Re: Large objects and savepoints - Snapshot reference leak

Andreas Lubensky wrote:
> Hi,
>
> I'm trying to read/write large objects via libpq. I encapsulated the
> operations in a transaction but I wanted to put a savepoint before doing
> any operations, so I can do a rollback in case anything fails without
> breaking the current transaction. Now, when sth. actually fails and the
> transaction is rolled back to the savepoint, the next commit results in
> a warning:
> Snapshot reference leak: Snapshot 0xb5e4b0 still referenced
> I'm not sure what to make of that. Can it be ignored? Is rolling back
> large object operations not possible?

What vresion are you running? I wonder if this can be attributed to a
bug fixed by this commit:

Author: Heikki Linnakangas <heikki(dot)linnakangas(at)iki(dot)fi>
Branch: master [357f75213] 2013-09-30 12:53:14 +0300
Branch: REL9_3_STABLE Release: REL9_3_1 [f609d0743] 2013-09-30 12:53:56 +0300
Branch: REL9_2_STABLE Release: REL9_2_5 [fc7a38f32] 2013-09-30 12:54:37 +0300
Branch: REL9_1_STABLE Release: REL9_1_10 [cd6c03b5c] 2013-09-30 12:55:57 +0300
Branch: REL9_0_STABLE Release: REL9_0_14 [c5c87f065] 2013-09-30 12:58:51 +0300
Branch: REL8_4_STABLE Release: REL8_4_18 [fef01d419] 2013-09-30 13:00:00 +0300

Fix snapshot leak if lo_open called on non-existent object.

lo_open registers the currently active snapshot, and checks if the
large object exists after that. Normally, snapshots registered by lo_open
are unregistered at end of transaction when the lo descriptor is closed, but
if we error out before the lo descriptor is added to the list of open
descriptors, it is leaked. Fix by moving the snapshot registration to after
checking if the large object exists.

Reported by Pavel Stehule. Backpatch to 8.4. The snapshot registration
system was introduced in 8.4, so prior versions are not affected (and not
supported, anyway).

--
Álvaro Herrera http://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Training & Services