Re: WIP: new system catalog pg_wait_event

On Sat, Aug 19, 2023 at 11:35:01AM +0200, Drouvot, Bertrand wrote:
> Hi,
>
> On 8/18/23 12:31 PM, Michael Paquier wrote:
> Thanks! Please find attached v9 fixing this typo.

I was looking at v8, and this looks pretty good to me.  I have
spotted a few minor things.

+  proretset => 't', provolatile => 's', prorettype => 'record',
This function should be volatile.  For example, a backend could add a
new wait event while a different backend queries twice this view in
the same transaction, resulting in a different set of rows sent back.

By the way, shouldn't the results of GetWaitEventExtensionNames() in
the SQL function be freed?  I mean that:
    for (int idx = 0; idx < nbextwaitevents; idx++)
        pfree(waiteventnames[idx]);
    pfree(waiteventnames);

+   /* Handling extension custom wait events */
Nit warning: s/Handling/Handle/, or "Handling of".
--
Michael