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 // Modified by:
   6 // Created:     29/01/98
   7 // RCS-ID:      $Id$
   8 // Copyright:   (c) 1999 Ove Kaaven, Robert Roebling, Vadim Zeitlin, Vaclav Slavik
   9 // Licence:     wxWindows license
  10 /////////////////////////////////////////////////////////////////////////////
  11
  12 // ============================================================================
  13 // declarations
  14 // ============================================================================
  15
  16 // ----------------------------------------------------------------------------
  17 // headers
  18 // ----------------------------------------------------------------------------
  19
  20 #ifdef __GNUG__
  21   #pragma implementation "strconv.h"
  22 #endif
  23
  24 // For compilers that support precompilation, includes "wx.h".
  25 #include "wx/wxprec.h"
  26
  27 #ifdef __BORLANDC__
  28   #pragma hdrstop
  29 #endif
  30
  31 #ifndef WX_PRECOMP
  32     #include "wx/intl.h"
  33     #include "wx/log.h"
  34 #endif // WX_PRECOMP
  35
  36 #ifdef __WXMSW__
  37     #include "wx/msw/private.h"
  38 #endif
  39
  40 #include <errno.h>
  41 #include <ctype.h>
  42 #include <string.h>
  43 #include <stdlib.h>
  44
  45 #include "wx/strconv.h"
  46
  47 // ----------------------------------------------------------------------------
  48 // globals
  49 // ----------------------------------------------------------------------------
  50
  51 #if wxUSE_WCHAR_T
  52     WXDLLEXPORT_DATA(wxMBConv) wxConvLibc;
  53     WXDLLEXPORT_DATA(wxCSConv) wxConvLocal((const wxChar *)NULL);
  54 #else
  55     // stand-ins in absence of wchar_t
  56     WXDLLEXPORT_DATA(wxMBConv) wxConvLibc, wxConvFile;
  57 #endif // wxUSE_WCHAR_T
  58
  59 WXDLLEXPORT_DATA(wxMBConv *) wxConvCurrent = &wxConvLibc;
  60
  61 // ----------------------------------------------------------------------------
  62 // headers
  63 // ----------------------------------------------------------------------------
  64
  65 #if wxUSE_WCHAR_T
  66
  67 #ifdef __SALFORDC__
  68     #include <clib.h>
  69 #endif
  70
  71 #ifdef HAVE_ICONV
  72     #include <iconv.h>
  73 #endif
  74
  75 #include "wx/encconv.h"
  76 #include "wx/fontmap.h"
  77
  78 // ----------------------------------------------------------------------------
  79 // macros
  80 // ----------------------------------------------------------------------------
  81
  82 #define BSWAP_UCS4(str, len) { unsigned _c; for (_c=0; _c<len; _c++) str[_c]=wxUINT32_SWAP_ALWAYS(str[_c]); }
  83 #define BSWAP_UTF16(str, len) { unsigned _c; for (_c=0; _c<len; _c++) str[_c]=wxUINT16_SWAP_ALWAYS(str[_c]); }
  84
  85 // under Unix SIZEOF_WCHAR_T is defined by configure, but under other platforms
  86 // it might be not defined - assume the most common value
  87 #ifndef SIZEOF_WCHAR_T
  88     #define SIZEOF_WCHAR_T 2
  89 #endif // !defined(SIZEOF_WCHAR_T)
  90
  91 #if SIZEOF_WCHAR_T == 4
  92     #define WC_NAME         "UCS4"
  93     #define WC_BSWAP         BSWAP_UCS4
  94     #ifdef WORDS_BIGENDIAN
  95       #define WC_NAME_BEST  "UCS-4BE"
  96     #else
  97       #define WC_NAME_BEST  "UCS-4LE"
  98     #endif
  99 #elif SIZEOF_WCHAR_T == 2
 100     #define WC_NAME         "UTF16"
 101     #define WC_BSWAP         BSWAP_UTF16
 102     #define WC_UTF16
 103     #ifdef WORDS_BIGENDIAN
 104       #define WC_NAME_BEST  "UTF-16BE"
 105     #else
 106       #define WC_NAME_BEST  "UTF-16LE"
 107     #endif
 108 #else // sizeof(wchar_t) != 2 nor 4
 109     // I don't know what to do about this
 110     #error "Weird sizeof(wchar_t): please report your platform details to wx-users mailing list"
 111 #endif
 112
 113 // ============================================================================
 114 // implementation
 115 // ============================================================================
 116
 117 // ----------------------------------------------------------------------------
 118 // UTF-16 en/decoding
 119 // ----------------------------------------------------------------------------
 120
 121 #ifdef WC_UTF16
 122
 123 static size_t encode_utf16(wxUint32 input, wchar_t *output)
 124 {
 125     if (input<=0xffff)
 126     {
 127         if (output) *output++ = (wchar_t) input;
 128         return 1;
 129     }
 130     else if (input>=0x110000)
 131     {
 132         return (size_t)-1;
 133     }
 134     else
 135     {
 136         if (output)
 137         {
 138             *output++ = (wchar_t) ((input >> 10)+0xd7c0);
 139             *output++ = (wchar_t) ((input&0x3ff)+0xdc00);
 140         }
 141         return 2;
 142     }
 143 }
 144
 145 static size_t decode_utf16(const wchar_t* input, wxUint32& output)
 146 {
 147     if ((*input<0xd800) || (*input>0xdfff))
 148     {
 149         output = *input;
 150         return 1;
 151     }
 152     else if ((input[1]<0xdc00) || (input[1]>=0xdfff))
 153     {
 154         output = *input;
 155         return (size_t)-1;
 156     }
 157     else
 158     {
 159         output = ((input[0] - 0xd7c0) << 10) + (input[1] - 0xdc00);
 160         return 2;
 161     }
 162 }
 163
 164 #endif // WC_UTF16
 165
 166 // ----------------------------------------------------------------------------
 167 // wxMBConv
 168 // ----------------------------------------------------------------------------
 169
 170 size_t wxMBConv::MB2WC(wchar_t *buf, const char *psz, size_t n) const
 171 {
 172     return wxMB2WC(buf, psz, n);
 173 }
 174
 175 size_t wxMBConv::WC2MB(char *buf, const wchar_t *psz, size_t n) const
 176 {
 177     return wxWC2MB(buf, psz, n);
 178 }
 179
 180 const wxWCharBuffer wxMBConv::cMB2WC(const char *psz) const
 181 {
 182     if (psz)
 183     {
 184         size_t nLen = MB2WC((wchar_t *) NULL, psz, 0);
 185         if (nLen == (size_t)-1)
 186             return wxWCharBuffer((wchar_t *) NULL);
 187         wxWCharBuffer buf(nLen);
 188         MB2WC((wchar_t *)(const wchar_t *) buf, psz, nLen);
 189         return buf;
 190     }
 191     else
 192         return wxWCharBuffer((wchar_t *) NULL);
 193 }
 194
 195 const wxCharBuffer wxMBConv::cWC2MB(const wchar_t *psz) const
 196 {
 197     if (psz)
 198     {
 199         size_t nLen = WC2MB((char *) NULL, psz, 0);
 200         if (nLen == (size_t)-1)
 201             return wxCharBuffer((char *) NULL);
 202         wxCharBuffer buf(nLen);
 203         WC2MB((char *)(const char *) buf, psz, nLen);
 204         return buf;
 205     }
 206     else
 207         return wxCharBuffer((char *) NULL);
 208 }
 209
 210 // ----------------------------------------------------------------------------
 211 // standard file conversion
 212 // ----------------------------------------------------------------------------
 213
 214 WXDLLEXPORT_DATA(wxMBConvFile) wxConvFile;
 215
 216 // just use the libc conversion for now
 217 size_t wxMBConvFile::MB2WC(wchar_t *buf, const char *psz, size_t n) const
 218 {
 219     return wxMB2WC(buf, psz, n);
 220 }
 221
 222 size_t wxMBConvFile::WC2MB(char *buf, const wchar_t *psz, size_t n) const
 223 {
 224     return wxWC2MB(buf, psz, n);
 225 }
 226
 227 // ----------------------------------------------------------------------------
 228 // standard gdk conversion
 229 // ----------------------------------------------------------------------------
 230
 231 #ifdef __WXGTK12__
 232
 233 WXDLLEXPORT_DATA(wxMBConvGdk) wxConvGdk;
 234
 235 #include <gdk/gdk.h>
 236
 237 size_t wxMBConvGdk::MB2WC(wchar_t *buf, const char *psz, size_t n) const
 238 {
 239     if (buf)
 240     {
 241         return gdk_mbstowcs((GdkWChar *)buf, psz, n);
 242     }
 243     else
 244     {
 245         GdkWChar *nbuf = new GdkWChar[n=strlen(psz)];
 246         size_t len = gdk_mbstowcs(nbuf, psz, n);
 247         delete[] nbuf;
 248         return len;
 249     }
 250 }
 251
 252 size_t wxMBConvGdk::WC2MB(char *buf, const wchar_t *psz, size_t n) const
 253 {
 254     char *mbstr = gdk_wcstombs((GdkWChar *)psz);
 255     size_t len = mbstr ? strlen(mbstr) : 0;
 256     if (buf)
 257     {
 258         if (len > n)
 259             len = n;
 260         memcpy(buf, psz, len);
 261         if (len < n)
 262             buf[len] = 0;
 263     }
 264     return len;
 265 }
 266
 267 #endif // GTK > 1.0
 268
 269 // ----------------------------------------------------------------------------
 270 // UTF-7
 271 // ----------------------------------------------------------------------------
 272
 273 WXDLLEXPORT_DATA(wxMBConvUTF7) wxConvUTF7;
 274
 275 #if 0
 276 static char utf7_setD[]="ABCDEFGHIJKLMNOPQRSTUVWXYZ"
 277                         "abcdefghijklmnopqrstuvwxyz"
 278                         "0123456789'(),-./:?";
 279 static char utf7_setO[]="!\"#$%&*;<=>@[]^_`{|}";
 280 static char utf7_setB[]="ABCDEFGHIJKLMNOPQRSTUVWXYZ"
 281                         "abcdefghijklmnopqrstuvwxyz"
 282                         "0123456789+/";
 283 #endif
 284
 285 // TODO: write actual implementations of UTF-7 here
 286 size_t wxMBConvUTF7::MB2WC(wchar_t * WXUNUSED(buf),
 287                            const char * WXUNUSED(psz),
 288                            size_t WXUNUSED(n)) const
 289 {
 290   return 0;
 291 }
 292
 293 size_t wxMBConvUTF7::WC2MB(char * WXUNUSED(buf),
 294                            const wchar_t * WXUNUSED(psz),
 295                            size_t WXUNUSED(n)) const
 296 {
 297   return 0;
 298 }
 299
 300 // ----------------------------------------------------------------------------
 301 // UTF-8
 302 // ----------------------------------------------------------------------------
 303
 304 WXDLLEXPORT_DATA(wxMBConvUTF8) wxConvUTF8;
 305
 306 static wxUint32 utf8_max[]=
 307     { 0x7f, 0x7ff, 0xffff, 0x1fffff, 0x3ffffff, 0x7fffffff, 0xffffffff };
 308
 309 size_t wxMBConvUTF8::MB2WC(wchar_t *buf, const char *psz, size_t n) const
 310 {
 311     size_t len = 0;
 312
 313     while (*psz && ((!buf) || (len < n)))
 314     {
 315         unsigned char cc = *psz++, fc = cc;
 316         unsigned cnt;
 317         for (cnt = 0; fc & 0x80; cnt++)
 318             fc <<= 1;
 319         if (!cnt)
 320         {
 321             // plain ASCII char
 322             if (buf)
 323                 *buf++ = cc;
 324             len++;
 325         }
 326         else
 327         {
 328             cnt--;
 329             if (!cnt)
 330             {
 331                 // invalid UTF-8 sequence
 332                 return (size_t)-1;
 333             }
 334             else
 335             {
 336                 unsigned ocnt = cnt - 1;
 337                 wxUint32 res = cc & (0x3f >> cnt);
 338                 while (cnt--)
 339                 {
 340                     cc = *psz++;
 341                     if ((cc & 0xC0) != 0x80)
 342                     {
 343                         // invalid UTF-8 sequence
 344                         return (size_t)-1;
 345                     }
 346                     res = (res << 6) | (cc & 0x3f);
 347                 }
 348                 if (res <= utf8_max[ocnt])
 349                 {
 350                     // illegal UTF-8 encoding
 351                     return (size_t)-1;
 352                 }
 353 #ifdef WC_UTF16
 354                 size_t pa = encode_utf16(res, buf);
 355                 if (pa == (size_t)-1)
 356                   return (size_t)-1;
 357                 if (buf)
 358                     buf += pa;
 359                 len += pa;
 360 #else // !WC_UTF16
 361                 if (buf)
 362                     *buf++ = res;
 363                 len++;
 364 #endif // WC_UTF16/!WC_UTF16
 365             }
 366         }
 367     }
 368     if (buf && (len < n))
 369         *buf = 0;
 370     return len;
 371 }
 372
 373 size_t wxMBConvUTF8::WC2MB(char *buf, const wchar_t *psz, size_t n) const
 374 {
 375     size_t len = 0;
 376
 377     while (*psz && ((!buf) || (len < n)))
 378     {
 379         wxUint32 cc;
 380 #ifdef WC_UTF16
 381         size_t pa = decode_utf16(psz, cc);
 382         psz += (pa == (size_t)-1) ? 1 : pa;
 383 #else
 384         cc=(*psz++) & 0x7fffffff;
 385 #endif
 386         unsigned cnt;
 387         for (cnt = 0; cc > utf8_max[cnt]; cnt++) {}
 388         if (!cnt)
 389         {
 390             // plain ASCII char
 391             if (buf)
 392                 *buf++ = (char) cc;
 393             len++;
 394         }
 395
 396         else
 397         {
 398             len += cnt + 1;
 399             if (buf)
 400             {
 401                 *buf++ = (char) ((-128 >> cnt) | ((cc >> (cnt * 6)) & (0x3f >> cnt)));
 402                 while (cnt--)
 403                     *buf++ = (char) (0x80 | ((cc >> (cnt * 6)) & 0x3f));
 404             }
 405         }
 406     }
 407
 408     if (buf && (len<n)) *buf = 0;
 409     return len;
 410 }
 411
 412 // ============================================================================
 413 // wxCharacterSet and derived classes
 414 // ============================================================================
 415
 416 // ----------------------------------------------------------------------------
 417 // wxCharacterSet is the ABC for the classes below
 418 // ----------------------------------------------------------------------------
 419
 420 class wxCharacterSet
 421 {
 422 public:
 423     wxCharacterSet(const wxChar*name) : cname(name) {}
 424     virtual ~wxCharacterSet() {}
 425     virtual size_t MB2WC(wchar_t *buf, const char *psz, size_t n) = 0;
 426     virtual size_t WC2MB(char *buf, const wchar_t *psz, size_t n) = 0;
 427     virtual bool usable() const = 0;
 428 public:
 429     const wxChar*cname;
 430 };
 431
 432 // ----------------------------------------------------------------------------
 433 // ID_CharSet: implementation of wxCharacterSet using an existing wxMBConv
 434 // ----------------------------------------------------------------------------
 435
 436 class ID_CharSet : public wxCharacterSet
 437 {
 438 public:
 439     ID_CharSet(const wxChar *name, wxMBConv *cnv)
 440         : wxCharacterSet(name), work(cnv) {}
 441
 442     size_t MB2WC(wchar_t *buf, const char *psz, size_t n)
 443         { return work ? work->MB2WC(buf,psz,n) : (size_t)-1; }
 444
 445     size_t WC2MB(char *buf, const wchar_t *psz, size_t n)
 446         { return work ? work->WC2MB(buf,psz,n) : (size_t)-1; }
 447
 448     bool usable() const
 449         { return work!=NULL; }
 450 public:
 451     wxMBConv*work;
 452 };
 453
 454
 455 // ============================================================================
 456 // The classes doing conversion using the iconv_xxx() functions
 457 // ============================================================================
 458
 459 #ifdef HAVE_ICONV
 460
 461 // VS: glibc 2.1.3 is broken in that iconv() conversion to/from UCS4 fails with E2BIG
 462 //     if output buffer is _exactly_ as big as needed. Such case is (unless there's
 463 //     yet another bug in glibc) the only case when iconv() returns with (size_t)-1
 464 //     (which means error) and says there are 0 bytes left in the input buffer --
 465 //     when _real_ error occurs, bytes-left-in-input buffer is non-zero. Hence,
 466 //     this alternative test for iconv() failure.
 467 //     [This bug does not appear in glibc 2.2.]
 468 #if defined(__GLIBC__) && __GLIBC__ == 2 && __GLIBC_MINOR__ <= 1
 469 #define ICONV_FAILED(cres, bufLeft) ((cres == (size_t)-1) && \
 470                                      (errno != E2BIG || bufLeft != 0))
 471 #else
 472 #define ICONV_FAILED(cres, bufLeft)  (cres == (size_t)-1)
 473 #endif
 474
 475 #define ICONV_CHAR_CAST(x)  ((ICONV_CONST char **)(x))
 476
 477 // ----------------------------------------------------------------------------
 478 // IC_CharSet: encapsulates an iconv character set
 479 // ----------------------------------------------------------------------------
 480
 481 class IC_CharSet : public wxCharacterSet
 482 {
 483 public:
 484     IC_CharSet(const wxChar *name);
 485     virtual ~IC_CharSet();
 486
 487     virtual size_t MB2WC(wchar_t *buf, const char *psz, size_t n);
 488     virtual size_t WC2MB(char *buf, const wchar_t *psz, size_t n);
 489
 490     bool usable() const
 491         { return (m2w != (iconv_t)-1) && (w2m != (iconv_t)-1); }
 492
 493 protected:
 494     // the iconv handlers used to translate from multibyte to wide char and in
 495     // the other direction
 496     iconv_t m2w,
 497             w2m;
 498
 499 private:
 500     // the name (for iconv_open()) of a wide char charset - if none is
 501     // available on this machine, it will remain NULL
 502     static const char *ms_wcCharsetName;
 503
 504     // true if the wide char encoding we use (i.e. ms_wcCharsetName) has
 505     // different endian-ness than the native one
 506     static bool ms_wcNeedsSwap;
 507 };
 508
 509 const char *IC_CharSet::ms_wcCharsetName = NULL;
 510 bool IC_CharSet::ms_wcNeedsSwap = FALSE;
 511
 512 IC_CharSet::IC_CharSet(const wxChar *name)
 513           : wxCharacterSet(name)
 514 {
 515     // check for charset that represents wchar_t:
 516     if (ms_wcCharsetName == NULL)
 517     {
 518         ms_wcNeedsSwap = FALSE;
 519
 520         // try charset with explicit bytesex info (e.g. "UCS-4LE"):
 521         ms_wcCharsetName = WC_NAME_BEST;
 522         m2w = iconv_open(ms_wcCharsetName, wxConvLibc.cWX2MB(name));
 523
 524         if (m2w == (iconv_t)-1)
 525         {
 526             // try charset w/o bytesex info (e.g. "UCS4")
 527             // and check for bytesex ourselves:
 528             ms_wcCharsetName = WC_NAME;
 529             m2w = iconv_open(ms_wcCharsetName, wxConvLibc.cWX2MB(name));
 530
 531             // last bet, try if it knows WCHAR_T pseudo-charset
 532             if (m2w == (iconv_t)-1)
 533             {
 534                 ms_wcCharsetName = "WCHAR_T";
 535                 m2w = iconv_open(ms_wcCharsetName, wxConvLibc.cWX2MB(name));
 536             }
 537
 538             if (m2w != (iconv_t)-1)
 539             {
 540                 char    buf[2], *bufPtr;
 541                 wchar_t wbuf[2], *wbufPtr;
 542                 size_t  insz, outsz;
 543                 size_t  res;
 544
 545                 buf[0] = 'A';
 546                 buf[1] = 0;
 547                 wbuf[0] = 0;
 548                 insz = 2;
 549                 outsz = SIZEOF_WCHAR_T * 2;
 550                 wbufPtr = wbuf;
 551                 bufPtr = buf;
 552
 553                 res = iconv(m2w, ICONV_CHAR_CAST(&bufPtr), &insz,
 554                             (char**)&wbufPtr, &outsz);
 555
 556                 if (ICONV_FAILED(res, insz))
 557                 {
 558                     ms_wcCharsetName = NULL;
 559                     wxLogLastError(wxT("iconv"));
 560                     wxLogError(_("Convertion to charset '%s' doesn't work."), name);
 561                 }
 562                 else
 563                 {
 564                     ms_wcNeedsSwap = wbuf[0] != (wchar_t)buf[0];
 565                 }
 566             }
 567             else
 568             {
 569                 ms_wcCharsetName = NULL;
 570
 571                 // VS: we must not output an error here, since wxWindows will safely
 572                 //     fall back to using wxEncodingConverter.
 573                 wxLogTrace(wxT("strconv"), wxT("Impossible to convert to/from charset '%s' with iconv, falling back to wxEncodingConverter."), name);
 574                 //wxLogError(
 575             }
 576         }
 577         wxLogTrace(wxT("strconv"), wxT("wchar_t charset is '%s', needs swap: %i"), ms_wcCharsetName, ms_wcNeedsSwap);
 578     }
 579     else // we already have ms_wcCharsetName
 580     {
 581         m2w = iconv_open(ms_wcCharsetName, wxConvLibc.cWX2MB(name));
 582     }
 583
 584     // NB: don't ever pass NULL to iconv_open(), it may crash!
 585     if ( ms_wcCharsetName )
 586     {
 587         w2m = iconv_open(wxConvLibc.cWX2MB(name), ms_wcCharsetName);
 588     }
 589     else
 590     {
 591         w2m = (iconv_t)-1;
 592     }
 593 }
 594
 595 IC_CharSet::~IC_CharSet()
 596 {
 597     if ( m2w != (iconv_t)-1 )
 598         iconv_close(m2w);
 599     if ( w2m != (iconv_t)-1 )
 600         iconv_close(w2m);
 601 }
 602
 603 size_t IC_CharSet::MB2WC(wchar_t *buf, const char *psz, size_t n)
 604 {
 605     size_t inbuf = strlen(psz);
 606     size_t outbuf = n * SIZEOF_WCHAR_T;
 607     size_t res, cres;
 608     // VS: Use these instead of psz, buf because iconv() modifies its arguments:
 609     wchar_t *bufPtr = buf;
 610     const char *pszPtr = psz;
 611
 612     if (buf)
 613     {
 614         // have destination buffer, convert there
 615         cres = iconv(m2w,
 616                      ICONV_CHAR_CAST(&pszPtr), &inbuf,
 617                      (char**)&bufPtr, &outbuf);
 618         res = n - (outbuf / SIZEOF_WCHAR_T);
 619
 620         if (ms_wcNeedsSwap)
 621         {
 622             // convert to native endianness
 623             WC_BSWAP(buf /* _not_ bufPtr */, res)
 624         }
 625     }
 626     else
 627     {
 628         // no destination buffer... convert using temp buffer
 629         // to calculate destination buffer requirement
 630         wchar_t tbuf[8];
 631         res = 0;
 632         do {
 633             bufPtr = tbuf;
 634             outbuf = 8*SIZEOF_WCHAR_T;
 635
 636             cres = iconv(m2w,
 637                          ICONV_CHAR_CAST(&pszPtr), &inbuf,
 638                          (char**)&bufPtr, &outbuf );
 639
 640             res += 8-(outbuf/SIZEOF_WCHAR_T);
 641         } while ((cres==(size_t)-1) && (errno==E2BIG));
 642     }
 643
 644     if (ICONV_FAILED(cres, inbuf))
 645     {
 646         //VS: it is ok if iconv fails, hence trace only
 647         wxLogTrace(wxT("strconv"), wxT("iconv failed: %s"), wxSysErrorMsg(wxSysErrorCode()));
 648         return (size_t)-1;
 649     }
 650
 651     return res;
 652 }
 653
 654 size_t IC_CharSet::WC2MB(char *buf, const wchar_t *psz, size_t n)
 655 {
 656 #if defined(__BORLANDC__) && (__BORLANDC__ > 0x530)
 657     size_t inbuf = std::wcslen(psz) * SIZEOF_WCHAR_T;
 658 #else
 659     size_t inbuf = ::wcslen(psz) * SIZEOF_WCHAR_T;
 660 #endif
 661     size_t outbuf = n;
 662     size_t res, cres;
 663
 664     wchar_t *tmpbuf = 0;
 665
 666     if (ms_wcNeedsSwap)
 667     {
 668         // need to copy to temp buffer to switch endianness
 669         // this absolutely doesn't rock!
 670         // (no, doing WC_BSWAP twice on the original buffer won't help, as it
 671         //  could be in read-only memory, or be accessed in some other thread)
 672         tmpbuf=(wchar_t*)malloc((inbuf+1)*SIZEOF_WCHAR_T);
 673         memcpy(tmpbuf,psz,(inbuf+1)*SIZEOF_WCHAR_T);
 674         WC_BSWAP(tmpbuf, inbuf)
 675         psz=tmpbuf;
 676     }
 677
 678     if (buf)
 679     {
 680         // have destination buffer, convert there
 681         cres = iconv( w2m, ICONV_CHAR_CAST(&psz), &inbuf, &buf, &outbuf );
 682
 683         res = n-outbuf;
 684     }
 685     else
 686     {
 687         // no destination buffer... convert using temp buffer
 688         // to calculate destination buffer requirement
 689         char tbuf[16];
 690         res = 0;
 691         do {
 692             buf = tbuf; outbuf = 16;
 693
 694             cres = iconv( w2m, ICONV_CHAR_CAST(&psz), &inbuf, &buf, &outbuf );
 695
 696             res += 16 - outbuf;
 697         } while ((cres==(size_t)-1) && (errno==E2BIG));
 698     }
 699
 700     if (ms_wcNeedsSwap)
 701     {
 702         free(tmpbuf);
 703     }
 704
 705     if (ICONV_FAILED(cres, inbuf))
 706     {
 707         //VS: it is ok if iconv fails, hence trace only
 708         wxLogTrace(wxT("strconv"), wxT("iconv failed: %s"), wxSysErrorMsg(wxSysErrorCode()));
 709         return (size_t)-1;
 710     }
 711
 712     return res;
 713 }
 714
 715 #endif // HAVE_ICONV
 716
 717 // ============================================================================
 718 // Win32 conversion classes
 719 // ============================================================================
 720
 721 #if defined(__WIN32__) && !defined(__WXMICROWIN__)
 722
 723 extern long wxCharsetToCodepage(const wxChar *charset); // from utils.cpp
 724
 725 class CP_CharSet : public wxCharacterSet
 726 {
 727 public:
 728     CP_CharSet(const wxChar* name)
 729         : wxCharacterSet(name)
 730         {
 731             m_CodePage = wxCharsetToCodepage(name);
 732         }
 733
 734     size_t MB2WC(wchar_t *buf, const char *psz, size_t n)
 735     {
 736         size_t len =
 737             MultiByteToWideChar(m_CodePage, 0, psz, -1, buf, buf ? n : 0);
 738         //VS: returns # of written chars for buf!=NULL and *size*
 739         //    needed buffer for buf==NULL
 740         return len ? (buf ? len : len-1) : (size_t)-1;
 741     }
 742
 743     size_t WC2MB(char *buf, const wchar_t *psz, size_t n)
 744     {
 745         size_t len = WideCharToMultiByte(m_CodePage, 0, psz, -1, buf,
 746                                          buf ? n : 0, NULL, NULL);
 747         //VS: returns # of written chars for buf!=NULL and *size*
 748         //    needed buffer for buf==NULL
 749         return len ? (buf ? len : len-1) : (size_t)-1;
 750     }
 751
 752     bool usable() const
 753         { return m_CodePage != -1; }
 754
 755 public:
 756     long m_CodePage;
 757 };
 758 #endif // __WIN32__
 759
 760 // ============================================================================
 761 // wxEncodingConverter based conversion classes
 762 // ============================================================================
 763
 764 #if wxUSE_FONTMAP
 765
 766 class EC_CharSet : public wxCharacterSet
 767 {
 768 public:
 769     // temporarily just use wxEncodingConverter stuff,
 770     // so that it works while a better implementation is built
 771     EC_CharSet(const wxChar* name) : wxCharacterSet(name),
 772                                      enc(wxFONTENCODING_SYSTEM)
 773     {
 774         if (name)
 775             enc = wxTheFontMapper->CharsetToEncoding(name, FALSE);
 776
 777         m_ok = m2w.Init(enc, wxFONTENCODING_UNICODE) &&
 778                w2m.Init(wxFONTENCODING_UNICODE, enc);
 779     }
 780
 781     size_t MB2WC(wchar_t *buf, const char *psz, size_t WXUNUSED(n))
 782     {
 783         size_t inbuf = strlen(psz);
 784         if (buf)
 785             m2w.Convert(psz,buf);
 786         return inbuf;
 787     }
 788
 789     size_t WC2MB(char *buf, const wchar_t *psz, size_t WXUNUSED(n))
 790     {
 791 #if ( defined(__BORLANDC__) && (__BORLANDC__ > 0x530) ) \
 792     || ( defined(__MWERKS__) && defined(__WXMSW__) )
 793         size_t inbuf = std::wcslen(psz);
 794 #else
 795         size_t inbuf = ::wcslen(psz);
 796 #endif
 797         if (buf)
 798             w2m.Convert(psz,buf);
 799
 800         return inbuf;
 801     }
 802
 803     bool usable() const { return m_ok; }
 804
 805 public:
 806     wxFontEncoding enc;
 807     wxEncodingConverter m2w, w2m;
 808
 809     // were we initialized successfully?
 810     bool m_ok;
 811 };
 812
 813 #endif // wxUSE_FONTMAP
 814
 815 // ----------------------------------------------------------------------------
 816 // the function creating the wxCharacterSet for the specified charset on the
 817 // current system, trying all possibilities
 818 // ----------------------------------------------------------------------------
 819
 820 static wxCharacterSet *wxGetCharacterSet(const wxChar *name)
 821 {
 822     // check for the special case of ASCII charset
 823 #if wxUSE_FONTMAP
 824     if ( wxTheFontMapper->CharsetToEncoding(name) == wxFONTENCODING_DEFAULT )
 825 #else // wxUSE_FONTMAP
 826     if ( !name )
 827 #endif // wxUSE_FONTMAP/!wxUSE_FONTMAP
 828     {
 829         // don't convert at all
 830         return NULL;
 831     }
 832
 833     // the test above must have taken care of this case
 834     wxCHECK_MSG( name, NULL, _T("NULL name must be wxFONTENCODING_DEFAULT") );
 835
 836     wxCharacterSet *cset;
 837
 838     if ( wxStricmp(name, wxT("UTF8")) == 0 || wxStricmp(name, wxT("UTF-8")) == 0)
 839     {
 840         cset = new ID_CharSet(name, &wxConvUTF8);
 841     }
 842     else
 843     {
 844 #ifdef HAVE_ICONV
 845         cset = new IC_CharSet(name);
 846 #else // !HAVE_ICONV
 847         cset = NULL;
 848 #endif // HAVE_ICONV/!HAVE_ICONV
 849     }
 850
 851     // it can only be NULL in this case
 852 #ifndef HAVE_ICONV
 853     if ( cset )
 854 #endif // !HAVE_ICONV
 855     {
 856         if ( cset->usable() )
 857             return cset;
 858
 859         delete cset;
 860         cset = NULL;
 861     }
 862
 863 #if defined(__WIN32__) && !defined(__WXMICROWIN__)
 864     cset = new CP_CharSet(name);
 865     if ( cset->usable() )
 866         return cset;
 867
 868     delete cset;
 869     cset = NULL;
 870 #endif // __WIN32__
 871
 872 #if wxUSE_FONTMAP
 873     cset = new EC_CharSet(name);
 874     if ( cset->usable() )
 875         return cset;
 876
 877     delete cset;
 878     cset = NULL;
 879 #endif // wxUSE_FONTMAP
 880
 881     wxLogError(_("Cannot convert from encoding '%s'!"), name);
 882
 883     return NULL;
 884 }
 885
 886 // ============================================================================
 887 // wxCSConv implementation
 888 // ============================================================================
 889
 890 wxCSConv::wxCSConv(const wxChar *charset)
 891 {
 892     m_name = (wxChar *)NULL;
 893     m_cset = (wxCharacterSet *) NULL;
 894     m_deferred = TRUE;
 895
 896     SetName(charset);
 897 }
 898
 899 wxCSConv::~wxCSConv()
 900 {
 901     free(m_name);
 902     delete m_cset;
 903 }
 904
 905 void wxCSConv::SetName(const wxChar *charset)
 906 {
 907     if (charset)
 908     {
 909         m_name = wxStrdup(charset);
 910         m_deferred = TRUE;
 911     }
 912 }
 913
 914 void wxCSConv::LoadNow()
 915 {
 916     if (m_deferred)
 917     {
 918         if ( !m_name )
 919         {
 920             wxString name = wxLocale::GetSystemEncodingName();
 921             if ( !name.empty() )
 922                 SetName(name);
 923         }
 924
 925         // wxGetCharacterSet() complains about NULL name
 926         m_cset = m_name ? wxGetCharacterSet(m_name) : NULL;
 927         m_deferred = FALSE;
 928     }
 929 }
 930
 931 size_t wxCSConv::MB2WC(wchar_t *buf, const char *psz, size_t n) const
 932 {
 933     ((wxCSConv *)this)->LoadNow(); // discard constness
 934
 935     if (m_cset)
 936         return m_cset->MB2WC(buf, psz, n);
 937
 938     // latin-1 (direct)
 939     size_t len = strlen(psz);
 940
 941     if (buf)
 942     {
 943         for (size_t c = 0; c <= len; c++)
 944             buf[c] = (unsigned char)(psz[c]);
 945     }
 946
 947     return len;
 948 }
 949
 950 size_t wxCSConv::WC2MB(char *buf, const wchar_t *psz, size_t n) const
 951 {
 952     ((wxCSConv *)this)->LoadNow(); // discard constness
 953
 954     if (m_cset)
 955         return m_cset->WC2MB(buf, psz, n);
 956
 957     // latin-1 (direct)
 958 #if ( defined(__BORLANDC__) && (__BORLANDC__ > 0x530) ) \
 959     || ( defined(__MWERKS__) && defined(__WXMSW__) )
 960     size_t len=std::wcslen(psz);
 961 #else
 962     size_t len=::wcslen(psz);
 963 #endif
 964     if (buf)
 965     {
 966         for (size_t c = 0; c <= len; c++)
 967             buf[c] = (psz[c] > 0xff) ? '?' : psz[c];
 968     }
 969
 970     return len;
 971 }
 972
 973 #endif // wxUSE_WCHAR_T
 974
 975