Re: BUG #8468: Create index on type tstzrange fail - Mailing list pgsql-bugs

From Heikki Linnakangas
Subject Re: BUG #8468: Create index on type tstzrange fail
Date
Msg-id 5242E129.4090903@vmware.com
Whole thread Raw
In response to BUG #8468: Create index on type tstzrange fail  (marian.krucina@gmail.com)
List pgsql-bugs
On 24.09.2013 14:42, marian.krucina@gmail.com wrote:
> CREATE INDEX ON tstzrange fail on 9.3.0 and 9.2.4 - default postgres
> configuration.
> It ate whole memory and was killed by oom.
>
>
> Example:
> postgres=# CREATE TABLE range_test AS SELECT tstzrange(t, t+'1min') tr FROM
> generate_series('2000-1-1'::TIMESTAMPTZ, '2010-1-1'::TIMESTAMPTZ, '1min') AS
> t1(t);
> SELECT 5260321
> postgres=# CREATE INDEX ON range_test(tr);
> WARNING:  terminating connection because of crash of another server process
> DETAIL:  The postmaster has commanded this server process to roll back the
> current transaction and exit, because another server process exited
> abnormally and possibly corrupted shared memory.
> HINT:  In a moment you should be able to reconnect to the database and
> repeat your command.
> The connection to the server was lost. Attempting reset: Failed.

A-ha, the comparison function of range datatypes, range_cmp(), detoasts
its arguments, but fails to free the detoasted copies. Functions are
normally not required to free such temporary copies - the memory is
usually leaked into a short-lived memory context that will be quickly
free'd anyway - but B-tree comparison operators are expected to not leak.

Committed a fix, it will appear in the next minor releases. Thanks for
the report!

- Heikki

pgsql-bugs by date:

Previous
From: Sandeep Thakkar
Date:
Subject: Re: Errors during PostgreSQL 9.2 and 9.3 installation
Next
From: os@ohmu.fi
Date:
Subject: BUG #8470: 9.3 locking/subtransaction performance regression