Re: 8.3.0 backend segfaults

On Tue, Mar 11, 2008 at 10:59 PM, Tom Lane <tgl@sss.pgh.pa.us> wrote:
> "Alex Hunsaker" <badalex@gmail.com> writes:
>  > Problem: Apparently random segfaults apparently query agnostic, seem
>  >   to be more frequent when a pg_dump is running
>
>  Hmm, seems from the backtrace that we're trying to do a replan with an
>  invalid ActiveSnapshot.  What sequence of operations is the connected
>  client performing, exactly?  (I'm not worried about the details of the
>  SQL per se, but of how it's preparing/issuing queries.)
>
>                         regards, tom lane
>

Its through perl-dbi using ->prepare, ->execute.. is that what you
mean?  Also this is with mod_perl and Apache::DBI.

So a typical transacion looks like

connect to postgres;
begin;
do stuff here;
commit or rollback;
(next page load)
begin;
....

so potentially a database connection/session is reused on the next
page load if that makes a difference.