Now, a curious thing happens. When inserting new data through a Recordset, everything works fine - except for columns defined as bigint. There are no exceptions thrown, but when you read the columns back they contain nothing but zeroes. Do the same to any other column type - I've tried integer, numeric, float, nvarchar and ntext so far, and they all seem to work just fine. It does not seem to be conversion-related either, since I've tested the exact same data to various column types. And using bigint on regular SQL Server works just fine.
The code involved is quite unspectacular, and simply switching the column types to integer would solve the immediate problem, but causes potential future issues since we normally store internal IDs in bigint columns, and the values may grow quite large.
This may indicate a problem with compatibility betweeen the 3rd party python ADO adaptor and the SQL CE OLEDB provider.
