Re: memory leak regression 9.1 versus 8.1

On 05/09/2012 10:01 PM, Tom Lane wrote:
> Joe Conway <mail(at)joeconway(dot)com> writes:
>> The attached one-liner seems to plug up the majority (although not quite
>> all) of the leakage.
>
> Looks sane to me. Are you planning to look for the remaining leakage?

Actually, now I'm not so sure there really are any other leaks. The last
test I ran, on 9.1 with the original data and plpgsql function, grew to:

VIRT RES SHR
540m 327m 267m

but then stabilized there through the end of the query, which
successfully returned:

count
----------
28847766
(1 row)

This was with:

report_log=# show shared_buffers;
shared_buffers
----------------
256MB
(1 row)

report_log=# show work_mem;
work_mem
----------
16MB
(1 row)

So I think those memory usage numbers look reasonable.

The bug appears to go back through 8.4 -- kind of surprising no one has
complained before.

Joe

--
Joe Conway
credativ LLC: http://www.credativ.us
Linux, PostgreSQL, and general Open Source
Training, Service, Consulting, & 24x7 Support