Nando,

At 09:48 PM 31/05/2004 +0200, you wrote:

>Helen,
>
>H> It's very late here (early? 2:40 a.m.) but tomorrow I'll prepare a demo for
>H> Jason to look at. I'll also check behaviour if I use IB_Query instead of
>H> IB_Cursor.
>
>sorry to jump in at this late point in the thread, but if the server
>crashes then the test case will be of use for the Fb developers as well
>as Jason, I believe.

I completely agree. But first, it will be most useful to find out why
TIBOQuery causes this behaviour, while TIB_Cursor doesn't. In the middle
is TIB_Query, which I'll test today.

A test case demonstrating all three will help Jason to isolate exactly what
it is about the client layer that is capable of crashing the server, i.e.
what API function is broken. It seems to relate to the default transaction
(gds__trans) that one could guess might have been broken on the client side
of the y-valve...I'm not familiar enough with the <default transaction>
mechanism to know whether the TDataset architecture handles it differently
to native IBO.

Helen