Re: pgsql: Basic partition-wise join functionality.

On Fri, Oct 6, 2017 at 12:47 PM, Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> wrote:
> Robert Haas <rhaas(at)postgresql(dot)org> writes:
>> Basic partition-wise join functionality.
>
> Buildfarm member prion seems quite unhappy with this.
>
> I'm not sure if that's a case of the plan choices being less
> platform-independent than you thought, or if it's triggered by the
> -DRELCACHE_FORCE_RELEASE option, in which case there's an actual bug
> here and not just poorly chosen test cases.

It's the bug one. Ashutosh pinged me off-list and he's working on it.
Somehow I thought it was OK for the planner to have pointers into the
relcache, but apparently we don't keep the relations open throughout
planning, only locked. So set_relation_partition_info and
find_partition_scheme needs to grow code to copy stuff instead of just
pointing to it.

--
Robert Haas
EnterpriseDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company