Hi,
On Tue, Oct 29, 2024 at 01:12:25PM +0900, Michael Paquier wrote:
> On Mon, Oct 28, 2024 at 07:17:28AM +0000, Bertrand Drouvot wrote:
> > I think that we can not be 100% sure that the s1 wait will finish before the
> > s2 detach (easy reproducible with gdb attached on s1 or an hardcoded sleep) and
> > that other OS could also report the test as failing for the same reason.
>
> Yes, the only safe thing we can do in this test is to let the wakeup2
> be last, as we are sure that the isolationtester is going to keep at
> s2 to finish the call of the wakeup function before moving on with
> checking the end of the wait.
Agree.
> > It's not ideal, but instead of removing this first permutation test what about
> > adding a "sleep2" step in it (doing say, SELECT pg_sleep(1);) and call this
> > new step before the detach2 one?
>
> There is no real guarantee of stability.
Right.
> Under a wait of N seconds,
> we could still have environments where the wait() could remain stuck
> more than N seconds between the moment the condition variable is woken
> up and the result of the wait() is reported back to the client. And
> hardcoded sleeps make the test slower even on fast machines.
Yeah, not ideal so better to keep the test as it is currently (after your
refactoring).
> What we have here seems like just contention of Cirrus with the
> FreeBSD hosts while there is much more stability with the linux hosts.
That sounds like it, agree.
Regards,
--
Bertrand Drouvot
PostgreSQL Contributors Team
RDS Open Source Databases
Amazon Web Services: https://aws.amazon.com
