6.3 spurious transaction aborted problem

Hello,

I have a product which uses Postgres 6.3.  I believe it really is 6.3 and
not 6.3.1 or 6.3.2.  I know this version is over a year old, and I've
tried to convince the other developers to upgrade, but they hadn't any
problems at all with the database and didn't want to introduce any new
variables.

A problem I've seen recently are "spurious" transaction aborted errors.
The exact message is: transaction aborted -- queries ignored until END.
I don't why they are happening, and I haven't yet been able to track down
what part of the code is emitting them.  They seem to happen randomly
and not all that frequently, so they are hard to reproduce. Is it possible
that multiple processes accessing the database could trigger this message?

Another quick question while I have your attention: several places in our
client code there are commits being done "early" with comments like
"clear read locks".  It seems, though, that you can never get blocked by a
transaction you started yourself, so I'm trying to figure out why the
previous set of developers felt it necessary to perform all of these
early commits.  For the most part our Postgres instance only has one
client, so the need to clear read locks is puzzling.

Any help with these issues would be greatly appreciated!

Erik Rantapaa
erantapa@securecomputing.com