]> git.saurik.com Git - wxWidgets.git/blobdiff - src/common/strconv.cpp
Allow wx parents a chance to handle events fired by non-wx children. (i.e. wxActiveX...
[wxWidgets.git] / src / common / strconv.cpp
index 792d2caa50795435846292a88d69facd6146399b..c55469c0a2be23b417715049b71c1d4b76b5844a 100644 (file)
@@ -1,5 +1,5 @@
 /////////////////////////////////////////////////////////////////////////////
-// Name:        strconv.cpp
+// Name:        src/common/strconv.cpp
 // Purpose:     Unicode conversion classes
 // Author:      Ove Kaaven, Robert Roebling, Vadim Zeitlin, Vaclav Slavik,
 //              Ryan Norton, Fredrik Roubert (UTF7)
 // headers
 // ----------------------------------------------------------------------------
 
-#if defined(__GNUG__) && !defined(NO_GCC_PRAGMA)
-  #pragma implementation "strconv.h"
-#endif
-
 // For compilers that support precompilation, includes "wx.h".
 #include "wx/wxprec.h"
 
 
 #if wxUSE_WCHAR_T
 
-#ifdef __WXMSW__
-    #include "wx/msw/private.h"
-#endif
-
 #ifdef __WINDOWS__
+    #include "wx/msw/private.h"
     #include "wx/msw/missing.h"
 #endif
 
     #define wxHAVE_WIN32_MB2WC
 #endif // __WIN32__ but !__WXMICROWIN__
 
-// ----------------------------------------------------------------------------
-// headers
-// ----------------------------------------------------------------------------
-
 #ifdef __SALFORDC__
     #include <clib.h>
 #endif
 #include "wx/utils.h"
 
 #ifdef __WXMAC__
+#ifndef __DARWIN__
 #include <ATSUnicode.h>
 #include <TextCommon.h>
 #include <TextEncodingConverter.h>
+#endif
 
 #include  "wx/mac/private.h"  // includes mac headers
 #endif
-// ----------------------------------------------------------------------------
-// macros
-// ----------------------------------------------------------------------------
 
-#define BSWAP_UCS4(str, len) { unsigned _c; for (_c=0; _c<len; _c++) str[_c]=wxUINT32_SWAP_ALWAYS(str[_c]); }
-#define BSWAP_UTF16(str, len) { unsigned _c; for (_c=0; _c<len; _c++) str[_c]=wxUINT16_SWAP_ALWAYS(str[_c]); }
+#define TRACE_STRCONV _T("strconv")
 
-#if SIZEOF_WCHAR_T == 4
-    #define WC_NAME         "UCS4"
-    #define WC_BSWAP         BSWAP_UCS4
-    #ifdef WORDS_BIGENDIAN
-      #define WC_NAME_BEST  "UCS-4BE"
-    #else
-      #define WC_NAME_BEST  "UCS-4LE"
-    #endif
-#elif SIZEOF_WCHAR_T == 2
-    #define WC_NAME         "UTF16"
-    #define WC_BSWAP         BSWAP_UTF16
+#if SIZEOF_WCHAR_T == 2
     #define WC_UTF16
-    #ifdef WORDS_BIGENDIAN
-      #define WC_NAME_BEST  "UTF-16BE"
-    #else
-      #define WC_NAME_BEST  "UTF-16LE"
-    #endif
-#else // sizeof(wchar_t) != 2 nor 4
-    // does this ever happen?
-    #error "Unknown sizeof(wchar_t): please report this to wx-dev@lists.wxwindows.org"
 #endif
 
 // ============================================================================
@@ -152,7 +121,7 @@ static size_t decode_utf16(const wxUint16* input, wxUint32& output)
         output = *input;
         return 1;
     }
-    else if ((input[1]<0xdc00) || (input[1]>=0xdfff))
+    else if ((input[1]<0xdc00) || (input[1]>0xdfff))
     {
         output = *input;
         return (size_t)-1;
@@ -270,15 +239,15 @@ const wxWCharBuffer wxMBConv::cMB2WC(const char *szString, size_t nStringLen, si
         }
 
         //Increment to next (sub)string
-        //Note that we have to use strlen here instead of nLen
-        //here because XX2XX gives us the size of the output buffer,
-        //not neccessarly the length of the string
+        //Note that we have to use strlen instead of nLen here
+        //because XX2XX gives us the size of the output buffer,
+        //which is not necessarily the length of the string
         szPos += strlen(szPos) + 1;
     }
 
     //success - return actual length and the buffer
     *pOutSize = nActualLength;
-    return theBuffer;  
+    return theBuffer;
 }
 
 const wxCharBuffer wxMBConv::cWC2MB(const wchar_t *szString, size_t nStringLen, size_t* pOutSize) const
@@ -313,7 +282,7 @@ const wxCharBuffer wxMBConv::cWC2MB(const wchar_t *szString, size_t nStringLen,
 
         //Increase the actual length (+1 for current null character)
         nActualLength += nLen + 1;
-        
+
         //if buffer too big, realloc the buffer
         if (nActualLength > (nCurrentSize+1))
         {
@@ -332,15 +301,15 @@ const wxCharBuffer wxMBConv::cWC2MB(const wchar_t *szString, size_t nStringLen,
         }
 
         //Increment to next (sub)string
-        //Note that we have to use wxWcslen here instead of nLen
-        //here because XX2XX gives us the size of the output buffer,
-        //not neccessarly the length of the string
+        //Note that we have to use wxWcslen instead of nLen here
+        //because XX2XX gives us the size of the output buffer,
+        //which is not necessarily the length of the string
         szPos += wxWcslen(szPos) + 1;
     }
 
     //success - return actual length and the buffer
     *pOutSize = nActualLength;
-    return theBuffer;  
+    return theBuffer;
 }
 
 // ----------------------------------------------------------------------------
@@ -356,8 +325,42 @@ size_t wxMBConvLibc::WC2MB(char *buf, const wchar_t *psz, size_t n) const
 {
     return wxWC2MB(buf, psz, n);
 }
+
+#ifdef __UNIX__
+
 // ----------------------------------------------------------------------------
-// UTF-7 
+// wxConvBrokenFileNames
+// ----------------------------------------------------------------------------
+
+wxConvBrokenFileNames::wxConvBrokenFileNames(const wxChar *charset)
+{
+    if ( !charset || wxStricmp(charset, _T("UTF-8")) == 0
+                  || wxStricmp(charset, _T("UTF8")) == 0  )
+        m_conv = new wxMBConvUTF8(wxMBConvUTF8::MAP_INVALID_UTF8_TO_OCTAL);
+    else
+        m_conv = new wxCSConv(charset);
+}
+
+size_t
+wxConvBrokenFileNames::MB2WC(wchar_t *outputBuf,
+                             const char *psz,
+                             size_t outputSize) const
+{
+    return m_conv->MB2WC( outputBuf, psz, outputSize );
+}
+
+size_t
+wxConvBrokenFileNames::WC2MB(char *outputBuf,
+                             const wchar_t *psz,
+                             size_t outputSize) const
+{
+    return m_conv->WC2MB( outputBuf, psz, outputSize );
+}
+
+#endif
+
+// ----------------------------------------------------------------------------
+// UTF-7
 // ----------------------------------------------------------------------------
 
 // Implementation (C) 2004 Fredrik Roubert
@@ -510,7 +513,7 @@ size_t wxMBConvUTF7::WC2MB(char *buf, const wchar_t *psz, size_t n) const
         }
 #ifndef WC_UTF16
         else if (((wxUint32)cc) > 0xffff)
-           {
+        {
             // no surrogate pair generation (yet?)
             return (size_t)-1;
         }
@@ -567,12 +570,19 @@ size_t wxMBConvUTF7::WC2MB(char *buf, const wchar_t *psz, size_t n) const
 static wxUint32 utf8_max[]=
     { 0x7f, 0x7ff, 0xffff, 0x1fffff, 0x3ffffff, 0x7fffffff, 0xffffffff };
 
+// boundaries of the private use area we use to (temporarily) remap invalid
+// characters invalid in a UTF-8 encoded string
+const wxUint32 wxUnicodePUA = 0x100000;
+const wxUint32 wxUnicodePUAEnd = wxUnicodePUA + 256;
+
 size_t wxMBConvUTF8::MB2WC(wchar_t *buf, const char *psz, size_t n) const
 {
     size_t len = 0;
 
     while (*psz && ((!buf) || (len < n)))
     {
+        const char *opsz = psz;
+        bool invalid = false;
         unsigned char cc = *psz++, fc = cc;
         unsigned cnt;
         for (cnt = 0; fc & 0x80; cnt++)
@@ -583,6 +593,15 @@ size_t wxMBConvUTF8::MB2WC(wchar_t *buf, const char *psz, size_t n) const
             if (buf)
                 *buf++ = cc;
             len++;
+
+            // escape the escape character for octal escapes
+            if ((m_options & MAP_INVALID_UTF8_TO_OCTAL)
+                    && cc == '\\' && (!buf || len < n))
+            {
+                if (buf)
+                    *buf++ = cc;
+                len++;
+            }
         }
         else
         {
@@ -590,7 +609,7 @@ size_t wxMBConvUTF8::MB2WC(wchar_t *buf, const char *psz, size_t n) const
             if (!cnt)
             {
                 // invalid UTF-8 sequence
-                return (size_t)-1;
+                invalid = true;
             }
             else
             {
@@ -598,32 +617,93 @@ size_t wxMBConvUTF8::MB2WC(wchar_t *buf, const char *psz, size_t n) const
                 wxUint32 res = cc & (0x3f >> cnt);
                 while (cnt--)
                 {
-                    cc = *psz++;
+                    cc = *psz;
                     if ((cc & 0xC0) != 0x80)
                     {
                         // invalid UTF-8 sequence
-                        return (size_t)-1;
+                        invalid = true;
+                        break;
                     }
+                    psz++;
                     res = (res << 6) | (cc & 0x3f);
                 }
-                if (res <= utf8_max[ocnt])
+                if (invalid || res <= utf8_max[ocnt])
                 {
                     // illegal UTF-8 encoding
-                    return (size_t)-1;
+                    invalid = true;
+                }
+                else if ((m_options & MAP_INVALID_UTF8_TO_PUA) &&
+                        res >= wxUnicodePUA && res < wxUnicodePUAEnd)
+                {
+                    // if one of our PUA characters turns up externally
+                    // it must also be treated as an illegal sequence
+                    // (a bit like you have to escape an escape character)
+                    invalid = true;
                 }
+                else
+                {
 #ifdef WC_UTF16
-                // cast is ok because wchar_t == wxUuint16 if WC_UTF16
-                size_t pa = encode_utf16(res, (wxUint16 *)buf);
-                if (pa == (size_t)-1)
-                  return (size_t)-1;
-                if (buf)
-                    buf += pa;
-                len += pa;
+                    // cast is ok because wchar_t == wxUuint16 if WC_UTF16
+                    size_t pa = encode_utf16(res, (wxUint16 *)buf);
+                    if (pa == (size_t)-1)
+                    {
+                        invalid = true;
+                    }
+                    else
+                    {
+                        if (buf)
+                            buf += pa;
+                        len += pa;
+                    }
 #else // !WC_UTF16
-                if (buf)
-                    *buf++ = res;
-                len++;
+                    if (buf)
+                        *buf++ = (wchar_t)res;
+                    len++;
 #endif // WC_UTF16/!WC_UTF16
+                }
+            }
+            if (invalid)
+            {
+                if (m_options & MAP_INVALID_UTF8_TO_PUA)
+                {
+                    while (opsz < psz && (!buf || len < n))
+                    {
+#ifdef WC_UTF16
+                        // cast is ok because wchar_t == wxUuint16 if WC_UTF16
+                        size_t pa = encode_utf16((unsigned char)*opsz + wxUnicodePUA, (wxUint16 *)buf);
+                        wxASSERT(pa != (size_t)-1);
+                        if (buf)
+                            buf += pa;
+                        opsz++;
+                        len += pa;
+#else
+                        if (buf)
+                            *buf++ = (wchar_t)(wxUnicodePUA + (unsigned char)*opsz);
+                        opsz++;
+                        len++;
+#endif
+                    }
+                }
+                else if (m_options & MAP_INVALID_UTF8_TO_OCTAL)
+                {
+                    while (opsz < psz && (!buf || len < n))
+                    {
+                        if ( buf && len + 3 < n )
+                        {
+                            unsigned char on = *opsz;
+                            *buf++ = L'\\';
+                            *buf++ = (wchar_t)( L'0' + on / 0100 );
+                            *buf++ = (wchar_t)( L'0' + (on % 0100) / 010 );
+                            *buf++ = (wchar_t)( L'0' + on % 010 );
+                        }
+                        opsz++;
+                        len += 4;
+                    }
+                }
+                else // MAP_INVALID_UTF8_NOT
+                {
+                    return (size_t)-1;
+                }
             }
         }
     }
@@ -632,6 +712,11 @@ size_t wxMBConvUTF8::MB2WC(wchar_t *buf, const char *psz, size_t n) const
     return len;
 }
 
+static inline bool isoctal(wchar_t wch)
+{
+    return L'0' <= wch && wch <= L'7';
+}
+
 size_t wxMBConvUTF8::WC2MB(char *buf, const wchar_t *psz, size_t n) const
 {
     size_t len = 0;
@@ -646,36 +731,67 @@ size_t wxMBConvUTF8::WC2MB(char *buf, const wchar_t *psz, size_t n) const
 #else
         cc=(*psz++) & 0x7fffffff;
 #endif
-        unsigned cnt;
-        for (cnt = 0; cc > utf8_max[cnt]; cnt++) {}
-        if (!cnt)
+
+        if ( (m_options & MAP_INVALID_UTF8_TO_PUA)
+                && cc >= wxUnicodePUA && cc < wxUnicodePUAEnd )
         {
-            // plain ASCII char
             if (buf)
-                *buf++ = (char) cc;
+                *buf++ = (char)(cc - wxUnicodePUA);
             len++;
         }
+        else if ( (m_options & MAP_INVALID_UTF8_TO_OCTAL)
+                    && cc == L'\\' && psz[0] == L'\\' )
+        {
+            if (buf)
+                *buf++ = (char)cc;
+            psz++;
+            len++;
+        }
+        else if ( (m_options & MAP_INVALID_UTF8_TO_OCTAL) &&
+                    cc == L'\\' &&
+                        isoctal(psz[0]) && isoctal(psz[1]) && isoctal(psz[2]) )
+        {
+            if (buf)
+            {
+                *buf++ = (char) ((psz[0] - L'0')*0100 +
+                                 (psz[1] - L'0')*010 +
+                                 (psz[2] - L'0'));
+            }
 
+            psz += 3;
+            len++;
+        }
         else
         {
-            len += cnt + 1;
-            if (buf)
+            unsigned cnt;
+            for (cnt = 0; cc > utf8_max[cnt]; cnt++) {}
+            if (!cnt)
             {
-                *buf++ = (char) ((-128 >> cnt) | ((cc >> (cnt * 6)) & (0x3f >> cnt)));
-                while (cnt--)
-                    *buf++ = (char) (0x80 | ((cc >> (cnt * 6)) & 0x3f));
+                // plain ASCII char
+                if (buf)
+                    *buf++ = (char) cc;
+                len++;
+            }
+
+            else
+            {
+                len += cnt + 1;
+                if (buf)
+                {
+                    *buf++ = (char) ((-128 >> cnt) | ((cc >> (cnt * 6)) & (0x3f >> cnt)));
+                    while (cnt--)
+                        *buf++ = (char) (0x80 | ((cc >> (cnt * 6)) & 0x3f));
+                }
             }
         }
     }
 
-    if (buf && (len<n)) *buf = 0;
+    if (buf && (len<n))
+        *buf = 0;
 
     return len;
 }
 
-
-
-
 // ----------------------------------------------------------------------------
 // UTF-16
 // ----------------------------------------------------------------------------
@@ -790,7 +906,7 @@ size_t wxMBConvUTF16straight::MB2WC(wchar_t *buf, const char *psz, size_t n) con
             return pa;
 
         if (buf)
-            *buf++ = cc;
+            *buf++ = (wchar_t)cc;
         len++;
         psz += pa * sizeof(wxUint16);
     }
@@ -850,7 +966,7 @@ size_t wxMBConvUTF16swap::MB2WC(wchar_t *buf, const char *psz, size_t n) const
             return pa;
 
         if (buf)
-            *buf++ = cc;
+            *buf++ = (wchar_t)cc;
 
         len++;
         psz += pa * sizeof(wxUint16);
@@ -1051,7 +1167,7 @@ size_t wxMBConvUTF32straight::MB2WC(wchar_t *buf, const char *psz, size_t n) con
     while (*(wxUint32*)psz && (!buf || len < n))
     {
         if (buf)
-            *buf++ = *(wxUint32*)psz;
+            *buf++ = (wchar_t)(*(wxUint32*)psz);
         len++;
         psz += sizeof(wxUint32);
     }
@@ -1164,6 +1280,19 @@ size_t wxMBConvUTF32swap::WC2MB(char *buf, const wchar_t *psz, size_t n) const
 
 #define ICONV_CHAR_CAST(x)  ((ICONV_CONST char **)(x))
 
+#define ICONV_T_INVALID ((iconv_t)-1)
+
+#if SIZEOF_WCHAR_T == 4
+    #define WC_BSWAP    wxUINT32_SWAP_ALWAYS
+    #define WC_ENC      wxFONTENCODING_UTF32
+#elif SIZEOF_WCHAR_T == 2
+    #define WC_BSWAP    wxUINT16_SWAP_ALWAYS
+    #define WC_ENC      wxFONTENCODING_UTF16
+#else // sizeof(wchar_t) != 2 nor 4
+    // does this ever happen?
+    #error "Unknown sizeof(wchar_t): please report this to wx-dev@lists.wxwindows.org"
+#endif
+
 // ----------------------------------------------------------------------------
 // wxMBConv_iconv: encapsulates an iconv character set
 // ----------------------------------------------------------------------------
@@ -1178,7 +1307,7 @@ public:
     virtual size_t WC2MB(char *buf, const wchar_t *psz, size_t n) const;
 
     bool IsOk() const
-        { return (m2w != (iconv_t)-1) && (w2m != (iconv_t)-1); }
+        { return (m2w != ICONV_T_INVALID) && (w2m != ICONV_T_INVALID); }
 
 protected:
     // the iconv handlers used to translate from multibyte to wide char and in
@@ -1193,108 +1322,147 @@ protected:
 private:
     // the name (for iconv_open()) of a wide char charset -- if none is
     // available on this machine, it will remain NULL
-    static const char *ms_wcCharsetName;
+    static wxString ms_wcCharsetName;
 
     // true if the wide char encoding we use (i.e. ms_wcCharsetName) has
     // different endian-ness than the native one
     static bool ms_wcNeedsSwap;
 };
 
-const char *wxMBConv_iconv::ms_wcCharsetName = NULL;
+// make the constructor available for unit testing
+WXDLLIMPEXP_BASE wxMBConv* new_wxMBConv_iconv( const wxChar* name )
+{
+    wxMBConv_iconv* result = new wxMBConv_iconv( name );
+    if ( !result->IsOk() )
+    {
+        delete result;
+        return 0;
+    }
+    return result;
+}
+
+wxString wxMBConv_iconv::ms_wcCharsetName;
 bool wxMBConv_iconv::ms_wcNeedsSwap = false;
 
 wxMBConv_iconv::wxMBConv_iconv(const wxChar *name)
 {
-    // Do it the hard way
-    char cname[100];
-    for (size_t i = 0; i < wxStrlen(name)+1; i++)
-        cname[i] = (char) name[i];
+    // iconv operates with chars, not wxChars, but luckily it uses only ASCII
+    // names for the charsets
+    const wxCharBuffer cname(wxString(name).ToAscii());
 
     // check for charset that represents wchar_t:
-    if (ms_wcCharsetName == NULL)
+    if ( ms_wcCharsetName.empty() )
     {
-        ms_wcNeedsSwap = false;
+        wxLogTrace(TRACE_STRCONV, _T("Looking for wide char codeset:"));
 
-        // try charset with explicit bytesex info (e.g. "UCS-4LE"):
-        ms_wcCharsetName = WC_NAME_BEST;
-        m2w = iconv_open(ms_wcCharsetName, cname);
+#if wxUSE_FONTMAP
+        const wxChar **names = wxFontMapperBase::GetAllEncodingNames(WC_ENC);
+#else // !wxUSE_FONTMAP
+        static const wxChar *names[] =
+        {
+#if SIZEOF_WCHAR_T == 4
+            _T("UCS-4"),
+#elif SIZEOF_WCHAR_T = 2
+            _T("UCS-2"),
+#endif
+            NULL
+        };
+#endif // wxUSE_FONTMAP/!wxUSE_FONTMAP
 
-        if (m2w == (iconv_t)-1)
+        for ( ; *names && ms_wcCharsetName.empty(); ++names )
         {
-            // try charset w/o bytesex info (e.g. "UCS4")
-            // and check for bytesex ourselves:
-            ms_wcCharsetName = WC_NAME;
-            m2w = iconv_open(ms_wcCharsetName, cname);
+            const wxString nameCS(*names);
 
-            // last bet, try if it knows WCHAR_T pseudo-charset
-            if (m2w == (iconv_t)-1)
-            {
-                ms_wcCharsetName = "WCHAR_T";
-                m2w = iconv_open(ms_wcCharsetName, cname);
-            }
+            // first try charset with explicit bytesex info (e.g. "UCS-4LE"):
+            wxString nameXE(nameCS);
+            #ifdef WORDS_BIGENDIAN
+                nameXE += _T("BE");
+            #else // little endian
+                nameXE += _T("LE");
+            #endif
+
+            wxLogTrace(TRACE_STRCONV, _T("  trying charset \"%s\""),
+                       nameXE.c_str());
 
-            if (m2w != (iconv_t)-1)
+            m2w = iconv_open(nameXE.ToAscii(), cname);
+            if ( m2w == ICONV_T_INVALID )
             {
-                char    buf[2], *bufPtr;
-                wchar_t wbuf[2], *wbufPtr;
-                size_t  insz, outsz;
-                size_t  res;
-
-                buf[0] = 'A';
-                buf[1] = 0;
-                wbuf[0] = 0;
-                insz = 2;
-                outsz = SIZEOF_WCHAR_T * 2;
-                wbufPtr = wbuf;
-                bufPtr = buf;
-
-                res = iconv(m2w, ICONV_CHAR_CAST(&bufPtr), &insz,
-                            (char**)&wbufPtr, &outsz);
-
-                if (ICONV_FAILED(res, insz))
-                {
-                    ms_wcCharsetName = NULL;
-                    wxLogLastError(wxT("iconv"));
-                    wxLogError(_("Conversion to charset '%s' doesn't work."), name);
-                }
-                else
+                // try charset w/o bytesex info (e.g. "UCS4")
+                wxLogTrace(TRACE_STRCONV, _T("  trying charset \"%s\""),
+                           nameCS.c_str());
+                m2w = iconv_open(nameCS.ToAscii(), cname);
+
+                // and check for bytesex ourselves:
+                if ( m2w != ICONV_T_INVALID )
                 {
-                    ms_wcNeedsSwap = wbuf[0] != (wchar_t)buf[0];
+                    char    buf[2], *bufPtr;
+                    wchar_t wbuf[2], *wbufPtr;
+                    size_t  insz, outsz;
+                    size_t  res;
+
+                    buf[0] = 'A';
+                    buf[1] = 0;
+                    wbuf[0] = 0;
+                    insz = 2;
+                    outsz = SIZEOF_WCHAR_T * 2;
+                    wbufPtr = wbuf;
+                    bufPtr = buf;
+
+                    res = iconv(m2w, ICONV_CHAR_CAST(&bufPtr), &insz,
+                                (char**)&wbufPtr, &outsz);
+
+                    if (ICONV_FAILED(res, insz))
+                    {
+                        wxLogLastError(wxT("iconv"));
+                        wxLogError(_("Conversion to charset '%s' doesn't work."),
+                                   nameCS.c_str());
+                    }
+                    else // ok, can convert to this encoding, remember it
+                    {
+                        ms_wcCharsetName = nameCS;
+                        ms_wcNeedsSwap = wbuf[0] != (wchar_t)buf[0];
+                    }
                 }
             }
-            else
+            else // use charset not requiring byte swapping
             {
-                ms_wcCharsetName = NULL;
-
-                // VS: we must not output an error here, since wxWidgets will safely
-                //     fall back to using wxEncodingConverter.
-                wxLogTrace(wxT("strconv"), wxT("Impossible to convert to/from charset '%s' with iconv, falling back to wxEncodingConverter."), name);
-                //wxLogError(
+                ms_wcCharsetName = nameXE;
             }
         }
-        wxLogTrace(wxT("strconv"), wxT("wchar_t charset is '%s', needs swap: %i"), ms_wcCharsetName, ms_wcNeedsSwap);
+
+        wxLogTrace(TRACE_STRCONV,
+                   wxT("iconv wchar_t charset is \"%s\"%s"),
+                   ms_wcCharsetName.empty() ? _T("<none>")
+                                            : ms_wcCharsetName.c_str(),
+                   ms_wcNeedsSwap ? _T(" (needs swap)")
+                                  : _T(""));
     }
     else // we already have ms_wcCharsetName
     {
-        m2w = iconv_open(ms_wcCharsetName, cname);
+        m2w = iconv_open(ms_wcCharsetName.ToAscii(), cname);
     }
 
-    // NB: don't ever pass NULL to iconv_open(), it may crash!
-    if ( ms_wcCharsetName )
+    if ( ms_wcCharsetName.empty() )
     {
-        w2m = iconv_open( cname, ms_wcCharsetName);
+        w2m = ICONV_T_INVALID;
     }
     else
     {
-        w2m = (iconv_t)-1;
+        w2m = iconv_open(cname, ms_wcCharsetName.ToAscii());
+        if ( w2m == ICONV_T_INVALID )
+        {
+            wxLogTrace(TRACE_STRCONV,
+                       wxT("\"%s\" -> \"%s\" works but not the converse!?"),
+                       ms_wcCharsetName.c_str(), cname.data());
+        }
     }
 }
 
 wxMBConv_iconv::~wxMBConv_iconv()
 {
-    if ( m2w != (iconv_t)-1 )
+    if ( m2w != ICONV_T_INVALID )
         iconv_close(m2w);
-    if ( w2m != (iconv_t)-1 )
+    if ( w2m != ICONV_T_INVALID )
         iconv_close(w2m);
 }
 
@@ -1309,7 +1477,7 @@ size_t wxMBConv_iconv::MB2WC(wchar_t *buf, const char *psz, size_t n) const
     //     as MB<->WC conversion would fail "randomly".
     wxMutexLocker lock(wxConstCast(this, wxMBConv_iconv)->m_iconvMutex);
 #endif
+
     size_t inbuf = strlen(psz);
     size_t outbuf = n * SIZEOF_WCHAR_T;
     size_t res, cres;
@@ -1328,7 +1496,8 @@ size_t wxMBConv_iconv::MB2WC(wchar_t *buf, const char *psz, size_t n) const
         if (ms_wcNeedsSwap)
         {
             // convert to native endianness
-            WC_BSWAP(buf /* _not_ bufPtr */, res)
+            for ( unsigned i = 0; i < res; i++ )
+                buf[n] = WC_BSWAP(buf[i]);
         }
 
         // NB: iconv was given only strlen(psz) characters on input, and so
@@ -1358,7 +1527,7 @@ size_t wxMBConv_iconv::MB2WC(wchar_t *buf, const char *psz, size_t n) const
     if (ICONV_FAILED(cres, inbuf))
     {
         //VS: it is ok if iconv fails, hence trace only
-        wxLogTrace(wxT("strconv"), wxT("iconv failed: %s"), wxSysErrorMsg(wxSysErrorCode()));
+        wxLogTrace(TRACE_STRCONV, wxT("iconv failed: %s"), wxSysErrorMsg(wxSysErrorCode()));
         return (size_t)-1;
     }
 
@@ -1371,8 +1540,9 @@ size_t wxMBConv_iconv::WC2MB(char *buf, const wchar_t *psz, size_t n) const
     // NB: explained in MB2WC
     wxMutexLocker lock(wxConstCast(this, wxMBConv_iconv)->m_iconvMutex);
 #endif
-    
-    size_t inbuf = wxWcslen(psz) * SIZEOF_WCHAR_T;
+
+    size_t inlen = wxWcslen(psz);
+    size_t inbuf = inlen * SIZEOF_WCHAR_T;
     size_t outbuf = n;
     size_t res, cres;
 
@@ -1381,13 +1551,13 @@ size_t wxMBConv_iconv::WC2MB(char *buf, const wchar_t *psz, size_t n) const
     if (ms_wcNeedsSwap)
     {
         // need to copy to temp buffer to switch endianness
-        // this absolutely doesn't rock!
-        // (no, doing WC_BSWAP twice on the original buffer won't help, as it
+        // (doing WC_BSWAP twice on the original buffer won't help, as it
         //  could be in read-only memory, or be accessed in some other thread)
-        tmpbuf=(wchar_t*)malloc((inbuf+1)*SIZEOF_WCHAR_T);
-        memcpy(tmpbuf,psz,(inbuf+1)*SIZEOF_WCHAR_T);
-        WC_BSWAP(tmpbuf, inbuf)
-        psz=tmpbuf;
+        tmpbuf = (wchar_t *)malloc(inbuf + SIZEOF_WCHAR_T);
+        for ( size_t i = 0; i < inlen; i++ )
+            tmpbuf[n] = WC_BSWAP(psz[i]);
+        tmpbuf[inlen] = L'\0';
+        psz = tmpbuf;
     }
 
     if (buf)
@@ -1425,8 +1595,7 @@ size_t wxMBConv_iconv::WC2MB(char *buf, const wchar_t *psz, size_t n) const
 
     if (ICONV_FAILED(cres, inbuf))
     {
-        //VS: it is ok if iconv fails, hence trace only
-        wxLogTrace(wxT("strconv"), wxT("iconv failed: %s"), wxSysErrorMsg(wxSysErrorCode()));
+        wxLogTrace(TRACE_STRCONV, wxT("iconv failed: %s"), wxSysErrorMsg(wxSysErrorCode()));
         return (size_t)-1;
     }
 
@@ -1899,7 +2068,7 @@ public:
 #if wxUSE_FONTMAP
     wxMBConv_cocoa(const wxChar* name)
     {
-        Init( wxCFStringEncFromFontEnc(wxFontMapper::Get()->CharsetToEncoding(name, false) ) ) ;
+        Init( wxCFStringEncFromFontEnc(wxFontMapperBase::Get()->CharsetToEncoding(name, false) ) ) ;
     }
 #endif
 
@@ -1944,9 +2113,9 @@ public:
 #if SIZEOF_WCHAR_T == 4
         UniChar* szUniCharBuffer = new UniChar[nOutSize];
 #endif
+
         CFStringGetCharacters(theString, theRange, szUniCharBuffer);
-        
+
         CFRelease(theString);
 
         szUniCharBuffer[nOutLength] = '\0' ;
@@ -1956,20 +2125,20 @@ public:
         converter.MB2WC(szOut, (const char*)szUniCharBuffer , nOutSize ) ;
         delete[] szUniCharBuffer;
 #endif
-    
+
         return nOutLength;
     }
 
     size_t WC2MB(char *szOut, const wchar_t *szUnConv, size_t nOutSize) const
     {
         wxASSERT(szUnConv);
-        
+
         size_t nRealOutSize;
         size_t nBufSize = wxWcslen(szUnConv);
         UniChar* szUniBuffer = (UniChar*) szUnConv;
 
 #if SIZEOF_WCHAR_T == 4
-        wxMBConvUTF16BE converter ;
+        wxMBConvUTF16 converter ;
         nBufSize = converter.WC2MB( NULL , szUnConv , 0 );
         szUniBuffer = new UniChar[ (nBufSize / sizeof(UniChar)) + 1] ;
         converter.WC2MB( (char*) szUniBuffer , szUnConv, nBufSize + sizeof(UniChar)) ;
@@ -1991,7 +2160,7 @@ public:
         {
             if (szOut != NULL)
                 CFStringGetCharacters(theString, CFRangeMake(0, nOutSize - 1), (UniChar*) szOut);
-            
+
             nRealOutSize = CFStringGetLength(theString) + 1;
         }
         else
@@ -2004,7 +2173,7 @@ public:
                     //0 tells CFString to return NULL if it meets such a character
                 false, //not an external representation
                 (UInt8*) szOut,
-                nOutSize, 
+                nOutSize,
                 (CFIndex*) &nRealOutSize
                         );
         }
@@ -2020,7 +2189,7 @@ public:
 
     bool IsOk() const
     {
-        return m_encoding != kCFStringEncodingInvalidId && 
+        return m_encoding != kCFStringEncodingInvalidId &&
               CFStringIsEncodingAvailable(m_encoding);
     }
 
@@ -2047,7 +2216,7 @@ public:
 #if wxUSE_FONTMAP
     wxMBConv_mac(const wxChar* name)
     {
-        Init( wxMacGetSystemEncFromFontEnc(wxFontMapper::Get()->CharsetToEncoding(name, false) ) ) ;
+        Init( wxMacGetSystemEncFromFontEnc(wxFontMapperBase::Get()->CharsetToEncoding(name, false) ) ) ;
     }
 #endif
 
@@ -2105,7 +2274,7 @@ public:
         // we have to terminate here, because n might be larger for the trailing zero, and if UniChar
         // is not properly terminated we get random characters at the end
         ubuf[byteOutLen / sizeof( UniChar ) ] = 0 ;
-        wxMBConvUTF16BE converter ;
+        wxMBConvUTF16 converter ;
         res = converter.MB2WC( (buf ? buf : tbuf) , (const char*)ubuf , n ) ;
         free( ubuf ) ;
 #else
@@ -2138,7 +2307,7 @@ public:
         ByteCount byteBufferLen = n ;
         UniChar* ubuf = NULL ;
 #if SIZEOF_WCHAR_T == 4
-        wxMBConvUTF16BE converter ;
+        wxMBConvUTF16 converter ;
         size_t unicharlen = converter.WC2MB( NULL , psz , 0 ) ;
         byteInLen = unicharlen ;
         ubuf = (UniChar*) malloc( byteInLen + 2 ) ;
@@ -2158,7 +2327,7 @@ public:
         if ( buf  && res < n)
         {
             buf[res] = 0;
-            
+
             //we need to double-trip to verify it didn't insert any ? in place
             //of bogus characters
             wxWCharBuffer wcBuf(n);
@@ -2210,7 +2379,7 @@ public:
     wxMBConv_wxwin(const wxChar* name)
     {
         if (name)
-            m_enc = wxFontMapper::Get()->CharsetToEncoding(name, false);
+            m_enc = wxFontMapperBase::Get()->CharsetToEncoding(name, false);
         else
             m_enc = wxFONTENCODING_SYSTEM;
 
@@ -2259,6 +2428,18 @@ public:
     DECLARE_NO_COPY_CLASS(wxMBConv_wxwin)
 };
 
+// make the constructors available for unit testing
+WXDLLIMPEXP_BASE wxMBConv* new_wxMBConv_wxwin( const wxChar* name )
+{
+    wxMBConv_wxwin* result = new wxMBConv_wxwin( name );
+    if ( !result->IsOk() )
+    {
+        delete result;
+        return 0;
+    }
+    return result;
+}
+
 #endif // wxUSE_FONTMAP
 
 // ============================================================================
@@ -2340,8 +2521,24 @@ void wxCSConv::SetName(const wxChar *charset)
     }
 }
 
+#if wxUSE_FONTMAP
+#include "wx/hashmap.h"
+
+WX_DECLARE_HASH_MAP( wxFontEncoding, wxString, wxIntegerHash, wxIntegerEqual,
+                     wxEncodingNameCache );
+
+static wxEncodingNameCache gs_nameCache;
+#endif
+
 wxMBConv *wxCSConv::DoCreate() const
 {
+#if wxUSE_FONTMAP
+    wxLogTrace(TRACE_STRCONV,
+               wxT("creating conversion for %s"),
+               (m_name ? m_name
+                       : wxFontMapperBase::GetEncodingName(m_encoding).c_str()));
+#endif // wxUSE_FONTMAP
+
     // check for the special case of ASCII or ISO8859-1 charset: as we have
     // special knowledge of it anyhow, we don't need to create a special
     // conversion object
@@ -2366,17 +2563,53 @@ wxMBConv *wxCSConv::DoCreate() const
 #endif // !wxUSE_FONTMAP
     {
         wxString name(m_name);
+        wxFontEncoding encoding(m_encoding);
+
+        if ( !name.empty() )
+        {
+            wxMBConv_iconv *conv = new wxMBConv_iconv(name);
+            if ( conv->IsOk() )
+                return conv;
+
+            delete conv;
 
 #if wxUSE_FONTMAP
-        if ( name.empty() )
-            name = wxFontMapper::Get()->GetEncodingName(m_encoding);
+            encoding =
+                wxFontMapperBase::Get()->CharsetToEncoding(name, false);
 #endif // wxUSE_FONTMAP
+        }
+#if wxUSE_FONTMAP
+        {
+            const wxEncodingNameCache::iterator it = gs_nameCache.find(encoding);
+            if ( it != gs_nameCache.end() )
+            {
+                if ( it->second.empty() )
+                    return NULL;
 
-        wxMBConv_iconv *conv = new wxMBConv_iconv(name);
-        if ( conv->IsOk() )
-            return conv;
+                wxMBConv_iconv *conv = new wxMBConv_iconv(it->second);
+                if ( conv->IsOk() )
+                    return conv;
 
-        delete conv;
+                delete conv;
+            }
+
+            const wxChar** names = wxFontMapperBase::GetAllEncodingNames(encoding);
+
+            for ( ; *names; ++names )
+            {
+                wxMBConv_iconv *conv = new wxMBConv_iconv(*names);
+                if ( conv->IsOk() )
+                {
+                    gs_nameCache[encoding] = *names;
+                    return conv;
+                }
+
+                delete conv;
+            }
+
+            gs_nameCache[encoding] = _T(""); // cache the failure
+        }
+#endif // wxUSE_FONTMAP
     }
 #endif // HAVE_ICONV
 
@@ -2397,7 +2630,7 @@ wxMBConv *wxCSConv::DoCreate() const
 #if defined(__WXMAC__)
     {
         // leave UTF16 and UTF32 to the built-ins of wx
-        if ( m_name || ( m_encoding < wxFONTENCODING_UTF16BE || 
+        if ( m_name || ( m_encoding < wxFONTENCODING_UTF16BE ||
             ( m_encoding >= wxFONTENCODING_MACMIN && m_encoding <= wxFONTENCODING_MACMAX ) ) )
         {
 
@@ -2440,7 +2673,7 @@ wxMBConv *wxCSConv::DoCreate() const
         // use "false" to suppress interactive dialogs -- we can be called from
         // anywhere and popping up a dialog from here is the last thing we want to
         // do
-        enc = wxFontMapper::Get()->CharsetToEncoding(m_name, false);
+        enc = wxFontMapperBase::Get()->CharsetToEncoding(m_name, false);
     }
 #endif // wxUSE_FONTMAP
 
@@ -2496,7 +2729,7 @@ wxMBConv *wxCSConv::DoCreate() const
                    m_name ? m_name
                       :
 #if wxUSE_FONTMAP
-                         wxFontMapper::GetEncodingDescription(m_encoding).c_str()
+                         wxFontMapperBase::GetEncodingDescription(m_encoding).c_str()
 #else // !wxUSE_FONTMAP
                          wxString::Format(_("encoding %s"), m_encoding).c_str()
 #endif // wxUSE_FONTMAP/!wxUSE_FONTMAP
@@ -2601,9 +2834,9 @@ WXDLLIMPEXP_DATA_BASE(wxMBConvUTF8&) wxConvUTF8 = wxConvUTF8Obj;
 WXDLLIMPEXP_DATA_BASE(wxMBConv *) wxConvCurrent = &wxConvLibcObj;
 WXDLLIMPEXP_DATA_BASE(wxMBConv *) wxConvFileName = &
 #ifdef __WXOSX__
-                                                    wxConvUTF8Obj;
+                                    wxConvUTF8Obj;
 #else
-                                                    wxConvLibcObj;
+                                    wxConvLibcObj;
 #endif
 
 
@@ -2616,5 +2849,3 @@ WXDLLIMPEXP_DATA_BASE(wxMBConv) wxConvLibc,
                                 wxConvUTF8;
 
 #endif // wxUSE_WCHAR_T/!wxUSE_WCHAR_T
-
-