Dear Helen,

Yeah. I'm using some sort of MIDAS-like middle tier. But I recreated a
test program and reproduced it. It shouldn't cause any problems since
it's pretty straight forward. I tried all 3 versions of firebird --
1.5.2, 1.5.3 and 2.0. They all exhibit the same kind of failures.

I'm guessing that firebird's dll (be it fbclient.dll or gds32.dll, of
which are the same since I manually replace gds32.dll with
fbclient.dll just to be sure) is causing some memory overrun that
causes IBO to fail. In fact, of all the tests that I've ran (same
test, just over and over again), gds32.dll ended up with a READ Access
Violation in quite a few of them.

This is kind of the same multithreading environment as with an ISAPI --
has anyone else reported this problem? Since I believe there should
be a lot of users using firebird + IBO in the ISAPI context, more so
than in MIDAS multithread mode.

I'm really scratching my head here. There must be something in
fbclient.dll that is not thread-safe. Enabling Hyperthreading will
cause strange failures as well, even though I'm only running one
thread (i.e. CommitRetaining clears the posted record from the dataset
occasionally while in non-HT mode, it retains them faithfully).

Seeing you're both the key person in firebird + IBO development, I'd
reckon you're the best person to have a look at this. Really
appreciate your time and patience. Sorry for throwing a bit of tantrum
in my earlier posts though. I was getting nowhere with our software
and deadlines are looming closer. :(

Thank you again.

--- In IBObjects@yahoogroups.com, Helen Borrie <helebor@t...> wrote:
>
> Regarding this and your previous posting, it looks as though you are
still
> trying to operate multiple local connection threads. Is this the
case?