> I did not find the getBlob(int) function but one called
> getBlob(boolean)

ups... yes, that's the function i meant.

> So I made the modifications, and compiled using ant. I tested out the
> new driver, and I got the same results.
>
> To me that function looks strange since it can be passed a value, but
> the value is not used in the function. So why pass anything?

i noticed the same. it seems to me that i had some ideas, but did not
implement them leaving the code somewhere in the middle... i'm really
sad about that and will fix it as soon as i get some free time.

> Anyhow I am pretty sure I made the modification suggested in the
> right place and do not know where to go or look from here.

exception you get is thrown in org.firebirdsql.jdbc.FBBlob class. you
can grep for that particular message. main idea was to prevent read
access to a newly created blob that has id 0 (nothing was created in
the database). but from what you write i conclude that id=0 can happen
even in case of correct code. i'm not sure if id=0 means that blob is
null or not, this must be found out experimentally.

> So if you or anyone else has any ideas I will swap out the code,
> compile, and test. As I am barely learning the C API, I do not think
> I can be much help other than as a tester. But I will do my part to
> test if needed.

it would be really helpful if you could create a test case that
reproduces this bug. test case should be able to reproduce the bug on
empty database. in other words it must create table, populate data and
actually reproduce data. this will allow us to concentrate on fixing
the bug and not on trying to reproduce it.

thanks!
roman