On Tue, Aug 15, 2017 at 5:15 PM, Tom Lane <tgl@sss.pgh.pa.us> wrote:
> The idea of the existing
> code seems to be "let's exercise what happens if the executor does
> EnterParallelMode/ExitParallelMode around any plan whatsoever, even a
> parallel-unsafe one"; which seems to me to be bogus as heck.
I think that is not what the current code is doing. If the plan is
diagnosed as parallel-unsafe, then parallelModeOK will be false and
nothing will happen. If the plan is actually parallel-unsafe but the
planner doesn't *think* it's parallel-unsafe, then what you are
talking about will happen, but that seems to me to be a good thing.
It lets you find planner bugs (or functions that a user has labelled
incorrectly).
--
Robert Haas
EnterpriseDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company