When grilled further on (Thu, 22 Jan 2004 14:27:19 -0500),
Tom Lane <tgl@sss.pgh.pa.us> confessed:
> "PostgreSQL Bugs List" <pgsql-bugs@postgresql.org> writes:
> > Description: message type 0x49 arrived from server while idle
>
> Hm. This would seem to have something to do with issuing empty queries
> (message type 0x49 = 'I' is EmptyQueryResponse). I couldn't duplicate the
> problem in some simple tests, however. Unless you can come up with more
> information, there's not a lot I can do ...
What more information would you be looking for? I'm thinking that maybe the
query that was executed (or not executed)? I looked at the log and there was no
info other than "unexpected EOF on client connection", which I get regularly
from my multi-process program. I'll turn on logging of all queries to see if
any happen to be empty. I don't believe I'm executing empty queries, and the
problem is not reproducible by running the same program multiple times. It just
shows up every now and then.
I was not getting this error while running 7.4Beta4, but that was running on
Solaris 2.8 and compiled with GCC 3.1. The same 7.3.3 PGSQL would of been used
in this instance.
>
> > The PGSQL version on Solaris is 7.3.3, but that shouldn't matter?
>
> No, that's *extremely* relevant, because it means the wire protocol is
> going to be 2.0 not 3.0. I couldn't get anything funny to happen by
> issuing empty queries with a 7.3 libpq and a 7.4 server, though.
I'll look into doing an install of the client side only (7.4.1) that I can
access when running my scripts, and re-install DBD::Pg against that version.
Probably won't happen 'till next week at the earliest though.
Cheers,
Rob
--
19:48:21 up 25 days, 9:35, 4 users, load average: 2.49, 2.43, 2.18