Re: failures on barnacle (CLOBBER_CACHE_RECURSIVELY) because of memory leaks

On 24.8.2014 18:01, Tom Lane wrote:
> Tomas Vondra <tv(at)fuzzy(dot)cz> writes:
>>> I stopped the already running test on addax and started the test on
>>> barnacle again. Let's see in a few days/weeks/months what is the result.
>
>> It seems to be running much faster (probably after removing the
>> randomization), and apparently it passed the create_view tests without
>> crashing, but then crashed at 'constraints' (again, because of OOM).
>
> I poked into the "constraints" test and soon found another leak just like
> the other one :-(, which I've now fixed. I lack the patience to let
> "constraints" run to conclusion with C_C_R on --- I let it run for about
> 20 hours and it still was only maybe 1/4th done --- but it got further
> than it did in your report and there was zero sign of leakage.

OK. There was some outage in the facility hosting this machine, and it
went up just a few hours ago. So I killed the current run, and it'll
start chewing on this new commit in a few minutes.

Regarding those leaks we've detected so far - is it the kind of leaks
that can happen only in testing with those specific flags, or is it
something that can happen in production too? (Assuming no one is running
with CLOBBER_CACHE_RECURSIVELY in production, of course ;-) That is,
does it seem worth the effort running those tests / fixing those leaks?

regards
Tomas