Re: pgsql: Add some regression tests that exercise hash join code.

On Wed, Dec 6, 2017 at 8:03 AM, Andres Freund <andres@anarazel.de> wrote:
> On 2017-12-05 12:59:21 +1300, Thomas Munro wrote:
>>  src/backend/commands/explain.c      |  58 ++++++++++++++------
>
> This didn't actually include nodeHash.h, ...

Ugh, I'd lost -Werror from my COPT.  Back in there it goes.

> Pushed, thanks for the patch. Let's see what the buildfarm says...

Thank you.  I'm keeping an eye on it.

By the way, the current coding of ExecSortRetrieveInstrumentation()
and now also ExecHashRetrieveInstrumentation() is wrong in the case of
rescans, since they get called multiple times but are written in a way
that assumes they'll be called only once:

https://www.postgresql.org/message-id/CAA4eK1JBj4YCEQKeTua5%3DBMXy7zW7zNOvoXomzBP%3Dkb_aqMF7w%40mail.gmail.com

Amit's patch would move ExecParallelRetrieveInstrumentation() so that
they are called only once, so that would fix that problem without any
change to this code.  More over on that thread.

-- 
Thomas Munro
http://www.enterprisedb.com