Re: pg_usleep for multisecond delays

On Thu, Feb 09, 2023 at 01:30:27PM -0800, Andres Freund wrote:
> On 2023-02-09 12:59:29 -0800, Nathan Bossart wrote:
>> I just found myself carefully counting the zeros in a call to pg_usleep().
>> Besides getting my eyes checked, perhaps there should be a wrapper called
>> pg_ssleep() than can be used for multisecond sleeps. Or maybe the
>> USECS_PER_SEC macro should be used more widely. I attached a patch for the
>> former approach. I don't have a strong opinion, but I do think it's worth
>> improving readability a bit here.
>
> pg_usleep() should pretty much never used for sleeps that long, at least in
> the backend - depending on the platform they're not interruptible. Most of the
> things changed here are debugging tools, but even so, it's e.g. pretty
> annoying. E.g. you can't normally shut down while a backend is in
> pre_auth_delay.
>
> So I'm not sure it's the right direction to make pg_usleep() easier to use...

Hm... We might be able to use WaitLatch() for some of these.

--
Nathan Bossart
Amazon Web Services: https://aws.amazon.com