At 09:53 PM 27/10/2004 -0400, you wrote:

>:: >Even if so shouldn't FB just err and not crash?
>::
>:: That error, if caught and handled, won't cause the server to
>:: crash. Because it is logged, you know that it has been
>:: caught at least once.
>
>Noooo - the error that the application gets is "Socket conenction aborted
>abruptly" . FB just CRASHES. There is no chance to handle it - in fact the
>whole server goes down - not just that DB.

There's some kind of misalignment of interpretation here. The firebird.log
is written to by the server, not "the database".

The server (== fbserver.exe, not the hardware) has caught at least one
instance of the TOO MANY SAVEPOINTS condition and had enough resources to
write that out to the log.

Perhaps or perhaps not it had enough resources to fill the error status
array and return the exception to the offending client. If the application
is swallowing exceptions, or the server never got far enough to deliver the
array back to the client, then either way the application won't see it.

By "the server never got far enough" I mean that Windows shut down the
service. A shut-down service has a certain level of *difficulty* in doing
what it oughta.

./h