On Fri, Oct 27, 2023 at 1:48 PM Alvaro Herrera <alvherre@alvh.no-ip.org> wrote:
> On 2023-Oct-27, Alexander Korotkov wrote:
>
> > Adjust parallel_schedule with event triggers on authenticated login
> >
> > Event triggers on authenticated login could catch the connection of the
> > concurrent test. In order to prevent this side effect we need to run
> > the event_trigger test alone.
>
> Hmm, not a fan, because the original lines of the event_trigger test,
> which had no problem running in parallel, no longer do so. Why not
> split out only the new part of the test to run in non-parallel mode?
> An effortless win of ~50 ms.
>
> After this patch:
>
> ok 211 + stats 644 ms
> # parallel group (2 tests): event_trigger oidjoins
> ok 212 + oidjoins 103 ms
> ok 213 + event_trigger 57 ms
> ok 214 - event_trigger_login 16 ms
> ok 215 - fast_default 66 ms
> ok 216 - tablespace 156 ms
> 1..216
> # All 216 tests passed.
>
>
> Before:
>
> ok 211 + stats 603 ms
> ok 212 - event_trigger 66 ms
> ok 213 - oidjoins 113 ms
> ok 214 - fast_default 71 ms
> ok 215 - tablespace 161 ms
> 1..215
> # All 215 tests passed.
OK, I'm good with this.
Do you like me to push this? Or do you like to push it yourself?
------
Regards,
Alexander Korotkov