Deadlocking: "xid loop detected"

Hello,

I have a deadlocking problem with PostgreSQL v6.3.2. The offending SQL
statement seems to be an UPDATE; when I try to SELECT from the same
table things block. After killing the postmaster I get the following
messages:

"Some other backend died abnormally and possibly corrupted shared
memory" (several times, probably proportional to the number of
postgres processes running)

NOTICE: LockReleaseAll: xid loop detected, giving up

and messages about shared memory and so on.

The table only uses standard TEXT, VARCHAR, INTEGER and TIMESTAMP
types and obtains default values for some attributes from a sequence.
The system is Linux kernel 2.0.34, libc5, with Apache + mod_perl +
Apache::DBI + DBD::Pg. I am running DBI in autocommit mode.

* is there any way to ask the backend what queries are running and
what locks are active on a given table?

* any advice as to detecting deadlock situations and recovering
gracefully in web/db applications?

--
Eric Marsden
emarsden @ mail.dotcom.fr
It's elephants all the way down