Hi,
Back in November 2001 (Doesn't time fly . . .) I had a problem with trying
to delete a persona after accessing Rdb. In a nutshell, If you'd started a
read-only transaction for a persona then you were unable to delete it and
were returned ss$_personadelpend even though you'd committed the txn and
gone back to your natural persona. Scary thing, from a VMS point of view,
was you were able to assume a dead persona. (Though I was more interested
in
the potential memory-leak/resource exhaustion of endless personae being
created)
At the time I had no idea if it was a VMS or Rdb problem, I was running
Alpha VMS 7.2 and Rdb 7.0.3 and no one (least of all Rdb engineering) was
interested in discussing the issue. Anyway, I was testing some new
features
the other day and I decided to revisit this functionality and guess what?
It's fixed! I'm now running Rdb 7.1-24 and if I rdb$setver back to 7.0 I
can
reproduce the problem. So it is clearly something Rdb was doing that
prevented the persona from dying. But what? I scanned through the Rdb
release notes and could not find anything. Surely this is a lot more
interesting then most of the crap that gets beaten to death in COV?
Most of you may not find this exciting, but for me, this is the dog's
bollocks! If only Rdb sup****ted Set Session Authorization Persona
:ws_integer; :-(
Norm, you've been very quite lately; any interest in this? What about you
VMS guys? What could Rdb have been doing that stopped the persona from
dying
immediately? I checked SDA> Show Proc/Persona but it had already gone
after
the reserve. I was hoping that RefCount might reveal something.
Anyway, I'll attach some test code as a reply along with the original
mail.
This really is pretty ****-hot stuff! Maybe if someone else asks for it we
may get a reply?
Regards Richard Maher
PS. On second thoughts let's resume normal programming - "This persona
stuff
could reveal a deep-rooted underlying lack of commitment to VMS!", "I
smell
smoke; is that a vibration?", "We're all going to die!!!"
|
