]> git.saurik.com Git - wxWidgets.git/blobdiff - tests/mbconv/mbconvtest.cpp
Restore ability to parse hours only with wxDateTime::ParseTime().
[wxWidgets.git] / tests / mbconv / mbconvtest.cpp
index 51eb23e96f1c189b98521b5d667bae1d96728751..93f3ea020f59b445249f60d37040cab39bbb7bbe 100644 (file)
@@ -1,10 +1,10 @@
 ///////////////////////////////////////////////////////////////////////////////
 // Name:        tests/mbconv/main.cpp
 // Purpose:     wxMBConv unit test
-// Author:      Vadim Zeitlin, Mike Wetherell
+// Author:      Vadim Zeitlin, Mike Wetherell, Vince Harron
 // Created:     14.02.04
 // RCS-ID:      $Id$
-// Copyright:   (c) 2003 TT-Solutions, (c) 2005 Mike Wetherell
+// Copyright:   (c) 2003 TT-Solutions, (c) 2005 Mike Wetherell, Vince Harron
 ///////////////////////////////////////////////////////////////////////////////
 
 // ----------------------------------------------------------------------------
@@ -23,6 +23,8 @@
 
 #include "wx/strconv.h"
 #include "wx/string.h"
+#include "wx/txtstrm.h"
+#include "wx/mstream.h"
 
 #if defined wxHAVE_TCHAR_SUPPORT && !defined HAVE_WCHAR_H
     #define HAVE_WCHAR_H
@@ -65,7 +67,21 @@ public:
 
 private:
     CPPUNIT_TEST_SUITE( MBConvTestCase );
+        CPPUNIT_TEST( UTF32LETests );
+        CPPUNIT_TEST( UTF32BETests );
         CPPUNIT_TEST( WC2CP1250 );
+        CPPUNIT_TEST( UTF7Tests );
+        CPPUNIT_TEST( UTF8Tests );
+        CPPUNIT_TEST( UTF16LETests );
+        CPPUNIT_TEST( UTF16BETests );
+        CPPUNIT_TEST( CP932Tests );
+        CPPUNIT_TEST( CP1252Tests ); // depends on UTF8 Decoder functioning correctly
+        CPPUNIT_TEST( LibcTests );
+        CPPUNIT_TEST( IconvTests );
+        CPPUNIT_TEST( Latin1Tests );
+        CPPUNIT_TEST( FontmapTests );
+        CPPUNIT_TEST( BufSize );
+        CPPUNIT_TEST( FromWCharTests );
 #ifdef HAVE_WCHAR_H
         CPPUNIT_TEST( UTF8_41 );
         CPPUNIT_TEST( UTF8_7f );
@@ -89,6 +105,77 @@ private:
     CPPUNIT_TEST_SUITE_END();
 
     void WC2CP1250();
+    void UTF7Tests();
+    void UTF8Tests();
+    void UTF16LETests();
+    void UTF16BETests();
+    void UTF32LETests();
+    void UTF32BETests();
+    void CP932Tests();
+    void CP1252Tests();
+    void LibcTests();
+    void FontmapTests();
+    void BufSize();
+    void FromWCharTests();
+    void IconvTests();
+    void Latin1Tests();
+
+    // verifies that the specified multibyte sequence decodes to the specified wchar_t sequence
+    void TestDecoder(
+        const wchar_t* wideBuffer,  // the same character sequence as multiBuffer, encoded as wchar_t
+        size_t         wideChars,   // the number of wide characters at wideBuffer
+        const char*    multiBuffer, // a multibyte encoded character sequence that can be decoded by "converter"
+        size_t         multiBytes,  // the byte length of the multibyte character sequence that can be decoded by "converter"
+        wxMBConv&      converter,   // the wxMBConv object that can decode multiBuffer into a wide character sequence
+        int            sizeofNull   // number of bytes occupied by terminating null in this encoding
+        );
+
+    // verifies that the specified wchar_t sequence encodes to the specified multibyte sequence
+    void TestEncoder(
+        const wchar_t* wideBuffer,  // the same character sequence as multiBuffer, encoded as wchar_t
+        size_t         wideChars,   // the number of wide characters at wideBuffer
+        const char*    multiBuffer, // a multibyte encoded character sequence that can be decoded by "converter"
+        size_t         multiBytes,  // the byte length of the multibyte character sequence that can be decoded by "converter"
+        wxMBConv&      converter,   // the wxMBConv object that can decode multiBuffer into a wide character sequence
+        int            sizeofNull   // number of bytes occupied by terminating null in this encoding
+        );
+
+#if wxUSE_UNICODE && wxUSE_STREAMS
+    // use wxTextInputStream to exercise wxMBConv interface
+    // (this reveals some bugs in certain wxMBConv subclasses)
+    void TestStreamDecoder(
+        const wchar_t* wideBuffer,  // the same character sequence as multiBuffer, encoded as wchar_t
+        size_t         wideChars,   // the number of wide characters at wideBuffer
+        const char*    multiBuffer, // a multibyte encoded character sequence that can be decoded by "converter"
+        size_t         multiBytes,  // the byte length of the multibyte character sequence that can be decoded by "converter"
+        wxMBConv&      converter    // the wxMBConv object that can decode multiBuffer into a wide character sequence
+        );
+
+    // use wxTextOutputStream to exercise wxMBConv interface
+    // (this reveals some bugs in certain wxMBConv subclasses)
+    void TestStreamEncoder(
+        const wchar_t* wideBuffer,  // the same character sequence as multiBuffer, encoded as wchar_t
+        size_t         wideChars,   // the number of wide characters at wideBuffer
+        const char*    multiBuffer, // a multibyte encoded character sequence that can be decoded by "converter"
+        size_t         multiBytes,  // the byte length of the multibyte character sequence that can be decoded by "converter"
+        wxMBConv&      converter    // the wxMBConv object that can decode multiBuffer into a wide character sequence
+        );
+#endif
+
+    // tests the encoding and decoding capability of an wxMBConv object
+    //
+    // decodes the utf-8 bytes into wide characters
+    // encodes the wide characters to compare against input multiBuffer
+    // decodes the multiBuffer to compare against wide characters
+    // decodes the multiBuffer into wide characters
+    void TestCoder(
+        const char*    multiBuffer, // a multibyte encoded character sequence that can be decoded by "converter"
+        size_t         multiBytes,  // the byte length of the multibyte character sequence that can be decoded by "converter"
+        const char*    utf8Buffer,  // the same character sequence as multiBuffer, encoded as UTF-8
+        size_t         utf8Bytes,   // the byte length of the UTF-8 encoded character sequence
+        wxMBConv&      converter,   // the wxMBConv object that can decode multiBuffer into a wide character sequence
+        int            sizeofNull   // the number of bytes occupied by a terminating null in the converter's encoding
+        );
 
 #ifdef HAVE_WCHAR_H
     // UTF-8 tests. Test the first, last and one in the middle for sequences
@@ -130,8 +217,9 @@ private:
 // register in the unnamed registry so that these tests are run by default
 CPPUNIT_TEST_SUITE_REGISTRATION( MBConvTestCase );
 
-// also include in it's own registry so that these tests can be run alone
+// also include in its own registry so that these tests can be run alone
 CPPUNIT_TEST_SUITE_NAMED_REGISTRATION( MBConvTestCase, "MBConvTestCase" );
+CPPUNIT_TEST_SUITE_NAMED_REGISTRATION( MBConvTestCase, "MBConv" );
 
 void MBConvTestCase::WC2CP1250()
 {
@@ -160,6 +248,953 @@ void MBConvTestCase::WC2CP1250()
     }
 }
 
+// Print an unsigned character array as a C unsigned character array.
+// NB: Please don't remove this function even though it's not used anywhere,
+//     it's very useful when debugging a failed test.
+wxString CByteArrayFormat( const void* data, size_t len, const wxChar* name )
+{
+    const unsigned char* bytes = (unsigned char*)data;
+    wxString result;
+
+    result.Printf( wxT("static const unsigned char %s[%i] = \n{"), name, (int)len );
+
+    for ( size_t i = 0; i < len; i++ )
+    {
+        if ( i != 0 )
+        {
+            result.append( wxT(",") );
+        }
+        if ((i%16)==0)
+        {
+            result.append( wxT("\n    ") );
+        }
+        wxString byte = wxString::Format( wxT("0x%02x"), bytes[i] );
+        result.append(byte);
+    }
+    result.append( wxT("\n};\n") );
+    return result;
+}
+
+// The following bytes represent the same string, containing Japanese and English
+// characters, encoded in several different formats.
+
+// encoded by iconv
+static const unsigned char welcome_utf7_iconv[84] =
+{
+    0x57,0x65,0x6c,0x63,0x6f,0x6d,0x65,0x20,0x74,0x6f,0x20,0x6f,0x75,0x72,0x20,0x63,
+    0x79,0x62,0x65,0x72,0x20,0x73,0x70,0x61,0x63,0x65,0x20,0x66,0x6f,0x72,0x63,0x65,
+    0x2e,0x20,0x20,0x2b,0x4d,0x46,0x6b,0x77,0x55,0x49,0x74,0x6d,0x57,0x39,0x38,0x77,
+    0x61,0x35,0x62,0x37,0x69,0x6e,0x45,0x77,0x6b,0x6a,0x42,0x5a,0x4d,0x49,0x73,0x77,
+    0x65,0x7a,0x42,0x47,0x4d,0x45,0x77,0x77,0x52,0x44,0x42,0x45,0x4d,0x47,0x63,0x77,
+    0x57,0x54,0x41,0x43
+};
+// encoded by wxWindows (iconv can decode this successfully)
+static const unsigned char welcome_utf7_wx[109] =
+{
+    0x57,0x65,0x6c,0x63,0x6f,0x6d,0x65,0x2b,0x41,0x43,0x41,0x2d,0x74,0x6f,0x2b,0x41,
+    0x43,0x41,0x2d,0x6f,0x75,0x72,0x2b,0x41,0x43,0x41,0x2d,0x63,0x79,0x62,0x65,0x72,
+    0x2b,0x41,0x43,0x41,0x2d,0x73,0x70,0x61,0x63,0x65,0x2b,0x41,0x43,0x41,0x2d,0x66,
+    0x6f,0x72,0x63,0x65,0x2e,0x2b,0x41,0x43,0x41,0x41,0x49,0x44,0x42,0x5a,0x4d,0x46,
+    0x43,0x4c,0x5a,0x6c,0x76,0x66,0x4d,0x47,0x75,0x57,0x2b,0x34,0x70,0x78,0x4d,0x4a,
+    0x49,0x77,0x57,0x54,0x43,0x4c,0x4d,0x48,0x73,0x77,0x52,0x6a,0x42,0x4d,0x4d,0x45,
+    0x51,0x77,0x52,0x44,0x42,0x6e,0x4d,0x46,0x6b,0x77,0x41,0x67,0x2d
+};
+// encoded by iconv
+static const unsigned char welcome_utf8[89] =
+{
+    0x57,0x65,0x6c,0x63,0x6f,0x6d,0x65,0x20,0x74,0x6f,0x20,0x6f,0x75,0x72,0x20,0x63,
+    0x79,0x62,0x65,0x72,0x20,0x73,0x70,0x61,0x63,0x65,0x20,0x66,0x6f,0x72,0x63,0x65,
+    0x2e,0x20,0x20,0xe3,0x81,0x99,0xe3,0x81,0x90,0xe8,0xad,0xa6,0xe5,0xaf,0x9f,0xe3,
+    0x81,0xab,0xe9,0x9b,0xbb,0xe8,0xa9,0xb1,0xe3,0x82,0x92,0xe3,0x81,0x99,0xe3,0x82,
+    0x8b,0xe3,0x81,0xbb,0xe3,0x81,0x86,0xe3,0x81,0x8c,0xe3,0x81,0x84,0xe3,0x81,0x84,
+    0xe3,0x81,0xa7,0xe3,0x81,0x99,0xe3,0x80,0x82
+};
+// encoded by iconv
+static const unsigned char welcome_utf16le[106] =
+{
+    0x57,0x00,0x65,0x00,0x6c,0x00,0x63,0x00,0x6f,0x00,0x6d,0x00,0x65,0x00,0x20,0x00,
+    0x74,0x00,0x6f,0x00,0x20,0x00,0x6f,0x00,0x75,0x00,0x72,0x00,0x20,0x00,0x63,0x00,
+    0x79,0x00,0x62,0x00,0x65,0x00,0x72,0x00,0x20,0x00,0x73,0x00,0x70,0x00,0x61,0x00,
+    0x63,0x00,0x65,0x00,0x20,0x00,0x66,0x00,0x6f,0x00,0x72,0x00,0x63,0x00,0x65,0x00,
+    0x2e,0x00,0x20,0x00,0x20,0x00,0x59,0x30,0x50,0x30,0x66,0x8b,0xdf,0x5b,0x6b,0x30,
+    0xfb,0x96,0x71,0x8a,0x92,0x30,0x59,0x30,0x8b,0x30,0x7b,0x30,0x46,0x30,0x4c,0x30,
+    0x44,0x30,0x44,0x30,0x67,0x30,0x59,0x30,0x02,0x30
+};
+// encoded by iconv
+static const unsigned char welcome_utf16be[106] =
+{
+    0x00,0x57,0x00,0x65,0x00,0x6c,0x00,0x63,0x00,0x6f,0x00,0x6d,0x00,0x65,0x00,0x20,
+    0x00,0x74,0x00,0x6f,0x00,0x20,0x00,0x6f,0x00,0x75,0x00,0x72,0x00,0x20,0x00,0x63,
+    0x00,0x79,0x00,0x62,0x00,0x65,0x00,0x72,0x00,0x20,0x00,0x73,0x00,0x70,0x00,0x61,
+    0x00,0x63,0x00,0x65,0x00,0x20,0x00,0x66,0x00,0x6f,0x00,0x72,0x00,0x63,0x00,0x65,
+    0x00,0x2e,0x00,0x20,0x00,0x20,0x30,0x59,0x30,0x50,0x8b,0x66,0x5b,0xdf,0x30,0x6b,
+    0x96,0xfb,0x8a,0x71,0x30,0x92,0x30,0x59,0x30,0x8b,0x30,0x7b,0x30,0x46,0x30,0x4c,
+    0x30,0x44,0x30,0x44,0x30,0x67,0x30,0x59,0x30,0x02
+};
+// encoded by iconv
+static const unsigned char welcome_utf32le[212] =
+{
+    0x57,0x00,0x00,0x00,0x65,0x00,0x00,0x00,0x6c,0x00,0x00,0x00,0x63,0x00,0x00,0x00,
+    0x6f,0x00,0x00,0x00,0x6d,0x00,0x00,0x00,0x65,0x00,0x00,0x00,0x20,0x00,0x00,0x00,
+    0x74,0x00,0x00,0x00,0x6f,0x00,0x00,0x00,0x20,0x00,0x00,0x00,0x6f,0x00,0x00,0x00,
+    0x75,0x00,0x00,0x00,0x72,0x00,0x00,0x00,0x20,0x00,0x00,0x00,0x63,0x00,0x00,0x00,
+    0x79,0x00,0x00,0x00,0x62,0x00,0x00,0x00,0x65,0x00,0x00,0x00,0x72,0x00,0x00,0x00,
+    0x20,0x00,0x00,0x00,0x73,0x00,0x00,0x00,0x70,0x00,0x00,0x00,0x61,0x00,0x00,0x00,
+    0x63,0x00,0x00,0x00,0x65,0x00,0x00,0x00,0x20,0x00,0x00,0x00,0x66,0x00,0x00,0x00,
+    0x6f,0x00,0x00,0x00,0x72,0x00,0x00,0x00,0x63,0x00,0x00,0x00,0x65,0x00,0x00,0x00,
+    0x2e,0x00,0x00,0x00,0x20,0x00,0x00,0x00,0x20,0x00,0x00,0x00,0x59,0x30,0x00,0x00,
+    0x50,0x30,0x00,0x00,0x66,0x8b,0x00,0x00,0xdf,0x5b,0x00,0x00,0x6b,0x30,0x00,0x00,
+    0xfb,0x96,0x00,0x00,0x71,0x8a,0x00,0x00,0x92,0x30,0x00,0x00,0x59,0x30,0x00,0x00,
+    0x8b,0x30,0x00,0x00,0x7b,0x30,0x00,0x00,0x46,0x30,0x00,0x00,0x4c,0x30,0x00,0x00,
+    0x44,0x30,0x00,0x00,0x44,0x30,0x00,0x00,0x67,0x30,0x00,0x00,0x59,0x30,0x00,0x00,
+    0x02,0x30,0x00,0x00
+};
+// encoded by iconv
+static const unsigned char welcome_utf32be[212] =
+{
+    0x00,0x00,0x00,0x57,0x00,0x00,0x00,0x65,0x00,0x00,0x00,0x6c,0x00,0x00,0x00,0x63,
+    0x00,0x00,0x00,0x6f,0x00,0x00,0x00,0x6d,0x00,0x00,0x00,0x65,0x00,0x00,0x00,0x20,
+    0x00,0x00,0x00,0x74,0x00,0x00,0x00,0x6f,0x00,0x00,0x00,0x20,0x00,0x00,0x00,0x6f,
+    0x00,0x00,0x00,0x75,0x00,0x00,0x00,0x72,0x00,0x00,0x00,0x20,0x00,0x00,0x00,0x63,
+    0x00,0x00,0x00,0x79,0x00,0x00,0x00,0x62,0x00,0x00,0x00,0x65,0x00,0x00,0x00,0x72,
+    0x00,0x00,0x00,0x20,0x00,0x00,0x00,0x73,0x00,0x00,0x00,0x70,0x00,0x00,0x00,0x61,
+    0x00,0x00,0x00,0x63,0x00,0x00,0x00,0x65,0x00,0x00,0x00,0x20,0x00,0x00,0x00,0x66,
+    0x00,0x00,0x00,0x6f,0x00,0x00,0x00,0x72,0x00,0x00,0x00,0x63,0x00,0x00,0x00,0x65,
+    0x00,0x00,0x00,0x2e,0x00,0x00,0x00,0x20,0x00,0x00,0x00,0x20,0x00,0x00,0x30,0x59,
+    0x00,0x00,0x30,0x50,0x00,0x00,0x8b,0x66,0x00,0x00,0x5b,0xdf,0x00,0x00,0x30,0x6b,
+    0x00,0x00,0x96,0xfb,0x00,0x00,0x8a,0x71,0x00,0x00,0x30,0x92,0x00,0x00,0x30,0x59,
+    0x00,0x00,0x30,0x8b,0x00,0x00,0x30,0x7b,0x00,0x00,0x30,0x46,0x00,0x00,0x30,0x4c,
+    0x00,0x00,0x30,0x44,0x00,0x00,0x30,0x44,0x00,0x00,0x30,0x67,0x00,0x00,0x30,0x59,
+    0x00,0x00,0x30,0x02
+};
+// encoded by iconv
+static const unsigned char welcome_cp932[71] =
+{
+    0x57,0x65,0x6c,0x63,0x6f,0x6d,0x65,0x20,0x74,0x6f,0x20,0x6f,0x75,0x72,0x20,0x63,
+    0x79,0x62,0x65,0x72,0x20,0x73,0x70,0x61,0x63,0x65,0x20,0x66,0x6f,0x72,0x63,0x65,
+    0x2e,0x20,0x20,0x82,0xb7,0x82,0xae,0x8c,0x78,0x8e,0x40,0x82,0xc9,0x93,0x64,0x98,
+    0x62,0x82,0xf0,0x82,0xb7,0x82,0xe9,0x82,0xd9,0x82,0xa4,0x82,0xaa,0x82,0xa2,0x82,
+    0xa2,0x82,0xc5,0x82,0xb7,0x81,0x42
+};
+
+#if wxBYTE_ORDER == wxBIG_ENDIAN
+    #if SIZEOF_WCHAR_T == 2
+        #define welcome_wchar_t welcome_utf16be
+    #elif SIZEOF_WCHAR_T == 4
+        #define welcome_wchar_t welcome_utf32be
+    #endif
+#elif wxBYTE_ORDER == wxLITTLE_ENDIAN
+    #if SIZEOF_WCHAR_T == 2
+        #define welcome_wchar_t welcome_utf16le
+    #elif SIZEOF_WCHAR_T == 4
+        #define welcome_wchar_t welcome_utf32le
+    #endif
+#endif
+
+void MBConvTestCase::UTF7Tests()
+{
+#if 0
+    wxCSConv convUTF7(wxFONTENCODING_UTF7);
+#else
+    wxMBConvUTF7 convUTF7;
+#endif
+
+    TestDecoder
+        (
+        (const wchar_t*)welcome_wchar_t,
+        sizeof(welcome_wchar_t)/sizeof(wchar_t),
+        (const char*)welcome_utf7_iconv,
+        sizeof(welcome_utf7_iconv),
+        convUTF7,
+        1
+        );
+    TestDecoder
+        (
+        (const wchar_t*)welcome_wchar_t,
+        sizeof(welcome_wchar_t)/sizeof(wchar_t),
+        (const char*)welcome_utf7_wx,
+        sizeof(welcome_utf7_wx),
+        convUTF7,
+        1
+        );
+#if 0
+    // wxWidget's UTF-7 encoder generates different byte sequences than iconv's.
+    // but both seem to be equally legal.
+    // This test won't work and that's okay.
+    TestEncoder
+        (
+        (const wchar_t*)welcome_wchar_t,
+        sizeof(welcome_wchar_t)/sizeof(wchar_t),
+        (const char*)welcome_utf7_iconv,
+        sizeof(welcome_utf7_iconv),
+        convUTF7,
+        1
+        );
+#endif
+    TestEncoder
+        (
+        (const wchar_t*)welcome_wchar_t,
+        sizeof(welcome_wchar_t)/sizeof(wchar_t),
+        (const char*)welcome_utf7_wx,
+        sizeof(welcome_utf7_wx),
+        convUTF7,
+        1
+        );
+}
+
+void MBConvTestCase::UTF8Tests()
+{
+    TestDecoder
+        (
+        (const wchar_t*)welcome_wchar_t,
+        sizeof(welcome_wchar_t)/sizeof(wchar_t),
+        (const char*)welcome_utf8,
+        sizeof(welcome_utf8),
+        wxConvUTF8,
+        1
+        );
+    TestEncoder
+        (
+        (const wchar_t*)welcome_wchar_t,
+        sizeof(welcome_wchar_t)/sizeof(wchar_t),
+        (const char*)welcome_utf8,
+        sizeof(welcome_utf8),
+        wxConvUTF8,
+        1
+        );
+}
+
+void MBConvTestCase::UTF16LETests()
+{
+    wxMBConvUTF16LE convUTF16LE;
+    TestDecoder
+        (
+        (const wchar_t*)welcome_wchar_t,
+        sizeof(welcome_wchar_t)/sizeof(wchar_t),
+        (const char*)welcome_utf16le,
+        sizeof(welcome_utf16le),
+        convUTF16LE,
+        2
+        );
+    TestEncoder
+        (
+        (const wchar_t*)welcome_wchar_t,
+        sizeof(welcome_wchar_t)/sizeof(wchar_t),
+        (const char*)welcome_utf16le,
+        sizeof(welcome_utf16le),
+        convUTF16LE,
+        2
+        );
+}
+
+void MBConvTestCase::UTF16BETests()
+{
+    wxMBConvUTF16BE convUTF16BE;
+    TestDecoder
+        (
+        (const wchar_t*)welcome_wchar_t,
+        sizeof(welcome_wchar_t)/sizeof(wchar_t),
+        (const char*)welcome_utf16be,
+        sizeof(welcome_utf16be),
+        convUTF16BE,
+        2
+        );
+    TestEncoder
+        (
+        (const wchar_t*)welcome_wchar_t,
+        sizeof(welcome_wchar_t)/sizeof(wchar_t),
+        (const char*)welcome_utf16be,
+        sizeof(welcome_utf16be),
+        convUTF16BE,
+        2
+        );
+}
+
+void MBConvTestCase::UTF32LETests()
+{
+    wxMBConvUTF32LE convUTF32LE;
+    TestDecoder
+        (
+        (const wchar_t*)welcome_wchar_t,
+        sizeof(welcome_wchar_t)/sizeof(wchar_t),
+        (const char*)welcome_utf32le,
+        sizeof(welcome_utf32le),
+        convUTF32LE,
+        4
+        );
+    TestEncoder
+        (
+        (const wchar_t*)welcome_wchar_t,
+        sizeof(welcome_wchar_t)/sizeof(wchar_t),
+        (const char*)welcome_utf32le,
+        sizeof(welcome_utf32le),
+        convUTF32LE,
+        4
+        );
+}
+
+void MBConvTestCase::UTF32BETests()
+{
+    wxMBConvUTF32BE convUTF32BE;
+    TestDecoder
+        (
+        (const wchar_t*)welcome_wchar_t,
+        sizeof(welcome_wchar_t)/sizeof(wchar_t),
+        (const char*)welcome_utf32be,
+        sizeof(welcome_utf32be),
+        convUTF32BE,
+        4
+        );
+    TestEncoder
+        (
+        (const wchar_t*)welcome_wchar_t,
+        sizeof(welcome_wchar_t)/sizeof(wchar_t),
+        (const char*)welcome_utf32be,
+        sizeof(welcome_utf32be),
+        convUTF32BE,
+        4
+        );
+}
+
+void MBConvTestCase::CP932Tests()
+{
+    wxCSConv convCP932( wxFONTENCODING_CP932 );
+    TestDecoder
+        (
+        (const wchar_t*)welcome_wchar_t,
+        sizeof(welcome_wchar_t)/sizeof(wchar_t),
+        (const char*)welcome_cp932,
+        sizeof(welcome_cp932),
+        convCP932,
+        1
+        );
+    TestEncoder
+        (
+        (const wchar_t*)welcome_wchar_t,
+        sizeof(welcome_wchar_t)/sizeof(wchar_t),
+        (const char*)welcome_cp932,
+        sizeof(welcome_cp932),
+        convCP932,
+        1
+        );
+}
+
+// a character sequence encoded as iso8859-1 (iconv)
+static const unsigned char iso8859_1[251] =
+{
+    0x05,0x06,0x07,0x08,0x09,0x0a,0x0b,0x0c,0x0d,0x0e,0x0f,0x10,0x11,0x12,0x13,0x14,
+    0x15,0x16,0x17,0x18,0x19,0x1a,0x1b,0x1c,0x1d,0x1e,0x1f,0x20,0x21,0x22,0x23,0x24,
+    0x25,0x26,0x27,0x28,0x29,0x2a,0x2b,0x2c,0x2d,0x2e,0x2f,0x30,0x31,0x32,0x33,0x34,
+    0x35,0x36,0x37,0x38,0x39,0x3a,0x3b,0x3c,0x3d,0x3e,0x3f,0x40,0x41,0x42,0x43,0x44,
+    0x45,0x46,0x47,0x48,0x49,0x4a,0x4b,0x4c,0x4d,0x4e,0x4f,0x50,0x51,0x52,0x53,0x54,
+    0x55,0x56,0x57,0x58,0x59,0x5a,0x5b,0x5c,0x5d,0x5e,0x5f,0x60,0x61,0x62,0x63,0x64,
+    0x65,0x66,0x67,0x68,0x69,0x6a,0x6b,0x6c,0x6d,0x6e,0x6f,0x70,0x71,0x72,0x73,0x74,
+    0x75,0x76,0x77,0x78,0x79,0x7a,0x7b,0x7c,0x7d,0x7e,0x7f,0x80,0x81,0x82,0x83,0x84,
+    0x85,0x86,0x87,0x88,0x89,0x8a,0x8b,0x8c,0x8d,0x8e,0x8f,0x90,0x91,0x92,0x93,0x94,
+    0x95,0x96,0x97,0x98,0x99,0x9a,0x9b,0x9c,0x9d,0x9e,0x9f,0xa0,0xa1,0xa2,0xa3,0xa4,
+    0xa5,0xa6,0xa7,0xa8,0xa9,0xaa,0xab,0xac,0xad,0xae,0xaf,0xb0,0xb1,0xb2,0xb3,0xb4,
+    0xb5,0xb6,0xb7,0xb8,0xb9,0xba,0xbb,0xbc,0xbd,0xbe,0xbf,0xc0,0xc1,0xc2,0xc3,0xc4,
+    0xc5,0xc6,0xc7,0xc8,0xc9,0xca,0xcb,0xcc,0xcd,0xce,0xcf,0xd0,0xd1,0xd2,0xd3,0xd4,
+    0xd5,0xd6,0xd7,0xd8,0xd9,0xda,0xdb,0xdc,0xdd,0xde,0xdf,0xe0,0xe1,0xe2,0xe3,0xe4,
+    0xe5,0xe6,0xe7,0xe8,0xe9,0xea,0xeb,0xec,0xed,0xee,0xef,0xf0,0xf1,0xf2,0xf3,0xf4,
+    0xf5,0xf6,0xf7,0xf8,0xf9,0xfa,0xfb,0xfc,0xfd,0xfe,0xff
+};
+// the above character sequence encoded as UTF-8 (iconv)
+static const unsigned char iso8859_1_utf8[379] =
+{
+    0x05,0x06,0x07,0x08,0x09,0x0a,0x0b,0x0c,0x0d,0x0e,0x0f,0x10,0x11,0x12,0x13,0x14,
+    0x15,0x16,0x17,0x18,0x19,0x1a,0x1b,0x1c,0x1d,0x1e,0x1f,0x20,0x21,0x22,0x23,0x24,
+    0x25,0x26,0x27,0x28,0x29,0x2a,0x2b,0x2c,0x2d,0x2e,0x2f,0x30,0x31,0x32,0x33,0x34,
+    0x35,0x36,0x37,0x38,0x39,0x3a,0x3b,0x3c,0x3d,0x3e,0x3f,0x40,0x41,0x42,0x43,0x44,
+    0x45,0x46,0x47,0x48,0x49,0x4a,0x4b,0x4c,0x4d,0x4e,0x4f,0x50,0x51,0x52,0x53,0x54,
+    0x55,0x56,0x57,0x58,0x59,0x5a,0x5b,0x5c,0x5d,0x5e,0x5f,0x60,0x61,0x62,0x63,0x64,
+    0x65,0x66,0x67,0x68,0x69,0x6a,0x6b,0x6c,0x6d,0x6e,0x6f,0x70,0x71,0x72,0x73,0x74,
+    0x75,0x76,0x77,0x78,0x79,0x7a,0x7b,0x7c,0x7d,0x7e,0x7f,0xc2,0x80,0xc2,0x81,0xc2,
+    0x82,0xc2,0x83,0xc2,0x84,0xc2,0x85,0xc2,0x86,0xc2,0x87,0xc2,0x88,0xc2,0x89,0xc2,
+    0x8a,0xc2,0x8b,0xc2,0x8c,0xc2,0x8d,0xc2,0x8e,0xc2,0x8f,0xc2,0x90,0xc2,0x91,0xc2,
+    0x92,0xc2,0x93,0xc2,0x94,0xc2,0x95,0xc2,0x96,0xc2,0x97,0xc2,0x98,0xc2,0x99,0xc2,
+    0x9a,0xc2,0x9b,0xc2,0x9c,0xc2,0x9d,0xc2,0x9e,0xc2,0x9f,0xc2,0xa0,0xc2,0xa1,0xc2,
+    0xa2,0xc2,0xa3,0xc2,0xa4,0xc2,0xa5,0xc2,0xa6,0xc2,0xa7,0xc2,0xa8,0xc2,0xa9,0xc2,
+    0xaa,0xc2,0xab,0xc2,0xac,0xc2,0xad,0xc2,0xae,0xc2,0xaf,0xc2,0xb0,0xc2,0xb1,0xc2,
+    0xb2,0xc2,0xb3,0xc2,0xb4,0xc2,0xb5,0xc2,0xb6,0xc2,0xb7,0xc2,0xb8,0xc2,0xb9,0xc2,
+    0xba,0xc2,0xbb,0xc2,0xbc,0xc2,0xbd,0xc2,0xbe,0xc2,0xbf,0xc3,0x80,0xc3,0x81,0xc3,
+    0x82,0xc3,0x83,0xc3,0x84,0xc3,0x85,0xc3,0x86,0xc3,0x87,0xc3,0x88,0xc3,0x89,0xc3,
+    0x8a,0xc3,0x8b,0xc3,0x8c,0xc3,0x8d,0xc3,0x8e,0xc3,0x8f,0xc3,0x90,0xc3,0x91,0xc3,
+    0x92,0xc3,0x93,0xc3,0x94,0xc3,0x95,0xc3,0x96,0xc3,0x97,0xc3,0x98,0xc3,0x99,0xc3,
+    0x9a,0xc3,0x9b,0xc3,0x9c,0xc3,0x9d,0xc3,0x9e,0xc3,0x9f,0xc3,0xa0,0xc3,0xa1,0xc3,
+    0xa2,0xc3,0xa3,0xc3,0xa4,0xc3,0xa5,0xc3,0xa6,0xc3,0xa7,0xc3,0xa8,0xc3,0xa9,0xc3,
+    0xaa,0xc3,0xab,0xc3,0xac,0xc3,0xad,0xc3,0xae,0xc3,0xaf,0xc3,0xb0,0xc3,0xb1,0xc3,
+    0xb2,0xc3,0xb3,0xc3,0xb4,0xc3,0xb5,0xc3,0xb6,0xc3,0xb7,0xc3,0xb8,0xc3,0xb9,0xc3,
+    0xba,0xc3,0xbb,0xc3,0xbc,0xc3,0xbd,0xc3,0xbe,0xc3,0xbf
+};
+
+// a character sequence encoded as CP1252 (iconv)
+static const unsigned char CP1252[246] =
+{
+    0x05,0x06,0x07,0x08,0x09,0x0a,0x0b,0x0c,0x0d,0x0e,0x0f,0x10,0x11,0x12,0x13,0x14,
+    0x15,0x16,0x17,0x18,0x19,0x1a,0x1b,0x1c,0x1d,0x1e,0x1f,0x20,0x21,0x22,0x23,0x24,
+    0x25,0x26,0x27,0x28,0x29,0x2a,0x2b,0x2c,0x2d,0x2e,0x2f,0x30,0x31,0x32,0x33,0x34,
+    0x35,0x36,0x37,0x38,0x39,0x3a,0x3b,0x3c,0x3d,0x3e,0x3f,0x40,0x41,0x42,0x43,0x44,
+    0x45,0x46,0x47,0x48,0x49,0x4a,0x4b,0x4c,0x4d,0x4e,0x4f,0x50,0x51,0x52,0x53,0x54,
+    0x55,0x56,0x57,0x58,0x59,0x5a,0x5b,0x5c,0x5d,0x5e,0x5f,0x60,0x61,0x62,0x63,0x64,
+    0x65,0x66,0x67,0x68,0x69,0x6a,0x6b,0x6c,0x6d,0x6e,0x6f,0x70,0x71,0x72,0x73,0x74,
+    0x75,0x76,0x77,0x78,0x79,0x7a,0x7b,0x7c,0x7d,0x7e,0x7f,0xa0,0xa1,0xa2,0xa3,0xa4,
+    0xa5,0xa6,0xa7,0xa8,0xa9,0xaa,0xab,0xac,0xad,0xae,0xaf,0xb0,0xb1,0xb2,0xb3,0xb4,
+    0xb5,0xb6,0xb7,0xb8,0xb9,0xba,0xbb,0xbc,0xbd,0xbe,0xbf,0xc0,0xc1,0xc2,0xc3,0xc4,
+    0xc5,0xc6,0xc7,0xc8,0xc9,0xca,0xcb,0xcc,0xcd,0xce,0xcf,0xd0,0xd1,0xd2,0xd3,0xd4,
+    0xd5,0xd6,0xd7,0xd8,0xd9,0xda,0xdb,0xdc,0xdd,0xde,0xdf,0xe0,0xe1,0xe2,0xe3,0xe4,
+    0xe5,0xe6,0xe7,0xe8,0xe9,0xea,0xeb,0xec,0xed,0xee,0xef,0xf0,0xf1,0xf2,0xf3,0xf4,
+    0xf5,0xf6,0xf7,0xf8,0xf9,0xfa,0xfb,0xfc,0xfd,0xfe,0xff,0x8c,0x9c,0x8a,0x9a,0x9f,
+    0x8e,0x9e,0x83,0x88,0x98,0x96,0x97,0x91,0x92,0x82,0x93,0x94,0x84,0x86,0x87,0x95,
+    0x85,0x89,0x8b,0x9b,0x80,0x99
+};
+// the above character sequence encoded as UTF-8 (iconv)
+static const unsigned char CP1252_utf8[386] =
+{
+    0x05,0x06,0x07,0x08,0x09,0x0a,0x0b,0x0c,0x0d,0x0e,0x0f,0x10,0x11,0x12,0x13,0x14,
+    0x15,0x16,0x17,0x18,0x19,0x1a,0x1b,0x1c,0x1d,0x1e,0x1f,0x20,0x21,0x22,0x23,0x24,
+    0x25,0x26,0x27,0x28,0x29,0x2a,0x2b,0x2c,0x2d,0x2e,0x2f,0x30,0x31,0x32,0x33,0x34,
+    0x35,0x36,0x37,0x38,0x39,0x3a,0x3b,0x3c,0x3d,0x3e,0x3f,0x40,0x41,0x42,0x43,0x44,
+    0x45,0x46,0x47,0x48,0x49,0x4a,0x4b,0x4c,0x4d,0x4e,0x4f,0x50,0x51,0x52,0x53,0x54,
+    0x55,0x56,0x57,0x58,0x59,0x5a,0x5b,0x5c,0x5d,0x5e,0x5f,0x60,0x61,0x62,0x63,0x64,
+    0x65,0x66,0x67,0x68,0x69,0x6a,0x6b,0x6c,0x6d,0x6e,0x6f,0x70,0x71,0x72,0x73,0x74,
+    0x75,0x76,0x77,0x78,0x79,0x7a,0x7b,0x7c,0x7d,0x7e,0x7f,0xc2,0xa0,0xc2,0xa1,0xc2,
+    0xa2,0xc2,0xa3,0xc2,0xa4,0xc2,0xa5,0xc2,0xa6,0xc2,0xa7,0xc2,0xa8,0xc2,0xa9,0xc2,
+    0xaa,0xc2,0xab,0xc2,0xac,0xc2,0xad,0xc2,0xae,0xc2,0xaf,0xc2,0xb0,0xc2,0xb1,0xc2,
+    0xb2,0xc2,0xb3,0xc2,0xb4,0xc2,0xb5,0xc2,0xb6,0xc2,0xb7,0xc2,0xb8,0xc2,0xb9,0xc2,
+    0xba,0xc2,0xbb,0xc2,0xbc,0xc2,0xbd,0xc2,0xbe,0xc2,0xbf,0xc3,0x80,0xc3,0x81,0xc3,
+    0x82,0xc3,0x83,0xc3,0x84,0xc3,0x85,0xc3,0x86,0xc3,0x87,0xc3,0x88,0xc3,0x89,0xc3,
+    0x8a,0xc3,0x8b,0xc3,0x8c,0xc3,0x8d,0xc3,0x8e,0xc3,0x8f,0xc3,0x90,0xc3,0x91,0xc3,
+    0x92,0xc3,0x93,0xc3,0x94,0xc3,0x95,0xc3,0x96,0xc3,0x97,0xc3,0x98,0xc3,0x99,0xc3,
+    0x9a,0xc3,0x9b,0xc3,0x9c,0xc3,0x9d,0xc3,0x9e,0xc3,0x9f,0xc3,0xa0,0xc3,0xa1,0xc3,
+    0xa2,0xc3,0xa3,0xc3,0xa4,0xc3,0xa5,0xc3,0xa6,0xc3,0xa7,0xc3,0xa8,0xc3,0xa9,0xc3,
+    0xaa,0xc3,0xab,0xc3,0xac,0xc3,0xad,0xc3,0xae,0xc3,0xaf,0xc3,0xb0,0xc3,0xb1,0xc3,
+    0xb2,0xc3,0xb3,0xc3,0xb4,0xc3,0xb5,0xc3,0xb6,0xc3,0xb7,0xc3,0xb8,0xc3,0xb9,0xc3,
+    0xba,0xc3,0xbb,0xc3,0xbc,0xc3,0xbd,0xc3,0xbe,0xc3,0xbf,0xc5,0x92,0xc5,0x93,0xc5,
+    0xa0,0xc5,0xa1,0xc5,0xb8,0xc5,0xbd,0xc5,0xbe,0xc6,0x92,0xcb,0x86,0xcb,0x9c,0xe2,
+    0x80,0x93,0xe2,0x80,0x94,0xe2,0x80,0x98,0xe2,0x80,0x99,0xe2,0x80,0x9a,0xe2,0x80,
+    0x9c,0xe2,0x80,0x9d,0xe2,0x80,0x9e,0xe2,0x80,0xa0,0xe2,0x80,0xa1,0xe2,0x80,0xa2,
+    0xe2,0x80,0xa6,0xe2,0x80,0xb0,0xe2,0x80,0xb9,0xe2,0x80,0xba,0xe2,0x82,0xac,0xe2,
+    0x84,0xa2
+};
+
+// this is unused currently so avoid warnings about this
+#if 0
+
+// a character sequence encoded as iso8859-5 (iconv)
+static const unsigned char iso8859_5[251] =
+{
+    0x05,0x06,0x07,0x08,0x09,0x0a,0x0b,0x0c,0x0d,0x0e,0x0f,0x10,0x11,0x12,0x13,0x14,
+    0x15,0x16,0x17,0x18,0x19,0x1a,0x1b,0x1c,0x1d,0x1e,0x1f,0x20,0x21,0x22,0x23,0x24,
+    0x25,0x26,0x27,0x28,0x29,0x2a,0x2b,0x2c,0x2d,0x2e,0x2f,0x30,0x31,0x32,0x33,0x34,
+    0x35,0x36,0x37,0x38,0x39,0x3a,0x3b,0x3c,0x3d,0x3e,0x3f,0x40,0x41,0x42,0x43,0x44,
+    0x45,0x46,0x47,0x48,0x49,0x4a,0x4b,0x4c,0x4d,0x4e,0x4f,0x50,0x51,0x52,0x53,0x54,
+    0x55,0x56,0x57,0x58,0x59,0x5a,0x5b,0x5c,0x5d,0x5e,0x5f,0x60,0x61,0x62,0x63,0x64,
+    0x65,0x66,0x67,0x68,0x69,0x6a,0x6b,0x6c,0x6d,0x6e,0x6f,0x70,0x71,0x72,0x73,0x74,
+    0x75,0x76,0x77,0x78,0x79,0x7a,0x7b,0x7c,0x7d,0x7e,0x7f,0x80,0x81,0x82,0x83,0x84,
+    0x85,0x86,0x87,0x88,0x89,0x8a,0x8b,0x8c,0x8d,0x8e,0x8f,0x90,0x91,0x92,0x93,0x94,
+    0x95,0x96,0x97,0x98,0x99,0x9a,0x9b,0x9c,0x9d,0x9e,0x9f,0xa0,0xfd,0xad,0xa1,0xa2,
+    0xa3,0xa4,0xa5,0xa6,0xa7,0xa8,0xa9,0xaa,0xab,0xac,0xae,0xaf,0xb0,0xb1,0xb2,0xb3,
+    0xb4,0xb5,0xb6,0xb7,0xb8,0xb9,0xba,0xbb,0xbc,0xbd,0xbe,0xbf,0xc0,0xc1,0xc2,0xc3,
+    0xc4,0xc5,0xc6,0xc7,0xc8,0xc9,0xca,0xcb,0xcc,0xcd,0xce,0xcf,0xd0,0xd1,0xd2,0xd3,
+    0xd4,0xd5,0xd6,0xd7,0xd8,0xd9,0xda,0xdb,0xdc,0xdd,0xde,0xdf,0xe0,0xe1,0xe2,0xe3,
+    0xe4,0xe5,0xe6,0xe7,0xe8,0xe9,0xea,0xeb,0xec,0xed,0xee,0xef,0xf1,0xf2,0xf3,0xf4,
+    0xf5,0xf6,0xf7,0xf8,0xf9,0xfa,0xfb,0xfc,0xfe,0xff,0xf0
+};
+// the above character sequence encoded as UTF-8 (iconv)
+static const unsigned char iso8859_5_utf8[380] =
+{
+    0x05,0x06,0x07,0x08,0x09,0x0a,0x0b,0x0c,0x0d,0x0e,0x0f,0x10,0x11,0x12,0x13,0x14,
+    0x15,0x16,0x17,0x18,0x19,0x1a,0x1b,0x1c,0x1d,0x1e,0x1f,0x20,0x21,0x22,0x23,0x24,
+    0x25,0x26,0x27,0x28,0x29,0x2a,0x2b,0x2c,0x2d,0x2e,0x2f,0x30,0x31,0x32,0x33,0x34,
+    0x35,0x36,0x37,0x38,0x39,0x3a,0x3b,0x3c,0x3d,0x3e,0x3f,0x40,0x41,0x42,0x43,0x44,
+    0x45,0x46,0x47,0x48,0x49,0x4a,0x4b,0x4c,0x4d,0x4e,0x4f,0x50,0x51,0x52,0x53,0x54,
+    0x55,0x56,0x57,0x58,0x59,0x5a,0x5b,0x5c,0x5d,0x5e,0x5f,0x60,0x61,0x62,0x63,0x64,
+    0x65,0x66,0x67,0x68,0x69,0x6a,0x6b,0x6c,0x6d,0x6e,0x6f,0x70,0x71,0x72,0x73,0x74,
+    0x75,0x76,0x77,0x78,0x79,0x7a,0x7b,0x7c,0x7d,0x7e,0x7f,0xc2,0x80,0xc2,0x81,0xc2,
+    0x82,0xc2,0x83,0xc2,0x84,0xc2,0x85,0xc2,0x86,0xc2,0x87,0xc2,0x88,0xc2,0x89,0xc2,
+    0x8a,0xc2,0x8b,0xc2,0x8c,0xc2,0x8d,0xc2,0x8e,0xc2,0x8f,0xc2,0x90,0xc2,0x91,0xc2,
+    0x92,0xc2,0x93,0xc2,0x94,0xc2,0x95,0xc2,0x96,0xc2,0x97,0xc2,0x98,0xc2,0x99,0xc2,
+    0x9a,0xc2,0x9b,0xc2,0x9c,0xc2,0x9d,0xc2,0x9e,0xc2,0x9f,0xc2,0xa0,0xc2,0xa7,0xc2,
+    0xad,0xd0,0x81,0xd0,0x82,0xd0,0x83,0xd0,0x84,0xd0,0x85,0xd0,0x86,0xd0,0x87,0xd0,
+    0x88,0xd0,0x89,0xd0,0x8a,0xd0,0x8b,0xd0,0x8c,0xd0,0x8e,0xd0,0x8f,0xd0,0x90,0xd0,
+    0x91,0xd0,0x92,0xd0,0x93,0xd0,0x94,0xd0,0x95,0xd0,0x96,0xd0,0x97,0xd0,0x98,0xd0,
+    0x99,0xd0,0x9a,0xd0,0x9b,0xd0,0x9c,0xd0,0x9d,0xd0,0x9e,0xd0,0x9f,0xd0,0xa0,0xd0,
+    0xa1,0xd0,0xa2,0xd0,0xa3,0xd0,0xa4,0xd0,0xa5,0xd0,0xa6,0xd0,0xa7,0xd0,0xa8,0xd0,
+    0xa9,0xd0,0xaa,0xd0,0xab,0xd0,0xac,0xd0,0xad,0xd0,0xae,0xd0,0xaf,0xd0,0xb0,0xd0,
+    0xb1,0xd0,0xb2,0xd0,0xb3,0xd0,0xb4,0xd0,0xb5,0xd0,0xb6,0xd0,0xb7,0xd0,0xb8,0xd0,
+    0xb9,0xd0,0xba,0xd0,0xbb,0xd0,0xbc,0xd0,0xbd,0xd0,0xbe,0xd0,0xbf,0xd1,0x80,0xd1,
+    0x81,0xd1,0x82,0xd1,0x83,0xd1,0x84,0xd1,0x85,0xd1,0x86,0xd1,0x87,0xd1,0x88,0xd1,
+    0x89,0xd1,0x8a,0xd1,0x8b,0xd1,0x8c,0xd1,0x8d,0xd1,0x8e,0xd1,0x8f,0xd1,0x91,0xd1,
+    0x92,0xd1,0x93,0xd1,0x94,0xd1,0x95,0xd1,0x96,0xd1,0x97,0xd1,0x98,0xd1,0x99,0xd1,
+    0x9a,0xd1,0x9b,0xd1,0x9c,0xd1,0x9e,0xd1,0x9f,0xe2,0x84,0x96
+};
+#endif // 0
+
+// DecodeUTF8
+// decodes the specified *unterminated* UTF-8 byte array
+wxWCharBuffer DecodeUTF8(
+    const void* data, // an unterminated UTF-8 encoded byte array
+    size_t size       // the byte length of data
+    )
+{
+    // the decoder requires a null terminated buffer.
+    // the input data is not null terminated.
+    // copy to null terminated buffer
+
+    wxCharBuffer nullTerminated( size+1 );
+    memcpy( nullTerminated.data(), data, size );
+    nullTerminated.data()[size] = 0;
+    return wxConvUTF8.cMB2WC(nullTerminated.data());
+}
+
+// tests the encoding and decoding capability of an wxMBConv object
+//
+// decodes the utf-8 bytes into wide characters
+// encodes the wide characters to compare against input multiBuffer
+// decodes the multiBuffer to compare against wide characters
+// decodes the multiBuffer into wide characters
+void MBConvTestCase::TestCoder(
+    const char*    multiBuffer, // a multibyte encoded character sequence that can be decoded by "converter"
+    size_t         multiBytes,  // the byte length of the multibyte character sequence that can be decoded by "converter"
+    const char*    utf8Buffer,  // the same character sequence as multiBuffer, encoded as UTF-8
+    size_t         utf8Bytes,   // the byte length of the UTF-8 encoded character sequence
+    wxMBConv&      converter,   // the wxMBConv object thta can decode multiBuffer into a wide character sequence
+    int            sizeofNull   // the number of bytes occupied by a terminating null in the converter's encoding
+    )
+{
+    // wide character size and endian-ess varies from platform to platform
+    // compiler support for wide character literals varies from compiler to compiler
+    // so we should store the wide character version as UTF-8 and depend on
+    // the UTF-8 converter's ability to decode it to platform specific wide characters
+    // this test is invalid if the UTF-8 converter can't decode
+    const wxWCharBuffer wideBuffer(DecodeUTF8(utf8Buffer, utf8Bytes));
+    const size_t wideChars = wxWcslen(wideBuffer);
+
+    TestDecoder
+        (
+        wideBuffer.data(),
+        wideChars,
+        multiBuffer,
+        multiBytes,
+        converter,
+        sizeofNull
+        );
+    TestEncoder
+        (
+        wideBuffer.data(),
+        wideChars,
+        multiBuffer,
+        multiBytes,
+        converter,
+        sizeofNull
+        );
+}
+
+
+WXDLLIMPEXP_BASE wxMBConv* new_wxMBConv_wxwin( const char* name );
+
+void MBConvTestCase::FontmapTests()
+{
+#ifdef wxUSE_FONTMAP
+    wxMBConv* converter = new_wxMBConv_wxwin("CP1252");
+    if ( !converter )
+    {
+        return;
+    }
+    TestCoder(
+        (const char*)CP1252,
+        sizeof(CP1252),
+        (const char*)CP1252_utf8,
+        sizeof(CP1252_utf8),
+        *converter,
+        1
+        );
+    delete converter;
+#endif
+}
+
+void MBConvTestCase::BufSize()
+{
+    wxCSConv conv1251(wxT("CP1251"));
+    CPPUNIT_ASSERT( conv1251.IsOk() );
+    const char *cp1251text =
+        "\313\301\326\305\324\323\321 \325\304\301\336\316\331\315";
+
+    const size_t lenW = conv1251.MB2WC(NULL, cp1251text, 0);
+    CPPUNIT_ASSERT_EQUAL( strlen(cp1251text), lenW );
+    wxWCharBuffer wbuf(lenW + 1); // allocates lenW + 2 characters
+    wbuf.data()[lenW + 1] = L'!';
+
+    // lenW is not enough because it's the length and we need the size
+    CPPUNIT_ASSERT_EQUAL(
+        wxCONV_FAILED, conv1251.MB2WC(wbuf.data(), cp1251text, lenW) );
+
+    // lenW+1 is just fine
+    CPPUNIT_ASSERT(
+        conv1251.MB2WC(wbuf.data(), cp1251text, lenW + 1) != wxCONV_FAILED );
+
+    // of course, greater values work too
+    CPPUNIT_ASSERT(
+        conv1251.MB2WC(wbuf.data(), cp1251text, lenW + 2) != wxCONV_FAILED );
+
+    // but they shouldn't write more stuff to the buffer
+    CPPUNIT_ASSERT_EQUAL( L'!', wbuf[lenW + 1] );
+
+
+    // test in the other direction too, using an encoding with multibyte NUL
+    wxCSConv convUTF16(wxT("UTF-16LE"));
+    CPPUNIT_ASSERT( convUTF16.IsOk() );
+    const wchar_t *utf16text = L"Hello";
+
+    const size_t lenMB = convUTF16.WC2MB(NULL, utf16text, 0);
+    CPPUNIT_ASSERT_EQUAL( wcslen(utf16text)*2, lenMB );
+    wxCharBuffer buf(lenMB + 2); // it only adds 1 for NUL on its own, we need 2
+                                 // for NUL and an extra one for the guard byte
+    buf.data()[lenMB + 2] = '?';
+
+    CPPUNIT_ASSERT_EQUAL(
+        wxCONV_FAILED, convUTF16.WC2MB(buf.data(), utf16text, lenMB) );
+    CPPUNIT_ASSERT_EQUAL(
+        wxCONV_FAILED, convUTF16.WC2MB(buf.data(), utf16text, lenMB + 1) );
+    CPPUNIT_ASSERT(
+        convUTF16.WC2MB(buf.data(), utf16text, lenMB + 2) != wxCONV_FAILED );
+    CPPUNIT_ASSERT(
+        convUTF16.WC2MB(buf.data(), utf16text, lenMB + 3) != wxCONV_FAILED );
+    CPPUNIT_ASSERT_EQUAL( '?', buf[lenMB + 2] );
+}
+
+void MBConvTestCase::FromWCharTests()
+{
+    wxCSConv conv950("CP950");
+    char mbuf[10];
+    // U+4e00 is 2 bytes (0xa4 0x40) in cp950
+    wchar_t wbuf[] = { 0x4e00, 0, 0x4e00, 0 };
+
+    // test simple ASCII text
+    memset(mbuf, '!', sizeof(mbuf));
+    CPPUNIT_ASSERT_EQUAL( wxCONV_FAILED, conv950.FromWChar(mbuf, 0, L"a", 1));
+    CPPUNIT_ASSERT_EQUAL( '!', mbuf[0]);
+
+    memset(mbuf, '!', sizeof(mbuf));
+    CPPUNIT_ASSERT_EQUAL( 1, conv950.FromWChar(mbuf, 1, L"a", 1));
+    CPPUNIT_ASSERT_EQUAL( 'a', mbuf[0]);
+    CPPUNIT_ASSERT_EQUAL( '!', mbuf[1]);
+
+    memset(mbuf, '!', sizeof(mbuf));
+    CPPUNIT_ASSERT_EQUAL( wxCONV_FAILED, conv950.FromWChar(mbuf, 1, L"a", 2));
+
+    memset(mbuf, '!', sizeof(mbuf));
+    CPPUNIT_ASSERT_EQUAL( 2, conv950.FromWChar(mbuf, 2, L"a", 2));
+    CPPUNIT_ASSERT_EQUAL( 'a', mbuf[0]);
+    CPPUNIT_ASSERT_EQUAL( '\0', mbuf[1]);
+    CPPUNIT_ASSERT_EQUAL( '!', mbuf[2]);
+
+    // test non-ASCII text, 1 wchar -> 2 char
+    memset(mbuf, '!', sizeof(mbuf));
+    CPPUNIT_ASSERT_EQUAL( wxCONV_FAILED, conv950.FromWChar(mbuf, 0, wbuf, 1));
+
+    memset(mbuf, '!', sizeof(mbuf));
+    CPPUNIT_ASSERT_EQUAL( wxCONV_FAILED, conv950.FromWChar(mbuf, 1, wbuf, 1));
+
+    memset(mbuf, '!', sizeof(mbuf));
+    CPPUNIT_ASSERT_EQUAL( 2, conv950.FromWChar(mbuf, 2, wbuf, 1));
+    CPPUNIT_ASSERT_EQUAL( '!', mbuf[2]);
+
+    memset(mbuf, '!', sizeof(mbuf));
+    CPPUNIT_ASSERT_EQUAL( wxCONV_FAILED, conv950.FromWChar(mbuf, 2, wbuf, 2));
+
+    memset(mbuf, '!', sizeof(mbuf));
+    CPPUNIT_ASSERT_EQUAL( 3, conv950.FromWChar(mbuf, 3, wbuf, 2));
+    CPPUNIT_ASSERT_EQUAL( '\0', mbuf[2]);
+    CPPUNIT_ASSERT_EQUAL( '!', mbuf[3]);
+
+    // test text with embedded NUL-character and srcLen specified
+    memset(mbuf, '!', sizeof(mbuf));
+    CPPUNIT_ASSERT_EQUAL( wxCONV_FAILED, conv950.FromWChar(mbuf, 3, wbuf, 3));
+
+    memset(mbuf, '!', sizeof(mbuf));
+    CPPUNIT_ASSERT_EQUAL( wxCONV_FAILED, conv950.FromWChar(mbuf, 4, wbuf, 3));
+    CPPUNIT_ASSERT_EQUAL( 5, conv950.FromWChar(mbuf, 5, wbuf, 3));
+    CPPUNIT_ASSERT_EQUAL( '\0', mbuf[2]);
+    CPPUNIT_ASSERT_EQUAL( '!', mbuf[5]);
+
+    memset(mbuf, '!', sizeof(mbuf));
+    CPPUNIT_ASSERT_EQUAL( wxCONV_FAILED, conv950.FromWChar(mbuf, 5, wbuf, 4));
+
+    memset(mbuf, '!', sizeof(mbuf));
+    CPPUNIT_ASSERT_EQUAL( 6, conv950.FromWChar(mbuf, 6, wbuf, 4));
+    CPPUNIT_ASSERT_EQUAL( '\0', mbuf[2]);
+    CPPUNIT_ASSERT_EQUAL( '\0', mbuf[5]);
+    CPPUNIT_ASSERT_EQUAL( '!', mbuf[6]);
+}
+
+WXDLLIMPEXP_BASE wxMBConv* new_wxMBConv_iconv( const char* name );
+
+void MBConvTestCase::IconvTests()
+{
+#ifdef HAVE_ICONV
+    wxMBConv* converter = new_wxMBConv_iconv("CP932");
+    if ( !converter )
+    {
+        return;
+    }
+    TestCoder(
+        (const char*)welcome_cp932,
+        sizeof(welcome_cp932),
+        (const char*)welcome_utf8,
+        sizeof(welcome_utf8),
+        *converter,
+        1
+        );
+    delete converter;
+#endif
+}
+
+void MBConvTestCase::Latin1Tests()
+{
+    TestCoder(
+        (const char*)iso8859_1,
+        sizeof(iso8859_1),
+        (const char*)iso8859_1_utf8,
+        sizeof(iso8859_1_utf8),
+        wxConvISO8859_1,
+        1
+        );
+
+    static const char nulstr[] = "foo\0bar\0";
+    static const size_t mbLen = WXSIZEOF(nulstr) - 1;
+    size_t wcLen;
+    wxConvISO8859_1.cMB2WC(nulstr, mbLen, &wcLen);
+    CPPUNIT_ASSERT_EQUAL( mbLen, wcLen );
+}
+
+void MBConvTestCase::CP1252Tests()
+{
+    wxCSConv convCP1252( wxFONTENCODING_CP1252 );
+    TestCoder(
+        (const char*)CP1252,
+        sizeof(CP1252),
+        (const char*)CP1252_utf8,
+        sizeof(CP1252_utf8),
+        convCP1252,
+        1
+        );
+}
+
+void MBConvTestCase::LibcTests()
+{
+    // The locale name are OS-dependent so this test is done only under Windows
+    // when using MSVC (surprisingly it fails with MinGW, even though it's
+    // supposed to use the same CRT -- no idea why and unfortunately gdb is too
+    // flaky to debug it)
+#ifdef __VISUALC__
+    LocaleSetter loc("English_United States.1252");
+
+    wxMBConvLibc convLibc;
+    TestCoder(
+        (const char*)CP1252,
+        sizeof(CP1252),
+        (const char*)CP1252_utf8,
+        sizeof(CP1252_utf8),
+        convLibc,
+        1
+        );
+#endif // __VISUALC__
+}
+
+// verifies that the specified mb sequences decode to the specified wc sequence
+void MBConvTestCase::TestDecoder(
+    const wchar_t* wideBuffer,  // the same character sequence as multiBuffer, encoded as wchar_t
+    size_t         wideChars,   // the number of wide characters at wideBuffer
+    const char*    multiBuffer, // a multibyte encoded character sequence that can be decoded by "converter"
+    size_t         multiBytes,  // the byte length of the multibyte character sequence that can be decoded by "converter"
+    wxMBConv&      converter,   // the wxMBConv object that can decode multiBuffer into a wide character sequence
+    int            sizeofNull   // number of bytes occupied by terminating null in this encoding
+    )
+{
+    const unsigned UNINITIALIZED = 0xcd;
+
+    // copy the input bytes into a null terminated buffer
+    wxCharBuffer inputCopy( multiBytes+sizeofNull );
+    memcpy( inputCopy.data(), multiBuffer, multiBytes );
+    memset( &inputCopy.data()[multiBytes], 0, sizeofNull );
+
+    // calculate the output size
+    size_t outputWritten = converter.MB2WC
+        (
+        0,
+        (const char*)inputCopy.data(),
+        0
+        );
+    // make sure the correct output length was calculated
+    WX_ASSERT_EQUAL_MESSAGE
+    (
+        ("while converting \"%s\"", multiBuffer),
+        wideChars,
+        outputWritten
+    );
+
+    // convert the string
+    size_t guardChars = 8; // to make sure we're not overrunning the output buffer
+    size_t nullCharacters = 1;
+    size_t outputBufferChars = outputWritten + nullCharacters + guardChars;
+    wxWCharBuffer outputBuffer(outputBufferChars);
+    memset( outputBuffer.data(), UNINITIALIZED, outputBufferChars*sizeof(wchar_t) );
+
+    outputWritten = converter.MB2WC
+        (
+        outputBuffer.data(),
+        (const char*)inputCopy.data(),
+        outputBufferChars
+        );
+    // make sure the correct number of characters were outputs
+    CPPUNIT_ASSERT_EQUAL( wideChars, outputWritten );
+
+    // make sure the characters generated are correct
+    CPPUNIT_ASSERT( 0 == memcmp( outputBuffer, wideBuffer, wideChars*sizeof(wchar_t) ) );
+
+    // the output buffer should be null terminated
+    CPPUNIT_ASSERT(  outputBuffer[outputWritten] == 0 );
+
+    // make sure the rest of the output buffer is untouched
+    for ( size_t i = (wideChars+1)*sizeof(wchar_t); i < (outputBufferChars*sizeof(wchar_t)); i++ )
+    {
+        CPPUNIT_ASSERT( ((unsigned char*)outputBuffer.data())[i] == UNINITIALIZED );
+    }
+
+#if wxUSE_UNICODE && wxUSE_STREAMS
+    TestStreamDecoder( wideBuffer, wideChars, multiBuffer, multiBytes, converter );
+#endif
+}
+
+// verifies that the specified wc sequences encodes to the specified mb sequence
+void MBConvTestCase::TestEncoder(
+    const wchar_t* wideBuffer,  // the same character sequence as multiBuffer, encoded as wchar_t
+    size_t         wideChars,   // the number of wide characters at wideBuffer
+    const char*    multiBuffer, // a multibyte encoded character sequence that can be decoded by "converter"
+    size_t         multiBytes,  // the byte length of the multibyte character sequence that can be decoded by "converter"
+    wxMBConv&      converter,   // the wxMBConv object that can decode multiBuffer into a wide character sequence
+    int            sizeofNull   // number of bytes occupied by terminating null in this encoding
+    )
+{
+    const unsigned UNINITIALIZED = 0xcd;
+
+    // copy the input bytes into a null terminated buffer
+    wxWCharBuffer inputCopy( wideChars + 1 );
+    memcpy( inputCopy.data(), wideBuffer, (wideChars*sizeof(wchar_t)) );
+    inputCopy.data()[wideChars] = 0;
+
+    // calculate the output size
+    size_t outputWritten = converter.WC2MB
+        (
+        0,
+        (const wchar_t*)inputCopy.data(),
+        0
+        );
+    // make sure the correct output length was calculated
+    CPPUNIT_ASSERT_EQUAL( multiBytes, outputWritten );
+
+    // convert the string
+    size_t guardBytes = 8; // to make sure we're not overrunning the output buffer
+    size_t outputBufferSize = outputWritten + sizeofNull + guardBytes;
+    wxCharBuffer outputBuffer(outputBufferSize);
+    memset( outputBuffer.data(), UNINITIALIZED, outputBufferSize );
+
+    outputWritten = converter.WC2MB
+        (
+        outputBuffer.data(),
+        (const wchar_t*)inputCopy.data(),
+        outputBufferSize
+        );
+
+    // make sure the correct number of characters were output
+    CPPUNIT_ASSERT_EQUAL( multiBytes, outputWritten );
+
+    // make sure the characters generated are correct
+    CPPUNIT_ASSERT( 0 == memcmp( outputBuffer, multiBuffer, multiBytes ) );
+
+    size_t i;
+
+    // the output buffer should be null terminated
+    for ( i = multiBytes; i < multiBytes + sizeofNull; i++ )
+    {
+        CPPUNIT_ASSERT( ((unsigned char*)outputBuffer.data())[i] == 0 );
+    }
+
+    // make sure the rest of the output buffer is untouched
+    for ( i = multiBytes + sizeofNull; i < outputBufferSize; i++ )
+    {
+        CPPUNIT_ASSERT( ((unsigned char*)outputBuffer.data())[i] == UNINITIALIZED );
+    }
+
+#if wxUSE_UNICODE && wxUSE_STREAMS
+    TestStreamEncoder( wideBuffer, wideChars, multiBuffer, multiBytes, converter );
+#endif
+}
+
+#if wxUSE_UNICODE && wxUSE_STREAMS
+// use wxTextInputStream to exercise wxMBConv interface
+// (this reveals some bugs in certain wxMBConv subclasses)
+void MBConvTestCase::TestStreamDecoder(
+                                       const wchar_t* wideBuffer,  // the same character sequence as multiBuffer, encoded as wchar_t
+                                       size_t         wideChars,   // the number of wide characters at wideBuffer
+                                       const char*    multiBuffer, // a multibyte encoded character sequence that can be decoded by "converter"
+                                       size_t         multiBytes,  // the byte length of the multibyte character sequence that can be decoded by "converter"
+                                       wxMBConv&      converter    // the wxMBConv object that can decode multiBuffer into a wide character sequence
+                                       )
+{
+    // this isn't meant to test wxMemoryInputStream or wxTextInputStream
+    // it's meant to test the way wxTextInputStream uses wxMBConv
+    // (which has exposed some problems with wxMBConv)
+    wxMemoryInputStream memoryInputStream( multiBuffer, multiBytes );
+    wxTextInputStream textInputStream( memoryInputStream, wxT(""), converter );
+    for ( size_t i = 0; i < wideChars; i++ )
+    {
+        wxChar wc = textInputStream.GetChar();
+        CPPUNIT_ASSERT_EQUAL_MESSAGE(
+            std::string(wxString::Format("At position %lu", (unsigned long)i)),
+            wideBuffer[i],
+            wc
+        );
+    }
+    CPPUNIT_ASSERT( 0 == textInputStream.GetChar() );
+    CPPUNIT_ASSERT( memoryInputStream.Eof() );
+}
+#endif
+
+#if wxUSE_UNICODE && wxUSE_STREAMS
+// use wxTextInputStream to exercise wxMBConv interface
+// (this reveals some bugs in certain wxMBConv subclasses)
+void MBConvTestCase::TestStreamEncoder(
+    const wchar_t* wideBuffer,  // the same character sequence as multiBuffer, encoded as wchar_t
+    size_t         wideChars,   // the number of wide characters at wideBuffer
+    const char*    multiBuffer, // a multibyte encoded character sequence that can be decoded by "converter"
+    size_t         multiBytes,  // the byte length of the multibyte character sequence that can be decoded by "converter"
+    wxMBConv&      converter    // the wxMBConv object that can decode multiBuffer into a wide character sequence
+    )
+{
+    // this isn't meant to test wxMemoryOutputStream or wxTextOutputStream
+    // it's meant to test the way wxTextOutputStream uses wxMBConv
+    // (which has exposed some problems with wxMBConv)
+    wxMemoryOutputStream memoryOutputStream;
+    // wxEOL_UNIX will pass \n \r unchanged
+    wxTextOutputStream textOutputStream( memoryOutputStream, wxEOL_UNIX, converter );
+    for ( size_t i = 0; i < wideChars; i++ )
+    {
+        textOutputStream.PutChar( wideBuffer[i] );
+    }
+
+    textOutputStream.Flush();
+
+    CPPUNIT_ASSERT_EQUAL( multiBytes, size_t(memoryOutputStream.TellO()) );
+    wxCharBuffer copy( memoryOutputStream.TellO() );
+    memoryOutputStream.CopyTo( copy.data(), memoryOutputStream.TellO());
+    CPPUNIT_ASSERT_EQUAL( 0, memcmp( copy.data(), multiBuffer, multiBytes ) );
+}
+#endif
+
+
 // ----------------------------------------------------------------------------
 // UTF-8 tests
 // ----------------------------------------------------------------------------
@@ -205,7 +1240,7 @@ void MBConvTestCase::UTF8Octal(const char *charSequence,
 //
 static wchar_t *wx_wcscpy(wchar_t *dest, const wchar_t *src)
 {
-    wchar_t *d = dest; 
+    wchar_t *d = dest;
     while ((*d++ = *src++) != 0)
         ;
     return dest;
@@ -215,7 +1250,7 @@ static wchar_t *wx_wcscpy(wchar_t *dest, const wchar_t *src)
 //
 static wchar_t *wx_wcscat(wchar_t *dest, const wchar_t *src)
 {
-    wchar_t *d = dest; 
+    wchar_t *d = dest;
     while (*d)
         d++;
     while ((*d++ = *src++) != 0)
@@ -223,6 +1258,28 @@ static wchar_t *wx_wcscat(wchar_t *dest, const wchar_t *src)
     return dest;
 }
 
+// in case wcscmp is missing
+//
+static int wx_wcscmp(const wchar_t *s1, const wchar_t *s2)
+{
+    while (*s1 == *s2 && *s1 != 0)
+    {
+        s1++;
+        s2++;
+    }
+    return *s1 - *s2;
+}
+
+// in case wcslen is missing
+//
+static size_t wx_wcslen(const wchar_t *s)
+{
+    const wchar_t *t = s;
+    while (*t != 0)
+        t++;
+    return t - s;
+}
+
 // include the option in the error messages so it's possible to see which
 // test failed
 #define UTF8ASSERT(expr) CPPUNIT_ASSERT_MESSAGE(#expr + errmsg,  expr)
@@ -236,12 +1293,12 @@ void MBConvTestCase::UTF8(const char *charSequence,
     const size_t BUFSIZE = 128;
     wxASSERT(strlen(charSequence) * 3 + 10 < BUFSIZE);
     char bytes[BUFSIZE];
-    
+
     // include the option in the error messages so it's possible to see
     // which test failed
     sprintf(bytes, " (with option == %d)", option);
     std::string errmsg(bytes);
-    
+
     // put the charSequence at the start, middle and end of a string
     strcpy(bytes, charSequence);
     strcat(bytes, "ABC");
@@ -268,8 +1325,8 @@ void MBConvTestCase::UTF8(const char *charSequence,
         wx_wcscat(expected, L"XYZ");
         wx_wcscat(expected, wideSequence);
 
-        UTF8ASSERT(wcscmp(widechars, expected) == 0);
-        UTF8ASSERT(wcslen(widechars) == result);
+        UTF8ASSERT(wx_wcscmp(widechars, expected) == 0);
+        UTF8ASSERT(wx_wcslen(widechars) == result);
     }
     else {
         // If 'wideSequence' is NULL, then the result is expected to be