On Mon, 22 Nov 2004, Paul Beach wrote:

> 1. An error like that would cause a crash of the process that was
> connected to the database - did that happen?

The client process (python program) did crash, because I didn't write
any error handling code. As for the server process I don't see any
way to find that out. It is the classic version of the Firebird
server which means one database per connection - when the connection
dies, the server process quits, too, AFAIK. If the server process
returned an error message and logged the same error in firebird.log,
then I'd assume it didn't crash. Or it crashed gracefully. I don't
know.

> 2. I think your problem (in Firebird 1.5.1) is related to the following bug that
> is fixed in Firebird 1.5.2 RC3
> The server could crash when execute_immediate was used to release or rollback
> a transaction to a non-existing savepoint. Not logged
> Solution Now fixed. N. Samofatov

I don't think it is related, because I doubt the server crashed,
and I didn't rollback any transaction.

> Can you download the following 1.5.2 RC3 build and retry the test?

I'll retry with the proposed build, but it will take several days
to fill up the table. It took almost 4 days before the error was
triggered.

-Ales