Michael Paquier <michael@paquier.xyz> writes:
> On Tue, Jul 09, 2019 at 02:30:51PM +1200, Thomas Munro wrote:
>> Yeah. I had obviously never noticed that test script. +1 for just
>> enabling hash joins the top of join_hash.sql in 12+, and the
>> equivalent section in 11's join.sql (which is luckily at the end of
>> the file).
> Right, I did not pay much attention to REL_11_STABLE. In this case
> the test begins around line 2030 and reaches the bottom of the file.
> I would actually add a RESET at the bottom of it to avoid any tests to
> be impacted, as usually bug-fix tests are just appended.
Agreed that the scope should be limited. But in 12/HEAD, I think the
relevant tests are all wrapped into one transaction block, so that
using SET LOCAL would be enough. Not sure if 11 is the same.
regards, tom lane