Re: GNU/Hurd portability patches

Hi,

On Wed, Sep 24, 2025 at 11:00:00PM +0300, Alexander Lakhin wrote:
> 24.09.2025 18:52, Michael Banck wrote:
> > I did that in the attached, so far my Hurd VM ran the stats test more
> > than 1000 times without a failure with it. I have the loop running till
> > 10000, I'll report back tomorrow.

(it ran for 7000 iterations without fault so far)

> If the stats test could be fixed this way, I wonder how to deal with
> regress/subscription.sql. When running:
> TESTS="$(printf "subscription %.0s" `seq 1000`)" make -s check-tests
>
> on the same Hurd VM, I'm observing:
> ...
> ok 986       - subscription                               53 ms
> not ok 987   - subscription                               53 ms
> ok 988       - subscription                               53 ms
> ...
> # 4 of 1000 tests failed.

I ran that five times now without a problem, both with and without the
Mach patch I mentioned earlier, and on 32 and 64 bit. Not sure what is
going on here.

> # The differences that caused some tests to fail can be viewed in the file
> "/home/demo/postgresql/src/test/regress/regression.diffs".
>
> $ cat "/home/demo/postgresql/src/test/regress/regression.diffs"
>
> --- /home/demo/postgresql/src/test/regress/expected/subscription.out 2025-09-24 19:49:53.000000000 +0100
> +++ /home/demo/postgresql/src/test/regress/results/subscription.out 2025-09-24 20:06:48.000000000 +0100
> @@ -70,7 +70,7 @@
>  SELECT :'prev_stats_reset' < stats_reset FROM pg_stat_subscription_stats WHERE subname = 'regress_testsub';
>   ?column?
>  ----------
> - t
> + f
>  (1 row)

I saw those issues frequently on the initial 32bit Hurd VM I started to
run the buildfarm code on, before I switched it to HPET timers. Since
then, I don't think I saw that particular error again, but 4 out 1000 is
not a lot of course.

Michael