On Mon, Jul 31, 2023 at 12:07:40PM +0530, Bharath Rupireddy wrote:
> We're not giving up and returning an unknown state in the v10 patch
> unlike v9, no? This comment needs to change.
Right. Better to be consistent with lwlock.c here.
>> No, it cannot because we need the custom wait event string to be
>> loaded in the same connection as the one querying pg_stat_activity.
>> A different thing that can be done here is to use background_psql()
>> with a query_until(), though I am not sure that this is worth doing
>> here.
>
> -1 to go to the background_psql and query_until route. However,
> enhancing the comment might help "we need the custom wait event string
> to be loaded in the same connection as .....". Having said this, I
> don't quite understand the point of having worker_spi_init() when we
> say clearly how to use shmem hooks and custom wait events. If its
> intention is to show loading of shared memory to a backend via a
> function, do we really need it in worker_spi? I don't mind removing it
> if it's not adding any significant value.
It seems to initialize the state of the worker_spi, so attaching a
function to this stuff makes sense to me, just for the sake of testing
all that.
--
Michael