Re: pg_stat_statements: more test coverage

On 31.12.23 10:26, Julien Rouhaud wrote:
> On Sun, Dec 31, 2023 at 2:28 PM Michael Paquier <michael(at)paquier(dot)xyz> wrote:
>>
>> On Sat, Dec 30, 2023 at 08:39:47PM +0100, Peter Eisentraut wrote:
>>> Ok, I have committed these two patches.
>>
>> Please note that the buildfarm has turned red, as in:
>> https://buildfarm.postgresql.org/cgi-bin/show_stagxe_log.pl?nm=pipit&dt=2023-12-31%2001%3A12%3A22&stg=misc-check
>>
>> pg_stat_statements's regression.diffs holds more details:
>> SELECT query FROM pg_stat_statements WHERE query LIKE '%t001%' OR query LIKE '%t098%' ORDER BY query;
>> query
>> --------------------
>> - select * from t001
>> select * from t098
>> -(2 rows)
>> +(1 row)
>
> That's surprising. I wanted to see if there was any specific
> configuration but I get a 403. I'm wondering if this is only due to
> other tests being run concurrently evicting an entry earlier than
> planned.

These tests are run in a separate instance and serially, so I don't
think concurrency is an issue.

It looks like the failing configurations are exactly all the big-endian
ones: s390x, sparc, powerpc. So it's possible that this is actually a
bug? But unless someone can reproduce this locally and debug it, we
should probably revert this for now.