X-Git-Url: https://git.saurik.com/wxWidgets.git/blobdiff_plain/394579cf6fd79c813687153bc84559c91974c556..7c71eb6a2607235a29ffb29e3c39bb4f1a74fff7:/tests/mbconv/mbconvtest.cpp?ds=inline diff --git a/tests/mbconv/mbconvtest.cpp b/tests/mbconv/mbconvtest.cpp index 51eb23e96f..6d385495ef 100644 --- a/tests/mbconv/mbconvtest.cpp +++ b/tests/mbconv/mbconvtest.cpp @@ -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 +67,18 @@ 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( FontmapTests ); #ifdef HAVE_WCHAR_H CPPUNIT_TEST( UTF8_41 ); CPPUNIT_TEST( UTF8_7f ); @@ -89,6 +102,74 @@ 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 IconvTests(); + + // 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 @@ -132,6 +213,7 @@ CPPUNIT_TEST_SUITE_REGISTRATION( MBConvTestCase ); // also include in it's 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 +242,812 @@ 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( _T("const static unsigned char %s[%i] = \n{"), name, (int)len ); + + for ( size_t i = 0; i < len; i++ ) + { + if ( i != 0 ) + { + result.append( _T(",") ); + } + if ((i%16)==0) + { + result.append( _T("\n ") ); + } + wxString byte = wxString::Format( _T("0x%02x"), bytes[i] ); + result.append(byte); + } + result.append( _T("\n};\n") ); + return result; +} + +// The following bytes represent the same string, containing Japanese and English +// characters, encoded in several different formats. + +// encoded by iconv +const static 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) +const static 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 +const static 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 +const static 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 +const static 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 +const static 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 +const static 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 +const static 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() +{ + TestDecoder + ( + (const wchar_t*)welcome_wchar_t, + sizeof(welcome_wchar_t)/sizeof(wchar_t), + (const char*)welcome_utf7_iconv, + sizeof(welcome_utf7_iconv), + wxConvUTF7, + 1 + ); + TestDecoder + ( + (const wchar_t*)welcome_wchar_t, + sizeof(welcome_wchar_t)/sizeof(wchar_t), + (const char*)welcome_utf7_wx, + sizeof(welcome_utf7_wx), + wxConvUTF7, + 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), + wxConvUTF7, + 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), + wxConvUTF7, + 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 +}; + +// 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 +}; + +// 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 + wxWCharBuffer wideBuffer((size_t)0); + wideBuffer = DecodeUTF8( utf8Buffer, utf8Bytes ); + size_t wideChars = wxWcslen( wideBuffer.data() ); + + TestDecoder + ( + wideBuffer.data(), + wideChars, + (const char*)multiBuffer, + multiBytes, + converter, + sizeofNull + ); + TestEncoder + ( + wideBuffer.data(), + wideChars, + (const char*)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 +} + + +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::CP1252Tests() +{ + wxCSConv convCP1252( wxFONTENCODING_CP1252 ); + TestCoder( + (const char*)CP1252, + sizeof(CP1252), + (const char*)CP1252_utf8, + sizeof(CP1252_utf8), + convCP1252, + 1 + ); +} + +void MBConvTestCase::LibcTests() +{ + // There isn't a locale that all systems support (except "C"), so leave + // this one disabled for non-Windows systems for the moment, until + // a solution can be found. +#ifdef __WXMSW__ + +#ifdef __WXMSW__ + setlocale( LC_ALL, "English_United States.1252" ); + const unsigned char* systemMB = CP1252; + size_t systemMB_size = sizeof(CP1252); + const unsigned char* systemMB_utf8 = CP1252_utf8; + size_t systemMB_utf8_size = sizeof(CP1252_utf8); +#else + setlocale( LC_ALL, "en_US.iso8859-1" ); + const unsigned char* systemMB = iso8859_1; + size_t systemMB_size = sizeof(iso8859_1); + const unsigned char* systemMB_utf8 = iso8859_1_utf8; + size_t systemMB_utf8_size = sizeof(iso8859_1_utf8); +#endif + wxMBConvLibc convLibc; + TestCoder( + (const char*)systemMB, + systemMB_size, + (const char*)systemMB_utf8, + systemMB_utf8_size, + convLibc, + 1 + ); + +#endif // __WXMSW__ +} + +// 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 + CPPUNIT_ASSERT( outputWritten == wideChars ); + + // 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( outputWritten == wideChars ); + + // 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( outputWritten == multiBytes ); + + // 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( outputWritten == multiBytes ); + + // 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( wc == wideBuffer[i] ); + } + 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] ); + } + CPPUNIT_ASSERT_EQUAL( (wxFileOffset)multiBytes, 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 // ---------------------------------------------------------------------------- @@ -223,6 +1111,28 @@ 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) @@ -268,8 +1178,8 @@ 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