made wxTextFile work in Unicode; also made it possible to use it with non seekable...

[wxWidgets.git] / src / common / strconv.cpp

diff --git a/src/common/strconv.cpp b/src/common/strconv.cpp
index 20ff6f49ef69c2e1ab68c2e1147599cd1bedab86..7b0e78a69db5d1eba0700ec2bde0c11ba8f66381 100644 (file)
--- a/src/common/strconv.cpp
+++ b/src/common/strconv.cpp
@@ -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)
 // Purpose:     Unicode conversion classes
 // Author:      Ove Kaaven, Robert Roebling, Vadim Zeitlin, Vaclav Slavik,
 //              Ryan Norton, Fredrik Roubert (UTF7)


@@ -20,10 +20,6 @@
 // headers
 // ----------------------------------------------------------------------------
 
 // 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"
 
 // For compilers that support precompilation, includes "wx.h".
 #include "wx/wxprec.h"
 


@@ -40,11 +36,8 @@
 
 #if wxUSE_WCHAR_T
 
 
 #if wxUSE_WCHAR_T
 

-#ifdef __WXMSW__
-    #include "wx/msw/private.h"
-#endif
-

 #ifdef __WINDOWS__
 #ifdef __WINDOWS__

+    #include "wx/msw/private.h"

     #include "wx/msw/missing.h"
 #endif
 
     #include "wx/msw/missing.h"
 #endif
 


@@ -60,10 +53,6 @@
     #define wxHAVE_WIN32_MB2WC
 #endif // __WIN32__ but !__WXMICROWIN__
 
     #define wxHAVE_WIN32_MB2WC
 #endif // __WIN32__ but !__WXMICROWIN__
 

-// ----------------------------------------------------------------------------
-// headers
-// ----------------------------------------------------------------------------
-

 #ifdef __SALFORDC__
     #include <clib.h>
 #endif
 #ifdef __SALFORDC__
     #include <clib.h>
 #endif


@@ -78,39 +67,19 @@
 #include "wx/utils.h"
 
 #ifdef __WXMAC__
 #include "wx/utils.h"
 
 #ifdef __WXMAC__

+#ifndef __DARWIN__

 #include <ATSUnicode.h>
 #include <TextCommon.h>
 #include <TextEncodingConverter.h>
 #include <ATSUnicode.h>
 #include <TextCommon.h>
 #include <TextEncodingConverter.h>

+#endif

 
 #include  "wx/mac/private.h"  // includes mac headers
 #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
     #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
 
 // ============================================================================
 #endif
 
 // ============================================================================


@@ -270,9 +239,9 @@ const wxWCharBuffer wxMBConv::cMB2WC(const char *szString, size_t nStringLen, si
         }
 
         //Increment to next (sub)string
         }
 
         //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;
     }
 
         szPos += strlen(szPos) + 1;
     }
 


@@ -332,9 +301,9 @@ const wxCharBuffer wxMBConv::cWC2MB(const wchar_t *szString, size_t nStringLen,
         }
 
         //Increment to next (sub)string
         }
 
         //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;
     }
 
         szPos += wxWcslen(szPos) + 1;
     }
 


@@ -360,39 +329,16 @@ size_t wxMBConvLibc::WC2MB(char *buf, const wchar_t *psz, size_t n) const
 #ifdef __UNIX__
 
 // ----------------------------------------------------------------------------
 #ifdef __UNIX__
 
 // ----------------------------------------------------------------------------

-// wxConvBrokenFileNames 
+// wxConvBrokenFileNames

 // ----------------------------------------------------------------------------
 
 // ----------------------------------------------------------------------------
 

-wxConvBrokenFileNames::wxConvBrokenFileNames()
+wxConvBrokenFileNames::wxConvBrokenFileNames(const wxChar *charset)

 {
 {

-    // decide which conversion to use for the file names
-
-    // (1) this variable exists for the sole purpose of specifying the encoding
-    //     of the filenames for GTK+ programs, so use it if it is set
-    wxString encName(wxGetenv(_T("G_FILENAME_ENCODING")));
-    encName.MakeUpper();
-    if ( !encName.empty() && encName != _T("UTF-8") && encName != _T("UTF8") )
-    {
-        m_conv = new wxCSConv(encName);
-    }
-    else // no G_FILENAME_ENCODING
-    {
-        if ( encName.empty() )
-            encName = wxLocale::GetSystemEncodingName().Upper();
-
-        // (2) if a non default locale is set, assume that the user wants his
-        //     filenames in this locale too
-        if ( !encName.empty() && encName != _T("UTF-8") && encName != _T("UTF8") )
-        {
-            wxSetEnv(_T("G_FILENAME_ENCODING"), encName);
-            m_conv = new wxMBConvLibc;
-        }
-        else
-        {
-            // (3) finally use UTF-8 by default
-            m_conv = new wxMBConvUTF8(wxMBConvUTF8::MAP_INVALID_UTF8_TO_OCTAL);
-        }
-    }
+    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
 }
 
 size_t


@@ -581,7 +527,7 @@ size_t wxMBConvUTF7::WC2MB(char *buf, const wchar_t *psz, size_t n) const
             {
                 // BASE64 encode string
                 unsigned int lsb, d, l;
             {
                 // BASE64 encode string
                 unsigned int lsb, d, l;

-                for (d = 0, l = 0;; psz++)
+                for (d = 0, l = 0; /*nothing*/; psz++)

                 {
                     for (lsb = 0; lsb < 2; lsb ++)
                     {
                 {
                     for (lsb = 0; lsb < 2; lsb ++)
                     {


@@ -647,6 +593,15 @@ size_t wxMBConvUTF8::MB2WC(wchar_t *buf, const char *psz, size_t n) const
             if (buf)
                 *buf++ = cc;
             len++;
             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
         {
         }
         else
         {


@@ -702,7 +657,7 @@ size_t wxMBConvUTF8::MB2WC(wchar_t *buf, const char *psz, size_t n) const
                     }
 #else // !WC_UTF16
                     if (buf)
                     }
 #else // !WC_UTF16
                     if (buf)

-                        *buf++ = res;
+                        *buf++ = (wchar_t)res;

                     len++;
 #endif // WC_UTF16/!WC_UTF16
                 }
                     len++;
 #endif // WC_UTF16/!WC_UTF16
                 }


@@ -723,7 +678,7 @@ size_t wxMBConvUTF8::MB2WC(wchar_t *buf, const char *psz, size_t n) const
                         len += pa;
 #else
                         if (buf)
                         len += pa;
 #else
                         if (buf)

-                            *buf++ = wxUnicodePUA + (unsigned char)*opsz;
+                            *buf++ = (wchar_t)(wxUnicodePUA + (unsigned char)*opsz);

                         opsz++;
                         len++;
 #endif
                         opsz++;
                         len++;
 #endif


@@ -735,11 +690,11 @@ size_t wxMBConvUTF8::MB2WC(wchar_t *buf, const char *psz, size_t n) const
                     {
                         if ( buf && len + 3 < n )
                         {
                     {
                         if ( buf && len + 3 < n )
                         {

-                            unsigned char n = *opsz;
+                            unsigned char on = *opsz;

                             *buf++ = L'\\';
                             *buf++ = L'\\';

-                            *buf++ = (wchar_t)( L'0' + n / 0100 );
-                            *buf++ = (wchar_t)( L'0' + (n % 0100) / 010 );
-                            *buf++ = (wchar_t)( L'0' + n % 010 );
+                            *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;
                         }
                         opsz++;
                         len += 4;


@@ -784,6 +739,14 @@ size_t wxMBConvUTF8::WC2MB(char *buf, const wchar_t *psz, size_t n) const
                 *buf++ = (char)(cc - wxUnicodePUA);
             len++;
         }
                 *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]) )
         else if ( (m_options & MAP_INVALID_UTF8_TO_OCTAL) &&
                     cc == L'\\' &&
                         isoctal(psz[0]) && isoctal(psz[1]) && isoctal(psz[2]) )


@@ -943,7 +906,7 @@ size_t wxMBConvUTF16straight::MB2WC(wchar_t *buf, const char *psz, size_t n) con
             return pa;
 
         if (buf)
             return pa;
 
         if (buf)

-            *buf++ = cc;
+            *buf++ = (wchar_t)cc;

         len++;
         psz += pa * sizeof(wxUint16);
     }
         len++;
         psz += pa * sizeof(wxUint16);
     }


@@ -1003,7 +966,7 @@ size_t wxMBConvUTF16swap::MB2WC(wchar_t *buf, const char *psz, size_t n) const
             return pa;
 
         if (buf)
             return pa;
 
         if (buf)

-            *buf++ = cc;
+            *buf++ = (wchar_t)cc;

 
         len++;
         psz += pa * sizeof(wxUint16);
 
         len++;
         psz += pa * sizeof(wxUint16);


@@ -1204,7 +1167,7 @@ size_t wxMBConvUTF32straight::MB2WC(wchar_t *buf, const char *psz, size_t n) con
     while (*(wxUint32*)psz && (!buf || len < n))
     {
         if (buf)
     while (*(wxUint32*)psz && (!buf || len < n))
     {
         if (buf)

-            *buf++ = *(wxUint32*)psz;
+            *buf++ = (wchar_t)(*(wxUint32*)psz);

         len++;
         psz += sizeof(wxUint32);
     }
         len++;
         psz += sizeof(wxUint32);
     }


@@ -1317,6 +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_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
 // ----------------------------------------------------------------------------
 // ----------------------------------------------------------------------------
 // wxMBConv_iconv: encapsulates an iconv character set
 // ----------------------------------------------------------------------------


@@ -1331,7 +1307,7 @@ public:
     virtual size_t WC2MB(char *buf, const wchar_t *psz, size_t n) const;
 
     bool IsOk() const
     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
 
 protected:
     // the iconv handlers used to translate from multibyte to wide char and in


@@ -1346,108 +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
 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;
 };
 
 
     // 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)
 {
 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:
 
     // 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

 
 

-            if (m2w != (iconv_t)-1)
+            wxLogTrace(TRACE_STRCONV, _T("  trying charset \"%s\""),
+                       nameXE.c_str());
+
+            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
     {
     }
     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
     {
     }
     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()
 {
     }
 }
 
 wxMBConv_iconv::~wxMBConv_iconv()
 {

-    if ( m2w != (iconv_t)-1 )
+    if ( m2w != ICONV_T_INVALID )

         iconv_close(m2w);
         iconv_close(m2w);

-    if ( w2m != (iconv_t)-1 )
+    if ( w2m != ICONV_T_INVALID )

         iconv_close(w2m);
 }
 
         iconv_close(w2m);
 }
 


@@ -1481,7 +1496,8 @@ size_t wxMBConv_iconv::MB2WC(wchar_t *buf, const char *psz, size_t n) const
         if (ms_wcNeedsSwap)
         {
             // convert to native endianness
         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
         }
 
         // NB: iconv was given only strlen(psz) characters on input, and so


@@ -1511,7 +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
     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;
     }
 
         return (size_t)-1;
     }
 


@@ -1525,7 +1541,8 @@ size_t wxMBConv_iconv::WC2MB(char *buf, const wchar_t *psz, size_t n) const
     wxMutexLocker lock(wxConstCast(this, wxMBConv_iconv)->m_iconvMutex);
 #endif
 
     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;
 
     size_t outbuf = n;
     size_t res, cres;
 


@@ -1534,13 +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
     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)
         //  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)
     }
 
     if (buf)


@@ -1578,8 +1595,7 @@ size_t wxMBConv_iconv::WC2MB(char *buf, const wchar_t *psz, size_t n) const
 
     if (ICONV_FAILED(cres, inbuf))
     {
 
     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;
     }
 
         return (size_t)-1;
     }
 


@@ -1636,7 +1652,28 @@ public:
         // own wxMBConvUTF7 doesn't detect errors (e.g. lone "+" which is
         // explicitly ill-formed according to RFC 2152) neither so we don't
         // even have any fallback here...
         // own wxMBConvUTF7 doesn't detect errors (e.g. lone "+" which is
         // explicitly ill-formed according to RFC 2152) neither so we don't
         // even have any fallback here...

-        int flags = m_CodePage == CP_UTF7 ? 0 : MB_ERR_INVALID_CHARS;
+        //
+        // Moreover, MB_ERR_INVALID_CHARS is only supported on Win 2K SP4 or
+        // Win XP or newer and if it is specified on older versions, conversion
+        // from CP_UTF8 (which can have flags only 0 or MB_ERR_INVALID_CHARS)
+        // fails. So we can only use the flag on newer Windows versions.
+        // Additionally, the flag is not supported by UTF7, symbol and CJK
+        // encodings. See here:
+        //     http://blogs.msdn.com/michkap/archive/2005/04/19/409566.aspx
+        //     http://msdn.microsoft.com/library/en-us/intl/unicode_17si.asp
+        int flags = 0;
+        if ( m_CodePage != CP_UTF7 && m_CodePage != CP_SYMBOL &&
+             m_CodePage < 50000 &&
+             IsAtLeastWin2kSP4() )
+        {
+            flags = MB_ERR_INVALID_CHARS;
+        }
+        else if ( m_CodePage == CP_UTF8 )
+        {
+            // Avoid round-trip in the special case of UTF-8 by using our
+            // own UTF-8 conversion code:
+            return wxMBConvUTF8().MB2WC(buf, psz, n);
+        }

 
         const size_t len = ::MultiByteToWideChar
                              (
 
         const size_t len = ::MultiByteToWideChar
                              (


@@ -1647,11 +1684,40 @@ public:
                                 buf,            // output string
                                 buf ? n : 0     // size of output buffer
                              );
                                 buf,            // output string
                                 buf ? n : 0     // size of output buffer
                              );

+        if ( !len )
+        {
+            // function totally failed
+            return (size_t)-1;
+        }
+
+        // if we were really converting and didn't use MB_ERR_INVALID_CHARS,
+        // check if we succeeded, by doing a double trip:
+        if ( !flags && buf )
+        {
+            wxCharBuffer mbBuf(n);
+            if ( ::WideCharToMultiByte
+                   (
+                      m_CodePage,
+                      0,
+                      buf,
+                      -1,
+                      mbBuf.data(),
+                      n,
+                      NULL,
+                      NULL
+                   ) == 0 ||
+                  strcmp(mbBuf, psz) != 0 )
+            {
+                // we didn't obtain the same thing we started from, hence
+                // the conversion was lossy and we consider that it failed
+                return (size_t)-1;
+            }
+        }

 
         // note that it returns count of written chars for buf != NULL and size
         // of the needed buffer for buf == NULL so in either case the length of
         // the string (which never includes the terminating NUL) is one less
 
         // note that it returns count of written chars for buf != NULL and size
         // of the needed buffer for buf == NULL so in either case the length of
         // the string (which never includes the terminating NUL) is one less

-        return len ? len - 1 : (size_t)-1;
+        return len - 1;

     }
 
     size_t WC2MB(char *buf, const wchar_t *pwz, size_t n) const
     }
 
     size_t WC2MB(char *buf, const wchar_t *pwz, size_t n) const


@@ -1766,6 +1832,33 @@ private:
         return s_isWin98Or2k == 1;
     }
 
         return s_isWin98Or2k == 1;
     }
 

+    static bool IsAtLeastWin2kSP4()
+    {
+#ifdef __WXWINCE__
+        return false;
+#else
+        static int s_isAtLeastWin2kSP4 = -1;
+
+        if ( s_isAtLeastWin2kSP4 == -1 )
+        {
+            OSVERSIONINFOEX ver;
+
+            memset(&ver, 0, sizeof(ver));
+            ver.dwOSVersionInfoSize = sizeof(ver);
+            GetVersionEx((OSVERSIONINFO*)&ver);
+
+            s_isAtLeastWin2kSP4 =
+              ((ver.dwMajorVersion > 5) || // Vista+
+               (ver.dwMajorVersion == 5 && ver.dwMinorVersion > 0) || // XP/2003
+               (ver.dwMajorVersion == 5 && ver.dwMinorVersion == 0 &&
+               ver.wServicePackMajor >= 4)) // 2000 SP4+
+              ? 1 : 0;
+        }
+
+        return s_isAtLeastWin2kSP4 == 1;
+#endif
+    }
+

     long m_CodePage;
 };
 
     long m_CodePage;
 };
 


@@ -2122,7 +2215,7 @@ public:
         UniChar* szUniBuffer = (UniChar*) szUnConv;
 
 #if SIZEOF_WCHAR_T == 4
         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)) ;
         nBufSize = converter.WC2MB( NULL , szUnConv , 0 );
         szUniBuffer = new UniChar[ (nBufSize / sizeof(UniChar)) + 1] ;
         converter.WC2MB( (char*) szUniBuffer , szUnConv, nBufSize + sizeof(UniChar)) ;


@@ -2258,7 +2351,7 @@ 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 ;
         // 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
         res = converter.MB2WC( (buf ? buf : tbuf) , (const char*)ubuf , n ) ;
         free( ubuf ) ;
 #else


@@ -2291,7 +2384,7 @@ public:
         ByteCount byteBufferLen = n ;
         UniChar* ubuf = NULL ;
 #if SIZEOF_WCHAR_T == 4
         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 ) ;
         size_t unicharlen = converter.WC2MB( NULL , psz , 0 ) ;
         byteInLen = unicharlen ;
         ubuf = (UniChar*) malloc( byteInLen + 2 ) ;


@@ -2412,6 +2505,18 @@ public:
     DECLARE_NO_COPY_CLASS(wxMBConv_wxwin)
 };
 
     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
 
 // ============================================================================
 #endif // wxUSE_FONTMAP
 
 // ============================================================================


@@ -2434,7 +2539,11 @@ wxCSConv::wxCSConv(const wxChar *charset)
         SetName(charset);
     }
 
         SetName(charset);
     }
 

+#if wxUSE_FONTMAP
+    m_encoding = wxFontMapperBase::GetEncodingFromName(charset);
+#else

     m_encoding = wxFONTENCODING_SYSTEM;
     m_encoding = wxFONTENCODING_SYSTEM;

+#endif

 }
 
 wxCSConv::wxCSConv(wxFontEncoding encoding)
 }
 
 wxCSConv::wxCSConv(wxFontEncoding encoding)


@@ -2493,12 +2602,29 @@ 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
 {
 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
     // 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

-    if ( m_encoding == wxFONTENCODING_ISO8859_1 )
+    if ( m_encoding == wxFONTENCODING_ISO8859_1 ||
+            m_encoding == wxFONTENCODING_DEFAULT )

     {
         // don't convert at all
         return NULL;
     {
         // don't convert at all
         return NULL;


@@ -2519,17 +2645,53 @@ wxMBConv *wxCSConv::DoCreate() const
 #endif // !wxUSE_FONTMAP
     {
         wxString name(m_name);
 #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 wxUSE_FONTMAP

-        if ( name.empty() )
-            name = wxFontMapperBase::Get()->GetEncodingName(m_encoding);
+            encoding =
+                wxFontMapperBase::Get()->CharsetToEncoding(name, false);

 #endif // wxUSE_FONTMAP
 #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
 
     }
 #endif // HAVE_ICONV
 


@@ -2769,5 +2931,3 @@ WXDLLIMPEXP_DATA_BASE(wxMBConv) wxConvLibc,
                                 wxConvUTF8;
 
 #endif // wxUSE_WCHAR_T/!wxUSE_WCHAR_T
                                 wxConvUTF8;
 
 #endif // wxUSE_WCHAR_T/!wxUSE_WCHAR_T

-
-