Can anybody tell me if I am doing something wrong here, or, can Microsoft
verify if this is a bug in SQL Server?

If I prepare a simple insert, for example, "INSERT INTO MYTABLE(MYCOL)
VALUES (?)", and then, bind an SQL_LONGVARCHAR column, for example:

SQLBindParameter(hstmt, 1, 1, 1, -1, 2147483647, 0, data, 0, &indicator);

then, if I set the ODBC environment attribute SQL_ATTR_ODBC_VERSION to
SQL_OV_ODBC2, I can choose between SQL_LEN_DATA_AT_EXEC(length) (and use
SQLPutData), 0, and SQL_NULL_DATA for the indicator, each time I call
SQLExecute (after SQLPrepare).

If I set SQL_OV_ODBC3, mixing these indicators causes "Invalid Cursor
State", when using SQLPrepare + SQLExecute (vs. SQLExecDirect).

1. Prepare the statement and bind.

2. Set the indicator to SQL_LEN_DATA_AT_EXEC(some length > 0) and use
SQLExecute, SQLParamData, SQLPutData, SQLParamData to send the bytes.

3. Now try to set the indicator to 0 (or, SQL_NULL_DATA). If that succeeds,
try sending another row with length > 0. With SQL Server 2000, 8.00.2039, I
get "Invalid Cursor State".

If I set SQL_LEN_DATA_AT_EXEC(0) I can then SQLPutData 0 bytes, but if I
want to send SQL_NULL_DATA, I have to either: (1) set
SQL_LEN_DATA_AT_EXEC(0), and SQLPutData SQL_NULL_DATA, which IMHO does not
follow the ODBC spec, or (2) use SQLExecDirect, or (3) use SQL_OV_ODBCV2.

Regards,
Jim Flood