalam ;
1975             break ;
1976         case wxFONTENCODING_MACSINHALESE :
1977             enc = kCFStringEncodingMacSinhalese ;
1978             break ;
1979         case wxFONTENCODING_MACBURMESE :
1980             enc = kCFStringEncodingMacBurmese ;
1981             break ;
1982         case wxFONTENCODING_MACKHMER :
1983             enc = kCFStringEncodingMacKhmer ;
1984             break ;
1985         case wxFONTENCODING_MACTHAI :
1986             enc = kCFStringEncodingMacThai ;
1987             break ;
1988         case wxFONTENCODING_MACLAOTIAN :
1989             enc = kCFStringEncodingMacLaotian ;
1990             break ;
1991         case wxFONTENCODING_MACGEORGIAN :
1992             enc = kCFStringEncodingMacGeorgian ;
1993             break ;
1994         case wxFONTENCODING_MACARMENIAN :
1995             enc = kCFStringEncodingMacArmenian ;
1996             break ;
1997         case wxFONTENCODING_MACCHINESESIMP :
1998             enc = kCFStringEncodingMacChineseSimp ;
1999             break ;
2000         case wxFONTENCODING_MACTIBETAN :
2001             enc = kCFStringEncodingMacTibetan ;
2002             break ;
2003         case wxFONTENCODING_MACMONGOLIAN :
2004             enc = kCFStringEncodingMacMongolian ;
2005             break ;
2006         case wxFONTENCODING_MACETHIOPIC :
2007             enc = kCFStringEncodingMacEthiopic ;
2008             break ;
2009         case wxFONTENCODING_MACCENTRALEUR :
2010             enc = kCFStringEncodingMacCentralEurRoman ;
2011             break ;
2012         case wxFONTENCODING_MACVIATNAMESE :
2013             enc = kCFStringEncodingMacVietnamese ;
2014             break ;
2015         case wxFONTENCODING_MACARABICEXT :
2016             enc = kCFStringEncodingMacExtArabic ;
2017             break ;
2018         case wxFONTENCODING_MACSYMBOL :
2019             enc = kCFStringEncodingMacSymbol ;
2020             break ;
2021         case wxFONTENCODING_MACDINGBATS :
2022             enc = kCFStringEncodingMacDingbats ;
2023             break ;
2024         case wxFONTENCODING_MACTURKISH :
2025             enc = kCFStringEncodingMacTurkish ;
2026             break ;
2027         case wxFONTENCODING_MACCROATIAN :
2028             enc = kCFStringEncodingMacCroatian ;
2029             break ;
2030         case wxFONTENCODING_MACICELANDIC :
2031             enc = kCFStringEncodingMacIcelandic ;
2032             break ;
2033         case wxFONTENCODING_MACROMANIAN :
2034             enc = kCFStringEncodingMacRomanian ;
2035             break ;
2036         case wxFONTENCODING_MACCELTIC :
2037             enc = kCFStringEncodingMacCeltic ;
2038             break ;
2039         case wxFONTENCODING_MACGAELIC :
2040             enc = kCFStringEncodingMacGaelic ;
2041             break ;
2042 //      case wxFONTENCODING_MACKEYBOARD :
2043 //          enc = kCFStringEncodingMacKeyboardGlyphs ;
2044 //          break ;
2045         default :
2046             // because gcc is picky
2047             break ;
2048     } ;
2049     return enc ;
2050 }
2051
2052 class wxMBConv_cocoa : public wxMBConv
2053 {
2054 public:
2055     wxMBConv_cocoa()
2056     {
2057         Init(CFStringGetSystemEncoding()) ;
2058     }
2059
2060 #if wxUSE_FONTMAP
2061     wxMBConv_cocoa(const wxChar* name)
2062     {
2063         Init( wxCFStringEncFromFontEnc(wxFontMapperBase::Get()->CharsetToEncoding(name, false) ) ) ;
2064     }
2065 #endif
2066
2067     wxMBConv_cocoa(wxFontEncoding encoding)
2068     {
2069         Init( wxCFStringEncFromFontEnc(encoding) );
2070     }
2071
2072     ~wxMBConv_cocoa()
2073     {
2074     }
2075
2076     void Init( CFStringEncoding encoding)
2077     {
2078         m_encoding = encoding ;
2079     }
2080
2081     size_t MB2WC(wchar_t * szOut, const char * szUnConv, size_t nOutSize) const
2082     {
2083         wxASSERT(szUnConv);
2084
2085         CFStringRef theString = CFStringCreateWithBytes (
2086                                                 NULL, //the allocator
2087                                                 (const UInt8*)szUnConv,
2088                                                 strlen(szUnConv),
2089                                                 m_encoding,
2090                                                 false //no BOM/external representation
2091                                                 );
2092
2093         wxASSERT(theString);
2094
2095         size_t nOutLength = CFStringGetLength(theString);
2096
2097         if (szOut == NULL)
2098         {
2099             CFRelease(theString);
2100             return nOutLength;
2101         }
2102
2103         CFRange theRange = { 0, nOutSize };
2104
2105 #if SIZEOF_WCHAR_T == 4
2106         UniChar* szUniCharBuffer = new UniChar[nOutSize];
2107 #endif
2108
2109         CFStringGetCharacters(theString, theRange, szUniCharBuffer);
2110
2111         CFRelease(theString);
2112
2113         szUniCharBuffer[nOutLength] = '\0' ;
2114
2115 #if SIZEOF_WCHAR_T == 4
2116         wxMBConvUTF16 converter ;
2117         converter.MB2WC(szOut, (const char*)szUniCharBuffer , nOutSize ) ;
2118         delete[] szUniCharBuffer;
2119 #endif
2120
2121         return nOutLength;
2122     }
2123
2124     size_t WC2MB(char *szOut, const wchar_t *szUnConv, size_t nOutSize) const
2125     {
2126         wxASSERT(szUnConv);
2127
2128         size_t nRealOutSize;
2129         size_t nBufSize = wxWcslen(szUnConv);
2130         UniChar* szUniBuffer = (UniChar*) szUnConv;
2131
2132 #if SIZEOF_WCHAR_T == 4
2133         wxMBConvUTF16 converter ;
2134         nBufSize = converter.WC2MB( NULL , szUnConv , 0 );
2135         szUniBuffer = new UniChar[ (nBufSize / sizeof(UniChar)) + 1] ;
2136         converter.WC2MB( (char*) szUniBuffer , szUnConv, nBufSize + sizeof(UniChar)) ;
2137         nBufSize /= sizeof(UniChar);
2138 #endif
2139
2140         CFStringRef theString = CFStringCreateWithCharactersNoCopy(
2141                                 NULL, //allocator
2142                                 szUniBuffer,
2143                                 nBufSize,
2144                                 kCFAllocatorNull //deallocator - we want to deallocate it ourselves
2145                             );
2146
2147         wxASSERT(theString);
2148
2149         //Note that CER puts a BOM when converting to unicode
2150         //so we  check and use getchars instead in that case
2151         if (m_encoding == kCFStringEncodingUnicode)
2152         {
2153             if (szOut != NULL)
2154                 CFStringGetCharacters(theString, CFRangeMake(0, nOutSize - 1), (UniChar*) szOut);
2155
2156             nRealOutSize = CFStringGetLength(theString) + 1;
2157         }
2158         else
2159         {
2160             CFStringGetBytes(
2161                 theString,
2162                 CFRangeMake(0, CFStringGetLength(theString)),
2163                 m_encoding,
2164                 0, //what to put in characters that can't be converted -
2165                     //0 tells CFString to return NULL if it meets such a character
2166                 false, //not an external representation
2167                 (UInt8*) szOut,
2168                 nOutSize,
2169                 (CFIndex*) &nRealOutSize
2170                         );
2171         }
2172
2173         CFRelease(theString);
2174
2175 #if SIZEOF_WCHAR_T == 4
2176         delete[] szUniBuffer;
2177 #endif
2178
2179         return  nRealOutSize - 1;
2180     }
2181
2182     bool IsOk() const
2183     {
2184         return m_encoding != kCFStringEncodingInvalidId &&
2185               CFStringIsEncodingAvailable(m_encoding);
2186     }
2187
2188 private:
2189     CFStringEncoding m_encoding ;
2190 };
2191
2192 #endif // defined(__WXCOCOA__)
2193
2194 // ============================================================================
2195 // Mac conversion classes
2196 // ============================================================================
2197
2198 #if defined(__WXMAC__) && defined(TARGET_CARBON)
2199
2200 class wxMBConv_mac : public wxMBConv
2201 {
2202 public:
2203     wxMBConv_mac()
2204     {
2205         Init(CFStringGetSystemEncoding()) ;
2206     }
2207
2208 #if wxUSE_FONTMAP
2209     wxMBConv_mac(const wxChar* name)
2210     {
2211         Init( wxMacGetSystemEncFromFontEnc(wxFontMapperBase::Get()->CharsetToEncoding(name, false) ) ) ;
2212     }
2213 #endif
2214
2215     wxMBConv_mac(wxFontEncoding encoding)
2216     {
2217         Init( wxMacGetSystemEncFromFontEnc(encoding) );
2218     }
2219
2220     ~wxMBConv_mac()
2221     {
2222         OSStatus status = noErr ;
2223         status = TECDisposeConverter(m_MB2WC_converter);
2224         status = TECDisposeConverter(m_WC2MB_converter);
2225     }
2226
2227
2228     void Init( TextEncodingBase encoding)
2229     {
2230         OSStatus status = noErr ;
2231         m_char_encoding = encoding ;
2232         m_unicode_encoding = CreateTextEncoding(kTextEncodingUnicodeDefault,0,kUnicode16BitFormat) ;
2233
2234         status = TECCreateConverter(&m_MB2WC_converter,
2235                                     m_char_encoding,
2236                                     m_unicode_encoding);
2237         status = TECCreateConverter(&m_WC2MB_converter,
2238                                     m_unicode_encoding,
2239                                     m_char_encoding);
2240     }
2241
2242     size_t MB2WC(wchar_t *buf, const char *psz, size_t n) const
2243     {
2244         OSStatus status = noErr ;
2245         ByteCount byteOutLen ;
2246         ByteCount byteInLen = strlen(psz) ;
2247         wchar_t *tbuf = NULL ;
2248         UniChar* ubuf = NULL ;
2249         size_t res = 0 ;
2250
2251         if (buf == NULL)
2252         {
2253             //apple specs say at least 32
2254             n = wxMax( 32 , byteInLen ) ;
2255             tbuf = (wchar_t*) malloc( n * SIZEOF_WCHAR_T) ;
2256         }
2257         ByteCount byteBufferLen = n * sizeof( UniChar ) ;
2258 #if SIZEOF_WCHAR_T == 4
2259         ubuf = (UniChar*) malloc( byteBufferLen + 2 ) ;
2260 #else
2261         ubuf = (UniChar*) (buf ? buf : tbuf) ;
2262 #endif
2263         status = TECConvertText(m_MB2WC_converter, (ConstTextPtr) psz , byteInLen, &byteInLen,
2264           (TextPtr) ubuf , byteBufferLen, &byteOutLen);
2265 #if SIZEOF_WCHAR_T == 4
2266         // we have to terminate here, because n might be larger for the trailing zero, and if UniChar
2267         // is not properly terminated we get random characters at the end
2268         ubuf[byteOutLen / sizeof( UniChar ) ] = 0 ;
2269         wxMBConvUTF16 converter ;
2270         res = converter.MB2WC( (buf ? buf : tbuf) , (const char*)ubuf , n ) ;
2271         free( ubuf ) ;
2272 #else
2273         res = byteOutLen / sizeof( UniChar ) ;
2274 #endif
2275         if ( buf == NULL )
2276              free(tbuf) ;
2277
2278         if ( buf  && res < n)
2279             buf[res] = 0;
2280
2281         return res ;
2282     }
2283
2284     size_t WC2MB(char *buf, const wchar_t *psz, size_t n) const
2285     {
2286         OSStatus status = noErr ;
2287         ByteCount byteOutLen ;
2288         ByteCount byteInLen = wxWcslen(psz) * SIZEOF_WCHAR_T ;
2289
2290         char *tbuf = NULL ;
2291
2292         if (buf == NULL)
2293         {
2294             //apple specs say at least 32
2295             n = wxMax( 32 , ((byteInLen / SIZEOF_WCHAR_T) * 8) + SIZEOF_WCHAR_T );
2296             tbuf = (char*) malloc( n ) ;
2297         }
2298
2299         ByteCount byteBufferLen = n ;
2300         UniChar* ubuf = NULL ;
2301 #if SIZEOF_WCHAR_T == 4
2302         wxMBConvUTF16 converter ;
2303         size_t unicharlen = converter.WC2MB( NULL , psz , 0 ) ;
2304         byteInLen = unicharlen ;
2305         ubuf = (UniChar*) malloc( byteInLen + 2 ) ;
2306         converter.WC2MB( (char*) ubuf , psz, unicharlen + 2 ) ;
2307 #else
2308         ubuf = (UniChar*) psz ;
2309 #endif
2310         status = TECConvertText(m_WC2MB_converter, (ConstTextPtr) ubuf , byteInLen, &byteInLen,
2311             (TextPtr) (buf ? buf : tbuf) , byteBufferLen, &byteOutLen);
2312 #if SIZEOF_WCHAR_T == 4
2313         free( ubuf ) ;
2314 #endif
2315         if ( buf == NULL )
2316             free(tbuf) ;
2317
2318         size_t res = byteOutLen ;
2319         if ( buf  && res < n)
2320         {
2321             buf[res] = 0;
2322
2323             //we need to double-trip to verify it didn't insert any ? in place
2324             //of bogus characters
2325             wxWCharBuffer wcBuf(n);
2326             size_t pszlen = wxWcslen(psz);
2327             if ( MB2WC(wcBuf.data(), buf, n) == (size_t)-1 ||
2328                         wxWcslen(wcBuf) != pszlen ||
2329                         memcmp(wcBuf, psz, pszlen * sizeof(wchar_t)) != 0 )
2330             {
2331                 // we didn't obtain the same thing we started from, hence
2332                 // the conversion was lossy and we consider that it failed
2333                 return (size_t)-1;
2334             }
2335         }
2336
2337         return res ;
2338     }
2339
2340     bool IsOk() const
2341         { return m_MB2WC_converter !=  NULL && m_WC2MB_converter != NULL  ; }
2342
2343 private:
2344     TECObjectRef m_MB2WC_converter ;
2345     TECObjectRef m_WC2MB_converter ;
2346
2347     TextEncodingBase m_char_encoding ;
2348     TextEncodingBase m_unicode_encoding ;
2349 };
2350
2351 #endif // defined(__WXMAC__) && defined(TARGET_CARBON)
2352
2353 // ============================================================================
2354 // wxEncodingConverter based conversion classes
2355 // ============================================================================
2356
2357 #if wxUSE_FONTMAP
2358
2359 class wxMBConv_wxwin : public wxMBConv
2360 {
2361 private:
2362     void Init()
2363     {
2364         m_ok = m2w.Init(m_enc, wxFONTENCODING_UNICODE) &&
2365                w2m.Init(wxFONTENCODING_UNICODE, m_enc);
2366     }
2367
2368 public:
2369     // temporarily just use wxEncodingConverter stuff,
2370     // so that it works while a better implementation is built
2371     wxMBConv_wxwin(const wxChar* name)
2372     {
2373         if (name)
2374             m_enc = wxFontMapperBase::Get()->CharsetToEncoding(name, false);
2375         else
2376             m_enc = wxFONTENCODING_SYSTEM;
2377
2378         Init();
2379     }
2380
2381     wxMBConv_wxwin(wxFontEncoding enc)
2382     {
2383         m_enc = enc;
2384
2385         Init();
2386     }
2387
2388     size_t MB2WC(wchar_t *buf, const char *psz, size_t WXUNUSED(n)) const
2389     {
2390         size_t inbuf = strlen(psz);
2391         if (buf)
2392         {
2393             if (!m2w.Convert(psz,buf))
2394                 return (size_t)-1;
2395         }
2396         return inbuf;
2397     }
2398
2399     size_t WC2MB(char *buf, const wchar_t *psz, size_t WXUNUSED(n)) const
2400     {
2401         const size_t inbuf = wxWcslen(psz);
2402         if (buf)
2403         {
2404             if (!w2m.Convert(psz,buf))
2405                 return (size_t)-1;
2406         }
2407
2408         return inbuf;
2409     }
2410
2411     bool IsOk() const { return m_ok; }
2412
2413 public:
2414     wxFontEncoding m_enc;
2415     wxEncodingConverter m2w, w2m;
2416
2417     // were we initialized successfully?
2418     bool m_ok;
2419
2420     DECLARE_NO_COPY_CLASS(wxMBConv_wxwin)
2421 };
2422
2423 // make the constructors available for unit testing
2424 WXDLLIMPEXP_BASE wxMBConv* new_wxMBConv_wxwin( const wxChar* name )
2425 {
2426     wxMBConv_wxwin* result = new wxMBConv_wxwin( name );
2427     if ( !result->IsOk() )
2428     {
2429         delete result;
2430         return 0;
2431     }
2432     return result;
2433 }
2434
2435 #endif // wxUSE_FONTMAP
2436
2437 // ============================================================================
2438 // wxCSConv implementation
2439 // ============================================================================
2440
2441 void wxCSConv::Init()
2442 {
2443     m_name = NULL;
2444     m_convReal =  NULL;
2445     m_deferred = true;
2446 }
2447
2448 wxCSConv::wxCSConv(const wxChar *charset)
2449 {
2450     Init();
2451
2452     if ( charset )
2453     {
2454         SetName(charset);
2455     }
2456
2457     m_encoding = wxFONTENCODING_SYSTEM;
2458 }
2459
2460 wxCSConv::wxCSConv(wxFontEncoding encoding)
2461 {
2462     if ( encoding == wxFONTENCODING_MAX || encoding == wxFONTENCODING_DEFAULT )
2463     {
2464         wxFAIL_MSG( _T("invalid encoding value in wxCSConv ctor") );
2465
2466         encoding = wxFONTENCODING_SYSTEM;
2467     }
2468
2469     Init();
2470
2471     m_encoding = encoding;
2472 }
2473
2474 wxCSConv::~wxCSConv()
2475 {
2476     Clear();
2477 }
2478
2479 wxCSConv::wxCSConv(const wxCSConv& conv)
2480         : wxMBConv()
2481 {
2482     Init();
2483
2484     SetName(conv.m_name);
2485     m_encoding = conv.m_encoding;
2486 }
2487
2488 wxCSConv& wxCSConv::operator=(const wxCSConv& conv)
2489 {
2490     Clear();
2491
2492     SetName(conv.m_name);
2493     m_encoding = conv.m_encoding;
2494
2495     return *this;
2496 }
2497
2498 void wxCSConv::Clear()
2499 {
2500     free(m_name);
2501     delete m_convReal;
2502
2503     m_name = NULL;
2504     m_convReal = NULL;
2505 }
2506
2507 void wxCSConv::SetName(const wxChar *charset)
2508 {
2509     if (charset)
2510     {
2511         m_name = wxStrdup(charset);
2512         m_deferred = true;
2513     }
2514 }
2515
2516 #if wxUSE_FONTMAP
2517 #include "wx/hashmap.h"
2518
2519 WX_DECLARE_HASH_MAP( wxFontEncoding, wxString, wxIntegerHash, wxIntegerEqual,
2520                      wxEncodingNameCache );
2521
2522 static wxEncodingNameCache gs_nameCache;
2523 #endif
2524
2525 wxMBConv *wxCSConv::DoCreate() const
2526 {
2527 #if wxUSE_FONTMAP
2528     wxLogTrace(TRACE_STRCONV,
2529                wxT("creating conversion for %s"),
2530                (m_name ? m_name
2531                        : wxFontMapperBase::GetEncodingName(m_encoding).c_str()));
2532 #endif // wxUSE_FONTMAP
2533
2534     // check for the special case of ASCII or ISO8859-1 charset: as we have
2535     // special knowledge of it anyhow, we don't need to create a special
2536     // conversion object
2537     if ( m_encoding == wxFONTENCODING_ISO8859_1 )
2538     {
2539         // don't convert at all
2540         return NULL;
2541     }
2542
2543     // we trust OS to do conversion better than we can so try external
2544     // conversion methods first
2545     //
2546     // the full order is:
2547     //      1. OS conversion (iconv() under Unix or Win32 API)
2548     //      2. hard coded conversions for UTF
2549     //      3. wxEncodingConverter as fall back
2550
2551     // step (1)
2552 #ifdef HAVE_ICONV
2553 #if !wxUSE_FONTMAP
2554     if ( m_name )
2555 #endif // !wxUSE_FONTMAP
2556     {
2557         wxString name(m_name);
2558         wxFontEncoding encoding(m_encoding);
2559
2560         if ( !name.empty() )
2561         {
2562             wxMBConv_iconv *conv = new wxMBConv_iconv(name);
2563             if ( conv->IsOk() )
2564                 return conv;
2565
2566             delete conv;
2567
2568 #if wxUSE_FONTMAP
2569             encoding =
2570                 wxFontMapperBase::Get()->CharsetToEncoding(name, false);
2571 #endif // wxUSE_FONTMAP
2572         }
2573 #if wxUSE_FONTMAP
2574         {
2575             const wxEncodingNameCache::iterator it = gs_nameCache.find(encoding);
2576             if ( it != gs_nameCache.end() )
2577             {
2578                 if ( it->second.empty() )
2579                     return NULL;
2580
2581                 wxMBConv_iconv *conv = new wxMBConv_iconv(it->second);
2582                 if ( conv->IsOk() )
2583                     return conv;
2584
2585                 delete conv;
2586             }
2587
2588             const wxChar** names = wxFontMapperBase::GetAllEncodingNames(encoding);
2589
2590             for ( ; *names; ++names )
2591             {
2592                 wxMBConv_iconv *conv = new wxMBConv_iconv(*names);
2593                 if ( conv->IsOk() )
2594                 {
2595                     gs_nameCache[encoding] = *names;
2596                     return conv;
2597                 }
2598
2599                 delete conv;
2600             }
2601
2602             gs_nameCache[encoding] = _T(""); // cache the failure
2603         }
2604 #endif // wxUSE_FONTMAP
2605     }
2606 #endif // HAVE_ICONV
2607
2608 #ifdef wxHAVE_WIN32_MB2WC
2609     {
2610 #if wxUSE_FONTMAP
2611         wxMBConv_win32 *conv = m_name ? new wxMBConv_win32(m_name)
2612                                       : new wxMBConv_win32(m_encoding);
2613         if ( conv->IsOk() )
2614             return conv;
2615
2616         delete conv;
2617 #else
2618         return NULL;
2619 #endif
2620     }
2621 #endif // wxHAVE_WIN32_MB2WC
2622 #if defined(__WXMAC__)
2623     {
2624         // leave UTF16 and UTF32 to the built-ins of wx
2625         if ( m_name || ( m_encoding < wxFONTENCODING_UTF16BE ||
2626             ( m_encoding >= wxFONTENCODING_MACMIN && m_encoding <= wxFONTENCODING_MACMAX ) ) )
2627         {
2628
2629 #if wxUSE_FONTMAP
2630             wxMBConv_mac *conv = m_name ? new wxMBConv_mac(m_name)
2631                                         : new wxMBConv_mac(m_encoding);
2632 #else
2633             wxMBConv_mac *conv = new wxMBConv_mac(m_encoding);
2634 #endif
2635             if ( conv->IsOk() )
2636                  return conv;
2637
2638             delete conv;
2639         }
2640     }
2641 #endif
2642 #if defined(__WXCOCOA__)
2643     {
2644         if ( m_name || ( m_encoding <= wxFONTENCODING_UTF16 ) )
2645         {
2646
2647 #if wxUSE_FONTMAP
2648             wxMBConv_cocoa *conv = m_name ? new wxMBConv_cocoa(m_name)
2649                                           : new wxMBConv_cocoa(m_encoding);
2650 #else
2651             wxMBConv_cocoa *conv = new wxMBConv_cocoa(m_encoding);
2652 #endif
2653             if ( conv->IsOk() )
2654                  return conv;
2655
2656             delete conv;
2657         }
2658     }
2659 #endif
2660     // step (2)
2661     wxFontEncoding enc = m_encoding;
2662 #if wxUSE_FONTMAP
2663     if ( enc == wxFONTENCODING_SYSTEM && m_name )
2664     {
2665         // use "false" to suppress interactive dialogs -- we can be called from
2666         // anywhere and popping up a dialog from here is the last thing we want to
2667         // do
2668         enc = wxFontMapperBase::Get()->CharsetToEncoding(m_name, false);
2669     }
2670 #endif // wxUSE_FONTMAP
2671
2672     switch ( enc )
2673     {
2674         case wxFONTENCODING_UTF7:
2675              return new wxMBConvUTF7;
2676
2677         case wxFONTENCODING_UTF8:
2678              return new wxMBConvUTF8;
2679
2680         case wxFONTENCODING_UTF16BE:
2681              return new wxMBConvUTF16BE;
2682
2683         case wxFONTENCODING_UTF16LE:
2684              return new wxMBConvUTF16LE;
2685
2686         case wxFONTENCODING_UTF32BE:
2687              return new wxMBConvUTF32BE;
2688
2689         case wxFONTENCODING_UTF32LE:
2690              return new wxMBConvUTF32LE;
2691
2692         default:
2693              // nothing to do but put here to suppress gcc warnings
2694              ;
2695     }
2696
2697     // step (3)
2698 #if wxUSE_FONTMAP
2699     {
2700         wxMBConv_wxwin *conv = m_name ? new wxMBConv_wxwin(m_name)
2701                                       : new wxMBConv_wxwin(m_encoding);
2702         if ( conv->IsOk() )
2703             return conv;
2704
2705         delete conv;
2706     }
2707 #endif // wxUSE_FONTMAP
2708
2709     // NB: This is a hack to prevent deadlock. What could otherwise happen
2710     //     in Unicode build: wxConvLocal creation ends up being here
2711     //     because of some failure and logs the error. But wxLog will try to
2712     //     attach timestamp, for which it will need wxConvLocal (to convert
2713     //     time to char* and then wchar_t*), but that fails, tries to log
2714     //     error, but wxLog has a (already locked) critical section that
2715     //     guards static buffer.
2716     static bool alreadyLoggingError = false;
2717     if (!alreadyLoggingError)
2718     {
2719         alreadyLoggingError = true;
2720         wxLogError(_("Cannot convert from the charset '%s'!"),
2721                    m_name ? m_name
2722                       :
2723 #if wxUSE_FONTMAP
2724                          wxFontMapperBase::GetEncodingDescription(m_encoding).c_str()
2725 #else // !wxUSE_FONTMAP
2726                          wxString::Format(_("encoding %s"), m_encoding).c_str()
2727 #endif // wxUSE_FONTMAP/!wxUSE_FONTMAP
2728               );
2729         alreadyLoggingError = false;
2730     }
2731
2732     return NULL;
2733 }
2734
2735 void wxCSConv::CreateConvIfNeeded() const
2736 {
2737     if ( m_deferred )
2738     {
2739         wxCSConv *self = (wxCSConv *)this; // const_cast
2740
2741 #if wxUSE_INTL
2742         // if we don't have neither the name nor the encoding, use the default
2743         // encoding for this system
2744         if ( !m_name && m_encoding == wxFONTENCODING_SYSTEM )
2745         {
2746             self->m_name = wxStrdup(wxLocale::GetSystemEncodingName());
2747         }
2748 #endif // wxUSE_INTL
2749
2750         self->m_convReal = DoCreate();
2751         self->m_deferred = false;
2752     }
2753 }
2754
2755 size_t wxCSConv::MB2WC(wchar_t *buf, const char *psz, size_t n) const
2756 {
2757     CreateConvIfNeeded();
2758
2759     if (m_convReal)
2760         return m_convReal->MB2WC(buf, psz, n);
2761
2762     // latin-1 (direct)
2763     size_t len = strlen(psz);
2764
2765     if (buf)
2766     {
2767         for (size_t c = 0; c <= len; c++)
2768             buf[c] = (unsigned char)(psz[c]);
2769     }
2770
2771     return len;
2772 }
2773
2774 size_t wxCSConv::WC2MB(char *buf, const wchar_t *psz, size_t n) const
2775 {
2776     CreateConvIfNeeded();
2777
2778     if (m_convReal)
2779         return m_convReal->WC2MB(buf, psz, n);
2780
2781     // latin-1 (direct)
2782     const size_t len = wxWcslen(psz);
2783     if (buf)
2784     {
2785         for (size_t c = 0; c <= len; c++)
2786         {
2787             if (psz[c] > 0xFF)
2788                 return (size_t)-1;
2789             buf[c] = (char)psz[c];
2790         }
2791     }
2792     else
2793     {
2794         for (size_t c = 0; c <= len; c++)
2795         {
2796             if (psz[c] > 0xFF)
2797                 return (size_t)-1;
2798         }
2799     }
2800
2801     return len;
2802 }
2803
2804 // ----------------------------------------------------------------------------
2805 // globals
2806 // ----------------------------------------------------------------------------
2807
2808 #ifdef __WINDOWS__
2809     static wxMBConv_win32 wxConvLibcObj;
2810 #elif defined(__WXMAC__) && !defined(__MACH__)
2811     static wxMBConv_mac wxConvLibcObj ;
2812 #else
2813     static wxMBConvLibc wxConvLibcObj;
2814 #endif
2815
2816 static wxCSConv wxConvLocalObj(wxFONTENCODING_SYSTEM);
2817 static wxCSConv wxConvISO8859_1Obj(wxFONTENCODING_ISO8859_1);
2818 static wxMBConvUTF7 wxConvUTF7Obj;
2819 static wxMBConvUTF8 wxConvUTF8Obj;
2820
2821 WXDLLIMPEXP_DATA_BASE(wxMBConv&) wxConvLibc = wxConvLibcObj;
2822 WXDLLIMPEXP_DATA_BASE(wxCSConv&) wxConvLocal = wxConvLocalObj;
2823 WXDLLIMPEXP_DATA_BASE(wxCSConv&) wxConvISO8859_1 = wxConvISO8859_1Obj;
2824 WXDLLIMPEXP_DATA_BASE(wxMBConvUTF7&) wxConvUTF7 = wxConvUTF7Obj;
2825 WXDLLIMPEXP_DATA_BASE(wxMBConvUTF8&) wxConvUTF8 = wxConvUTF8Obj;
2826 WXDLLIMPEXP_DATA_BASE(wxMBConv *) wxConvCurrent = &wxConvLibcObj;
2827 WXDLLIMPEXP_DATA_BASE(wxMBConv *) wxConvFileName = &
2828 #ifdef __WXOSX__
2829                                     wxConvUTF8Obj;
2830 #else
2831                                     wxConvLibcObj;
2832 #endif
2833
2834
2835 #else // !wxUSE_WCHAR_T
2836
2837 // stand-ins in absence of wchar_t
2838 WXDLLIMPEXP_DATA_BASE(wxMBConv) wxConvLibc,
2839                                 wxConvISO8859_1,
2840                                 wxConvLocal,
2841                                 wxConvUTF8;
2842
2843 #endif // wxUSE_WCHAR_T/!wxUSE_WCHAR_T
2844
2845