Hi,
I am having trouble with stuffing in an getting out unicode strings
into/from my database. I will briefly describe what I am doing, hoping that
someone can point me to a mistake and end my weeks of trying...

* Oracle 9i client and server, connecting via OCI from a C++ (MS Dev Studio
2005) application
* Client OS: German Windows XP, SP2
* Server: Oracle9.2.0.1 Enterprise on Windows XP SP2, NLS_LANGUAGE=AMERICAN
* Server character set (determined via "select value from
nls_database_parameters where parameter='NLS_CHARACTERSET';"): WE8MSWIN1252
* Client character set GERMAN_GERMANY.WE8MSWIN1252 (determined via "sqlplus
/nolog" and @.[%NLS_LANG%])

What I am doing in my program to stuff unicode in is this:

std::wstring strUnicode(L"Âb穾ü??déñf??");
....
wchar_t* pValue = (pWS_ == NULL) ? NULL : _wcsdup( strUnicode.c_str() );
size_t nSize = (pValue == NULL) ? 0 : (sizeof(wchar_t) * (1 +
wcslen(pValue)));
....
swdReturnCode = OCIBindByPos (
(OCIStmt*) m_pOCIStatement,
(OCIBind**) &pBindHandle,
(OCIError*) m_pConnection->_getOCIError(),
(ub4) columnIndex_,
(dvoid*) pValue,
(sb4) nSize,
(ub2) SQLT_STR,
(dvoid*) NULL,
(ub2*) NULL,
(ub2*) NULL,
(ub4) NULL,
(ub4*) NULL,
(ub4) OCI_DEFAULT
);
ub2 csid = OCI_UTF16ID;
swdReturnCode = OCIAttrSet(
(void *) pBindHandle
, (ub4) OCI_HTYPE_BIND
, (void *) &csid
, (ub4) 0
, (ub4) OCI_ATTR_CHARSET_ID
, m_pConnection->_getOCIError()
);
....

The string that I am stuffing in corresponds to the following sequence of
integers in my program's main memory:

194,98,231,169,190,252,1046,950,100,233,241,102,17 15,1492

When I select I get the following back (displayed as "Âb穾ü¿¿déñf¿¿"):

194,98,231,169,190,252,191,191,100,233,241,102,191 ,191

So it turns out that some characters have been transformed to be
upsode-down question marks. I see the same thing when I dont fetch with my
application but with the Oracle Enterprose Manager.

What is wrong here? Am I missing out some important conversion on the way?
Thanks in advance,
André