reverted the change of r56246 and keep the old behaviour of MB2WC/WC2MB; document it even more clearly and correct code using these functions incorrectly

git-svn-id: https://svn.wxwidgets.org/svn/wx/wxWidgets/branches/WX_2_8_BRANCH@56394 c3d73ce0-8a6f-49c7-b76d-6d57e0e08775

docs/changes.txt

@@ -91,10 +91,6 @@ Major new features in 2.8 release
 2.8.10:
 -------
 
-All:
-
-- Fixed off by one bug in wxMBConv::MB2WC/WC2MB() functions.
-
 All (GUI):
 
 - Fixed wxHTML's pagebreaks computation in tables (D.J.Stauffer).
@@ -110,6 +106,7 @@ All (Unix):
 
 wxMSW:
 
+- Fixed bug with using non default encoding in wxTextCtrl in ANSI build.
 - Restored generation of events for accelerators when there is no menu bar.
 - Fixed wxArtProvider::GetSizeHint() to return 16x16 for wxART_FRAME_ICON.
 

src/common/strconv.cpp

@@ -332,9 +332,7 @@ wxMBConv::FromWChar(char *dst, size_t dstLen,
 
 size_t wxMBConv::MB2WC(wchar_t *outBuff, const char *inBuff, size_t outLen) const
 {
-    // add 1 to available buffer length because MB2WC() parameter counts the
+    size_t rc = ToWChar(outBuff, outLen, inBuff);
-    // number of non-NUL characters while ToWChar() counts everything
-    size_t rc = ToWChar(outBuff, outLen + 1, inBuff);
     if ( rc != wxCONV_FAILED )
     {
         // ToWChar() returns the buffer length, i.e. including the trailing
@@ -347,12 +345,10 @@ size_t wxMBConv::MB2WC(wchar_t *outBuff, const char *inBuff, size_t outLen) cons
 
 size_t wxMBConv::WC2MB(char *outBuff, const wchar_t *inBuff, size_t outLen) const
 {
-    const size_t nulLen = GetMBNulLen();
+    size_t rc = FromWChar(outBuff, outLen, inBuff);
-
-    size_t rc = FromWChar(outBuff, outLen + nulLen, inBuff);
     if ( rc != wxCONV_FAILED )
     {
-        rc -= nulLen;
+        rc -= GetMBNulLen();
     }
 
     return rc;

src/mac/carbon/textctrl.cpp

@@ -2252,8 +2252,10 @@ void wxMacMLTEControl::SetTXNData( const wxString& st, TXNOffset start, TXNOffse
 #else
     wxMBConvUTF16 converter ;
     ByteCount byteBufferLen = converter.WC2MB( NULL, st.wc_str(), 0 ) ;
-    UniChar *unibuf = (UniChar*)malloc( byteBufferLen ) ;
+    wxASSERT_MSG( byteBufferLen != wxCONV_FAILED,
-    converter.WC2MB( (char*)unibuf, st.wc_str(), byteBufferLen ) ;
+                  _T("Conversion to UTF-16 unexpectedly failed") );
+    UniChar *unibuf = (UniChar*)malloc( byteBufferLen + 2 ) ; // 2 for NUL in UTF-16
+    converter.WC2MB( (char*)unibuf, st.wc_str(), byteBufferLen + 2 ) ;
     TXNSetData( m_txn, kTXNUnicodeTextData, (void*)unibuf, byteBufferLen, start, end ) ;
     free( unibuf ) ;
 #endif

src/msw/textctrl.cpp

@@ -947,14 +947,14 @@ wxTextCtrl::StreamIn(const wxString& value,
         return false;
 
 #if wxUSE_WCHAR_T
-    wxWCharBuffer wchBuf(len);
+    wxWCharBuffer wchBuf(len); // allocates one extra character
     wchar_t *wpc = wchBuf.data();
 #else
     wchar_t *wchBuf = (wchar_t *)malloc((len + 1)*sizeof(wchar_t));
     wchar_t *wpc = wchBuf;
 #endif
 
-    conv.MB2WC(wpc, value, value.length());
+    conv.MB2WC(wpc, value, len + 1);
 #endif // wxUSE_UNICODE_MSLU
 
     // finally, stream it in the control

tests/mbconv/mbconvtest.cpp

@@ -807,41 +807,48 @@ void MBConvTestCase::BufSize()
     CPPUNIT_ASSERT( conv1251.IsOk() );
     const char *cp1251text =
         "\313\301\326\305\324\323\321 \325\304\301\336\316\331\315";
-    const size_t cp1251textLen = strlen(cp1251text);
 
-    const size_t lenW = conv1251.MB2WC(NULL, cp1251text, cp1251textLen);
+    const size_t lenW = conv1251.MB2WC(NULL, cp1251text, 0);
-    CPPUNIT_ASSERT( lenW != wxCONV_FAILED );
+    CPPUNIT_ASSERT_EQUAL( strlen(cp1251text), lenW );
-    wxWCharBuffer wbuf(lenW);
+    wxWCharBuffer wbuf(lenW + 1); // allocates lenW + 2 characters
+    wbuf.data()[lenW + 1] = L'!';
 
-    // lenW-1 is not enough
+    // 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 - 1) );
+        wxCONV_FAILED, conv1251.MB2WC(wbuf.data(), cp1251text, lenW) );
 
-    // lenW is just fine
+    // lenW+1 is just fine
-    CPPUNIT_ASSERT(
-        conv1251.MB2WC(wbuf.data(), cp1251text, lenW) != wxCONV_FAILED );
-
-    // of course, greater values work too
     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(_T("UTF-16"));
+    wxCSConv convUTF16(_T("UTF-16LE"));
     CPPUNIT_ASSERT( convUTF16.IsOk() );
     const wchar_t *utf16text = L"Hello";
-    const size_t utf16textLen = wcslen(utf16text);
 
-    const size_t lenMB = convUTF16.WC2MB(NULL, utf16text, utf16textLen);
+    const size_t lenMB = convUTF16.WC2MB(NULL, utf16text, 0);
-    CPPUNIT_ASSERT( lenMB != wxCONV_FAILED );
+    CPPUNIT_ASSERT_EQUAL( wcslen(utf16text)*2, lenMB );
-    wxCharBuffer buf(lenMB + 1); // it only adds 1 for NUL on its own, we need 2
+    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 - 1) );
+        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) != wxCONV_FAILED );
+        convUTF16.WC2MB(buf.data(), utf16text, lenMB + 2) != wxCONV_FAILED );
     CPPUNIT_ASSERT(
-        convUTF16.WC2MB(buf.data(), utf16text, lenMB + 1) != wxCONV_FAILED );
+        convUTF16.WC2MB(buf.data(), utf16text, lenMB + 3) != wxCONV_FAILED );
+    CPPUNIT_ASSERT_EQUAL( '?', buf[lenMB + 2] );
 }
 
 WXDLLIMPEXP_BASE wxMBConv* new_wxMBConv_iconv( const wxChar* name );